Land Posters

Modelling of land use may be improved by incorporating the results of habitat suitability modelling and habitat suitability modelling may be upgraded a variable of the process-based variable of forest cover change or accessibility of forest from human settlement is included. Forest cover data derived from satellite images Landsat TM and ETM+ for the years 1988, 2000 and 2007 with resolution of 28.5 m. This work presents the results of spatially explicit analyses of the changes in forest cover from 2000 to 2007 derived from remote sensing using the method of Geographically Weighted Regression (GWR) and of the habitat suitability for protected species of Lynx lynx martinoi, Ursus arctos using Generalized Linear Models (GLMs). The methodological approach is separately searching for aparsimonious model for forest cover change and species habitat suitability for the entire territory of Albania. The findings of this work show that modelling of land change and habitat suitability is indeed value-added by showing higher values of model selection of Corrected Akaike Information Criterion and evaluation metrics of Receiver Operating Characteristic Curve and cross-validation. These results provide evidences on the effects of process-based variables on species habitat modelling and on the performance of habitat suitability modelling of species as well as show an example of the incorporation of estimated species habitats in a land change modelling. 

Community-based conservation (CBC) is nowadays widely accepted as an approach to conservation. However, success is not always present and there is a debate on which factors make CBC work. Tenure devolution on the one hand and a strong legal and institutional framework are thought to enable conservation outcomes, however, empirical proof is missing. In this study, I did a quantitative statistical analysis to compare the annual deforestation rate of various tenure regimes of community-based forest managed areas (CBFMAs) in Africa, to see, whether a devolved tenure system is related to successful conservation. Moreover, I did a qualitative comparative analysis (QCA) to see, whether a strong legal and institutional framework enables the success of CBFMAs. The non-significant results of the quantitative analysis support the view that tenure devolution on itself is not enough to enable successful conservation outcomes. The results of the QCA support the fact that devolved tenure is not necessary, but rather that the combination of strong environmental and human rights legislations, low corruption, an at least medium human development and national policies supporting CBC enable ecologically effective CBFMAs. Even though methodological constraints cause the results to only giving a limited evidence base of the examined relations, and further research is needed to strengthen the findings, this study represents a basis for understanding how tenure and institutional settings are related to the ecological effectiveness of CBFMAs.

The SMOS and SMAP satellite missions each produce global soil moisture products using L-band radiometry. Both missions begin with the same fundamental equations in developing their soil moisture retrieval algorithm but implement it differently due to design differences of the instruments. SMOS with over five years of observations and experience now provides a mature product that is of great value in algorithm refinement and validation of SMAP that was launched in 2015. Over the course of the SMAP Calibration/Validation Phase, which ends May 1, 2016, SMOS data will be combined with a robust set of core validation site observations and sparse network measurements to assess SMAP products. These analyses will also enhance the validation of SMOS soil moisture by expanding the geographic and climate domains of in situ resources.

 

Our Earth’s environment is experiencing rapid changes due to natural variability and human activities. To monitor, understand and predict environment changes to meet the economic, social and environmental needs, use of long-term high-quality satellite data products is critical. The Global LAnd Surface Satellite (GLASS) product suite, generated at Beijing Normal University, currently includes 12 products, including leaf area index (LAI), broadband shortwave albedo, broadband longwave emissivity, downwelling shortwave radiation and photosynthetically active radiation, land surface skin temperature, longwave net radiation, daytime all-wave net radiation, fraction of absorbed photosynetically active radiation absorbed by green vegetation (FAPAR), fraction of green vegetation coverage, gross primary productivity (GPP), and evapotranspiration (ET). Most products span from 1981-2014. The algorithms for producing these products have been published in the top remote sensing related journals and books. More and more applications have being reported in the scientific literature. The GLASS products are freely available at the Center for Global Change Data Processing and Analysis of Beijing Normal University (http://www.bnu-datacenter.com/), and the University of Maryland Global Land Cover Facility (http://glcf.umd.edu). The basic characteristics of the GLASS products and an overview of their applications will be presented.

We conducted a remote sensing study to extract features and patterns of salt affected soils. Hyperspectral data from Hyperion are used in this work due to their very high spectral resolution and a large observation swathe. In order to study saline soils in southern Tunisia, samples were collected for ground truthing in the investigated region. A remote sensing classification of the saline soil was made using several spectral indices.

Soil properties, land use and the ground water table were selected and analyzed to understand the spatial distribution of the different classes of soil salinity. Computed accuracy from the classification varied between 60% and 82%.  

The spectral analysis of different surface features of saline soil confirms that the increased reflectivity corresponds to the increased soil salinity. 

Studied area is located on northern part of Nile Delta from Matruh Gov. in western Nile Delta. Studied area is considered as one of the most populated region which has environmental complicated risks on water and soil, in addition to suffering from water management problem. The problem of surface and groundwater quality deterioration, land salinization and degradation, and water use mal practices are environmental hazards that should be focused and deeply investigated. The socio-economic and health impacts are other sides of the problem which needs sustainable water management issue. 

Undoubtedly, the need for water resources becomes a vital issue in the development schemes of environmentally risked areas. Surface and groundwater resources evaluation, building up a digital geographic information system (GIS), with computerized updated maps made by remote sensing, which comprise all information about surface and groundwater resources are the main target of the present research. Setting-up a sustainable water resources management planning is an urgent needed in order to favor requirements for all activities, to enable mitigation of water-related hazards, and to maintain the water resources without deterioration.

According to the constructed digital data bases, geospatial GIS modeling techniques will be performed to determine the suitable areas and sites for performing sustainable water resources planning. These areas will be determined according to the availability and potentiality of water, soil resources as well as geotechnical behavior.

constructed digital data bases, geospatial GIS modeling techniques will be used as a core base in order to establish database for risk analysis/assessment to emphasize different scenarios for risk management in different aspects; water, soil, socio-economical, human health,….etc.

The work plan will depend on collection of all previous and available works about water and land resources and risks, which are enormous and need to be documented in the planned digital GIS of the project. Remote Sensing (RS) will be used as an effective imaging, mapping and monitoring techniques, where ETM landsat, Spot 4 images, Radar images, Egyptsat-1 images, Quikbird and others will be provided and interpreted for extracting valuable data about drainage system and surface and groundwater resources potentialities. Updating and upgrading the GIS-ready databases will be a good starting point for the project execution. The final stage will be the production of digital up-to-date maps, which will be in the form of hard copy and digital GIS formats capable to be used for future data update and ready for decision makers and planners. The project fund will be used for performing a full planning for surface and groundwater sustainable management with different alternatives, techniques and methods. Sufficient fund needed for supplying materials and collecting databases and running watershed modeling techniques. These data includes purchasing sophisticated satellite images with excellent spatial and optical resolution, using advanced modeling software in watershed modeling and surface/groundwater assessment, purchasing commercial high resolution Digital Elevation Models (DEM) (30 m resolution), Radar images, etc. These data and equipments need adequate fund to be available and applied in our project. Also, in parallel, collecting of the previous social and economical studies will take part using all required data, statistics and reports in order to scan the previous and current socio economic situations regarding the water risk and its effect on all social and economical aspects as human health risks and economical results which will be the corner base for promoting the socio-economical and environmental expecting scenarios.

Soil moisture monitoring and accurately acquiring is great significance for explore the mechanism of global water cycle, construct climate and hydrological model,monitorcrop growth , forecast drought disasters and so on. Microwave remote sensing has become one of the most effective means to monitor soil moisture rely on the advantages of all-weather, day and night observation, ability to penetrate surface etc.

The study of this paper mainly concentrated on the north of China’s Henan plain region which in front of Taihang Mountains, and carried out the validation/calibration and application research with SMOS L2 soil moisture data products. The results and innovations are mainly in the following aspects:

（1）The average-average and node-site validation methods were carried out to SMOS L2 soil moisture products. The results shows that in the 3 pixels and the 9 sites, the correlation coefficient of the SMOS and the Insitu are mainly concentrated in 0.20 ~ 0.40, also the existence of dry bias mainly concentrated in the 0.06 ~ 0.14, the changes and improvements of soil dielectric model from Dobson to Mironov make the difference between the SMOS and Insitu decreases, and SMOS soil moisture has a seasonal performance, especially in summer.

（2）Analyze the influence factors on the quality of SMOS soil moisture produncts: ① Precipitation, the correlation coefficient of the Precipitation between SMOS and Insitu are 0.25, 0.23 respectively. Continuous precipitation affect SMOS observation, when precipitation reaches in torrential rain, it has a persistent effect on soil moisture. ②Land cover, the proportional of pixel land cover have an effect on the difference between the SMOS and Insitu; the typical regional heterogeneity influence on soil moisture retention. ③ RFI, RFI have a slightly increasing trend; the different geographical position, the influence of RFI is also different.

（3）Especially for RFI influence factors, the contaminated SMOS L2 soil moisture data were filtered with the filter abnormal value criteria and filter RFI formula, then the validation shows that in the pixel 1 and 4 sites the correlation coefficient of the filtered SMOS and the Insitu are mainly concentrated in 0.10 ~ 0.32，the filtered data in alleviating numerical wave, but fail to improve on the verification accuracy.

The utilisation of UAVs for the acquisition of ultra-high imagery has heavily increased during the past years. Once the hardware is purchased images can be recorded almost any time at very low cost. The image parameters can be determined in terms of spectral channels, image overlap and geometric resolution. Depending on the flight altitude and the camera setting, the geometric resolution can reach 1 cm or even better. The overlap between the images enables stereoscopic image processing.

In this study, an area of approximately 175 ha was covered by a UAV mission. The area is located 15 km southeast of Jena, Germany. It is covered with planted forest of different ages; the main species are pine, fir, larch and birch. In total a number of 1750 RGB images (sony NEX-7) and 5200 MS images (Tetracam miniMCA) were recorded. The flight altitude was 100 m above the tree tops. The overlap in flight direction was 80% and 60% between parallel tracks. The cameras were mounted on an octocopter (Logo-Team Geocopter X8000). As the flight duration is limited the study area was subdivided into seven subsectors. For one of the subsectors TLS (Riegl VZ 1000) was recorded during the UAV mission.

The single images were processed with the Software Photoscan 1.1.6 (Agisoft). Processing includes image alignment, point cloud generation, mosaicking, and orthorectification. The resulting data was georeferenced using DGPS data. For the further processing the point clouds (geometric features) and the image mosaics (spectral features) were used. The point clouds feature approximately 80,000,000 points per subsite which corresponds to 300 points/m². This data was used to delineate forest structure parameters such as canopy height, canopy gaps, tree density (after single tree detection), and tree height. Using allometric equations the stem volume was estimated. The tree species were derived from the spectral information. The delineated forest parameters were in good agreement with the TLS data based forest parameters.

Based on the gathered forestry data homogenous forest plots were detected. The plots will be used in a following study aiming at investigating the impact of forest structural parameters on SAR backscatter (Sentinel-1, ALOS-2 PALSAR-2). Besides the impact of forest structure also the impact of soil moisture will be investigated. For this purpose the study site was equipped with a soil moisture network (240 probes).

We estimate the synergetic potentials of Sentinel-1 and Sentinel-2 for a supervised monitoring of deforestation sites and consecutive land uses. As surrogates we fuse RADARSAT-2 (RS2) and Landsat 8 (LS8) time series using an approach based on Markov Random Fields (MRF). The time series consist of 13 and 22 scenes, respectively, which were acquired between 2015-01-02, and 2015-12-20.

The study region is located in the South West of Pará, Brazil, in close proximity to the current deforestation frontier. Deforestation in the study region is a consequence of the growing demand for pasture land, and its northward relocation from Mato Grosso to Pará. It was enabled by the increasing development and paving of the longitudinal highway BR-163 since the beginning of the 1990's. As a consequence, various stages of post-deforestation land cover classes can be identified today.

For this study we use MRF to generate high accuracy land cover maps for each acquisition instance, which allows us to track changes of highly dynamic land cover classes. Using probabilistic land cover maps as input, the MRF performs a sophisticated multi-temporal smoothing, which respects the spatial and temporal interactions between each two classes. This is achieved through integrating expert knowledge about the land cover trajectories of the study region into the temporal transition matrices of the MRF. This method can be transfered to fuse time series data of any number of sensors, and due to the properties of the inference algorithm used in this study (Loopy Belief Propagation), it can also be tiled to allow the integration of artificially large data sources. Reference data is derived utilizing high resolution time series of RapidEye as well SPOT-5 to gather reliable land cover information over the same 2015 acquisition period.

As a result of the extensive data basis and the good characteristics of the developed classification scheme, the presented approach of combining SAR and multispectral time series allows for the detection of short termed land cover changes, not only with regards to forest classes, but with enough confidence to track various other carbon relevant land cover types (e.g. pasture areas, secondary regrowth). Synergetic potentials of multi-sensoral input (RS2+LS8) are discussed as results are compared to mono-sensoral model runs relying solely on input of either RS2 or LS8. We furthermore discuss the degree to which LS8, and hence Sentinel-2 classifications, can benefit when they are linked with C-band classifications (such as Sentinel-1) via an MRF.

Land covers in Saudi Arabia are generally described as salty soils with sand dunes and sand sheets. Waterlogging and higher soil salinity are major challenges to sustain agricultural practices in Saudi Arabia principally within closed drainage basins. Agricultural practices in Saudi Arabia were flourishing in the last two decades. New reclaimed land were added annually and distributed all over the country. Irrigation techniques are mostly modernized to fulfil water saving strategies. Nevertheless, water resources in Saudi Arabia are under stress and groundwater levels are depleted rapidly due to heavy abstraction that may exceed crop water requirements in most of the cases due to high evaporation rates. The excess use of irrigational water leads to severe soil salinity problems. Applications of remote sensing technique in agricultural practices became widely distinctive and cover multidisciplinary principal interests on both levels locally and regionally.  The most important remote sensing applications in agricultural practices are vegetation indices which are related to vegetation and water especially in arid environment. Soil salinity mapping in an arid ecosystem using remote sensing data is a demanding task. Several soil salinity indices were implemented and evaluated to detect soil salinity effectively and quantitatively. Thematic maps of soil salinity were satisfactory produced and assessed.

There is an increasingly growing interest in observing the Earth’s surface from space.  This can be done in several ways, such as using visible and radar imaging - the latter significantly less popular amongst members of the international scientific community. The purpose of this study is the exploitation of the advantages of radar satellite data, namely independence from cloud cover and solar illumination, for land cover classification. The properties of the recorded radar data heavily depend on the type of land cover, season and weather conditions. Therefore, it is possible to utilize the variability of these factors, in order to develop various methodologies and innovative techniques that can be used for classifying land surfaces. Although the study focuses on four main land cover types in Europe (urban, mountainous, agricultural-low vegetation and forested areas), the same principles can be extended worldwide, concluding to useful results for designing future satellite missions or as a guideline for researchers interested in an in-depth studying of a specific land cover type. 

We present an integrated monitoring framework that uses all available Landsat data and a continuous stream of community-based monitoring data to characterize forest changes in an Afromontane forest system in southern Ethiopia. To demonstrate this framework, we trained random forest models using a suite a spectral-temporal variables derived from the Landsat time series data as covariates and forest disturbance reports from local forest rangers as training data. We classified deforestation and degradation with out-of-bag (OOB) class accuracies of 74% and 69%, respectively, with an overall OOB accuracy rate of 71%. While these models did not show improvements for the deforestation class compared to previous studies in our site, our approach succeeded in mapping diffuse forest degradation with greater certainty than previously achieved. Forest change classification accuracies improved as more ground-based observations from local rangers became available, demonstrating the utility of such a continuous data stream. The random forest models also revealed that short-wave infrared bands, or indices based on these bands, consistently provided the most important information for distinguishing deforestation, degradation and no-change classes. Given the continuous acquisition of satellite-based and in situ observations, this framework provides a flexible approach to monitoring forest changes which could be used to ingest observations from newly launched sensors such as Sentinel-1 and 2 combined with continued Landsat-7 and 8 acquisitions. Furthermore, with modifications to the parameters measured by local forest rangers, this framework could be scaled up to monitor a broader range of relevant forest variables.

Water resources management in Sinai Peninsulais vital needed goal for the different developmental activities. A detailed hydro-morphometric analysis for Wadi Fierran basin in Southern Sinaiwas introduced through the present work to throw lights on new potential or priority areas suitable for the water holding capacity which will have its own bearing on the groundwater recharge. The sub-watersheds' boundaries and hydro-morphometric parameters were delineated using the the ASTER DEM and Spot-4 satellite image mosaic by the WMS 8.0^© Software. Subsequently, six thematic layers were used to construct a multi-parametric weighted spatial probability model (WSPM) by the  ArcGIS 10.2^© software. They are represented by the bifurcation ratio (Rb), drainage density (Dd), drainage texture (Dt), stream frequency (Fs), length of overland flow (Lg) and basin infiltration number (If). The resulted prioritization map classifiedW. Fierran into four priority classes ranging from the poor to very high for the water holding capacity. Accordingly, El-Akhdar, El-Rahaba, El-Shiekh, Solaf, Nesriene, sub-watersheds were categorized as high to very high in water holding capacity, because they have low to moderate values of Dd and high Rb, moderate values of  Lg, low values of If and very low values of Fs. The high and very high priority classes in these sub-watersheds occupy most of W. Fierran, which constitute about 72% of its total area.

The land surface temperature (LST) is an extremely significant parameter in order to understand the processes of energetic interactions between the Earth's surface and the atmosphere. This knowledge is significant for various environmental research questions, particularly with regard to climate change. The current challenge is to reduce the higher deviations during daytime especially for bare areas with a maximum of 5.7 Kelvin. These temperature differences are time and vegetation cover dependent. This study shows an innovative approach to retrieve land surface emissivity (LSE) and LST by using thermal infrared (TIR) data from satellite sensors, such as SEVIRI and AATSR.

So far there are no methods to derive LSE/LST particularly in areas of highly dynamic emissivity changes. Therefore especially for regions with large surface temperature amplitude in the diurnal cycle such as bare and uneven soil surfaces but also for regions with seasonal changes in vegetation cover including various surface areas such as grassland, mixed forests or agricultural land different methods were investigated to identify the most appropriate one.

The LSE is retrieved by using the day/night Temperature-Independent Spectral Indices (TISI) method, while the Generalised Split-Window (GSW) method is used to retrieve the LST. Nevertheless different GSW algorithms show that equal LSEs lead to large LST differences. For bare surfaces during daytime the difference is about 6 Kelvin. Additionally LSE is also measured using a NDVI-based threshold method (NDVI^THM) to distinguish between soil, dense vegetation cover and pixel composed of soil and vegetation. The data used for this analysis were derived from MODIS TIR.

The analysis is implemented with IDL and an intercomparison is performed to determine the most effective methods. To compensate temperature differences between derived and ground truth data appropriate correction terms, by comparing derived LSE/LST data with ground-based measurements, are developed. One way to calibrate LST retrievals is by comparing the canopy leaf temperature of conifers derived from TIR data with the surrounding air temperature (e.g. from synoptic stations).

Prospectively, the derived LSE/LST data become validated with near infrared data obtained from an UVA with a TIR camera (TIRC) onboard, and is also compared with ground-based measurements.

This study aims to generate an appropriate method to eventually obtain a high correlation between LSE/LST, TIRC and ground truth data by integrating developed correction terms.

The effectiveness of Rapideye high-resolution multispectral imagery for vegetation mapping in Madeira Island Natural Park (Archipelago of Madeira, Portugal) was assessed, by using an image acquired in August 2011. Geometric and atmospheric corrections were applied to all five Rapideye image spectral bands. Masks of both clouded and strongly shadowed areas were built. A classification scheme of 26 land cover and vegetation classes was set up based on a dataset of 1341 training sites. A segmentation procedure was developed by using a region-based algorithm. The spectral separability of both pixel-based and segmentation-based classification schemes was assessed. A strong spectral separability between most relevant vegetation classes was achieved. After applying a parametric supervised classifier (Maximum Likelihood), the accuracy of both classification schemes was also assessed and compared. Although both pixel-based and object-based supervised classifications showed strong agreement and good accuracy at overall and vegetation class level, the segmentation-based classification scheme proved to be significantly more accurate. Therefore this methodological approach applied to Rapideye high-resolution multispectral imagery was confirmed as a cost-effective procedure for mapping and monitoring vegetation of Madeira Island Natural Park (Archipelago of Madeira, Portugal).

Sandification refers to land degradation in sandy areas.  Considerable attention has been given to sandification processes in China since vast areas of sandy deserts are located in the north of the country within arid and semi-arid climatic zones.  The current paper is aimed at assessing the land-use/land-cover spatial and temporal dynamics over the Mu Us Sandy Land, China, via change detection methodology based on spaceborne images.  Two biophysical variables, NDVI, positively correlated with vegetation cover, and albedo, positively correlated with cover of exposed sands, were computed from a time series of merged NOAA-AVHRR and MODIS images (1981 to 2010).  Generally, throughout the study period, NDVI increased and albedo decreased.  Improved understanding of spatial and temporal dynamics of these environmental processes was achieved by using the Change Vector Analysis (CVA) technique applied to NDVI and albedo data extracted from four sets of consecutive Landsat images, several years apart.  Changes were detected for each time step as well as over the entire period (1978 to 2007).  CVA created four categories of land-cover change – vegetation, exposed sands, water bodies, and wetlands.  The CVA’s direction and magnitude result in pixel-based maps of the change rather than broad qualitative classes, such as slight-, moderate-, or severe land degradation that previously presented for this region.  Each of the four categories has a biophysical meaning that was validated in selected hot-spots, employing very high spatial resolution images (e.g., Ikonos).  Careful selection of images, taking into account inter and intra annual variability of rainfall, enables differentiating between short-term conservancies (e.g., drought) and long-term alterations.  NDVI and albedo, although comparable to tasseled cap’s brightness and greenness indices, have the advantage of being computed using reflectance values extracted from various Landsat platforms since the early 1970s.  It is shown that, over the entire study period, the majority of the Mu Us Sandy Land area remained unchanged.  Part of the area (6%), mainly in the east, was under human-induced rehabilitation processes, in terms of increasing vegetation cover.  In other areas (5.1%), bare sands were found to expand to the central-north and the southwest of the area.

Reference

Karnieli, A., Qin, Z., Wu, B., Panov, N.  Yan, F. 2014. Spatio-temporal dynamics of land-use and land-cover in the Mu Us Sandy Land, China, using the change vector analysis technique. Remote Sensing. 6, 9316-9339. (doi:10.3390/rs6109316).

Due to deforestation, intact tropical forest areas are increasingly transformed into a mixture of remaining forest patches and human modified areas. These forest fragments suffer from edge effects, which cause changes in ecological and ecosystem processes, undermining habitat quality and the offer of ecosystem services. Even though detailed and long term studies were developed on the topic of edge effects at local scale, understanding edge conditions in fragmented forests on larger scales is crucial for predicting habitat loss and developing management options. In order to evaluate forest conditions in forest edges for large areas, we analyzed 11 LANDSAT Tree Cover (LTC) scenes (180 x 185 km each, 8 in the tropics and 3 in temperate forested areas) using tree cover as an indicator of habitat quality and measured its values in relation to edge distance. In addition, we did a temporal analysis of LTC in a smaller area in the Brazilian Amazon forest where one larger forest fragment (25.890 hectares) became completely fragmented in 5 years. Our results show that tree cover LTC near to fragment edges for all 11 scenes show great variation, becoming more stable as distance to forest edge increases. Depending on the scene, maximum distances where edge-interior differences in LTC were detected ranged from 125 to more than 500 meters from the edge. Our temporal analysis also confirms that after 5 years of fragmentation, close to the edge (between 50 and 100 meters) the amount of low LTC areas increased. Although it is still unclear which are the main causes of LTC edge variability, this generalized pattern suggests that 1. Edges can be described as areas where forests are exposed to intense spatial and temporal variability; 2. Maximum extent of edge effect penetration (between 125 – 500 meters) varies greatly depending on the studied region; and 3. Small fragments dominated by edges will be exposed to this intense variability. Finally, our results confirm LTC as a consistent high-resolution indicator of forest conditions, providing more information than forest-non forest maps.

CLIMATE CHANGE IMPACTS ON SMALLFARMERS AND LIVELIHOOD IN          LEVUBU COMMUNITIES, VHEMBE DISTRICT, SOUTH AFRICA.

                               Authors: Zongho-kom & N.S Nethengwe

                                Email: zogokom@yahoo.fr and nethengwe1@yahoo.com

                             Department of Geography & Geo-Information sciences,

                                       University of Venda, South Africa

                                              Abstract

Climate change has been considered as one of the biggest challenges of the 21st century because it affects all countries in the World. It has major crisis on the society due to it high significant negative impact Worldwide. The main aim of the study would be to investigate the major climatic change trends impacts and establish the interrelationship between agriculture and communities livelihood in Levubu, Vhembe district, also in order to suggest better and desirable management strategies to better the life of the local population in the study area.

This research design is multidimensional to achieve the main objective of the study, so a triangulation of different data sources and data collection instruments would be employed.  Primary data would be collected from household interview, focus group discussion and key informant interview. Secondary data would be obtained from literature and other secondary sources. This thesis would use semi-structured interview to garner data from local society, government officials, consultants, and secondary data from published and unpublished sources. Data analysis would be systematically analyses, using both qualitative and quantitative analysis. Both qualitative and quantitative data would be analyse by means of coding, and taxonomies, visual representation and Statistical Package for Social Sciences (SPSS) and it would generate various form of description statistic like; charts, graphs and frequencies tables. This study would use levubu area in Vhembe district, South Africa as a case study and determine the climate trends and impacts on agriculture and on the livelihood of the locality.

 

 

Abstract

Vegetation is one of the most important renewable natural resources, which play a role in the preservation of the environment and biodiversity. Various land use activities such as urbanisation, population growth, as well as climatic change have been some of the major drivers which alter vegetation cover and contribute to biodiversity loss. This paper aim using remote sensing and Geographical Information Systems (GIS) to quantify vegetation changes and analyses the spatio-temporal changes and adopted NDVI index. The NDVI (Normalised Difference Vegetation Index) taxonomy is used to detect the spatio-temporal change in vegetation cover in Makhado Alti-villas portion of town. The NDVI value above zero is further classified into three groups viz 0.0-0.20 (bare surface) 0.20-0.46 and 0.46-0.74 (dense forest and cropland vegetation). The monitoring of vegetation change can play a vital role in knowledge generation, best practices, as well as Environmental Monitoring and Evaluation which can abate in the near future. This paper main contribution was to illustrated vegetation cover change and some drivers of vegetation degradation in Makhado urban area from 2007 to 2012.

Keywords: Makhado, GIS, remote sensing, NDVI, land use/land cover, Environmental Monitoring, Alti-villas, biodiversity.

Wetlands provide invaluable ecosystem services and contribute significantly to food security around the world. To maintain these functions, wetlands need to be protected from rapid transformation and overuse. Spatially-explicit information is required for sustainable wetland management. Development of wetland maps based on remote sensing requires a clear-cut definition of wetlands. Within the GlobE- and SWOS-projects, we try to improve the understanding of these habitats from a remote sensing perspective and to determine available wetland map layers for the East African countries of Kenya, Rwanda, Tanzania and Uganda.

A multi scale approach is being utilized in acquisition of information which will give insight into better understanding and subsequently managing the use of the wetlands. At the regional scale spanning the four countries, MODIS NDVI satellite data time series are exploited to develop seasonal maps for the wetland locations in East Africa. Within these regions, the catchment scale is under study at Kilombero in Tanzania. Stacked RapidEye images are in use in classification of the area so as to obtain a land cover map. From this map, areas within the wetland that have potential of food production can be identified. The LULC map will be an input in hydrological models increasing knowledge on the hydrologic cycle in the wetland. At the flood plain scale, SENTINEL-1, TerraSAR-X and RadarSAT-2 images will be utilized in analyzing in field soil moisture variation with time. Soil moisture is an essential parameter in growth of crops and thus its variation over time will aid in dictating the most suitable crops that will thrive well given the environmental conditions and give high yields. In addition, very high spatial resolution UAV flight campaigns have been carried out. From these, digital elevation models and vegetation indices have been derived giving an overview of the dynamics of vegetation in time. These indices will aid in monitoring phenological changes in the plants and furthermore be exploited in crop yield estimation.

Dubai is a modern city which is growing at a fast pace, constantly changing and developing. The application presented in this paper is part of the SAFIY project, a project to develop satellite data applications to help the government in Dubai to get up-to-date information on features such as water bodies, vegetation areas, buildings and roads. In this paper we present an application to automatically update the road network map for the city of Dubai using very high resolution satellite imagery, Deimos-2 and DubaiSat-2 at 0.75 cm resolution. Deimos-2 and DubaiSat-2 have been recently included as third party contributors to the Copernicus constellation.

A review of the state of the art is included to give an overview of the latest techniques and data used for road extraction. The algorithms implemented use supervised classification techniques to extract the road pixels. Vector data of the map to be updated is available and is used as training data. The classified pixels are grouped and the road orientations and intersections are formed. Special road objects such as roundabouts or bridges are also detected using algorithms based on feature descriptors, designed to match the specific characteristics of those objects. Finally, the system identifies the changes in roads and objects from the current road network map in order to update the map

The city of Dubai is 4,114 km² and has many different types of neighbourhoods from suburbs to very densely populated areas. The use of very high resolution data is both advantageous and challenging in that the high level of detail that is captured can show the different types of roads, surfaces, widths, etc. but also can become noisey  for the algorithms, for example cars or high variability between what is described as a road for the algorithms. Shadows, parts of the roads hidden by tall buildings and some of the features such as bridges and tunnels are also a challenge when building a consistent road network.

Since automatic algorithms do not often achieve the 100% accuracy, added manual steps will be needed in order to perfect the results obtained by the machine. Our work includes the design of a strategy and a user interface so that human interaction is as positive and rewarding as possible, as opposed to just blindly fixing classification errors.

Overall, we propose a novel end to end system that automatically updates a road network map and allows a user to easily improve this so that the final resulting map is usable.

EO information and related services are on the rise, but not yet appropriately applied to planning and commercial sectors. Within the EOworld program (ESA) key aspects of World Bank topics were addressed in order to raise awareness of capabilities and the potential of EO information services to support the decision making process of financing institutions (World Bank) and assist in analysis and planning activities of different scale. The project was designed to set-up EO information services following thematic requirements for planning activities, construction, appropriate management of investment activities, combined with detailed spatial information for broader modelling of potential future developments.

Depending on the thematic scope of the World Bank representatives and the related stakeholder a range of urban services, tailored solution as well as standardized GIS analysis were provided for implementation. The results will be presented and conclusions summarized which are input for a wider spread opalization of a range of baseline EO information service clusters to meet the users requirements, depending on their background to remote sensing.

 

    The southern steppe of Algeria is subject to very ancient climatic and anthropogenic disturbances in recent decades, from the 1970s, the changes recorded in the steppes are important. The major observation now is that of a decrease in area of this steppe, which is the consequence of the degradation sometimes extreme. This situation due to the phenomenon of desertification in this region has caused degradation of biodiversity, including wildlife and flora, threatening to disappearance of natural resources hard renewable.

      Thus, the last thirty years, there is a real dynamic change of land cover with intensive degradation of the natural vegetation especially in arid zone. Indeed, the adverse effects of drought periods from the year 1970 combined with population growth and economic conditions experienced by the country in the 1980s have greatly upset the delicate balance of the natural environment. These adverse effects may result in partial or total disappearance of some natural ecosystems. However, the location of the most significant different changing sectors in space and time, allows specialists planning and local leaders understand these spatial changes that affect natural ecosystems in Algeria. In addition, there are relatively few studies using a long time series of data to determine land cover changes at local scale in Algeria.

      The objective of this work is to study the distribution of plant formations that constitute the ecosystem typical of west Algeria and their dynamics in time and space, as well as to develop a method to monitor the degradation process and a system capable of effectively protecting areas classified for their plant and animal species. To this end, it is necessary to make an assessment of the state of the land in general, and of plant cover in particular, which should be carried out as soon as possible. As part of this work, we have studied the vegetation dynamics of steppe by contrasting the present situation with that prevailing 35 years ago (1979 and 2014).

      Using remote sensing, this study identified and discriminated plant groups in the steppe with the analysis of remotely sensed data (LANDSAT 8), digital processing and field observations, the study aimed to establish vegetation types, in terms of sensor satellite perception. The LANDSAT satellite images were used to map the vegetation of the study area at a scale of 1:200,000. A comparison was then made between the map obtained from satellite images (LANDSAT 8) in 2014 and the vegetation map established in 1979.

          The results show the following main trends in the distribution patterns of plant specie, a strong decrease of land occupied by steppe of Stipa tenacissima and steppe of lygeum spartum, witch replaced by three taxa  Diplotaxis houra, Gynandiris sisynchuim and Peganum harmala. Steppe of Artemisia herba alba has been transformed by steppe of Diplotaxis houra and Gynandiris sisynchuim. Woody species such as Quercus ilex  and Juniperus phoenicea are characterized by a large regression.

    

Long-term global satellite-based and reanalysis soil moisture products have been available for several years. Comprehensive evaluation of these products is significant before using them. In this study, in-situ soil moisture measurements from 2008 to 2012 over Southwest China are used to examine the reliability of four satellite-based and one reanalysis soil moisture products. This study evaluates satellite data products (AMSR-E, ASCAT, ESA-CCI, SMOS) and reanalysis data (ERA-Interim) over Southwest China using new in situ soil moisture data. Evaluation of soil moisture absolute values and anomalies shows that all the products except for AMSR-E and SMOS can capture well the temporal dynamics of in-situ soil moisture. The bias and noise in AMSR-E and SMOS are probably due to the severe effects of radio frequency interference (RFI) over this region. In general, the ERA-Interim and CCI SM perform the best compared to the in situ data. The accuracy levels are comparable to validations over other regions worldwide. Therefore, local hydrological applications and water resources managements are expected to benefit a lot from the long-term ERA-Interim and CCI SM soil moisture products.

Microwave remote sensing has been largely applied to retrieve soil moisture (SM) from active and passive sensors. The obvious advantage of microwave sensor is that SM can be obtained regardless of atmospheric conditions. However, existing global SM products only provide observations at coarse spatial resolutions, which often hamper their applications in regional hydrological studies. Therefore, various downscaling methods have been proposed to enhance the spatial resolution of satellite soil moisture products. The aim of this study is to investigate the validity and robustness of a simple Vegetation Temperature Condition Index (VTCI) downscaling scheme over different climates and regions. Both polar orbiting (MODIS) and geostationary (MSG SEVIRI) satellite data are used to improve the spatial resolution of the European Space Agency's Water Cycle Multi-mission Observation Strategy and Climate Change Initiative (ESA CCI) soil moisture, which is a merged product based on both active and passive microwave observations. The results from direct validation against soil moisture in-situ measurements, spatial pattern comparison, as well as seasonal and land use analyses show that the downscaling method can significantly improve the spatial details of CCI soil moisture while maintain the accuracy of CCI soil moisture. The application of the scheme with different satellite platforms and over different regions further demonstrate the robustness and effectiveness of the proposed method. Therefore, the VTCI downscaling method has the potential to facilitate relevant hydrological applications that require high spatial and temporal resolution soil moisture.

Fire disturbances affect large areas of the Canadian boreal forest each year, with impacted areas expected to increase under future climatic conditions. Forest fire disturbances can be monitored on an annual basis for disturbances using remotely sensed data to provide seamless, consistent, and synoptic reporting across multiple jurisdictions. While not mapped, the return of vegetation towards forest re-establishment following harvest is typically captured via field visits as a component of sustainable forest management practices.  In contrast to harvesting, other forest disturbances (e.g., fire) are not necessarily subject to the same practices, with the return of forest vegetation not systematically captured. As a result, there is an opportunity to use remote sensing to track the return of vegetation following disturbance and to further inform on forest re-establishment trends over time. Our objective is to demonstrate the diversity of spectral trajectories that represent forest regrowth after stand replacing fire disturbances, first spatially over the Canadian Boreal and Taiga Shield ecozones, and then temporally using time series techniques.Landsat data products were acquired from the National Terrestrial Ecosystem Monitoring System dataset, providing a continuous and gap free Landsat surface reflectance dataset covering the entirety of both ecozones and temporally between 1985 and 2012, with an additional disturbance classification data layer. The spectral data were restricted to fire disturbances and spectral trajectories were tracked for five years post disturbance, grouping each year’s post disturbance regrowth trajectories together for each year from 1985 through 2007. Thus a time series of initial regrowth trajectories was created and then used to calculate spectral trajectory metrics.Spectral trajectory metrics indicate that forests overwhelmingly show signals of regrowth after disturbance; however, post disturbance spectral trajectories vary spatially and temporally across the Boreal and Taiga Shield ecozones. Regrowth metric differences between each ecozone are greater than the differences that are observable over time in each ecozone, though several years show concurrent trends within multiple ecozones.These results confirm the applicability of Landsat data to monitor large areas with high levels of detail and the utility of long annual time series to generate meaningful and unique information. The spatial variability of post disturbance spectral trajectories may reflect the broader inherent differences between ecozones. However the temporal differences within each ecozone show the effects of a variable site conditions on disturbance events, and in particular how single large fire disturbance events can drive post disturbance spectral trajectories for a number of years after the initial disturbance.  

Forest operationas are generally performed in large-scale and difficult mountainous terrains in Turkey. Therefore, remote sensing technologies potentially provide important advantages in forest operations by reducing the workload of forestry, evaluating the large areas with less time, and making straightforward plans. The cost minimization is usually common goal and necessity for intensive forestry activities. Recently, satellites and other aerial vehicles combined with different sensors provide important remotely sensed data that have intensive usage in the forestry such as land use/land change detection, forest transportation, and the forest fire management. Landsat, MODIS, SPOT and IKONOS are common observation satellites used for environmental assessments in which thermal sensors are one of the most preferred tools used for monitoring forest resources. Aerial thermal imaging applications are performed with thermal sensor mounted on most of the aerial platforms such as planes, helicopters and unmanned air vehicles. Pros and cons of satellite and aerial imagining may be disputable, but thermal reflection data still play important role in forestry related researches in Turkey. Especially freely available Landsat thermal imagery offering bulk gallery for land time series provide researches and practitioners with great data sources for conducting studies on digital image processing. This study aims to evaluate potential positive and negative impacts of forest operations on residual stands by using Landsat thermal imagery. As a result of this study, it is concluded that thermal imagery can be effectively used for detecting impacts of forest operations on residual forests.

This study relates to part of the area of Guéra, located between the latitudes 12°5' and 12°20' and longitudes 18°20' and 18°50'.¶It lies within a more general scope of lithostructural mapping of central massif of Chad for the comprehension of the geodynamic evolution and the bonds between the various fields of the Pan African chains in the North of the Congo craton and the West of the West Africa craton.¶

¶The principal aims of this work are the lithological and structural mapping of the Precambrian basement of Mongo. ¶The exploitation of the hybrid data resulting from the merging of Landsat 8 and PALSAR provided us an invaluable help for the lithostructural mapping. ¶Indeed, specific treatment of the images (spectral signature, Decision tree, ratio of images, etc.) ¶allowed to raise and identify lithological edge (diorite, calc-alkali granite, calc-alkali granite Tardi and Post-tectonics, granodiorite, very thick formations of recent pediment and old alluvia).¶The application of the directional filters of Sobel, those of Yésou and Prewitt made it possible to raise the features on the new-channels resulting from the analysis in principal component. ¶The integration of the images treated with other data ¶in the GIS facilitated the interpretation and the mapping of the various lithological features and the fractures of this part of the Precambrian basement of Guéra. ¶The rose diagram of the fractures present a NE-SW major orientation. ¶

Land surface type plays an important role in controlling Earth’s environment: altering surface roughness, albedo and exchange rates of heat, water vapor, CO₂ and other green house gases between land surface and the atmosphere, and consequently affecting major components of the water, energy and carbon cycles of the weather and climate systems. Depending on time scale, monitoring land surface type changes is also increasingly important for natural disaster assessment and natural resources management. The literature has demonstrated that rapid and slow land surface type changes could be detected using daily observations of optical and microwave satellites of various national space agencies such as NASA, ESA, NOAA, EUMETSAT and JAXA. The study to be presented explores the feasibilities of detection major land surface type changes, such as active fire and burned areas, active flooded areas, urban expansion and deforestation, using daily observations from the Visible Infrared Imaging Radiometer Suite (VIIRS) on current Suomi National Polar-orbiting Partnership (Suomi NPP) and future Joint Polar Satellite System (JPSS) and the second Advanced Microwave Scanning Radiometer (AMSR2) on the 1^st Global Change Observing Mission (GCOM-W1). The all-weather observations from AMSR2 can be used to detect possible significant changes of surface soil moisture (SM) or vegetation water content (VWC) at course spatial resolution which may indicate biomass burning, flooding or deforestation within a time period from a day to multiple years. Once a potential surface type change is detected at the coarse resolution in a region, data from VIIRS are collected and processed for the region to find out the specific surface type change and location (e.g. flooded areas at 375m spatial resolution).  The simple change vector approach is applied to these satellite data. Testing results for the flooding occurred in early October of 2015 in South Carolina of United States will be presented as a prototype. Implementation of the approach to operationally detecting the major land surface type changes will be discussed and demonstrated with examples.

 

Phenology is the study of recurring life-cycle events that are initiated and driven by environmental factors. Changes in plant phenology constitute solid evidence for the influence of global climate change on different species and ecosystems, therefor it is highly important to monitor phenology and study the influence of climate on it. Previous phenological studies focused mainly on exploring the shifts in the timing of phenological events as a response to various climatic changes. Another developing research field is phenology modeling using climatic factors for explanatory and predictive purposes. Forests are interesting and important ecosystems for phenological modeling, due to their extensive global cover, their long life span, and their functioning as carbon storing ecosystems. In Israel, planted and natural forests are spread along a strong climatic gradient, from arid in the southern parts of it, through semi-arid in the center and up to dry sub humid and humid in the northern parts. Although the influence of climatic events on forested areas in Israel had been studied before, there are still some uncertainties about how climate in the last decades has influenced on the phenology of these forests. Specifically, it is still unclear how important particular environmental factors are in determining forest phenology, and what is the ability to predict future phenological cycles from climatic variables. These questions can be addressed by establishing models that quantify the impact of climate variables on phenological cycles. The primary objective of this research was to explore the relations between three main growth limiting factors (precipitation, temperature and radiation) and two vegetation indices (normalized difference vegetation index (NDVI) and normalized difference water index (NDWI)) in four forests located along a climatic gradient in Israel. MODIS based NDVI and NDWI were used as a proxy to the phenology of the forests, and statistical time series modeling approach was used in order to establish best explanatory model. Specific objectives were to examine which vegetation index (NDVI or NDWI) is more suitable for phenological modeling, and to create a time series prediction model that quantifies the effect of potential climatic constraints to plant growth (e.g., precipitation, temperature, radiation) on vegetation indices. This model strives to explain and predict continuous multi-year phenological cycles, rather than specific phenological events. Time series analysis showed the decreasing influence of precipitation as a limiting factor from the southern arid site to the northern dry sub-humid site; in the southern Yatir site precipitation was the main factor dictating the phenology of the forest. However, in the three others sites (Jerusalem corridor, Carmel and Galilee) the influence of temperature and radiation on phenology was either similar to that of precipitation or higher. The explanatory power of the models that were fitted for each site was relatively high ranging from adjusted R² results of 0.76 to 0.87 for NDVI, and 0.67 to 0.84 for NDWI (depending on the site) (Fig. 1-2). However, the predictive abilities of the models, which were tested on a validation period, showed limited functioning (Fig. 1-2). 

Our presentation focuses on the evaluation of the potential of Landsat and SAR C band sentinel1 radar optical images and their complementarities for the historical analysis of deforestation and mapping of land cover in tropical forest ecosystems. The area of interest of this study is the region immediately bordering and inside the Virunga National Park (VNP) in DRC. The VNP sits alongside the equator, on the Albertine Rift volcano chain in North Kivu, eastern DRC along the borders of Rwanda and Uganda. VNP is the oldest African national park having been created in 1925. The interest of this site resides in its ecological diversity in addition to the multiple threats facing its ecosystem (slash and burn and illegal agriculture, charcoal production, illegal logging, armed conflicts, the installation of war-torn displaced populations, artisanal mining of mineral resources).
To perform the historical analysis of deforestation, available low cloud cover Landsat satellite images were used. The following three reference dates were chosen: 1995, 2005 and 2015. A supervised classification (SVM classification algorithm with RBF kernel) was applied to each Landsat satellite image to separate forest from non-forest areas. A forest /non forest map was then produced for each reference date. Classifications were made using the Orfeo Toolbox software. Deforestation maps and subsequent spatial analysis of deforestation was completed for the periods 1995-2005 and 2005-2015. For 2015 (most recent Landsat 8 satellite image), a reference map with 6 land cover stratum (forest, savannah, agricultural areas, bare soil, settlements and watercourses) was produced using the same classification approach.
To assess the complementarity of optical Landsat 8 satellite images and radar sentinel 1 C-band, we used Sentinel 1 satellite images acquired from February 2015. The first step is to undertake different pretreatments in order to make the sentinel 1 satellite images exploitable for thematic analysis. Sentinel 1 images were calibrated radiometrically, through the transforming of the amplitude of the signal (numerical count) to a backscattering coefficient (σo) for a ground type which is supposed to be flat. Geocoding was performed in order to correct geometric distortion due to relief. This geocoding process requires the use of a digital elevation model (DEM). The SNAP software (Sentinel ToolBox) makes such processing possible by automatically downloading the DEM (SRTM 30 m) available on the whole of the images.
The images were then georeferenced in the WGS-84 International Geodetic System. This processing treatment enabled the spatial harmonization and comparison with Landsat 8 data. In a second step, photo-interpretation by visual analysis was conducted on the result of the pretreatment of Sentinel 1 images to identify land cover classes extracted by supervised classification of Landsat 8 images.
The historical analysis results obtained from the Landsat Image show very significant deforestation dynamics in the region. The annual rate of deforestation around the VNP is greater than 2.5% over the period 1995-2005 and around 2% over the period 2005-2015. The land cover map based on 6 stratums helped identify each land cover stratum with producer accuracy greater than 96%. Through visual analysis and photo interpretation, the same stratums were identified and digitized on the sentinel 1 satellite.
Automatic extraction tests of land use stratum were performed on sentinel 1 images at different levels of resolution obtained by spatial filtering with filters available in the Sentinel Toolbox. From here to the date of the symposium, we will also test the temporal filtering on newly acquired Sentinel 1 images in the area.

Invasive plant species represent a serious threat to both biodiversity and modern landscapes, and can cause damage to human health as well as have serious socioeconomic consequences. Such species spread rapidly, outcompete native flora and their eradication is problematic. To successfully fight plant invasions, new methods enabling fast and efficient monitoring are needed. By improving opportunities for early detection of invading plants, remote sensing approach can make their management more efficient and less expensive. In an ongoing project in the Czech Republic, we aim at developing innovative methods of mapping invasive plant species (semi-automatic detection algorithms) by using purposely designed unmanned aircraft (UAV) combined with aerial and satellite optical RS data. We examine possibilities for detection of several invasive species: giant hogweed, black locust, tree of heaven and knotweed. Our aim is to establish fast, repeatable and efficient computer-assisted method of timely monitoring applicable over large areas, reducing the costs of extensive field campaigns. We test imagery of different origin (satellite, aerial and UAV), spectral (panchromatic, multispectral and color), spatial (very high to medium) and temporal resolution, and various technical approaches (object-, pixel-based and combined approaches) to choose the best strategies for invasive species monitoring. Variety of data tested enables us to choose the most suitable phenological stage and spatial resolution for the species recognition. In our study, we address trade-offs between spectral, spatial and temporal resolutions required for balance between the precision of detection and economic feasibility. The resulting data enable us to monitoring the efficiency of eradication efforts, assess the invasibility of different types of habitats, model the potential species distribution and identify the drivers of spread. This knowledge will serve as a basis for prediction, monitoring and prioritization of management targets.

In the paper there is presented a review of remote sensing studies carried out for investigation of environmental changes in Albania. Using, often simple methodologies and general purpose image processing software, and exploiting free Internet archives of satellite imagery, significant results were obtained for hot areas of environmental changes. Such areas include sea coasts experiencing sea transgression, temporal variations of vegetation and aerosols, lakes, landslides and regional tectonics. Internet archives of European Space Agency ESA and US Geological Service USGS are used.

An accurate agricultural land use map is essential for many agro-environmental assessments such as irrigated water management. Enhancement of the accuracy of remote sensing based land use maps is still an ongoing process since the development of the first classification algorithms for satellite databases in the 1970s. With the rapid advances in computer technology, Earth Observation sensors and geographical information system, object based (OB) image analysis evolved along with development of many machine learning algorithms. Studies showed that regardless of availability of different classification methods and algorithms no particular method have universal applicability and acceptability. This study aimed to compare different classification methods (object-based and pixel based (PB) algorithms) to contribute to improve agriculture land use mapping using a case study in the arid irrigated croplands of Khorezm in northern Uzbekistan. The comparison is made using two robust non-parametric machine learning algorithms, random forest (RF) and support vector machine (SVM), and a classical parametric algorithm maximum likelihood (MLC) based on the freely available multitemporal Landsat 8 OLI imagery and open source software EnMAP Box (www.enmap.org) and Interactive Data Language (IDL) Program (www.exelisvis.com). Accuracy assessment showed a significant higher overall accuracy (OA) of the machine learning OB-RF algorithm (87.69%) and OB-SVM algorithm (89.23%) over the PB-RF algorithm (78.28%), PB-SVM (79.23%) and PB- MLC (78.51%). The lowest OA occurred with OB-MLC (66.87%). The specific crop class accuracies varied from18.5% to 100%. The OB-RF produced visually appealing agriculture land use map of the area. The results indicate that the OB based machine learning robust non-parametric algorithms have good potential for extracting land use information from satellite imagery captured over spatially heterogeneous irrigated croplands.

Wetlands provide a wide array of ecosystem services; erosion protection in coastlines, habitat services (nursing, breeding and migratory habitats for many species), clean water and recreational services among others. However, wetlands have experienced major transformations in recent history. Strong efforts have been invested in finding out the causes and effects of this global trend. A rather unexplored parameter for studying wetland dynamics is the Land Surface Temperature (LST). LST is one of the most important variables in physical processes of surface energy and water balance from local to global scales (Anderson et al. 2008; Brunsell & Gillies 2003; Karnieli et al. 2010; Kustas & Anderson 2009; Zhang et al. 2008) and understanding wetlands' hydrological processes is fundamental for their effective conservation (Mitsch & Gosselink 2000). Thus, LST has a big potential to act as a continuous and global indicator or proxy of the status of wetlands, their hydrological and evapotranspiration regimes, and how changes in land use and land cover might affect such regimes.

LST retrieval requires several input parameters such as land use specific emissivities and knowledge of the atmospheric profiles at the time of the pass of the satellite. The Satellite-based Wetland Observation Service (SWOS) Horizon 2020 Project, offers a unique possibility of developing a methodology and toolset for retrieving LST operationally, and to study the spatiotemporal dynamics of energy balance and evapotranspiration dynamics of wetlands.

Two of the most popular methods of LST retrieval are tested using the Jimenez-Munoz and Sobrino algorithm (Sobrino et al 2004, Jimenez-Munoz et al 2009) in two study sites; The Camargue, a large coastal wetland in Southern France, and the Lagoon of Fuente de Piedra, a small wetland in Southern Spain. The first one is an NDVI-based method. It is built on the statistical relationship between the NDVI derived from VNIR and TIR bands. With this approach, Land Surface Emissivities (LSE) are unknown. The second method assumes that LSEs are known. For that, the study area is classified using the same Landsat imagery and the SWOS geoclassifier, a tool designed to standardize monitoring and increase comparability of results which will be further developed during the runtime of SWOS. Land cover specific LSE (endmembers) are retrieved from MODIS and ASTER emissivity libraries. Out of the endmembers and land use types identified, we calculate for each pixel a weighted average emissivity value using spectral linear unmixing.

Both LST retrieval methods and compared, showing the relevance of knowing the specific land cover patterns in LST retrieval. Thereby, we give evidence of the importance of continuous monitoring and accurate classification of Earth surface parameters to study global and local energy balances and water dynamics. An operational monitoring of LST supports the development of new indicators and provides inputs for modeling approaches.

Currently there is a lack of knowledge on spatio-temporal patterns of land surface dynamics at medium spatial scale across post-Soviet Central Asia. Specifically, relatively little is known about vegetation variability across climatic gradients (from sub-humid to (semi-)arid) and its response to different intensity of land use (from intensively used irrigated croplands to natural rangelands).  

In this study, we analysed vegetation dynamics at a regional level using the 10-years (2000-2009) of medium spatial resolution (250m) MODIS-EVI time-series data. To explain patterns of vegetation dynamics, we have linked key vegetation parameter (i.e. overall greenness) with its possible direct and indirect drivers across management and climatic gradients using fixed- and random-effects and population-averaged linear regression models. We calculated key seasonal parameters such as (mean, peak and timing of peak) from gridded rainfall, temperature and snow occurrence time-series data using best fit harmonic regressions. In total, 12 regression models were analyzed.

The linear models’ fit was above average for all models explaining the variance in vegetation dynamics of 75 % for irrigated croplands in the northern, sub-humid climate zone and 87 % for natural vegetation in the central, semi-arid to arid climate zone. The results showed the broad control of climatic factors on managed and unmanaged vegetation cover in quantitative terms across all climatic zones. The most varied response of vegetation to climatic variables was observed for intensively managed irrigated croplands, while  almost no difference in response to these variables was observed for managed (non-irrigated croplands) and unmanaged (natural vegetation) vegetation across all climate zones. Natural vegetation and rainfed crop dynamics was mainly associated with temperature parameters, whereas this association was not that pronounced for the irrigated croplands. 

Mapping of hydrological and hydrological elements is critical for understanding the hydrological setting in an arid region. In this study, an integrated approach has been developed, which uses a combination of remote sensing data and geographic information system (GIS) to map factors controlling groundwater recharge, discharge and quality across Abu Dhabi Emirate. The Spectral Angel Mapper (SAM) algorithm, which uses a n-D angle to match the pixels to reference spectra, was used to map new water –bearing rocks and the deterministic eight-node (D8) algorithm, which allows flow to one of only eight neighbors based on the direction of steepest descent, was used to map paleochannels. The flood basin algorithm was used to simulate seawater intrusion from DEM. New lithology, normalized difference vegetation index (NDVI), paleochannels were derived and interpreted indicated that the area was produced by fluvial and eolian process and recharged by local, intermediate and regional flow. The results showed that the Omanand HafeetMountainsare the natural sources of groundwater recharge and HCO₃, Ca, Na and Mg in groundwater. The mapped factors were spatially correlated with hydrologic anomalies observed in groundwater wells. The integrated approach is timely and cost effective and can be used in arid regions for numerical modelling and water balance analysis. 

Fire regimes in Canada’s northern boreal forest are changing, affecting available habitat for barren ground caribou across their southern range. Multiple recent studies have employed successional vegetation modelling to estimate available habitat in the future given different fire regimes. Here, we present an alternative method for assessing caribou habitat recovery post-fire by examining caribou behaviour pre-fire and post-fire across five decades. We expect complete avoidance by caribou immediately post-fire, followed by recovery of travel focused behaviour then foraging behaviour. Disturbances were mapped using best available pixel Landsat data from 1985 to 2012 with proxy pixel values used to fill data gaps and random forests used to attribute disturbance types. Caribou behaviour was examined using step length and turning angle from 274 animals across five herds. Behaviour post-fire was assessed using general additive mixed models. Pre/post fire analysis demonstrated strong avoidance of burned locations, confirming our first hypothesis. Behavioural recovery, however, occurred much earlier and in a more complex fashion than hypothesized. Burned locations were used in spring, fall, and winter as early as three years post fire. In spring and fall, travel did recover prior to foraging; however, behaviour in winter showed little response to time since fire. Our approach allows for a more direct estimation of caribou habitat recovery post fire while considering different behaviours and seasons. Importantly, our results indicate that use occurs earlier than predicted in successional modelling, suggesting an overestimation of habitat loss when using successional models to project future caribou habitat.

In rural landscapes, grasslands are a major source of biodiversity. They provide many ecosystem services. In particular, they are the main livestock feeding resource, through grazing and forrage production. Semi-natural grasslands host a rich species composition that is impacted by the management practices. Indeed, the anthropic events on the grasslands, like mowing and/or casual grazing, disturb the natural cycle and the structure of the vegetation. Therefore, it is essential to identify the management practices in each parcel in order to predict their effect on biodiversity and related ecosystem services.

In this context, remote sensing appears to be an adapted tool to characterize grasslands at the landscape scale, because of their large spatial coverage and their revisit frequency. However, the reflected signal of the grasslands species mix and diversity is more difficult to interpret compared to mono-specific lands like crops. Furthermore, grasslands are relatively small elements of the landscape (in the order of the hectare), which require high spatial resolution data to be distinguishable. Given their phenological cycle and the punctuality of the anthropic event (i.e., mowing), very dense time series through the vegetation cycle are necessary to identify the management types.

Until recently, satellites mission offering high frequency of revisit had low spatial resolution (i.e., MODIS), and high spatial resolution mission did not offer dense time series. Now the sensor capacities have been improved in the field of terrestrial observation, in particular through the ESA Copernicus Programme. Thus, these new missions offer new possibilities for grasslands monitoring.

The aim of this study is to identify the management practices of the grasslands through a satellite image time series (SITS). We used a Formosat-2 (8-m spatial resolution) series composed of 17 multispectral images through the year 2013. The study site is located in south-west France, near Toulouse, in a semi-rural area where livestock farming is in decline. The dataset is composed of about 50 parcels with diverse management methods. The management practices were determined from interviewing the farmers. We identified 4 management types during this vegetation cycle: grassland mown once, grassland mown twice, grazing and mixed management (mowing then grazing). We used them as classes for the classification. To remove the noise in the SITS, we smoothed the NDVI of the pixels applying the Whittaker filter. Then we processed a statistical supervised classification using Support Vector Machines.

The classification performed well, with an overall accuracy of 71% and a kappa of 0.58. The results were improved by grouping the 4 classes into 2 similar classes: mown grasslands (independent of the number of cuts) and grazed grasslands. The overall accuracy was 82% with a kappa of 0.65.

We concluded that it is possible to discriminate the grassland management types with a SITS of 17 images. We suppose that the classification could be improved using a more dense SITS like Sentinel-2 will present. Moreover, we performed the analysis on the NDVI, that is limited to the effect of only 2 bands (Red and NIR). Sentinel-2 will offer more bands, including one in the red-edge that is useful in vegetation characterization. The work will be continued in the aim of determining the mowing and grazing dates. We assume that the future Sentinel-2 time series will allow more precision in the detection of these events and the measurement of their impact on vegetation structure and composition.

Near-surface air temperature (T_a) is a critical variable in the energy and water cycle at earth surface and have immense significance in agro meteorological applications such as crop growth modeling, water requirement mapping, pest-diseases surveillance and climate change studies. Air Temperature is usually obtained from meteorological observatories but paucity of dense network of such met station limit regional applications. Remote Sensing data can help to solve this problem particularly in non-weather station areas at global and regional scale. Over the past few decades, various approaches viz., statistical approach, Energy balance approach, TVX (Temperature and Vegetation) approach and Neural Network approach are being developed with use of satellite observations. In the present study, we explored applicability of TVX (Temperature and Vegetation) approach for estimating air temperature at satellite overpass in Gujarat state. Satellite data comprised combination of 8-day composite of surface reflectance and land surface temperature (LST) products from MODIS on selected seven periods in year (Julian days 032, 129, 265 and 305) of year 2008. Land surface temperature from day-time satellite pass was used in this study. Observed air temperature at satellite pass time from four ISRO AWS station was used for validation purpose. The TVX technique (T_air = a+b*NDVI_max)  employed  on 9x9 moving window and corresponding  intercept and slope of relationship between LST and normalized difference vegetation index was obtained.  The relationship obtained for each moving 9x9 window was extrapolated to maximum NDVI (NDVI_max) obtained for that window in order to compute air temperature.  Maps of air temperature at Terra satellite overpass were generated for four Julian day period. For validation, air-temperature values extracted corresponding to geographical coordinates of each of four stations. Overall results on instantaneous air temperature mapping were satisfactory for initial three months i.e. January, February and March owing to presence of
adequate vegetation cover. The error in estimates of air temperature goes high when agricultural crops get harvested in April. Overall comparison between observed and estimated values showed a root mean square error (RMSE) of 1.1°C.

The goal of this research is to test the hypothesis that narrow-band spectral data is superior for the classification of Arctic vegetation species and functional groups, as well as vegetation pigments and chemistry at the plot, landscape and ecosystem scale. Using high spectral resolution ground-based spectroscopy I asked the following questions, (1) how consistent species spectral responses are among and between vegetation communities; (2) how consistent the background or non-vegetated spectral response is, and (3) do functional groups, species associations, or community types exhibit distinct spectral signals. Ground-based spectral data was related to the following biophysical properties: chlorophyll α, chlorophyll β, carotenoids, leaf nitrogen concentrations, leaf carbon concentrations, and leaf area index. A spectral mixing technique was then used to determine how biophysical properties are affected by the spectral response of other ecosystem and optical characteristics such as water content, shadow and vegetation composition. We expect that the biophysical properties of chlorophyll α, chlorophyll β and leaf area index will be easily identifiable in highly mixed pixels. Carbon and nitrogen content we expect will be a more difficult to identify with increasing heterogeneity. The results of this research will inform current and future remote sensing missions on how to assess vegetation heterogeneity and biophysical properties of the Arctic under a changing climate regime.

-          Because of the fine spectral and spatial resolution, improved atmospheric correction, wide geographical cover, and three-day repeat cycle, MERIS has become a valuable sensor for capturing measurements and monitoring of terrestrial environments at regional and global scale, and accordingly two land biophysical variables has been designed specifically for MERIS sensor, MERIS Terrestrial Chlorophyll Index (MTCI), and MERIS Global Vegetation Index (MGVI), those variables play an important role in ecosystem modelling and environmental studies and contribute to our understanding of biogeochemical fluxes and global climate. Hence, Proper usage of such products requires that the corresponding uncertainty information needs to be provided, furthermore. Validation of MERIS land variables is important across a range of vegetation types, spatial scales, and different environmental conditions. As the performance of MERIS land variables needs to be fully understood to generate consistent products between sensors. For the launch of the ocean and land color instrument (OLCI) on board of Sentinel-3 mission expected by late 2015, providing continuity of MERIS measurements. Investigate MERIS capabilities will encourage the use of OLCI data also for land applications like monitoring of vegetation dynamics over long time periods.

Evaluating MERIS products MTCI and MGVI at global scale, to quantify the uncertainty corresponds to each of the hey biomes at  global scale, in order to provide further understanding on the sensitivity of these variables in the preparation for the launch of Sentinel3 mission. By estimate MTCI and MGVI distribution for each biome using BELMANIP2 network of sites at global scale for the period of MERIS sensor life time 2002-2012. As well as investigate the temporal stability (consistency, continuity) of MTCI and MGVI over the study period 2002-2012. That has led to evaluate the global correlation between MTCI and MGVI. Also using temperature data as independent climatic variable to evaluate MERIS products on global scale from 2002-2012.

Different biome types, seasons, and latitudes has dominated the variation and uncertainty of MERIS variables. Both variables displayed good temporal continuity over the study time ideally with no gap. MGVI tend to show more consistency in the reflectance values over tome than MTCI. MERIS variables showed great correlation in the northern hemisphere. Both variables showed high dependency on temperate as climactic variables in the northern hemisphere. The study highlighted both potential and challenges of applying MTCI and MGVI in further global studies relating to spatial cover, latitude, vegetation season, and climactic variable, with assessment of the use of CEOS-BELMANIP2 network of sites. This methodology, analysis, and results could be used as framework for the evaluating of the products coming from OLCI on board of Sentinel-3 mission and will be based on MERIS MTCI heritage.

In recent years severe drought events have become more intense and frequent and their impacts have been exacerbated due to increasing water demands. Especially in semi-arid and arid regions of Africa the duration and intensity of the rainy season determines the success or failure of crop harvests. Long lasting dry spells have devastating consequences for smallholder farmers and although crop insurance instruments exist, they are suffering from significant constraints like a lack of appropriate weather station coverage and adequate yield information. Hence, the introduction of crop insurance index products is limited and their spatial applicability is reduced. Remote sensing techniques, however, are offering new opportunities for the development of innovative insurance instruments for indexation on a global scale.

Here we present a crop insurance index product that is based on microwave observations of soil moisture derived from the European Remote Sensing Satellite (ERS-1 & ERS-2) and the Advanced SCATerometer (ASCAT) for four regions across Senegal. The insurance index product design was implemented to consider information on planting, sowing and harvesting periods of the main staples namely groundnut, millet and maize in specific agro-ecological zones as provided by the Food and Agriculture Organization (FAO) Crop Calendar. The long-term average soil moisture (1992-2000 and 2004 - present) was related to the yearly soil moisture measurements to derive the annual soil moisture deficit for different periods of the crop-specific life cycle, e.g. sowing, vegetation and harvesting period. The soil moisture deficit was then correlated with the annual crop harvest to determine the period that is essential for crop growth and harvest. Our analysis revealed that for groundnut and maize the vegetation period was essential (R=0.7 and R=0.64), while millet was sensitive to changes in soil moisture during sowing and planting period (R=0.88). Historic payouts were generated based on the long-term mean and minimum values of the soil moisture deficit and corresponded well with severe drought years in the region. To give an example, our indicator covered well the severe production and yield losses in groundnut for the 2007/2008 cropping season as well as below-average harvests for the year 2014. Using the average historical pay-outs, the Expected Loss Curve (ELC) was derived for each region and crop type assuming a value of 6% for groundnut, 8% for maize and 4% for millet, respectively.

Although our preliminary results are only based on four test sites in Senegal, we are confident that our surface soil moisture based crop insurance index provides a good estimate on a regional scale. Our preliminary results are promising, but further in-depth research is needed to confirm the reliability of such an approach and should be steered towards the integration of Sentinel-1 datasets which will allow the application on the local level.

The terrestrial ecosystems of the northern high latitudes are subject to significant changes during the last century and highly vulnerable to modifications in the global climate system. Predictions from Global Climate Models (GCM) have shown increasing trends of global temperature for the 21th century. Changes in temperature, as well as precipitation, are having an impact on snow cover, vegetation productivity and coverage, vegetation seasonality, surface albedo, and permafrost dynamics. Especially the recruitment of trees into the northern regions, which are controlled by summer temperature and the length of the growing season, is of high importance for the Global Climate System.

The coupled climate-land surface changes in the arctic are thought to be a positive feedback in the Earth system, which can potentially further accelerate global warming. In contrast to other regions, the northern high latitudes are supposed to face the strongest and largest feedbacks for the upcoming decades. Hence, the arctic region has become a major focus in Earth system science and climate related research topics. The identification of significant modifications of the state and spatio-temporal dynamics of arctic climate and land surface parameters, such as temperature, precipitation, snow and vegetation is a challenging issue. Earth observation information from various sources provides a useful tool to observe and monitor essential climate parameters while retrieving spatial information over large areas and compensates for the lack of ground measurements in remote areas. Therefore, the use of remote sensing information has gained high importance for climate change related research studies during the last decades.

The co-occurrence of temperature, precipitation, snow cover (ESA DUE GlobSnow), and vegetation greenness trends in the pan-arctic region between 1981 and 2012 has been analysed using Earth observation products and ground based measurements.

Based on coarse resolution remote sensing time series data as well as in-situ based data products it was possible to identify regions showing congruent and divergent trends of various essential climate parameters. In general, precipitation significantly increases during summer and fall. Temperature conditions in winter are increasing between 1981 and 2012. The parameter snow water equivalent has shown the highest trends during spring and fall. Vegetation greening trends are characterized by a constant increase during the vegetation growing period. Largest dependencies between the parameters are found in spring, fall and winter. The co-occurrence of snow water equivalent and temperature trends in the transition month between summer and winter indicate to have the largest impact to the arctic environment during the last decades.

High spatial resolution Earth observation data have been utilized to identify land surface changes in the taiga tundra ecotone for a selected test region in northern Siberia between 1973 and 2012. This region was identified to be characterized by the most significant and largest trends during the last 40 years. By using high spatial resolution remote sensing data, an intensification of the woody vegetation cover have been found at the arctic tree line near the Taymyr peninsula. These climatic and ecosystem changes represent the multi-scale feedbacks in the arctic environment.

Hydrological conditions including fresh water supplies in mountainous areas are mainly controlled by runoff regimes and its components. For many semi-arid and arid regions, the falling precipitation is allocated to different runoff components including surface and subsurface flow. Runoff generation in mountainous basins are the primary water sources for the large populations and critical components for the continuum of the ecosystems. Therefore determining the amount of runoff using sophisticated techniques is an important benchmark and a necessity in order to use and manage the present and future water resources.

The recent trends in hydrological modeling resulted in the improvement of fully distributed snow models. Spatially distributed hydrological models attempt to quantify the runoff processes by subdividing the catchment into a number of units. The main objective of this study was to carry out a quantitative analysis of catchment runoff generation at basin scale in a semi-arid Mediterranean environment. Goksu Watershed located at the Eastern Mediterranean Part of Turkey is selected as the study area.

The methodology consisted five steps; i) spatial and temporal data analysis, ii) model calibration, iii) implementing process-based J2000 modelling suite, iv) model validation, v)  assessing the spatio-temporal variability of runoff dynamics by interpreting the modelling results.

The J2000 is a conceptual hydrological model. It requires air temperature, vapour pressure, wind speed, precipitation, solar radiation, terrain and land cover variables, such as slope, aspect, soil texture, forest cover to simulate hydrological Dynamics at different scales. For the long-term, seasonal, distributed runoff generation of the Goksu Watershed, only a distributed conceptual modeling approach is feasible, given the availability of input data. The main advantage of the model is the ability to initiate model runs and directly compare model predictions with products derived from remote sensing and field measurements. Hydrological quantities such as runoff components are critical factors that can affect the regional ecosystems. Thus, this study has a significant importance to estimate these factors quantitatively that can directly contribute to the regional planning strategies.

For the purpose of agricultural monitoring, the study of vegetation phenology by means of remote sensing has been largely addressed by the analysis of temporal variations in vegetation indices acquired by optical sensors. These indices are formed by combining different spectral bands which exhibit specific sensitivities to vegetation parameters, acting as a proxy for monitoring plant growth. Optical imagery has gone some way to supporting the precision farming industry, however particularly over the UK, cloud prohibits regular delivery of information required by the farmer. To date, the use of radar sensors has been explored in the literature but is less developed for operational agricultural monitoring activities. Many theoretical and experimental studies have demonstrated the sensitivity of microwave backscattering to geometric and dielectric properties of vegetation, which can be linked to parameters such as biomass and crop type. However, due to the costs associated with data and the infrequency of data acquisition, there is huge failure to commercially exploit this technology for agricultural applications. ESA’s Sentinel-1 mission now offers the ideal platform to support application development, with focus on reliability, operational stability, global coverage and zero cost.

Work has been conducted to explore the potential of Sentinel-1 data for the retrieval of information on the phenology of cereal and oilseed crops. Specific activities have included analysis of SAR backscatter coefficients and the exploitation of polarimetry and interferometry techniques, correlating these observables to extensive ground truth data collected in South-West England (approximately 700 data points) from March 2015 up to July 2015.  A total of 43 SAR images were acquired from February to July, spread across four different tracks, two ascending and two descending. In particular, both the TOPS Level-1 GRD and SLC products have been employed. The ESA open-source Sentinel-1 and PolSARpro toolboxes have been exploited to create processing chains and bash LINUX code has been written to enable automation of some of the processing, delivering:

Backscattering coefficients (sigma0 and gamma0), both VV and VH polarisation.

Interferometric coherences.

Entropy, Anisotropy, combinations of Entropy-Anisotropy, and Entropy-Shannon dual polarimetric observables.

Once generated, the SAR derived observables were averaged over 24 different fields (winter wheat, winter oil seed rape, winter barley, summer barley) and analysed in relation to temporal behaviour and sensitivity to crop parameters. By using VHR optical data, the field’s shapes were carefully selected to avoid the inclusion of objects which are not crop vegetation, such as trees, power lines or sheds. Analysis of the SAR derived observables with respect to the ground data demonstrated the following:

geographically consistent temporal trends that can be distinctly attributed to different crop types;

different sensitivities from multiple polarizations that can be identified as growth parameters;

good discrimination of certain phenological change from coherence measures;

good discrimination of certain phenological change from entropy and entropy shannon parameters which supplement backscattering information;

further understanding of phenological state by accounting for different backscattering mechanisms and

consistency in results from Sentinel-1 TOPS mode when compared with stripmap mode images.

Consideration of these parameters has led to reliable identification within the data, of significant growth stages that farmers rely on identifying for accurate chemical application.  This has important consequences for the precision agriculture market and potential enhanced service delivery building on information already obtained from optical satellite observations.

Satellite imagery has been used since the 1960s to monitor agricultural activities at a government and international level. It is only now that the image quality and of cost feasibility have made satellite imagery available to individual farmers in application to crop quality analysis and other measures relating to precision farming. The underlying reasoning for the propulsion forward of this technology to the level of individual producers has been in the name of food security; to use agricultural intensification through precise monitoring of field factors including plant rigor, pests, water needs, weeds, soil quality, etc. However, when looking at the goals of common agricultural policy at large, satellite imagery availability contributes greatly to some, while detracting from important aspects such as maintaining small farms and rural infrastructure, as well as encouraging agrobiodiversity and it’s acknowledgement as a main facet of global food security. Through a technological assessment, both the actors and supply chains that enable satellite imagery to be available to individuals, as well as the social impact and ultimate impact on global food security of this new development will be presented.

It is with this in mind that this study will, in the first section, explain the development of the privatisation and eventual commercialisation of satellite imagery in agriculture in Germany. Second, present an overview of applications currently available for small farmers and channels of communication for adoption. Third, an examination of the oppositional relationship between agrobiodiversity and precision agriculture. Then finally, a summation of the implications of this new practice on a social level with regard to small farmers and rural infrastructure, as well as agrobiodiversity and food security. 

The method study is a classical technology assessment outlining potential issues and background workings to inform policy making with regard to satellite imagery availability and regulation for individual agricultural use. As such, although the study is focused on the German context, the study maintains global perspective and is future-oriented.

Key findings are:
-Satellite imagery analysis has successfully been used to include measures of agrobiodiversity, as demonstrated by Karl Zimmerer’s research using remote sensing to study potato varieties in South America. While this technology and method exists for research purposes, it has not yet been either marketed or incentivised by precision farming companies or through agricultural policy.

-The use of satellite imagery for field and crop analysis has not yet been divorced from industrial agriculture. While NDVI imagery can reveal spots on a field that require fertilizer, pesticides, or herbicides, it is still mainly with synthetic and chemical use. There is no reason why this can’t be a gateway to enable the application of organic fertilizers, pesticides, and herbicides as by definition, less field inputs makes it more affordable, and the practice of precision agriculture is advertising environmental sustainability as a main feature and benefit of use.

In this study spatial forest resource planning model developed by systems analysis and human ecology approaches and applied in the Chigou - Bideleh traditional region in the forests of central Zagros. This model has been done in 5 stages, including: Analysis of traditional region situation, assessment of ecosystem services, assessment of demand of forest resources, selection of management- services units. Sustainability and determine rural developing showed this region is in unstable and developing. Identification of forest function and services was done base on separation structure of natural and human sector, 54 ecosystem services was classified to supplying, regulating, habitat and cultural macro services. Three priorities of forest services that in this research were assessed including: water production, soil conservation and outdoor recreation. Outdoor recreation ability model was assessed by 7 criteria and 14 indicators in silviculture criteria group and 4 criteria and 7 indicators in physical resources criteria group.   Results showed with cover canopy percent increasing, soil erosion is reduced, average Annual soil erosion in forest with 52, 35, 15, 3 cover canopy percent is in order 0.53, 3.3, 7.7 and 26 tons per hectares. According to results in a per hectares of forest land, low density forest, moderate forest and high density forest 1, 105, 150 and 188 m3 Water is produced. Results of demand forest resources showed, average wood consume for a rural household in one year is 5.66 m3. Annual household income and the degree of rural development were used two impellent factors for determine of demand forest resources, with household income and the degree of rural development increasing, demand for wood consume is reduced. Increasing of water production and decrease of soil erosion with response to local people demand were   optimization objectives that done in this research.  

Current and future terrestrial ecosystems interact strongly with climate by feedback mechanisms linked to their biophysical and biogeochemical functioning, including their carbon balance, and the exchange of water and energy between the land-surface and the atmosphere. Ecosystem models are an integral part of monitoring and managing terrestrial carbon stocks, but their predictions generally suffer from high uncertainty, principally resulting from model parameterisation errors and forest initial condition errors. Information on forest structure and composition has traditionally come from ground-based inventories of the plant canopy within small sample plots. In contrast, remote sensing offers the promise of large scale, spatially-consistent information on key aspects of ecosystem state such as above-ground ecosystem composition that can be used to constrain regional and global scale simulations of the terrestrial biosphere. Yet, remote sensing is usually used to define single metric attributes per pixel, e.g. forest height, above-ground-biomass, average trunk diameter, coarse forest composition. First, radar- and lidar-derived measurements of above-ground biomass and canopy height are used to successfully constrain the predictions of the Ecosystem Demography (ED2) model at the La Selva tropical forest site in Costa Rica (~8% reduction in net carbon flux error compared to simulations initialized with ground measurements and from a potential vegetation simulation). A closer constraining of ecosystem dynamics in ED2 model, though, resulted first from inclusion of forest composition and included sub-pixel information on trunk spatial density. Subsequently, waveform lidar in conjunction with imaging spectroscopy were used at the temperate Harvard Forest, MA, to define the full fine-scale forest structure and composition (tree size class distribution with plant functional type). Initializing using the ED2 model reduced net carbon flux errors by 50% compared to flux tower observations. This fusion of lidar and imaging spectroscopy also fared better than widely used lidar and radar-derived forest structure methods with an increase in net carbon flux uncertainty of 6-23%. These results suggest that terrestrial biosphere model simulations can utilize modern-remote sensing data on vegetation composition and structure to improve their predictions of the current and near-term future functioning of the terrestrial biosphere.

The knowledge of the radar response of terrain at different incidence angles would be used to develop discrimination approaches to land cover classification from radar sensors. Due to multi-angular SAR measurements increase the observing dimension, analysis of extracted parameters from multi-incidence angle datasets has been extensively investigated. It makes possible to distinguish different forest types through the use of their characteristic extracted parameters in multi-incidence angles properly.

Formulaic analyzing methods for polarimetric synthetic aperture RADAR (PolSAR) data such as scattering or covariance matrix show polarimetric information just in a confined number of polarization bases, whereas radar responses has information on wide range of polarizations.

Polarimetric signatures which illustrate the extracted parameters at all combinations of transmit and receive polarizations could be solved the problem. This illustration of the target responses could be investigated details of physical information from target backscattering in various polarization bases appropriately.

In the first part of this paper, the influence of incidence angle on the polarimetric signatures in different forest types was examined and then Extremely Randomized Trees (ERT) classification approach based on the polarimetric signatures of multi-angular PolSAR datasets is proposed.

Polarimetric signatures for various PolSAR features extracted from the parameters which obtained directly by original data (Scattering matrix elements, Coherency matrix elements, Covariance matrix elements), well-known decomposition methods (Pauli, H/A/Alpha, Freeman-Durden) and the SAR discriminators (Co-polarization Phase Difference, Span, Depolarization Index).

The Multi-angular Radarsat-2 datasets including FQ4, FQ9, FQ14 and FQ18 were acquired at C band in Fine Quad polarization mode with a nominal swath width 25km, spatial resolution 5.2 m in range × 7.6 m in azimuth. Also, polarimetric signatures from different classes, including white pine (Pw), red oak (Or), black spruce (Sb), urban (Ur) water (W) and ground vegetation (GV) was analyzed by the proposed method.

In the first part, the response for various polarization basis obtained by polarization basis transformation of covariance matrix. Then, the polarimetric signature represented by a two dimensional matrix. The ellipse angles are varying with increments of 10◦ (−45◦ to +45◦ for the Ellipticityand 0◦ to 180◦ for the Orientation Angle), the size of polarimetric signature matrix is 10×19. For easier use of them in pattern recognition methods, these matrices are treated as vectors consisting of 190 components. For comparing the influence of incidence angle on the polarimetric signatures the similarity value of polarimetric signature of a pixel in one of the images to polarimetric signature of the same pixel in another image is calculated by the Normalized Euclidean Distance and Normalized Signature Correlation Mapper between the polarimetric signature of the one pixel in one image with the corresponding signature of the same pixel in another image. Then in order to investigate the potential applications of these comparison results, ERT classification algorithm has been proposed. It essentially consists of randomizing both attribute and cut-point choice while splitting a tree node. It builds completely randomized trees whose structures are independent of the output values of the learning sample, in the extreme case. Besides accuracy, the main strength of the resulting algorithm is computational efficiency which is appropriate for polarimetric signatures of multi-angular dataset.

The results of compariosion part, show polarimetric signatures of various PolSAR features in different incidence angles, introduce new concepts of various targets which are appropriate for interpreting forest types. In classification part, the proposed ERT algorithm with using polarimetric signatures in multi-angular PolSAR data, has a better results in compare with using parameters which obtained directly by original data.

We estimate the total land water storage change between 2003 and 2013 using the global water mass budget as well as the sea level budget approach. We compare the ocean mass change from GRACE space gravimetry to the sum of the main water mass components of the climate system: glaciers, Greenland and Antarctica ice sheets, atmospheric water and land water storage (the latter being the unknown quantity to be estimated). We compute the mean trend over the study period and estimate it to be -115 +/- 68 km³/yr in terms of net land water storage decrease equivalent to +0.32 +/-0.19 mm/yr sea level rise, respectively. To check the consistency of the results we also use the sea level budget approach in which the ocean mass component is estimated using the ESA CCI-based global mean sea level corrected for the steric component (using Argo and the ORAS4 reanalysis). A very good agreement with GRACE-based ocean mass is found, thus on the net land water storage trend. In another step, we account for acceleration terms seen in several components of the budget equation (e.g., ocean mass, Greenland, Antarctica and glaciers mass balances). We derive a time series of land water storage rate and find significant acceleration over the study period. The computed mean rate in global land water storage mainly represents direct anthropogenic effects on land hydrology, i.e. the net effect of groundwater depletion and impoundment of water in man-made reservoirs, and to a lesser extent the effect of naturally-forced land hydrology variability. Our results compare well with independent estimates of human-induced changes in global land hydrology.

Desertification, a phenomenon of loss of productivity of the land is both a matter of Environment and Development (Cornet, 2002). It is linked to the anthropogenic action and to climate variability but also to changes in biodiversity, in particular the Maghreb (Hobbs et al., 1995). The desertification of the steppe areas of North Africa (Algeria, Morocco and Tunisia) is considered of special concern by the specialists in these regions. Desertification, Climate Change and the erosion of biodiversity are the central issues for the development of arid, semi-arid. In this region, the combination of two factors, climatic and anthropogenic, has fostered a deterioration of the vegetation cover, soil erosion and the scarcity of water resources. The climate of this region is characterized by periods of recurring droughts since the 1970s. The anthropogenic pressure is the result of a combination of factors among which the strong demographic growth, the intensification and extension of production systems agro-pastoral or still further the concentration of a growing livestock on smaller spaces. In this study, the criteria for classification and identification of physical parameters for spatial ecological analysis of vegetation in the steppe region to determine the degradation and vulnerability vegetation formations and how to conduct to phenomenon of desertification. So we use some satellite data in different dates (LANDSAT) in order to determine the ecological of steppe formation and changes in land cover, sand moving and forest deterioration. The application of classification and some arithmetic combination with NDVI and MSAVI2 through specific processes was used to characterize the main steppe formations. An ecological analysis of plant communities and impact of sand move describe the nature of the desertification phenomenon and allow us to determine the impact of factors of climate and entropic activity in the Algerian steppe.

With agriculture remaining as one of the key aspects of a strong society, the ever decreasing crop yields, attributed to climate change, urbanization and the accompanying human practices, have been quite of a concern. In its 2012 Economic Survey Report, the Kenya National Bureau of Statistics (KNBS) estimated that there was a decrease in maize production by nearly over one million bags and an increase in the maize import by about 56.5 percent. There is therefore the need to emphasize and optimize reliable agricultural production within the country. One the main objective in determining the maize crop yield in Trans Nzoia region of Kenya is to extract the Leaf area index and biomass values from remote sensing imagery. Remote sensing is being utilized as it provides larger ground area coverage in a shorter amount of time as opposed to direct ground measurements. The high resolution RapidEye satellite offers spatial resolution of 5 meters making it appropriate for the at field scale assessment. Images covering the sampled maize fields for the months of April to November will be analyzed as these cover the main maize growing seasons in Trans Nzoia. From the ground referencing information, the sampled maize fields differ in terms of the sowing times, crop development, crop management and harvesting times. Satellite derived information such as vegetation indices will be evaluated against field scale crop development indicators namely the Leaf Area Index and biomass from satellite imagery acquired at a specific phenological stage of the plants cycle dictated by the BBCH code. Subsequently, the indicators will be applied as inputs parameters to initialize the crop model (DSSAT V4.5 or APSIM). The crop models aid in the prediction of the maize crops yield in the different fields in Kenya. The validity of the model’s performance will be tested by applying it in other maize growing fields. This information is crucial in decision making with respect to ability of the local production to feed its population.

The use of satellite soil moisture data is limited in the field of watershed hydrology due to lower spatial resolution. In this study, we have disaggregated the surface soil moisture data from Soil Moisture and Ocean Salinity (SMOS) using two fine resolution datasets of evaporative fraction and land surface temperature (LST). Evaporative fraction (EF) is estimated by applying surface energy balance system (SEBS) using spatial land surface parameter products from Moderate-Resolution Imaging Spectroradiometer (MODIS) and point in situ meteorological datasets. We applied a linear downscaling method to disaggregate SMOS dataset. Disaggregation of SMOS data were done from September of 2011 to end of 2012. The disaggregated SMOS soil moisture data were validated with 10 and 22 in situ soil moisture monitoring stations located in Mizewa, Lake Tana basin, Ethiopia and Remedhus, Zamora, Spain, respectively. The result indicated that, liner downscaling of SMOS data with evaporation fraction improved the SMOS data, it has captured the in situ soil moisture variation with  6% error and the SMOS downscaled with a land surface temperature captured the in situ measurements with 20% error and with a higher variation in dry and wet seasons.

While surface albedo retrieval from polar-orbiting satellites is a well-established field of science, it is commonplace to utilize data from only a single satellite or a family of similar satellites for this task. We propose that combining the observations from different satellite families (such as AVHRR and MODIS) is physically justifiable and that the approach we present will lead to surface albedo retrievals at higher temporal resolution than the state of the art, with comparable or better accuracy. This study is carried out in the World Meteorological Organization (WMO) Sustained and coordinated processing of Environmental Satellite data for Climate Monitoring (SCOPE-CM) project SCM-02.

Our approach is to first spectrally homogenize the observation datasets , thus ensuring that the observations are comparable. Then, both datasets are atmospherically corrected with a homogeneous atmospheric profile and algorithm. The resulting surface reflectance observations are then used as input to a surface bidirectional reflectance distribution function (BRDF) inversion algorithm, the results of which may be integrated to yield various surface albedo quantities. A key principle here is that the larger number of valid surface observations with multiple satellites allows us to invert the BRDF coefficients within a shorter timespan, thus enabling the monitoring of relatively rapid surface phenomena such as snowmelt.

Here, we present results from a demonstrator dataset of a global combined AVHRR-GAC and MODIS dataset covering the year 2010. The retrieved surface albedo is compared against quality-monitored in situ albedo observations from the Baseline Surface Radiation Network (BSRN) and the Greenland Climate Network (GC-Net). Additionally, the combined retrieval dataset is compared against AVHRR-only and MODIS-only albedo datasets to assess the quality of the multiplatform approach against current state of the art.

The proposed multiplatform approach is expected to bring benefits in particular to the observation of the albedo of the polar regions, where persistent cloudiness and long atmospheric path lengths present challenges to satellite-based retrievals. Following a similar logic, the retrievals over tropical regions with high cloudiness should also benefit from the approach.

The proposed approach is not limited to AHVRR and MODIS observations. Provided that the spectral homogenization produces an acceptably good match, any instrument observing the Earth's surface in the visible and near-infrared wavelengths could, in principal, be included to further enhance the temporal resolution and accuracy of the retrievals. The SCOPE-CM initiative provides a potential framework for such expansion in the future.

Large wetlands are an important component of the global water cycle and the knowledge of water flow and storage dynamics within these regions is valuable for many applications such as flood risk assessment and water availability studies. Most of the inundation areas are remote regions without significant infrastructure, especially without in-situ gauging observations. Remote sensing techniques can help to provide highly valuable information for hydrological questions.

Radar altimetry had been designed to provide highly accurate measurements of sea surface heights over open oceans on the global scale. However, nowadays, satellite altimetry provides also inland water level heights, mainly for lakes, reservoirs, and rivers. It can also be used to monitor water heights in inundation areas. Remote sensing images from different frequency bands (optical as well as SAR) can provide information on the surface water extent in inundation areas. Combining both measurements types allows for the quantification of water volume and its temporal changes.

In this contribution, we derive water level heights from radar altimetry and surface water extent from SAR images in the Pantanal region. The Pantanal is one of the largest wetlands worldwide. It is located in South America and comprises an area of about 400 by 250 km. The region is affected by strong seasonal inundation and desiccation with phases with standing water and phases with subsurface water table below the rooting zone. Basic methodology for the derivation of water level variations as well as water extent variations will be presented, and results for the years between 2006 and 2011 will be shown, analyzed, and compared with each other.

The TanDEM-X mission, consisting of two twin X-band synthetic aperture radar (SAR) satellites operated by the German Aerospace Center (DLR), has been active since 2010. Over the years, the satellites have operated in several configurations tailored to suit the different mission phases, such as the generation of the digital elevation model of the Earth’s surface. In 2014, the TanDEM-X mission entered the science phase aimed at supporting the creation of innovative SAR data processing techniques and applications. The configurations include a period were the satellites are separated by physical baselines in excess of those available before. The long baseline phase of the TanDEM-X mission provide access to unprecedented SAR data. Across-track InSAR acquisitions with effective baselines of up to 3 km enable the investigations of volumetric decorrelation effects in short vegetation. These investigations, in turn, help to set up a foundation for future applications focused on the monitoring of agricultural crops and grasslands. Since the launch of the TanDEM-X mission great advances have been made in estimating forest parameters from polarimetric InSAR data. Techniques developed based on the short baseline TanDEM-X data can now be adapted, where possible, for longer baselines. Ability to detect the height of crops and grasslands in the future from SAR data would be invaluable for governments and farmers alike.

This study presents the first results from a campaign carried out in the summer of 2015 in Estonia. Times series of TanDEM-X acquisitions with effective baselines ranging from 1300 to 1800 m are analyzed in respect to vegetation parameters collected during a field survey covering six grassland and eight crop fields. Crop fields included three winter wheat fields, three rape seed fields, and two maize fields. The grasslands were covered with different vegetation species and exhibited different levels if species homogeneity. The field survey provided data about the vegetation height, wet and dry above-ground biomass, plant relative water content, and soil moisture. All instances of the field survey were conducted simultaneously with the TanDEM-X acquisitions.

In this study, the possibility to detect the vegetation height from the complex correlation coefficient calculated from the long baseline TanDEM-X data is investigated. A total of 55 acquisitions in three configurations are analyzed spanning the time between April and September. The given TanDEM-X configuration provides height of ambiguity values ranging from 1.8 to 4.2 m, allowing, in theory, height inversion form the complex correlation coefficient. Several approaches are discussed, including the use of the Oriented Volume over Ground model as well as the “sinc” approximation. The different components making up the complex correlation coefficient are discussed.

EUSTACE (EU Surface Temperature for All Corners of Earth) is a Horizon2020 project that will produce a spatially complete, near-surface air temperature (NSAT) analysis for the globe for every day since 1850.  The analysis will utilise data from both in situ and satellite observations, which will be blended using novel statistical techniques over all Earth surfaces.  The inclusion of satellite data will enable EUSTACE to capture NSAT variability over regions that are currently poorly observed in situ, and are therefore not well represented in existing global NSAT analyses.  As satellite surface temperature observations are sensitive to the Earth’s skin temperature rather than NSAT, the satellite data must be transformed into an NSAT estimate before being ingested into the EUSTACE analysis.  Over land, land surface temperatures (LST) from ESA’s GlobTemperature project will be used, primarily from the MODIS and SEVIRI sensors.  Although LST and NSAT are often closely coupled, the transformation from LST to NSAT is non-trivial.  Two methods will be trialled within EUSTACE: an established empirical statistical approach, and a new physical approach that could, in theory, be applied consistently over all surfaces.  This presentation will describe both methods, together with the associated uncertainties in the satellite NSAT estimates and a characterisation of the skin-air temperature relationship over land.  

Spatially explicit forest cover maps are an important tool for observing extent and changes in forest distribution. These products provide local stakeholders with important baseline data for the development of sustainable management strategies. In this work we investigate the potential of C‑band Synthetic Aperture Radar (SAR) data acquired by Sentinel-1 as well as multi-spectral optical data acquired by Sentinel-2 sensor for forest cover mapping in Mexico at national level. Forests of Mexico are very diverse and contain dry and humid tropical rainforests, mangrove and temperate forests as well as woodlands. Therefore, we analyze performance of SAR and optical data, separately and in the combined use for forest cover delineation in these different forest types. From the very high resolution airborne LiDAR data (G-LiHT sensor) acquired by NASA in 2013, we extract forest areas (i.e., areas >0.5 ha with tree crown >10% and vegetation high >5m), which are then used as training datasets in a machine learning classification algorithm (e.g., random forests). Finally, the generated Sentinel-1/-2 based forest/non-forest maps (F/NF) are validated using independent Mexican forest inventory data or airborne LiDAR samples, which were not used for model training. The produced F/NF maps are also compared with the Landsat based Tree canopy cover product (Hansen et al. 2013) as well as with the CCI Land Cover product. Freely available remote sensing data from airborne G‑LiHT LiDAR as well as from Sentinel-1 and -2 sensors are processed and analyzed using open‑source software (e.g., Sentinel Application Platform (SNAP), Fusion, R), which further contributes to open science as well as to operational use of these data for forest monitoring. An overall aim of the study is the investigation of capabilities and limitations of Sentinel-1 and -2 data, and the potential synergistic use of these systems for forest cover mapping in diverse landscapes of Mexico.

ESA’s Soil Moisture and Ocean Salinity (SMOS) mission has been providing valuable measurements of soil moisture and salinity over global land and oceans surfaces since January 2010. Level 2 algorithms are used to estimate Soil Moisture (SM) and Ocean Salinity (OS) based on SMOS observed brightness temperature (BT). A set of parameters, in particular SM and OS, is iteratively retrieved by minimizing the least square difference between SMOS observed BTs and BTs modeled by means of an L-Band (1.4 GHz) forward model [1]. The modeling at LBand needs to account for various surface features, like water, bare soil, forest, and their states like frozen versus non-frozen. External data sets are used to define the target composition with respect to each cover type, their relative sizes, as well as their states. Currently the Level 2 SM algorithm uses a combination of external maps which includes ECOCLIMAP II, ESRI's DCW, and ESA's GlobeCover to obtain the static land cover information required in the modeling process [1]. However the ECOCLIMAP II landcover, built from SPOT4/VEGETATION and CORINE II data for the year 2000, is aging and attempts are made to find a suitable replacement [3]. The Land Cover (LC) product of the European Space Agency (ESA)’s Climate Change Initiative (CCI) at 300m resolution [4], released in October 2014, offers some promising features including a water body map derived from the ENVISATASAR dataset, and is considered as a replacement candidate. This talk aims at evaluating the CCI LC product, epoch 2010, in conjunction with MODIS IGBP, developed by Broxton [3], and ECOCLIMAP II to assess their suitability as the land cover auxiliary data for the SMOS operational processors. In doing so there are several issues that need to be addressed including different scales, resolutions, and taxonomies. The SMOS Level 2 SM algorithms rely on classifications provided in the ECOCLIMAP II dataset to distinguish subclasses of forests, vegetation, water, and other surfaces, and hence other external data sets need to be pre-processed to adhere to these conventions. The talk will examine these issues. The resulting maps are compared in terms of their impacts on the decision tree configurations used in the SMOS soil moisture algorithm. Global and local performance statistics, such as the number of successful retrievals, are used to gain better insight into the suitability of each map with respect to the modeling process. Finally in situ data from known SMOS validation sites are used to evaluate the performances of each product with respect to soil moisture retrieval accuracy.

References

[1] Yann H Kerr, Philippe Waldteufel, Philippe Richaume, Jean Pierre Wigneron, Paolo Ferrazzoli, Ali Mahmoodi, Ahmad Al Bitar, Franc¸ois Cabot, Claire Gruhier, Silvia Enache Juglea, Delphine Leroux, Arnaud Mialon, and Steven Delwart, “The SMOS Soil Moisture Retrieval Algorithm,” IEEE Transactions on Geoscience and Remote Sensing, vol. 50, pp. 1384–1403, May 2012.

[2] Broxton, P.D., Zeng, X., Sulla-Menashe, D., Troch, P.A., 2014a: A Global Land Cover Climatology Using MODIS Data. J. Appl. Meteor. Climatol., 53, 1593–1605.

[3] Mahmoodi, A., Richaume, P., Kerr, Y., Mialon, A., Bircher, S., & Leroux, D. (2015). Evaluation of 1063 MODIS IGBP land cover data on the SMOS Level 2 Soil Moisture Retrievals. In ESA (Ed.), ESA SMOS 1064 Workshop. Villanfranca de la Canada, Madrid Spain: ESA

[4] Land Cover CCI, Product User Guide, Version 2, CCI-LC-PUG, Issue 2.4, 2014-09-02.

Cloud-induced gaps in multispectral remote sensing imagery are a major constraint for the derivation of land cover products. According to the US Department of Energy about 52 % of global land surfaces are covered by clouds on average. In addition, many regions of acute scientific interest, such as tropical forests, coastal areas, and urban sprawls tend to be especially affected by seasonal or full-year cloud coverage. Therefore, handling cloud cover is of special interest in multispectral land cover mapping. Many approaches have been proposed to address this problem, including filtering, histogram matching, band substitution, or the use of complementary images of sensors which are not affected by atmospheric distortions.

Our study region covers the refugee camp Domeez and its surroundings in the Kurdish region of Northern Iraq. The camp was set up in 2012 and hosts more than 45 000 persons in 2015. Due to the conflicts in Syria and Iraq an increase of refugee numbers can be expected. While the environmental impact of refugee camps is still discussed, remote sensing can provide information about land-use and landcover and hereby assist planning and decision-making by local authorities or governments in order to achieve a long-term management of land-use and resources. In this context, cloud filling is an important task to assure the quality of the products and the monitoring of developments.

We propose a post classification method based on Markov Random Fields (MRF) to perform the filling of gaps. Loopy Belief Propagation (LBP) is an established algorithm for inference on general graphs. It utilizes message passing to let each pixel propagate its assumptions on land cover to its direct neighbors. This method allows to pass messages into unclassified areas, and thus can be utilized to fill cloud-induced gaps. Our method is formulated over a multi-temporal MRF, thereby introducing the possibility to utilize imagery from different sensors to furthermore increase outcome quality. The outcomes are compared to a Closest Spectral Fit (CSF) technique which fills clouded pixels by a simple matching algorithm based on ancillary SAR data.

We use a cloud-free scene of Landsat TIRS/OLI and complementary data from Sentinel-1. Training and validation is performed through manually assigning clouds over areas of special interest. These areas include regions of especially homogeneous or heterogeneous land cover, regions which show either good or bad classification accuracies in the direct vicinity of the clouds, and other areas of elevated interest. We furthermore evaluate the potentials of utilizing the MRF's energy as a measure of confidence for the gap filling.

The proposed approach is a promising alternative to the established algorithms. While the quality of the filling is comparable to that of CSF, its ability to be implemented on a single scene, and the possibility to derive confidences make the method very reliable and offer a transparent measure for interpretation or further processing.

This study presents an empirical methodology to improve modeling mapping of air temperature and precipitation using methods that combine multiple regression and residual interpolation using inverse distance in a GIS.

So, series of ad-hoc experiments have been designed in order to study the role of factors that a priori have a strong weight in developing digital models of temperature and precipitation, such as solar radiation and land cover . The aim is to understand what roles these two factors play to incorporate them into the process of generating mapping of temperature and rainfall.

We have developed a multiple regression analysis between these meteorological variables as the dependent ones (Temperature and rainfall), and some geographical variables: altitude (ALT), latitude (LAT), continentality (CON) and solar radiation (RAD) as the independent ones. The dependent variables used in all experiments relate to data from meteorological stations precipitation (PL) mean temperature (MT), average temperature minimum (MN) and maximum average temperature (MX). These data were obtained from the AEMET (Agencia Estatal de Meteorología). Data series of stations cover the period between 1950 and 2010.

The main study Area is the Iberian Peninsula, but some experiments have been done specifically in the territory of Catalonia and the Catalan Pyrenees.

The first experiment based on the following question: What is the influence of solar radiation on the temperature of the air from a quantitative point of view? The difficulty in answering this question lies in the fact that there are lots of weather stations located in areas where solar radiation is similar. This suggests that the role of the radiation variable remains "off" when, instead, we intuitively think that would strongly influence the temperature. The idea  to detect the role of radiation is a design ad hoc, based on a sample of more equitable space statistician.

In case of the experiment with land covers, we have used the NDVI index as a proxy of land covers and added this variable in the independents to improve the models.

The results show how the role of solar radiation does not improve all models, it only benefits the results under certain conditions and areas, specially in the Pyrennes. The vegetation index NDVI and therefore the land cover on which the station is located, helps develop rainfall and temperature patterns, obtaining various degrees of improvement in terms of molded variables and months."

Land cover information is important in order to support policy regarding environmental sustainability and biological conservation. Up to date and accurate land cover information as well as information related to the changes over the years are useful for proper land resources management and habitat conservation. Especially for Protected Areas, land cover data represent key information for biodiversity conservation and alterations of the cover and structure over time, allowing for the assessment of their impact on changes in the biodiversity.  Samaria (Lefka Ori) National Park is located on the Southwest part of island of Crete (Greece, East Mediterranean). It was declared as a National Park via a Royal Decree in 1962. It is a mulidesignated area and specifically a National Park, Landscape of Outstanding Beauty, Natura 2000 site (SCI and SPA codes GR 4340008 and GR4340014) a UNESCO Biosphere Reserve (MAB) and has been awarded the European Diploma of protected areas by the Council of Europe. It is characterized by high mountains (more than 50 peaks over 2000 m. a.s.l) with subalpine zones covered by snow for a long period, pastoral areas mixed with cultivations, various shrubland types, forests, several gorges and karstic phenomena and an extended coastline, on a predominantly limestone substrate.

In the framework of ECOPOTENTIAL (www.ecopotential-project.eu), a H2020 funded research project, Samaria National park is one of the selected protected areas for the application of remote sensing tools and methods aiming to an effective management of resources and biodiversity conservation, as well as, for the identification of ecosystem functions and services with emphasis on the Copernicus services. Remote Sensing has a crucial role in the extraction of land cover information as it can provide accurate and regular information over large areas, within a specific predefined time period, e.g. every 16 days. Nowadays, given the availability of the Landsat image archive dated back to 1982, we can detected the changes that have occurred over a long term period, both temporal or permanent and to continuously monitor the desired area. The objective of this work is to detect the changes in the land cover of the Samaria (Lefka Ori) National Park since 1982 and to identify potential reasons for these (desertification, overgrazing, urbanization, climate change, natural hazards and alterations due to change of land use). An additional goal is to prepare an up-to-date land cover dataset for the National Park, meaningful for use in dedicated models (e.c. hydrological modeling, species distribution modeling, etc.). The classification of the archive will be based on the database of training data that FORTH has been prepare by analyzing VHR imageries (WVII & Aerial Color Images), field data and machine learning algorithms (SVM, RF). Moreover, the developed methodology will be applied in the forthcoming Sentinel 2 satellite data in combination with the Landsat 8 for a more frequent monitoring (weekly) of this important protected area.

Even though forest degradation is responsible for a large share of carbon emissions due to land use, land use change and forestry activities (LULUCF) the monitoring of degradation activities is technically more challenging than detecting deforestation events. In the context of REDD+ (Reduction of Emissions from Deforestation and Forest Degradation) the long-term reduction of biomass due to anthropogenic activities in forests remaining forests is the most prominent aspect of forest degradation. Depending on location and forest type various degradation drivers occur in Southeast Asia with different effects on the forest ecosystems. In our study we analyze forest disturbance mainly caused by selective logging in a tropical evergreen forest in Cambodia (Kampong Thom province) by monitoring changes in canopy cover over time. For this purpose, two cloud- and haze-free Landsat 8 scenes from January 2014 and April 2015 are selected, the former serving as base and the latter as secondary scene. A comparison of the top of atmosphere (TOA) reflectance values of the two scenes is used to derive information about disturbances in canopy cover, involving the following steps of image processing: Firstly, we generate an evergreen forest mask using a fully automated pixel-based classification algorithm from a series of Landsat imagery from October 2013 and to April 2014. Forest disturbance events occurring from 2014 to 2015 are then detected and analyzed within that mask. Secondly, in order to compensate for non-anthropogenic differences in the spectral values of the vegetation most probably caused by varying soil water content or other seasonal effects, a relative normalization of the secondary scene to the base scene is applied. This is achieved by adding a local correction matrix to the secondary scene, which is derived by calculating the difference of both median-filtered scenes. Finally, changes in the canopy cover are derived by calculating the difference between the normalized burnt ratios (NBR) of both scenes. The NBR is calculated similar to the NDVI but replacing the red band by the short-wave infrared band (SWIR 2 at 2.11 - 2.29 µm). While values around 0 indicate areas of no forest disturbance, any deviations can be interpreted as openings (positive values) or closings (negative values) of the crown cover. In our study only crown cover openings are analyzed. Visual comparisons with spatially high-resolution RapidEye scenes from February 2014 and March 2015 confirm the capability of the approach to identify small-scale or even sub-pixel changes in crown cover closure. The result shows that by applying the difference between two normalized scenes an almost complete reduction of topographic effects can be achieved, which enables in contrast to single-scene approaches the monitoring of forests in mountainous areas. Furthermore, the relative normalization derives consistent and comparable results, thus allowing a time series analysis over longer periods. Using Sentinel-2 imagery from the coming dry season 2015/16 we expect to derive an even more detailed detection of small-scale disturbance events.

Earth observation and more specifically, spaceborne radar remote sensing, had made much progress toward its high potential to retrieve Soil Moisture (SM) at different scales. However, for a single sensing system there always exists a trade-off between spatial and temporal resolution of the observations. While scatterometer-derived SM products can well describe temporal soil moisture dynamics, they lack of spatial details. They do not facilitate analysis of local hydrological patterns, such as effects from convectional rains and topography and thus miss the requirements of many users. Contrary, SM products from SAR sensors can resolve dynamics at this level. However, they observe individual locations less frequently and are thus not suitable for acquisition of short-term variations.

To overcome these spatial and temporal scale gaps, a data fusion approach is developed that fuses SM derived from the MetOp ASCAT scatterometer (12.5 km pixel spacing) and SM derived from the Sentinel-1 SAR at Interferometric Wide Swath mode (S-1 IWS, 40 m), yielding new SM data benefitting from both their high temporal and spatial resolution, respectively. The so-called SCAT-SAR SWI (Soil Water Index) describes water content in eight soil layers of the upmost metre, with a frequency of 1 day and spatial resolution of 1km. With these features, the SCAT-SAR SWI is part of the Copernicus Global Land Services product portfolio for monitoring continental ecosystems. 

This study shows first results from in Near-Real-Time generated SCAT-SAR SWI estimates, examining a 2-year-dataset over Austria. Results are validated against alone-standing ASCAT and S-1 SWI estimates, as well as against in-situ measurements at the Hydrological Open Air Laboratory (HOAL) in Petzenkirchen, Lower Austria. Additionally, insights from comparison against remotely sensed vegetation-, land cover-, and precipitation data are discussed in respect to the above mentioned spatio-temporal scale gaps.

Wetlands are an important component of the hydrologic and carbon cycles. They play an important role in ground water recharge, coastal protection and flood control (Richardson, 1994). They provide habitat for fish and wildlife and are thus important part of local economies. To conserve existing wetland ecosystems for biodiversity and ecosystem services and goods, it is important to inventory and monitor wetlands. Wetlands can act as carbon sink and as carbon source at the same time. While wetland vegetation takes up the carbon dioxide (CO₂) from the atmosphere and stores it as carbon, its respiration and decomposition releases CO₂ and methane (CH₄) to the atmosphere. Within the last years the importance of wetlands in global biogeochemical modelling has been recognized, with regard to the current role of the wetlands and in particular with its future change as they are believed to be sensible to climate change.

Despite several national to global mapping initiatives, the wetland area remains one of the largest uncertainties in the global methane budget (Prigent et al., 2001; Grunwald et al., 2012). Today, a number of global land cover products from global earth observation missions exist that include wetlands in their classification scheme. However, those maps show high heterogeneity in the spatial coverage and patterns of wetlands as well as in the underlying classification system.

In the present study, we investigated the consistencies between and uncertainties in four different global land cover products for the year 2000 or later (GLC2000, GlobCover, MODIS LandCover, GLC-Share) for the area of continental Europe including European Russia. Reference data were available from three different case studies in the boreal zone of Russia and Scandinavia where the largest part of the wetlands occurs. The land cover legends were transformed to a uniform key based on the Ecosystems Types (Level 2) classification system of the European Environment Agency (10 classes). The focus of our work was on the analysis of spatial patterns and statistics of wetland areas from national to pixel block level.

The results confirm the hypothesis of a high heterogeneity of class statistics at national level between the products and also within the products for different years (e.g. GlobCover 2005 and 2009). For European Russia, which holds about 2/3 of the wetlands in the study area, the coefficient of variation is 39 % for all products and 26.5 % for the MODIS Land Cover product (2001 to 2012). At the level of the test sites we observed high inconsistencies between the products with cross validation accuracies for the class wetland ranging from 2 % (MODIS, GLC2000) to 75 % (MODIS, GLC-Share). Finally we computed uncertainty estimates based on the confusion matrix for the three test sites after Olofsson et al. (2013). The results underline the substantial differences between the mapped area and the estimated area based on the class accuracy measures. In all but one case, the mapped area is outside the 95 % confidence interval of the estimated area; in several cases, there is a discrepancy of the mapped area with rations (mapped / estimated) ranging from 0.12 to 0.75. In all cases, the wetland is underestimated in the global land cover products.

In summary, the results highlight the importance of proper validation of existing land cover maps for ecological modelling and the need for further development and harmonization of global land cover products.

References:

Grunwald, D., Fender, A.-C., Erasmi, S. and H. Jungkunst (2012) Towards improved bottom-up inventories of methane from the European land surface. Atmospheric Environment, 51, 203-211

Olofsson, P.; Foody, G.; Stehman, S.V.; Woodcock, C.E. (2013): Making better use of accuracy data in land change studies. Estimating accuracy and area and quantifying uncertainty using stratified estimation. In: Remote Sensing of Environment 129, S. 122–131.

Prigent, C., Matthews, E., Aires, F. and W.B. Rossow (2001) Remote sensing of global wetland dynamics with multiple satellite data sets. Geophysical research Letters, 28, 24, 4631-4634  

Richardson, C.J. (1994) Ecological functions and human values in wetlands: A framework for assessing forestry impacts. Wetlands, 14, 1, 1-9

There is an ongoing need to understand and quantify the state and dynamics of forest ecosystems at local scales to support operational research including land use/land cover change monitoring and sustainable forest management. Forest cover is one of the key biophysical parameters used in prediction models to assist with ecosystem monitoring & and in geospatial tools to assist with forest management.

In a wide range of novel research studies data collected by Synthetic Aperture Radar (SAR) satellite systems have been used to enhance the identification & mapping of forest structure parameters including forest cover in remote areas where access is difficult, or with persistent cloud-cover. SAR data haven’t been used as widely in operational research to assist forest managers, geospatial practitioners, and forest scientists with decision making thus awareness needs to be raised in understanding the SAR data meaning & the processing flows required to use SAR data in information extraction systems. This poster is a key part of a European Space Agency (ESA) & Canadian Space Agency (CSA) sciences and operational application research (SOAR-2) project with the title ‘Studying Some Coniferous Forests' Characteristics with RadarSat-2 Data in Bulgaria’ which aims to assess the capability of backscatter data collected by the Canadian satellite RadarSat-2 to assist with forest cover mapping to support the sustainable management in two study areas; Parangalitza; United Nations Educational, Scientific and Cultural Organization (UNESCO) Man and Biosphere reserve, and Govedatzi, both located in the Rila mountain, Bulgaria.

Overall the ESA Cat-1 project has the following objectives:

Design & implement a SAR data collection plan to assist with dual-pol (HH, HV) RadarSat-2 (C-band) data collection during dry periods to mitigate effects of precipitation to the SAR data, and using ascending & descending data collection modes to support the mitigation of the extreme topographic relief found in Rila Mountain.

Pre-process & enhance the RadarSat-2 data to analysis level using the ESA Sentinel Application Platform (SNAP) and the RadarSat-2 toolbox including a very high resolution Digital Elevation Model (DEM).

Using the RadarSat-2 backscatter data establish their sensitivity to forest cover as quantified by field measurements or a forest cover raster product (25m pixel spacing) which was derived using Advanced Land Observing Satellite (ALOS) Phased Array L-band Synthetic Aperture Radar (PALSAR) data.

In this poster we present the bespoke process flow devised & implemented during objective (1) which was used to order & collect the RadarSat-2 data to fit the purpose of analyzing radar backscatter data in an area with high topographic relief and precipitation, and we present & describe the RadarSat-2 backscatter data which will be used in objective (3) in future analyses.

The results presented in this poster will assist forest managers, geospatial experts, and forest modelers to gain insights on the SAR data design & collection workflows required for forested environments and on the meaning of the backscactter SAR data collected by RadarSat-2 to support information extraction.

A stark difference exists in the micro-climate of urban and rural environmnets due to their differential surface energy budgets. Urban areas are characterised by impervious materials that absorb more heat and tall buildings that obstruct wind movement. In addition to this urban areas also release great amount of energy called anthropogenic heat. All these factors cumulatively alter the micro-climate of the city and give rise to the phenomenon of Urban Heat Island (UHI).

For long scientists have been trying to accurately model the urban climates. And in the current scenarios when climate change is already affecting populations across the world, urban climate modelling becomes all the more crucial. UrbClim is an urban climate model that simulates and studies the UHI effect and other urban climate variables at the local scale with spatial resolution of few hundred meters. It is designed to generate results at finer resolutions consuming minimal amount of computational power and is also performs faster dynamical downscaling  in comparison to other global and regional climate models. UrbClim is a coupled model consisting of a land surface scheme and a 3-D atmospheric boundary layer module tied to synoptic-scale meteorological fields through lateral and top boundary conditions. The urban climate variables generated from UrbClim could be further used to derive indicators such as the magnitude of the average urban heat island effect, the daily cycle of the effect or spatially explicit maps of heat stress indicators or heat wave days/intensity.

The UrbClim projection for Delhi will be presented through this paper. UrbClim has been successfully validated for various European cities that include Antwerp, Bilbao, Ghent, Toulouse and Almada to mention a few. Delhi will be one of the first Asian cities, where UrbClim climate simulations are attempted. Delhi is a tropical city and the capital of India. The city is characterised with humid subtropical climate according to Koppen climate classification. Delhi experiences extreme summers beginning in early April and continuing till the end of June.

The city has highly heterogenous pattern of land use and getting a land use land cover map for Delhi, that is as extensive as, e.g., the European CORINE land cover database is a challenge. To address this issue, a Local Climate Zone (LCZ) based land cover mapping is being done for the city using WUDAPT (World Urban Database and Access Portal Tools) methodology that employs an advanced classification algorithm. LCZ classification consists of 17 land cover types, 10 (Compact high-rise, Compact mid-rise, Compact low-rise, Open high-rise, Open mid-rise, Open low-rise, Lightweight low-rise, Large low-rise, sparsely built and Heavy industry) of these are based on urban and other 7 (Dense trees, Scattered trees, Bush scrub, Low plants, Bare rock or paved, Bare soil or sand and Water) are based on vegetation cover. This classification is built on logical division of the landscape.

The LST outputs from the UrbClim runs would be validated against satellite derived LST for the corresponding timings. The air temperature results will help in identifying and mapping the heatwave vulnerable locations and the city can hence plan and prepare for future heatwave events.

In 2002, narrow directional scattering over agricultural fields was identified in ERS-2 – ENVISAT ASAR Tandem (EET) pairs acquired within 30 minutes over the same area.  Examples of backscatter differences larger than 6dB were observed in EET pairs for look directions differing by less than 1 deg in the aspect angle. The presence of such poorly characterised anomalies of this type can have severe implications for the confidence with which we assimilate satellite SAR data. Therefore, directional scattering effects were further investigated (Wegmüller et al., 2006, Wegmüller et al., 2011) including scatter modeling (Mattia, 2011).

 

The work done so far addressed directional scattering at like-polarization. Now, with ALOS PALSAR-2 and Sentinel-1 cross-polarized backscatter data are becoming much more widely available, and so investigating, characterizing and modeling directional scattering also at cross-polarization may, respectively should, be addressed.

 

Split-beam techniques applied to dual-polarization PALSAR-2 clearly reveal that strong narrow directional scattering over agricultural fields is also present at cross-polarization – and this not only for fields with cultivation directions perpendicular to the line of sight, but also for fields with different cultivation directions. In our contribution we confirm the presence of strong directional scattering with a very narrow directionality over some agricultural fields at cross-polarization. We also compare the effects observed to the corresponding effects at like-polarization. Furthermore, the like to cross-polarization phase offset was considered to potentially obtain further understanding of the underlying scatter mechanisms that cause the directional scattering. A complete explanation of the directional scattering mechanism for the fields with cultivation directions not perpendicular to the line of sight, was not achieved.

 

 

References:

Mattia, F. Coherent and incoherent scattering from anisotropic tilled soil surfaces. Waves in Random and Complex Media, 21(2), 278–300, 2011

Wegmüller U., R. A. Cordey, C. Werner, P. J. Meadows, “Flashing fields” in nearly simultaneous ENVISAT and ERS-2 C-band SAR images, IEEE Trans. Geosci. Remote Sensing, Vol. 44, No. 4, pp. 801-805, 2006.

Wegmüller U., M. Santoro, F. Mattia, A. Balenzano, G. Satalino, P. Marzahn, G. Fischer, R. Ludwig, and N. Floury, “Progress in the understanding of narrow directional microwave scattering of agricultural fields,” Remote Sensing of Environment, Vol. 115, pp. 2423-2433, 2011.

 

Land-use/land-cover (LU/LC) change detection using satellite data has gained momentum in the past decades due to the need of detecting changes in Earth ecosystems caused by climate change or antropogenic drivers. With the advance of the Copernicus program, one of the greatest potentials of the large amount of earth observation data to be generated from the Sentinels, is studying global, regional, and local changes through time. The Compact High Resolution Imaging Spectrometer (CHRIS) mounted onboard the Project for OnBoard Autonomy (PROBA-1) spacecraft is a hyperspectral sensor which multi-angle viewing capability was found useful for vegetation structure assessment. The study area, “Bistrishko Branishte” UNESCO MAB reserve, is located in the upper part, between 1430 and 2282 m a.s.l., of Bistrica river basin in the North-East Vitosha Mountain. It was designated a biosphere reserve by UNESCO Man And Biosphere (MAB) Program in 1977. More than a half of the reserve territory is covered by an old-grove (120-150 years) Norway spruce (Picea abies L.) forest stands. Following a wind throw in 2001, a European Spruce Bark Beetle (Ips typographus) infestation outbreak in the most affected areas by the wind throw. The dead tree stands, which were not removed due to the strict regime of protection, were almost completely burned out in a wildfire occurred in the summer of 2012. Present study aims at studying the LU/LC change of “Bistrishko Branishte” UNESCO MAB reserve using CHRIS/PROBA Mode 1 satellite data acquired in 2012 and 2015. In this mode the satellite acquires images in five viewing angles, including nadir. The methods used in the study can be grouped into: 1) Pre-processing and data manipulation - CHRIS/PROBA satellite data were subject to data collection, noise correction, atmospheric and geometric correction, georeferencing, conversion to reflectance, cross-track illumination correction, terrain correction, spatial sub-setting, and design of a geodatabase in ArcGIS for subsequent spatial analysis; 2) Data analysis - LU/LC change detection is done using spectral difference images of nadir-looking mode CHRIS/PROBA Mode 1 images as well as difference between LU/LC maps produced after Maximum Likelihood Classification (MLC) method; 3) Data visualization – production of statistics (zonal), charts, and maps. The results obtained show a significant shift of the reserve’s LU/LC classes between pre and post-wildfire event and a slow vegetation regrowth at the fire scar of 2012 wildfire, represented mainly by grassland and deciduous shrubs, detected on CHRIS/PROBA image acquired in 2015. Finally, the results are compared with outputs from previous studies. The study is part of ‘Model for Assessment of Coniferous Vegetation Stress using Multispectral and Hyperspectral Satellite Data’ ESA Category-1 Third Party Mission (TPM) project (ID 8072).

In 2013, DLR extend the imaging capability of Terra-SAR, adding new products in term of spatial resolution, and off course corresponding swath, the Wide Scan SAR and the Starring Spot Light, SPL,  modes. By the way, TerraSAR offered the possibility of acquiring at resolutions ranging from 30m to 0.25 meter, with intermediate ones at 16m, 3.5m and 1m. The covered areas were ranging depending of the mode from 40 000 km² to 12 km² with intermediate surfaces of150 km² and 50 km². Complementarities in term of resolution versus swath were investigated over a specific target, water bodies’ recognition/extraction and monitoring over two tests sites presenting embedded sensitive areas. One test site is located under the tropics, it corresponds to the Poyang lake, the first fresh water resource in China, an hydro system that is under monitoring since 15 years exploiting both Middle and High resolution satellite images. Over this test site, all the modes have been investigated, from the Wide ScanSar allowing to capture in a single image the 24 000 km2 of the AOI, focusing the high resolutions, Stripe Map and Spotlight (SM & SL) of the Poyang Lake Natural Reserve (PLNR) and for highest one, Starring Spot Light mode, over Meixi HU and Bang Hu, the most sensitive elements and rich ones in term of biodiversity, of the PLNR.  The Rhine flood plain is a typical large inland flood prone of a temperate region. In this case it is the higher resoluted modes that have been exploited, ie Stripe map, Spot Light and Starring Spot Light. Two sensitive ecosystems, ie Rohrmatten wetlands and Muttersholtz grasslands, have been captured individually by the SPL, being covered within a single image by the others modes.

 

Both over Chinese and Alsatian sites, data acquisition, in HH polarization, were requested to be in a shortest at possible time interval, ie within 3 to 5 days and this every month or twice a month from June 2014 to June 2015 over China, and October 2014  to May 2015 over the Alsatian Flood plain. In term of acquisition, over each site, the SPL failed over one of the two smallest areas of interest, Bang Hu and Rohrmatten but functioned very well on the other one, ie Meixi Hu and Muttersholtz.

 The multitemporel approach was as promising as expected, and thanks to the very high resolution and revisit, it was possible to derived, waters masks from each image and then  water occurrence maps. The multiresolution approach was in addition exploited for validating the water extraction results from one coarser image towards better resolute ones. In a global view, resultants are very convenient, therefore this approach also highlight the limits of the validation approach.   Indeed, even if it is plain flood phenomena, characterized by a slow dynamic, that are targeted, the results shown that even on a few days (ie 3 to 5) water redraw can be very noticeable limiting the possibility of comparison in term of accuracy between data acquired at 3 days interval.  

 The results  highlight the importance of accessing to high revisiting data, as surface variations with are sensitive to river discharge and water table fluctuations, varies positively or negatively , with a within a few days (5 to 10) but also as water bodies are relative small (few ha to few km square) to high resolute imagery. This confirm and enhance the previous results obtained mostly based on successive mono sensors approach, SPOT4 Take Five, Pleiades HR, CSK.  Further steps will correspond to the synergy between SAR constellation, ie TerraSAR and Sentenl1, both in term of revisit and accuracy, and secondly with optical constellation such as SPOT6-7 and Sentinel2.

Identifying regions of change in remotely sensed images acquired over the same area at different times is critically important for applications related to landuse and environmental monitoring. With a growing number of optical satellite missions, the frequency at which any geographical area can be imaged is expected to increase to daily coverage. The effective exploitation of increased data volumes from frequent multi-sensor acquisitions requires development and implementation of novel approaches and algorithms in automated change detection and analysis.

This study developed software technology for identifying and analysing changes in land cover using multi-sensor imagery. A number of pre-processing steps were executed to minimize the detection of spurious changes due to variations in sensor geometry and environmental conditions, including spatial co-registration, radiometric and atmospheric correction as well as cloud and cloud shadow detection. Image co-registration is based on the Fourier phase correlation method with a registration accuracy of up to 1/10 of a pixel. Several basic and advanced change detection algorithms were implemented, including change vector analysis (CVA), cross-correlation analysis, principle component analysis, as well as band ratioing and differencing.  Object-based image analysis and change detection methods were used to identify specific targets of interest in very high resolution (VHR) imagery. Time series of multi-temporal images were analyzed using support vector machines, neural networks and discriminant analysis to characterize detected changes. The identification of change points simplifies the problem of threshold selection and minimizes the probability of false positives due to gradual changes. Rapid changes are identified by the software and a mask is generated of land cover changes. The implemented processes for change detection and analysis allow for the automated update of existing land cover and land use classifications. An interactive graphical user interface facilitates semi-automated quality control and verification of the results.

The approach implemented during this investigation was applied to several test study sites in Argentina and Canada to monitor natural disasters and anthropogenic activity using time series of multispectral imagery acquired by LANDSAT, RapidEye and other satellites. The comparison of different change detection algorithms demonstrated a good performance of CVA for the automated analysis of agricultural areas as well as the detection of deforestation and burnt areas. Flood detection and analysis using near-infrared band rationing demonstrated a high level of accuracy in mapping water extent changes after heavy rainfalls. The analysis of VHR imagery demonstrated high efficiency in detecting large vehicles and material stockpiling in rural areas. The developed image analysis approaches will be used with the growing satellite missions such as Planet Labs and other constellations planned for launch during 2016 and the following years.

Our understanding of the linkage between ecological succession and biophysical variables, that in turn can be observed and quantified via remote sensing, is highly biased towards temperate ecosystems, with little or no information on the linkages between such variables in tropical environments. In this context our knowledge is more limited in Tropical Dry Forests (TDFs) that are under-represented on most validation/calibration sites of emerging remote sensing products. Tropical dry forests are seasonal ecosystems with well-defined wet and dry seasons, with a mean annual temperature of 25oC and precipitation between 900 mm and 2000 mm per year. Dry season lasts between 5 to 6 months. Currently, TDFs cover 47% of all tropical environments, and are considered one of most endangered tropical ecosystem given its high rate of deforestation. Today, of the total extent of TDF in the Americas just 40% is left.  The current landscape of TDFs in the Americas is highly fragmented, under different levels of ecological succession, and surrounded by extensive agricultural areas given the fact that its soils are highly fertile.

 

In this paper we present for the first time an in-depth analysis of near-real time fPAR, NDVI and EVI information acquired by one of the largest Wireless Sensor Networks (WSN) currently deployed on a TDF worldwide. We evaluate linkages between those biophysical variables and ecosystem succession across early, intermediate and late TDFs. This work is conducted at the Santa Rosa National Park Environmental Monitoring Super Site, Guanacaste, Costa Rica. This site is part of the Central American “Dry Corridor”, a significant fraction of the Pacific Ocean facing land in the isthmus. Our work takes advantage of a large spatially distributed wireless sensor networks that measures temperature, relative humidity, soil moisture and Photosynthetic Active Radiation for a total monitoring area of 2,000 sq. meters. Information is collected in real time every 15 min and analysed using IBM’s stream analytics software. Information collected provides real time information for validation of emergin FPAR, NDVI and EVI. Information is presented in the context of one of the largest droughts affecting the Central American Dry Corridor, where it is estimated that 1M+ people are currently under extreme drought conditions.

 

Our results suggest that TDFs in the Central American Dry Corridor are currently under extreme drought conditions with overall Net Primary Productivity down by 80% when 2015 data is compared with 2013, a normal year. In addition, or results demonstrate that current NDVI and EVI spaceborne observations are highly accurate for TDFs but fail in the context of quantifying inter-annual drought effects because of their temporal resolution. Furthermore, our work suggests that current fPAR observations are highly inaccurate with significant mismatch between ground observations and spaceborne observations.

 

Our works suggest the strong need to develop and implement WSNs across the Central American Dry Corridor to support drought monitoring in the context of both ENSO and climate change. In addition, our work provides a solid background for the accurate estimation of ecosystems services such as carbon and water production; two ecosystem services provided by TDFs that are critical to the sustainability of large populations leaving along the Central American Dry Corridor.

Numerous satellite sensor systems have been launched during the last twenty years and satellite data are increasingly being used in regional or global mapping and monitoring. The observation from multiple satellites requires much effort to ensure continuity and compatibility due to differences in sensor characteristics and product generation algorithms. More recently the launch of Landsat-8 OLI and Sentinel-2 make the compatibility problem even more difficult as the very new spectral bands that both sensors use, need to be simulated and tested before proceed to any further comparison. Both sensors quantize the acquired information in 12-bit and thus they both present an increased radiometric resolution in comparison with the classic 8-bit sensors like Landsat-7 or previous. Furthermore, the new and quite narrow spectral bands of Sentinel-2 like bands five or six with a bandwidth of 15 nm are more similar to the bands of hyperspectral sesnors like Hyperion than to the classic bands of multispectral sensors.

The use of remote sensing data for target discrimination, whether the target is mineral, vegetation, man-made is based on the reflectance spectrum of the target. Every material has a characteristic spectrum based on its chemical composition. When sunlight strikes a target, certain wavelengths are absorbed by the chemical bonds; the rest are reflected back to the sensor. It is, in fact, the wavelengths that are not returned to the sensor that provide information about the imaged area. Specific wavelengths are also absorbed by gases in the atmosphere (H₂O vapor, CO₂, O₂, etc.). If the atmosphere absorbs a large percentage of the radiation, it may become impossible to identify a specific target. In order to overpass these difficulties the band rationing was proposed and extensively used with multispectral data. Band ratios are typically used to enhance spectral differences between bands. Dividing one spectral band by a second band produces an image that can be used to determine relative band intensities. Combining three ratios into a false colour composite (FCC) image allows determination of the approximate spectral shape for each pixel spectrum. Different combinations sensitive on mineral (TM5/7, TM5/4, TM3/1) or hydrothermal anomalies (TM5/7, TM3/1, TM4/3) detection were used for the detection and mapping of minerals or alteration zones.

A preliminary comparison of multispectral data from Landsat 8 OLI to the  respective data from Sentinel-2 is performed and the results are presented in this study. The behavior of different band ratios which, although they have a similar algebraic structure, they may not produce similar results when applied to data from different sensors (like OLI or Sentinel-2), is examined in qualitative and in quantitative terms. In order to assess the performance of these band ratio images different quantitative criteria are used such as, the standard deviation of the image, the signal to noise ratio, the autocorrelation function and the coefficient of variation of each pixel. The standard deviation (stdev) of the image histogram is a measure of the contrast of the image. A high standard deviation means a good contrast. The signal to noise ratio (SNR) is a measure of how clearly a target of interest (for example a quarry) is expressed and in what extent the noise is suppressed. The autocorrelation function and the coefficient of variation of each pixel may provide information about the spatial variation and the texture of the band image.

The European Space Agency (ESA) and the United States National Aeronautics and Space Agency (NASA) jointly sponsored an aircraft campaign to support the ESA Earth Explorer 8 candidate mission, the Fluorescence Explorer (FLEX), referred to as FLEX-US. This campaign was conducted in North Carolina (NC), USA in the fall of 2013. The campaign co-manifested a high-performance imaging spectrometer package called HyPlant with a NASA/Goddard Space Flight Center (GSFC) instrument package, the GSFC’s Lidar Hyperspectral Thermal Airborne Imager (G-LiHT). G-LiHT is a compact and portable integrated system comprised of off the shelf LiDAR, hyperspectral, and thermal components to measure canopy structure, vegetation composition, and surface temperature, respectively. HyPlant is an instrument sponsored by Forschungszentrum Jülich GmbH and built by Specim (Spectral Imaging Ltd., Finland) and developed to serve as the prototype instrument package for the FLEX mission concept. HyPlant has the capability to observe solar induced fluorescence (SIF) as well as visible and infrared (380-2500 nm) spectra for plant health, composition, and biodiversity.

The campaign was conducted at two forested NC sites having active eddy covariance flux towers: the Parker Tract (Plymouth, NC) and the Duke Forest (Raleigh, NC). Parker Tract is a managed loblolly pine plantation comprised of different age classes. The Duke Forest’s tower is located in a protected mixed deciduous forest, with isolated loblolly pine stands. Ground-based measurements for calibration purposes were made to support the airborne overflights at both sites in late September and again in late October. In situ canopy and leaf-level measurements were obtained in late September/early October during the early senescence fall timeframe at the Duke site on 8 tree species – seven deciduous and loblolly pine. We show the retrieved SIF, lidar, thermal, and VNIR data from a number of the flight lines acquired at both sites, and provide a summary of the leaf-level measurements. We provide the synthesis of the HyPlant/G-LiHT SIF, thermal, and VNIR information to describe vegetation responses during the senescence period, and show the benefit of additional canopy structure information. The airborne data show that SIF is influenced by the surface temperature, and differentiates young versus old stands. The leaf level data show that the SIF information differentiated species during senescence and could be related to PSII photosynthetic function. This project assisted the evaluation of the HyPlant instrument performance, supported the evaluation and development of SIF airborne high resolution data retrieval approaches, corroborated hypothesis of the relationship of SIF to photosynthetic variables, and most importantly – supported the integrated spectral concept advanced by the FLEX mission.

Accurate representation of actual terrestrial surface types at regional to global scales is an important element for a wide range of applications, such as land surface parameterization, modeling of biogeochemical cycles, and carbon cycle studies. Since the 1990s, a variety of global surface type products with different spatial resolutions and legend definitions have been introduced into the scientific community. The best-known land cover products include International Geosphere-Biosphere Program (IGBP) DISCover, created using 1992-1993 1-km resolution Advanced Very High Resolution Radiometer (AVHRR) data, the University of Maryland (UMD) land cover product, which was also based on the 1-km AVHRR data, Global Land Cover 2000 by the Joint Research Centre of the European Commission (JRC), which was generated using the 2000 VEGETATION sensor onboard Satellite Pour l'Observation de la Terre 4 (SPOT4) satellite, and the Moderate-resolution Imaging Spectroradiometer (MODIS) land cover product using MODIS-based surface reflectance data. With the launch of the Suomi National Polar-orbiting Partnership (S-NPP) satellite on 28 October 2011, and the data from the Visible Infrared Imaging Radiometer Suite (VIIRS) onboard S-NPP having become available since early 2012, the National Oceanic and Atmospheric Administration (NOAA) Joint Polar Satellite System (JPSS) land surface type team began generating a new global surface type (aka land cover) classification map using the new VIIRS data. The first VIIRS data generated global surface type map has been released to public in early 2015, which may be accessed from ftp://vct.geog.umd.edu/st/. The new VIIRS surface type map basically followed the processing flow used by the MODIS land cover map with the similar C5.0 decision tree algorithm. International Geosphere-Biosphere Program (IGBP) classification scheme (including 17 classes) was used for the VIIRS global surface type map. The surface type team is also investigating other classification algorithms for future creations of VIIRS surface type maps. Meanwhile, for some other studies, different classification types are required. For instance, the retrieval of global leaf area index (LAI) and fraction of photosynthetically active radiation absorbed by the vegetation (fPAR) relies on accurate discrimination of surface type map in six major biome classes based on their canopy structures, which include: Grass/Cereal Crops, Shrubs, Broadleaf Crops, Savannas, Broadleaf Forests, and Needleleaf Forests. Therefore, in this study, the original VIIRS surface type classification map in IGBP types is further being developed into Biome classification legends, and the preliminary global map in biome types has been generated based on a LUT conversion from IGBP legend with the ancillary data. In the meantime, alternative methodology in creating global surface type map in Biome scheme, such as direct classification, has also been explored and compared. The global biome classification maps are under evaluation and validation results will be presented.

To date, mapping anthropogenic forest disturbances has largely been focused on distinct delineations of events of deforestation using optical satellite images. This method however faces a number of challenges: (i) in the tropics, cloud cover and other atmospheric effects hinder timely monitoring, and (ii) optical signals saturate in dense forests, hindering the detection of subtle changes in forest structure. This is particularly relevant for forest degradation, i.e. the anthropogenic reduction of forest cover or woody biomass in areas that remain defined as forests thereafter. Studies that quantify carbon emissions or the extent of degradation indicate that it is widespread across the tropics and affecting land area similar to deforestation alone. However, these disturbances are often spatially and temporally dynamic, characterized by frequent reversibility and repeatability over short time periods, which are challenging to detect using optical signals. 

To overcome these challenges, our study instead provides an algorithm to detect the spatially and temporally continuous processes of deforestation and forest degradation using radar backscatter (obtained from ALOS PALSAR), mapping them as disturbances affected by reversibility and fluctuations over successive observation periods. The results hence provide a characterization of the actual land use processes causing forest disturbances. The study covers parts of Tahuamanu and Tambopata provinces of Madre de Dios in the Amazon basin, recognized as a tropical “Capital of Biodiversity” (Peruvian law No. 26311) and a conservation priority. By analyzing the magnitude of radar signal changes, we are able to characterize successional forest dynamics occurring during and post-disturbance, a process that has been deemed impossible or largely ignored in previous studies. Also, by allowing for the detection of fast-recovering and slow-recovering changes in radar signal, we are able to pick up small-scale and diffuse forest degradation, in addition to deforestation. Comparisons of our results to widely used optical-based deforestation datasets of Asner et al. (2010) and Hansen et al. (2013) indicated that the total area of disturbance detected by radar is over twice their estimates. It was further noted that spatially delineating disturbed areas with abrupt boundaries risks ignoring areas in transitional disturbance. This highlighted both the continuous gradient of land cover change and the possibility of degradation preceding and accompanying deforestation. Further, initial and recovered backscatter values were related to the size of land cleared for each land use type. For example, the clearing of large areas for gold-mining or pastures was more common in lands with low above-ground biomass to begin with. These large areas are more likely to be fully cleared or with very low biomass immediately after disturbance, and also subsequently recover backscatter faster in the years after disturbance, possibly since they are more likely to be abandoned. The results suggest that rapid cover and biomass accumulation post-disturbance are crucial successional dynamics. 

In summary, the main contribution of our work is presenting the utility of radar for detecting deforestation and forest degradation, and monitoring forest dynamics, in tropical rainforests. In Madre de Dios, our key results suggest (i) satellite-based radar can specifically detect degradation related disturbances to a significant extent in tropical forests; (ii) the total disturbed area in our study region is more than double that detected in previous, optical-based, studies; (iii) large-scale deforestation is most likely in low biomass areas; and (iv) rapid canopy cover recovery or biomass accumulation post-disturbance, which is an important sink for atmospheric carbon, can be monitored using radar. The presented radar-based detection algorithm is spatially and temporally scalable, and can support monitoring degradation and deforestation in tropical rainforests with the use of products from ALOS-2 and the future BIOMASS mission.

Invasive plant species are among the top five threats to biodiversity at a global scale. However, mapping of plants and plant communities over large areas is difficult, time consuming and expensive when using traditional field surveys. For management purposes, the early detection of invasive species is crucial. High resolution hyperspectral and LiDAR data offer a great potential to map and monitor invasive plant species, especially so at an early detection stage, and their impact on ecosystems. By combining presence-only data of the target species with airborne remote sensing data, our goals were (1) to develop an approach that allows to efficiently map biotic invasions; and (2) to create distribution maps for invasive plant species in two study areas in Western Europe, based on the aforementioned approach. This will offer the basis for further analyses of the impact of invasions on ecosystem functioning and will thus allow to infer possible management options.

For this purpose we collected vegetation data on 120 plots with a size of 3 m × 3 m on the island of Sylt (Northern Germany). The plots had different cover fractions of the invasive moss Campylopus introflexus and the invasive shrub Rosa rugosa. In the forest of Compiègne (Northern France), we sampled a total of 50 plots with a size of 25 × 25 m, targeting the invasive tree Prunus serotina. In both study areas, independent validation datasets containing presence and absence points of the target species were collected. Airborne hyperspectral data were acquired for both study areas in summer 2014 by means of the APEX imaging spectrometer, providing 285 spectral bands (350-2500 nm) with a pixel size of 1.8 and 3 m, respectively. For classification, we used maxent.

The results for Sylt showed that mapping invasive species using one-class classifiers is possible. For C. introflexus, the overall detection accuracy was 72%, for R. rugosa this was 92%. Most of the misclassified validation plots either had a very low cover percentage of the target species, a high cover of spectrally similar species, or were located less than one pixel away from predicted presences. In this presentation, we want to show the results of our study and evaluate the potential for mapping invasive plant species with hyperspectral and LiDAR data, especially the potential for early detection and the transferability of our models to similar study areas. Furthermore, we will discuss possible error sources connected to the collection of field and remote sensing data, as well as in the data analysis and interpretation process.

Traditionally Soil Effective Temperature,  Teff , is considered as a secondary intermediate variable in the microwave radiative transfer theory but its systematic impact on the microwave radiometry tends to be  comparable to those by vegetation cover, soil surface roughness and dielectric constant. In this study, we evaluate the magnitude of such impact by using the MERRA-Land soil moisture and temperature profiles to reproduce global  Teff  data sets with single layer (e.g. current SMOS (T2 inscheme), two-layer (e.g. the text)Beta version SMAP (T1) scheme) and multilayer  Teff schemes (e.g. Lv’s scheme(TM) ). The result reveals that difference in  Teff  among these schemes could reach 8K, which is comparable to the errors resulted from other well identified factors. The comparison result between the Wigneron’s and Lv’s scheme indicates that for areas where Wigneron’s scheme is valid the difference is small (RMSD<1.5K) while for those areas not applicable for Wigneron’s scheme the difference is relatively big (RMSD>5K) . The RMSD between the Beta SMAP  Teff  scheme and Lv’s scheme can reach around 8K, especially at 6 p.m. Such Teff   difference leads to the difference in emissivity of 0.15 or more at incidence angles of 42.5° (the incidence angle of SMAP is about 40° ). Regions with big RMSD ( Lv’s vs. Beta SMAP scheme) include the Central Asia, the North America and subtropics. It is suggested to reconsider the    schemes and their functionalities in soil moisture retrievals.

The central topic of this study is the monitoring of winter wheat phenology and the detection of anthesis (flowering) using remotely sensed data as well as crop growth modeling.

Accurate tracking of phenology in crops is of utmost importance for crop growth and yield modeling, as the degree of carbon allocation into different plant compartments (root, stem, leaf, fruit) changes with the phenological stages, thus directly influencing yield formation. Anthesis is a crucial stage for the crop performance, as the plant faces an increased risk of infection by Fusarium funghi at this stage. This Fusarium infection can contaminate the grain with the mycotoxin Deoxynivalenol (DON), making it unacceptable for certain end uses and induces yield reduction.

It is not possible to directly observe the flowering of wheat with optical satellite sensors. Thus, an approach that combines crop growth modeling with remote sensing data covering optical and microwave spectral ranges was developed. This was done in three steps:

The hydro-agroecological land surface model PROMET was first run in a stand-alone version for selected sites distributed throughout Bavaria using only static input parameters (e.g. soil map) and current meteorological data as driving factors. The phenological simulation then was validated based on in-situ data that had been collected for the modelled sites. The Bavarian State Research Center for Agriculture provided an extensive data set of weekly in-situ measurements for 63 and 71 observation sites for the years 2014 and 2015 respectively. Based on this information, the performance of the model could successfully be tested. However, some deviation between modelled and observed crop phenology still remained.

Thus, multi-temporal information from optical remote sensing data was assimilated into the model runs in a second step to improve the accuracy of the results. For this, green Leaf Area Index (LAI) was derived from high-resolution optical imagery through inversion of the radiative transfer model SLC (Soil-Leaf-Canopy). Green Leaf Area Index should be in its maximum at the stage of anthesis for winter wheat. Both the time-series of satellite information on their own as well as the outputs of the enhanced model were validated against the Bavarian in-situ-data with good results. The deviation of the model results from the actual phenological development could be reduced by several days.

Finally, the use of radar data for anthesis detection in winter wheat was tested using Sentinel-1 data of 2015 in dual polarization mode (VV+VH). Based on literature, it is assumed that the green LAI maximum correlates with a minimum in C-Band backscatter, and that the anthesis stage can thus be detected. For this analysis, an approach was developed to quantify and eliminate the influence of soil moisture dynamics on the SAR signal by establishing site-specific regressions of simulated surface soil moisture and bare soil backscatter. In these soil moisture corrected backscatter time-series the minimum is identified as proxy for flowering. The results were validated against the Bavarian in-situ data, though the results weren’t as convincing as for the optical satellite information.

Concluding, the best results for phenology monitoring and detection of flowering in winter wheat in Bavaria were achieved by assimilating high-resolution optical remote sensing data into the crop growth model. It is expected that with the availability of Sentinel-2 and thus with improved temporal, spectral and spatial resolution, it will be possible to map the crop cycle of winter wheat in even more detail. With a stable high-quality monitoring of crop development through optical sensors, it becomes a realistic goal to predict the onset of anthesis with the PROMET model using two-week weather forecasts as meteorological drivers

Currently world economy is dealing with the agricultural problems. Increasing population and decreasing fertile lands require researchers to find better crop management practices. For this purpose remote sensing lately became an important technique to monitor agricultural areas. Additionally with the use of polarimetric properties of the Synthetic Aperture Radar (PolSAR) data, it is possible to detect physical properties of the plants within different polarimetric channels. A recent study presented by the authors provides the global sensitivity analysis of a crop morphology based electromagnetic scattering model and shows the importance of crop height and structural density through the growth cycle ^[1].  Therefore, to have the maximum efficiency from PolSAR data in crop monitoring studies, it is important to identify the sensitivity of the different polarimetric descriptors throughout the growth cycle of the crops.

This study focuses on the selection of PolSAR parameters for an efficient growth stage determination scheme. To this aim PolSAR parameters are related to crop height and growth stage using a Polynomial Chaos Expansion (PCE) metamodel ^[2]. Based on the provided high degree polynomial relation between PolSAR and biophysical parameters, Global Sensitivity Analysis (GSA) is performed by efficiently calculating the variance-decomposition-based Sobol’ indices. Specific to this study, Total Sobol’ indices provide a measure of overall sensitivity of each of the PolSAR parameters to the change of the corresponding biophysical parameter.

In this work 2 different datasets are analyzed: 12 TerraSAR-X co-polar data over the Isla Major (Seville, Spain) with a central frequency of 9.65 GHz and an incidence angle of 29^o and 5 RADARSAT-2 quad-pol data over the Ipsala (Turkey) with a central frequency of 5.40 GHz and an incidence angle of 31^o.In accordance with the PolSAR data, crop height and the growth stage in terms of Biologische Bundesanstalt, Bundessortenamt und CHemische Industrie (BBCH) scale is measured. With respect to each growth phase of rice fields (e.g. early and late vegetative, early and late reproductive, maturative), GSA provides information about the most sensitive PolSAR parameters that are available from PolSAR data.  Based on the analysis, a set of parameters are emphasized for each phase in the crop monitoring. Finally, the effect of biophysical changes over the sensitive PolSAR parameters are discussed for different frequencies.

[1]    O. Yuzugullu, S. Marelli, E. Erten, B. Sudret, I. Hajnsek, Global sensitivity analysis of a morphology based electromagnetic scattering model, in: Geoscience and Remote Sensing Symposium (IGARSS), 2015 IEEE International, pp. 1017–1020, 2015.

[2]    G. Blatman and B. Sudret, “Efficient computation of global sensitivity indices using sparse polynomial chaos expansions”, Reliability Engineering & System Safety, vol. 95, no. 11, pp. 1216–1229, 2010. 

In wetlands, hydrological regimes determine values of functions and ecosystem services. If hydrological regimes of wetlands are now widely handled by environmental managers, its impacts on both floods and vegetation dynamics are still unexplored. Simultaneously, the new Earth observation Sentinel-1 constellation, allowing all weather radar acquisitions with high temporal and spatial resolution appears as a promising opportunity for the monitoring of wetlands. In the same time, this study aims at (i) evaluating the ability of a Sentinel-1 time for the detection and monitoring of flood areas at 1:50 000 scale across a 100 000 hectares marsh; (ii) exploring the relation between the hydrological dynamic and ecological processes. 

Information on the spatial distribution, quality and amount of forest resources is of great interest to many disciplines. Different methods for extracting forest related information (e.g. above ground biomass, forest cover, tree height) from remotely sensed data, either passive or active, have been developed in recent years. Especially methods using data acquired by active sensors (SAR) operating in the microwave domain of the electromagnetic spectrum are well suited for above ground biomass mapping. One method for above ground biomass estimation has been developed within the ESA funded BIOMASAR-I and II projects, which makes use of hyper-temporal C-band Envisat ScanSAR data to estimate above ground biomass using a water-cloud like model with gaps individually for each scene before a final map is generated from a weighted averaging of the individual estimates. Analysis of the individual above ground biomass maps showed different levels of agreement with reference data, which has been addressed to differences in environmental conditions.

Radar backscatter from forested areas is controlled by both system (e.g. wavelength, polarization, incidence angle) and ground parameters (e.g. tree density, tree height, species composition). Additionally, environmental factors such as soil moisture under forest may have an impact on radar backscatter from forest and thus on above ground biomass mapping. This phenomenon has gained only little attention in investigations dealing with the retrieval of above ground biomass estimations from SAR data so far. This is mainly due to a lack of spatially extended in-situ soil moisture reference data acquired regularly over an extended period of time under forest.

To further investigate the influence of varying soil moisture conditions under forest on the estimation of above ground biomass from hyper-temporal SAR data, measurements of a in-situ soil moisture monitoring network are used that was installed by the University of Jena in spring 2015 in different forest stands near the city of Jena, Germany. In-situ soil moisture is measured at 30 locations at two depths and an interval of 15 minutes. Together with additional information on the vertical and horizontal structure of the forest floor, a comprehensive database has been established and is continuously extended by incoming regular soil moisture measurements.

Individual above ground biomass maps derived from C-band Sentinel-1a data using the hyper-temporal approach introduced above, will be compared to environmental data, e.g. in-situ soil moisture under forest or precipitation data to further investigate the external factors influencing above ground biomass mapping of forest areas using C-band data.

Satellites offer products like forest gross or net primary productivity on a pixel scale (e.g. Modis). Such products are highly post-processed signals, which also include observations from other sources. These post-processed products face a certain uncertainty due to many aspects. One of these aspects might be species diversity and forest structure. Forest structure can be characterized by several attributes e.g. LAI (which is included in many post-process) or maximal tree height and tree height heterogeneity in a forest patch (which both are not included so far). To investigate the influence of forest structure and diversity on productivity, we developed a method which generates millions of forests patches, using the well-established, individual- and process-based forest growth model FORMIND. Every forest patch arise from a given species mixture and a given tree size distribution. In addition, trees crowns occupy a certain space and cannot overlap. For every forest patch we calculate the productivity of forests under different climates. We remove those forests, which include trees with negative productivity. We than analyze the relationship between productivity and various characteristics of forest structure and the influence of tree species diversity. For instance, LAI shows a positive relationship with productivity whereas increasing tree height heterogeneity has a negative relationship. Additional, we show that structure attributes of forest are more important for the productivity estimation than tree species richness.

The taiga-tundra ecotone extents over 13400km around the northern hemisphere and marks the transition zone between closed forest and tundra. Since this area is sensitive to climate change, human activities and natural disturbances such as wildfires, monitoring of this region is extremely important. Dealing with the characterization of this transition zone, very few studies are available on a global scale. Existing approaches lack the thorough characterization of the ecotone as they only capture a part of the complex transition zone.

The objective of this research was the development of a model for detecting the taiga-tundra interface for a designated study area across northern Siberia. Therefore, optical and radar remote sensing data products from 2007 to 2010 were used, whereby the spatial resolution of all datasets were unified to 1km. In order to cover the full complexity of the taiga-tundra ecotone, as many different products as possible were used as input data for the model. Therefore, data from ALOS PALSAR, MODIS, AMSR-E were used in addition to digital elevation model data and the Percent Tree Cover product by Hansen et al. (2013). The datasets were used for computing several inter-annual and intra-annual statistics. Using random forest variable importance measure, most relevant statistics were selected and used for different model setups. The final random forest model setup included statistics of land surface temperature (LST), albedo, NDVI, EVI, Vegetation Continuous Fields (VCF), PALSAR HV-Polarization and LAI. Snow water equivalent (SWE) statistics had to be excluded to compensate for artifacts.

Since sharp distinctions between forest and tundra do not occur in nature, class probability values were computed. Those allowed the unique allocation per class for each grid cell. The model setup showed high performance values, which were cross validated, with accuracy of 0.956 ± 0.015 and kappa of 0.930 ± 0.023. The model result was validated using high resolution Google Earth imagery, where an overall map accuracy of 93% and a Kappa of 0.89 were achieved. The assessment of the input data products revealed that statistics of LST, SWE and albedo were most important.

An advantage of this approach over existing circumpolar products is the derivation of continuous probability values for the taiga-tundra transition zone. The model may be applicable for the entire circumpolar region and thus, provide the basis for a monitoring effort aiming at detecting changes over time. This would help to understand the dynamics of ecosystem changes with respect to influencing factors.

More than a decade ago, Palmer et al. (2002) proposed the Spectral Variation Hypothesis (SVH). This hypothesis states that the spectral heterogeneity of neighboring pixels covering mixed vegetation stands is positively correlated with the diversity of these stands. The spectral variation as quantified in image data has since then frequently been used as a proxy for various measures of biodiversity. Even though in theory, species with similar spectral properties may lead to a homogeneous spectral signal of otherwise diverse vegetation stands, most studies testing the SVH came to promising results. However, no consensus on the set up of the approach exists and the available studies thus differ fundamentally in their results. These differences include the I) measure of biodiversity (e.g., species numbers, quantitative species richness, or functional diversity), II) the measures of spectral heterogeneity, III) the spectral information considered, as well as IV) the spatial resolution of the image data. Each of these parameters may have a considerable influence on the observed relationship. A systematic analysis of their influence is missing so far and hampers the definition of a 'golden standard' for the SVH. We hence ask

* Do different measures of spectral heterogeneity concur in their results?

* How does the measure of biodiversity affect the outcome of the SVH?

* Is the SVH sensitive to the spectral and spatial resolution of the image data?

To answer these questions, we simulated 500 mixed grassland stands, each covering an area of 10 m x 10 m, based on simplified assembly rules. These stands included 3 to 30 different species from up to five life forms. The species varied in their trait combination, resulting in a variable spectral dissimilarity. Subsequently, we used the PROSAIL model to simulate the canopy reflectance of the stands in eight different spatial resolutions (0.2, 0.5, 1, 1.25, 1.67, 2, 2.5, and 3.3 m). Using these data sets, we quantified the biodiversity of the stands with different measures and calculated the spectral heterogeneity with common approaches described in published SVH studies for all spatial resolutions. In these calculations, we additionally varied the spectral information content of the data using a principal component analysis based approach. We then tested for the variation in spectral heterogeneity induced by different measures, spatial resolutions, and the spectral information content. Likewise, we analyzed the influence of the respective biodiversity measure on the quantified diversity of the stands. Finally, we evaluated the strength of the relationship between biodiversity and spectral heterogeneity considering the variable set ups.

The results show that in particular the measure of biodiversity (for example, whether we test for taxonomic or functional diversity) has a severe influence on the strength of the observed relationships. The spectral information content, spatial resolution, and measure of spectral heterogeneity had a less prominent effect. The different measures of spectral heterogeneity were in general highly inter-correlated. We discuss these observations considering the trait combinations of the species under investigation and the compositional evenness of the simulated vegetation stands. The outcome of this study enables a better understanding of the SVH and helps to define a golden standard for its use. This may help to take full advantage of new sensors for biodiversity research.

Passive microwave remote sensing at L-band (1.4 GHz) has been widely acknowledged as the most promising technique to measure global soil moisture from space. Therefore, several passive microwave missions at L-band have been launched in the past 5-years. The first soil moisture dedicated satellite was the Soil Moisture and Ocean Salinity (SMOS) mission led by the European Space Agency (ESA), launched in November 2009 and continuing to outlive its 5-year design life. SMOS carries an innovative interferometric radiometer that gives it the unique capability to observe brightness temperatures at 40 km resolution with multiple incidence angles for the same location. The second soil moisture dedicated satellite was the Soil Moisture Active Passive (SMAP) mission launched by the National Aeronautics and Space Administration (NASA) in January 2015. SMAP has a traditional real aperture scanning antenna measuring brightness temperatures at 40 km resolution and a fixed incidence angle of 40°. In addition the Aquarius satellite, launched in June 2011 for ocean salinity measurement, had three L-band radiometers with fixed real aperture antenna measuring brightness temperature at incidence angles of 28.7°, 37.8°, and 45.6° with footprint sizes of 74 km × 94 km, 84 km × 120 km, and 96 km × 156 km respectively. As its power and attitude control system for the spacecraft stopped working, Aquarius ceased operation in June 2015. For the purpose of developing a long-term consistent global soil moisture product, their brightness temperature observations need to be carefully inter-compared and the derived soil moisture products validated. Airborne Polarimetric L-band Multi-beam Radiometer (PLMR) brightness temperature observations were collected at 1 km spatial resolution coincident with the three satellites during the Fourth Soil Moisture Active Passive Experiment (SMAPEx-4) conducted in May 2015. The PLMR brightness temperature observations were processed to match the 3-dB footprints of each satellite and compared with coincident satellite brightness temperature observations. The main result presented in this research is an evaluation of the brightness temperature inter-comparison between PLMR and the different satellites.

Semi-arid ecosystems with mixed canopies and strong vegetation dynamics are challenging for satellite based landcover (LC) mapping. This is reflected in relatively low map accuracies in global LC datasets, particular over Africa. Improvements could be achieved with dense multiannual satellite time series, which represent inter-seasonal variability and seasonal vegetation patters. With almost daily observations, Proba-V provides dense time-series at 100 m spatial resolution, which were previously available only at coarser resolutions. While this is a step towards Sentinel-2, it also provides a compromise between spatial detail and data volume.

Here, we present time series based LC-mapping with Proba-V data for Africa. The processing of at continental level is achieved using a scaleable and reproducible workflow implemented with open source software installed on cloud based resources provided by the ESA Research and Service Support service (RSS CloudToolbox).

We derive and apply a method for (1) multitemporal cloud-filtering, (2) extracting seasonal dynamics and (3) subsequent LC classification. For (2) we explore methods based on statistical metrics, seasonal models and time series similarity measures. For (3) we use state-of-the-art machine learning algorithms such as Random Forest. Nine general LC classes are used for classification:  Forest, Shrubland, Grassland, Cropland, Wetland, Urban, Bare, Water and Snow/Ice. The classification outcome will be calibrated and validated with  datasets  available through GOFC-GOLD reference data portal and Geo-Wiki platform and compared to global LC datasets at different resolutions.

First tests show the potential of multitemporal cloud-filtering for Proba-V tiem series: With a moving-window filter, about 80% of the un-flagged clouds could be removed for a sample of Proba-V NDVI time-series, according to visual examination. A first classification based on a harmonic linear model fitted to the NDVI time-series of 141 calibration samples showed a high out-of-bag error rate (> 30%) with high confusion of the crop- and shrubland classes. Multitemporal cloud filtering allowed us to retrieve smoother time-series. The low accuracies of the first classification will be tackled with advanced time-series analysis methods. The ESA RSS CloudToolbox proved to be a valuable tool: the saleable resources (RAM and CPU) make it possible to adapt to the applications needs.

Rural development is an important process of improving the quality of life and economic well-being of people living in sparsely populated areas in Central Asia. Therefore, multilateral development banks (MDBs) have financed multiple projects in order to support rural development activities in the region for a long time.

Recently, collaboration programmes such as the EOTAP (Earth Observation for a Transforming Asia Pacific)initiative from the European Space Agency (ESA) have been launched for demonstrating the value of Earth Observation-based products supporting MDBs for planning, implementation, monitoring and evaluation for their specific project activities.

With the main goal to build up to 50,000 private homes within Uzbekistan, a large-scale, widely dispersed national residential construction program has been financed and implemented by the Asian Development Bank (ADB). This integrated rural development project in Central Asia is served now with information products derived from Earth Observation datasets through a dedicated EOTAP project.  Focussing on selected housing complexes financed by ADB, optical PLÉIADES datasets with very high spatial resolution have been used for monitoring and mapping of housing construction and related infrastructure.  The temporal construction progress has been documented using archived very high-resolution optical SPOT-5 datasets for change detection analysis.

The project resulted in a comprehensive building and infrastructure inventory with over 20,000 buildings, 31 thematic land use/land cover classes and 400km of road network mapped in high spatial and thematic accuracy, presented in map format as well as on a dedicated web platform designed for this project.

According to the latest analytical report of a reputable consultancy agency  Euroconsult- "Remote Sensing: Market Prospects for 2023", the number of satellites increased during the next decade by public and commercial organizations, will more than doubled and shall reach 353 satellites, providing income estimated by the analysts equal to 35.8 billion USD within the period from 2013 to 2023 [1].

Following the growing number of member-states of the "space club" and the quantity of remote sensing satellites (ERS SC), the volume and quality of geospatial information is growing, distributed on a free-of-charge (or shareware) basis.

Such information, diverse in its spectral and spatial characteristics, opens new capabilities for a system of applied tasks solution, such as agromonitoring.

Taking into consideration that agriculture is characterized almost everywhere by instability and high dependence on weather conditions, the use of geospatial space information is gaining a strategic importance.

Ukraine possesses already a practice of using data from the Sentinel satellites, designed under the Copernicus program, in the following areas:

- Inventory and mapping of farmlands;

- Monitoring of areas and the crop state;

- Monitoring of irrigation areas;

- Monitoring of crop rotation violations and illegal crops identification.

Using time series of optical and radar images an efficient method was developed for crop classification [2], [3]. Evaluation of crop areas at regional level based on obtained classification map. Classification maps based on time series of Landsat-8 and Sentinel-1 images were provided at NUTS2 level for Kyiv, Odessa and Khmelnitsky regions. Also official statistics were compared to evaluated crops area for those regions. Crop maps based on satellite data are objective and reliable for use at governmental and regional revel.

Thus, given the large amount of data, obtained under the Copernicus program (Sentinel-1 and Sentinel-2), a capability has come up to carry out operational agricultural monitoring at national and regional levels.

http://www.euroconsult-ec.com/earthobservation

S. Skakun, N. Kussul, A. Y. Shelestov, M. Lavreniuk, O. Kussul, “Efficiency Assessment of Multitemporal C-Band Radarsat-2 Intensity and Landsat-8 Surface Reflectance Satellite Imagery for Crop Classification in Ukraine,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2015, DOI: 10.1109/JSTARS.2015.2454297.

N. Kussul, S. Skakun, A. Shelestov, M. Lavreniuk, B. Yailymov, and O. Kussul, “Regional Scale Crop Mapping Using Multi-Temporal Satellite Imagery,” Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XL-7/W3, pp. 45–52, 2015.

The University of Natural Resources and Life Sciences (BOKU) in Vienna, in cooperation with the National Drought Management Authority (NDMA) in Nairobi, has setup an operational processing chain for mapping drought occurrence and strength for the territory of Kenya using MODIS NDVI data at 250 m ground resolution from 2000 onwards. The processing chain employs a modified Whittaker smoother providing consistent NDVI “Monday-images” in near real-time (NRT) at a 7-daily updating interval. The approach constrains temporally forecasted NDVI values based on reasonable temporal NDVI paths.

Contrary to other competing approaches, the processing chain also provides a modelled uncertainty range for each pixel and time step. The uncertainties are calculated by a hindcast analysis of the NRT products against an “optimum” smoothing.

To detect and quantify the droughts and the drought strength, the vegetation condition index (VCI) is calculated at pixel level from the filtered NDVI data. Starting from weekly temporal resolution, the indicator is also aggregated for 1- and 3-monthly intervals considering available uncertainty information (Figure 3). For administrative purposes, the VCI’s are further averaged over administrative boundaries, by taking into account the land cover of interest (here pastures). Analysts at NDMA use the spatially/temporally aggregated VCI for their monthly drought analysis. Based on the provided biophysical indicators as well as a number of socio-economic indicators, disaster contingency funds (DCF) are released by NDMA to sustain counties in drought conditions.

The paper shows the successful application of the drought products within NDMA by providing a retrospective analysis applied to droughts. 

Land surface parameters from remote sensing observations are critical in monitoring and modeling of global climate change and biogeochemical cycles. Current methods for estimating land surface parameters are generally parameter-specific algorithms and are based on instantaneous physical models, which result in spatial, temporal and physical inconsistencies in current global products. Besides, optical and Thermal Infrared (TIR) remote sensing observations are usually separated to use based on different models , and the Middle InfraRed (MIR) observations have received little attention due to the complexity of the radiometric signal that mixes both reflected and emitted fluxes.

 

In this paper, we proposed a unified algorithm for simultaneously retrieving a total of five land surface parameters, including Leaf Area Index (LAI), Fraction of Absorbed Photosynthetically Active Radiation (FAPAR), surface albedo, Land Surface Temperature (LST), and surface emissivity, by exploiting remote sensing observations from visible to TIR domain based on a common physical Radiative Transfer (RT) model and a data assimilation framework. The coupled PROSPECT-VISIR and 4SAIL RT model were used for canopy reflectance modeling. MODIS visible to TIR time series datasets were used as input datasets of this algorithm. At first, LAI was estimated using a data assimilation method that combines MODIS visible to MIR observations and a phenology model. The estimated LAI values were then input into the RT model to simulate surface emissivity and surface reflectance. Using data from the RT model simulation, the black-sky, while-sky and actual albedo were calculated. Besides, the background albedo and the transmittance of solar radiation down to the background, and the canopy albedo were also calculated to produce FAPAR values at the given wavelengths, which were then used to produce FAPAR over the range of 400-700 nm. At last, LST were estimated by exploiting MODIS TIR data and the simulated emissivity using the split-window algorithm.

 

These five parameters were validated over several representative sites with different biome type, and compared with MODIS and GLASS land surface product. Results demonstrated that this unified inversion algorithm can retrieve temporally complete and physical consistent land surface parameters with high accuracy.

Bush encroachment describes a global problem which is especially facing the savanna ecosystem in Africa. Invasive species and woody vegetation spread out in areas where they are not naturally occurring and suppress endemic vegetation, especially grasses. Livestock is directly affected by decreasing grasslands and inedible invasive species which are a result of the process of bush encroachment. For many small scale farmers in developing countries livestock represents a type of insurance in times of crop failure and droughts. Among that bush encroachment is also becoming more and more a problem for crop production. Studies on the mapping of bush encroachment so far focus on small scales using high-resolution data such as mainly aerial photography, and rarely provide information that goes beyond the local or national level. Therefore a process chain was developed using a multi-scale approach to detect bush encroachment for the whole African continent. This bush encroachment map is calibrated with field data provided by experts in Southern, Eastern and Western Africa. Supervised classification links location data, which represent the training samples for the up-scaling, to the respective pixel of remote sensing data. The location specific information is up-scaled with the integration of remote sensing imagery from two different sensors. The first level uses Landsat 5 and Landsat 8 with 30m resolution. For the second level the MODIS surface-reflectance product (MOD09A1) with 500m resolution is taken into account. Based on these data a map is developed that shows potential and actual areas of bush encroachment in Africa and thereby provides an innovative approach to map bush encroachment on the regional scale. The classification technique is based on random forests and regression trees, a machine learning classification approach which is programmed in R. In addition to the map on bush encroachment a second output is generated which focuses on the probability of bush encroachment occurrence based on possible causes such as fire occurrence based on MODIS data (product MCD14DL) or soil moisture information by ESA (CCI SM v02.1). By this, possible areas for bush encroachment occurrence based on their pre-conditions and risk factors are identified.

This innovative approach includes multiple datasets derived from earth observation data to detect bush encroachment in Africa, a severe and ongoing global problem. The identification of bush encroachment and the probability of its occurrence can help to prevent further grassland decrease and identify those regions where land management strategies are of high importance to sustain livestock keeping and thereby also secure livelihoods in rural areas.

Selected as European Space Agency’s seventh Earth Explorer in May 2013, the BIOMASS mission will provide crucial information about the state of our forests and how they are changing [Le Toan et al., 2011]. This mission is being designed to provide, for the first time from space, P-band Synthetic Aperture Radar measurements to determine the amount of biomass and carbon stored in forests [Le Toan et al., 1992]. The data will be used to further our knowledge of the role forests play in the carbon cycle.

Earth Observation platforms play an essential role in the new European Space Agency Earth Observation strategy and in the evolution of the ground segment. Earth Observation Platforms are gradually appearing in Europe, at national level or at European scale, and outside Europe. For example, the Thematic Exploitation Platforms currently developed at European Space Agency are a category of exploitation platforms dedicated to geophysical themes. The Thematic Exploitation Platforms under development are instrumental as pilots to formulate future technical and economical approaches for networking exploitation platforms (European ecosystem of Earth Observation Platforms), which is a central idea of the Earth Observation ground segment evolution.

In this context of an innovative sensor and a changing ground segment, the concept of user exploitation platform dedicated to the BIOMASS mission is proposed. This exploitation platform will be a virtual open and collaborative environment. The goal is to bring together data centre (Earth Observation and non- Earth Observation data), computing resources and hosted processing, collaborative tools (processing tools, data mining tools, user tools, …), concurrent design and test bench functions, application shops and market place functionalities, accounting tools to manage resource utilisation, communication tools (social network) and documentation.

This platform will give the opportunity, for the first time, to build from a community of user of this new Earth Observation mission around this innovative concept.

[Le Toan et al., 2011]  T. Le Toan, S. Quegan, M. Davidson, H. Balzter, P. Paillou, K. Papathanassiou, S. Plummer, F. Rocca, S. Saatchi, H. Shugart and L. Ulander, “The BIOMASS Mission : Mapping global forest biomass to better understand the terrestrial carbon cycle”, Remote Sensing of Environment, Vol. 115, No. 11, pp. 2850-2860, June 2011.

[Le Toan et al., 1992]  T. Le Toan, A. Beaudoin,  et al., “ Relating forest biomass to SAR data”, IEEE Transactions on Geoscience and Remote Sensing, Vol. 30, No. 2, pp. 403-411, March 1992.

Leaf Area Index, major tree species or plant functional type, forest height, growing stock volume, stem basal area, site fertility and land cover type were predicted over two boreal forest sites in Finland. Both sites comprised an area of 200 km by 200 km. The principal satellite data were from Landsat 8. The land cover classification was augmented in cloudy regions with Sentinel-1 based land cover map. Additionally clear cuts and other rapid changes were analysed using two-temporal Landsat-8 and multi-temporal Sentinel1- data. The results of the change detection will be reported in another paper for the Living Planet Symposium.

The images were analysed with the fuzzy estimation method Probability by VTT. The method combines unsupervised and supervised approaches and outputs all the predictions as continuous values. The blue, red, near-infrared and short wave infrared band of Landsat-8 data were used as input as features.  With Sentinel-1 data amplitude mean and temporal variability of multi-temporal of VV and VH polarized data were applied. The temporal variability was computed as the standard deviation of logarithm of amplitude data.

The ground reference data were 3173 ha of stand-wise forest maps on the southern study site and 24451 ha on the northern site, respectively, with attribute data from Metsähallitus that manages the government forests of Finland. The data set at both sites was split geographically in two parts and the other half was used exclusively to assess the accuracy. A regular point grid of 200 m by 200 m was placed over the test site and the predictions and the stand variable values were read at these points.

The land cover mapping was assessed by placing a systematic sample plot network on Pleiades data and interpreting the plots visually.

The test results of land cover classification are not available at the time of writing this abstract but will be computed until the end of the year 2015.  Preliminary accuracy figures for stem volume, tree height, tree species and site type maps have been computed. The root mean square error for stem volume was 63 m³/ha (39 % relative root mean square error) for the site in Southern Finland and 28 m³/ha (48 %) for the site that was located in Northern Finland. For the tree height the corresponding figures were 5.9 m (40 %) in southern site and 5.1 m (45 %) in the northern site.  In the northern site the bias of the estimates was small (5 m³/ha and 1.3 m overestimation). In southern site the overestimation for stem volume was larger but for the tree height the same as in the northern site (13 m³/ha and 1.3 m). The proportion of the observations where the predicted main tree species was the same as the main species in the ground data was 79 % in the southern site and 95 % in the northern site. The northern site with the higher accuracy was very pine-dominated. The site fertility was classified into seven classes. The proportion of the observations where the predicted site type was the same as the site type in the ground data was 71 % in the northern site and 65 % in the southern site. The approach in the testing could not consider within stand heterogeneity because in the reference data a stand had only one value for a forest variable which likely decreased the accuracy figures.

The improved radiometric sensitivity of Landsat-8 compared to the earlier Landsat missions improved the estimation of the high growing stock level up to 300 m³/ha. The results are promising for Sentinel-2 due to its unforeseen radiometric resolution and higher spatial resolution compared to Landsat. A major challenge remained in closed young canopies in which the growing stock tend to be overestimated.

The predictions were used in a Dynamic Vegetation Model of the University of Sheffield and in a semi-empirical carbon assimilation model of the University of Helsinki. In these models the satellite image predictions were either direct input parameters or they served validation of the parameters computed by the models.

This study is part of the North State Framework 7 project of the European Union.

Agricultural monitoring was one of the first applications of remote sensing. For a long time it remained mostly in the public sector but over the past decade the interest from the private sector has been steadily growing. Since a few years the insurance sector is picking up on the use of remote sensing.

Satellite images provide useful information for crop monitoring and damage assessment. This will be illustrated with results from ESA's Gazelle service trial in Morocco which used MODIS 250m data and with examples of the use of SPOT-VGT/Proba-V 1km data for comparable studies in Russia and Ukraine. Typically, damage maps are generated, showing the insurance company in which region problems occur. This allows them to organize their field visits in a more efficient way. The maps can be used to check if a field for which a claim is received is located in a problem area or not. The availability of long time series of satellite images allows the insurer to assess historical losses and to generate risk maps, which can be used for premium calculation.

Satellite images can also be used for index insurance. In this case, payouts are based on a regional index (a vegetation index, rainfall, soil moisture or evapotranspiration estimate) that is correlated with farmers’ losses. When the index exceeds a certain threshold, all farmers in the region receive a payout. In the frame of a study for the Weather Risk Management Facility (WRMF), a joint initiative of the International Fund for Agricultural Development (IFAD) and the World Food Programme (WFP), VITO developed index insurance products for Senegal for groundnut, millet and maize, to provide cover against drought. The products are based on SPOT-VGT/Proba-V 1km fAPAR products combined with TAMSAT rainfall estimates. The methodology, findings and lessons learned so far will be briefly presented.

Time series of SPOT-VGT/Proba-V and MODIS products have already proved their usefulness, both for damage and risk assessment as for index insurance. Further improvements are expected from the use of more detailed Proba-V 100-300m and Sentinel-2 time series. Index insurance products could also become more accurate if combined with an accurate crop type mask, as could be provided by Sentinel-2.

While the mineral exploration activities in Greenland are increasing, the demand for geophysical datasets to support these activities is increasing as well. The application of the remote sensing technology to assist the geological mapping and mineral exploration projects in the Arctic regions of Greenland is of interest because of the vast territory and the logistic difficulties of the field work in Greenland. In addition, the particular geological features and processes that occur in Greenland offer the possibility of particular geological applications of remote sensing, that are of broader scientific interest and not only because of the Arctic latitudes of the study area.

Although, the traditional Landsat data have lower spectral resolution and moderate spatial resolution than other type of imagery that has successfully been applied for mineral exploration in Greenland (Bedini 2009; 2011), the Landsat imagery due to the large area coverage can be used at reconnaissance stage of mineral exploration especially to detect mineral alteration zones that could be associated with mineralization.

In this poster are shown results of the analysis of Landsat data from study areas in East Greenland and West Greenland. The analysis is based on traditional band ratio techniques applied to Landsat data in order to highlight the alteration zones. This is complemented by specific data preprocessing in order to mask the vegetation, water and ice based on the spectral properties of these types of land cover. In general, better results are obtained in the central East Greenland environment due to the more pronounced alteration systems in these areas. A number of the detected alteration zones constitute exploration targets.  

Recently, the European Space Agency (ESA) has launched the Sentinel-2 satellite. The large swath (290 km), availability and the 10 m spatial resolution of VNIR bands make the Sentinel-2 imagery attractive for reconnaissance stage mineral exploration activities in Greenland. The applicability of Sentinel-2 imagery to mineral exploration in Greenland is in the detection of mineral alteration zones and for the mapping of lineaments that could be related to the structural grain of a given study area. Due to the extended snow season in Greenland, the first geologically useful Sentinel-2 images of the Greenland ice-free strip of land are expected to be acquired in the summer of 2016.  

 

References

Bedini E. (2011). Mineral mapping in the Kap Simpson complex, central East Greenland using Hymap and ASTER remote sensing data. Advances in Space Research, 47, 60-73.

Bedini E. (2009). Mapping lithology of the Sarfartoq carbonatite complex, southern West Greenland, using HyMap imaging spectrometer data. Remote Sensing of Environment, 113, 1208-1219.

The aim of this paper is to identify the phase contribution in SAR interferograms due soil moisture and Snow Water Equivalent (SWE). Both phase contributions have characteristics different from those related to simple propagation as in the case of atmosphere phase delay or of the displacements occurred in the observed scene.

The importance of knowledge of the spatial distribution of soil moisture is well known and has many applications in agriculture. SWE provides a measure of the amount of snow cover. Physically, it corresponds to the volume of the water which it could be obtained is a unit-volume of snow volume is completely melt. The SWE is directly related to the snow depth (SD). Furthermore, high resolution information about both soil moisture and SWE is relevant also for Numerical Weather Models.

As a first step we are not interested in retrieving soil moisture levels and SWE with interferometric techniques, but to predict their effects on the interferograms and the coherence maps to be able to compensate for it and isolate additional signals like atmosphere and terrain deformation.

The main idea is that the differential propagation of the electromagnetic waves into the soil and snow causes the interferometric effects, by affecting the vertical wavenumber in a thin layer close to the surface, in dependence on the moisture level.

The capability of SAR interferometry to provide information on the vertical distribution of scatterers in the ice will be also investigated to find a relationship between and the interferometric coherence and penetration depth or snow accumulation. Absorption by melt water will also reduce interferometric coherence introducing limitations to interferometric methods.

The interferometric phase contribution due soil moisture and snow water content will be determined computing the phase inconsistencies in all sets of three images. Starting from the three SAR image three interferograms will be computed and spatially averaged. Then for each pixel the three interferometric phases will be summed up following the scheme described in [1]. A phase inconsistencies is observed when sum is different from zero (modulo 2p). The phase inconsistency will be mapped to soil moisture and snow water content maps. The time series of phase inconsistencies will be computed using two strategies: a) choosing the master image using a spatial baseline criterium and keeping it constant along the time while varying the other two SAR images used to compute the map of phase inconsistencies; b) choosing the three images as three consecutives images, starting from the first acquired image of the stack.

The methodology will be applied on SAR interferograms obtained in different frequencies bands: L (ALOS and ALOS-2) and C (ERS-1/2, ENVISAT, SENTINEL-1) Results will be compared with those obtained by traditional methodologies applied to the amplitude SAR images and, when available, with in-situ measurements.

 

F. De Zan, A. Parizzi, P. Pratz-Iraola, P. Lopez-Dekker, “A SAR interferometric model for soil moisture”, IEEE Transactions on Geoscience and Remote Sensing, 52(1), 418-425, 2014.

 

Global measurements of the Earth’s surface temperature, particularly the oceans, but increasingly Land, lakes and Ice are key indicators of climate change as well as being of importance for operational meteorological applications. Observations from Satellite are a crucial component of the global observing system for these and many other parameters.  However, whilst every effort is made to provide an optimum SI traceable pre-flight calibration of the instrument and validation of associated algorithms it is only in-flight that the sensors measurands and derived geo-physical variables can be fully validated.  This requires robust, SI traceable measurements of the geo-physical variable under test that can be fully correlated to the satellite observation: time, location and footprint whilst accounting for differences in environmental factors that might affect either observation differently. 

 

The criticality of such validation measurements has long been recognised by all space agencies and given its global nature, international cooperation is essential to ensure a cost effective yet comprehensive process. This is why it is one of the core activities of the Committee on Earth Observation Satellites (CEOS), the space arm of Group on Earth Observation (GEO), through its Working Group on Calibration and Validation (WGCV).  More recently, with the increased attention to climate and the need to establish multi-decadal climate data records, a more rigorous requirement and consequential definition for a sub-set of high accuracy reference measurements (Fiducial Reference Measurements, (FRM)) has been established by the Sentinel 3 Validation team and is becoming widely adopted:

 

“The suite of independent ground measurements that provide the maximum Return On Investment (ROI) for a satellite mission by delivering, to users, the required confidence in data products, in the form of independent validation results and satellite measurement uncertainty estimation, over the entire end-to-end duration of a satellite mission.”

 

The definition places a strong emphasis on the need for rigorous, internationally harmonised, measurement and analysis processes and calibration of associated instrumentation. Whilst this does not mean that all instruments and methods should be the same, it does suggest that there is a place for some underpinning internationally coordinated best practises to provide guidance on the key elements needed to make an FRM and provide advice on how this can be achieved. It also implies the need for comparisons to provide clarity on declared uncertainties enabling potential biases to be evaluated and ideally corrected.

 

Over the last couple of decades CEOS WGCV has run a number of formal comparison exercises to facilitate international harmonisation. Comparisons of instruments measuring ‘brightness temperature’ of Sea surfaces have been particularly prevalent with a so called ‘Miami series’ of comparisons. As the series progressed an increased sophistication and rigour in terms of uncertainty assessment and SI traceability has been achieved.  Triggered in part by the forthcoming Sentinel 3 launch CEOS WGCV has embarked on plans for a new comparison campaign, not only for Sea Surface Temperature (SST) but also to include the other Surface Temperature domains (Land, Lakes, Ice). 

 

To facilitate this CEOS WGCV project the European Space Agency (ESA) has provided resources to support this exercise through a project called FRM4STS (http://www.frm4sts.org).  In delivering this project a team of European experts led by the UK national metrology laboratory, National Physical Laboratory (NPL) will not only carry out a set of comparisons but will also look to draft, in consultation with the community, a set of best practise guidance for each of the domains. Such guidance will be invaluable to ensure that the measurement systems and appropriate underpinning comparisons can be cost effectively sustained into the future as a key element of long term operational programs like Copernicus.

 

The core of the project will see the Worlds surface temperature validation community take part in a set of laboratory and field (Land and/or water) comparisons.  The comparisons are designed to evaluate variances in instrument calibration and their performance/traceability ‘end to end’ in both ideal and operational conditions where the influence of the environment may have additional impacts on their uncertainty. 

The structure of forested areas is a key element for assessing the characteristics of the forest including biodiversity, variations such as degradation/regrowth, as well as distribution and quality of forest resources. The derivation of the 3D structural information is important since physical parameters such as biomass can be estimated based on its allometric relation to the variables like tree height. Therefore, information on vertical structure of the forest allows development of accurate and robust estimates for such parameters.

 

Dealing with a volumetric object, remote sensing data that interacts with the forest components under the canopy provides very valuable information on the forest structure, hence, facilitates the assessment of parameters such as above ground biomass, tree height or forest density. In order to infer that information, Synthetic Aperture Radar (SAR) sensors operating in the microwave range of the electromagnetic spectrum are commonly used. Furthermore, an advanced processing technique referred as SAR tomography has proven to be a unique tool to retrieve the 3 dimensional forest information. SAR tomograms provide estimations to reconstruct the forest in 3D, since different scattering characteristics at different heights can be mapped by combining multiple SAR observations from different perspectives.

 

In this study, the investigation of forest structure by SAR tomography is carried out in order to assess different parameters, e.g. tree height, forest density and biomass, and to investigate the impact of different factors, such as undergrowth/regrowth and soil moisture, on such assessments.  The airborne SAR data at L-band acquired by DLR’s F-SAR from multiple tracks will be utilized to study the forest structure characteristics by SAR tomograms over a forest in Thuringia, Germany. The test site includes managed forest stands covering different structural stand conditions. Stand structure ranges from single layered to multi-layered, even aged forest stands. The forest consists mainly of pine and spruce plus some larch and birch trees.

To further investigate the impact of different factors including seasonal changes, state of undergrowth and soil moisture, multi-temporal F-SAR data composed of mid-summer and autumn acquisitions will be used.  Together with the information delivered by the F-SAR observations, a comprehensive and continuously extended database is established to improve and to verify the interpretations based on SAR tomography. The collection of data ranges from forest inventory and in-situ soil moisture time series, to regular acquisitions of Sentinel-1 data for hyper-temporal biomass estimations, which will be compared to the tomography based estimations.

 

The objectives of the DeMo-Wetlands project are to develop improved methods and tools for detection and monitoring of tropical wetlands and to demonstrate and implement the developed methods on the national scale in Rwanda. The project results and developed methods and products are expected to directly feed into a national wetland information system of Rwanda and to directly support international organizations and initiatives. To reach these goals, a good cooperation with academic and agency partners in Rwanda and with international partnering conventions and initiatives will be established.

 

The project activities are designed to directly support the Ramsar Convention on Wetlands of International Importance by making the development of national wetland inventories much easier for tropical countries. This is expected to improve the availability and knowledge about the extent and status of wetlands in the tropics, which supports the Ramsar Convention in the development of its ‘State of the World's Wetlands’ reports. The project is also linked to the Group on Earth Observations (GEO) and will support the water and biodiversity/ ecosystems  activities of GEO. Further, the project results will directly feed into the development of a Global Wetland Observation System (GWOS) which is developed in the framework of GEO together with other projects and partners.
Besides the Ramsar Convention, the developed tools and products potentially also support global monitoring and assessment endeavours in the post 2015 development framework like the Sustainable Development Goals and especially target 6.6 which is to “protect and restore water-related ecosystems, including mountains, forests, wetlands, rivers, aquifers and lakes”.

 

The envisaged working approach of DeMo-Wetlands will be concerned with collection and analysis of Copernicus-related sensor system datasets and the automatization and implementation of results on the national demonstrator Rwanda. Sentinel-1 data will be used to identify wetland locations based on geomorphological conditions, followed by a verification of wetland sites based on a high spatial and temporal resolution optical dataset (Sentinel-2). This detection of wetlands will be augmented by a monitoring component which considers inundation and land cover/ land use of wetlands. The multisensoral working approach combined with a multitemporal analysis and continuous automatization to enable a sustainable use of the demonstrator will be presented besides the general objectives of DeMo-Wetlands. We intend to highlight and discuss potentials for cooperation with other ongoing wetlands related projects to identify common goals and based on this, harmonize working activities.

We propose a new method to produce time series of global maps of surface mass variations by progressive integration of daily geopotential variations measured by orbiting satellites. In the case of the GRACE mission (2002-2012), these geopotential variations can be determined from very accurate inter-satellite K-Band Range Rate (KBRR) measurements of 5-second daily orbits. In particular, the along-track gravity contribution of hydrology is extracted by removing de-aliasing models for static field, atmosphere, oceans mass variations (including periodical tides), as well as polar movements. Our determination of surface mass sources consists of two successive dependent Kalman filter stages. The first one consists of reducing the satellite-based potential anomalies by adjusting the longest spatial wavelengths (i.e. low-degree spherical harmonics less than 5-6). In the second stage, the residual potential anomalies from the previous stage are used to recover surface mass density changes - in terms of Equivalent-Water Height (EWH) - over a global network of juxtaposed triangular elements. These surface tiles of ∼40,000 km x km are imposed to be identical and homogeneously-distributed over the terrestrial sphere, however they can be adapted to the local geometry of the surface mass. Our global approach was tested by inverting simulated hydrology-related geopotential data, and successfully applied to estimate time-varying surface mass densities from real GRACE-based residuals. This strategy of combined Kalman filter-type inversions can also be useful for exploring the possibility of reaching better time and space resolutions for hydrology, that would be hopefully brought by future low altitude geodetic missions.

Net primary productivity (NPP) is a key environmental indicator. It provides information about vegetation productivity and carbon sequestration. NPP time-series allow for monitoring of meteorological and human impact on the environment.

Monitoring of NPP is of special interest in regions that are quickly changing such as China, a country that shows immense and rapid development. The Republic of China is the world’s third largest country in terms of area and – with respect to population – the most populous country in the world. Economic growth and migration trends put pressure on ecological resources. Observation of NPP dynamics helps to understand possible impacts on the environment and to observe changes in productivity of natural vegetation and agricultural areas.

In this study, we model NPP time-series with 1 km spatial resolution for entire China for 14 years (1999–2012). We apply the Biosphere Energy Transfer Hydrology (BETHY/DLR) model, which is driven by remote sensing based and meteorological data. The usage of remote sensing derived leaf area index (LAI) as one of nine major input parameters allows not only analysis of climate effects but also observation of human induced, local, and temporary effects on vegetation productivity.

Annual and monthly NPP patterns are presented and spatio-temporal NPP distributions analysed. This includes mean annual NPP distribution and mean productivities for different land cover classes. Monthly data provide information about temporal patterns of vegetation productivity for different regions in China and different vegetation types. Analyses of interannual NPP variability revealed considerable differences in the development of annual vegetation productivity within the analysed time period for different provinces.

Further, two focus areas have been identified to demonstrate usability and relevance of the results. In Shanghai province, an area of rapid urban growth, we observe a strong impact of urban development on NPP (reduction in NPP of at least 10%). This decrease in NPP shows the strong influence of one of Asia’s fastest growing megacities on the environment.

Further, the NPP time-series was analysed for a region that is affected by forest disturbances in Northeast China. The results show that NPP data are suitable for identification and monitoring of forest disturbance and regrowth.

The 14-year NPP time-series for China provides important information for understanding environmental change. The examples demonstrate that the data can be successfully applied for monitoring human and temporal impacts on the environment. The retrieved information is important for understanding impact of urban growth and ecological disturbances.

The analyses and results presented in this study already show the large potential of NPP time-series for environmental monitoring. The use of vegetation phenology information based on Sentinel-3 OLCI data as model input will allow for more detailed analyses due to a higher spatial resolution in future research activities.

The emerging field of movement ecology is highly challenging due to its interdisciplinary nature. Including earth observation data and products into the analysis of animal movement for conservation purposes indeed requires a sound understanding of the needs and obstacles in conservation as well as mastering the technical challenges associated with animal movement and remote sensing analyses. Bridging the gaps between these disciplines is at the heart of AniMove – a collective of international researchers committed to support the emergence of interdisciplinary and collaborative work at the interface between ecology, conservation and remote sensing.

The main focus of Animove is training in a variety of spatial modelling techniques that allow the combined use of animal movement data and remote sensing information. All the shared technical expertise is embedded in conservation frameworks to promote a high level of integration between applied ecologists and remote sensing scientists. AniMove training sessions rely exclusively on OpenSource software such as R, GRASS or QGIS. This makes participants independent of their institutions’ software support strategy and empowers them to sustainably use the acquired knowledge.

Training is designed to cover coursework as well as project work. The opportunity to study theoretical examples while working on the students’ own datasets has proven to be highly valuable and appreciated by the attendees. Allowing the students to work with their own data sets also ensures that practical problems can be discussed and solved during the training session. This promotes the establishment of a long-lasting professional and social network among students and lecturers and a high motivation for the participants to continue working in this interdisciplinary field.

The high demand for AniMove.org courses resulted in training sessions currently being organized in North-America, Africa and South East Asia. Moreover specialized OpenSource software packages designed to support the course are being developed, while other forms of outreach activities such as workshops and symposia are untertaken.

Phenology refers to the timing of biological events. Vegetation phenology has important year-to-year variation due to climatic variability, and spatial variation that is determined by both environmental conditions and plant species. Phenological analysis through the use of time series data can provide important input to biodiversity assessment and agricultural monitoring. For example, phenology is proposed as one of the candidate Essential Biodiversity Variables (EBVs). However, spatially-continuous analysis of phenology has been largely limited to coarse resolutions (>250m) due to the requirement of frequent observations of plant photosynthetic activity. This high frequency is needed to closely follow the vegetation’s development while providing sufficient cloud-free observations. New opportunities for monitoring phenology at higher spatial resolutions have emerged with the launch of Sentinel-2 and the related third-party missions RapidEye and SPOT-5 (Take 5).

As part of the ESA Innovators-III funded project RS4EBV (Remote Sensing for Essential Biodiversity Variables), this study examines options and challenges for extracting phenological metrics from Sentinel-2-type data. Three sites are selected in the project, i.e. 1) the Bavarian Forest National Park, dominated by deciduous and evergreen forest, 2) salt marshes on National Park Schiermonnikoog in the north of the Netherlands, and 3) North Wyke Farm Platform in Devon, UK; a managed grassland site. Between April and September 2015, the three sites were observed as part of the SPOT-5 Take 5 experiment. In addition, approximately 10 RapidEye acquisitions were made per site between March and September 2015, and imagery from Sentinel-2A is available since its launch date 23 June 2015.

To evaluate if the multi-temporal satellite data allow for effective capturing of phenological transition dates, a total of 30 digital cameras were installed at fixed locations at the three sites capturing photos from May/June 2015 onwards. We selected the relatively low-cost Bushnell TrophyCam cameras, which we programmed to take ten photos per day of the same scene at 30-minute intervals. From the photo series we calculated the so-called green chromatic coordinate (GCC: brightness in green divided by total brightness) and composited these in time by taking the 90^th percentile GCC over three consecutive days.

We will present comparisons of temporal profiles from camera-derived GCC and from the normalized difference vegetation index (NDVI) derived from the atmospherically-corrected and cloud-screened satellite images for pixels corresponding to the regions-of-interest defined on the camera images. These comparisons illustrate if and where the temporal frequency of the satellite data is sufficient for capturing moments of strong transitions in greenness. Subsequently, we show results of fitting mathematical models (e.g. double hyperbolic tangent) to both GCC series and satellite-derived NDVI series. Based on these models, we then present the first high-resolution maps of phenological metrics for these sites, including onset of greenness, and start of senescence. Finally we will summarize the main challenges with regard to Sentinel-2-based phenological monitoring and validation efforts focusing on temporal frequencies obtained, site-characteristics, and technical considerations.

The development of advanced remote sensing satellite systems, application software for digital image processing and GIS facilitate operational obtaining reliable information about crop moisture provision. Relevance of research is caused by necessity of improvement of the traditional monitoring system of agricultural landscapes in changing climate conditions, including diagnosis of such crises as droughts and operational information providing for business entities. Using modern satellite remote sensing systems contributes to obtaining more timely and accurate information about the crop status, including moisture conditions. R. Jackson (1986), S. Idso (1982), B. Gao (1995), G. Roerink (2000), F. Kogan (2001), S. Zwart (2004), W. Bastiaanssen (2005),  T. Shadchyna (2001) V.  Antonenko (2002), O. Voynov (2005) made a significant contribution to the assessment of vegetation state by remote methods. The aim of the article. To determine the relationship between the crops spectral characteristics of and crop moisture content, to improve space remote sensing methods to evaluate moisture provision of agricultural crops. Methods. The method of determining of the crop moisture content is based on the integrated use of satellite data of different spatial differentiation and in-situ ground data. It is proposed to use water index obtaining according to MODIS / Terra images for the regional monitoring. The local level monitoring is proposed to carry out by moisture model developed by satellite data of high spatial resolution. Results. During the work the effectiveness of modern satellite remote sensing systems for rapid assessment of moisture agrophytocenoses was proved. To determine the agrophytocenosis moisture content at the local level it was proved the possibility of replacement of satellite imagery of high spatial resolution (RapidEye, SICH-2), on medium spatial resolution (Landsat, MODIS/Terra), which are freely available. The relations between the moisture content of winter wheat plants and spectral indices were found based on the results of experimental field research. The mathematical models to determine the plant moisture content by given indices were developed. The maps of the moisture content in winter wheat plants in test sites by obtained models were constructed using modern GIS technology.

Current land use and land-use change have a direct impact on the global carbon cycle and climate through the emissions or removal greenhouses gases resulting from anthropogenic activities. Land use, and its influence on land cover, is also of critical importance for biodiversity and acts as a major driver of the distribution and functioning of ecosystems, and thus in the delivery of ecosystem services.

With more than a third the European agricultural area, grasslands are an important land use in Europe with essential functions for feed and regulating ecosystem services (e.g. reduction of erosion and of water contamination by fertilizers and pesticides) supply. Grasslands also support biodiversity and cultural services through its contribution to region’s cultural heritage and to recreational values.

Land-use change and intensification are causing further fragmentation and homogenization of forests and agro-ecosystems. In areas where an intensification of agricultural practices is observed, there is an increasing pressure on grassland ecosystems through their conversion into arable land. This increasing pressure results in an intensification of grassland management. Grassland management practices (mowing or grazing) and the intensification level of these practices (number of mowings per year, livestock density, level of N fertilization) have an impact on the biodiversity  and the different services associated to this agro-ecosystem such as for example carbon sequestration, nitrate and pesticides leaching risk or pollination support.

Land use science has so far mainly focused on broad land cover conversions while the spatial patterns in the intensity of cropland, grazing, and forestry systems remain highly unclear for most world region. Appropriate integration of remote sensing technologies into ecosystem services concepts and practices, through an assessment of land use intensity, could therefore lead to potential practical benefits for the protection of biodiversity and the promotion of sustainable management of Earth’s natural assets.

Grassland use intensity can be assessed in mowed grasslands based on several features such as the biomass, the floristic composition or the mowing calendar.

The present study, representing a step of a wider approach aiming to assess land use intensity for the specific case study of grasslands, is focused on the detection of mowings by remote sensing in Belgium. As mowings are punctual events characterized by temporal differences due to climate and farm management, the use of multi-temporal image series ideally with a high temporal frequency is therefore required. To face the limited use of optical images in Belgium arising from a frequent cloud cover, mowing calendar have been estimated based on SAR (ERS-2) data.

As a first step, a procedure to discriminate mowed from grazed parcels has been set up based on the difference in backscattering coefficients distributions and the effect of water content on backscattering coefficients. The procedure allows a good discrimination with an overall classification accuracy around 80%. In a second step, a methodology has been set up to detect mowings. The methodology uses a temporal approach and is on the relationship between the differences of backscattering between 2 successive SAR images against the differences of water content. Results for mowings detection are a bit lower as half of the mowings (14 mowings over 29) were correctly identified. Considering the ERS-2 sensor repeat cycle, the methodology can be however considered as promising in particular in the perspective of very dense SAR time series.

In this context, a study is currently underway based on observations collected during 2015 growing season and SENTINEL-1 data to validate the approach. 

Lots of geospatial web portals exist around the globe but most of them focus on providing only metadata, visualization and in some cases also the download of datasets relevant for their users. In general spatial and non-spatial information exist for study areas that are described by the web portals but the link between spatial and non-spatial data is still lacking. The data portal and brokering system from the EU-H2020 funded project Satellite-based Wetland Observation Service (SWOS, http://swos-service.eu/) will overcome this issue and provide not only spatial datasets that are created within the project but especially the collection and provision of further available wetland-related information and datasets, such as scientific papers, links, documents, and images. Connections to existing wetland-specific databases (e.g., the Biofresh database) and other metadata catalogue systems (e.g., the GEOSS Common Infrastructure) will be linked to each wetland area. This should lead to a next-generation “knowledge hub” for wetland areas. The data portal will bring a wide range of wetland-relevant information together with spatial products and services. Visualization and access to specific satellite data (e.g., relevant MODIS, Landsat and Sentinel data) and derived products (e.g., products based on the Copernicus Global Land Service) will be provided. Web-based services and tools will allow analysis and processing of data directly in the portal.

From a technological point of view linked data will be used to connect all information. External systems will be searched, harvested and linked to each wetland area. This approach allows an easy human- and machine-readable discovery of available data and information based on the wetland study areas of the SWOS project. Any kind of discovery, visualization, download, and processing functionalities will be provided with web services compliant to the Open Geospatial Consortium secured by a backend administration interface. The system is also in line with the GEO Data Management Principles (e.g., metadata, web services, data documentation, data provenance). According to the new proposals of the GEO Data Sharing Principles the data collected and created during the project lifetime will be freely accessible as open datasets.

Within the SWOS project a strong focus is set on exchange and cooperation with activities of GEO and their “system of systems”, GEOSS. Any kind of data that is available in the new SWOS knowledge hub will therefore be available within the GEOSS metadata catalogue system. Furthermore, within the project it will be demonstrated how GEOSS could also benefit from the applied linked data approach in terms of organizing and structuring metadata. The SWOS data and brokering system could be identified as a wetland-specific “system of systems”. According to the GEO Work Program 2016 the SWOS project will contribute to the upcoming Architecture Implementation Pilots to enhance the GEOSS Common Infrastructure and to show examples how the GEOSS system can be integrated into other applications.

The concept of the new SWOS Data Portal & Brokering System leading to a new SWOS “knowledge hub” with interconnections to existing databases and catalogue systems will be explained. Examples how the GEOSS Common Infrastructure can be used and integrated in applications will be demonstrated. The presented SWOS infrastructure builds the technical baseline for the implementation of a Global Wetland Observation System (GWOS). Together with other partners like e.g. other relevant projects, GEO and the Ramsar Convention on Wetlands this activity aims for developing a sustained and operational global service that provides information, data and knowledge on wetlands in a user-friendly and open way, tailored to the requirements of policy- and decision-makers as well as other users. 

In the last decades biomass estimation utilizing remote sensing data has moved into the center of attention of interdisciplinary research for the quantification and validation of complex carbon storage. The latter is the most complex part of the carbon cycle and has therefore a huge impact in the analysis of the global climate change. 

To help in this issue, the ENVISAT ASAR based BIOMASAR maps for 2005 and 2010 with a spatial resolution of 1 km give information about Growing Stock Volume (GSV) in Northeast China. GSV refers to the stem/bore volume of living trees for all living species, including bark and excluding branches and stumps. The BIOMASAR maps were produced within ESA Forest-DRAGON 2. The  follow up Forest-DRAGON 3 project focuses on the validation of these products. 

To help addressing this goal, a pure- and mixed pixel analysis for forest areas of the BIOMASAR maps was undertaken. To detect pure-and mixed forest pixels, the National Land Cover Dataset (NLDC) of China for the years 2000 and 2010 as well as ESA's Climate Change Initiative (CCI) Land Cover products for the years 2005 and 2010 were utilized. All further analysis steps were applied to both land cover products and compared in the end. 

For the pure pixel analysis at first a reclassification of the land cover products focusing on forest classes was applied. Afterwards continuous forest pixels according to the land cover products for both years 2005 and 2010 were selected. To ensure purity, the Pixel-Purity-Index (PPI) was calculated on all selected pixels for both years using Landsat 5 Surface Reflectance images and followed by a rejection of all pixels not being 100 percent pure. Furthermore, the NDVI was calculated and applied with the condition of being >= 0.2 to be free of cloud cover problems in the PPI calculation. By this, it was also ensured that pure non-forest areas (due to possible misclassification of the LC products) are not falsely included into further analysis steps. 

Mixed pixel analysis was performed with the same datasets as used for the pure pixel investigation. Mixed forest classes were set up in 25 percent intervals and are based on the percentage share of non-forest classes, e.g. “0-25% shrub & 75-100% forest”. For both land cover datasets (NLCD & CCI) an a priori reclassification was undertaken featuring the three main classes shrub, crop and grassland, ending up with 12 mixed-forest classes in total (3 types á 4 classes). Statistics for each class were calculated over the whole study area on the base of the 1km BIOMASAR grid. 

First results of the pure pixel analysis on a small subset reveal some interesting areas with very low GSV values for both years. However, it has to be further investigated whether this error relies on the BIOMASAR maps or the land cover products. 

The mixed pixel analysis showed similar reasonable results for the mixed classes based on grassland and crop with a decreasing GSV value when featuring a higher percentage of non-forest elements. However, the mixed-forest classes based on shrub showed unexpected behaviors with a non-consistent trend for an increase of non-forest area. For NLCD, no mixed-forest class based on shrubs could be taken out as the number of detected pixels was too small. 

A further investigation of the undertaken analysis is needed with a special focus on the detected unreasonable results of both pure-and mixed pixel analysis. 

This work was undertaken during the Young Scientist Exchange between ESA and NRSCC at Chinese Academy for Forestry (CAF) in China, to contribute to the tasks of Phase 1 and 2 of the ESA-NRSCC/CAF Project "Forest-DRAGON 3”.

Being one of the world’s hotspots for biodiversity and endemism, Myanmar is currently
undergoing enormous political and economic transformations which are likely
to result in increased pressure on its already endangered forest ecosystems. Of particular
relevance in the scope of forest conservation in Myanmar are thereby mangrove
ecosystems, since they are among the Earth’s most imperiled tropical environments.

The highly productive and biologically rich mangrove forests not only possess intrinsic
value due to their uniqueness, but also deliver numerous ecosystem services
to humanity: Mangrove forests provide coastal protection, water pollution filtering,
prevent soil erosion and support the livelihoods of numerous coastal communities
by delivering seafood, construction materials, firewood, timber or charcoal. Because
mangrove forests provide such a wide range of essential ecosystem services, mangroves
are generally valued at 100,000 $ to 277,000 $ per km2.

Even though mangrove forests represent one of the Earth’s most diverse and
valuable ecosystems, unsustainable exploitation and mismanagement of mangrove
forests is steadily increasing at alarming rates. The considerable deforestation of
Myanmar’s mangroves is mainly caused by extensive extraction of firewood, timber
for construction or charcoal production and by the encroachment of agriculture and
infrastructure into the habitat of mangrove forests.

To ensure the sustainable management of mangrove forests, it is necessary that
inventories are undertaken on a regular basis. Satellite remote sensing offers a relatively
cost efficient and rapid method to regularly monitor the extent and changes of
mangrove forests. However, the numerous varying sensor types as well as processing
and classification methods which are available nowadays make it very difficult to
choose the suitable satellite imagery and analysis method which can provide the
best possible results.

This study comparatively examines Landsat 8 and RapidEye satellite imagery obtained in
2014 by investigating three different influencing elements: First, sensor specifications
(spatial and spectral resolution); second, different classification models (Random
Forest, Maximum Likelihood, SVM...); and third different predictors ranging from
spectral bands to moisture and vegetation indices. Performance statistics were calculated
for each influencing element and sensor type in order to test their accuracy
and suitability for mangrove mapping in the context of biodiversity conservation.
The resulting patterns are valuable to analyze the importance of Sentinel-2 data for
the monitoring of mangrove ecosystems.

Soil moisture is one of the main variables in hydrology, meteorology and agriculture, as it determines the partitioning of both incoming solar radiation into latent and sensible heat and precipitation into surface runoff and infiltration. Nevertheless, its spatio-temporal patterns in agriculturally used landscapes, that are affected by multiple natural (rainfall, soil, topography etc.) and agronomic (fertilisation, soil management etc.) factors, are often not well known. The aim of this study is to derive a long term time series of surface soil moisture patterns from SAR data in order to create a soil moisture climatology, to observe long term surface soil moisture pattern development and to identify the main drivers for these patterns in a mesoscale catchment (2354 km²). A time series of 85 single-channel C-band ERS SAR scenes from 1995-2003 with a spatial resolution of 20 m was used to create this soil moisture climatology.

Our research area, the Rur catchment, is located in Western Germany and is part of SFB/TR32 “Patterns in soil, vegetation, atmosphere systems: monitoring, modeling and data assimilation” (www.tr32.de). It consists of two major landscape units, with the northern part belonging to a mainly agriculturally used fertile loess plain and the southern part belonging to the low mountain range of Eifel, dominated by forests and pasture. Surface soil moisture maps of agricultural areas and pasture were calculated, using a semi-quantitative soil moisture index, grouping the backscatter values from SAR scenes into different classes of soil moisture content. With this approach a simple and robust method to derive soil moisture classes ranging from dry to wet is developed which is particularly useful in the absence of ground truth data. The radar backscattering coefficient (s₀) is sensitive to vegetation, surface roughness and soil moisture content. To eliminate the influence of vegetation and surface roughness, the data was sorted into classes as follows. Using land use maps for each year (derived from optical remote sensing) the data was grouped by agricultural land use class (cereals, sugar beet, maize, bare soil, pasture) and month. Within each class the influence of biomass and surface roughness is assumed to be constant. Thus, changes in backscatter intensity can be accounted to changes in surface soil moisture.

Given a sufficiently long time series of satellite images, we assume that pixels exist within each group which represent a wet and a dry soil moisture state. The qualitative index is derived by grouping backscatter values into classes ranging from dry to wet. By using soil parameters (porosity, field capacity and wilting point) according to the soil texture of the given location, the qualitative index can be converted into volumetric soil moisture.   

The approach was tested i) using the full resolution 20 m pixels and ii) using medians of the agricultural fields. In order to validate the approach, resulting soil moisture maps were compared to antecedent precipitation index patterns. The surface soil moisture maps were then used to create a soil moisture climatology, which depicts the typical, average state of soil moisture for every month. The climatology can be used as a reference for comparison to upcoming soil moisture studies in the Rur catchment.

Today, dual-and full-polarimetric SAR data at C- and L-band from ALOS-2, Radarsat-2 and Sentinel-1 provide high quality radar backscatter and polarimetric observables. Based upon previously developed semi-empirical retrieval algorithms we use these data in combination with a concise ground truth data set on soil moisture and vegetation parameters to estimate surface soil moistures and high resolution soil moisture patterns in the Rur catchment for the vegetation period of 2015. The comparison of the different soil moisture maps provides information on the advancement in soil moisture retrieval methods.

One of the main goals of ESA´s Data User Element (DUE) is to bridge the gap between research activities and the sustainable provision of Earth Observation products, which in turn aim to satisfy the operational needs of larger user communities. Consistently with this goal, BlackBridge and the University of Wageningen are working together on the development of products and solutions to support monitoring of deforestation and forest degradation for REDD+ Measuring, Reporting and Verification (MRV). The Forestry Department of FAO, as the worldwide technical expert of the UN-REDD program, is the end user - or direct beneficiary - of the results achieved by this project.

The ForMoSa project focuses on the development of methods for mapping and quantifying deforestation and forest degradation, based on the integrated use of available remote sensing satellites, such as Landsat 7 and 8, RapidEye, Sentinel-2 and SPOT-5. Project goals and outcomes were defined by the project team in close cooperation with FAO, as the expert user, to ensure that the results achieved are in full compliance to the REDD+ MRV monitoring requirements. Three demonstration or study areas in different continents, in Vietnam, Ethiopia and Peru, were chosen to represent the natural global diversity and to reflect the complexity of the REDD programs around the world.

In this project, forest natural variability and dynamics are modelled on a per-pixel basis, to detect departures from normal conditions as potential indicators of different degrees of forest canopy disturbance. These pixel-based models will be complemented with forest mapping protocols that will serve as framework for forest delineation.

To achieve the necessary robustness, models require long series of reflectance data. A sufficient amount of accurate data is also necessary for the subsequent monitoring period. Remote sensor interoperability is the necessary answer to the vast amount of accurate remote sensing data required in any REDD program. A combination of various information sources not only results in more data but also enables a complementarity, by which different characteristics of sensors can be combined to act synergistically for the improvement of the solutions to be developed. Different sensors are to be normalized to one another by applying spectral band adjustment factors (SBAFs), and thus enabling the building of a solid database.

The main outcome of the ForMoSa project is an operational concept for forest monitoring to be used in national REDD+ MRV programs. This includes protocols for the pre-processing of images to produce long, consistent data series, and the production of forest cover and forest cover change maps. A measure of the intensity of forest canopy changes at each location can be re-signified into REDD relevant activity data (forest degradation and deforestation), and the subsequent gas emission quantification.

Water is a major concern for environmental policies, and for maintaining ecosystems. The water transportation makes use of piping and open tunnels systems. These transportation systems use infrastructures which suffer from aging and are not frequently replaced. The wear of the infrastructure causes water losses, which might reach high percentages. Water leakages from distribution networks are as high as 50% in certain areas of Europe (European Environment Agency).

To reduce water loss caused by transportation leakage, the detection at the earliest time is particularly important. The water leak in the urban area is easy to detect, but not the case in the open field. To detect water leak in open spaces outside the urban area, the remote sensing technology can play a big role. Remote sensing has been used in a few reported cases to detect leaks in aqueducts, either directly through changes to soil and surface water content or indirectly through effects on the vigor and health of overlying vegetation. The direct remote sensing of water leaks is based on the detection of the soil moisture produced by the leak and can be based both on optical and thermal sensors as well as on active microwaves. The principle for water leaks detection with SAR data is their ability to detect soil moisture and their capability to penetrate dry soils and thus to detect underground moisture. 

In our study, Synthetic Aperture Radar (SAR) remote sensing images, including Envisat ASAR data and Sentinel 1 data, are used for this detection. 

The studied network is located in Tarragona (Catalunya), and it belongs to the Consorci d’Aigues de Tarragona (CAT). CAT infrastructure includes almost 400 km of pipes and is managed by a remote management system with a high degree of automation that centralizes the operation and control of a large number of facilities. They maintain a comprehensive leak database that includes location, duration, estimated leak start date and repair date that has been used for this study. 

With the prior knowledge of the leakage, we found the consistency with the result shows in remote sensing data. After data processing, the backscatters of the images were analyzed for several leakages, then leak detection methodologies were studied to detect the leakage along the water transportation network. The leakage can be seen by all the method, but along with other “potential” leakages within the network, which are the noise to be eliminated. Since the water leakage of transportation network only changes the image backscatter in a quite small scale, the analysis should be in pixel size, making the result sensitive to noise. However, with the combination of different bands (VV and VH) conditions, lots of noisy points can be removed, and improvement could be done by reducing surface roughness and vegetation cover effects, and combining data from different sensors.

Today, the certification industry for organic farming makes limited and superficial use of mobile technology and smartphone services, and no use of Earth Observation data. The developed demonstrator addresses specific challenges in the field of certification and aims to provide a new innovative tool making in-field inspections more efficient and effective. It offers incentives for developing, testing and implementing new approaches for organic certification that make use of satellite technologies in general and especially Copernicus.

To support the field audits and give the auditors a source of information independent of the farmers’ documentation and book-keeping, optical high-resolution satellite data (Landsat-8, RapidEye, Sentinel-2) is used to derive indicators about the conformity of individual fields with the EU regulation for organic farming. These indicators span from basic tests of data consistency (are the field boundaries given correct?) to actual interpretation of the management of the fields (do the biomass and chlorophyll values indicate a treatment with high fertilizer amounts?). Within myEOrganics, the system was developed and validated for different cereals as well as for corn.

The crop parameters derived from the EO data are then translated into a simple traffic light color system (green = everything is okay, yellow = results are indecisive, red = warning flag to check the field) and fed into a web- server from where it can be downloaded into a mobile app by the auditor to use during his on-site audits. The service is designed to be integrated seamlessly into the field audit. It is a support tool for the auditor, not meant to replace him, and a warning flag does not necessarily point toward a  non-conformity with the EU regulations. It does, however, show an outlier and can, as such, trigger a targeted technical discussion   with the farmer.

After having shown the principle validity of the EO approach within EOrganic (a feasibility study funded by ESA, cntr.no. 4000100515/10/l-AM), for farms in Eastern Germany and France, the service was demonstrated for several farms in Bavaria in 2015 and validated together with the auditors and organic farmers of the respective farms with very good results. Outliers were detected and flagged. These were checked with the farmer by the certifier. Resulting, all outliers were confirmed but had reasons that could be explained, that may have required updates in terms of crop changes. Non-conformities were not found within the test panel of farmers.

One of the spectral indicators that are used is the chlorophyll content of the fields which is a proxy for the nitrogen fertilization level. Organic treated fields usually have lower nitrogen input. Sensitivity studies showed that additional bands in the red edge bands are favorable for chlorophyll content retrieval. Thus it is expected and demonstrated with first Sentinel-2 images from summer 2015 that Sentinel-2 will improve this indicator derivation.  

This large scale demonstrator (funded by the EU within the EMMIA 2 initiative, cntr.no. SI2.648479) is a first step towards showing how to efficiently use COPERNICUS and GNSS technology to develop innovative mobile services for organic certifiers in green farming. It can be used as a blueprint for other applications and users. Supporting organic certification contributes to the societal challenges of sustainable agriculture, food security and consumer’s safety. myEOrganics thus focuses on EU’s Common Agricultural Policy: promoting a safe, clean, environmentally friendly, competitive and sustainable agriculture in Europe, and uses satellite technologies as a globally available, reliable and effective means to do so.

The Cerrado (Brazilian savanna) is the second largest biome in South America (204 million hectares), occupying about 24% of the Brazilian territory. It is considered the most biologically rich savannah ecosystem in the world and a global biodiversity hotspot for conservation priorities. The Cerrado is Brazil’s primary agricultural production area. Since the 1970s, the Cerrado has drastically changed through land use intensification by agriculture, with far reaching implications on the ecosystem and its biodiversity. Extensive cattle ranching has resulted in deforestation and an increase in fire frequency. As in most savanna ecosystems, fire plays a major role in the Cerrado and while some are natural fires, humans start the great majority thereof. Fire return intervals are very short and are reported between one to five years. The spatial pattern of fine fuels from the surface grass/herbaceous layer is the major determinant for fire patterns.

The Brazilian-German Cooperation Project “Prevention, control and monitoring of bushfires in the Cerrado” aims at implementing integrated fire management (IFM) in the Brazilian savannah. The main objective of IFM is to reduce the negative impacts from unplanned fires on people, property and biodiversity. An important component of IFM is managing the fuel load through prescribed burning beginning early in the dry season with the goal to minimize high intensity fires in the late dry season. Planning and implementation of IFM and prescribed burning activities (where, when and how to burn?) require reliable information on the distribution and amount of fuel loads.

The objective here is to demonstrate how remote sensing based fuel load mapping benefits the planning, implementation and evaluation of controlled early burning activities. Estimates of green and dry biomass based on Landsat 8 OLI data were provided as an indicator for fuel load in support of the IFM zoning and used in the planning of controlled early burning. 37 OLI scenes have been acquired in total during the observation period between January 2014 and August 2015. A threshold of 50% cloud cover over the focus area was applied, resulting in 17 scenes that could be used in this study. The Mixture Tuned Matched Filtering (MTMF), was applied to the data, whereby high-resolution RapidEye data were additionally used to find optimal spectral endmembers and to monitor the controlled burning.

The final fuel load maps were qualitatively validated through field visits and quantitatively validated by the use of pre- and post-fire collected biomass data [kg/m²]. The fuel load maps not only highlight areas with high fuel loads, but also reflect the recent fire history of the Cerrado. The matched fractions of dry vegetation (NPV), green vegetation and soil from the MTMF were analyzed over time in order to assess the different characteristics of the fuel load in different Cerrado types. The fuel load plots not only allow monitoring of the fuel loads over time, but were also used to identify and further analyze fire prone locations. The comparison of the biomass data and the matched fractions of NPV showed a coefficient of determination of r²=0.80. Besides this good fit, the scatterplot shows that very low biomass values below ~0.25 kg/m² have matched fractions of zero and cannot be further differentiated. A negative correlation was found between soil fractions and biomass with an r²=0.50.

These fuel load maps were provided to park managers as a planning tool for controlled early burning. Due to the demonstrated wide-ranging benefits of the fuel load maps, an operational provision of these maps through Brazilian authorities is envisaged. Sentinel-2 data further add benefit to an operational monitoring of fuel loads due to its optimal spectral band set-up and short revisit times.

Disturbances are a fundamental component of forest ecosystems. Spatially and temporally-explicit databases of disturbance events are useful to assess recovery patterns and rates, and analyze the effects of disturbance intensity and time since disturbance on ecosystem processes and services, e.g., protection from gravitational hazards. We compiled a disturbance database for the Valle d’Aosta region of Italy, and used it to assess current vegetation cover in direct protection forests disturbed by abiotic or biotic agents.
To do this, we georeferenced disturbance polygons (wildfire, insects, storm damage, and avalanches) for the period 1961-2010, and superposed them to a map of direct protection forests, based on slope and the existence of vulnerable targets. We classified and validated land cover classes (forest, grassland, bare soil) from a 2011 Landsat TM5 image, using a fuzzy approach. Finally, we applied our classification rules to pixels in disturbed protection forests. 
The result was a map of vegetation cover classes in existing protection forests of the region. This can be used by managers in order to prioritize restoration activities where protection from gravitational hazards is needed but not currently provided due to past disturbance events.

Sustainable forest management requires flexible, reliable and up-to-date information. Possible solution for updating forest attributes is Object-Based Image Analysis (OBIA), which are presently being investigated and optical multispectral sensors are being considered as a data source to cover large areas rather than the conventional at plot-level forest inventory methods. The OBIA approach persists to reveal its effectiveness in remote sensing data analysis, which provides paradigms that integrate analyst’s expert knowledge to generate semantically meaningful image-segments in way that emulate human perception. These segments might contribute for the reduction of problems associated with the analysis of discrete spectral value of pixel, such as illumination and shaded tree crowns. However, the challenge in this paper is to introduce OBIA as a sophisticated framework toward the automation of forest structural attributes estimate. Aster scene was examined to develop models for the estimation of stand density, crown height, basal area, and volume. The integration of high-resolution satellite imagery offer exciting possibilities to enhance the measurement, mapping, and modeling of forest attributes. In this context, the higher resolution RapidEye scene was introduced to test the feasibility of finer resolution. Moreover, one of the addition values of this work is to predict a vertical attribute (i.e. crown height), which it has been rarely investigated based on optical sensors. Analyses were performed over a forested selected site in Blue Nile region of Sudan. The framework of the present research involves the following main pillars; generation of meaningful segment, integration of ground truthing gis samples, forest attribute generalization, correlation, regression model, model validation, and mapping structural attributes based on the selected models. The rationale for incorporating these analyses are to offer semi-automatic OBIA metrics estimate from which forest attribute is acquired through automated segmentation algorithms at the delineated tree crowns or clusters of crowns level (i.e., mean segment size of 0.04 to 0.09 ha). Correlation and regression analyses were applied to identify the relation between wide range of spectral, textural, contextual image metrics from one side (e.g., Digital Number, grey-level co-occurrence matrix, Object Layer Value, etc.), and the field derived forest attributes from the other side. Based on the highest determination coefficient (R²) and the lowest Residual Mean Squire Error (RMSE) criteria, forest structural attribute estimation results acquired from our OBIA framework reveal strong relationships and precise estimates. Subsequently, the best fitted models were cross-validated with independent set of field derived samples. The result confirmed that there was a good agreement between those points and the predicted forest stand volume achieved from the best fitted models (R² = 0.80 and 0.76 for the both resolutions of RapidEye and Aster data respectively). An important question is how the spatial resolution and spectrum used affect the quality of the developed model will be also discussed based on the different sensors examined.

Surface Temperature (ST) changes in the Polar Regions are predicted to be more rapid than either global averages or responses in lower latitudes. Observations increasingly confirm these findings, their urgency, and their significance in the Arctic. It is, therefore, particularly important to monitor Arctic climate change.

Satellites are particularly relevant to observations of Polar Regions as they are well-served by low-Earth orbiting satellites. Whilst clouds often cause problems for satellite observations of the surface, in situ observations are much more sparse. The ATSRs are accurate infra-red satellite radiometers, designed explicitly for climate standard observations and particularly suited to surface temperature observations. ATSR radiance observations have been used to retrieve sea and land surface temperature for a series of three instruments over a period greater than twenty years. This series will be extended with the launch of SLSTR on Sentinel 3, which has the same key design features necessary for providing climate quality surface temperature datasets.

We have combined land, ocean and sea-ice surface temperature retrievals from ATSR-2 and AATSR to produce a new surface temperature dataset for the Arctic; the ATSR Arctic combined Surface Temperature (AAST) dataset. The method of cloud-clearing, use of auxiliary data for ice classification and the ST retrievals used for each surface-type will be described. We will establish the accuracy of sea-ice and land-ice retrievals with recent results from validation against in situ data and how they compare with other retrieval algorithms over land ice and sea ice. We will also discuss the results from the calculation and propagation of uncertainties in the AAST dataset. Time series of ST anomalies for each surface type will be presented. The time series for open ocean in the Arctic Polar Region shows a significant warming trend during the AATSR mission. Time series for land, land-ice and sea-ice show high variability as expected but also interesting patterns.

Overall, our purpose is to present the state-of-the-art for ATSR observations of surface temperature change in the Arctic and hence indicate the confidence we can have in temperature change across all three domains, and in combination. Currently there is no plan to provide sea-ice ST as a core product from Sentinel 3. We make the case for a near real time Arctic surface temperature product from SLSTR which would include surface temperatures for all three domains: land, sea and sea-ice.

Land cover classification by means of remote sensing data is of fundamental importance for a broad range of commercial and scientific uses. In particular, the detection of vegetated areas is of great interest for global change research and for applications in agriculture, cartography, geology, forestry, as well as for regional planning. At a global level, only in the last years high-resolution forest classification maps have been produced [1], [2]. This paper presents a method to generate forest/non-forest maps from TanDEM-X interferometric SAR data. The TanDEM-X mission comprises the two twin satellites TerraSAR-X and TanDEM-X which fly in close orbit formation of a few hundred meters distance and act as a large single-pass SAR interferometer (InSAR), with the main goal of producing a global and consistent Digital Elevation Model (DEM) with an unprecedented accuracy [3]. The radar backscatter and the interferometric coherence are both sensitive to the type of land cover under illumination, and may therefore be exploited for classification purposes. In particular, the interferometric coherence represents a key quantity for the assessment of InSAR performance, and describes the amount of noise affecting the interferogram. Among the several error sources which deteriorate interferometric data, the coherence loss caused by volume scattering represents the contribution which is predominantly affected by the presence of vegetation. The method for estimating the volume decorrelation contribution from TanDEM-X interferometric data and its use for the identification of vegetated areas is detailed in [4].

Since the beginning of the TanDEM-X mission (end of 2010), about half a million high-resolution bistatic InSAR scenes have been acquired covering all the Earth’s land masses. A single bistatic scene has typically a ground extension of about 30 km in range by 50 km in azimuth. From each of them, quicklook images for several SAR and InSAR quantities (like the coherence matrix or the roughly calibrated DEM) are generated by applying a spatial averaging process to the corresponding operational TanDEM-X interferometric data in full resolution. They have a ground pixel spacing of about 50 m × 50 m, and are taken into account as input data base for generating quicklook mosaics. Hence, large-scale mosaics of the radar backscatter and of the volume decorrelation of TanDEM-X quicklooks data are used as input for applying a classification method based on the c-means fuzzy clustering algorithm, whose potentials in the context of the TanDEM-X mission have been already shown in [5]. During the whole mission duration, the global land masses have been acquired at least twice. In order to further improve the performance, many densely forested areas have been covered up to four times in a time frame of several years (from 2010 to 2015). Such a unique, timely, and manifold data set can be exploited to get up-to-date information as well as for detecting possible changes in the forest cover. To improve the classification capabilities, invalid areas, such as water bodies, urban settlements, as well as regions affected by strong geometrical distortions need to be opportunely filtered out. The obtained results are then quantitatively validated with existing forest/non-forest maps as well as by means of external land cover classification data.

The BACI (Biosphere Atmosphere Change Index) project (http://baci-h2020.eu/index.php/) aims to develop new methods to combine observations from a wide range of sources, from the point to the regional scale, and at a range of temporal and spatial scales, to provide indicators (index) of change in essential ecosystem variables (EEVs) and biodiversity properties, and so-called essential biodiversity variables (EBVs) [1]. The approach combines machine learning (ML) methods with Earth Observation (EO) and field data to identify changes in land surface properties which may indicate changes with implications/impacts on ecosystem properties, particularly related to biodiversity. The aim is primarily to develop optimal ways to detect change, rather than to attribute potential causes of, or mechanisms explaining observed changes.

BACI is exploiting multiple existing and scheduled space-borne EO data streams to detect changes in ecosystem functioning and services that have repercussions for EBVs, land use potentials, and land-atmosphere interactions. A particular focus of BACI is the exploitation of EO data from multiple platforms, including the potential for combining optical, microwave and potentially thermal EO data. This approach poses a number of challenges for merging data across spectral domains and temporal and spatial scales, but with the potential for significant benefits. Chief among these is the development of an optimally sensitive ‘surface state vector’ that may be applied generally to indicate change, and then tuned locally with specific observations of e.g. plant traits, species abundance, ecosystem type etc. A data assimilation (DA) framework is the ideal tool for approaching this application, as it provides a framework with which a set of heterogeneous observations, together with ancillary observation can be combined into an inference of surface state change, fully qualified by uncertainty[2]

Here, we present results from applying the ESA EO Land Data Assimilation Scheme (EOLDAS, http://www.eoldas.info/, http://www.assimila.eu/eoldas/) to observations from multiple sensors over a set of core sites identified within the BACI project. EOLDAS was developed specifically to allow DA incorporating multiple sensors, particularly ESA Sentinels, and uses computationally efficient radiative transfer (RT) model inversion schemes to enable assimilation of large amounts of data. The test sites used here were selected due to their importance and likely sensitivity to change; we show that the EOLDAS DA scheme is capable of modelling surface state change over these core sites from optical data from different sources, notably MODIS, MERIS and SPOT-VGT. We discuss the assumptions and developments required to extend the approach, both in terms of wavelength, but also in terms of spatial and temporal sampling. We show how incorporating observations and time series information from microwave (particularly ESA Sentinel 1A, ENVISAT ASAR, ALOS PALSAR and ALOS 2 PALSAR 2) may provide additional constraints on the surface state vector. We present the first application of this DA approach to identifying EEV and EBV change hotspots.

Current and historic data records of optical EO satellites have already proven to be an excellent source for transparent, timely and consistent information on agricultural productivity at field, regional or national scale. The Austrian-funded CropMon project prepares for the exploitation of next generation satellite systems, particularly the European Copernicus Sentinel-2 mission, to set up operational crop monitoring services based on a range of vegetation indices and bio-physical indicators. The global coverage of S-2 at unprecedented spatio-temporal resolution will facilitate the identification and precise mapping of crop areas, crop type (i.e. summer and winter crop) and crop status, that can serve as an indicator for productivity.

The near-real time provision of these parameters is useful for monitoring crop condition and soil water balance. Such observations can also be used in crop growth models to predict crop yield. The combined historical and near-real time information provides quantitative information for sustainable development and management of agriculture.

The full exploitation of Sentinel-2 observations in the context of operational crop monitoring systems requires preparatory activities to enable a dedicated processing and dissemination capacity. In the CropMon project, supercomputer-ready processing chains to identify crop areas and types, as well as for condition monitoring, are being set up. The mid-term aim is to establish an online, interactive crop monitoring service order and delivery platform enabled by the Earth Observation Data Centre for Water Resources Monitoring (EODC), where users can access these services independently via web and mobile enabled data channels.

Megacities are symbols of highly dynamic urbanization processes. To keep record of urban dynamics, dimensions and complexity, earth observation plays a central role. Hyperspectral remote sensing data offer the opportunity to map urban characteristics in detail.

The upcoming hyperspectral imaging spectrometer EnMAP will provide a detailed look at the physical conditions and the distribution of urban surfaces worldwide. This allows supplementation of urban mapping products, e.g. Urban Atlas (GMES) and Global Urban Footprint (DLR), with enhanced information about details on the material characteristics of urban surfaces in a high spatial resolution. Though, increasing data dimensionality and high redundancy between individual bands might cause problems during data analysis.

Recent work has demonstrated the potential of methods from the field of machine learning, which turn out to be useful for high dimensional hyperspectral data. While more recent machine learning algorithms, e.g. support vector machines, perform well in hard image classification tasks, the potential of their probabilistic outputs for sub-pixel land cover mapping has not been examined sufficiently yet.

The overall objective of this study is to explore the potential of the EnMAP mission for urban areas, as well as to develop concepts and processes for multiscale sub-pixel mapping and unmixing of relevant urban landcover and land use classes. This study focuses on the quantification of sub-pixel information using simulated EnMAP data acquired over the city of Berlin and probabilistic outputs of kernel-based Import Vector Machine (IVM) and Support Vector Machine (probSVM) classifiers.

The study area depicts a subset of Berlin’s urban rural gradient and covers an area of approximately 6 by 23 km. Two HyMap images with 3.6 and 9 m spatial resolution were acquired during the HyEurope flight campaign provided by the DLR. A spaceborne EnMAP scene with 30 m resolution was simulated based on the 9 m HyMap dataset using the EnMAP end-to-end simulation tool (EeteS). A spectral library consisting of various natural and anthropogenic landcover classes was derived from the 3.6 m image by the German Research Centre for Geosciences. In addition, very high resolution orthophotos are available for validation purposes.

The experimental setup incorporates a modified hyperparameter selection approach for finding optimal kernel- and regularization parameters to improve the description of land cover fractions by means of probability outputs. This procedure comprises synthetically mixed map spectra, which serve as an independent source of reference data during model selection.

The results show significant correlations between modeled and reference land cover fractions for both IVM and probSVM. The methodological workflow is also applied to multispectral image data of varying spatial resolution (Landsat, SPOT-5), and results will be compared.

Our findings indicate that probabilistic machine learning algorithms are convenient for urban sub-pixel analysis based on hyperspectral EnMAP data. It can be concluded that the proposed methodology provides a surplus for evaluation of EnMAP data and presents a promising alternative to conventional unmixing and regression approaches.

Bucharest is a very dynamic capital-city in terms of economy and demography and thus its urban extension, along new built-up areas and changes of land use/land cover, have a direct influence on many environmental issues, for example the variation of spatial distribution of temperatures. The main objective of this research is to demonstrate the relationship between land cover/land use classes and temperature variations, depicted from satellite imagery calculation. Data from several meteorological stations and also field measurements were used in order to establish the years in which the highest values were recorded during summer and also to calibrate the temperature variable resulted from satellite image processing. Four Landsat 5 TM and one Landsat 8 OLI scenes acquired in the second part of July, from the years 1987, 2001, 2007, 2012 and 2015 were selected for this study, emphasizing the use of thermal infrared band in assessing spatial distribution of temperatures and studying the urban heat island phenomenon. Also, land cover/land use classifications were conducted. Data validation was done based on in-situ instruments (LUCAS), other LCLU databases and field campaigns. In order to demonstrate the correlation between land use/land cover changes and LST estimation, a multi-temporal analysis was conducted. Over the past 25 years, the urban landscape in Bucharest has been changing as new elements were newly constructed or transformed: numerous residential neighborhoods in the city and mostly in its peripheral areas, vast commercial areas, business centers etc.

The GREENHOUSE project (Generating Regional Emissions Estimates with a Novel Hierarchy of Observations and Upscaled Simulation Experiments, http://www.greenhouse-gases.org.uk/projects/greenhouse) is developing observation and modelling frameworks to combine UK field data on GHG emissions with EO and other spatial measurements at a range of scales. GREENHOUSE is aiming to produce accurate UK GHG inventories, not just CO₂, but also other smaller but important (from a climate perspective) biogenic fluxes of CH₄ and N₂O. In addition, GREENHOUSE is integrating observations into two large-scale land surface models to improved model-based estimates of GHG emissions, particularly in the light of agricultural management. Using field-scale knowledge to infer estimates of emissions at large scales is a major challenge [1]. The challenge arises because of inherent spatial and temporal variability in the underlying processes driving GHG exchanges [2]. Improved attribution of terrestrial fluxes, at regional scales is crucial in developing GHG inventories that reflect the consequences of current land management practice and allow assessment of the impact of climate change/variability, and land use and land management change on emissions. 

Addressing these issues requires combining data and model outputs to integrate surface fluxes over a range of scales [3]. In particular, using Earth Observation (EO) data allows for improved parameterisation of large scale models as well as constraining model operation, particularly when used in a data assimilation (DA) framework [4]. This latter approach is potentially extremely flexible and powerful for upscaling model-based predictions of GHG estimates across the landscape (with uncertainty), as well as for testing model predictions. A DA approach requires EO data with well-characterised uncertainty, as well as some form of observation operator (radiative transfer model) to translate EO data into properties the model can ingest into its state vector. Here, we present application of an EO Land Data Assimilation Scheme (EOLDAS, http://www.eoldas.info/, http://www.assimila.eu/eoldas/) developed explicitly for land surface modelling. We show how improved (optimally smoothed in space and time) EO-derived estimates of land cover and albedo can improve calibration of large-scale land surface models (LSMs, namely the Joint UK Land Earth Simulator, JULES; and the Carbon-coupled Tiled ECMWF Scheme for Surface Exchange over Land, CTESSEL) over core agricultural sites in the UK, covering crops, pasture and some managed forestry. We combine time-series of observations from MODIS, SPOT-VGT and MERIS data, to derive model parameters and driving data. We also explore how higher spatial resolution data (field-measured, airborne, Sentinel-2) can be combined in the DA scheme to address upscaling of model-based GHG estimates, with the potential to provide estimates of GHGs over wider agricultural landscapes. 

The launch of Sentinel-2 holds great promise for a range of applications related to mapping and monitoring of Earth surface processes. This includes assessment of vegetation, land-cover, land-use, agriculture, urban change, flooding, and much more. Through provision of high quality multispectral imagery, with frequent revisit times, straightforward access and rapid dissemination, there is great potential for the development of a range of end user products. This research presents the results of a feasibility study, which is evaluating Sentinel-2 imagery for operational provision of land-cover products at national-scale, considering the case of Scotland, UK. This explores the feasibility of generating national-level products, and examines some of the main considerations and challenges, highlighting potential areas for future development.

The EC-ESA Copernicus earth observation (EO) programme presents a step-change in EO provision and access not just across Europe, but globally. Perhaps one of the most promising developments is the successful launch of Sentinel-2A. As a medium resolution multispectral system, this offers significant potential across a range of application fields in the context of many ongoing global societal challenges related to food and water security, climate change, and disaster mitigation. The high frequency of data capture and the medium spatial resolution make Sentinel-2 ideal for amongst other things, monitoring of land use and land cover change, as well as assessing the impact of extreme events and supporting policy decision making. The presented research considers the development of an operational workflow for generation of (initially) basic land-cover products, and potential for using this as a foundation to derive more focussed outputs, through integration with complementary datasets (e.g. digital soil mapping, agricultural data, etc.).

Scotland, like much of northern Europe, is affected by high levels of cloud cover, which to-date has significantly hampered the usefulness of optical satellite imagery, particularly in relation to regular monitoring of land surface processes. However, the high revisit frequency of Sentinel-2 (five days with Sentinel-2A and 2B) offers enhanced opportunities in this regard. This research evaluates current provision through Sentinel-2A (10 day revisit), simulating likely frequency of coverage once the full system is operational (2016 onwards). Additionally, challenges associated with accessing, archiving and processing typical data volumes are investigated. Established methodologies for land-cover classification are explored and compared, reporting on a representative test site, and comparing proprietary and open source software solutions. This delivers land-cover classification, with emphasis on achieving multi-temporal change detection (land-cover monitoring) on a monthly basis. The data will is validated through accompanying ground truth measurements to assess classification accuracy as well as reliability of detected changes. From the outset, the research engages with likely stakeholders in government, public bodies and other agencies, informing on end user and operational requirements.

Considering these aspects, the research will report on the challenges and opportunities associated with scaling this up to national (Scotland) level, considering optimal classification methodologies; revisit frequency; computational resources, and potential for associated spin-off products. This reveals requirements for investment in data infrastructure and data processing capacity to generate consistent and robust data products at a national level, at a frequency that is relevant for monitoring processes and policy-relevant assessment.

The Landsat Albedo Product provides higher resolution albedo values by coupling  30m Landsat surface reflectances with concurrent coarser resolution (500m) MODIS Bidirectional Reflectance Distribution Functions (BRDF) Products to fully capture the surface anisotropy (Shuai et al., 2011, 2014, Wang et al 2015).  Here we present further development of this method as well as its extension to the newly released sample data from the 10m resolution Sentinel-2A MSI sensor.  

The Landsat Albedo Product uses Landsat 5, 7 and 8 to generate time series albedo from 2000 to present. The use of higher resolution surface reflectance data paired with temporally concurrent daily BRDF information allows for the generation of an accurate snowfree albedo product (RMSE 0.0191). The availability and the increased radiometric fidelity (12 bit) of Landsat-8 OLI and the newly launched Sentinel 2A data now allows for the inclusion of high-quality, unsaturated surface reflectance and albedo calculations over snow covered surfaces (RMSE 0.0426 for Landsat-8).  Here we present the Landsat-8 albedo product validated over snow and snow-free land surfaces (Fig 1) in North America.   In addition, we present newly implemented processing protocols which provide more temporally and spatially consistent albedo generation through the automated and optimized mosaicking and resizing of MODIS scenes to ensure landscape-scale accuracy of the applied BRDF adjustments.  The new processing protocol is being validated at 9 tower sites in the United States and Canada over a diverse range of land covers. 

We also present a preliminary albedo product for the newly released Sentinel-2A images and compare it with both available tower data and the Landsat-8 albedo product.  The 12-bit radiometric fidelity of the Sentinel sensor and the similar band configurations allow for the implementation of this method on the Sentinel: Landsat: MODIS coincident bands.   The ability to produce higher spatial resolution albedo values at increased temporal resolutions will be critical in studying land surface changes over the dynamic and heterogeneous landscapes most susceptible to climate change (such as arctic, coastal, and high-elevation zones).  

REFERENCES

Shuai, Y., Masek, J. G., Gao, F., & Schaaf, C. B. (2011). An algorithm for the retrieval of 30-m snow-free albedo from Landsat surface reflectance and MODIS BRDF. Remote Sensing of Environment, 115(9), 2204–2216. http://doi.org/10.1016/j.rse.2011.04.019

Shuai, Y., Masek, J. G., Gao, F., Schaaf, C. B., & He, T. (2014). An approach for the long-term 30-m land surface snow-free albedo retrieval from historic Landsat surface reflectance and MODIS-based a priori anisotropy knowledge. Remote Sensing of Environment, 152, 467–479. http://doi.org/10.1016/j.rse.2014.07.009

Román, M. O., Schaaf, C. B., Lewis, P., Gao, F., Anderson, G. P., Privette, J. L., … Barnsley, M. (2010). Assessing the coupling between surface albedo derived from MODIS and the fraction of diffuse skylight over spatially-characterized landscapes. Remote Sensing of Environment, 114(4), 738–760. http://doi.org/10.1016/j.rse.2009.11.014

Wang, Z., Erb, A.M., Schaaf C.B., Sun, Q., Liu, Y., Yang, Y., Roman, M.O., Shuai, Y., Casey, K. (2015). Early spring post-fire snow albedo dynamics in high latitude boreal forests using Landsat-8 OLI data.  Remote Sensing on Environment. In press.

 

Mt. Etna (Italy) is a composite strato-volcano and one of the most active volcanoes in the world. Eruptions occur almost every year and there is a persistent degassing activity at the summit craters. In the last 100 years it produced in average 10⁷m³ of new lava per year.

The main goal of our work is to discriminate different lava flows as well as detect land cover changes over the volcano that occurred between 2009 and 2012.

For this purpose, we analyze two Hyperion L1T scenes acquired on 8/10/2009 and 14/7/2012. The time span of the two scenes encompasses several eruptions, especially at the summit “New Southeast” crater. Both images have near nadir acquisition angles and low cloud coverage. Level 1T products are already geometrically corrected. However, additional co-registration is performed in order to ensure that both datasets are superimposed with accuracy at sub-pixel scale.  Preprocessing also includes atmospheric corrections and noise reduction. For the latter, the 67 noisiest bands are removed from further processing. For additional noise reduction in the retained bands, two transforms are tested: inverse Principal Component Analysis (PCA) and inverse Minimum Noise Fraction (MNF) rotation.

The NDVI map is then calculated in order to delineate and differentiate vegetated and non-vegetated areas for both dates. An unsupervised classification map using the ISODATA method is produced using four datasets for each date. These datasets are (i) the atmospherically corrected image, (ii) the first seven PC components composite image, (iii) noise-reduced image after inverse PCA and (iv) the corresponding one after inverse MNF rotation. The assessment of the classification results is performed through comparison with geological maps and previous published work, including field spectral measurements in specific parts of the volcano. For this purpose, five different sites are selected for the interest of their particular spectral features in the general land cover classes: vegetation, urban areas, bare soil and lava flows.

The examination of these areas shows that the most realistic classification results are provided by the classified image issued from the inverse PCA rotation. The specific classification maps are retained for further analysis.

The comparison between the two classified images shows that the spectral and spatial differentiation of the older than 2009 lavas, located at the slopes and near the center of the volcano, is difficult. The most significant lava changes after 2009 are detected in the southern part of the volcano, where several lava flows that occurred during the 3-year time period.

Concerning the vegetated areas, including vegetation installed on older lavas, a significant difference in the spatial extent is observed between the two dates. This is probably due to both (i) the vegetation seasonal behavior combined with the different acquisition dates and (ii) the vegetation growth between the three-year time span in some areas and new lava depositions in others. Older lavas under vegetated areas cannot be distinguished. On the other hand, no spectral changes are observed in urban areas. However, slight differences in their spatial extent are observed between the two dates.

The first results of this study show that Hyperion can be useful for lava flow mapping and change detection in a complex environment such as the Etna volcano. The results are promising considering the relatively coarse pixel size, the number of noisy bands, the scarcity of lava field spectral measurements and the limited published work for this study period. Future investigations include field work, testing of clustering techniques such as SAM and the complementarity of Hyperion to other satellite data.

The a priori knowledge of the Earth surface reflectance is required for the operational exploitation of space-borne missions (both active and passive) measuring the electromagnetic radiation reflected by the surface in the solar domain, as well as for the preparation of future missions and processing chains. In this prospect, a monthly climatology, ADAM (A Surface Reflectance Database for ESA’s Earth Observation Missions), has been designed in order to simulate realistic spectro-directional variations of the Earth surface reflectance at the global scale with a 0.1°x0.1° resolution. A calculation toolkit associated to the ADAM database allows simulating the spectral BRDF over the 240-4000 nm spectral domain (with a spectral resolution of 1 nm) and in any observation geometry.

For land surfaces (soil/vegetation and snow), the ADAM database is made of a climatology of normalized reflectances (corrected from the directional effects) in seven wavebands, processed by the FondsDeSol processing chain from MODIS MOD09 products. For ocean surfaces, the database consists in climatologies of ocean colour (chlorophyll content derived from SeaWiFs) and wind speed (from QuickScat).

For the soil/vegetation pixels, the spectral interpolation/extrapolation of the MODIS normalized reflectances between 240 and 4000 nm is performed using EOF (Empirical Orthogonal Functions) derived from spectral reflectance databases of soil/vegetation/leaf optical properties. The Ross-Li-Maignan kernel based BRDF model is then used to calculate the reflectance spectrum in any observation geometry. A specific processing is applied for snow covered surfaces; It relies on the model of Kokhanovsky, fitted to the MODIS normalized reflectances, to simulate the spectro-directional variations of snow reflectance. Moreover, it is possible to calculate the uncertainty attached to land surface reflectance: the calculation relies on the variance covariance matrix of the reflectance values between the seven MODIS bands for each 0.1°x0.1° pixel. Over ocean, the surface reflectance is generated by the combination of  three components: i) the water column that mainly shapes the spectral variation of reflectance depending on the chlorophyll content, ii) the specular component  and iii) foam that essentially drive the directional variation of ocean reflectance depending on the wind speed.

Recent updates of the ADAM database and calculation tools have been performed. They consist essentially in an improved gap filling over ocean and lakes (with an accounting of sea ice) and an extension of the models in the UV (between 240-300 nm when the original version was limited to 300 nm). We will also present the latest evaluations of the ADAM database/toolkit to simulate realistically land surface BRDFs: 1) relying on POLDER/PARASOL top of canopy spectro-directional reflectance products over a selection of pixels spanning the main soil/vegetation types, and 2) using CALIPSO LIDAR measurements at 532 and 1064 nm to validate the backscatter (hot spot) increase factor simulated by the BRDF model.

Human life is affected by precipitation more directly in contrast to the other atmospheric phenomena. The most recent studies revealed the precipitation mean and extreme increased in the large part of the Netherlands. The significant changes for different extreme indices, show precipitation, especially extreme precipitation events did not distribute homogeneously over the study area. Moreover, the detected  trend of extreme precipitation is greater than the annual total amount over the country. The aim of the study was to detect the climatological differences of the maximum precipitation and the land cover signal influences on the extreme precipitation patterns. The precipitation probability and its distribution were assessed by block maxima approach through the comparison of observation over different land cover and timescales. The related signal of extreme precipitation change was investigated by using different proxies of land covers, which were defined based on the remote sensing data. The difference between the categories of stations has been extracted by the spatial gridding method for quantifying the possible effects of land cover. The comparison exposes the urban signal in the maximum daily precipitation. The consistent evidence shows that the extreme precipitation trend over the urban area has a greater change than that over the nonurban areas, during the winter half-year for the last 53 years.

Terrestrial Laser Scanner has the potential to capture the complex 3D structure, and in combination with 3D tree reconstruction models would allow us to model the shape of the trunk and main branches of trees. This is a step further into a more precise determination of whole-tree architecture and branching patterns; which would lead us into a better understanding of scaling exponents and metabolic rate at branch and whole-tree level in tropical forest trees without the need of destructive sampling. For this study, we extracted three trees from a TLS pointcloud data acquired during November 2013 in the Peruvian amazon rainforest. Quantitative structure model was used to calculate branch length, diameters and architecture from the individual trees. These parameters were used in the WBE plant-scaling model. This model calculated the following exponents: length ratio scaling, radii ratio scaling and estimated metabolic rate scaling and compared to the theoretical values. The theoretical exponent expected from WBE for branch length scaling is 0.3 and for branch radii scaling is 0.5. Across our samples, the calculated branch-level length scaling exponent varied from 0.04 to 0.11 and the calculated branch-level radii scaling exponent ranged from 0.30 to 0.32 (Table 1). The calculated (estimated) metabolic rate scaling exponent was 0.72, 0.71 and 0.76 for Tachigali polyphylla, Jacaranda copaia and Sclerolobium bracteosum (expected to be 0.75 from the WBE model). Estimations of tree scaling metabolism derived from architecture via TLS scans showed consistent and comparable values to the model predictions for all scaling exponents. Since the scanned trees were different species, these results provide evidence to support the WBE assumption of similarities in branching structure and common set of branching rules across trees. To conclude, tree scaling metabolism derived from TLS evidenced that (1) length ratio exponent, radii ratio exponent and architecture estimated metabolic rate converge between the tropical trees analysed, and (2) length ratio exponent, radii ratio exponent and estimated metabolic rate from the analysed samples are comparable with the predicted values.

Nearly 19.65 percent of the terrestrial surface worldwide is defined as degraded land as of the beginning of the 21st century. One major aspect of land degradation is the loss of top soils due to erosional processes, causing multiple environmental and socio-economic consequences. The key objective of this study is to assess the capability of using simulated EnMAP imagery to derive essential land cover information in a regional context, specifically photosynthetic active vegetation (PV), non-photosynthetic active (NPV, including standing dry vegetation and litter) and bare soil (BS, including stones and rocks) fractional cover, that can be integrated into soil erosion models as the cover-management C-factor.

This study investigates the capability of EnMAP imagery to map fractional cover in a region near San Jose, Costa Rica, characterized by spatially extensive coffee plantations and grazing in a mountainous terrain. Simulated EnMAP imagery is based on airborne hyperspectral HyMap data. Fractional cover estimates are derived in an automated fashion by extracting image endmembers to be used with a Multiple End-member Spectral Mixture Analysis approach. The C-factor is calculated based on the fractional cover estimates determined independently for EnMAP and HyMap. A detailed qualitative and quantitative comparison of the EnMAP and HyMap results derived using the same methodological approach give valuable insights into the capacities of the upcoming EnMAP satellite to provide useful information related to estimating soil erosion potential at a regional scale.

Results demonstrate that with EnMAP imagery it is possible to extract quality endmember classes with important spectral features related to PV, NPV and soil. The estimated relative cover fractions of PV, NPV and soil for both airborne HyMap and simulated satellite EnMAP imagery produce similar overall spatial patterns and similar quantitative values of fractional coverage. The key differences were found at a local scale, where the distribution of surface materials commonly occurs below that of the 30 m pixel size of EnMAP. However, spectral information is critical to separate BS and NPV which greatly can impact the C-factor derivation. The spectral capabilities of EnMAP were shown to be robust and emphasises that from a regional perspective we can use EnMAP to provide good fractional cover estimates that can be integrated into soil erosion modeling.

Overall the study demonstrates the capability of EnMAP to continue to provide hyperspectral satellite data compared to existing missions with similar sensor properties, and therefore, facilitate a continuation of exiting research focusing on long term monitoring and mapping of various environmental applications.

The FLuorescence EXplorer (FLEX) mission, candidate for ESA’s 8th Earth Explorer, aims to  globally measure the sun-induced chlorophyll fluorescence emission from terrestrial vegetation. In the frame of FLEX mission, several industrial and scientific projects have studied specific aspects of the instrument design, image processing algorithms or vegetation photosynthesis. In this respect, a common tool is needed to address the overall FLEX mission performance by combining each single factor. For this reason, an End-to-End Mission Performance Simulator (E2ES) has been developed for the FLEX mission (FLEX-E). This work will describe the FLEX-E software design, which combines the simulation of complex synthetic scenes with an advanced modeling of the instrument behavior and the full processing scheme up to the final fluorescence product. The results derived from FLEX-E simulations indicate that the instrument and developed image processing algorithms retrieve the sun-induced fluorescence with an accuracy below the 0.2 mW·m^-2

In the recent years a couple of operational urban mapping activities for international financial institutions have been implemented by Gisat in South-East Asia and Pacific region. Mapping services have been delivered in frame of EOWORLD2 and its precursor EOWORLD, both joint initiatives of ESA and World Bank, and PUMA project, follow-up initiative of World Bank. Service cases have aimed to contribute to understanding the extensive urban growth in various metropolitan areas including e.g. Karachi, Mumbai, Dhaka, Colombo or Surabaya by assessment of urban land use development.

The service is based on two pillars. First, Earth observation data and techniques are utilized for extraction of both retrospective and up-to-date information on urban land use and subsequently, assessment and comparison of extracted information and their links to the statistical figures are provided within dedicated web-based platform.

Land use status and changes are extracted by analysis of optical satellite imagery. Based on previous experience and throughout the course of the projects object-based image analysis techniques of detecting complex land use classes have been developed and tuned up using multi-resolution data. Semi-automated workflows preceding manual enhancements support consistent operational implementation of the service for large urban areas. Depending on recency of requested retrospective land use high or very high resolution data are used as baseline for information extraction.

The platform for Urban Management and Analysis (PUMA) represents web-based geospatial software for exploration and analysis of spatial data. PUMA adapts open-source software and allows users with no prior GIS experience to access, explore, visualize, analyse, integrate and share local, regional and global urban spatial data from a variety of sources in an interactive and customizable way. It supports the objectives of Global Urban Growth Data initiative: it helps the World Bank and its clients to develop a shared understanding of the long-term spatial, economic and environmental implications of land use by assessment of harmonized, comparable urban reference datasets. Trends of urban development and its patterns could be analysed both individually and between the cities mapped in South-East Asia.  

Services have been defined in a way to be operationally extendable in short or long term update time frame in order to serve as a base for further monitoring in future. The service components have also been proven in frame of urban risk domain in Copernicus Emergency Management Service or ESA’s initiative EO for a Transforming Asia Pacific.

Soil moisture matters. It is well known that soil moisture influences hydrological and agricultural processes, runoff generation, and drought development and impacts on the climate system through atmospheric feedbacks. It is also well known that soil moisture is a source of water for evapotranspiration over the continents, and is involved in both the water and the energy cycles. The importance of having access to a global soil moisture data set is reflected by the 1800+ users that have registered, in the last 3 years, to obtain the soil moisture product generated under ESA’s Climate Change Initiative. This product, with the latest product release – ECV SM 02.2 – being available from the end of 2015 on the ESA CCI soil moisture (SM) web site, is the most complete and most consistent global soil moisture data record, based on active and passive microwave sensors, to date with this global product providing 36 years of daily data from 1979 onwards.

ESA’s CCI ECV SM project is now in its first year of its second phase, led by the Earth Observation Data Centre for Water Resources Monitoring (EODC) and strongly supported by Earth Observation Science, Climate Research, and Systems Engineering groups. The project sees the evolution of the ECV Soil Moisture system from its prototype in 2014 towards a sustainable version via an operational framework for the production of this Soil Moisture Climate Data Record (CDR). The SM ECV production chain, classified as being at an early operational stage in 2014 (Bates Maturity Index 3), is being integrated into the virtual cloud computing environment of the EODC infrastructure.

Our key users and science team, drawn from the broad spectrum of scientific experts, interested in soil moisture information, have been working with, and validating the latest product release since early 2015, and provided positive feedback. This paper presents the results of the feedback from these and previous validation activities on the currently published dataset. In brief, alongside a series of positive validation studies, some studies have revealed that the soil moisture data set is very suitable for the detection of anomalies and trends, which makes this product highly suitable for climate studies. It was also found that the soil moisture dataset shows good performance with models and in situ data.  The novel gap filling technique devised for the product generation further improves performance. Issues related to extreme events and the integration of other datasets are targeted for the next product release. Significant progress has also been made on the integration of SMOS within the CCI data record. ESA recently supported an additional study to facilitate a seamless integration of SMOS data in the CCI soil moisture CDR climate data record, and different scenarios have been applied and show promising results. In the next product release, SMOS will become part of CCI soil moisture data record.

 Many of these research highlights, summarised in this paper, can be found in a special issue of the International Journal of Applied Earth Observation and GeoInformation that highlights the “Advances in the Validation and Application of Remotely Sensed Soil Moisture”.

This study aims at investigating the potential of Sentinel-1 (S1) and its high temporal resolution for retrieving surface soil moisture (SM) using a continuous change detection approach, in which each additional scene aids in further calibrating the retrieval model from which SM can be computed for every image acquisition time step.

Global retrieval of SM using spaceborne sensors has found wide applicability using active scatterometers (e.g. ASCAT aboard MetOp) and passive radiometers (e.g. SMOS) from which operational products are derived and regularly provided to the scientific community. Although having proven to accurately provide SM information from space, these two techniques are limited in both spatial resolution (on the scale of tens of kilometers) and land cover characteristics (radiometers were found to be more accurate over bare soil, while scatterometers yielded better results over vegetated areas). The latter disadvantage resulted in the launch of the Soil Moisture Active Passive Mission (SMAP), whose active Synthetic Aperture Radar (SAR) failed in 2015 after only a few months of operation.

SAR offers a much higher spatial resolution than scatterometers and radiometers, yet applicability to SM retrieval was in the past limited by data scarcity. Hence, although much research was conducted investigating the potential of SAR for SM retrieval, this research was mostly limited to exploiting single scenes and required a reliable ground truth reference for estimating soil roughness. An interesting alternative not reliant on ancillary roughness information is by investigating image time series assuming a linear correlation between SAR backscatter and relative SM between estimated dry and wet references of a given soil. This methodology, originally developed and operationally implemented using scatterometer data, has also proven to be applicable to SAR data using Envisat ASAR on 1 km spatial resolution. At the time of writing this methodology has not been applied to any other satellite’s data, most likely due to the lack of sufficiently large time series. For this reason this study aims at collecting all available S1 datasets for selected test sites and further complementing this set with additional data from Envisat ASAR and ERS in order to reliably calibrate the model and accurately compute dry and wet references. Once radiometric differences between the datasets are accounted for and the SM reference values are computed, a relative soil moisture index can be retrieved. An experimental study investigating the pre-launch potential of S1, has found this satellite to be well capable of delivering global SM information on a 1 km spatial resolution. It is yet to be investigated whether this resolution is realistic or can even be improved, which is a key research question of this study.

The work presented here is part of the EOTAP (Earth Observation for a Transforming Asia Pacific) initiative, a collaboration between the European Space Agency (ESA) and the Asian Development Bank (ADB) to develop a set of twelve small scale projects with the purpose to produce, deliver and assess the benefits of information services based on Earth Observation (EO), in support of ongoing ADB development projects in the Asian Region.

The project "Establishment of Climate-Resilient Rural Livelihoods in Mongolia " (JFPR 9164-MON), developed and implemented by the Ministry of Food and Agriculture of Mongolia, is one of the projects included in this initiative. The project focuses on the livestock sector, one of the main economic pillars in Mongolia, accounting for about 20% of Gross Domestic Product (GDP), in order to overcome the productivity and income loss problems, related to over-grazing and climate change, occurred in recent years.

The EO services developed within the EOTAP activity in support of this project primarily aim at enriching the existing environmental database maintained by the National Remote Sensing Center (NRSC) in Mongolia and sustaining the collaborative pasture management. The geographic area covered by the EOTAP services is Bayankhongor province, in western Mongolia region, with two main services: drought monitoring at the provincial level for the year 2014 and LULC and changes mapping for three districts of this province (Buutsagaan, Dzag and Khureemaral) for the years 2013, 2014 and 2015.

The drought monitoring service is based on the determination of 2014 MODIS NDVI monthly anomalies with a spatial resolution of 250m (using the MOD13Q1 16-day product), by comparing the calculated NDVIs for that year with the average NDVI for the 2001-2013 time series. The study of the instantaneous and accumulated anomalies allows for the identification of areas possibly affected by drought (with significant negative anomalies), completed by a correlation analysis of those values with precipitation and temperature EO derived products, to determine the main driving factors. Furthermore, high spatial resolution satellite images (22m) from the Deimos-1 and DMCii sensors, acquired during the pasture vegetation growth cycle (from April to October, twice per month), are used to complement this analysis and look with further spatial detail into the vegetation growth and productivity patterns.

In parallel, LULC and respective LULC change maps are produced for the years 2013, 2014 and 2015 using the different Landsat 7 bands and derived vegetation indices as classification parameters. The defined LULC classes are: a) Water bodies; b) Bare Soil Rock; c) Bare soil Sand; c) Pasture Classes: Desert Steppe, Pasture steppe; High, Medium and Low Mountain Steppe; Mountain Steppe; River Valley and Floodplains Meadows. Field campaign data, provided by the local users are used to train and validate the classification, reaching overall accuracies between 70% and 80%, depending on the year. It should be stated that the accuracy in attempting to discriminate between the different pasture classes is significantly lower (between 60 and 70%). In the opinion of the authors this is due to a lack of a sufficient training/validation points and the need to introduce additional classification parameters, able to provide information regarding possible differences in seasonal phenological cycle between pasture types (e.g., NDVI time-series). Nevertheless, with this classification method, it is possible to establish a LULC reference map for those years and help understand the possible dynamics of LULC change.

Information from the two EO-based services is a first basis on which to build tools to more efficiently monitor the effects of the sustainable pasture management practices introduced in the three above mentioned districts of the Bayankhongor province in the scope of the project “Establishment of Climate-Resilient Rural Livelihoods in Mongolia”, funded by the Japan Fund for Poverty Reduction (JFPR), which started in 2012 and covers a 4 year period.

Mapping transportation road conditions is an important issue for city and state authorities worldwide. Today, pavement safety is assessed by specific assemblies based on a mechanical wheel device, a method that is limited in its potential product, operation and cost. In this study, we examined the possibility of harnessing remote reflectance spectroscopy and imaging spectroscopy to predict asphalt’s dynamic friction coefficient, thereby enabling the identification and mapping of road conditions. We used near-infrared analysis techniques to evaluate several prediction models designed to assess the friction coefficient of asphaltic roads solely from spectral readings. We used an Artificial Neural Network (ANN) modeling technique which resulted with an accuracy of R=0.845 and high significance of P < 0.0001 between actual and predicted friction values for the best model. We have developed a method for applying the aquired models on both point and image hyperspectral data. This method was used on the airborne Aisa-FENIX and Hyvista-HYMAP hyperspectral sensors and on data obtained using an ASD FieldSpec Pro field spectrometer. The advantages of the suggested technology over the traditional systems are apperant, enabling rapid and large coverage of urban or intra-urban areas, providing high spatial and temporal resolution and providing the information in a minimal cost. 

The Xingu River basin in the State of Para, Brazil is home to a high diversity of freshwater organisms.  There are more than 420 fish species alone, many of which are endemic (approx. 30 have yet to be scientifically described) to the 1640 km long river that flows from the Brazilian Shield into the Amazon lowlands. Complex morphological characteristics such as shallow braided rapids, riffles and Riparian zones with a range of substrate types from fine sand to pebbles and boulders creates unique habitat types for several very specialized fishes. The middle portion of the river between Vittoria do Xingu and the confluence of the Iriri river will be permanently altered due to the construction of the Belo Monte dam, the third largest hydroelectric dam in the world and concurrent gold mining of the Volta Grande section proposed by Belo Sun.  It is anticipated that due to the loss of their habitat, several highly specialized species such as the Hypancistrus zebra might become extinct in the wild. Therefore, a very detailed baseline map for fish habitats is needed.

High spatial resolution satellite imagery, such as Pleiades, SPOT, Worldview-2, Worldview -3, in addition to UAV based photography/videography are well suited for mapping small scale features such as individual tree crowns, agricultural fields and detecting fine-scale features (e.g. rock outcrops). Small unmanned aerial vehicles (UAVs), are important in remote sensing studies requiring high spatial resolution (<1 m) and precise temporal imagery. These deployable systems carry small payloads with interchangeable camera systems (e.g. optical, thermal) and as such have several advantages over conventional imaging from high altitude fixed-wing aircraft, rotorcraft or satellite platforms. These systems are often more cost effective than other forms of remotely sensed data (at high resolutions), can be deployed with high temporal frequency and are safe to operate as their navigational software allows the user to enter flight characteristics such as altitude, geographical waypoints and speed for near automated flights. Moreover, concurrently paired GPS enabled photographic cameras allow for rapid characterization of land and water features as the information embedded in the metadata of the photographs contain location, altitude and orientation information in addition to the camera raw file information (e.g. lense characteristics). For two weeks in September 2015, we acquired videography at 4000 pixel resolution using a UAV and GPS-photo data in order to develop a baseline fish habitat map for specific areas in the Xingu river. We compared our results with a classification of the same small areas using Pleiades. Here we present the results of the 2-D and 3-D reconstructions of the habitats and discuss the main advantages and challenges of the proposed methodology. Overall, our approach provides a framework for a rapid assessment of riverscapes applicable also to landscapes under human pressure.

Soil moisture and precipitation are essential components of the water cycle that intervene simultaneously in the infiltration, runoff and evaporation processes. Whereas rainfall provides the amount of available water at the surface, soil moisture partly controls the partitioning of rainfall into runoff and infiltration. The knowledge of both complementary variables is hence critical for achieving efficient and sustainable water management. Soil moisture is highly variable both in space and time, mainly as a result of the alternation between intermittent rainfall/irrigation events and dry down periods, and of the heterogeneity in land cover, topography and soil properties.

Satellite remote sensing is often the most practical and effective method to observe the land surface over large geographical areas. Currently, several Surface Soil Moisture (SSM) data sets are available at global scale but with a spatial resolution that is generally too coarse for agricultural applications. The Surface Soil Moisture retrieved from microwave radiometers such as C-band AMSR-E (Advanced Microwave Scanning Radiometer-EOS) and L-band SMOS/SMAP (Soil Moisture and Ocean Salinity / Soil Moisture Active and Passive) has a spatial resolution of about 60 km and 40 km, respectively. In this context, downscaling methodologies have been developed to improve the spatial resolution of readily available passive microwave-derived Surface Soil Moisture data. DISPATCH (DISaggregation based on Physical And Theoretical scale Change) estimates the Surface Soil Moisture variability within a 40 km resolution SMOS pixel at the target 1 km resolution using MODIS data. However, the temporal resolution of DISPATCH data based on SMOS and MODIS data is limited 1) by the data gaps in MODIS images due to cloud cover and 2) to a lesser extent by the temporal resolution of SMOS (global coverage every 3 days).

In this context, this paper examines a methodology for interpolating DISPATCH data using auxiliary high-frequency precipitation data and a simple model of the soil water budget at the near surface. This approach is based on the assimilation of both 40 km resolution SMOS and the first-guess 1 km resolution (DISPATCH) Surface Soil Moisture into the dynamic soil model forced by meteorological data including precipitation and potential evaporation. The validation is performed using ground measurements of soil moisture and precipitation measurements set up by the Joint International Laboratory TREMA in the Tensift Al Houz region around Marrakech.

Previous studies have shown the effectiveness of SAR (Synthetic Aperture Radar) remote sensing data in identifying and discriminating different crop types. Thanks to the capability of SAR systems to penetrate clouds, to work under all-weather conditions and to be independent from the solar illumination, numerous researches focused on the investigation of their use for crop mapping [1].

Several studies have demonstrated the importance of time series of SAR data to map different crop types at, C-band [2]; L-band [3]; and X-band [4], [5], [6].

One of the most recent SAR systems, launched in 2007, is the Italian constellation COSMO-SkyMed^® (CSK^®) working in X-band (frequency 9.6 GHz, wavelength 3.1 cm). COSMO-SkyMed^® is an attractive information source for classifying different crop types, due to its dense revisit time (which can be lower than 12 hours) [7] and its capability to operate in different imaging modes.  

The main goal of this work is to present a study that analyzes i) the effectiveness of single polarization multi-temporal X-band COSMO-SkyMed^® for identifying and discriminating different crop types over an agricultural area, and ii) the importance of the use of different polarizations (VV. HH, VH, HV) as additional source of information. The study has been developed on a test site located in the Austrian region of Marchfeld, near the city of Vienna. This is an interesting testing area due to its crop types heterogeneity. The experiment was focused on the main summer crops grown in this region, namely carrot, corn, potato, soybean and sugar beet.

The CSK^® dataset includes 29 imagery acquired during the whole crop growing period, from April 1^st to October 15^th 2014, at different polarizations (VV. HH, VH, HV) and acquisition modes(Stripmap HIMAGE and Stripmap PING PONG).

Ground-truth information were gathered over the surveyed fields and many ground measurements were obtained during the field campaign: crop status, crop type, crop phenological stage, leaf area index (LAI), leaf dimension, plant height, field locations.

In order to identify the different agricultural crops, a pixel-based classification system based on the support vector machine (SVM) was fed with time series of CSK backscattering coefficient values.

The classification accuracy provided by the SVM system was assessed in terms of overall accuracy, user’s accuracy, producer’s accuracy and Kappa statistics of the error matrix.

Classification results obtained by using multi-temporal COSMO-SkyMed^® data at different polarizations (VV, HH, VH, HV) and at diverse acquisition mode will be presented. Moreover, in order to evaluate the possibility to monitor crops at the start of crop growth stage, the results of the CSK^® classification carried out on the whole season dataset will be compared to the CSK^® classification conducted on few scenes acquired at the beginning of the season.

[1] Shewalkar, P., Khobragade, A. and Jajulwar, P.K., “Review Paper on Crop Area Estimation Using SAR Remote Sensing Data” IOSR Journal of Electrical and Electronics Engineering (IOSR-JEEE) e-ISSN: 2278-1676, p-ISSN: 2320-3331, 9(2), 97-98 (2014).

[2] Jiao, X., Kovacs, J.M., Shang, J., McNairn, H., Walters, D., Ma, B. and Geng, X., “Object-oriented crop mapping and monitoring using multi-temporal polarimetric RADARSAT-2 data” International Society for Photogrammetry and Remote Sensing (ISPRS). Papers 96(10), 38-46 (2014).

[3] McNairn, H., Shang, J., Jiao, X. and Champagne, C., “The Contribution of ALOS PALSAR Multipolarization and Polarimetric Data to Crop Classification” IEEE Transactions on Geoscience and Remote Sensing, 47, 3981-3992 (2009).

[4] McNairn, H., Krossa, A., Lapena, D., Cavesb, R. and Shanga, J., “Early season monitoring of corn and soybeans with TerraSAR-X and RADARSAT-2,” International Journal of Applied Earth Observation and Geoinformation 28 252–259 (2014).

[5] Villa, P., Stroppiana D., Fontanelli, G., Azar R. and  Brivio P.A., "In-Season Mapping of Crop Type with Optical and X-Band SAR Data: A Classification Tree Approach Using Synoptic Seasonal Features” Remote Sensing, 7(10), 12859-12886; doi:10.3390/rs71012859 (2015).

[6] Sonobe, R., Tanib, H., Wangb, X., Kobayashic, N. and Shimamurad, H., “Random forest classification of crop type using multi-temporal TerraSAR-X dual-polarimetric data”, Remote Sensing Letters 5(2), 157-164 (2014).

[7] Italian Space Agency (ASI) - Cosmo SkyMed Mission. COSMOSkyMed SAR System Description & User Guide [online]. Available: http://www.asi.it/it/flash/osservare/cosmoskymed.

The European Space Agency’s Earth Observation for Oil & Gas (EO4OG) initiative aimed at providing a base for the future development of earth observation (EO) guidelines for the on-shore and offshore oil and gas (O&G) sector. Comprising the offshore element of EO4OG, it was the objective of this study was to analyze the capabilities of EO technologies for offshore oil and gas operations, highlight capability gaps and identify opportunities to close these gaps and foster a more widespread use of EO within the O&G industry.

 Following the comprehensive identification of geo-information requirements of offshore oil and gas operations, a framework for relevant EO-based product categories and service scenarios was established. This formed the basis for a detailed analysis of EO capabilities available to address different user needs and the comparison with actual use of EO within the O&G industry. Capability and utilization gaps were further characterized by carrying out an analysis of strengths, weaknesses, opportunities and threats (SWOT) at the level of product categories. This was followed by identifying R&D priorities and recommended actions to close gaps in EO capability and utilization. 

Approximately 57% of the EO-based products identified in this study are considered important by O&G stakeholders. These products are being used within the industry in accordance with their respective levels of technical maturity. Significant capability gaps of EO-derived information remain in the areas of wave and surface current retrieval, the assessment of local weather phenomena, the distribution and abundance of seabirds and marine mammals and the interaction between gas flares and seabirds. Several factors have been identified to play a role in the under-utilization of mature EO capacities in the O&G industry, including varied levels of EO expertise in user organizations, ineffective communication of EO capabilities to key decision-makers competition with non-EO approaches and limitation in current EO capabilities.

It is recommended to continue the dialogue between O&G and EO communities, build awareness of EO capacity within O&G industry and work towards the industry-wide adoption of best practices regarding the use of EO technologies. The policy free and open access to comprehensive, global EO data coverage provided by the Sentinel missions constitutes a significant opportunity for the oil and gas sector to use EO to the fullest of its capabilities. A preliminary roadmap for the further development of EO within the offshore oil and gas sector was formulated comprising the establishment of an industry-led task force, the execution of pilot studies, the engagement of O&G stakeholders through outreach and awareness building, and the continued alignment with existing and emerging practices within the O&G sector.

The SMOS mission was launched in November 2009 and allows measuring the surface soil moisture over continental land, covering the entire surface in 3 days. The multi-angular algorithm also enables to estimate the vegetation opacity which is directly related to the water content of the canopy.
However, the SMOS radiometer operates within the Earth Exploration Satellite Service passive band at 1400-1427 MHz . Since its launch in November 2009, SMOS images have been strongly impacted by Radio Frequency Interference (RFI). So far approximately 500 RFI sources distributed worldwide have been detected. Some of the strongest RFI sources might mask other weaker RFI underneath, hence it is expected the total number of RFI detected may increase as strong ones are progressively located and switched off. Most RFIs are located in Asia and Europe, which together hold approximately 80% of the active sources and more than 90% of the strongest interference. The areas affected by RFI may experience either an underestimation in the retrieval values of soil moisture and ocean salinity or data loss, with the associated detrimental impact in the scientific return. ESA and the teams participating in SMOS mission have put in place different strategies to alleviate this RF interference situation with significant successes in Europe.
We will present some detection and localisation approaches together with specific examples of detection in complex environments.
The Aquarius and SMAP missions results in terms of RFI detection will be compared to that of SMOS

The Touareg shield and its Algerian part the Hoggar have been structured by the amalgamation of twentieth terrane during panafrican orogeny (850-550 Ma, Caby, 2003).

The 4°50’ accident which separates the Central Hoggar from the Occidental Hoggar constitutes a major suture zone in this mobile belt (Caby, 2003). It separate the micro continent LATEA, constituted by four terrane, archeo-paleoproterozoic, dominated by amphibolitic to granulitic basement assumed Paleoproterozoic; the panafrican terrane of Iskel which corresponds to a whole island arc showing a low degree metamorphism (green schist to amphibolite).
This accident besides being considered as a lithospheric suture zone has played an important role in post collisional transpressif movement that led the dislocation of LATEA in many terranes (Liégeois et al., 2003).
The aim of this work is to image the 4°50’ accident and the other mega shear zones close to it by using a multi-source study. Indeed, a geological map was demonstrated by the utilization of many treatments of Landsat 8 multispectral images (colour composite, ratios bands, directional filter, mnf…). The result includes a lineament mapping accompanied a density map of these lineaments and rosette directions.
Excepting this surface cartography, an E-W magnetotelluric section covering the entire crustal thickness has been highlighted, the high geophysical resolution data are of a paramount importance to identify the deep structure as well as the dynamic of the Hoggar lithosphere. The magnetotelluric method is the ideal tool for imaging the deep structures on either side of the shear zone 4°50’. Many recognition profiles have been realized in the Hoggar with relatively low resolution (spacing between stations is of the order of 40 to 50 Km) showing that there’s no conductive regional anomaly (Bouzid et al., 2014). The profile overlapping the 4°50’ shear zone realized by the magnetotelluric team of CRAAG in 2013 has been densified (2 Km between every station) at Ablessa in order to enhancing the lateral resolution of data and to get more details.
After the inversion of the apparent resistivity data, a 2D model has been obtained (investigation depth of 15 Km), showing the conductive structures especially the 4°50’ and the 4°30’ shear zones. The rest of the  interpretation will be discussing during this symposium.

Over recent decades, a range of satellite-derived vegetation products have emerged, facilitating the systematic monitoring of vegetation at regional to global scales.  To be of real use in environmental decision making, it is vital to ensure these products are of high quality and consistency.  The evaluation of satellite-derived vegetation products remains a challenging task as a result of their moderate spatial resolution.  In areas of homogeneity, estimates of gross primary productivity are often obtained from eddy covariance flux tower data, providing a measure of photosynthetic activity against which vegetation products can be compared.  Uncertainties are associated with such an approach however, particularly in areas of heterogeneity, where logistically-challenging field campaigns are instead required. 

In recent years, near-surface remote sensing has emerged as a promising alternative.  Although not calibrated instruments, digital cameras enable the greenness of vegetation to be characterised at a high temporal resolution and with a greater degree of spatial integration than traditional in-situ techniques.  In this paper, a range of vegetation products derived from MERIS data are evaluated using near-surface remote sensing data obtained from the PHENOCAM network.  17 study sites are examined in the United States, covering a range of vegetation types.  Moderate to strong relationships between the vegetation products and measures of greenness derived using near-surface remote sensing are demonstrated, highlighting the potential of the approach.  Differences in performance are discussed, both between vegetation products and study sites.  Issues associated with near-surface remote sensing are identified, as are potential solutions, including the use of narrow band-pass filters to provide increased sensitivity to the biophysical variables of interest.

Time series of remote sensing data are an important and effective way to assess land surface phenology trends across spatial and temporal scales. Estimates of land surface phenology from satellite data are sensitive to the pixel resolution, temporal resolution, phenology extraction method and atmospheric contamination. Relatively few studies have provided a complete evaluation of the effect of these sensitivities on estimated phenological parameters, such as start of season and end of season.

The purpose of this study was to evaluate the sensitivity of satellite-derived phenological variables to choice of temporal (compositing period) and spatial resolution. MERIS MTCI data (Level 2) for 2005 to 2011 were acquired for the United Kingdom. These data sets were compiled at 4, 8, 10 and 16 day composites using the BEAM open-source toolbox and development platform at 250 m resolution, and then resampled to 500 m, 1 km, 2 km, 4 km, and 8 km. Discrete Fourier transformation was applied to create smoothed time-series of MTCI across the UK and an inflection based method was used to estimate the start of season (SOS) and end of season (EOS) for each temporal and spatial resolution. The results highlighted the variation of phenological parameters in comparison to the six dominant land cover type with variation in both spatial and temporal resolution. 

The variation in composite period highlighted that the 16 day composite period predicted a later SOS, and the 4 day composite period predicted a later EOS estimate. In addition coarse spatial resolution predicted a later SOS compared with finer spatial resolutions. The satellite-derived estimates were then compared to in-situ observations from Natures Calendar UK, and near-surface digital camera photography from CEH, to evaluate the mangnitude of sensitivity to spatial and temporal resolution. 

The remnants of native vegetation still available in Azores (Portugal) are composed by unique communities (e.g. Juniperus brevifoliaand broad-leaf Laurel Forests) that support most of the endemic fauna and flora of the archipelago. Flores Island (Azores, Portugal), a UNESCO Biosphere Island, has one third of its territory also classified as Natura 2000 site. The presence of some very aggressive invasive alien species (IAS) such as Pittosporum undulatum or Acacia melanoxylon (both native from Australia), and Hydrangea macrophylla or Hedychium gardneranum (both native from Asia) can cause serious damage on these ecosystems. Their direct competition with native species results in the decline of native species and native ecosystems degradation. The aim of this paper is to assess the effect of using different land use / land cover (LULC) maps in the calculation of landscape metrics and indices that present correlation to habitat quality and biodiversity in Flores Island. Several landscape metrics and indices as well as different LULC mapping legends were tested. Special attention was given to the effects that the different sources of information (thematic maps and high resolution multispectral Rapideye satellite imagery), as well as the spatial resolutions, might have on the use of the LULC maps as monitoring tools for these areas. Regarding the land use assessment, satellite imagery showed to be the best option since it allowed a more accurate representation of the landscape patterns.  Simplifications of the thematic maps generated by different classification methods, like the maximum likelihood and the standard vector machine, are needed to reduce artifacts generated in the classification, like the salt and pepper effect that can have a great impact in metrics and indices.  Landscape metrics derived from LULC maps generated from satellite imagery processing can be assumed as an effective landscape and LULC monitoring tool in order to obtain accurate and updated information on the land use typology, characteristics and structure. The association of the data generated by the landscape metrics and indices extracted from the different LULC thematic maps with data related to habitat quality and biodiversity can be a method to define the ideal classification function to be applied to the satellite imagery. This method can also help to define the level of simplification to be applied to the resultant LULC map in order to establish the highest possible correlation and to ensure that the selected LULC map has the specifications needed to constitute the base of a monitoring system. This method could also make possible to define the ideal LULC map for the region were the monitoring system is to be applied respecting its singular characteristics and empowering the Copernicus Programme implementation for ecosystem monitoring and biodiversity conservation.

ENVIRONMENTAL RISK AQUATIC ECOSYSTEMS CONTAMINATION

Olga S. Reshetnyak

 

Environmental issues and risks are becoming more relevant in recent years. Compliance with environmental safety of natural ecosystems at various levels (local, regional and global) is one of the priorities of the environmental policies of many countries. A lot of attention is paid to the identification and prediction of potential environmental threats, environmental risk assessment in the event of environmental emergencies. Environmental risk - the probability of habitat degradation or its transition into an unstable state (the environmental crisis and environmental disasters) as a result of human activities and the possibility of loss of control occurring environmental events. Environmental risk is a measure of the environmental hazard [1, 2].

The risk assessment of aquatic ecosystems contamination is the process of identifying possible negative consequences as a result of structural organization violations of the ecosystem and the presentation of these violations in quantitative terms [3].

Environmental risk assessment of water pollution is a definition the likely changes in ecosystems under anthropogenic impact. It is necessary to determine the reliability features that characterize a possible variability in the aquatic communities development due to deterioration of water quality, beyond which the system loses its resilience. Changes in the structural organization of hydrobiocenoses are the ecosystem response and the risk of anthropogenic impact by increasing internal ecosystem processes such as the anthropogenic eutrophication or ecological regression [4, 5].

To assess the Environmental risk on aquatic ecosystems it is necessary to assess the anthropogenic impact effect on the biota (i.e, which process dominates in the ecosystem) and the level of prevailing process. The results of the environmental risk assessment for aquatic ecosystems estuarine areas of the major rivers of Russia in terms of ecological regression are presented in the table.

It is found that the aquatic community of studied ecosystem functioning in a state of anthropogenic stress with elements of anthropogenic ecological regression (the Lena, the Northern Dvina and the Volga Rivers), or in a state with elements of ecological regression (the Don, the Selenga and the Amur Rivers).

References

1. Schmal AG Factors environmental hazard & environmental risks. Publisher: MP "ICC BNTV", 2010, Bronnitsy, 191p.

2. EPA / 630 / R-00/002. Supplementary Guidance for Conducting Health Risk Assessment of Chemical Mixtures / Risk Assessment Forum / US Environmental Pro-tection Agency. - Washington, DC, 20460, 2000. - 209 p.

3. P 52.24.776-2012. Recommendation. Assessment of anthropogenic load and the risk to river mouth area with regard to their regional characteristics. Roshydromet, FGBI "GHI", Rostov-on-Don, 2012. - 32 p.

4. P 52.24.661-2004. Recommendation. Risk assessment of anthropogenic impact to priority contaminants in the surface water. M .: Meteoagenstva Roshydromet, 2006 – 26 p.

5. Reshetnyak O.S. Environmental Risk Assessment impact on estuarine ecosystems of large rivers / The 15th International Scientific and Industrial Forum "Great Rivers 2013". Proceedings of the Congress of 2 volumes. Vol.1 / Nizhegorod. state. arhit.-building. Univ; Ans. Ed. S.V.Sobol. - Nizhniy Novgorod: NNGASU, 2013. pp. 148-150.

 

Table – Spatial variability of the environmental risk level of contamination to estuarine ecosystems

In arid regions, such as Egypt, irrigation is the main source of water consumption and freshwater resources are getting scarcer. Therefore, the development of adequate irrigation water practices becomes a necessity. Soil moisture, in particular, plays a key role in any efficient water use strategy for agriculture. Active microwave sensors such as the synthetic aperture radar (SAR) have been widely used to provide soil moisture estimates at large scale. Sentinel-1 carries a single C-band synthetic aperture radar instrument operating at a centre frequency of 5.405 GHz, providing data at adequate spatial resolution suitable for the larger holdings of the new cultivated areas in Egypt. Also, it will provide a repeat cycle of 6 days, which is appropriate for monitoring soil moisture content especially when used in adjusting the irrigation scheduling. Vegetation is regarded as an important factor to consider when using radar data to estimate soil moisture content in vegetated areas. Accordingly, optical data has been intensively employed with radar data to retrieve soil moisture content over vegetated areas to improve the quality of the soil moisture estimations. The main objective of this study is to develop a protocol for processing microwave data supported by optical data to provide reliable and accurate estimation of soil moisture content. The study area is located about 84 km to the northwest of Cairo and covers an area of about 25 Km2. The study area is considered a newly reclaimed area, covered mostly by orchards mainly peaches, mango, pear, pomegranates and citrus in addition to small area cultivated with field crops or occupied by nurseries or bare soils. The soil surface is configured by low roughness due to soil texture and management practices and irrigated using drip irrigation systems.The fieldwork was conducted from 1/4/2015 to 1/10/2015 and sixty locations were selected for observation. The fieldwork was organized to repeatedly collect soil water content measurements within 2-3 hours of Sentinel-1 overpass based on its schedule. Other data were collected including soil texture and field capacity and vegetation information such as type information on crop development stage, plant height, and irrigation practices through an in-situ survey. The Sentinel-1 data was downloaded free of charges via ESA's Scientific Data Hub andlevel-1 data acquired at Interferometric Wide swath mode. The data were collected as Single Look Complex (SLC) and processed in ESA's Sentinel-1 Toolbox. Landsat 8 data surface reflectance data were collected in the dates as close as possible to the Sentinel-1and both the Normalized Difference Vegetation Index (NDVI) and Land Surface Water Index (LSWI) were calculated. These data were staked together into one image, resized to the Sentinel-1 data, for each date. Additionally, the elevation was also added to these images. The neural network was designed combining radar, optical and elevation data to develop a soil moister prediction model. Additionally, field data was also included in the neural network. The neural network analysis was performed utilizing cross-validation, where 70% of the data was used for training, 15% for validation and 15% for testing. The preliminary results indicated that the correlation between the predicted and actual soil moisture content was more than 0.9.

Aboveground biomass (AGB) is an important variable in carbon accounting and climate science. The estimation of AGB in the boreal forests is of special concern as it constitutes the largest biome in the world and has substantial carbon accumulation capability. Russia, as a country with the largest forest area in the world (809 million ha (FAO 2010)), provided more than 90% of the carbon sink of the world’s boreal forests between the years 2000 and 2007 (FAO 2012). Despite this importance, Russia’s boreal forest has the highest (Pan et al. 2011) uncertainty (Nilsson et al. 2007) in carbon stock calculations. This is mostly due to poor measurements of biomass stocks, forest degradation, deforestation, and forest growth. Additionally, due to the lack of financial support, some forested regions in Siberia have not been inventoried for more than 20 years (Shvidenko et al. 2011). Therefore, there is a strong need for earth observation (EO) methods of biomass monitoring, which will reduce costs and improve the estimations.

The objective of this study was to use the multi-frequency Synthetic Aperture Radar (SAR) L-band and C-band data for improved AGB estimation in Siberian forests. The L-band ALOS-2 PALSAR-2 and C-band RADARSAT-2 data in single (HH) and dual (HH and HV) polarizations in Single Look Complex format (SLC) were used in this study. The data have been obtained within the ALOS Kyoto and Carbon Initiative Project (K&C) and the RADARSAT-2 Science and Operational Applications Research and development program (SOAR2). The backscattering coefficient was calculated taking into account local incidence angles. The SAR data were used as predictor variables for AGB retrieval. As the response data the forest inventory data were used. The datasets were provided as a Geographical Information System (GIS) database by the Russian State Forest Inventory. The database contains information of growing stock volume (GSV) at stand level. In order to convert GSV into AGB a non-linear relation was developed employing the regional data from the International Institute for Applied Systems Analysis (IIASA) Russian live biomass plot database (IIASA 2007).

The SAR and forest inventory data were used as inputs for a non-parametric data fusion machine learning algorithm – Random Forests (Breiman 2001). The Random Forests are widely used for classification in ecology (Prasad, Iverson, and Liaw 2006; Hüttich et al. 2011; Cutler et al. 2007) as well as for AGB estimation (Baccini et al. 2008; Simard et al. 2011; Houghton et al. 2007; Walker et al. 2007; Avitabile et al. 2012; Cartus et al. 2012; Cartus et al. 2014; Wilhelm et al. 2014). Recent study showed (Fassnacht et al. 2014) that the Random Forests is superior to other methods such as support vector machine (SVM), k-nearest neighbour (KNN), Gaussian processes (GP), step-wise linear models. The AGB was estimated for an area located in Krasnoyarskiy Kray in the Southern part of Central Siberia, Russia, approximately 120 km North-East of the city Krasnoyarsk – part of the Bolshe Murtinsky forest enterprise.

The AGB retrieval was done at 0.5 ha. The estimation error was calculated of approximately 25% by validation against an independent dataset. In the previous studies using only L-band or C-band data the reported errors were in the range of 30-40% for Siberian boreal forests (Santoro et al. 2006; Wilhelm et al. 2014; Rodriguez-Veiga et al. 2014; Chowdhury, Thiel, and Schmullius 2014).

Avitabile, Valerio, Alessandro Baccini, Mark A. Friedl, and Christiane Schmullius. 2012. “Capabilities and Limitations of Landsat and Land Cover Data for Aboveground Woody Biomass Estimation of Uganda.” Remote Sensing of Environment 117 (February). Elsevier Inc.: 366–80. doi:10.1016/j.rse.2011.10.012.

Baccini, a, N Laporte, S J Goetz, M Sun, and H Dong. 2008. “A First Map of Tropical Africa’s above-Ground Biomass Derived from Satellite Imagery.” Environmental Research Letters 3 (October). doi:10.1088/1748-9326/3/4/045011.

Breiman, Leo. 2001. “Random Forests.” Machine Learning 45 (1): 5–32. doi:10.1023/A:1010933404324.

Cartus, Oliver, Josef Kellndorfer, Markus Rombach, and Wayne Walker. 2012. “Mapping Canopy Height and Growing Stock Volume Using Airborne Lidar, ALOS PALSAR and Landsat ETM+.” Remote Sensing 4 (11): 3320–45. doi:10.3390/rs4113320.

Cartus, Oliver, Josef Kellndorfer, Wayne Walker, Carol Franco, Jesse Bishop, Lucio Santos, and José Fuentes. 2014. “A National, Detailed Map of Forest Aboveground Carbon Stocks in Mexico.” Remote Sensing 6 (6): 5559–88. doi:10.3390/rs6065559.

Chowdhury, Tanvir Ahmed, Christian Thiel, and Christiane Schmullius. 2014. “Growing Stock Volume Estimation from L-Band ALOS PALSAR Polarimetric Coherence in Siberian Forest.” Remote Sensing of Environment, June. Elsevier Inc. doi:10.1016/j.rse.2014.05.007.

Cutler, D. Richard, Thomas C. Edwards, Karen H. Beard, Adele Cutler, Kyle T. Hess, Jacob Gibson, and Joshua J. Lawler. 2007. “Random Forests for Classification in Ecology.” Ecology 88 (11): 2783–92. doi:10.1890/07-0539.1.

FAO. 2010. Global Forest Resources. Assesment 2010. Main Report. Rome. http://www.fao.org/docrep/013/i1757e/i1757e00.htm.

FAO. 2012. The Russian Federation Forest Sector Outlook Study to 2030. Rome.

Fassnacht, F.E., F. Hartig, H. Latifi, C. Berger, J. Hernández, P. Corvalán, and B. Koch. 2014. “Importance of Sample Size, Data Type and Prediction Method for Remote Sensing-Based Estimations of Aboveground Forest Biomass.” Remote Sensing of Environment 154 (November): 102–14. doi:10.1016/j.rse.2014.07.028.

Houghton, R. A., D. Butman, A. G. Bunn, O. N. Krankina, P. Schlesinger, and T. A. Stone. 2007. “Mapping Russian Forest Biomass with Data from Satellites and Forest Inventories.” Environmental Research Letters 2 (045032). doi:10.1088/1748-9326/2/4/045032.

Hüttich, Christian, Martin Herold, Ben J. Strohbach, and Stefan Dech. 2011. “Integrating in-Situ, Landsat, and MODIS Data for Mapping in Southern African Savannas: Experiences of LCCS-Based Land-Cover Mapping in the Kalahari in Namibia.” Environmental Monitoring and Assessment 176 (1-4): 531–47. doi:10.1007/s10661-010-1602-5.

IIASA. 2007. “Russian Forests & Forestry. Live Biomass & Net Primary Production – Measurements of Forest Phytomass in Situ.” http://webarchive.iiasa.ac.at/Research/FOR/forest_cdrom/english/for_prod_en.html.

Nilsson, S., A. Shvidenko, M. Jonas, I. McCallum, A. Thomson, and H. Balzter. 2007. “Uncertainties of a Regional Terrestrial Biota Full Carbon Account: A Systems Analysis.” Water, Air and Soil Pollution: Focus 7: 425–41. doi:10.1007/s11267-006-9119-1.

Pan, Yude, Richard A. Birdsey, Jingyun Fang, Richard Houghton, Pekka E. Kauppi, Werner A. Kurz, Oliver L. Phillips, et al. 2011. “A Large and Persistent Carbon Sink in the World’s Forests.” Science (New York, N.Y.) 333 (6045): 988–93. doi:10.1126/science.1201609.

Prasad, Anantha M., Louis R. Iverson, and Andy Liaw. 2006. “Newer Classification and Regression Tree Techniques: Bagging and Random Forests for Ecological Prediction.” Ecosystems 9 (2): 181–99. doi:10.1007/s10021-005-0054-1.

Rodriguez-Veiga, Pedro, Martyna Stelmaszczuk-Górska, Christian Hüttich, Christiane Schmullius, Kevin Tansey, and Heiko Balzter. 2014. “Aboveground Biomass Mapping in Krasnoyarsk Kray (Central Siberia) Using Allometry, Landsat, and ALOS PALSAR.” In Proceedings RSPSoc Annual Conference. Aberystwyth, Wales: Remote Sensing and Photogrammetry Society.

Santoro, M, L Eriksson, J Askne, and C Schmullius. 2006. “Assessment of Stand-Wise Stem Volume Retrieval in Boreal Forest from JERS-1 L-Band SAR Backscatter.” International Journal of Remote Sensing 27 (16): 3425–54. doi:10.1080/01431160600646037.

Shvidenko, A. Z., D. G. Schepaschenko, E. A. Vaganov, A. I. Sukhinin, Sh. Sh. Maksyutov, I. McCallum, and I. P. Lakyda. 2011. “Impact of Wildfire in Russia between 1998–2010 on Ecosystems and the Global Carbon Budget.” Doklady Earth Sciences 441 (2): 1678–82. doi:10.1134/S1028334X11120075.

Simard, Marc, Naiara Pinto, Joshua B. Fisher, and Alessandro Baccini. 2011. “Mapping Forest Canopy Height Globally with Spaceborne Lidar.” Journal of Geophysical Research 116 (G04021). doi:10.1029/2011JG001708.

Walker, Wayne S, Josef M Kellndorfer, Elizabeth Lapoint, Michael Hoppus, and James Westfall. 2007. “An Empirical InSAR-Optical Fusion Approach to Mapping Vegetation Canopy Height” 109: 482–99. doi:10.1016/j.rse.2007.02.001.

Wilhelm, Sebastian, Christian Hüttich, Mikhail Korets, and Christiane Schmullius. 2014. “Large Area Mapping of Boreal Growing Stock Volume on an Annual and Multi-Temporal Level Using PALSAR L-Band Backscatter Mosaics.” Forests 5 (8): 1999–2015. doi:10.3390/f5081999.

Spaceborne mapping of forest biomass is an active research topic due to the lack of accurate estimates of global forest carbon stock, as well as the lack of suitable tools for deforestation and forest degradation monitoring [1, 2]. In May 2013, BIOMASS was selected by the European Space Agency (ESA) to become the 7^th Earth Explorer Mission [3, 4]. Equipped with the first P-band (435 MHz) synthetic-aperture radar (SAR) in space, a major objective of BIOMASS will be estimation of tropical forest biomass and disturbance. During a dedicated phase at the beginning of the mission, BIOMASS will operate in a repeat-pass tomographic mode, giving information about the vertical distribution of backscattering in forests. However, due to ITU (International Telecommunication Union) regulations, BIOMASS coverage will be restricted over Europe and North America, leaving large areas of boreal forests unmonitored. To improve boreal forest mapping, ESA is investigating the implementation of a passive companion satellite to the Argentine SAOCOM L-band (1.275 GHz) SAR satellite. The SAOCOM companion satellite (SAOCOM-CS) mission will focus on mapping of the global boreal forests using SAR tomography [5].

One of the greatest challenges in forest parameter estimation from SAR imagery is the mitigation of moisture effects. An uncompensated moisture signal will introduce bias in biomass estimates, making them less reliable. Thus, it is of great importance to improve the understanding of these effects and to develop correction schemes.

In tropical forests, moisture effects in a French Guianese forest have been addressed during the TropiScat campaign [6, 7], in which temporal changes were studied using a tomographic radar with 20 antennas mounted on a 55-m high tower. Diurnal cycles and rain episodes were observed and it was concluded that temporal coherence is highest at dusk and dawn. In the AfriScat campaign, a similar experiment is being conducted in an African tropical forest in Ghana.

BorealScat, developed by the Radar Remote Sensing group from Chalmers University of Technology in Gothenburg, Sweden in close co-operation with the Division of Forest Remote Sensing at the Swedish University of Agricultural Sciences (SLU) in Umeå, Sweden, will investigate moisture and other temporal effects in a boreal forest. Campaign objectives include:

understand weather effects in P- and L-band radar data,

predict weather effects in P-band BIOMASS and L-band SAOCOM-CS data,

investigate algorithms which mitigate moisture variations in P- and L-band data.

In the early spring of 2016, a dedicated 50-m high tower will be built within the test site Remningstorp, situated in southern Sweden. Measurement equipment will consist of a 20-channel vector network analyser (VNA), a switch network, and 20 log-periodic, wideband antennas with radomes. During the first phase of the campaign, the system will use a similar antenna geometry to that used in the TropiScat campaign, with 20 antennas mounted in 4 columns (5 antennas in each column, separate columns for transmission and receiving and for horizontal and vertical polarisations ). Measurements will be conducted at both P- and L-bands. Multiple antennas in the vertical direction and multiple polarisations will be used for tomography and polarimetry.

During the first years of measurements, when neither BIOMASS nor SAOCOM-CS will be available, existing airborne and spaceborne SAR data will be analysed, mainly P- and L-band airborne SAR data acquired over Remningstorp during the BioSAR campaigns, and L-band SAR data acquired by the ALOS PALSAR and ALOS-2 PALSAR-2.

A dedicated weather station will be installed in close proximity of the tower. Moreover, the forest will be monitored using surveillance cameras. The system will be connected to the internet via a 4G broadband link. A dedicated webpage (www.borealscat.se) will be maintained, featuring a live video stream from the forest, live weather report, and a blog. Moreover, periodic in situ and terrestrial lidar scanning (TLS) surveys will be conducted by specialists from SLU.

 

[1]       IPCC, Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press, 2013.

[2]       C. Parker, A. Mitchell, M. Trivedi, and N. Mardas, The Little REDD Book: A guide to governmental and non-governmental proposals for reducing emissions from deforestation and degradation: Global Canopy Programme, John Krebs Field Station, Oxford OX2 8QJ, UK, 2008.

[3]       F. Heliere, F. Fois, M. Arcioni, P. Bensi, M. Fehringer, and K. Scipal, "Biomass P-band SAR interferometric mission selected as 7th Earth Explorer Mission," in 10th European Conference on Synthetic Aperture Radar (EUSAR), Berlin, Germany, 2014, pp. 1152-1155.

[4]       ESA, "BIOMASS, Report for Mission Selection," European Space Agency (SP-1324/1)2012.

[5]       N. Gebert, B. Carnicero Dominguez, M. W. J. Davidson, M. Diaz Martin, and P. Silvestrin, "SAOCOM-CS -- A passive companion to SAOCOM for single-pass L-band SAR interferometry," in 10th European Conference on Synthetic Aperture Radar (EUSAR), Berlin, Germany, 2014, pp. 1251-1254.

[6]       C. Albinet, P. Borderies, T. Koleck, F. Rocca, S. Tebaldini, L. Villard, et al., "TropiSCAT: A Ground Based Polarimetric Scatterometer Experiment in Tropical Forests," IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 5, pp. 1060-1066, June 2012.

[7]       D. H. T. Minh, S. Tebaldini, F. Rocca, T. L. Toan, P. Borderies, T. Koleck, et al., "Vertical Structure of P-Band Temporal Decorrelation at the Paracou Forest: Results From TropiScat," IEEE Geoscience and Remote Sensing Letters, vol. 11, pp. 1438-1442, Aug. 2014.

 

 

The Copernicus programme is a European initiative for the implementation of information services dealing with environment and security, mainly based on observation data received from Earth Observation (EO) satellites. In the frame of this programme, ESA launched the Sentinel-2A optical imaging mission on 23th June 2015. This satellite delivers a new generation of optical data products designed to directly feed downstream services mainly related to land monitoring, emergency management and security. To ensure the highest quality of service for the Sentinel-2 mission, ESA set up the Sentinel-2 Mission Performance Centre (MPC) supported by several Expert Support Laboratories (ESL). This presentation reports early results of validation of the Sentinel-2 L2A processor Sen2Cor performed by the ESL-L2A.

 Sen2Cor Level 2A processing is applied to Top-Of-Atmosphere (TOA) Level 1C ortho-image reflectance products. Sen2Cor L2A outputs are a scene classification image, aerosol and water vapour maps and the ortho-image Bottom-Of-Atmosphere (BOA) corrected reflectance product. Scene classification is generated together with Quality Indicators for cloud and snow probabilities.

 In this poster we present the calibration and validation datasets used for our on-going L2A cal/val activities, covering different land cover types, different atmospheric conditions and including different latitudes in order to cover various solar angles and seasons. The already performed and on-going in-situ campaigns for collecting atmospheric and ground reference data to support the L2A products validation are also described.

 The methodologies of calibration and validation of the atmospheric correction processor are presented under the form of a workflow chart. This shows the  validation approach for aerosol optical thickness and water vapour estimates and for the validation of BOA reflectance product, the validation approach for cloud screening and scene classification (stratified random sampling), and the interactions with the L2A processor calibration activities. Details are given about the specifics of each method.

 Finally, the results of Sen2Cor L2A products validation activities (at the time of the LPS 2016 conference) will be provided for the cloud screening and scene classification, as well as for the atmospheric correction.

The SWOT satellite by interferometry radar (Surface Water and Ocean Topography; NASA and CNES mission, with a CSA and UKSA contribution) will be launched in 2020 and will provide data on the water level with an high spatial resolution: 1 km for the oceans and 50-100 m of width for the rivers, 250m² for the lakes. With such resolution, SWOT will have the capacity to measure changes in water levels of 68% of the lakes and many rivers, including those of small size that were previously inaccessible from other satellites. However, with a number of passages from 2 to 7 per cycle of 22 days, it is necessary to understand the hydrological variability modes that will be recorded or not by SWOT.

To answer to the question of the SWOT use in the french rivers, estuaries (Seine, Garonne, Gironde, Loire, Rhône) and coastal zones (Channel and Atlantic), we use two approachs. The first approach is to simulate SWOT data without error and with error (white noise) according to the SWOT orbit to study the effect of the number of passages of SWOT to this capacity to record the hydrological temporal variability. We use statistical and signal processing methods (wavelets) to compare the in situ water level and the simulated SWOT water level. The second approach is the modeling of the spatial and temporal variability of water level by the T-UGO model and the use of model data as input data of SWOT simulator HR (High Resolution).

The results of the first approach show, for the 4 studied rivers, SWOT reproduces well the main hydrological variability patterns (2y = NAO mode, 1y = hydrological cycle, 1,5-3 months = flood period), and the wavelet coherence between the in-situ data and the SWOT simulated data is from 91 to 99%. For the estuaries (Seine and Gironde) and coastal zones (The Channel, Altantic), SWOT reproduces well the water levels in the estuary upstream and less in the estuary downstream and in the coastal zones (the variability modes of 2 and 1 y are not well reproduced, and the mode of 2-4 months is overexpressed), and the wavelet coherence between SWOT simulated data and in situ data decreases from downstream to the upstream estuary, from 90 to 60 %.

Regarding modeling, T-UGOm model used in the Seine estuary reproduces well the temporal hydrological variability and the water level amplitude. The water level maps obtained by modeling in the Seine estuary show that the water levels are spatialy, highly variable in different hydrodynamic conditions and also in the precise hydrodynamic condition and this high spatial variation can be observed over distances of less kilometer and this shows the importance of the high spatial resolution of SWOT to see these transitions in these environments.

These T-UGOm model data were used as imput data in the SWOT simulator HR. The result shows that the majority of the SWOT measure points are located in the channel and with low water level error, but some are outside and the error can be plurimetric for points outside or on the edge of the channel.

Crop rotation has been used for thousands of years. This agronomic practice was developed to produce higher yields by replenishing soil nutrients and breaking disease and pest cycles. The current consensus is still that crop rotation increases yield and profit and supports sustained production. With increased availability of open agricultural data and in particular time series of detailed crop parcel data it becomes possible to determine consistent, but regionally diverse, crop rotation practices. We use data sets in the Netherlands to focus on potato rotation pattern in major production areas.
Potato prices fluctuate yearly and is strongly dependent on supply. Detailed and early crop occurrence maps, coupled with annual trends analysis and actual seed sales records, can provide important production information to potato growers and   the potato processing industry. The new availability of frequent Sentinel-1 and -2 imagery under a free and open data license is providing unique opportunities to derive early area estimates from a combination of rule-based image classification steps which are conditioned by a priori probability masks derived from multi-annual crop rotation mask.
Based on the analysis of historical data in combination with regional knowledge on soil and potato varieties, we have created a model which predicts which parcels in the Netherlands will have the highest probability to be used for potato production in the subsequent season. We use both inter-crop rotation knowledge (e.g. 3 or 4 year rotation for the same crop, i.e. potato) as well as intra-crop rotations (e.g. typical inter-annual sequences of different crop types, i.e. potato and sugar beet). We create probability maps that enumerate the likelihood, from 0 to 1, that potato will be grown.

These probabilities are further fine-tuned with simple masking operations derived from early season Sentinel-1 and -2 time series that highlighted expected surface conditions for potato cultivation, which are, again dependent on local soil types and management practices. We will show practical application of the method for the 2015 potato campaign, and will present early results for 2016. We will also discuss how our method can be generalised to other crop rotation practices.

The Joint Polar Satellite System (JPSS) provides continuity for NOAA’s Polar-orbiting Operational Environmental Satellites (POES) on the afternoon orbit. Vegetation Index (VI) and Green Vegetation Fraction (GVF) products derived from the Suomi National Polar-orbiting Partnership (NPP) Visible Infrared Imaging Radiometer Suite (VIIRS) instrument are being generated operationally, and distributed through the NOAA Comprehensive Large Array-data Stewardship System (CLASS). The Suomi NPP VIIRS VI Environmental Data Record (EDR) consists of two products, the Top Of Atmosphere (TOA) Normalized Difference Vegetation Index (NDVI), and the Top Of Canopy (TOC) Enhanced Vegetation Index (EVI). The VI EDR is generated for each granule at 375 meter resolution at nadir. The VIIRS VI EDR products have been compared to MODIS VI products and good consistency was found. Explicit validation is performed against VI derived from FLUXNET radiation measurements. The current operational Suomi NPP Vegetation Index EDR algorithm has been updated to include the TOC NDVI, additional quality flags, and improved high quality definitions for the three indices and will be deployed and transitioned to operations before the launch of the JPSS-1 satellite.
The GVF product consists of a weekly global and a regional GVF composite over North America, updated daily, at 4 km and 1 km resolutions respectively. These GVF products have been shown to meet the JPSS program requirements, and are currently tested for operational use in land surface and numerical weather prediction models at the National Centers for Environmental Prediction (NCEP). The VIIRS GVF product was validated against Landsat-based GVF data. The calculated accuracy and precision statistics were 0.079 and 0.109 for the global GVF product, and 0.065 and 0.126 for the regional GVF product. The JPSS land algorithm developers are currently embarking in the planning and development of common vegetation index algorithms and products from JPSS, GOES-R, and other non-NOAA missions such as Sentinel-3 and Himawari, to better serve the needs of operational users and the scientific community. This development also is aimed at including additional variables in the operational vegetation product suite, such as fraction of Photosynthetically Active Radiation (fPAR) and Leaf Area Index (LAI).

Plant stress detection is an important issue in the management and conservation of natural and anthropogenic environments. Remote sensing is a tool that can monitor large areas to detect plant stress. Most of research on plant stress detection has been focusing on visible-near infrared (VISNIR). However, the thermal infrared (TIR) seems to contain valuable information on Leaf Water Content (LWC) and structural and microstructural traits of leaves, which could be used as proxies for plant stress detection.

This study explores changes in the TIR in relation to imposed plant stress. Sixty plants of Rhododendron cf. catawbiense were used in an two way factorial experiment with water and temperature stress. In total 75 leaves were tracked and measured after the experiment to identify changes in leaf traits and spectral changes in the TIR (1.4 to 16 µm). All stress treatments (dry-cold, wet-cold, dry-ambient) showed significantly different spectra compared to leaves from normal growing conditions (wet-ambient). Leaves of stressed plants displayed similar changes in LWC with a reduction to 41.1% in the most extreme conditions (dry-cold) compared to 61.8% of LWC in the control (ambient-wet). Also stressed leaves tended to have thicker leaves and upper cuticles, increases in stomata density and reductions in stomata size as strategies to cope with dehydration and extreme temperatures. Multinomial models were fit using leaf trait changes as proxies to determine plant stress.

The stress leaves after the experiment, also had different spectra in the TIR compared to the control plants (ambient-wet conditions). Descriptive statistics, PLSR analysis and multinomial models were conclusive in the identification of LWC, leaf thickness, lignin content, leaf area and stomata density, as the leaf traits more affected by the stress treatments and which can be used as proxies for stress detection. PLSR optimized models (PLSR_opt) were fitted to predict changes in leaf traits using changes in distinct bands of the spectra. Most bands selected for these predictive models on the most important leaf traits (e.g. lignin, cellulose and LWC) have been reported as molecular vibrations related to those compounds in other studies. 

The importance of Earth Observation (EO) techniques and the benefits they can bring to agricultural production and food security in general are already well proven facts. However, their scope and impact were usually severely limited due to data availability issues, low coverage, or different sensor or mission design constraints. Fortunately, we are currently experiencing a quantum leap in the field of spaceborne EO, which will hopefully proliferate and broaden current applications, create new ones, and finally bring them closer to everyday life and small and medium businesses. One of the EO aspects that has particular significance for management and monitoring of agricultural production on different levels of decision making is timely production of accurate and precise land-cover and land-use thematic maps, in particular crop classification maps. Therefore, there is a constant interest and an ongoing activity aimed at further development and improvement of existing remote sensing techniques. Although it is almost certain that the expected data richness in the near future will solve many of the previous issues and result in the immediate progress of currently used classification techniques, it will also create the need to adequately address a new working environment created by measuring instruments with higher spatial, spectral and temporal resolutions. In the line with these efforts, aim of this paper is to study and analyze a standard pixel-based supervised crop classification task, but in the new scenario characterized by the presence of dense multispectral time series. It will investigate different methods for exploiting and incorporating additional information provided by dense time sampling and analyze applicability of new feature descriptors in the given context. Dense time sampling will certainly enable better characterization of the plant phenology dynamics and eventually better discrimination between different crop types, but the open question is how the feature extraction should be done in order to achieve more efficiency and better utilize available information. The former is the aspect that will be addressed through the analysis of information available in the spectral and temporal domain at the level of each pixel and over the extent of available time series. Quality of the obtained crop classification maps will be assessed numerically and by visual comparison on the reference ground-truth dataset. The study will utilize the satellite data generously provided by the Spot-5 Take5 initiative, which was organized by CNES and ESA between the April and September 2015 with the aim of simulating the data that will be available from the fully operational ESA’s Sentinel-2 mission in the near future.

Over the last eleven years, five national parks were established in Germany. Country’s newest national park Hunsrück-Hochwald with protected area of 10.000 hectares was established in March 2015 in two federal states Rhineland Palatinate and Saarland. As a priority, national parks protect and preserve biological diversity. Especially forest conversion to natural species composition of stands focused on European Beech (Fagus sylvatica L.), which is the native tree species for this region. Therefore, accurate mapping of tree species within the national park is of great importance for long-term monitoring programs, sustainable forest planning and management.

The potential of Sentinel-2 with its wide swath observation geometry, high spatial resolution and extended band set is considerable for forest inventory and forest management application. Due to high repetition rate there is ability to multi-temporal analysis of phenological development and growing stages. As long as Sentinel-2 images are not available for the study area, we used simulated Sentinel-2 data from airborne hyperspectral campaigns. For the simulation of Sentinel-2 data high resolution hyperspectral data from NEO's HySpex were used. HySpex is a hyperspectral pushbroom imaging spectrometer working in the visible (HySpex VNIR-1600) and SWIR (HySpex SWIR-320m-e) spectral regions. The combination of the VNIR and SWIR sensors leads to a continuous spectral coverage of the range 416 to 2500 nm, allowing simulation of all 13 Sentinel-2 bands.

This study aimed to classify five common tree species (European Beech, Pedunculate and Sessile Oak, Norway Spruce, Douglas Fir and Scots Pine) in the national park area and their development stages (qualification, dimensioning, maturing). To classify tree species from simulated Sentinel-2 data we examined three discriminative models (AdaBoost, Support Vector Maschines and Self Organazing Maps) and one generative method (Gaussian Maximum Likelihood). Image classification was performed on the synthetic multi-temporal Sentinel-2 data as well as on a high resolution airborne imaging spectrometer data to be used for comparison. Evaluation of the accuracy was carried out by assessing confusion matrices. It can be stated that all discriminative classification models outperformed the generative model. Image classification products of this study will be integrated into forest monitoring concept of the national park Hunsrück-Hochwald.

The ability of satellite data to monitor agriculture reliably is dependent on several factors but is constrained fundamentally by the satellite spatial resolution relative to the field spatial dimensions. The recently launched Sentinel-2 MultiSpectral Instrument (MSI) and Landsat-8 Operational Land Imager (OLI) sensors have 10m and 30m visible and near-infrared bands respectively and offer an opportunity to monitor agriculture over regions with small and complex fields that cannot be resolved using coarser resolution 250m MODIS data. This study examines the spatial characteristics of these sensors in Africa considering regions of small holder and commercial agriculture that provide examples of small/complex and large/uniform fields respectively.  The Sentinel-2 data were reprojected into alignment with the Landsat-8 data Web Enabled Landsat Data (WELD) that are defined in tiles in the MODIS product global sinusoidal projection. Visual examination and spatial profile analysis of contemporaneous MODIS, Landsat-8 and Sentinel-2 normalized difference vegetation index data are used to demonstrate the potential improvements that 10m Sentinel-2 data will make for agricultural monitoring in small holder dominated agricultural regions that are difficult to monitor but are common in Africa and Asia.      

Forestry is an important component in achieving national Kyoto Protocol (and its successor) greenhouse gas (GHG) emission reduction targets as net sinks of atmospheric CO₂. The Republic of Ireland’s forest areas have increased from 1% of land at the start of the 20^th century to over 10% by the end. However, many forest plots are small and fragmented, and often under one hectare in area. A reliable system is needed for quantifying land use changes for global GHG models and reporting requirements. The majority of Irish forests are for commercial wood production, and can be subject to both planned (planting, thinning, and clearcut harvesting) and unplanned disturbances, e.g., deforestation or storm damage. Optical remote sensing of forest disturbance in Ireland can be confounded by frequent cloud cover and shadows, limiting the availability of medium resolution Landsat imagery, particularly in upland areas. Furthermore, most forest cover disturbance events occur in areas smaller than a MODIS pixel (250 × 250 m), requiring the use of higher-resolution, but lower temporal resolution sensors. Additionally, cloud cover results in reduced MODIS pixel reliability for Ireland, making MODIS time series especially noisy. Because acquiring Landsat imagery from the same times of year (± one week in differing years) can be difficult, and the temporal importance of vegetation phenology, detecting disturbances requires creative strategies. The cloud cover issue is circumvented by utilizing both land cover classification and NDVI time-series for unclouded, unshadowed pixels, as determined from Fmask cloud mask layers. The changes in land cover denote major events, such as clearcutting, planting, and developmental stages, and NDVI can be used to determine thinning events. In contrast to optical remote sensing, synthetic aperture radar (SAR) can penetrate cloud cover and being an active sensor, it is independent of solar illumination. The Advanced Land Observing Satellite (ALOS) Phased Array L-band SAR (PALSAR) launched in 2006 has been successfully used for forest disturbance mapping, however very few studies have investigated the potential of SAR for forest monitoring in sparse and fragmented forest landscapes. After pre-processing and masking of settlements, a Random Forest machine learning approach was adopted to generate forest/non-forest maps using backscatter and Grey Level Co-occurrence Matrix texture measures. Techniques to further refine the forest class into stands of different ages and different dominant species were explored to aid discrimination of disturbance events that may have occurred. Using the backscatter of different forest patches, their biomass and hence carbon content can be estimated. The methodologies will be transferred and tested on the recently launched Sentinel-2A and ALOS-2 data. Along with other new optical and SAR sensors, there is potential for annual forest monitoring in Ireland on a national scale.

Land surface dynamics are induced by anthropogenic and/or climatic drivers and can have a major impact on ecosystem functioning. During the last decades, Ukraine has experienced immense institutional and environmental changes. Considering the importance of land surface dynamics, this research aims to explore long term and seasonal trends and variability of land-use and land-cover across Ukraine for over three decadal time span (1982-2013). To analyse and evaluate the temporal trends of inter-annual and seasonal changes, this study was conducted in two consecutive steps. First, Global Inventory Modelling and Mapping Studies (GIMMS) Normalized Difference Vegetation Index (NDVI) time series were used for the trend analysis at different aggregation level. Second, variation of essential phonological variables (start/end of the season, peak and small integral of the season) over the whole area of Ukraine were studied. We tested several most used methods for trend analysis including Mann-Kendall trend analysis, Breaks For Additive Season and Trend (BFAST).
The obtained results were different depending on used method. Based on annual trends, around one third of the area was characterised with positive trends. Meanwhile, the use of breakpoints increased the number of detected significant negative trends. Around 26 % of the area was characterised with negative breaks. The breakpoints with negative magnitude were mainly pronounced in Chernihiv, Zhytomyr, Sumy and Kiev regions. The analysis of phenological metrics revealed the shifts in seasonality. This was particularly evident in the south eastern regions of the country (Odessa, Mykolayiv, Crimea). In order to relate the detected changes with disturbances and to check the robustness of temporal distribution of breakpoints, the validation is carried out based on reference areas of changes. 
Trajectories of these vegetation metrics derived from a full-length remote sensing datasets can lead to better understanding of land surface dynamics in general and specifically of vegetation degradation and recovery, the potential causes of these dynamics and their underlying processes. In addition, the use of different methods (e.g. parametric and non-parametric trend analysis, change point analysis) enables a more reliable identification of ecosystem dynamics.

With the globally decreasing amount of in-situ stations, satellite altimetry based water levels are an important supplement to obtain continuous time series of the worlds inland water.

In this presentation we demonstrate two new services, that are related to inland water and altimetry. The first is Altimetry for inland water (AltWater), which is a new open service, that provides altimetry based time series for inland water. Currently, the service includes data from cryoSat-2, but we intend to add other missions in future versions.
The service is developed under the FP7 project Land and Ocean Take Up from Sentinel-3 (LOTUS) and is available at www.altwater.dtu.space. The web page contains a google map with the available inland water bodies and the user can access the data by clicking on the specific target of interest. For each target the user can download both the along-track water levels and the estimated water level time series.

The second "tsHydro" is a software package, that is implemented in the open source environment "R". The package enables the user to easily construct water level time series for lakes and rivers based on along-track altimetry data. The implemented model, that predicts the temporal evolution of the water level, is a novel state-space model, that accounts for erroneous data in a robust manner. The package is available at https://github.com/cavios, where installations instructions and package documentation also are found.

The services are here demonstrated for various targets around the world

In this study, which is part of the FP7 project Land and Ocean take up from Sentinel-3 (LOTUS), we demonstrate the potential SAR altimetry. We consider lakes at various sizes and evaluate the CryoSat-2 derived lake levels in terms of along-track precision and agreement with in-situ data. As a reference we compare our CryoSat based results with conventional altimetry such as Envisat.

We find that the precision of the along-track mean water level is a few cm, even
for lakes with a surface of just 9 km². The high precision makes it possible to detect water level variation below the decimeter level.

To derive lake level time series we apply a state-space model with a robust handling of erroneous data. Instead of attempting to identify and remove the polluted observations we use a mixture distribution to describe the observation noise, which prevents the polluted observations from biasing our final reconstructed time series.

These results demonstrate the promising possibilities of the upcoming mission Sentinel-3, which potentially will be able to provide accurate time series for small lakes.

Vulnerability to climate variability and human impacts put wetlands into the focus of interest of water and environmental managers both in arid and humid areas. Environmental services of wetlands highly depend on their habitats, so monitoring this latter has strong societal relevance.

A number of global land cover maps (e.g., IGBP-DIS, GlobCover2009, GLC2000 and GLCNMO) have been developed using satellite remote sensing data at a spatial resolutions ranging from 1 km to 300 m in which wetlands are treated as one of the individual land cover classes. However, the spatial distribution and habitat changes of wetlands have not been fully characterized in most of these products. This is because of the limitations related to the spatial and temporal resolutions of the used satellite data as well as the classification methodologies applied. Regional efforts, focusing on more detailed wetland mapping provided some more detailed results (e.g. GlobWetland II), but the applied tools (e.g. proprietary software) and the highly interactive methods hamper the direct use of the results in the daily practice of wetland managers.

Habitat mapping is more complex than land cover mapping. On the one hand, a dominant vegetation cover may comprise several different habitats, without having significantly different spectral characteristics. On the other hand, combination of different data types (optical and microwave remote sensing time series, in situ environmental data, land use information, conceptual models and spatial indicator parameters) broaden the possibilities of identifying the habitats in and around the wetlands.

In the GlobWetland Africa project of the TIGER Initiative, (semi-) automatic workflows are being developed for wetland habitat mapping with the focus on Africa. In our presentation, we are reviewing the challenges and pitfalls of automation of the different steps. A systematic approach to historic base-line mapping will be outlined using medium resolution optical images (Landsat, SPOT, etc.) based on the wetland typologies defined by the Ramsar Convention. Regular data provided by the Sentinels forms the basis for mapping the recent status. Classification methods need to be robust and flexible even in the extremes of African conditions, so decision tree algorithms with given structures will be presented, which can be parameterized for the needs of specific regions. Detection of land cover change is proposed to be based on a binary change mask approach where multivariate alternation detection (MAD) analysis is combined with the re-classification of changed areas and post-classification comparison methods. Indicators will be used for quantifying the changes.

The presentation will focus on the theoretical background and the scientific challenges, although it is worth mentioning that an open source software environment, following the logic of the Water Observation Information System (WOIS, developed in the TIGER-NET project of ESA) will be used for the implementation of the (semi-) automatic workflows.

Recent climate warming has shifted the timing of spring and autumn vegetation phenological events in the temperate and boreal forest ecosystems of Europe. In many areas, spring phenological events start earlier and autumn events switch between earlier and later onset. Consequently, the length of growing season in mid and high latitudes of Europe has been extended. However, lagged effects (the impact of a warm spring or autumn on subsequent phenological events) on vegetation phenology and productivity are less explored. In this study, we have (1) characterized extreme warm spring and extreme warm autumn events in Europe during 2003-2011, and (2) investigated if direct impact on forest phenology and productivity due to a specific warm event is translated into a lagged effect in subsequent phenological year. To achieve these objectives, the study used Envisat MERIS Terrestrial Chlorophyll Index (MTCI) time series data to characterize the leaf phenology of forested areas in Europe that experienced extreme warm spring and autumn events during 2003-2011. Further, time series temperature data from the European Climate Assessment and Dataset (ECAD) were used to estimate the anomalous temperature and hence, the two scenarios- extreme warm spring and extreme warm autumn sites; from these sites samples of broadleaf deciduous forest (BLDF), needleleaf deciduous forest (NLDF), needleleaf evergreen forest (NLEF) and mixed forest (MF) collected. The study essentially derived forest phenology and productivity variables for normal and anomalous temperature conditions. The phenological variables were onset of growth (OG) and end of senescence (EOS) whereas all the MTCI values of the area under curve when MTCI was plotted against day of year (DOY) were summed up and called integrated MTCI (I-MTCI) to estimate forest productivity. The phenological and productivity variables for normal and anomalous (extreme warm events) temperatures were compared across the various forest types.

We found that warmer events in spring occurred extensively in high latitude Europe producing a significant earlier OG in NLDF, and MF. However, this earlier OG did not show any significant carry-over effects to autumnal leaf senescence. BLDF, NLEF, NLDF, and MF significantly delayed EOS as a result of extreme warm autumn events; and in the following year’s spring phenological events, OG started significantly earlier. Extreme warm spring events directly led to significant (p=0.0189) increases in the productivity of NLDF. It was only the MF that experienced significant (p=0.0132) increase in productivity of the following year’s spring phenology in response to the preceding warmer autumn experienced. 

Included in the larger Guiana Shield ecosystem, Suriname, Guyana, French Guiana and the Brazilian state of Amapá possess one of the largest continuous tracts of pristine forest in the world. Under little threat until fifteen years ago, deforestation and forest degradation are of increasing concern in the region. Gold mining activities driven by the sustained increase of gold price has experienced a significant boom and represents nowadays one of the main driver. The pollution of rivers and streams by mercury used in small-scale gold mining is also expanding, which increases risks to local population health and freshwater biodiversity.

In 2010, the French National Forest Office (ONF) showed by using optical satellite images at medium and high resolution (HR) that gold mining activities’ impacts on forest cover and freshwater increased approximately by a factor three in the region between 2001 and 2008. More recently, Alvarez-Berríos et al. (2015) pointed out a sustained acceleration of deforestation caused by gold mining in the Guiana shield between 2007 and 2013. However, this study which was performed using low resolution data at the scale of South America has limited capacity to detect gold mining, especially in the high forest cover of Guiana Shield where small- and medium-scale operations account for most of the deforestation.  

To overcome this limitation, the REDD+ for the Guiana Shield project conducted a study co-funded by WWF Guianas to update for 2014 the ONF 2001-2008 results, using optical multi-sensors data at medium and high resolution. The study was carried out following a unique collaborative and participatory approach involving a team of experts from the forestry and environmental services of each territory, namely SEMA (Amapá-Brazil), ONF (French Guiana-France), GFC (Guyana), and SBB (Suriname). The results confirmed the rapid expansion of the activity in the region where more than 92,000 ha were newly deforested between 2008 and 2014, compared to approximately 46,000 ha during the period 2001-2008. In 2014, more than 9,000 km of waterways were in direct contact with mining sites, which is approximately 6.5 times more than in 2001.

Although a reliable, accurate and robust regional methodology has been developed and operationally implemented, the frequent and widespread cloud cover of the Guianan moist forest region represents a challenge for the use of optical HR data. The need to process time series of satellite images in most areas to reduce cloud cover is time-consuming. Despite processing more than two hundreds images, 3.6% of the study area remained masked by clouds.

The recent free access to SAR HR Sentinel-1 data offers great opportunities to improve the process. SAR sensors can peer through clouds and their sensitivity to soil moisture can help to better detect small-scale mining sites. Therefore, the REDD+ for the Guiana Shield project has started to build capacities in the region on SAR image interpretation and processing using the Sentinel Application Platform (SNAP). A first mosaic of Sentinel-1 data covering Suriname, Guyana, French Guiana and the Brazilian state of Amapá has been created and automated pre-processing steps have been developed. The integration of Sentinel-1 data in the regional gold mining monitoring system has been successfully tested in four study sites, one in each country. The coming free access to optical HR Sentinel-2 data opens even more perspectives towards the development of cost-effective monitoring systems in the region, especially valuable in the context of REDD+.

This paper first presents the results of the impact of gold mining activities on the forest cover and freshwater for 2014 and shows the evolution since 2001. Secondly, it provides the first outcomes towards the development of time- and cost-efficient forest monitoring systems in the region.

Oil palm is a major cash crop in Ife local government area however, oil palm is not presently been cultivated in the study area but grows in the wild hence the need to identify suitable land for its cultivation and also  increase the crop production to generate more revenue from agriculture.Geo-spatial assessment of Land suitability based on the FAO Land Framework was carried for oil palm cultivation in the Ife Central Local Government Area using GIS integration with soil diagnostics survey, Remote Sensing, and Multi-Criteria Evaluation (MCE). The aim of this study is to carry out an objective geospatial assessment and mapping of land suitability for oil palm in terms of basic physical and chemicals factors, evaluation of the spatial variation in these factors and eventual identification of suitable and unsuitable areas using Soil Diagnostic Survey, Remote Sensing and GIS with MCE. The study area was partitioned into grids of equal dimensions and soil samples collected from the centroid of each the grids. The physio-chemical properties of the soil samples collected were analyses in laboratory using standard methods and spatial attributes assigned to soil properties; other important factors with known spatial attributes such as climate, topography and land use land cover were assessed using standard geospatial methods. All the factors were integrated into GIS geodatabase from which spatial analysis was carried out and MCE was applied to arrive at the land suitability mapping. The result showed that the area has moderately suitable climate and highly suitable topography for the oil palm cultivation. The MCE indicates that the area is suitability for oil palm with 12.10% (17.18km²) of the area highly suitable, 20.07% (28.50km²) moderately suitable while 24.08% (37.46km²) of the study area is marginally suitable for oil palm production. The major problems affecting oil palm suitability in the study area are insufficient rainfall, high pH, low nitrogen, phosphorus and cation exchange capacity (C.E.C)   Hence, this study recommends effective crop residue management, increased use of leguminous plants as well as use of nitrogen and phosphatic fertilizers in order to enhance the suitability of land and increase oil palm yield in the Ife central LGA. This study concludes that farmers and other stakeholders in agriculture should be encouraged to cultivate oil palm in the study area.

To assess the North American high-latitude vegetation response to the rising temperature, we derived NDVI trend for 91.2% of the non-water, non-snow area of Canada and Alaska using the peak-summer Landsat surface reflectance data of 1984-2012. Our analysis indicated that 29.4% and 2.9% of the area of Canada and Alaska showed statistically significant positive (greening) and negative (browning) trends respectively, at significance level p < 0.01, after burned forest areas were masked out. The area with greening trend dominated over that with browning trend for all land cover types.  The greening occurred primarily in the tundra of western Alaska, along the north coast of Canada and in northeastern Canada; the most intensive and extensive greening occurred in Quebec and Labrador.  The browning occurred mostly in the boreal forests of eastern Alaska. The Landsat-based greenness trend is broadly similar to the AVHRR-based trend for all vegetation zones. However, for tundra, the Landsat data indicated much less extensive greening in Alaska North Slope and much more extensive greening in Quebec and Labrador, and substantially less extensive  browning trend in the boreal forests that were free of fire disturbances. These differences call for further validation of the Landsat reflectance and the AVHRR NDVI datasets. Correlation study with local environmental factors,  such as topography, glacial history and soil condition,  will be needed to  understand the heterogeneous greenness change at the Landsat scale. 

Volga is a great Russian river flowing into the Caspian sea. Together with Akhtuba brook it feeds the one of largest wetland territory in Russia occupying an area about 9000 km3. It is called Volga-Akhtuba floodplain (VAF). Volga runoff is regulated by Hydro-Electric Power station (HEP) cascade closed by Volzhskaya HEP located in the origin of Akhtuba brook.

In arid environment Volga-Akhtuba wetland ecosystems are very sensitive to Volga runoff which currently depend on downstream Volzhskaya HEP downstream discharge. Volga's natural hydrological cycle is approximately decennial which result in extremely dry year appearance every 10 years including 2015. And there are more long-term hydrological regime alterations as a background.

It was observed that many inudable sites within VAF were not flooded in spring 2015. This resulted in wetland plant cover breakdown in many field reference monitoring points and complete dry-up of many model water bodies.

As VAF is a very large and geterogenous territory, the remotely sensed data analysis is the only way to analyze its ecosystems as a a whole and localize and quantify plant cover and hydrographic network changes.

We performed the joint analysis of remotely sensed data along with field geobotanical and hydrological observations. The remotely sensed data chosen for comparative analysis was for dry years 1975, 1984, 1996, 2006, 2015 as well as for wet years 2007 and 2013.

Analysis of field and remotely sensed data for 2015 and other years revealed the drastic changes in water bodies, soil moisture and plant communities in 2015 summer season on a vast areas within VAF compared to wet and even other dry years.

Thus all the upper (widest) VAF segment situated between Volgograd and Akhtubinsk together with some downstream territories became a place of unexampled ecological catastrophe. Desertification processes observed in previous years were found to expand to most VAF territory.

Mean decrease of healthy wetland ecosystem areas in 2015 for the upper VAF segment (Volgograd-Akhtubinsk) calculated from Landsat imagery data was 37%. Local reduction of wetland areas at test polygons varied from10 to 50% compared to such dry years as 2014, 1996,1984 and decreased up to 80-85% compared to wet 2007 and 2013. Vegetation peak was abnormaily shifted towards the flood peak. Open water areas and soil moisture changes indicate that most of permanent water bodies or marshes distant from Akhtuba completely dried up in 2015. So wetland ecosystems were replaced by bare soils or ruderal plant communities. The magnitude of wetland damage decreases downstream. Field data revealed that large water bodies dry-up preceded by its shallowing up to 0,4-1,5m deep in last years, indicate an active sludge accumulation at the bottom and confirm the decrease of open water areas.

We conclude that the main backgrounds of the ecological catastrophe are as follows:

Maximum HEP discharge during the flood 2015 was the lowest over the entire period of regulated Volga runoff (16000 m3/s instead of usual approx.26000 m3/s).The long-term low flow period in Volga and Akhtuba preceded to 2015 as all other dry years such as 2006, 1996, 1984 and 1975 occurred within higher flow periods. They were also foregone by wet years 2005, 1995, 1983, 1974. But 2015 was preceded by dry 2014. 

This catastrophe also caused the significant decrease of flooded areas at flood peak in 2015 (less by 70-80%) compared to 2013. The flooding period 2015 was also short (about 3 weeks) as its normal length is at least 4 weeks. All these factors leaded to the breakdown or degradation of many wetland ecosystems in upper VAF segment and severe ecosystem damage in its downstream part.

Sentinel-2 (S2), a new operational Earth observing satellite mission of ESA, was launched in June 2015. S2 offers unique capabilities for temporal monitoring of vegetation characteristics, as it provides several spectral bands of high spatial resolutions, especially in the red edge region of the electromagnetic spectrum (i.e. 675–760 nm). A number of scientific studies conducted in last three decades has indicated that the red edge reflectance is particularly sensitive to changes in plant eco-physiological parameters such as leaf area index (LAI) and leaf chlorophyll a+b content (Cab). Therefore, objective of this study was to assess the potential of the red edge spectral bands and the red edge inflection point (REP) computed from the upcoming Sentinel-2 multispectral data for quantitative estimations of Cab and LAI.

We used Discrete Anisotropic Radiative Transfer (DART) model to simulate spectral databases of top-of-canopy bidirectional reflectance factor (BRF) of four vegetation types with an increasing structural complexity: i) a maize agricultural field, ii) an orchard of citrus trees, iii) a beech forest, and iv) a spruce forest stand. In the first phase of the study, we applied a machine learning approach called support vector regression (SVR) on the simulated spectral databases of all four vegetation types to investigate contribution of individual S2 spectral bands and to test importance of reconstructed REP in retrievals of LAI, Cab and canopy chlorophyll content (CCC; defined as product of LAI*Cab). In the second phase, we applied the SVR retrieval machines trained on DART BRF simulations on S2 data that were simulated from existing airborne hyperspectral images. These images were for the maize site acquired with a CASI instrument (Itres, Ltd., Canada) during the SEN3EXP/ESA campaign, and for the beech and spruce sites with an AISA Eagle instrument (Specim, Ltd., Finland) in frame of CzechGlobe field/airborne campaigns.

SVR outcomes of the first phase using DART simulated BRF data demonstrated that the single REP wavelength retrieved from reconstructed red edge S2 vegetation reflectance is important for retrieval of CCC, slightly less important for estimation of Cab and rather unimportant for retrieval of LAI. Despite of this, retrievals of all three variables performed with and without presence of REP were statistically indifferent. Unlike the single REP wavelength, contribution of spectral bands located in the red edge region was found to be crucial for retrievals of all three parameters of interest, especially for Cab and CCC. In the second phase, retrievals were applied on S2 data simulated from real airborne hyperspectral images. Pattern of S2-based results across the tested vegetation types was rather inconsistent. Since we had no suitable field data for direct validation of the S2 estimates, the only way of an indirect validation was the comparison with similar retrievals obtained from the original airborne hyperspectral images. The S2 Cab predictions were in general overestimated, whereas the S2 LAI values were underestimated when compared to the results from airborne data. Regarding the evaluated ecosystem types, the least reliable results were obtained for the architecturally most complex spruce forest, which seems to still represent a challenge for radiative transfer modelling. Although the retrieval results from simulated S2 images were less encouraging, they do not explicitly imply that S2 data are not suitable for Cab and LAI mapping. Since our DART spectral simulations of S2 data provided reasonable Cab and LAI outputs, we incline to conclude that the S2 image simulations created from airborne data suffered from residual reflectance inaccuracies causing the mismatch with the correctly validated airborne maps.

Estimates of global forest vegetation carbon underlie the global carbon budget. The accuracy of these estimates is challenged by uncertainties in spatial and temporal patterns of processes critical to the budget. These include sequestration and respiration rates, heterogeneous forest distribution, and increasingly complex anthropogenic land use and management practices. Forest carbon stocks are among many resources that can be estimated indirectly at broad spatial scales from remote sensing by utilising empirical relationships with key biophysical variables. Extrapolation of carbon estimates from site-specific plot observations to regional and continental scales using remote sensing allow for the generation of accurate and detailed baseline estimates of forest carbon, distribution, and structure. By exploring the relationships among remotely sensed and observed data for resources of interest, change can be systematically assessed over time using robust and repeatable models to apply at multiple time steps.

This study focuses on a case study in south-eastern Australian temperate forests. The aim was to examine how forest structural information derived from airborne light detection and ranging (lidar) data and spatial environmental variables can be used to scale limited field observations of forest carbon to a regional scale. This study developed novel allometrics for the estimation of biomass in the region. It also provides improved estimates of carbon among multiple forest components addressing uncertainties of within plot level carbon estimates for the study area. A novel approach to analysing lidar data was also explored. Lidar data was reconfigured using hyperspectral remote sensing pre-processing approaches to maximise information from the full vertical structural profile. The result was a carbon prediction model and map for the area that accounts for the characterisation of vegetation density, structure, and indirectly, composition. The structural complexity identified by lidar resulted in a higher spatial resolution carbon prediction model for the study area compared to existing regional and continental predictions.

Future work in this area will use Australia as an ideal case study to expand the regional approach to a continental scale. It has a broad range of ecoregions, from arid shrublands to tropical rainforests, and associated carbon stock variability. Recently developed functional type specific allometrics from vegetation across Australia can be used to develop empirical relationships with variables from Global Ecosystems Dynamics Investigation (GEDI) lidar data. This may allow for systematic forest carbon estimates to be scaled from the plot to continental scale. Established empirical relationships between lidar and individual vegetation carbon could allow for repeat continental, or potentially global, assessment of forest carbon with each satellite acquisition.

Abstract:

Rapid Urbanisation and urban growth is one of the crucial issues of the global change in the 21st century affecting the physical dimensions of the city. This is increasing, and is one of the most powerful and visible anthropogenic force that has brought about changes in land use land cover (LULC). Major reason for urbanisation is the rapid growing population and economic development which is leading to different environmental issues. The study of the impact of urban growth can be quantified using various factors- physical, environmental, economic or social. In this study, the spatio-temporal patterns and process of urban growth of Pune, is investigated from 1999 to 2013 by using GIS/RS and statistical analysis.
Three different land cover maps derived from Landsat TM image of 1999, 2006 and 2013, were used to understand the change in the LULC in the study area. Apart from the LULC change, the study also considers Demography, Environmental (ground water aquifer depth, water quality and air quality) and Climate (Temperature and Rainfall). Firstly, the seven parameters were individually analysed using remote sensing GIS/RS and graphical representation. Secondly, the Landsat images were analysed; using the change detection technique the change was identified and quantified between 1999-2006 and 2006-2013 images. Finally the changes in LULC were correlated with the changes in the other seven parameters and the potential urban area had been delineated. Using Geographically Weighted Regression (GWR) the driving forces for the changes in the peripheral areas were identified.

The major challenges of urbanisation in India are unprecedented in scale and significance. It can never be understood by the proportion of population in India, and the lack of social and physical infrastructure required catering to the needs of the population. The fast growing metropolitan cities in India have contributed negatively in the development process through different issues.
The rapid and haphazard growth of major cities in India generates numerous problems in the country. The growth of the cities normally extends more to the fringe areas and it creates the unplanned development of the cities. The development of the cities in the concentric pattern or radial pattern had raised the challenges for planning of the cities. Pune, the cultural and education centre of Maharashtra, is also facing these problems. In the same context, because of the imbalance land use pattern, Pune city is facing problems of unequal distribution of physical and social infrastructure in the city.
The research aims to study and quantify the impact on urban development in the peri-urban areas of Pune city considering various development parameters using GIS and remote sensing techniques.

This communication presents an applying of multi-temporal Landsat TM/ETM+ images of vegetation cover for its temporal change detection. Data were collected for the last 20 years (15 scenes) for the Minusinsk intermountain Hollow in the Altai-Sayan Mts, southernSiberia. Vegetation indices NDVI and SWVI were calculated for an each scene to obtain time-series of these indices. An image differencing algorithm was applied to detect changes in the vegetation distribution. The results were interpreted as both temporal land use change and consequences of climate change. 

Climate variability and associated extreme weather events such as drought is an important aspect of the climate system that affects society and our economy. New science is emerging on the probabilistic increase of climate extremes under a warmer Earth. Increasing variability is likely to profoundly affect ecosystems, as many ecological processes are more sensitive to climate extremes than to changes in the mean states. The recent extreme climate events of El Niño in 1997 and La Niña in 2010-11 illustrate the significant consequences of climatic extremes for the terrestrial biosphere, and some of the negative and positive impacts on Australia and globally. However, the response of vegetation to these changes is one of the largest uncertainties in projecting future climate, carbon sequestration, and water resources. This remains a major limitation for long term climate prediction models integrating vegetation dynamics that are crucial for modelling the interplay of water, carbon and radiation fluxes. Satellite remote sensing data, such as that from the MODIS, Landsat and Sentinel missions, are the only viable means to study national and global vegetation trends. Highly accurate in-situ data is critical to better understand and validate our satellite products.

Here, we developed a fully automated hyperspectral monitoring system installed on a flux monitoring tower at a mature Eucalypt forest site. The monitoring system is designed to provide a long-term (May 2014 - ongoing) and high temporal characterisation (3 acquisitions per day) of the proximal forest canopy to an unprecedented level of detail. The system comprises four main instruments: a thermal imaging camera and hyperspectral line camera (spectral ranges 7.5-14 µm and 400-1000 nm, respectively), an upward pointing spectrometer (350-1000 nm), and hemispherical camera. The time series of hyperspectral and thermal imagery and flux tower data provides a unique dataset to study the impacts of logging, nutrient, and heat stress on trees and forest. Specifically, the monitoring system can be used to derive a range of physiological and structural indices that are also derived by satellites, such as PRI, TCARI/OSAVI, and NDVI. The Tumbarumba site is part of the Australian SuperSite network spanning a wide range of environmental conditions. Each site located in a significant biome. The network collects a vast assortment of coincident in-situ, air-, and space-borne data in a nationally consistent manner, enabling a better understanding of how key ecosystems will respond to future environmental change.

The monitoring system, to our knowledge, is the first fully automated data acquisition system that allows for spatially resolved measurements of PRI, thus enabling us to test the hypothesis that PRI is tree or canopy height dependent. Preliminary results indicate that this is the case. Further analysis will help establish if a relationship exists between photosynthetic rate and light use efficiency dependent upon canopy position and height. However, challenges remain to properly interpret the PRI values. We found due to the similar spectral characteristics of different species, between-tree variability of indices is largely influenced by structural effects. Therefore, accounting for different structural and illumination conditions is integral for future work interpreting and scaling these findings. The monitoring system will also be used to enhance our understanding of BRDF effects on the highly sensitive PRI values among other indices. Panoramic hyperspectral images capturing a range of inclination angles at different times of day and year will be used for this purpose. The ultimate aim of this work is to significantly advance our understanding of the impacts of lagged climate effects on vegetation by assimilating relevant remotely sensed data streams into a dynamic-vegetation-enabled land surface model (CABLE) at the regional, continental and global scale.

All countries with different agriculture culture in the world especially in the developing countries need reliable techniques which can help in predicting ahead the crop outcome in order to increase production, to meet the demand for food and to reduce the import of crops.  In this research, remote sensing and other technologies including software such as METRIC which is based on the equation of energy balance are used to map evapotranspiration, potato crop and to  estimate crop yield. A private potato industry has planted a large area in the valley. The above ground biomass of this area is estimated using remote sensing and two different techniques one is based on the energy balance equation and the another based on extreme temprature diffrences. The outcomes of the two techniques are verified using an installed Bowen ratio station and the result showed that energy balance is more reliable to estimate crop yield. The comparison between the estimated potato crop yield and the actual production showed about 98% agreement.  

After climate change and global security, flooding has become a major issue that concerns most of international organizations. Ranked among the deadliest natural disasters in the world, flooding affects nearly the five continents such as in Africa (Madagascar, january 2015), in Australia (january 2011), in Asia (China, 2002), in Europe (Germany, 2000) and in America (Canada, 1996). The risk management problem during the flooding period is based mainly in the difficulty of data collecting on ground. This is due to the inaccessibility in certain areas during the flooded period. To overcome this problem, we have used satellite imagery to deduce information about the concerned areas. In this present study, we have used RADAR images, since they offer spatial observations of floods with high spatial resolution. In addition, the quality of the acquired images is not affected by cloudiness since flooding is mostly accompanied with a sky significantly cloudy covered. We have experimented the techniques and methods used for the detection of flooded areas from the RADAR Sentinel-1A images, using the steps in the "flood mapping" algorithm, which is predefined in the WOIS (Water Observation and Information System) open software developed by ESA (European Space Agency). This technique is combined with the change detection approach which is based in two images acquired in different dates in order to distinguish the permanent waters from flooding areas. The choice of this technique is due to the objective of the study that is the delineation of flooded areas and the establishment of the areas map that are prone to be flooded. The processing flow chart is divided into four blocks such as the pretreatment steps including geometric correction and filtering, classification, delimitation of risk areas and the separation of permanent water and flooded areas. For the application and use of these processing systems, we have chosen for study area the region of Analamanga, Madagascar as this region was one of the victims of the passage of the CHEZDA hurricane in january 2015. In order to delineate the flooded areas, we have used two Sentinel-1A images, one in flooded period (January 22^nd, 2015), and one in a dry season (May 16^th, 2015). These RADAR datas were acquired in an IWS (Interferometric Wide Swath) mode in the C-band wavelength of 5.6 cm, with a spatial resolution of 5 m by 20 m and a swath length of 250 km. We have used a DEM image with 30m x 30m of spatial resolution for geometric correction of our images. This has served also for the determination of the drainage network, the direction of drainage, watersheds, the length of slopes, slope and HAND (Height Above the Nearest Drainage) parameter that are all essential for delineating risk areas during the flood period. The results we have obtained map properly the flooded areas in this region.

A novel methodology for tropical forest monitoring based on InSAR is developed. The idea is to monitor forest height changes. Height decreases result from logging and represent a decrease in biomass and carbon stock. Height increases result from tree growth and increasing biomass and carbon stocks. The method is developed for monitoring in REDD+. A Reference Emission Level (REL) is derived from height changes between the SRTM in 2000 and Tandem-X around 2012. Careful error removals and corrections for wavelength specific penetration differences are included. Annual, or bi-annual, coaverages with Tandem-X or similar satellite missions in the coming years will provide the MRV data to be compared to the REL. The method is based on several research studies published in peer-reviewed journals. The company Forest Vision is established in order to set the method into operational use. By now we have processed Uganda wall-to-wall and obtained a REL. The mean height change in Uganda during 2000 - 2012 was a decrease of 31 cm, corresponding to an annual decrease of 2.6 cm. This is further recalculated to an estimated CO2 emission estimate of 20.7 million tons.

The FLEX/Sentinel 3 tandem mission will provide for the first time vegetation fluorescence in the oxygen A and B band at a high spatial resolution of 300m. On the basis of sun-induced fluorescence (SIF) as level 2 products, FLEX ultimately aims at delivering higher level products such as photosynthesis rate, vegetation stress or gross primary production (GPP).  In the processing chain that relies fluorescence radiances to photosynthetic products, fluorescence quantum efficiency (FQE) is a key parameter representative of the link between optical variables and plant physiology. FQE is defined as the ratio between the number of photons emitted by fluorescence and the absorbed photons.

However, fluorescence radiance as provided by level -2 FLEX products is not an accurate proxy of FQE, because of its interference with canopy structure and illumination or viewing geometry.

Recent studies in the framework of the FLEX preparatory activities have shown that FQE can be retrieved from the combination of two-peaks SIF and reflectance data using statistical model inversion techniques.  Though promising, these techniques need experimental validation at the canopy level to be implemented in the FLEX processing scheme.

Active fluorescence measurements provide a direct way to assess FQE variability at the canopy level because illumination and viewing geometries are well characterized and can be kept constant.

In this study, we present a dual sensor approach that combines simultaneous measurements of SIF and laser-induced fluorescence (LIF). SIF can be conceptualized as the product of FQE and two other factors: the absorbed photosynthetically radiation (APAR) and a transmission factor (TF) that results from the radiative transfer of fluorescence through the canopy. In LIF measurements, these two later factors can be kept constant, thus LIF give access to FQE and the combination of SIF and LIF can be used as an experimental validation of FQE retrieval algorithms. On the other hand, gross primary production (GPP) can be expressed as the product of APAR and light use efficiency (LUE). In this perspective, the SIF/LIF combination brings valuable information to better characterize the relationship between LUE and FQE at canopy level for integrated fluorescence and photosynthesis models like SCOPE.

We described here two new active sensors: (i) a ground sensor that uses electroluminescent diodes to monitor fluorescence at distances up to 10 meters, and (ii) an airborne LIF sensor that measures the two peaks of fluorescence and the vertical structure of the canopy by waveform analysis.

The ground sensor can capture the full diurnal cycle of FQE including the night level (F0) at a high sampling rate, thus allowing the analysis of the fluorescence dynamic over a large range of temporal scales from minutes to months. The airborne sensor allows the scaling up of FQE at larger spatial scales including the footprint of eddy fluxes towers, the FLEX pixel size (300 m) or landscape scale.

We discuss the potential benefit of the synergy between ground and airborne sensors in the framework of FQE calibration and validation.

Today’s available large-scale forest area and biomass products are primarily based on optical data. However, due to the increasing availability of high resolution synthetic aperture radar (SAR) data, numerous studies have also been dedicated to SAR-based forest mapping and monitoring. Promising results were obtained using SAR polarimetry, interferometry or radargrammetry. Also SAR backscatter itself was found to be sensitive to biomass changes. However, one important limiting factor to using backscatter data is the frequency dependent saturation level; C-Band backscatter data have often been found to saturate at about 100 to 300 m³/ha. Nonetheless, the important advantage of SAR backscatter at C-Band is the relatively high available spatial and temporal coverage from previous (Envisat ASAR) as well as current (Sentinel-1) sensors. Furthermore, hyper-temporal combination of backscatter measurements was found to improve the results of vegetation related parameters retrievals over forests and was successfully used to derive the growing stock volume (GSV) from Envisat ASAR at a coarse resolution [1].

Within the first year of Sentinel-1 data availability, high temporal coverage of interferometric wide swath mode data over Europe became available. Thus, the objective of this study is to analyse the backscatter behaviour in V/V and V/H polarization over forested areas within a growing season. A test area in the central part of the Austrian’s federal state Burgenland with an extent of about 400 km² was selected for this study. The forests in this area consist of different species of coniferous and deciduous trees. According to the national forest inventory, the average GSV of Burgenland is 256 m³/ha with values up to 650 m³/ha within the test area. As a reference, small-footprint, full-waveform LiDAR data were collected over the region in April 2010, under leaf-off conditions. The LiDAR data were processed in Opals software [2] and parameters describing the forest structure were derived. These include e.g. canopy density, canopy height or a so-called vegetation-amplitude portion. The latest was calculated as an amplitude sum of the non-ground echoes normalized by amplitude sum of all echoes (ground and non-ground) within 2 by 2m cells. The raster values close to 0 indicate open-land areas or vegetation which is highly penetrable (e.g. sparse and deciduous forest), whereas the values close to 1 indicate less-penetrable vegetation (e.g. coniferous forest).

All available Sentinel-1 data were processed over the region, resulting into a coverage of about 90 images within the first year (10/2014 to 09/2015). Statistical parameters as well as monthly averages were derived to analyse the variation of backscatter in both polarizations. The comparison with LiDAR data revealed differences in backscatter temporal variations as well as differences in the statistical parameters for various forest types, as characterized by LiDAR parameters. Especially the ratio between average backscatter over leaf-off (December to March) and leaf-on (Mai to August) conditions corresponded well to the LiDAR vegetation-amplitude portion raster (Pearson correlation coefficient of 0.63 for spatial resolution of 100m) and reflected the variability of forest structure in different areas.

The results demonstrate the influence of the different forest structure on C-Band SAR backscatter in both polarizations as well as a strong variation of backscatter within a year, especially in case of the VH polarization.

[1] Santoro, M., Cartus, O., Fransson, J.E., Shvidenko, A., McCallum, I., Hall, R.J., Beaudoin, A., Beer, C., Schmullius, C., 2013. Estimates of Forest Growing Stock Volume for Sweden, Central Siberia, and Québec Using Envisat Advanced Synthetic Aperture Radar Backscatter Data. Remote Sens., 5: 4503-4532.

[2] Pfeifer, N., Mandlburger, G., Otepka, J., Karel, W., 2014. OPALS – A framework for Airborne Laser Scanning data analysis. Computers, Environment and Urban Systems 45, 125-136.

Soil moisture is very important for estimates the vegetation growing conditions. Backscattering coefficient - sigma (σ˚) calculated from Sentinel1 images was acquired and analyzed along with ground truth (in-situ) measurements. At the same time NDVI  has been obtained from SPOT5 for wetlands area. Biomass, LAI, has been measured  for agriculture and wetlands habitats. It was examined the differences in backscatter values during some period of time and the differences in soil moisture and  differences in NDVI and biomass.  For grasslands the difference in backscatter has been related to the differences in soil moisture as it was assumed that the surface roughness has not  changed. For dry conditions when the soil moisture did not varied , the difference in backscatter corresponded with the difference of phenological development stage of vegetation. It was also examined the impact of soil moisture on backscatter and the impact of biomass expressed in vegetation indices on backscatter. The relationship between LAI and σ˚ was adversely proportional when the soil was covered by short vegetation, after the threshold the vegetation index was proportional  to the backscatter. Vegetation attenuates the signal until some amount of biomass (varied for different vegetation). The models have been elaborated  for analyses of the relationship between backscatter and soil moisture under different density of vegetation.

In order to reveal the backscattering coefficients sensitivity to the soil moisture and to the canopy descriptor (as LAI, NDVI ), the gradient of the function σ^o of the two independent variables - soil moisture and canopy descriptor – was calculated.

The polarization of VV and VH has been considered and examined. Soil moisture varied for these two test sites : soil moisture at wetlands was rather high with the exception of 2015 while soil moisture for the agriculture field was much lower, changing during the period of the measurements. Soil moisture dominated vegetation contribution to backscatter  in C band.  It was also shown up that with the increase of canopy cover, the sensitivity of radar signal to dry soil conditions was low. Due to possibilities of obtaining SPOT5 and information from SWIR band, it was possible to examine the relationship between backscatter and vegetation moisture.

Sentinel1 and Landsat8 has been applied for  distinguishing the crop types. During the S-1 acquisition the ground measurements has been carried out. The research has been done for the areas of northern-eastern test site of wetlands and western Poland for agriculture .

From the passive L-band microwave radiometer onboard the Soil Moisture and Ocean Salinity (SMOS) space mission global surface soil moisture data is retrieved every 3 days. Thus far, the empirical L-band Microwave Emission of the Biosphere (L-MEB) radiative transfer model applied in the SMOS soil moisture retrieval algorithm is exclusively calibrated over test sites in dry and temperate climate zones and the included dielectric mixing model relating soil moisture to permittivity accounts only for mineral soils.However, soil moisture monitoring over the higher northern latitudes is crucial since these regions are especially sensitive to climate change and a considerable feedback is expected due to carbon liberated from thawing ground of these extremely organic soils. Due to differing structural characteristics and thus varying bound water fractions, the permittivity of organic material is lower than the one of most mineral soils at a given water content. This assumption was verified by means of measurements in organic and mineral substrates from various sites in Denmark, Finland, Scotland and Siberia. For this purpose, conventional soil moisture sensors were used as well as weak perturbation and waveguide techniques in order to infer effective soil permittivity at the microwave L-band (1-2 GHz). Based on these data, a generic L-band soil moisture – permittivity relation for organic soils was derived and validated with dielectric mixing model runs as well as literature data. Furthermore, the derived function was tested in the L-MEB model. Results showed that modeled data agreed with measurements from a tower-based passive L-band microwave radiometer observing organic-rich soil over a 2 months period in a highly controlled set-up. The generic «organic» empirical model was then implemented in the SMOS Prototype Algorithm to retrieve soil moisture over a site in Northern Finland. The validation with in situ soil moisture observations calibrated for organic soils showed a distinct improvement in the agreement between the satellite and ground datasets when using the «organic» instead of the operational SMOS processor version. This analysis is to be continued in more detail and the validation effort needs to be expanded over as many regions with abundant soil organic matter content as possible. Appropriate in situ observations are currently available from various sites in Alaska, Canada, and the Netherlands. In this communication, first the derivation of the generic L-band «organic» soil moisture-permittivity model will be presented. Focus will then be on the comparison of «organic» SMOS soil moisture retrievals with corresponding operational SMOS products as well as in situ observations over all available sites.

The actual yield that can be harvested at any given location is limited by several factors, most notably among them water and nutrient availability. For many questions, both societal (“feeding 9 billion”) as well as local (“what is the target yield I should aim for when calculating my nitrogen fertilizer inputs?”), knowing what is currently produced on a field is not enough, though. In these cases, it is necessary to calculate the so-called yield potential, which is defined as the yield that can be reached with optimal nutrient supply and under given circumstances, such as longitude and latitude, soil properties and climate conditions.

Within the EU FP7 project MELODIES (Grant Agreement 603525), VISTA is developing an open data and EO-based method to determine this yield potential as part of a service for decision support in land management.

For this, in a first step, time-series of high-resolution optical EO data are used to extract persistent structure in biomass for each field. Via the use of multi-year data and a dedicated geostatistical method, patterns caused by the specific management of one year are excluded from the results, leaving the patterns caused by the soil properties as result. These patterns are relative values per field, though, and not yet crop type specific.

Thus, in a second step the absolute values for the yield potential are modelled using the crop production model PROMET. This physically-based, raster-oriented model calculates the allocation of carbon into the different plant compartments (roots, leaves, stem, fruit) depending on the location (longitude, latitude, exposition, soil properties, climate conditions) for the whole growing season. Boundary conditions can be set to allow modelling of either actual yield or assume optimal nutrient supply and /or no water stress for the calculation of yield potential. To capture the climate conditions and not just the meteorological conditions of one year, the model is run for at least ten years and the results are averaged.  

Usually, the input data to capture the heterogeneity of the fields in the model runs, are not globally available in the very high resolution needed to capture the full extent of the variation. Thus, the assimilation of the high resolution long-term biomass structures derived from EO into the modelling, is a globally applicable way of allowing a high resolution spatially distributed calculation of the yield potential. The results are validated against actual yield information and the German soil inventory.

One concrete example where the knowledge of yield potential supports land management decisions is the implementation of the new regulation called “Greening” within the EU Common Agriculture Policy (CAP). “Greening” means that every farm in the EU has to take areas out of production to set them aside as so-called “ecological focus areas”. VISTA is using the yield potential as one input for the development of a support tool for farmers that takes into account both the economic potential of a given area, as well as its proximity to different ecologically valuable resources (e.g. rivers, forests, FFH areas), to decide which of his fields (or parts of fields) are best taken out of production for both economic as well as ecological reasons. A prototype of this service will be shown at the symposium.

This EO-data-based service can directly support a modern, data-driven and resource-efficient agriculture which in turn is one step towards using our global resources in a sustainable way, both for the preservation of nature and towards the goal of future food security.

One of the key parameters for the description of forest ecosystems is the tree species.  Detailed and regularly updated information about the distribution of tree species is important for conservation issues, forest management and assessment of wildlife habitats. The changing growth conditions caused by climate change further increase the relevance of information about species occurrence.

Remote sensing data can provide suitable information for large areas with the possibility for regular updates. In this study we compare the potential of the new Sentinel-2 data (10 - 20m) and two common satellite sensors with both, higher (WorldView-2) and lower spatial resolution (Landsat-8). The Sentinel-2 data, as well as Landsat-8 (30m) and WorldView-2 (2m) scenes were acquired in summer (near vegetation peak) under cloudless conditions (Figure 1).

For the analysis all data sets were atmospherically corrected. Using this data set we focused on the spectral separability of Central European tree species based on mono-temporal satellite data. An object-based approach at forest stand level together with a Random Forest (RF) classification was applied to assess tree species separability.  Our results clearly demonstrate the potential of Sentinel-2 for mapping tree species

Introduction. Approximately 30-40 million hectares of arable land has been abandoned over the past two decades in Russia [Agroecological state and prospects of the use of lands retired from active agriculture in Russia, 2008]. At present, these lands undergo changing in response to regional peculiarities and abandonment duration. The assessment of fallow lands is of importance in identifying the activities on reclamation and melioration of these lands when converting them again into agricultural use. The main object of our study is finding a way of satellite optical and radar images combined application for abandoned agricultural lands mapping.

Test site. The Volga Delta is a region of a special interest. It is located in a semi-arid climate zone, with rather fertile soils, but success of the agriculture here is highly dependent on spring flooding and water availability. Since the 1960s, an irrigated crops and vegetables production based on artificial irrigation has become the major agricultural sector and achieved its peak development by the mid-1980s. Both environmental and socio-economic factors caused the suspension of the agricultural activity and irrigation by now, so a considerable part of the arable lands, protected from flooding and watering by banks, turned out to be exposed to degradation processes i.e. shrub invasion, salinization, pasture digression, halophytic vegetation spreading.

Data used. Russian topographic maps of the late 70's-early 80's and archive of Landsat images obtained from 1984 till present were used for mapping and studying changes in agricultural land use at the delta. Multi-temporal polarimetric Radarsat-2 images obtained in summer 2014 in conjunction with synchronous ground truth observations of the site were used for evaluation of fallow lands current state. The ground truth data is particularly useful in evaluating the radar data processing results accuracy.

Methods. A vector mask of agricultural fields created from topographic maps fixed the maximal areas of arable lands for the early 1980’s. Land use/land cover (LULC) patterns for each representative year (1984, 1993, 2002, 2008, 2014) were created using Landsat multi-seasonal composites. The period of fallow lands abandonment was determined basing on multi-year LULC patterns comparison.

SAR data were expected to be helpful for assessing the state of abandoned agricultural lands owing to sensitivity of radio waves to geometric features and moisture content of a terrain. A series of experiments on selecting the optimal radar data type and processing parameters revealed that polarimetric data have significant advantages over single-polarized data for the examined region. The polarimetric data let 93% accuracy in discriminating overgrown and not overgrown fallows while HV-polarized image – 67%. The following processing chain gave the most accurate result: polarimetric decomposition (Yamaguchi) followed by supervised classification (Support Vector Machine with 3 classes extraction) of volume scattering component expressed in decibels.

Results and discussions. The main result is a map of the Volga delta abandoned agricultural lands where for each fallow field period of abandonment and overgrowth level is determined. A comparison of the radar data processing results and map based on optical data shows that there is no a distinct correspondence between fallow lands overgrowth degree and duration of their abandonment period. Apparently, environmental features (such as water availability, soil type and etc.) have a much greater impact on fallow lands state.

Acknowledgments. The authors thank the specialists of VNIIOB for their support in field works organization. Landsat data was downloaded from USGS archive [http://earthexplorer.usgs.gov/]. MDA and CSA provided the Radarsat-2 data in the framework of SOAR-EI project #5137.

SMOS has been delivering data since January 2010. Three ground segments have been developed among which the CATDS (Centre Aval de Traitements des Données SMOS) by the French space agency CNES. This ground segment is providing the users with level 3 data that are a bit different from ESA level 1 and level 2 as these products : i) are temporally aggregated as daily, during 3-day and 10-day time window and monthly averages ii) are projected on the EASE Grid version 2 iii) using the netcdf format iv) are derived using a modified algorithm taking advantage of 3 consecutive SMOS overpassesv) brightness temperatures are averaged by bin of incidence angles in the Earth reference frame, i.e. H and V polarizations.

Since the beginning of the mission the algorithm to derive the SMOS data has evolved due to improvements of the models and also due to better quality SMOS brightness temperatures. A reprocessing campaign has been done in fall 2015 to deliver the entire time series (2010-spring 2015) using the latest versions of the processors and with a more accurate brightness temperatures (improved bias reconstruction).

The data are available on line (http://www.catds.fr/sipad/startPage.do) through the SIPAD (Système d’Information, de Préservation et d’Accès aux Données) which is a user-friendly dissemination service that allows the users to download their own products according to their criteria (as area of interest, period of time, subset of fields...).

The aim of this communication is then to present the last updates and version of the products delivered by the CATDS that is the Version 3. Derived soil moisture values are validated by comparing to SMOS level 2 and in situ measurements acquired at various climate conditions.

High resolution soil moisture is critically needed for the improvement of forecasts and early warning of a variety of plagues. SMELLS is an ESA Innovators-III project to develop innovative EO products and services in response to authoritative requirements from end-user organisations.

Specifically, this project addresses the use of soil moisture to preventive management of Desert locust Schistocerca gregaria, whose plagues have threatened agricultural production in Africa, the Middle East and Asia for centuries and regularly affect up to one-tenth of the world’s human population. Preventive management aims to prevent or to limit crop damage by controlling populations before they can reach high densities and form mass migrating swarms.

SMELLS implements an innovative approach to combine Sentinel-1 SAR data with thermal disaggregated SMOS-derived soil moisture to derive a soil moisture product at both high-spatial and high-temporal resolution to provide a new tool for decision-makers in the Desert locust preventive control system.

In this presentation we will show the validation results based on SMOS-derived soil moisture at 1 km.

Sentinel-2A (S2A) will offer an unprecedented data stream for the establishment of global geophysical products on high temporal and spatial scale. Red edge vegetation indices derived from simulated S2A images have proven to boost accuracy in estimating Leaf Area Index (LAI) for agricultural crops. However, the capabilities to improve LAI estimates with red edge bands for forest ecosystems remain untested. This study aims at doing so with three independent approaches: (1) testing sensitivity of red edge indices with canopy radiative transfer models (RTM), including dedicated forest RTMs as well as comparing S2A derived red edge indices with (2) established global LAI products and (3) ground truth data collected during an intense campaign in an old-growth beech forest in the Netherlands. The results will help to gauge the added value of S2A red edge bands in high spatial resolution LAI retrievals.

Mapping forests is an important process in forest management. The complexity of the terrain and the characteristics and nature of the forest define the degree of difficulty to achieve this process. At present and due to spectral resolution limitations, multi-spectral satellite images are limited in separating between different forest species. In contrary, advances in remote sensing technology have provided hyperspectral images as a solution for accurate determination of forest species. Two different satellite images are used to map different areas in Lebanon. The first one is based on processing an archived hyperspectral Hyperion image provided by National Aeronautics and Space Administration (NASA) for specific year, and the second one is a Hyperspectral image (CHRIS-Proba) provided by European Space Agency (ESA).  ASD Spectroradiometer is used in the classification process in order to increase the accuracy and the efficiency of the results. Several issues related to the use of hyperspectral images are investigated and analyzed such as spectroradiometer recorded spectral signatures resampling, atmospheric correction and pre-processing issues associated with the quality enhancement of the images. The results of classification and verification of the hyperspectral images for two different areas show that the stone pine forest map created using CHRIS-Proba image has an accuracy of 82% and it shows that stone pine forest map created using Hyperion has an accuracy of 92%. 

Food insecurity has been a topic of concern for a long time particularly for the developing countries. Population increase and changing climate have been cited to be the common culprits affecting increased food production. The potential of utilizing wetlands as agricultural production zones is being investigated. The choice of carrying out the research in the wetlands is driven by the fact that it has consistent water availability throughout the year. The study area selected is the Ifakara wetlands in Tanzania. The main food produced in the region is rain fed rice during the long rains falling between March and May. In the short rain season between November to January, maize is the preferred crop.     In order to determine whether the utilization of the wetland for increased food production is viable, there was a need to analyze the land uses in this region in different months of the year to better understand the dynamics of existing vegetation.

It is with this background that the current research is undertaken. Remote sensing imagery, RapidEye images spanning from 2013 to 2015 were acquired and used in the research. Support Vector Machine was used to classify a series of RapidEye images covering the area to establish the dynamics of changing vegetation all year round. The land cover was consistent with the seasonal changes. Bare land coverage gives an indication of the potentially available land that can be utilized for crop growth. The region lies at the equator and experiences a heavy presence of cloud coverage and this was evident in the images classified. As a remedial action to cater for this short coming, Sentinel 1 radar images were also utilized to aid in giving a complete coverage of the land classes as they are not affected by the presence of clouds. However the classified radar images are subject to satellite passes since it was launched. From the results obtained a great percentage of land remains bare. Quantification of this from the classified images forms a basis upon which the agricultural ministry in the region can make decisions on developing a strategic plan of increasing production in the region sustainably.

Wetlands are one of the fastest declining ecosystem types worldwide, while at the same time they are hot spots of biodiversity and provide diverse and valuable ecosystem services; such as water supply, hydrological buffering against floods and droughts, and climate regulation through carbon storage. Information on wetlands extent, its ecological character and their services is often scattered, underestimated and difficult to find and access, which leads to the fact that wetlands are only partially covered worldwide by policies and management practices.

In this respect, SWOS (a Horizon-2020 project funded by the European Union) provides monitoring tools and information on wetland ecosystems, mainly derived from Earth Observation data.

One of the key mechanisms for making earth observation data available for wetland monitoring activities is the development of a SWOS software toolbox. Several optical and SAR-based data sources are used for the implementation of SWOS /preferable free available data) and will be implemented in the course of the project, such as Landsat (MSS to OLI) Sentinel-1/2/3, ERS, ENVISAT (MERIS, ASAR).

The toolbox is a further development of the ESA Globwetland-II toolbox and integrates processing functions for the main satellite-based wetland products of the SWOS project, namely: land use / land cover and change, surface wetness dynamics and indicators on the status and trends of the wetland ecological character. New SAR-based map products as well as newly developed mapping tools will be integrated or linked, such as soil moisture, water quality, surface temperature. One of the central features is the provision of a harmonized legend system according to the EU framework on Mapping and Assessment of Ecosystems and their Services (MAES).

The SWOS software toolbox will be published as freeware at the end of the project. The design of the toolbox and the corresponding modules supports scripting of the separate modules and an easy integration of third party software as well as an easy integration into third party software. A graphical user interface with a very user-friendly handling is part of the SWOS software toolbox. Therefore one of the used concepts is “keep it simple”.

All necessary modules for the processing from the prepared Geo-Tiff satellite data to the defined end products of the SWOS project will be included. For the integration of third party tools and the integration into third party software a simple concept with well-defined interfaces will be used to connect modules. The toolbox is part of the SWOS portal (http://swos-service.eu) and will contribute to a wetland-specific knowledge hub for users.

This paper provides an overview of the functions and implemented method of the SWOS toolbox. The different service lines of the project will be demonstrated through mapping examples and indicator reports for exemplary threatened Ramsar wetlands.

The goal of this study is to analyze the sensitivity of X-band SAR (TerraSAR-X) signals as a function of different physical bare soil parameters (soil moisture, soil roughness), and to demonstrate that it is possible to estimate of both soil moisture, roughness and texture from the same experimental campaign, using a single radar signal configuration (one incidence angle, one polarization). Discussions are based on experimental campaigns acquired on North Africa (Merguellil site, Tunisia) with ground measurements over more than fifteen test fields, simultaneously to seven TerraSAR-X images acquisitions at HH polarization and 36° incidence angle.

Firstly, we analyzed statistically the relationships between X-band SAR (TerraSAR-X) backscattering signals function of soil moisture and different roughness parameters (the root mean square height Hrms, the Zs parameter and the Zg parameter) at HH polarization and for an incidence angle about 36°, over a semi-arid site in Tunisia (North Africa). Results have shown a high sensitivity of real radar data to the two soil parameters: roughness and moisture. A linear relationship is obtained between volumetric soil moisture and radar signal. A logarithmic correlation is observed between backscattering coefficient and all roughness parameters. The highest dynamic sensitivity is obtained with Zg parameter. Then, we proposed to retrieve of both soil moisture and texture using these multi-temporal X-band SAR images. Our approach is based on the change detection method and combines the seven radar images with different continuous thetaprobe measurements. To estimate soil moisture from X-band SAR data, we analyzed statistically the sensitivity between radar measurements and ground soil moisture derived from permanent thetaprobe stations. Our approaches are applied over bare soil class identified from an optical image SPOT / HRV acquired in the same period of measurements. Results have shown linear relationship for the radar signals as a function of volumetric soil moisture with high sensitivity about 0.21 dB/vol%. For estimation of change in soil moisture, we considered two options: (1) roughness variations during the three-month radar acquisition campaigns were not accounted for; (2) a simple correction for temporal variations in roughness was included. The results reveal a small improvement in the estimation of soil moisture when a correction for temporal variations in roughness is introduced. Finally, by considering the estimated temporal dynamics of soil moisture, a methodology is proposed for the retrieval of clay and sand content (expressed as percentages) in soil. Two empirical relationships were established between the mean moisture values retrieved from the seven acquired radar images and the two soil texture components over 36 test fields. Validation of the proposed approach was carried out over a second set of 34 fields, showing that highly accurate clay estimations can be achieved.

 

Satellite data fusion is a one of techniques which is used in various studies related to land cover monitoring. Combining images of various spectral bands allows to increase the differentiation of objects on the Earth’s surface and thereby improve e.g. classification accuracy. One of examples of applications where the data obtained from various sensors are used is monitoring of agricultural lands.
In our study, we focused on crops recognition with optical and radar data. Our study main objective was to develop a workflow which allow to monitor crops types and their spatial pattern with overall accuracy higher than 80%.
The study area was located in the Żywiec Basin, a part of the Polish Carpathians. The advantage of this small agriculture region was that it is covered by a mosaic of different size cultivated fields (from 0,01 to 36 ha). We used Landsat 8 multispectral images from OLI sensor (7 June 2014) and Sentinel-1A SAR images (2 August 2014) here. In first step of our workflow we tested various methods of optical and radar data integration (e.g. IHS, PCA, Brovey, Ehlers). Then we applied different algorithms (e.g. SVM, Random Forest) on pixel- and object-based classification. Finally we assessed accuracy using information collected during the field survey (interviews with farmers).
The results demonstrated that use of integrated optical and radar data is effective for crops classification – the highest overall accuracy achieved in this study was equal to 87,9%. In five cases, we achieved overall accuracy of crops classification above 80%. In summary, it should be emphasised that a wide range of factors can greatly affect crops types maps accuracy – the choice of data and methodology is essential in such studies (e.g. polarization of SAR data, classification algorithm) and have to be considerate during the workflow preparation. The accuracy of classification is also influenced by the region characteristic - farmed plants, as well as fields size and their fragmentation.

Land surface temperature (LST) has a long heritage of satellite observations which have facilitated our understanding of land surface and climate change processes. These observations have been acquired from a variety of satellite instruments on platforms in both low-earth orbit and in geostationary orbit, with numerous products available for exploitation by the user community. Although many LST products in operational production have accompanying uncertainty fields, these are frequently derived following different principles with insufficient information to enable propagation to higher levels required in many applications, or for cross-comparison.

Within the framework of the EU Horizon2020 project EUSTACE (EU Surface Temperatures for All Corners of Earth) a new sensor independent method of estimating uncertainty for LST observations has been developed in support of the blending of in situ and satellite observations to generate spatially complete near-surface air temperature over all Earth surfaces. The method for LST uncertainty estimation is consistent with those for sea and ice surface temperatures also being developed within EUSTACE.

For each LST pixel three components of uncertainty are supplied, representing the uncertainty from effects whose errors have distinct correlation properties: random, whereby there are no correlation of error components between pixels; locally systematic, whereby there are correlation of error components between pixels within the correlation length of the property; and systematic, whereby there are correlation of error components between pixels on a large-scale. This three-component model is equally valid for any sensor from which LST is derived, and applies to all satellite processing levels. These LST products are also being made available to the LST community via the framework of the ESA DUE GlobTemperature Project.

The full potential of ALOS PALSAR L-band interferometric (InSAR) coherence data for the estimation of forest growing stock volume (GSV) in the boreal forest has rarely been investigated. Moreover, ALOS PALSAR backscatter and InSAR coherence have yet to be used together to delineate GSV. Due to the observation strategy and the high acquisition success rate over Eurasia, a large amount of high quality ALOS PALSAR L-band data is available over Siberia. Consequently, this paper investigates the capability of ALOS PALSAR backscatter and InSAR coherence for the estimation of GSV in Central Siberia, Russia. The potential of backscatter and coherence are directly compared using the same inventory data. Altogether, 87 PALSAR images are used and eleven forest inventory sites are investigated.

Based on this large dataset it was observed that InSAR coherence acquired in frozen conditions offers the highest potential for GSV estimation. The saturation level for single coherence images was on average 230 m³/ha, with an average R² between coherence and GSV of 0.58. PALSAR backscatter acquired in unfrozen conditions could also estimate GSV; however, the saturation levels (75‑100 m³/ha) and the average R² (0.42 - 0.48) were lower. HV backscatter offered only a slightly greater potential than HH backscatter.

A simple inversion approach aiming at the delineation of forest GSV maps based on the multitemporal SAR data was developed and applied to five forest inventory sites. This approach combines HV backscatter data acquired in unfrozen conditions and InSAR coherence data acquired in frozen conditions. In general, the produced maps feature a corrected relative RMSE_corr of < 30% which was similar to the accuracy of the forest inventory data. The R² between inventory data and SAR data based maps varied between 0.54 and 0.83.

CRUCIAL is an ESA/STSE funded project investigating innovative land and inland water applications from Cryosat-2 with a forward-look component to the future Sentinel-3 and Jason-CS/Sentinel-6 missions. The high along-track sampling and resolution of Cryosat-2 altimeter in SAR mode (18 KHz) offers the opportunity to recover high frequency signals over much of the Earth’s land surface, enhancing the inland water height retrieval capability. To perform this study we use the samples of SAR Full Bit Rate (FBR) data from Cryosat-2 acquired over a few of these land surfaces; however, for Sentinel-3 the SAR mode will be deployed widely over land. This paper will summarise the CRUCIAL aims and objectives and present the theoretical approach to analysis of the FBR L1A Doppler beams to form a product using ground cell gridding, beam steering and beam stacking from which inland water heights are derivable from the retracked Cryosat-2 altimetric waveforms.  Results over the Mekong, Brahmaputra and Amazon will use the along-track rms as a measure of consistency across the river with further validation against in situ and other satellite data where possible.

The thophic state of lakes and reservoirs has been studied for years since water quality can deteriorate due to the eutrophication phenomenon. This is a critical issue in the nexus of water supply, environmental management, and ecosystem conservation. Remote sensing techniques have been shown useful to estimate water quality such as chlorophyll-a, total suspended particles and water transparency. Different algorithms have been recently proposed for the estimation of these variables using Landsat Thematic Mapper (TM) data (30-m spatial resolution, and 16-day revisit time) and Deimos-1 (22-m spatial resolution, and around 3-day revisit time). These algorithms were derived and tested using ground measurements from a set of lakes and reservoirs in Spain.

Sentinel-2 offers the opportunity to continue with this inland water quality monitoring task, thanks to its 5-day revisit cycle and 10-30 m spatial resolution. In the framework of the Sentinel-2 preparatory activities ESA developed the SPOT5 take 5 experiment. From early April to the end of August 2015, SPOT-5 satellite was relocated in a 5-day orbit, before being decommissioned. Based on the spectral matching between both VNIR sensors, SPOT-5 was used to simulate Sentinel-2 products and show the benefits of its high spatial resolution to monitor small water bodies.

In this work we focused on the Jucar river basin, located in the central part of the Mediterranean coast of Spain, in the province of Valencia. The area of interest is centered at around 39.30 ºN, -0.45º W. A few reservoirs plus the Albufera Lake fall within the SPOT5 scenes provided. Several experimental campaigns were carried out concurrent with SPOT-5 overpasses or close in date. Data collected in these campaigns completed the registers from the automatic stations network. Chlorophyll-a concentration, transparency and seston concentration were measured. Genetic programming models were used to generate nonlinear regression equations between ground measurements and reflectance values from the SPOT-5 spectral bands. More than 20 SPOT-5 images were available for the period April-September 2015. Maps of Chlorophyll-a concentration, transparency and seston concentration were generated, and the evolution of the trophic state of the Albufera Lake and some other reservoirs was analyzed in terms of these parameters.

These results give confidence to the use of Sentinel-2 images as a basic tool for the operational monitoring of variables of the trophic state of lakes and reservoirs, avoiding the necessity of continuous and tedious ground sampling.

This paper presents the results of a comparison of soil moisture obtained from the standard algorithms of SMOS, GCOM-W1 with soil moisture retrieved from SMOS multi-angular brightness temperature observations using dielectric model specially developed based on soil samples collected at the test site. As a test site, an area close to Marresale Weather Station (MWS) on the Yamal peninsula was chosen. This test site is located on the coast of the Kara Sea, coordinate of the under satellite point of interest (PoI) is (69.7115N, 66.8197E). The landscape at the test site is a typical well-drained polygonal tundra, comprised of grasses, moss, lichens, and prostrate dwarf shrubs. This choice is due to the following factors. First, the temperature dependent multi-relaxation spectral dielectric model (TD MRSDM) earlier developed in [1] is based on the soil samples collected at the test site. Second, soil moisture measured in situ, during the nine day mission on August, 2015 from 12 to 20 at the test site is available. Third, daily soil temperature and air temperature (2m above the ground) measured in situ by MWS appeared to be easy available, as well as the SMOS and GCOM-W1 data. For soil moisture retrieval, full polarization brightness temperature product of the Centre Aval de Traitement des Données SMOS (CATDS) [2] was used. Standard soil moisture data was obtained from SMOS L3 SM products, provided by the CATDS and GCOM-W1 L2 SMC products, provided by the Japan Aerospace Exploration Agency (JAXA) [3]. Method to retrieve soil moisture is based on solving inverse problem by minimizing norm of the residuals between the observed and predicted values of SMOS brightness temperatures. The calculation of brightness temperatures were performed using semi-empirical model of radiothermal emission [4], which was modified and takes into account an attenuation of the microwaves in canopy and TD MRSDM for an organic-rich tundra soil [1]. As a result, the values of soil moisture obtained from standard SMOS and GCOM-W1 products at the test site was found to be very understated compared to in situ measured soil moisture values. Retrieved values of soil moisture based on our approach with using specific dielectric model much more close to in situ measured soil moisture values. This study shows that for the Arctic soils rich in organic matter, existing soil moisture products of SMOS and GCOM-W1 do not meet the criteria of measurement accuracy of soil moisture. Solve this problem allows the use of a dielectric model specially developed for the typical organic soil collected at the test site. Validation of standard soil moisture products of the SMOS and GCOM-W1 is required over the Arctic territory.

Forests are a critical component of the global carbon cycle, storing significant amounts of carbon, split between living biomass and dead organic matter.The carbon budget of forests is the most uncertain component of the global carbon cycle – it is currently impossible to quantify accurately the carbon source/sink strength of forest biomes due to their heterogeneity and complex dynamics.

It has been a major challenge to generate robust carbon budgets across landscapes due to data scarcity. Models have been used for estimating carbon budgets, but outputs have lacked an assessment of uncertainty, making a robust assessment of their reliability and accuracy challenging.

Here a Metropolis Hastings - Markov Chain Monte Carlo (MH-MCMC) data assimilation frameworkhas been used to combine remotely sensed leaf area index (MODIS), biomass (where available) and deforestation estimates, in addition to forest planting information from the UK's national forest inventory,an estimate of soil carbon from the Harmonized World Database(HWSD)and plant trait informationwith a process model(DALEC) to produceaconstrained analysiswith a robust estimate of uncertaintyofthe UK forestrycarbon budgetbetween 2000 and 2010.

Our analysis estimates the mean annual UK forest carbon sink at -3.9 MgC ha^-1 yr^-1 with a 95 % confidence interval between -4.0 and -3.1 MgC ha^-1 yr^-1. The UK national forest inventory (NFI) estimates the mean UK forest carbon sink to be between -1.4 and -5.5 MgC ha^-1 yr^-1. The analysis estimate for total forest biomass stock in 2010 is estimated at 229 (177/232) TgC, while the NFI an estimated total forest biomass carbon stock of 216 TgC. Leaf carbon area (LCA) is a key plant trait which we are able to estimate using our analysis. Comparison of median estimates for (LCA) retrieved from the analysis and a UK land cover map show higher and lower values for LCA are estimated areas dominated by needle leaf and broad leaf forests forest respectively, consistent with ecological expectations. Moreover, LCA is positively and negatively correlated with leaf-life span and allocation of photosynthate to foliage respectively, supported by field observations.

This emergence of key plant traits and correlations between traits increases our confidence in the robustness of this analysis. Furthermore, this framework also allows us to search for additional emergent properties from the analysis such as spatial variation of retrieved drought tolerance. Finally our analysis is able to identify components of the carbon cycle with the largest uncertainty e.g. allocation of photosynthate to wood and wood residence times, providing targets for future observations (e.g. the BIOMASS mission). Our Bayesian analysis system is ideally suited for assimilation of multiple biomass estimates and their associated uncertainties to reduce both uncertainty in the state of the system but also process parameters (e.g. wood residence time).

As a result of government policies irrigated rice cultivation in the Senegal River Valley has shifted to from the wet to the dry season. Moreover, private and government investments have led to doubling of the total rice acreage since 2007 and cropping intensity has increased from1 to 2 crops per year. However, temperature extremes (both hot and cold) during the rice growing season frequently lead to yield losses or failure. Better knowledge is required to where and when rice is cultivated and which areas are mostly affected. The use of multi-temporal satellite imagery can provide useful information to track the dynamics of rice production, including extent of cultivated area per season, agricultural practices such as cropping intensity, and crop establishment and harvest periods. This information will ultimately be linked to temperature data bases to assess zones that are frequently exposed to climatic risks. Improved spatial knowledge on climatic risks will be exploited by extension services to target recommendations on the choice of rice varieties and optimal planting dates.

In this context, inter-annual variation of agricultural practices in the Senegal River Valley were analyzed by applying the PhenoRice algorithm (Boschetti et al., 2014) to time series of 16-days composite vegetation indexes 250m resolution MODIS images (Products MOD13Q1 and MYD13Q1) acquired between 2003 and 2014. The algorithm allows to identify the main rice cultivated zones and to estimate the dates of occurrence of the main phenological stages (crop establishment, crop emergence, flowering) on the basis of analysis of spectral indices (SIs) time series. Estimated acreages of rice cultivated in wet and dry seasons areas were compared with official statistics, while rice planting and harvest estimates were analyzed with respect to a database of farmer surveys that includes more than 1000 reports of seeding or transplanting and harvest dates for a period of 10 years. Interannual changes of cultivated rice areas statistics showed a clear correlation with those estimated from PhenoRice (Hot season R² =0.89; Cold season R² =0.50; all R² =0.76 ), and clearly highlighted a strong increase of cultivation of rice in the Senegal River Valley during the hot and dry season in the last years (Sowing dates between February and April). This result testifies the suitability of the PhenoRice algorithm to track spatio-temporal changes in rice cultivation practices, and its potential usefulness for rice monitoring in regions lacking high quality information on cultivation practices.

 

Acknowledgement:

The research leading to these results received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement n° 606983 (ERMES Project)

This work focuses on the pixel level uncertainty of forest variable estimates of boreal forest. Forest variables of interest are continuous essential climate model variables, such as forest biomass and tree height. The results are demonstrated over a 16km x 10km area in Southern Finland with source data from Metsähallitus in Finland and Landsat 8 images from the summer 2014.

The users of remote sensing products are not solely interested about the overall accuracy or resolution of the estimates but they are interested about the uncertainty of the estimates as well. Predictive variance of an estimate is a usable form of uncertainty. In addition, pixel-level uncertainty information has potential benefits for remote sensing method development as well, as this will highlight where more or additional modelling is needed.  This work demonstrated that pixel-level uncertainty estimates are possible over large areas of boreal forests.

The biomass predictions were computed by Gaussian processes regression. The GP regression method is a supervised learning method. A key assumption in GP modelling is that our data is sampled from a multivariate Gaussian distribution. One of the benefits for applying GP regression for biomass estimation is that from GP regression both predictive mean and a predictive variance can be obtained.

A Landsat 8 satellite image was selected a source data. The image has been acquired on July 23rd of 2014. Forest attribute data was received from Metsähallitus in Finland as ground reference data. The data comprised 3173 hectares of stand-wise forest maps, which were divided into training set of 1884 hectares and test set 1288 hectares.

Accuracy assessment of the estimates was carried out based on grid sampling. The root mean square error of the estimates in the test set were 63 m3/ha (42% relative error). The average standard deviation of the estimates were 50 m3/ha in the training set and 54 m3/ha in the test set. The estimate bias was 2.3 m3/ha, overestimated.

The results have been compared already with the ones produced by Probability method. Initial comparisons show that the estimated results are as good as the ones produced by existing methods with much less overestimated bias. GP regression provides information about the standard deviation in pixel level, which is huge benefit. The results are promising because very little a priori information was used in the modelling and the estimates were good.

This study is part of the North State Framework 7 project (grant no. 606962) of the European Union.

Surface temperatures derived from remote sensing data over heterogeneous, non-isothermal land surfaces depend on the viewing and solar angles mainly due to variations in sunlit and shaded fractions of the different elements in the field of view. The near-simultaneous dual-view capability of the Advanced Along-Track Scanning Radiometer (AATSR) can be used to estimate differences in brightness surface temperatures between the nadir (satellite zenith angle of 0°-23°) and forward views (around 55°) in the 11 and 12 μm bands. Brightness surface temperature (BST) is defined as the temperature corresponding to the radiance at surface level (that is, corrected for the atmospheric absorption and emission) through the black-body Planck’s function. We developed an accurate, pixel-by-pixel atmospheric correction method to obtain the BST for the 11 and 12μm bands at nadir and forward views. The atmospheric correction method uses: (1) Atmospheric profiles from the NCEP reanalysis product, provided at 1°×1° spatial resolution every six hours; (2) the MODTRAN 5 radiative transfer model; and (3) a GTOPO30 digital elevation model with spatial resolution of 1 km. The method provides the atmospheric transmittance and upwelling radiance at the AATSR resolution (1 km) for the two bands and views, taking into account the geographical coordinates, altitude and zenith observation angle of each pixel, and the AATSR overpass time. The method defines a tridimensional grid with horizontal resolution of 1°×1° (corresponding to the location of the NCEP data) and 12 altitudes from sea level to the maximum height in the AATSR scene. Each NCEP profile is modified in order to adjust the lower atmospheric levels to the 12 predefined altitudes, which is done by interpolation and/or extrapolation from the original NCEP levels and ignoring the layers below the prescribed altitude. These profiles are used as inputs of MODTRAN 5 to calculate the spectral atmospheric transmittance and radiance for zenith viewing angles from 0° to 60°, which are integrated to the AATSR 11 and 12 μm bands using the spectral response functions. The dependence of transmittances and radiances on zenith angle (θ) is modeled with linear functions of 1/cosq, which are obtained from regression. For a given pixel, the four closest 1°×1° nodes and the two closest altitude levels are selected for which the transmittance and radiance are calculated for the zenith angle corresponding to the pixel in the nadir and forward views. This is done for the two closest profiles to the AATSR acquisition time, and the resulting transmittances and radiances are linearly interpolated to that time. Finally, the values corresponding to a given pixel are obtained by horizontal and vertical interpolation. We used weighted horizontal interpolation, the weight being the inverse squared distance between the pixel and the grid node, and linear vertical interpolation. The method was applied to several daytime AATSR scenes over central-east Spain during 2007. The nadir-forward BST differences were usually positive, due to the larger fraction of sunlit elements in the nadir view and, as expected, the largest values were found in summer. For a given date, the largest differences were found for surfaces with low to medium vegetation covers, due to the large change in the proportions of sunlit and shaded elements observed by the two views. For such cases in summer, differences up to 8 K were found, with average values of 3-4 K. For near-full vegetation cover, the angular difference was about 0.5 K. This methodology is applicable to the forthcoming SLSTR sensor onboard Sentinel-3 and can be useful for analyzing the impact of different factors (season, land cover, LAI) on the directional variability of land surface temperatures. 

Synthetic aperture radar (SAR) sensors play an important role for the monitoring of wetlands due to their high sensitivity to soil and surface water and their day-and-night, all-weather operability. Due to the higher capability to penetrate vegetation L-band has been described as the preferred tool for wetlands application. On the other hand, data from C-band SAR sensors such as Sentinel-1 have become available at high temporal and spatial resolutions. However, performance is limited by factors such diverse as observation geometry, vegetation density/structure and wetland water height. These factors, which are often difficult to quantify, therefore represent potential sources of missed alarms when classifying flooded areas. Furthermore, the complex interactions between radar backscattering, hydrological processes and vegetation that occur in wetlands are still not well understood.

Recently, Yuan et al. (2015) reported changing positive and negative correlations between L-band backscatter and altimeter-derived water height depending on the total range of water level increase during flooding and the density of vegetation cover. The goal of the presented study is to assess if similar findings can be obtained in a comparison between multi-temporal C-band data from the ENVISAT Advanced Synthetic Aperture Radar (ASAR) with altimetry-derived water height and backscatter data. In addition, L-band data from the Advanced Land Observing Satellite (ALOS) are used as reference. The study area is a large wetland in central Zambia covering an area of more than 6000 km² with diverse vegetation types and where large-scale flooding occurs annually at the end of the rainy season and persists for several months.

Preliminary results show that for some regions similar relationships between C-band SAR backscatter and water height can be found as in the case of Yuan et al. (2015) for L-band. The results also highlight the potential value of altimeter backscatter for identifying floods under dense vegetation canopy which side-looking SAR sensors – using either frequency – are seemingly not able to penetrate.

Moreover, using only single scenes, open water can be considerably underestimated by ASAR in comparison to L-band data. This is mostly due to the influence of emerging vegetation and waves on the water surface induced by wind. Meanwhile, using multi-temporal data, at least the maximum annual flood extent can be estimated with sufficient reliability. Furthermore, harnessing the full time series of ASAR data comprising more than 200 scenes it is possible to identify areas in which ASAR shows low sensitivity towards flooding based on their backscatter signatures. The findings also hold relevance for the monitoring of wetlands using current missions like Sentinel-1 and the altimetry missions like SARAL and the future Sentinel-3 constellation.

Yuan, T., Lee, H., Jung, H. (2015) Toward Estimating Wetland Water Level Changes Based on Hydrological Sensitivity Analysis of PALSAR Backscattering Coefficients over Different Vegetation Fields. Remote Sens., 7, 3153–3183.

Former military training areas in Germany are often biodiversity hotspots. When the vegetation of these areas is left to natural succession processes without maintenance measures the variety of special plant societies is lost. Our study area, the Döberitzer Heide, west of Berlin with an area of about 60 square kilometers, is a heathland with an open landscape, which is dominated by dry sandy heaths, semi-natural grassland, humid meadows and wetlands. These habitats are protected by the European Habitat Directive (92/43/EEC). Therefore, a monitoring report about the conservation status of the protected habitats has to be prepared every six years. As large parts of the study area are inaccessible because of remnants of military munitions, remote sensing data is very useful to fulfill the monitoring task.

The aim of this study is to create a map with different classification scales of heathland habitats. The habitats of heathlands concerning the Habitat Directive are small-scaled and spectrally very similar to each other. Therefore, we used the different phenological characteristics over one vegetation period using a time-series with twelve scenes from March to October of the year 2011. Some habitats are unique in their phenological curve so that they can be classified with a high accuracy on a small-scaled level. However, other habitats are just separable by the valuable species and the spectral reflectance between two habitats can be very similar. Such habitats usually have low classification accuracies. In this case a coarser classification level is used.

For this reason we choose a hierarchical classification approach. The reference data contain information about the composition, the dominated species, the coverage and the moisture content for each plant community.

The first level of the hierarchical classification is the moisture gradient that includes the four classes dry, fresh, moist and wet. Based on the first level classification we classified alliances with 16 classes for the second level. The third level shows the plant society with 25 classes.

As classifier we used the machine learning technique Import-Vector-Machine (IVM). The IVM algorithm is based on the Kernel Logistic Regression Model. We used the probabilistic output of the IVM classifier to estimate how accurate the classification for each class is.

The results show that the overall accuracy of 82% for the coarse first level is reasonably good. The accuracy decreases with an increase of the level of detail from 70 percent for the alliance level (2^nd level) and 57 percent in the plant society level (3^rd level). However, some individual classes achieve high classification accuracy in the 3^rd level classification. Therefore, the hierarchical classification approach has proven very useful for our research question. Moreover, the dependencies of the classes are getting visible. For example the classified moisture gradient for the study area Döberitzer Heide shows the gradual changes and transition areas of the plant societies. In conclusion it is difficult to deal with a high number of classes which are spectrally very similar to each other. With the hierarchical classification approach it is possible to classify all habitats in their best classification scale.

Biomass is an Essential Climate Variable (ECV) required by the Global Climate Observing System (GCOS) to support the work of the United Nations Framework Convention on Climate Change (UNFCCC) and the Intergovernmental Panel on Climate Change (IPCC). Forests sequester carbon through photosynthesis and store it primarily as living aboveground biomass of trees (AGB). As 50% of AGB is carbon, it is a key parameter for monitoring carbon allocation in terrestrial ecosystems. The future BIOMASS mission from the European Space Agency (to be launch in 2020) aims to determine the amount and spatial distribution of AGB to improve resource assessment, carbon accounting and carbon models as well as monitoring AGB changes globally. The sensor will deliver three primary geophysical products: maps of forest AGB and height at 200 m scale every six months with 20% uncertainty, and annual maps of severe forest disturbances at 50 m scale with 90% classification accuracy. Methods to develop AGB maps with comparable accuracy levels and using readily available satellites are required for the period before the mission launch. The ESA DUE GlobBiomass project is developing state-of-the art approaches to map AGB and change at regional and global level for the period from 2005 to 2015.

Mexico was selected as regional case study by the GlobBiomass consortium. The GlobBiomass regional approach for Mexico is based on a data mining approach using Landsat surface reflectance multi-temporal composites, dual polarization ALOS PALSAR, and elevation data from the SRTM v3 imagery to produce a map of AGB density and its associated uncertainty. The country is covered by approximately 65 million ha of forests and 20 million ha of other wooded land (43% of the national territory). The country’s topography and climate are extremely varied, leading to a wide variety of different major habitats in which the forest biomass is distributed. The Yucatan peninsula is mostly covered by dry and moist tropical forest. The same approach is also tested in the Iberian Peninsula, which is a Temperate Mediterranean region characterized by a variety of forests, scrublands and sclerophyllous vegetation. Extensive in-situ national forest inventory datasets are available and will be used for training and validation of the maps.

The Soil Repository is a regional based collection of products developed under the Operational Plattform for Provision and Processing of Sentinel-data in support of Copernicus geoinformation services (OPUS_GMES) project. The primary objective of the Soil Repository is to supply value added information about soils at three levels: 1) the spatial distribution of exposed soils; 2) temporal statistics of those soils; and, 3) mapping of soil properties. Presently, the Soil Repository is designed for temperate climatic regions, such as central Europe or western Canada, that comprise areas of extensive crop based agriculture where soils are commonly covered by vegetation. To build the Soil Repository satellite based multi-temporal optical imagery is used to generate two composite images that reflect maximum vegetative and exposed soil coverage on a per-pixel basis using spectroscopic information. Subsequently, temporal characteristics of different land cover/use types are used to generate the level 1 binary map showing the spatial distribution of all exposed soils over the time frame of data availability. With the binary map in place level 2 temporal statistical information of exposed soils based on the original data stack are used to assess land use characteristics, such as seasonal information related to frequency in which individual agricultural fields are seeded and harvested. This information can also be used in the determination of longer term land use changes, such as from forest to agriculture and from agriculture to urban. At a third level the soil composite image derived from the full time frame of data availability allows for further assessment of soil properties over the spatially extended coverage of exposed soils. The resulting products can be updated as new data becomes available where ongoing monitoring and assessment of a region using information derived from the Soil Repository has direct links to important environmental (e.g. carbon flux) and societal (economic productivity) impacts. The generation of this Soil Repository has been designed towards free and open access high spatial muti-spectral Sentinel 2 data which will allow for 5 day repeat coverage (10 days with one satellite) with 13 spectral bands through the visible, near-infrared and shotwave infrared portions of the spectrum at a spatial resolution of 10, 20 and 60 m. The Soil Repository is also applicable to satellite based hyperspectral systems in the future, such as EnMAP, which will allow for a higher level of spectral information related to soil properties. As Sentinel II data will only become available in 2016, the Soil Repository processing chain is presently being applied to freely available archived Landsat (4,5,7) imagery from 1986 to present for all of Germany. Level 1 and 2 products are presented for all of Germany, whereas the level 3 product is focused on the state of Bavaria where the Landsat products are validated using soil information from different soil surveys and permanent soil observation sites overseen by the Bavarian Environment Agency (LfU) and the Bavarian State Institute for Forestry (LWF).

Soil moisture is a key parameter that plays a critical role in the surface energy balance at the soil atmosphere interface with direct influence on the evaporation, the runoff generation and the percolation of the water into the soil. The main goal of this paper is to estimate the surface soil moisture of an agricultural area affected by soil salinization and erosion. The second goal is to relate soil moisture with spatio-temporal dynamics of land degradation using Synthetic Aperture Radar (SAR) interferometry. The experimental analysis is carried out on data acquired in the joint ESA-CSA SOAR Europe 16605 scientific proposal over the North Braila Terrace agricultural area during 2014-2015 from the RADARSAT2 in Quad Fine mode. The chosen test area - Braila Plain has the special particularities such as: dry climate, high annual average temperatures (9-11⁰C), very dry and hot summers which cause a large potential evapotranspiration and conduct to a moisture deficit in soil, alkaline soils, winter winds with an average speed of 2.7 - 3.4 m/s. The soil type and climate conditions favor the culture of maize (50%), wheat and successive crops (16%), alpha-alpha (18%), sugar beet (6%), sunflower (7%), vegetables and other crops (3%).

Polarimetric Synthetic Aperture Radar allows as to extract information for identification and classification of different natural features since each polarization is sensitive to different surface characteristics (shape and orientation) and properties (soil moisture, surface roughness and vegetation cover). Single polarimetric channel cannot resolve unambiguously of the bare surface scattering problem. One approach is to use SAR polarimetry for scattering mechanism decomposition in order to characterize and subtract the volume from a ground component (surface / dihedral). Therefore, PolSAR technique was applied to invert the soil moisture over bare fields and to decompose the signal on the scattering mechanism components. Also, the InSAR, PolSAR and multi-temporal SAR analysis were performed for polarimetric SAR signal interaction monitoring. Taking into account the soil type, climate conditions and geomorphological characteristics of the studied area, Oh and Dubois semi-empirical scattering models were applied for the volumetric soil moisture and surface roughness estimation. The preliminary results indicate that some SAR polarimetric parameters such as: alpha angle, entropy, anisotropy and differential reflectivity are sensitive to crop growth stages for wheat and ripe. Soil moisture estimated from full polarimetric RADARSAT 2 data shows a deficit in the moisture content during autumn-spring season with an improvement in harvest stage due to the irrigation measures.  Since the decreasing in phase is correlated to an increasing in soil moisture, a coherent and non-coherent multi-temporal analysis was performed. Soil moisture and land degradation could be used as conservation planning tools in the Braila agricultural area, especially in identifying degradation prone areas in relation to hydrological conditions.

The estimation of soil moisture for agricultural areas is an urgently needed parameter for a variety of applications, e.g. for yield estimation or the monitoring of the requirements of the water framework directive (WFD). Therefore, an indicator-based model is tested for implementing different sensors for predicting the water content in open and vegetation covered riparian soils. Since the soil moisture content is not stable over time, the estimations were repeated for several time-steps within the vegetation periods of 2014 and 2015. The presented study is focusing on the Peene area close to the city Demmin in North-East Germany.

A set of 22 RapidEye images and 17 Landsat 8 images was acquired throughout the vegetation periods of 2014 and 2015. The images were co-registered and atmospherically corrected with ATCOR. In-situ measurements of 50 vegetation plots (including coverage and dominating species, soil moisture, soil and surface temperature) were available for up to seven different time-steps. Subsequently, a robust set of indicators and the sensor data can be applied to derive maps, which can be used independently from the time of the year and the utilized sensor.For the linkage between the sensor data and the soil moisture measurements several vegetation indices will be tested for their potential to detect indicator species as proxies for soil moisture. Since only a limited number of in-situ sample points is available for each class and to ensure a transferable method to other test sites, a robust one-class-classification approach (Maximum Entropy algorithm (Maxent) (Sethna, 2006; Shannon, 1948)) is used to determine the classes individually (Stenzel et al. 2014). For each class an unique classification run was done using its points of occurrence and 10.000 random generated background points as pseudo-absence data. Subsequently the resulting logistic probabilities of occurrence (ranging from 0 to 1) of each class were merged to a common map. For this task a specific cut-off criteria (threshold) was used for each classification output to decide if a predicted class is present or absent. Finally the class allocation for each pixel is determined by the highest probability of all present classes. First results indicates that the one-class-classification approach reached a higher model fit than a multi-class approach (SVM) when the NDVI is used.

In a further step, the above mentioned method should be applied with several vegetation indices and randomly selected acquisitions to allocate influential phenological phases in which the indicator species can be detected most accurately.

 

References

Stenzel, S., Feilhauer, H., Mack, B., Metz, A., & Schmidtlein, S., 2014. Remote sensing of scattered natura 2000 habitats using a one-class classifier. International Journal of Applied Earth Observation and Geoinformation, 33(1), 211–217.

Sethna, 2006. Statistical Mechanics: Entropy, Order Parameters, and Complexity. Oxford University Press, Oxford.

Shannon, 1948. A mathematical theory of communication. Bell Syst. Tech. J. 27, 379–423.

Coastal dunes are fragile systems threated by erosion. On many islands in the European Wadden Sea, the shrub Rosa rugosa, the Japanese Rose, has been introduced to stabilize the sandy substrate. The species is native to northeastern Asia and is well adapted to the conditions on extreme and salty soils. Due to these characteristics, Rosa has subsequently started to invade the adjacent ecosystems, thereby severely affecting their compositional and functional properties. In consequence, the negative impact of the species on the dune ecosystems thwarts the stabilizing effect on the substrate and the species is nowadays subject to eradication programs. Building quantitative records of its distribution, status, and dynamism is essential for handling plant invasions. Maps of the distribution of Rosa are thus urgently needed.

Hyperspectral remote sensing is an effective and versatile technique for mapping and detecting the spread of invasive species. To fully exploit its potential, the use of one-class classifiers that require only calibration data on the occurrence of the invasive species is promising. We thus test in this study whether Rosa rugosa can be detected in an early stage of the invasion with hyperspectral data and a one-class classifier approach.

The study area is the island Sylt in the Wadden Sea in northwestern Germany. Hyperspectral image data with 1.8 m x 1.8 m GSD were collected in summer 2014 with an airborne APEX sensor covering 0.4-2.5 μm spectral range with 285 bands. Ground truth data on the presence of Rosa were sampled across the whole island on 3 x 3 m plots for calibration. In addition a validation data set consisting of presence and absence information on Rosa occurrences was sampled. The data sets will be subjected to One-Class Support Vector Machine (SVM) classification, a machine learning technique, for a concise delineation of Rosa in the spectral feature space and the detection of Rosa occurrences in the image data. The predicted distribution patterns will be validated against actual observations in the field. Results on the potential of imaging spectroscopy and One-Class SVM for an early detection of an invasive species in a dune ecosystem will be presented.

More than 80% of agricultural land in Ireland is grassland, providing a major feed source for the pasture based dairy farming and livestock industry. In grassland management and the livestock business, grazing capacity and intensity are key factors, and for sustainable farming need to be monitored consistently in order to optimize resources and avoid grassland degradation. This work investigates the application of optical time series imagery to estimate the grassland biomass at two sites in the Republic of Ireland using both statistical linear regression and state of the art machine learning algorithms. Artificial neural networks (ANN) are one of the most commonly used machine learning algorithms which have the ability to learn from complex patterns in a dataset, and fuzzy logic approaches have the power to reason and generate rules from the dataset. Adaptive-neuro fuzzy inference systems (ANFIS) are the integration of both ANN and fuzzy logic, combining the power of both methods to provide an approach with improved predictive or approximation ability.

In situ and satellite data covering 12 years for the Moorepark and 6 years for the Grange study sites were used to predict grassland biomass using multiple linear regression, ANN and ANFIS models. The results demonstrate that a dense (8-day composite) MODIS image time series, along with high quality in situ data, can be used to retrieve grassland biomass with high performance (R² = 0.86, RMSE = 11.07 for Moorepark). Due to the combined features of ANN and fuzzy logic, the ANFIS has the ability to accurately model complex and chaotic systems, and the results concur with those of the literature which report a high predictive power of ANFIS compared to the ANN. The ANFIS model allows multiple inputs to produce a single output, however to achieve a higher level of accuracy at a regional scale a large sample size from different locations would be required to drive the model.

The model for Grange was modified to evaluate the synergistic use of vegetation indices derived from remote sensing time series and the accumulated Growing Degree-Days (GDD) information. As GDD is strongly linked to the plant development, or phenological stage, an improvement in biomass estimation would be expected. Daily minimum, maximum and average temperature data from an on-site weather station were used to calculate the GDD for the Grange study site. It was observed that using the ANFIS model the biomass estimation accuracy increased from R²=0.76 to R²=0.81 and root mean square error reduced by 2.72%, however for large scale mapping a denser sampling of weather data is required in order to minimize the effects of different micro-climatic zones.

Overall, the work presented in this study has demonstrated the potential of dense remote sensing and weather data time series to predict grassland biomass using machine-learning algorithms, where high quality ground data were used for training. In future, with an increased availability of in situ samples and a higher spatial and temporal frequency of optical imaging systems, this strategy should generate more robust estimates of grassland biomass at a national scale. The proposed approach complements the current paradigm of Big Data in Earth Observation, and illustrates the feasibility of integrating data from multiple sources. In future, this framework can be used to build an operational decision support system for retrieval of grassland biophysical parameters based on data from long term missions such as Landsat and Sentinel. 

In arid, semi-arid and Mediterranean environments, most surface waters exhibit a high inter- and intra-annual variability primarily in response to precipitation and evapotranspiration. A better knowledge of the spatial and temporal distribution of surface water in these areas is critical for many applications including water management, ecosystem assessment and biodiversity conservation. The Moderate Resolution Imaging Spectrometer (MODIS) provides global measurements of the land surface with high frequency repeat coverage, and is therefore well suited for long-term mapping and monitoring surface water at high temporal resolution.  However, the relatively coarse spatial resolution of MODIS may easily miss out on small water bodies, which in arid, semi-arid and Mediterranean areas are distributed throughout the landscape.

To address this problem, this study proposes options for estimation of surface water fraction from MODIS 16-day Bidirectional Reflectance Distribution Function (BRDF) corrected surface reflectance image composites. Surface water fraction refers to the proportion of surface that is covered by water within each MODIS pixel. Multi-temporal Landsat imagery of three locations in Spain were selected, spanning different years (i.e. wet and dry years) and seasons. They were first classified into water/non water classes using water index (e.g. NDWI) thresholding method and then aggregated to 500m spatial resolution yielding a continuous representation of the surface water from 0-100 percent. These classified Landsat images will be used as training and validation data for estimating the fractions of surface water in the corresponding MODIS 500m pixels.

We will compare different models (e.g. empirical models and regression tree model) to predict water fraction from MODIS imagery. As input predictor variables we use the surface reflectance for individual MODIS bands, as well as MODIS-derived spectral indices, including the Normalized difference vegetation index (NDVI), Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI), Tasseled Cap brightness index (TCBI), and Tasseled Cap wetness index (TCWI). An additional input is the Topographic wetness index (TCW) derived from the 30-m elevation model obtained from the Shuttle Rader Topography Mission (SRTM) . The accuracy of the models will be assessed with root mean square error (RMSE) and r². Subsequently the model with the highest accuracy will be applied to produce high temporal resolution (i.e. every 16 days) maps of surface water fraction for Spain from 2000 to 2015. Our study will provide more detailed spatial and temporal information on surface water which is useful for monitoring and assessing the seasonal, inter-annual and long-term variability of water resources. 

The UN FAO Africover project was active from 1995 to 2000, during this period it produced the most accurate map of land cover/use for Central-Eastern African countries.

The project was specifically dedicated to map agricultural areas in Central Africa. Since 2000 this map has never been updated. The present study aims at demonstrating how necessary an updating of this product really is. Our case study is the Kenya’s Grain Basket (Rift valley); and we are using Landsat 8 and (when continuously available) Sentinel 2 imagery. We are going to set up an automatic monitoring system able to classify agricultural areas and detect land use changes. The FIHS (Fast Intensity Hue Saturation) pan-sharpening technique was used to increase Landsat 8 imagery resolution, from 30m to 15m, in order to overcome the limitation imposed by a low resolution for mapping the small and medium size fields that characterized Kenya’s agricultural  areas; A multi-temporal object-based classification technique, that implements both spectral and phonological information (extracted from NDVI time series), was used for identifying  the current agricultural areas within the area of interest. Deforestation, erosion, reforestation and urban sprawl were been the main natural drivers to the lost of agricultural areas, on the other hand major policy changes have deployed several programs for recovering agro-pastoral and forest areas and for introducing alternative technique of soil preservation; those areas are faced with increased health and resource issues, it is important to look at which extent the land surface is changing over time; the crop growth stage represents essential information for irrigation scheduling, fertilizer management and evaluating crop productivity, monitoring over time food security. Testing sites on the ground have been selected to evaluate how the land cover/use map has evolved during the last decade.

Wetlands are among the most threatened ecosystems in the Mediterranean countries by human activities. During the last decades, their protection and conservation, by the implementation of monitoring programs (e.g. those related to the WFD in the European countries), became an important priority for many national and international organizations and governments. Additionally, wetlands delineation and inventorying is a key parameter for the monitoring of this ecosystem. Despite this, very few countries around the Mediterranean Basin have a comprehensive inventory of their wetlands. This is partially due to a lack of efficient tools that allow management authorities to implement wetland inventories at a wide scale (river basin and/or country level).

In the frame of the Horizon-2020 SWOS project, a new approach is tested in order to provide an innovative EO-based tool that can map and delineate potential wetland areas at a wide scale (large river catchments and/or country scale) and can be used as a map baseline for finer resolution field wetland inventories. This tool is developed by improving existing methodologies (e.g. those of the GlobWetland-II project). Most of these were based on water detection using optical or SAR imageries. Indeed, during the GlobWetland-II project, the spectral bands of the Landsat Thematic Mapper (TM 4, 5 and 7) in the Near Infrared (NIR) and Short Wave Infrared (SWIR) regions, combined with multi-spectral and multi-seasonal indices (NDVI, SAVI, NDWI and MNDWI), were used to map water surface dynamic (permanent water and temporally flooded areas) and vegetated wetlands.

In this new proposed approach, the mapping process will follow three main steps:

a) The first one is the production of a binary wetlands map with two classes (potential wetlands and no wetlands) using water surface dynamic maps derived from SAR imagery (Sentinel-1) and vegetated wetlands map derived from SAR and optical data (Sentinel-1, Sentinel-2 and Landsat-8). These maps represent a full annual dynamic of water extent and aquatic vegetation cycles and their production needs to use very high temporal resolution satellite data (e.g. 2015 full hydrological year in our case).

b) Beside the multi-temporal satellite imagery products, the information on surface water storage capacity of the watersheds is obtained from DEM analysis (calculation of Topo-climatic Wetness Index) and used during the downscaling process of the water surface dynamic and aquatic vegetation map to produce maps of potential wetland areas with different levels of occurrences (from 0 to 100%).

c) The last step is a simple Land Use/Land Cover classification (with high level classes, e.g. the 2^nd level of CLC nomenclature) using the implemented SWOS wetland habitat mapping tool to provide a more accurate quantitative (surface and delineation) and qualitative (wetland type: natural or human-made habitats) information on wetlands at a river basin scale.

At the end of the SWOS project, this new tool will be integrated within the SWOS mapping software. It could be used to provide additional information on wetlands delineation derived from EO data that can help the implementation of wetlands inventories at local and/or national scales.

The water regime of soils in highland agriculture areas is strongly impacted by high soil heterogeneity due to pedogenesis (e.g. parent material, terrain configuration) as well as improper land consolidation processes. As a result, large fields of arable land can encompass soil zones with shallow leaching profiles and hence different water and nutrition regimes. Identification of such zones is necessary for economically efficient and ecologically sensitive management. In addition, most of these areas function as important underground drinking water recharge sources.

This study aimed at: 1/identification of water-vulnerable zones in crystalinicum soils of the Czech-Moravian Highlands from different spatial/spectral/time resolution remotely sensed imagery data and ancillary GIS data, namely Land Parcel Identification, soil maps, digital elevation model derived from map contours scale 1:10 000; 2/comparison of results based on remote sensing and GIS analyses with those derived from soil maps.

We hypothesized that dry vegetation seasons cause higher variability in reflectance of hyperspectral/multispectral data from monoculture crops on arable land due to different soil water availability for plants. That is why selection of appropriate dates for satellite and airborne data acquisition was based on meteorological conditions.

On a detailed scale we used two sets of airborne hyperspectral data acquired over two watersheds (cca 12 km2) by AISA Eagle scanner:  2010 a normal or rather above average precipitation year in vegetation season (May 60 mm, June 65 mm); 2011 dry spring (April 26 mm, May 21)   in order to estimate spatial heterogeneity (SH) within each crop/field. Multilayer analyses on results from unsupervised classification of post-processed hyperspectral data, vegetation indices and topographic parameters were used to characterize SH and identify vulnerable zones.

On a larger scale we chose three dates of ASTER scenes: 2003 – dry spring (5.5. 2003) and dry autumn (26.9. 2003); 2007 – “normal” spring (10.6.2007).  Data procedures similar to detailed scale were carried out under a mask of arable land parcels greater than 10ha, in territory approximately200 km2.

Our results show that vulnerable zones identified from remote sensing and ancillary data display a 52% overlap with soil properties zones coming from soil maps on detail scale and 65% for larger scale. We hypothesize that better performance in latter case is due to combination of bare/vegetated field cover and SWIR availability from ASTER.

Here we present a new method for retracking of SAR waveforms over rivers and lakes. Satellite altimetry offers frequent and global sampling across borders, which can be used to validate and calibrate hydrological models in remote areas where in situ measurements are scarce.

 

The method was developed using CryoSat-2 20Hz SAR data, but due to the similarities between the Sentinel-3 SRAL altimeter and the SIRAL altimeter on-board CryoSat-2 an adaption of the method will be straightforward.

 

The MWaPP retracker is based on a sub-waveform retracker, but takes the shape of adjacent waveforms into account before selecting the sub-waveform belonging to nadir. This is new compared to primary peak retrackers, and alleviates a lot of snagging due to off-nadir bright targets, but also topography challenges.

 

The results from the MWaPP retracker show a significant decrease in the standard deviation of the mean of various lake and river crossings throughout the world. Results are presented for rivers and lakes, such as The Brahmaputra River, The Amazon River, The River Thames, and for Lake Vänern in Sweden and Lake Okeechobee in Florida, US. 

Forests play an important role in regulation of the global climate; moreover, they provide human beings with a whole range of ecosystem services. Forest health and ecosystem functioning have been influenced by anthropogenic activities and their consequences, such as air pollution, surface mining, heavy metal contamination, and other biotic and abiotic stress factors, which had an especially serious effect on central Europe. Modern remote sensing has become a novel tool not only for detecting target materials but also for monitoring dynamic processes and physical-property induced changes. The use of multispectral imagery has been demonstrated to effectively map the distribution of ecosystem types and vegetation systems; however, the low spectral resolution of multispectral imagery is a major limitation. On the other hand, imagery with higher spectral resolution (e.g., hyperspectral) provides sufficient spectral resolution to describe diagnostic absorption signatures and allows sufficiently detailed species discrimination and biochemical differentiation. Hyperspectral satellite data will be available in the near future (e.g. EnMap in 2018), however a new superspectral satellite, Sentinel-2 (ESA), was launched in June 2015.

Sentinel-2 mission combines the advantage of high temporal resolution of common satellite multispectral data with higher number of spectral bands important for vegetation analysis as there are 5 bands within the 650 – 850 nm domain compared with only 2 bands of sensors like Landsat, ASTER, ALI etc. As so it is possible to apply some of the spectral indicators of vegetation properties developed originally for hyperspectral imagery even in a generalized form. One of these indicators is so called Red Edge Position (REP) defined as the inflection point of the reflectance curve within the red-edge domain. As it was reported by numerous studies, REP has been widely used as an indicator of vegetation health status as it is changing with the crucial biophysical and biochemical parameters such as for example chlorophyll content, leaf area index etc.

In the presented study we utilized the set of airborne hyperspectral data (HyMap multi-date and APEX data acquired over Sokolov basin and Ore Mountains, the Czech Republic) which Sentinel-2 simulated imagery were derived from. We focused on assessing several available REP calculation techniques (e.g. maximum first derivative, 4-point interpolation, linear extrapolation, polynomial fitting) as well as on comparing the REP values simulated using a RT model, extracted from hyperspectral imagery as well as calculated from the Sentinel-2 simulations. The obtained REP values were then correlated with leaf (C_ab) and canopy (C_ab×LAI) level chlorophyll content retrieved by laboratory analysis of Norway spruce needles collected during the hyperspectral data acquisitions and LAI estimation based on processing digital hemispherical photographs. It is thus possible to demonstrate which of the used REP calculation techniques has the highest potential for estimation of basic biochemical/biophysical vegetation parameters using the Sentinel-2 level data. 

After 6 years in orbit, the European Space Agency (ESA) Soil Moisture and Ocean Salinity mission (SMOS) continues to provide high-quality, global maps of soil moisture and ocean surface salinity. Its payload MIRAS (Microwave Imaging Radiometer with Aperture Synthesis) is an L-band interferometric radiometer which achieves unprecedented resolution. Since MIRAS is the first in its kind in orbit, an exhaustive monitoring of its behaviour in space and its measurements is periodically conducted by the mission quality control team.

This paper provides a summary of the long term trends of the key parameters of the instrument health and the SMOS measurements conducted by SMOS Quality Control (QC) Operations. The present analysis covers calibration trends since the beginning of the mission with the latest processing baseline, v62x. Long-term instrument stability, from brightness temperature measurements over Antarctica, is also assessed. In addition, Hovmöller diagrams of the biases are provided, in order to assess latitudinal and temporal stability. Operational QC of SMOS is carried out under the Instrument Data quality Evaluation and Analysis Service (IDEAS+).

Over the last years, the frequency and intensity of flood events evidently increased all over the world. Especially floods which pass abandoned open pit mines, industrial areas, intensively used agricultural fields and urban structures may be enriched with various pollutants, such as heavy metals. Natural or near-natural floodplain ecosystems are particularly exposed to flood events. Here, polluted sediment loads and dissolved heavy metals accumulate in the soil and gradually enrich in the floodplain vegetation. Since floodplains are often subjected to extensive land use, e.g. as hay meadows, it is necessary to monitor these contaminations.

Therefore, the aim of this project is the spatial acquisition and assessment of the current state of heavy metal contamination with hyperspectral remote sensing data. The focus of this study is the development of hyperspectral indices to apply spectral information as indicator for the detection of heavy metal induced plant stress. Usefulness of established indices will be tested. Further methods, like determination of the Red-Edge position and spectral reflectance normalization techniques will be applied. Ecotoxicological effects of heavy metals are derived directly and indirectly from spectral soil and plant properties, and semi-automatic algorithms are used for parametrization and separation of stress features in floodplain vegetation.

Within this study, three overflights were conducted with a HySpex full range [0.35 - 2.5 µm] hyperspectral scanner in two floodplain areas in Central Germany at the Elbe River in May, June and July 2015. Simultaneously, a total of 42 soil and 98 vegetation samples were collected. As contaminant accumulation highly depends on surface morphology, a high resolution laser scanner digital elevation model was utilised for an object based classification approach to determine sampling locations in the field. Thereby, relational object units characterized as sinks, terraces and plateaus were derived and sampling points defined. Along a 30 m transect spectrometric ground measurements were carried out for each sampling point using an ASD FieldSpec Pro [0.35 - 2.5 µm]. In addition, various plant parameters, e.g. vegetation height and chlorophyll content, were assessed. Soil and vegetation samples will be analysed in the laboratory for Pb, Cd, Cr, Cu, Ni, Zn, As and Hg contents. Also the pH-values and water- and nutrient contents (PO₄, K⁺) as well as grain size and organic matter of soils will be determined. Results indicate correlations between heavy metal contents, spectrometric measurements and various soil and vegetation parameter. The observed relationships will afterwards be transferred to Hyspex data and then be used for vegetation stress detection and for the derivation of contamination maps of selected floodplain areas.

Upcoming research will focus on the transformation of the algorithm to simulated EnMAP data which is based on the acquired hyperspectral airborne images.

The project is founded by the Federal Ministry for Economic Affairs and Energy (BMWi), project number: 50 EE 1346

Global trends of declining biodiversity are widely acknowledged, which have a direct impact on ecosystems’ functioning and their provisioning of services. Systematic mapping and monitoring of biodiversity patterns, e.g. of biological communities and their turnover, are thus needed. Earth observation data, coupled with suitable methods for analysis, have great potential for characterising these patterns in an accurate manner. Previous studies have shown that hyperspectral data systematically collected at repeated times are able to provide very detailed information on the Earth’s surface, and this way are suitable for characterising complex ecological systems. The highly dynamic and heterogeneous Brazilian Cerrado, while remaining largely understudied, holds ca. 5% of the world’s species. However, economic pressures and relaxed conservation laws result in the continuous destruction of this globally important system thus constituting a global biodiversity hotspot. In this study we used time series of hyperspectral (EO-1 Hyperion) and multispectral (Landsat) data to map spatial transitions in woody plant communities transitions, based on ground-based inventory data. We applied a Sparse Generalised Dissimilarity Modelling approach, capable of analysing such complex data. We further assessed the trade-offs between the spectral and temporal domains for describing biodiversity patterns in our study area. We conclude that our approach is suitable to systematically monitor changes in plant community patterns.

Satellite image time series enable an enriched insight into the crop phenological status and seasonal growth cycles and as such provide a unique and valuable tool for crop type differentiation and condition monitoring. In addition to spectral values of individual bands, a comprehensive understanding of vegetation behaviour and its biophysical characteristics can be obtained by analysing vegetation indices over time. Those indices, most often formed as the mathematical combination of several spectral bands, are an effective and simple means of extracting and enhancing information contained in spectral reflectance data and of reducing the amount of input data included in analysis.

 

Vegetation indices have often theoretically been defined in relation to a specific problem, specific research field or sensor with a common purpose to minimize the effect of mixture of vegetation, atmosphere, shadow, soil colour and moisture. However, it has been suggested in the literature that in practise there are few differences between the multitudes of vegetation indices that have been developed in the last three decades. As the red and near infrared band often contain more than 90% of the information related to vegetation, a variety of vegetation indices predominantly addresses the reflectance in these parts of spectrum, implying a high statistical correlation and similarity. Thus, not all of them may be beneficial for the crop type classification performance, both in terms of accuracy and computation time. With an advent of not only high spatial but also high spectral and temporal resolution satellites like Sentinel‑2, the amount of data to be processed is increasing exponentially. Therefore, it is important to perform preliminary evaluations of the most significant vegetation indices for crop differentiation in order to minimize data redundancy.

 

The objective of the presented study is to explore and understand the multitemporal correlation between a variety of vegetation indices in order to reduce the computation time of crop classification and to obtain efficient, meaningful and prominent indices suitable for crop type identification. Five major crop types (wheat, maize, rapeseed, barley and triticale) and two categories of vegetation indices (spectral ratio indices and orthogonal indices) are analysed. The importance level of each of the indices for crop type separability observed over one growing season is compared to the contribution of individual spectral bands and particularly to NDVI, the most widely used vegetation index which is also often considered as the reference of the newly developed indices.

 

In terms of crop separability analysis, special attention is given to the following questions: Is the correlation between indices computed on an entire satellite image comparable to the correlation computed on the areas of each crop type? Is the correlation between indices stable throughout the growing season? Which indices are important measures for crop separability, why, and in what part of the growing season? Which vegetation indices can equally contribute to classification performance? Is the correlation comparable for different sensors? As each index should present the vegetation in its unique way, are there any indices that highlight any of the crop features specifically and which one? First preliminary results indicate strong correlation of many indices with NDVI.

 

Tests are being performed on the RapidEye and SPOT5 time series data for a period between April and November 2013 and April through September 2015, respectively. The access to SPOT5 data over the agriculturally important region and study site in NE Slovenia was guaranteed through the SPOT5 (Take5) experiment.

Abstract. The aim of the presented work is to determine applicability of high-resolution satellite images for monitoring forest stands located in a temperate zone, characterized by a wide variability of environmental conditions, in context of future usage of Sentinel-2 data. 10-meter SPOT images collected in 2015 vegetation season for forest area located in northeastern Poland within SPOT 5 Take 5 experiment were applied for studying differentiation of forest stands due to species type, forest site type and impact of meteorological conditions throughout growing season. Results of analysis of vegetation indices derived from SPOT 5 data were cross-compared with those obtained from analysis of Landsat 8 images acquired for the same area in 2015 growing season. Analyses of forest stands based on vegetation indices derived from high-resolution satellite data were supported with ground measurements of parameters characterizing tree conditions. The following ground parameters were collected in the course of field campaigns: pigment content, fluorescence level and spectral characteristics of vegetation obtained with the use of ASD Field Spec 3 spectrometer. The analysis of satellite images was also supported with ancillary data – the detailed forest inventory maps prepared by forest service. High temporal resolution of SPOT 5 Take 5 images (5 day’s cycle) and their high ground resolution – both features equivalent to those which characterize Sentinel-2 constellation – predestinate them for the detailed analysis of forest areas in spatial and temporal context. The already achieved results of the analysis demonstrate variability of vegetation indices derived from satellite data due to type of species and type of forest site. They also reveal some relationships between the derived indices and meteorological conditions existing within study area throughout the growing season. Cross-comparison with indices derived from Landsat 8 OLI images demonstrated better sensitivity of SPOT-based indices to changes of environmental conditions due to higher spatial resolution, but also pointed out the need to apply narrowband-based indices derived from red-edge spectral region to better characterize subtle differences in vegetation types and growth conditions. This option will be possible to develop through the use of Sentinel-2 data, which offer information in narrow, red-edge spectral bands. The results of satellite-based analysis was supported with conclusions based on analysis of ground spectroradiometric measurements, which pointed out superiority of narrowband vegetation indices for monitoring conditions of forest stands, as compared to broadband indices.

The presented work is a part of the research WICLAP project, conducted within Polish-Norwegian Research Programme, managed by the National Centre for Research and Development (NCBiR). The project is aimed at assessment of impact of climate changes and air pollution on forest and tundra ecosystems, through application of multi-resolution EO-based approach, including high-resolution data acquired by new-generation satellites.

Sustaining forest resources in Romania requires a better understanding of forest ecosystem processes, and how management decisions and climate and anthropogenic change may affect these processes in the future. Due to high variation in forest communities, forest structure and the fragmentation of the forested area in Romania, satellite based biophysical parameters information for forest state analysis needs to meet particularly high quality requirements. Forest vegetation and climate interact through a series of complex feedbacks, which are not very well understood. The patterns of forest vegetation are largely determined by temperature, precipitation, solar irradiance, soil conditions and carbon dioxide (CO₂) concentration. Vegetation impacts climate directly through moisture, energy, and momentum exchanges with the atmosphere and indirectly through biogeochemical processes that alter atmospheric CO₂ concentration. Changes in forest vegetation land cover/use alter the surface albedo and radiation fluxes, leading to a local temperature change and eventually a vegetation response. This albedo (energy) feedback is particularly important when forests mask snow cover.Forest vegetation-climate feedback regimes are designated based on the temporal correlations between the vegetation and the surface temperature and precipitation. The different feedback regimes are linked to the relative importance of vegetation and soil moisture in determining land-atmosphere interactions.Forest vegetation phenology constitutes an efficient bio-indicator of impacts of climate and anthropogenic changes and a key parameter for understanding and modeling vegetation-climate interactions.Spatio-temporally explicit, quantitative assessment for forest vegetation bio-geophysical characteristics are a requirement in a variety of ecological and forest applications. Optical Earth observing satellites, endowed with a high temporal resolution, enable the retrieval and hence monitoring of forest phenology and key bio-geophysical variables (NDVI-normalized difference vegetation index, LAI-leaf area index, FPAR- fraction of photosynthetically active radiation). Super-spectral Copernicus Sentinel-2 and and Sentinel-3 missions are able to provide  data stream for forest land cover monitoring.Spatio-temporal forest vegetation dynamics have been quantified as the total amount of vegetation (mean NDVI) and the seasonal difference (annual NDVI amplitude) by a time series analysis of NDVI LAI satellite images over 2000 – 2015 period for a forest ecosystem placed in the North-Eastern part of Bucharest town, Romania, from MODIS Terra/Aqua, LANDSAT TM/ETM and Sentinel satellite and meteorological data. A climate indicator (CI) was created from meteorological data (precipitation over net radiation). The relationships between the vegetation dynamics and the CI have been determined spatially and temporally. The driest test regions prove to be the most sensitive to climate impact. The spatial and temporal patterns of the mean NDVI are the same, while they are partially different for the seasonal difference. For investigated test area, considerable NDVI decline was observed for drought events during 2003, 2007 and 2010 years. Under stress conditions, it is evident that environmental factors such as soil type, parent material, and topography are not correlated with NDVI dynamics. Specific aim of this paper was to assess, forecast, and mitigate the risks of climatic changes on forest systems and its biodiversity as well as on adjacent environment areas and to provide early warning strategies on the basis of spectral information derived from satellite data regarding atmospheric effects of forest biome degradation.EO-based estimates of forest biophysical variables were shown to be similar to predictions derived from forest field inventories.

An ESA Project has been launched in Poland to develop and validate advanced methodologies [1, 2] for deriving biophysical variables from the synergetic use of ESA optical sensors missions, with focus on Proba-V, Sentinel-2 and Sentinel-3. The methodologies will be validated according to standard protocols and the products quality assessment will be carried out following QA4EO (Quality Assurance for Earth Observation) guidelines.

New validation data will be collected as part of this project resulting from dedicated field campaigns in the Poland region. The field campaigns will include a wide range of ground measurements, allowing full characterization of carbon and water balance of the studied area. Field measurements will include in particular: Leaf Area Index (with LAI-2000 and LAI 2200 Plant Canopy Analyzer) Soil Moisture (using TRIME Field Measurement Devices), APAR (with AccuPar instrument), Carbon Balance (using chamber method and Eddy-Covariance method), wet and dry biomass, type of vegetation cover and its development stage. In addition to the field measurements, data from existing networks for global validation of land biophysical products, such as BELMANIP-2, will be used. Cross-comparison with available dataset of biophysical variables, such as the ones provided in NRT by the Copernicus Global Land service, will be additionally considered for quality verification. As a result of these validation activities, the suitability of ESA optical sensors products for the target land applications will be additionally assessed.

The validated biophysical products will be ultimately used for land monitoring studies of wetland and agricultural areas in Poland, this includes: vegetation status mapping, vegetation hazards (drought, flood) mapping, crop classification and crop yields estimation. Maps describing vegetation condition for the areas of interest, e.g., maps of wetland habitats and cover type, maps of vegetation indices (NDVI, EVI, GEMI), maps of vegetation hazards, maps of heat fluxes, maps of LAI and GLAI, maps of evapotranspiration will be made available to the users community.

[1] Katarzyna Dabrowska-Zielinska, et al., (2014) Monitoring Wetlands Ecosystems Using ALOS PALSAR (L-Band, HV) Supplemented by Optical Data: A Case Study of Biebrza Wetlands in Northeast Poland. Remote Sens., 6, 1605-1633

[2] Katarzyna Dabrowska-Zielinska, et al. (2004) Biophysical properties of wetlands vegetation retrieved from satellite images. Geoscience and Remote Sensing Symposium, 2004. IGARSS '04. Proceedings. 2004 IEEE International (Volume 7)

Evaluating the available forage biomass on a territory is nowadays becoming compulsory for a good farm management, furthermore on an island context with multiple constraints. On the one hand, the import of forage or food supplements represents an additional cost due to the distance of the production and supply centres. On the other hand, an increase in local production is hardly possible because of the impossibility of extending agricultural land.

From all these aspects, Reunion Island represents a particularly interesting “laboratory”. Its location, its volcanic origin and its rugged terrain lead to a multitude of microclimates on a very small territory (2 500 km²). In general, we can observe a tropical climate on the coast and a temperate one in altitude. It is also these distinctive features that bring Reunion Island as a study site rather complex to analyse.

In order to better understand the grass growth on Reunion Island pastures within this climatic diversity, different versions of a grass growth model[1] have been developed at CIRAD, remaining perfectible regarding the tropical grass. As on other models it uses a variable named Leaf Area Index (LAI), expressed by the leaf area (m²) per m² on the ground.

The objective of our study is to establish an empirical model between the LAI and NDVI (Normalized Difference Vegetation Index) for our study sites. If proved, this result will be used in future work in order to force the existing growth models so the predictions would be improved.

The LAI measurements are carried out on the ground thanks to a ceptometer which measures the solar radiation received on and under the canopy. They are carried out every five days so the delay between the measurement and the SPOT5 satellite images’ acquisition is very short. Thanks to the SPOT5Take5 program, we have been able to know the exact date of a satellite pass and therefore coordinate the field measurements with the images: in average, we got a maximum of three-day delay between the measurements and the images. Furthermore, the high frequency of image acquisition allowed us to take an important number of measurements (430) in a short period of time. Several plots have been selected on the island (9) in order to cover the two types of climate and then to be able to elaborate a model for plots with temperate grass species in highlands and plots with tropical grass species in lowlands.

From the SPOT5Take5 temporal series’ advantages, we can confirm the correlation between the LAI and NDVI on our study sites. This correlation is even strong on the exponential model as we obtain a coefficient of 0.87 (RMSE = 0.82) for temperate grass and 0.92 (RMSE = 0.81) for tropical grass.

[1] C. Detaille, 2013 - Creation of a new grassland growth module for the Global Assessment Model for Evaluating the sustainability of Dairy Enterprises

J. Vayssières, F. Guerrin, J.M. Paillat, P. Lecompte GAMEDE: A global activity model for evaluating the sustainability of dairy enterprises Part I - Whole-farm dynamic model

Pinus mugo is one of the most important and dominant species above the tree line in the Krkonoše Mts. National park (KRNAP). Its abundance was influenced by many natural and human factors/activities during the last centuries (avalanches, fires, colonization, development, agricultural activities of mountain huts inhabitants, artificial planting, cutting as a protective arrangement of the national park administration). Remote sensing is a suitable tool for Pinus mugo and other vegetation abundance development monitoring and using remote sensing valuable inputs for the future tundra vegetation management can be produced. The paper deals with the development of vegetation above the tree line in the Krkonoše mountains national park in the period between 1950 and 2013. We focused especially on the development of Pinus mugo abundance in tundra ecosystem evaluation. Two types of multispectral data were used for the assessment: 1) freely available Landsat surface reflectance images  from the years 1984, 1992, 2002 and 2013 with spatial resolution 30 m; 2) scened orthophotos from 1950s with the spatial resolution 50 cm and orthophotos including near-infrared band from 2012 with very high spatial resolution (12.5 cm). Four classes were classified from Landsat images: Pinus mugo scrub dense, Pinus mugo scrub sparse, bare land and other (closed alpine grasslands dominated by Nardus stricta, other grasslands, subalpine Vaccinium vegetation, alpine heathlands, wetlands and peat bogs). Two classification methods were used for Landsat images analysis: Maximum Likelihood Classification (MLC) and Support Vector Machine classifier (Radial Basic function; SVM - RBF).  Ortophotos were classified using object based classification (the example-based approach and SVM algorithm). Four classes were classified from orthophotos: Pinus mugo scrub, Picea abies stands, bare land and other (including same categories as for Landsat data). The classifications were performed for the whole area of tundra vegetation in KRNAP (ca 3,591 ha) in case of Landsat data, in case of orthophotos the classifications were performed only in the subset of eastern part of tundra on area of ca 1,200 ha. The first results from the Landsat images classification (MLC) shows that the Pinus mugo was the most extended class in the year 2002 and 2013, the area of Pinus mugo scrub sparse was 1447 ha (2002), 1352 ha (2013) and area of Pinus mugo scrub dense was 800 ha (2002) and 839 ha (2013). Minimal area was detected in the year 1992 (1098 ha Pinus mugo scrub sparse and 715 ha Pinus mugo scrub dense). The analysis of the Landsat images shows that in the period 1984 – 2013 the area of Pinus mugo in the western tundra decreased, perhaps because of the planed cut downs around Labská bouda. But Pinus mugo area in the eastern tundra was increasing throughout the whole period 1984 - 2013. This finding was confirmed based on orthophotos.  Classification outputs of orthophotos show that from 1950 till 2012 Pinus mugo area in the eastern part of tundra rapidly increased by almost 160 ha (13 %). The accuracy assessment of all results and the quantification of uncertainty will be added.

 

Acknowledgement: This research was made possible by the Charles University in Prague project GAUK No. 938214.  

Before Arab Spring revolution, Greater Cairo had been the focus of urban growth studies due to its huge increase of population in the last decades, also using remote sensing satellites. Probably, the change in its urban extent may have been affected by the Arab Spring activity since it started in 2011, as  it has been discussed by international organisations for which they are trying to get the answer to this phenomenon.

In order to analyse the urban changes in Greater Cairo from 2010 to 2015, this work uses the European Envisat and Sentinel-1A satellites as well as the American Landsat 7 and 8 for creating pre-revolution (2010) and post-revolution (2015) land use maps by combining the different SAR and Multi-Spectral (MS) sensors. The selected input parameters for creating the land use cover (LUC) maps are: (i) 4 bands with SAR amplitude statistics of one year SAR acquisitions, (ii) the interferometric coherence map with lower perpendicular baseline nearest to the MS acquisition date, (iii) 5 bands of the MS data. These LUC maps are created by using a supervised classifier based on neural networks, specifically the multi-layer perceptron, identifying only 4 general classes such as water, desert, urban and vegetated areas. For the 2010 LUC map it is used a combination of Envisat ASAR SLC data with Landsat 7 and for the 2015 LUC map, a combination of Sentinel-1A IW SLC with Landsat 8. By comparing the generated LUC maps, it is possible to identify the urban changes that occurred during the past 5 years, giving an answer to the question of the quantification of the urban increase in Greater Cairo.

In the future, with the new higher resolution satellites, it will be possible to create LUC maps with higher temporal and spatial resolutions, becoming an ideal tool for monitoring urban changes, fulfilling authority’s needs.

Earth Explorers are the backbone of the science and research element of ESA’s Living Planet Programme, providing an important contribution to the understanding of the Earth system.  Following the User Consultation Meeting held in Graz, Austria on 5-6 March 2013, the Earth Science Advisory Committee (ESAC) has recommended implementing Biomass as the 7th Earth Explorer Mission within the frame of the ESA Earth Observation Envelope Programme. This paper will give an overview of the overarching mission requirements, the satellite system and its payload.

The primary scientific objectives of the Biomass mission are to determine the distribution of aboveground biomass in the world forests and to measure annual changes in this stock over the period of the mission to greatly enhance our understanding of the land carbon cycle. The Biomass mission shall provide global maps of forest biomass stocks at a spatial resolution in the order of 4 ha, about twice a year over the life of the five-year mission. To achieve these objectives, the Biomass sensor will consist of a P-band (435 MHz) Synthetic Aperture Radar (SAR) in side-looking geometry with full polarimetric and interferometric capabilities

The Biomass space segment comprises a single low Earth orbit satellite platform carrying the SAR instrument. The SAR antenna is based on a large deployable reflector (12 m circular projected aperture) with an offset feed array and a single-beam. The spacecraft carrying the P-band SAR is planned for launch by Vega in 2020 and will operate in a near-polar, sun-synchronous quasi-circular frozen orbit.  The mission entails two mission phases, a tomographic phase during the first year of the mission which will provide one full global coverage, and an interferometric phase during the remaining four years of the mission life time providing full global coverage every six month. The orbit is designed to enable repeat pass interferometric acquisitions throughout the mission’s life and to minimise the impact of ionospheric disturbances.

This paper will present the mission and implementation status at Phase-B2 and all scientific support activities.

L-band radiometry is a prominent technology for remote sensing of surface soil moisture from space. Currently two satellite missions, European Space Agency’s (ESA) Soil Moisture and Ocean Salinity (SMOS) satellite, launched in 2009 [1] and Soil Moisture Active and Passive (SMAP) satellite of National Aeronautics and Space Administration (NASA) in 2015 [2] utilize a 1.4 GHz radiometer.

At the spatial scale of SMOS (35 km for nadir observations) the Earth’s surface typically includes several types of land usage and vegetation cover in one resolution cell. Accuracy of soil moisture estimate based on remote sensing is decreased e.g. by the existence of several land cover types within a radiometer pixel (mixed pixel problem). For such mixed pixel areas the inversion algorithm used to extract soil moisture from SMOS measurements often has to take into account several surface types, which complicates the parameter retrieval as well as validation and calibration of the results. [3, 4, 5]

This paper is to address the mixed pixel problem, namely differences in moisture distribution between agricultural and forested areas and their effects to brigthness temperature, through two complementary approaches: airborne measurements with HUT2D interferometric radiometer [6] and soil moisture simulations with soil moisture model ofWatershed Simulation and Forecasting System (WSFS) [7], augmented with small-scale in-situ measurements. The Watershed Simulation and Forecasting System
(WSFS) is a hydrological model used e.g. in climate change research and flood forecasting in Finland [8, 9].

The Hypöistenkoski catchment, located in the upper part of the Aurajoki watershed in Southwest Finland was chosen as a test site for this experiment as the area offers a possibility to combine airborne measurements with both calibrated soil moisture data by the WSFS and in situ measurements of soil and vegetation. The area has also been used as a test site for SMOS level 2 data validation [10].

The experiment was conducted between August 2013 and May 2014. Airborne measurements were conducted with HUT2D, a 2-dimensional synthetic aperture radiometer, an airborne L-band interferometric radiometer operating at the same frequency (1.4 GHz) as MIRAS, and designed, manufactured and operated by Helsinki University of Technology (now Aalto University) [6]. Simultaneously with the flights soil and humus layer moisture and temperature were measured on the ground in six subsites.

The WSFS data and in situ measurements were combined to produce a map of soil water content over the test area. The simulated soil moisture map was compared to soil moisture retrieved from HUT2D measurements and consequent SMOS measurements. Our results show that water content of different land usage classes exhibited significant spatial differences, which could be observed in high resolution HUT2D images. We also show that the accuracy was improved by taking soil properties into account for accordingly, when HUT2D data was processed to retrieve moisture of a mixed pixel according to the SMOS DGG grid.
REFERENCES
[1] K. D. McMullan, M. A. Brown, M. Martin-Neira, W. Rits, S. Ekholm, J. Marti, and J. Lemanczyk, “SMOS: The payload,” IEEE Transactions on Geoscience and Remote Sensing, vol. 46, no. 3, pp. 594–605, March 2008.
[2] D. Entekhabi, E.G. Njoku, P.E. O’Neill, K.H. Kellogg, W.T. Crow, W.N. Edelstein, J.K. Entin, S.D. Goodman, T.J. Jackson, J. Johnson, J. Kimball, J.R. Piepmeier, R.D. Koster, N. Martin, K.C. McDonald, M. Moghaddam, S. Moran, R. Reichle, J.-C. Shi, M.W. Spencer, S.W. Thurman, Leung Tsang, and J. Van Zyl, “The soil moisture active passive (smap) mission,” Proceedings of the IEEE, vol. 98, no. 5, pp. 704–716, May 2010.
[3] T. Pellarin, J.-P. Wigneron, J.-C. Calvet, and P. Waldteufel, “Global soil moisture retrieval from a synthetic l-band brightness temperature data set,” Journal of Geophysical Research, vol. 108 (D12), pp. 4364, 2003.
[4] A.A. Van de Griend, J.-P. Wigneron, and P. Waldteufel, “Consequences of surface heterogeneity for parameter retrieval from 1.4-GHz multiangle SMOS observations,” IEEE Transactions on Geoscience and Remote Sensing, vol. 41, no. 4, pp. 803–811, April 2003.
[5] Rachid Rahmoune, Paolo Ferrazzoli, Yann H. Kerr, and Philippe Richaume, “Testing a new model for the L-band radiation of moist leaf litter,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 6, no. 3, pp. 1430–1439, June 2013.
[6] K. Rautiainen, J. Kainulainen, T. Auer, J. Pihlflyckt, J. Kettunen, and M. Hallikainen, “Helsinki University of Technology L-band airborne synthetic aperture radiometer,” IEEE Transactions on Geoscience and Remote Sensing, vol. 46, no. 3, pp. 717–726, March 2008.
[7] B Vehviläinen, “Snow cover models in operational watershed forecasting,” Publications of water and environment research institute, vol. 11, 1992.
[8] B. Vehviläinen, M. Huttunen, and I. Huttunen, “Hydrological forecasting and real time monitoring in Finland: The watershed simulation and forecasting system (wsfs),” in Innovation, Advances and Implementation of Flood Forecasting Technology, conference papers, Tromsø, Norway, 17-19 October 2005, 2005.
[9] N. Veijalainen, Estimation of climate change impacts on hydrology and floods in Finland, Doctoral dissertation, Aalto University, 2012.
[10] J. Jakkila, T. Vento, T. Rousi, and B. Vehviläinen, “Smos soil moisture data validation in the aurajoki watershed, finland,” Hydrological Research, vol. 45, no. 4-5, pp. 684–702, 2014.

Institute of Geodesy and Cartography (IGiK) possess extensive experience in application of medium and HR satellite data for the agriculture and energy biomass sectors. IGiK conducts two projects, covering the above subjects, which are financed by the European Space Agency.

Project ASAP – “Advanced Sustainable Agriculture Production – satellite data your tool in agriculture management” (01.09.2015 – 31.08.2015), being developed under IAP ARTES 20 Programme, aims at development of Complex Service for agriculture management, based on actual information derived from satellite data. The Service is designed for individual Users – Farmers applying precision farming as well as companies supporting and supplying agriculture market and administration.

SERENE “Bioenergy as the key to economic growth of the regions - EO Based Service Supporting Energy Crops Cultivation” is funded under 1^st Call under Polish Industry Incentive Scheme. The project aims at development of the service for energy crops plantations monitoring – biomass production and energy estimation on the basis of satellite data.

In both projects, high spectral, spatial and temporal resolution satellite data are very demanded.

SPOT 5 data, obtained through the SPOT (Take5) experiment, has been successfully applied for both projects in order to deliver the following products:

SERENE:

¾    classification of the energy crops (Figure 1) – delimitation of the energy crops plantations (corn and poplar) (Figure 2)

¾    monitoring of the identified energy crops plantations:

monitoring of energy crops conditions, applying various indices based n Infrared and SWIR. Figure 3 presents the distribution of NDVI which increases during the growing season;

detection of plantations spots with lower biomass production (Figure 4)

ASAP:

¾    delimitation of the fields - boundaries of the fields with single crop (Figures 5, 6)

monitoring of the selected fields with NDVI index (Figures 7, 8)

The above listed products will be subsequently expanded. The models for biomass increase (between following months) as well as biomass estimations and energy productivity for the corn and poplar plantations will be run. The decrease in potential biomass in result of lower production in several spots (caused by incorrect planting or game etc.), classified on SPOT 5, are estimated. It is very important for the Users to know the status of the plantations and percentage of damages. The comparison between the products obtained from Landsat 8 and SPOT 5 will be presented. The Report on SPOT 5 applicability for energy sector will be presented to the National Authorities, such as Energy Regulatory Office and Agency of Restructuration and Modernization of Agriculture (ARMA) as well as widely disseminated through the cooperating webportals: farmer.pl and reo.pl, very well recognized in the Energy Sector in Poland.

Under the ASAP Project, the following products will be subsequently delivered: the classification of crops, delimitation of homogenous polygons (comparison with the results of electromagnetic scanning performed for these fields by the IGiK partner in ASAP), early prediction of yield furtherly adjusted to the actual conditions on the basis of SPOT 5  supported by Landsat 8 (surface temperature) and NOAA.AVHRR data (drought prediction and monitoring with model developed by IGiK) and yield estimation. These products will be delivered on the basis of archive data, but the modelled delivered will be used for the Sentinel 2 based models development.

The SPOT 5 data delivered under the SPOT (Take5) experiment are found to have significant potential for both energy plants and agriculture crops monitoring. It is then anticipated that Sentinel 2 data with higher spectral resolution will be widely used in both above projects and will be prime data of the ASAP Service in the future.

Soil moisture retrieval from remote sensing data is very useful for a number of applications, and even specific missions have been devised to this aim. When high-resolution soil moisture maps are needed, use of polarimetric SAR data is the obvious choice. However, soil moisture retrieval from SAR data is not an easy task, especially in presence of a vegetation cover, because radar return depends not only on the soil dielectric constant (and hence soil moisture) but also on several other parameters describing soil roughness and vegetation. Accordingly, in recent years some methods for soil moisture retrieval under vegetation cover have been developed [1-4]. They rely on model-based or hybrid polarimetric target decomposition techniques. Many of these decomposition techniques, in their original formulations, suffer from the so-called negative power problem, which is due, on one side, to poor modelling of surface scattering, so that the whole cross-polarization effect is attributed to volumetric scattering, which is thus overestimated; and, on the other side, to the poor modelling of the vegetation scattering contribution itself. The approach of [1-2] (named Polarimetric Two-Scale Two-Component Model, PTSTCM) focuses on the former problem and tries to solve it by using a more refined surface scattering model [5] that accounts for de- and cross-polarization due to surface roughness; the price to be paid is the need of ignoring double bounce contributions and still using a simplified vegetation scattering model. Conversely, the approach of [3-4] (named Iterative Generalized Hybrid Decomposition, IGHD) focuses on improving the modelling of vegetation scattering, at the cost of still using a simplified, non-depolarizing ground scattering model. It turns out that PTSTCM provides the best results for moderately vegetated fields (vegetation height lower than 50 cm, or cross-polarized ratio smaller than 0.1, and negligible double-bounce component [2]), whereas IGHD provides the best results in the other cases and shows a wider range of validity [2-4].

In this work we first of all recall and analyse the results of the above mentioned two methods, and show a method to combine them by choosing pixel by pixel in an adaptive way the more suitable one, based on the values of the cross-polarized ratio (or, if known, of the vegetation height) and on the signum of the real part of the co-polarized correlation. It turns out that the combination of the two approaches covers most of the vegetation cover conditions. The only critical situation is the case of dominant surface scattering and secondary, non-negligible, dihedral component. Then, we propose a method to try to fill this gap, to be used when the co-polarized correlation coefficient is significantly smaller than unity and the cross-polarized ratio is very small, so that the decreased correlation coefficient is not justified by roughness or volumetric effects and it is most likely due to the dihedral component. In this case, we suggest that the more refined surface model of PTSTCM is firstly used to compute the volumetric component, so that the latter is not overestimated. Then, with this estimate of the volumetric component, soil moisture can be retrieved by recurring to one of the usual model-based or hybrid decompositions.

References

[1] A. Iodice, A. Natale, D. Riccio, "Soil moisture retrieval in moderately vegetated areas via a Polarimetric Two-Scale Model", Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, pp. 759-762, Melbourne, Australia, 21-26 July 2013.

[2] G. Di Martino, A. Iodice, A. Natale, D. Riccio, "Polarimetric Two-Scale Two-Component Model for the Retrieval of Soil Moisture under Moderate Vegetation via L-Band SAR Data", submitted to IEEE Transactions on Geoscience and Remote Sensing.

[3] T. Jagdhuber, I. Hajnsek, and K.P. Papathanassiou: “Refined Soil Moisture Estimation by means of L-Band Polarimetry,” Proc. of IEEE International Geoscience and Remote Sensing Symposium, pp. 2325-2328, July 21-26, Melbourne, Australia, 2013.

[4] T. Jagdhuber, I. Hajnsek, and K.P. Papathanassiou: “An Iterative Generalized Hybrid Decomposition for Soil Moisture Retrieval Under Vegetation Cover Using Fully Polarimetric SAR", IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol.8, no.8, pp. 3911- 3922, 2015.

[5] A. Iodice, A. Natale, D. Riccio, "Retrieval of Soil Surface Parameters via a Polarimetric Two-Scale Model", IEEE Transactions on Geoscience and Remote Sensing, vol.49, no.7, pp. 2531-2547, 2011.

Geophysical parameters derived from space-borne Earth Observation Systems are either assigned to discrete points on a fixed Earth grid (e.g. regular lat/lon grid) or located on orbital point nodes with a customized arrangement, often in-line with the instrument's measurement geometry. The driving factors of the choice and structure of the underlying grid are typically spatial resolution, instrument geometry, instrument measurement technique and application. In this context, the most practical solution for the choice of the sampling distance represents supporting the Nyquist–Shannon sampling theorem. In other words, the spatial sampling rate should be half of the spatial resolution establishing a sufficient condition to capture all the information from the continuous-spatial signal.

In this study we propose a new global fixed Earth grid, the so-called Equi1 grid, and demonstrate its structure, usability and field of application. The grid construction follows the aforementioned condition of the Nyquist-Shannon sampling theorem, by applying a nearly-constant spatial sampling rate. The general appearance is similar to a Reduced Gaussian Grid, which represents a gridded coordinate system often used for scientific modeling on the Earth surface. The properties of the Equi1 grid are compared to common Earth observation grid systems (e.g. EASE-grid, Gaussian Grid) discussing the advantages and disadvantages. Finally, a real-world example will be presented based on surface soil moisture information derived from MetOp ASCAT.

Food demand and processes of land conversion from natural to cultivated land have been accelerating in East Africa over the past decades. Land cover information of the region at reasonable spatial scales is often inconsistent, outdated, incomplete or not available.

The Kilombero floodplain is a huge wetland that experiences pressure from increasing agricultural use and land use changes. Traditional crop is rain-fed rice. Recently, cultivation of cash crops such as sugarcane, and other commercial land uses such as teak plantations increased. Irrigation recently appeared in parts of the floodplain. The floodplain is surrounded by mountain ranges. The mountains are dominated by forests whereas the flat floodplain is dominated by agricultural land, savanna and grassland.

Remote sensing is a valuable means to assess land cover and land use changes since the datasets cover the most dynamic period of the last decades. Inconsistent datasets with gaps due to frequent cloud coverage are common for the tropics. However, compositing of moderate resolution data such as Landsat (Griffiths et al. 2013) is an option to generate cloudfree datasets. Most of the compositing approaches select cloudfree observations that are closest to a predefined day of year (DOY). In case no appropriate observation is available, data from other years are considered as well. In the present study we make use of an alternative compositing approach. Instead of selecting observations we consider all cloudfree observations of predefined periods and calculate multitemporal metrics such as maximum, minimum, mean, and median per pixel and per band. The approach is applied on Landsat and RapidEye data of the Kilombero floodplain, Tanzania. Landsat data are used to generate land cover and land use maps for 1984, 1994, 2004, and 2014. Each period refers to at least one water year to allow for the assessment of flood extent and non-permanent land cover classes such as the flooded grasslands. For a local scale study in the vicinity of the city of Ifakara we used RapidEye data from 2013 to 2015. The multitemporal metrics of each period were subjected to a supervised Random Forest classification. Training samples were taken during a field trip in 2015. Data of areas that haven’t changed over time were used as training input for the classification of the historical datasets. Results confirm that land cover and land use changes accelerated over the past decade. Only remote and inaccessible areas remained unchanged. The anthropogenic impact spreads from the settlements into the floodplain and in the adjacent mountains.

Palaeochannels are dried up old rivers, and are important sources of groundwater, economic minerals and placer deposits. Investigating palaeochannels through traditional ground survey based methods is resource intensive, whereas remote sensing approaches can help us in better and more efficient delineation of such features. Exploiting such resources as palaeochannels through remote sensing would involve their identification / mapping for subsequent exploration and water management utility. The Cholistan desert is the largest desert in the Punjab province of Pakistan, and lies along the eastern border of Pakistan with India. This paper presents a study in which multi-sensor SAR (Envisat ASAR) and Optical (Landsat 8 OLI / TIRS) images of the Cholistan desert of Pakistan were processed and analyzed to identify and map Hakra River palaeochannels. C-band VV-polarization Ellipsoid Geocoded Intensity images SAR data from Envisat ASAR was used in an overall dry season and the dry sand penetration properties of SAR data were utilized in this study to get backscatter from the buried Hakra palaeochannel under the desert sand. After pre-processing of both SAR (calibration, mosaicking, speckle filtering, etc.) and optical data (mosaicking, calibration, atmospheric correction, etc.), image fusion using PCA technique was performed with Landsat 8 data to identify and delineate the Hakra river palaeochannel in the study area. The generated results from this data processing and analysis are shown in the attached images. From the results, our preliminary deductions are that upper segment of the palaeochannel lying in northern portion of Cholistan desert still possesses freshwater in abundance due to its coexistence with Punjab canal system and underground recharging from perennial channel of old Indian Saraswati river. Future work entails validation of the identification and alignment of identified palaeochannel by comparisons with geophysical ground measurements (electrical resistivity & conductivity surveys) and historical evidences.

In recent years, water level variations of inland water bodies such as lakes, reservoirs, and rivers measured by satellite altimetry got well established. Most inland water level time series are only assembled from measurements of one pass of one single satellite mission. Only a few multi-mission approaches combine different missions and passes over lakes and reservoirs in order to increase the accuracy and temporal resolution of the time series. This is possible because the lake surface can be considered to be constant everywhere at a given time. However, it is not possible so far to combine different altimeter missions and passes over rivers.

We developed a new methodology to combine altimetry data from different missions in a statistical robust way along the river. The methodology is based on kriging which is an interpolation method originating from geostatistics. We expanded the concept to spatio-temporal kriging along the river. The interpolation is a weighted average of available measurements based on empirical correlations not only in the spatial domain but in the temporal domain as well. The higher the correlation, the more weight a measurement obtains in the average. With this approach we are able to combine data not only along the river at a given time or a given location but also data at another location at another time. We developed an advanced statistical model to describe the dependencies between different measurement locations; a prerequisite for the kriging algorithm. These statistical models reflect the changing flow velocity of the river as well as dependencies caused by the catchment areas.

We employed the kriging method on altimeter measurements of the Mekong River in Southeast Asia. We incorporated data of the Envisat, Jason-2, SARAL/AltiKa, and Cryosat-2 mission. With this we are able to achieve a higher temporal resolution time series at any given location. The kriging method smooths the resulting time series for which reason outliers are detected and can be removed. This allows to better judge the quality of each individual time series, e.g., whether they fit with other measurements.

And finally we can combine data from different altimeter missions with which we are able to close the data gap between the end of the Envisat and the start of the SARAL/AltiKa mission with Cryosat-2 data. Cryosat-2 has a repeat time of only 369 days but a dense spatial distribution and so far could not be used to estimate water level time series of rivers. The resulting estimated time series are compared to in-situ data from gauging stations along the river and show a high agreement with these.

Multispectral and Hyperspectral image understanding plays a key role numerous areas ranging from climate change tracking, to security and surveillance, fuelling the research in the introduction of machine learning approaches in image pixel classification. Typically, such pixel classification schemes exploit the high spectral content of such images in order to assign each pixel to the single most representative class. In order to achieve this goal, supervised learning algorithms are employed, where features extracted from labeled training examples are utilized in order to generalize the appearance of specific classes and correctly predict future examples. In general, the performance of the classification process primarily depends on two factors, namely the learning capacity of the classifier and the characteristics of the extracted features.

The effects of the feature extraction process are particularly evident in computer vision tasks, where carefully designed, hand-crafted features, such as Scale Invariant Feature Transform (SIFT) and Histogram of Oriented Gradients (HOG) have shown great effectiveness in a variety of tasks. Despite their impressive performance, the main drawback of these descriptors is that significant human intervention is required during their design. Furthermore, such features are highly domain-specific and have limited generalization ability. This motivates the need for efficient feature representations extracted automatically from raw data through Representation Learning, a set of techniques which aim to learn useful (i.e. discriminative, robust, smooth) representations of the input data for use in higher level tasks such as classification and recognition, minimizing the dependency of learning algorithms on feature engineering.

In this work, we investigate feature learning as a preprocessing step which seeks to identify the underlying explanatory factors hidden in low-level multi and hyperspectral satellite data for the purpose of multi-label classification, where each pixel may belong to multiple classes simultaneously. We focus on a particularly successful unsupervised feature learning approach of Sparse AutoEncoders (SAE), a type of artificial neural network which employs nonlinear codes and imposes sparsity constraints for representing the input data. While theoretically a one-hidden-layer ANN can approximate any function to arbitrary precision, this approach becomes impractical due to need for exponential increase in the number of the units.

Inspired by the human cognitive system which exhibits a hierarchical structure and learns in a layer-wise fashion, Deep Learning architectures seek to incorporate depth into learning algorithms, which allows them to extract compact and increasingly more abstract representations. Nevertheless, despite the encouraging theoretical results, in practice, up until very recently it has been impossible to train sufficiently deep architectures, since gradient-based optimization methods starting from random initial weights tended to get fixated near poor local optima. Deep learning was revolutionized in the past few years, when the strategy of greedy layer-wise unsupervised ``pretraining'' followed by supervised fine-tuning was introduced.

In this work we concentrate our investigation on the effects of deep learning using Stacked SAE on the performance of land cover map estimation using time-series of multi and hyperspectral imagery. To achieve this goal, we consider real data from the CORINE dataset of the European Environmental Agency and investigate the prediction accuracy, using available imagery from different instruments including MODIS, EO-1 Hyperion and Senstinel-2 MSI. 

Mapping and monitoring vegetation biophysical and biochemical parameters is valuable for biodiversity modeling, but also for a diversity of other applications including agricultural, ecological, and meteorological studies. The spatial and temporal distribution of vegetation parameters are used as inputs in models quantifying the exchange of energy and matter between the land surface and the atmosphere. Among the many vegetation parameters, leaf area index (LAI) and chlorophyll are of prime importance as they provide vital information for biodiversity assessment and have been proposed as Essential Biodiversity Variables (EBVs). Significant efforts to estimate and quantify these parameters using Radiative Transfer Models (RTM) have been carried out in the last two decades. However, most of these studies rely on hyperspectral measurements. Although hyperspectral data have demonstrated to be accurate and suitable for retrieval of vegetation parameters, they are usually expensive and not frequently available. New generation of multispectral sensors with high spatial resolutions such as Sentinel-2 and RapidEye which provide spectral information at the red-edge region have provided further opportunities for estimating these parameters. As part of the ESA: DUE, Innovators-III funded project RS4EBV (Remote Sensing for Essential Biodiversity Variables), the retrieval of LAI and chlorophyll is investigated utilizing RapidEye images and different RTMs. Several RapidEye images were acquired between March and September 2015, for each of the three study sites of the project, i.e. 1) the Bavarian Forest National Park, 2) the National Park of Schiermonnikoog in the north of the Netherlands, and 3) the North Wyke Farm Platform in Devon, UK. In situ measurements of a large number of biophysical and biochemical parameters including LAI and chlorophyll were collected concomitant with the time of image acquisitions. The widely used canopy radiative transfer models: PROSAIL (SAILH and the PROSPECT) and INFORM (Invertible Forest Reflectance Model), were investigated for retrieval of LAI and chlorophyll in grasslands (including the saltmarsh) and forest study sites, respectively. The two RTMs were first parameterized based on the spectral band settings of RapidEye. Consequently, large look-up tables (LUTs) were generated for each study sites accounting for available prior information related to the distribution and range of the vegetation characteristics in each site. The LUTs were then inverted using the spectral reflectance obtained from the images. To assess the performance of the model inversion and analyze the suitability of the models, the normalized RMSE and R² between independent in situ measurements and estimated parameters were used. Our results will demonstrate the potential and drawbacks of model inversion for estimating vegetation biophysical and biochemical parameters at three different types of biomes in Europe using new generation of multispectral satellite data.

Canopy radiative transfer (RT) models are computer programmes simulating interactions between canopy elements and incident light of visible and near infrared wavelengths. Since they provide a physical link between canopy biochemical and structural properties and canopy reflectance, they are often used to interpret remote sensing (RS) data for vegetation properties. Still, retrieval of quantitative properties of coniferous forests can be a challenging task due to the complex canopy and shoot architecture that significantly modulates forest light scattering, particularly in the near infrared spectral domain.

In this study we used the Discrete Anisotropic Radiative Transfer (DART) model, which is capable of simulating complex 3D forest scenes, including Norway spruce (Picea abies /L./ Karst.) canopies, with a very high detail. However, when we used default tree 3D models of DART, which are based on simplified turbid cells, we yielded simulated top-of-canopy bidirectional reflectance factor (BRF) that was not readily comparable with reflectance of real RS images. One of the possible reasons is insufficient simulation of light scattering between single needles within shoots. Computation of light scattering within the turbid tree model uses the Leaf Angle Distribution (LAD) that describes angularity of shoots without taking into account the actual angular distribution of needles. This insufficient modelling of the light scattering in-between needles impacts negatively accuracy of simulated canopy BRF. Thus we hypothesize that an improvement of RT reflectance modelling of spruce canopy in DART could be achieved by using a more realistic facet-based 3D tree model instead of the default turbid one.

The process of building a 3D spruce tree model was divided into three phases: 1/ creation of 3D shoot model, 2/ creation of basic tree wooden structures (i.e. stem and main branches), and 3/ distribution of green biomass (i.e. 3D shoot models) within a tree crown. Several shape types of shoots are typically present in a single tree. We, therefore, built three shoot types representing different spatial formations of needles along twigs. We used an automatic algorithm forreconstruction of a spruce tree wooden structure that is based on ground laser scanning (LiDAR) data. 3/ Distribution of green biomass (i.e. 3D shoot models) into a tree crown. We developed new algorithm based on L-systems theory using the same laser scanning data from the previous phase. Distribution of shoots within a tree crown follows morphological appearance of Norway spruce, i.e. it is dependent on age, vertical position, light distribution regime inside crown, distance from the stem, random natural disturbances, etc. These parameters have also dictates distribution of different types of needle shoots in the whole crown. Such an ecologically correct 3D spruce representation is expected to reduce the existing mismatch between DART simulated and measured RS BRF of spruce canopies.

One of the main challenges of hydrological modeling is the poor spatio-temporal coverage of in situ discharge databases, which are declining steadily over the past few decades. It has been demonstrated that altimetry-derived water height over rivers can sensibly be used to deal with the growing lack of in situ discharge data. However, the altimetric discharge is often estimated from a single virtual station suffering from data outages and high noise level. In this study we implement a stochastic process model to (1) deal with the data outages in altimetric discharge, (2) provide a scheme for data assimilation of multiple altimetric discharge and (3) smooth the discharge estimation. The model benefits from the cyclostationary behavior of the discharge and is combined with multiple altimetric discharge time series and available in situ measurements to form a linear dynamic system. The dynamic system is then solved using the Kalman filter that provides an unbiased discharge with minimum variance. Our process model comprises the covariance and cross-covariance information of altimetric and measured river discharge belonging to different gauges. We evaluated our method over Niger River, where we have access to in situ discharge data from GRDC and altimetric water level time series at different virtual stations along the river. We validated our method at 18 gauges along the Niger River against available in situ data. Our validation shows correlation coefficient of above 0.9 for all gauges, Nash-Sutcliffe efficiency coefficient of above 0.5 for 18 gauges and average relative RMSE of around 15%.

For many applications a circumpolar representative and consistent wetland map is required. It is important for upscaling of carbon fluxes and pools as well as for climate modeling or wildlife habitat assessments. However there is still a lack of sufficient accuracy or thematic detail in many arctic regions in currently available datasets. Synthetic Aperture Radar (SAR) data has already been shown to be suitable for wetland mapping and an easy way to use C-band SAR data in order to derive a circumpolar wetness classification map has been introduced in a previous study. By exploring circumpolar ENVISAT ASAR GM backscatter (1km resolution) data of arctic environments it could be found that winter minimum backscatter values reflect vegetation physiognomy units of certain wetness regimes. Low winter backscatter values are mostly found in areas vegetated by plant communities typically for wet regions in the tundra biome, due to low roughness and low volume scattering caused by the predominant vegetation.

Flux studies often distinguish between three classes: dry, mixed and wet. It could be shown before that the Envisat ASAR GM derived wetness classes (‘dry’, ‘medium’, ‘wet’ and ‘other’, which is characterized by very low backscatter found at flat surfaced or sandy soils) are related to wetland classes of conventional vegetation maps, indicating its applicability.

With heterogeneity being a major challenge in the arctic, higher resolution products are essential. In this study we therefore investigate the potential applicability of this approach at site scale using ENVISAT ASAR WS data (~120m resolution). These higher resolution ASAR WS maps have been produced for some study sites throughout the Arctic and compared to high resolution land cover maps.

For the study site of Kytalik it can be demonstrated that all fen classes correspond to at least 50% of the SAR product class ‘wet’. Sedge fens even include 70%. In case of the site Zackenberg more than 60% of the class ‘fen’ coincides with ‘other’. This class corresponds however to only 2.3% of land cover in Zackenberg. Constraints for the used approach can be found for the Yamal study site, where salix vegetation with stem diameters of more than 5 cm and heights over 1.5 m can be encountered. These shrub patches often feature higher soil moisture. However volume scattering leads to higher backscatter in these regions and are therefore classified as dry area.

Nevertheless it could be shown that a medium resolution C-band SAR based wetness level map can be derived for tundra regions where no scattering due to tree trunks hampers the used method.  

Accurate short and mid-term forecasting of the flow generated from snowmelt in mountain basins is an essential component for water management and hydropower activities in several areas at a global scale. Due to the remoteness and difficult access to basins in certain cold weather conditions, where in-situ snow monitoring and meteorological stations are scarce, flow model predictions are not easily implemented.

INTOGENER (INTegration of EO data and GNSS-R signals for ENERgy applications) feasibility and demonstration are projects funded by the European Space Agency (ESA) under its ARTES 20 / IAP programme, aiming at delivering improved streamflow predictions in specific points of interest of remote mountain areas.

INTOGENER aims at improving the water flow predictions in remote mountainous catchments were most of the hydroelectric production is located. For this purpose the service concept developed by Starlab and Future Water uses near real-time geophysical information from in situ and remote sensing space based instruments to feed a hydrological model producing water flow forecast.

The in situ measurements identified as relevant for flow prediction are: temperature, precipitation, solar radiation, water level, soil moisture. Reservoir water level is estimated using a novel technology based on the measurements of Global Navigation Satellite System (such a s GPS and Galileo) signals reflected by the lake surface (GNSS-R). If required, in situ information can be sent by satellite communications, in order to provide real-time data to the model.

The space based Earth Observations identified as relevant for flow prediction are temperature maps and snow cover. Snow cover and temperature maps can be measured by optical instruments, with the limitation of cloud cover that occurs regularly during the melting season, and coarse resolution not suitable for hydrological modelling. Whereas temperature can be completed by in situ measurement, extrapolated and downscaled using a digital elevation model, snow cover is more difficult to downscale and extrapolate. For this reason the proposed service includes the acquisition of SAR images which are cloud immune and have a resolution suitable for the hydrological model. Using state of the art algorithms, optical and SAR images can be fused in order to produce high resolution realistic maps of the snow cover.

The hydrological model is initialized with digital elevation model, land use and soil type maps, and calibrated using historical meteorological and flow data. It is then capable of producing short term (days) to long term (seasons) water flow forecasts at points of interest by taking as starting points an updated state of the basin (from updated geophysical information) and a climatology modulated by El Niño Southern Oscillation (ENSO) index as prediction of the meteorological variables.

All the information (from in situ instruments, satellite observations, historical data, and model output) is centralized on a secure server running a database as well as services required for data flow and automation. Routines on this server will produce water flow reports, including quality indicators, as well as system health indicators available by the service team.

The output of the service is a complete and much improved forecast of the water-flow based on near real time physical observations delivered automatically to the user on a weekly basis. This forecast is computed at points of interest of the basin, that match input variables in the operational model, and therefore usable directly in their management practices.

Rapidly changing urban agglomerations in developing and emerging countries face social and environmental challenges. Urban population growth and socioeconomic developments manifest in changes in the existing built-up areas as well as in the land use of the urban-rural transition zone. The City of DaNang is the largest city in Central Vietnam with a population of 1,046,876 (2015). Its population is predicted to reach 1.6 million by 2020. In the years 1997 to 2013, the city’s economy grew at an average rate of 11 %, resulting in significant investments in the cities infrastructure. The peri-urban districts of the city are experiencing a rapid land conversion. Strong touristic development leads to dynamic building activity along the eastern beach of the city. Industry and technology parks arebeing created. Flooding and natural disasters, as well as a strong investment in mid- to high-income development projects led to the resettlement of poor households. While the pressure on undeveloped land is increasing, the city administration aims to protect and manage vulnerable ecosystems, such as the Ba Na forest zone and the Son Tra peninsula. Hence, qualitative and quantitative information on the land use development is necessary to provide reliable and continuous data for spatial planning processes. Multi-temporal RapidEye data provide the possibility of change detection of the land use and monitoring of protected areas.

Five RapidEye scenes (05/05/2010; 04/20/2012; 04/04/2013; 03/02/2014 and 04/02/2015; Level 3A) are the initial data for the proposed study. Relative radiometric normalization was conducted on the scenes using the Iteratively Reweighted-Multivariate Alteration Detection (IR-MAD; Canty & Nielsen 2008) method. The algorithm selects invariant pixels for bi-temporal image pairing. The RapidEye scene acquired in 2010 served as the reference scene. Subsequently, an object-based image segmentation was applied to each scene. The objects were then classified based on both their spectral properties as well as their spatial and shape-related attributes to distinguish between 9 classes (water, dense and sparse vegetation, clouds, built-up area, roads, construction sites, beach, river banks, agricultural and undeveloped land). For the most part, expert knowledge was applied in a rule-based classification approach. Due to highly similar spectral properties, the distinction between some very bright areas of undeveloped land and built-up area proved to be difficult. Here, a support vector machine (SVM) algorithm was applied. For each individual scene, training samples were selected to train the algorithm to distinguish between both classes. For the detection of change, a post-classification, pixel-based approach was used to quantify land use changes from one class to another on a per area basis between each consecutive year and for the entire period between 2010 and 2015.

The transferability of the classification from the reference image to the subsequent images is robust, but needs some adjustments. Although the applied radiometric normalization procedure improves the comparability between scenes, the persisting spectral variation of the same class in different scenes requires the rulesets to be individually modified. This might be largely due to seasonally dependent vegetation cover at the time of image acquisition, ranging from March to May. The change detection reflects the general developments of the city. It identifies clearly the investments in infrastructure and the residential expansion. The main land use change is from agricultural managed and undeveloped to built-up areas.

About 19% of the Republic of Ireland can be associated with upland biomes, accounting for several priority Annex 1 habitats under the EU Habitats Directive (Council Directive 92/43/EEC). Most of these large semi-natural uplands have not been sufficiently mapped or described and are under pressure due to land abandonment and grazing management, culminating in scrub encroachment and biodiversity loss. Regular Earth Observation (EO) monitoring of these extensive upland areas can therefore support biodiversity conservation and land management, as well as carbon storage accounting, even at small-scales. However, good quality high resolution EO data are limited by persistent cloud contamination of upland areas, introducing difficulties in capturing seasonal variabilities of vegetation using spectral information in the optical and NIR domain.

In order to overcome these challenges, one cloud free (2010-04-12) and two cloud contaminated (2010-05-23 and 2010-08-15) 5m spatial resolution RapidEye scenes were acquired over the Galtee Mountains (approx. 54 km²) in the south-east of Ireland. The objective of this study was to determine a robust approach to discriminating upland vegetation types by carrying out a comparison between object-oriented and pixel-based Random Forest (RF) classification approaches. The parameter setting of the region growing segmentation i.segment, as implemented in the open-source GRASS GIS software, was iteratively optimised using an accuracy assessment within a semi-automated process-chain. Field survey data, provided by the National Parks & Wildlife Service (NPWS) was used for the RF model calibration and validation. In addition to a multi-temporal time series, all single acquisitions were processed using combinations of different derived variables, including vegetation indices and texture measures, as well as ancillary data (elevation derived information, soil and bedrock data).

72 different combinations of variables were analysed, resulting in Overall Accuracy (OA) values ranging between 54.2 and 95.6% (with corresponding Kappa () coefficients ranging from 0.49 to 0.95). The inclusion of ancillary datasets improved the classification accuracies between 6 and 27% for the pixel-based approach, but for the object-based scenarios the improvement is less or can even cause greater confusion. Without considering ancillary information, the segmentation approach outperforms the pixel-based results for single acquisitions, however for the time series, the OA and  results are very similar. These results were supported for the single date acquisitions by the McNemar’s test (e.g. single image acquired on April 2010, 12.7 ≤  ≤  32.7,  0.0001 ≤   ) but not for the time series results ( 4.9 ≤  ≤  11.6,  0.0001 ≤    ≤ 0.008). When comparing the pixel-based time series results with the results provided by the single cloud-free (April 2010) acquisition, the multi-temporal approach improved the OA by about 5%, which is confirmed by the McNemar’s test. For the same variable combinations, the object-oriented time series and single acquisition (April 2010) revealed marginal differences in OA and , which are not confirmed by a statistically significant difference.

The results indicated the benefit of a multi-temporal approach to overcoming persistent cloud contamination in upland areas, and, from an accuracy perspective the unimportance of the classification strategy. The comparison of the single date pixel- and object-based results suggests that the dense bracken class, which acts as an indicator habitat for land abandonment in the uplands, has more realistic boundaries in the segmented output. However, the more natural look of the segmented time series is offset by artefacts introduced by the cloud mask and loss of transition zones. Thus, given the increased calculation effort required for segmentation of the time series, it would be recommended to use a pixel based approach. However, when only a single, cloud-free image is available the segmentation approach proved the more appropriate strategy as less importance was attached to the spectral signature of a single pixel.

Accurate and spatially distributed estimates of evapotranspiration (ET) are increasingly important with the growing global population and economy putting strain on fresh water resources and food supplies. The utility of ET maps has been demonstrated in a variety of applications, ranging from water rights management, through drought and food shortage monitoring, to more efficient use of land and water in agriculture. The most efficient methods for deriving such maps make use of remote sensing observations, particularly from satellite-based sensors since they provide a synoptic view at regular time intervals. Majority of remote sensing models that estimate actual (as opposed to potential) ET do so through the estimation of land surface energy fluxes. This requires both knowledge of the state of vegetation (e.g. leaf area index or fraction of green vegetation) and of land surface temperature (LST) which acts as one of the model boundaries. Therefore, remote sensing observations using both visible/near-infrared and thermal infra-red sensors serve as an important source of input data for such models.

The Two-Source Energy Balance (TSEB) modelling scheme, which splits the land surface energy fluxes between canopy and soil components, has been widely used with satellite derived observations of LST. When LST observation from only one angle is available the model makes an initial assumption of vegetation transpiring at potential rate. The accuracy of outputs derived with this assumption depends on accurate estimation of radiation available for transpiration which in turn requires estimates of fraction of vegetation that is green (f_g). However, there does not yet exist a well-established methodology for estimating f_g from satellite observations. On the other hand, if LST observations from two or more angles are available then it is possible to solve TSEB equations without making any assumptions about canopy transpiration. However, in that case it is important to accurately model the transfer of thermal radiation through the canopy while accounting for such parameters as leaf angle orientation.

The recent launch of the Sentinel-2 and up-coming launch of Sentinel-3 satellites present new opportunities for accurate and operational modelling of actual ET with the use of remote sensing data. In particular the MultiSpectral Instrument (MSI) on board Sentinel-2, with its high spatial resolution and red-edge spectral bands, could allow for accurate retrieval of parameters such as leaf area index, fraction of vegetation that is green or surface albedo. On the other hand, the Sea and Land Surface Temperature Radiometer (SLSTR) thermal sensor on board Sentinel-3, with its unique dual angle LST observation capability, could be utilised for obtaining a more accurate split between vegetation transpiration and soil evaporation.

In this study we propose to couple the PROSPECT leaf radiative transfer model and the 4SAIL canopy radiative transfer model to TSEB evapotranspiration modelling scheme with the aim of 1) deriving f_g by inverting 4SAIL and PROSPECT models driven by multispectral observations from the MSI sensor, and 2) inverting 4SAIL model driven by dual-angle LST observations from the SLSTR sensor and thus obtaining the canopy and soil temperatures needed in TSEB.

We will compare the two TSEB approaches (single- and dual-angle LST) for deriving ET and assess synergies present between them. Additionally we will investigate methods for bridging the spatial scale gap between MSI (spatial resolution on the order of tens of meters) and SLSTR (spatial resolution on the order of kilometres) observations. The model outputs will be compared against field-based ET measurements in agricultural sites located in temperate and semi-arid climatic zones. The outcomes will contribute to operational and accurate estimation of ET with the use of Sentinel data.

Spruce forest decay due to bark beetle attack in The Šumava Mountains is highly discussed topic for a more than 30 years. Our study shows a potential of very high spatial resolution of UAS images for forest ecosystem classification in particular to identify forest regeneration after two types of forest disturbance. We expected that using multispectral cameras with very high spatial resolution (several cm) allows to identify small seedlings and spectrally distinguishable from surrounding vegetation. It also envisages the use of UAS images to classify land cover in forests accurately. The methodology was tested on two types of plots (20 x 20 m): a) the area of spruce forest decay due to bark beetle, b) clear cutting area with subsequent planting of new trees. Spruce forest disturbances in above mentioned plots happened in 1997. There was used multispectral camera ADC Tetracam for aerial imaging, which was carried on unmanned systems Hexacopter XL. As a second camera we used DSLR camera Canon with prime lens for photogrammetry purposes. The imagery includes visible and near infrared (NIR) bands. The height of the imaging was 50 m. The UAS data were acquired always in May and September to avoid the vegetation season because in this time the reflectances of spruce and understory vegetation are similar. All images were photogrammetrically processed and converted into the UTM coordinate system. Very important is also field radiometric calibration of data. Final classification was done by Maximum likelihood method and it is based on the surface reflectance. For radiometric calibration we used spectral radiometr ASD HandHeld 2. Calibration is based on measuring wery well known areas with typical reflectance. For this we used spectral plates with reflectance 95 % and 50 % or natural areas (e.g. roads). The results of classification shows that seedlings with crown projection higher than 20 cm can be identified in both spring and autumn images. For smaller diameters of crowns has been substantially reflected the influence of the surrounding vegetation. The images from the spring can be used for distinguish of trees from blueberries and other surrounding vegetation. This is not possible on images from autumn. Conversely, there is better identification of the dead wood on the autumn's images. We also compared the images with the distribution of size categories of seedlings. These results can be used not only for modeling natural forest regeneration, but also as a useful supplement for forest inventory monitoring.

There is ongoing effort to establish a global terrestrial monitoring of chlorophyll fluorescence emission within the framework of the European Space Agency’s (ESA’s) Fluorescence Explorer (FLEX) mission. As a part of preparation study of this mission, an extensive literature review has been conducted in order to determine the information contained in the chlorophyll fluorescence signal under steady-state light conditions within the red and far-red spectral region to indicate and predict vegetation stress.  More than 70 peer-reviewed papers including both passive and active methods of chlorophyll fluorescence retrieval have been analyzed using Comprehensive Meta-Analysis (CMA) and other statistical approaches. Common, mostly abiotic, stressors occurring in nature were considered, such as water deficit, temperature extremes, and nitrogen limitation. While there is a high variability in the response of chlorophyll fluorescence to stress, which is affected by intensity, duration, scale of observation (leaf vs. canopy levels), and methodology (active vs. passive methods), some general conclusions are discernible. Water deficit, for example, is generally associated with a decline in both red and far-red fluorescence intensity. High temperature extremes produce a decline in the red to far-red fluorescence ratio, whereas chilling stress produce an increase in both fluorescence bands, while nitrogen limitation resulted in an increase of red to far-red ratio. Based on the results of meta-analysis we developed a mathematical approach for stress quantification and identification that utilizes indicators based on spectral band indices and comparisons to baseline or control values. Recommendations are discussed for applications of these indicators, taking into account potential sources of variability and error arising from biological or environmental factors. Crucially, we concluded that both the red and far-red spectral bandsare necessary for a successful application of chlorophyll fluorescence for detection of stress in terrestrial vegetation ecosystems.

Proba-V was successfully launched on May 2013 to continue the 15 years long-term data set of SPOT-VGT sensors, bridging the gap on global land cover until the launch of Sentinel-3. The main goal of Proba-V mission is to ensure vegetation monitoring on a global scale at a spatial resolution of 1km and 300m. Additionally, central camera measurements at 100m resolutions and with 5 days revisit time are also provided to Proba-V users. As any other optical sensor, Proba-V measurements are affected by atmospheric contaminations and clouds coverage, which limit their usability for Land Remote Sensing. It is therefore common practice to use multi-temporal compositing techniques in order to generate clouds-free and spatially continuous images for a pre-defined time interval. The main challenge of such techniques is to select the best quality pixels, preserving the information on land surface parameters, while minimizing the noise induced by residuals atmospheric contamination, undetected clouds (and clouds shadows) as well as remaining directional (BRDF) effects. The choice of the compositing method is crucial in order to detect phenological or land cover changes.

For Proba-V, the Maximum Value Composite (MVC) method is used, which is based on the selection of pixels having the Maximum Normalised Difference Vegetation Index (NDVI) under the assumption that those pixels will be the least contaminated by the atmosphere and with the most favourable observation geometry. The choice of this method is mostly induced by the need of continuity with the SPOT-VGT data series, which was generated using this method. Most of the NRT vegetation monitoring services, currently supported by Proba-V, are in fact based on the detection of NDVI anomalies with respect to the SPOT-VGT long-term average values. The limitations of the MVC approach are however known, in particular the tendency to favour large viewing angles, with resulting increased uncertainty in atmospheric correction [1]. Several methods have been proposed in the past, each one having advantages and drawbacks, the choice of the method is highly dependent on the target application. A comprehensive review of existing methods is available here [2].

The aim of this paper is to analyse the performances of the currently adopted Proba-V compositing algorithm in relation to alternative methodologies. Different methods based on minimization of reflectances in the visible channels, minimization of viewing angle or statistical approaches will be tested on Proba-V 300m atmospherically corrected data. Several quality criteria will be used to assess the performances of each method, namely: image texture, residual clouds contamination, observation geometry distribution, spatial consistency and temporal stability of reflectances values and NDVI. The impact of the different methods for vegetation phenology monitoring will be ultimately discussed.

[1] van Leeuwen, Wim JD, at el. "MODIS vegetation index compositing approach: A prototype with AVHRR data." Remote Sensing of Environment 69.3 (1999): 264-280.

[2] Dennison, Philip E., et al. "Spectral shape-based temporal compositing algorithms for MODIS surface reflectance data." Remote Sensing of Environment 109.4 (2007): 510-522. 

Mapping and monitoring of deforestation in the Brazilian Amazon has been subject to numerous studies based on a wide range of available earth observation systems. However, studies assessing deforestation in the Brazilian Cerrado biome are few and only recent. The Cerrado Biome is considered the most extensive and diverse ecosystem and one of the top biodiversity hotspots in Brazil. But, exploitation of natural resources within the past decade has transformed the landscape of the Cerrado from homogeneous forest to a fragmented landscape resulting in a severe loss of carbon sequestration in those regions. This loss is not only related to the total decrease in forest area but also assumed to be a consequence of increasing fragmentation at the forest edge (Pütz et al. 2014). The trees near the forest edge are more susceptible to wind and sunlight, which increases tree mortality and consequently affects carbon storage capacity. This loss is often not accounted in global carbon loss calculations, especially on a larger scale analyses (Pütz et al. 2014). Hence, all monitoring and reporting activities that are supported by remote sensing data not only have to account for total forest loss but also for landscape fragmentation.

The vegetation of Cerrado consists of different physiognomic units that cover a wide range of plant communities from dry grasslands (campo limpo) characterized by herbaceous plants (principally grasses) without trees and shrubs, grasslands with scattered shrubs and small trees (< 15%) (campo cerrado) to forest formations dominated by trees and shrubs often 3-8 m tall and more than 30% crown cover (cerrado sensu stricto) and almost closed forest with crown cover between 50% to 90% often 8-15 m or taller (cerradão). Therefore, the mapping and monitoring of patches and degradation of the Cerrado types at large scale is a challenge for satellite image analysis.

Radar sensors have the advantage to penetrate through cloud cover which supports all monitoring activities in tropical regions. Further, they are able to record dielectric and geometrical characteristics of land cover properties without saturation. This is an important issue when the study area is a Tropical Forest (Amazon, Cerrado), because in these Biomes the saturation of optical sensors is specifically high.

In our study, we evaluate the potential of optical and SAR satellite images from the recently available Sentinel-Missions (-1/-2) for mapping land cover, forest loss and forest fragmentation in the Cerrado Biome in Mato Grosso. In the first step, the Cerrado types will be assessed using the map of native vegetation types in 2013 from IBAMA and the mapping will be refined based on available Sentinel-data. At the level of the Cerrado types, we delineate classes of forest fragmentation based on a predefined set of landscape metrics, e.g. number of patches (NP), total fragment size in hectares and total core area in hectares (TCA). In the second step, vegetation indices will be generated from Sentinel-2 (MultiSpectral Instrument - MSI) or equivalent data. This allows for the determination of biophysical properties of the vegetation. In the third stage, the biophysical parameters of the Cerrado types will be extracted using radar images from Sentinel-1A C-band satellite. We will evaluate the potential of the freely available Sentinel-1 data with regard to estimating differences in vegetation structure measures for different Cerrado types that can further be related to biomass and hence carbon storage. Here, we will use both, the intensity and polarisation of the backscattered signal (backscattering coefficient) as well as the correlation between the signals of two subsequent radar images (coherence) by means of repeat-pass SAR interferometry.

The poster presentation will outline the concept and background of the ongoing study and report first results of the Cerrado mapping from available Sentinel-data with focus on the precise delineation and parametrization of the forest edge.

Reference:

Pütz, S.; Groeneveld, J.; Henle, K.; Knogge, C.; Martensen, A.C.; Metz, M.; Metzger, J.P.; Ribeiro, M.C.; Paula, M.D.de; Huth, A. (2014): Long-term carbon loss in fragmented Neotropical forests. Nature Communications 5,  (2014). doi:10.1038/ncomms6037

Soil surface roughness (SSR) is a crucial parameter in the assessment and modelling of soil erosion in agricultural landscapes. Still, in recent modelling efforts, roughness is usually treated as a static parameter or parametrized from literature, leading to strong simplification and data uncertainty in the description of these physical processes and the derivation of hydrological quantities. However, this simplification is not only due to the lack of theoretical process knowledge, but rather refers to the lack of appropriate roughness input data, as it is very complex to measure roughness at field scale under natural conditions.

To overcome the current limitations, the performance of microwave remote sensing acquisitions is investigated to derive SSR dynamics for a whole vegetation period over several agricultural fields as potential input for eco-hydrological models.

The proposed approach utilizes air- and space-borne SAR data, acquired at C- and L-Band (e.g. 5.6 GHz and 1.3 GHz) for the derivation a variety of potential roughness estimators. In addition an extensive ground truth database of photogrammetrically measured roughness samples is used to establish soil surface roughness retrieval models and to validate the results across scales. To characterize the in-field measurements the RMS-height s – which is the standard deviation of the heights to a reference height – as well as the autocorrelationlength l based on an autocorrelation function was chosen. Using the best fit approach, a highly accurate assessment of SSR at field-scale could be achieved by deriving s using a linear model from the real part of the circular coherence (Re[ρRRLL]) derived from L-Band acquisitions.

In this presentation, we show the database of the proposed approach acquired in the framework of two different major campaigns. Results will be shown from the AgriSAR 2006 campaign funded by the European Space Agency, ESA, as well as results from the Wallerfing test-site using data acquired by DLRs F-SAR system and Radarsat2 provided within the SOAR-EU CSA-ESA Initiative. We will show results acquired at different spatial resolution and discuss the impact of the different frequencies on the retrieval scheme.

In addition we will discuss the results in context of soil erosion research and in the framework of a future utilization of this approach by using operational and future SAR satellites such as Sentinel-1, Radarsat2, Alos-2 and TanDEM-L.

Soil moisture is a climate variable that has an important impact on environment and climate system. Many studies exploited remote sensing imagery (especially spaceborne Synthetic Aperture Radar (SAR)) for the retrieval of soil moisture in agricultural areas [1]-[3]. However, most of the available approaches don’t perform well in mountain areas and only few studies, specializing on such areas, are reported. Topographic features, heterogeneity in land-cover and presence of vegetation complicate the analysis of data signatures in this complex environment.

The objective of this work is to investigate the synergy of multi-sensor observations for soil moisture retrieval in grassland mountain areas and to discuss its added value. The Kernel method, Gaussian Process Regression (GPR) [4] is used as the retrieval technique. A Synergy of SAR, optical, topographic, in-situ, and ancillary (solar radiation) data is tested. C-band SAR features, i.e. backscattering coefficients and local incidence angles, are acquired by the Envisat ASAR sensor in Wide Swath mode with 150 m resolution. Optical data are represented by daily MODIS surface reflectance (res. 500 m), 250 m resolution normalized vegetation index (NDVI) and leaf area index (LAI) with 1km resolution. Terrain information, such as slope, elevation and aspect are extracted from a 10 m resolution Lidar DEM. The solar data are calculated using the model SPECMAGIC [5]. The top 5-cm volumetric soil moisture measurements are collected daily in eight fixed stations, three located in meadows and four located in pastures. A high resolution (25 m) land-cover map derived from orthophotos is used for the identification of pasture and meadow areas.

The fusion of different features, i.e. combinations of different data sources was examined.

In the case when radar images are not available an approximate soil moisture estimate can be obtained with the joint knowledge of the land class, vegetation parameters (LAI) and surface reflectance at 2130 nm, which is close to the SWIR water absorption band. For this scenario, corresponding RMSE and correlation coefficient (r) equal 6.03 % Vol. and 0.85 respectively. It should be emphasized that both surface reflectance and LAI are derived from optical images. This is as an important observation for the Sentinel 2 applications. Estimation further improves by including solar radiation data. When the information about the global daily irradiance is included as a predictor, soil moisture is estimated with the RMSE and r of 5.23 % Vol. and 0.89 respectively.

 

If only SAR and terrain data (classified by the land use map) are used in the retrieval, GPR estimates moisture with a RMSE and r of 7.85 % Vol. and 0.73 respectively. Estimation is just slightly improved by including LAI and leads to a RMSE of 7.65 % Vol. and r that equals to 0.75. A strong performance improvement for SAR retrieval of soil moisture is obtained by including the information about direct solar radiation additionally to the previous case. The RMSE and r for that case are 5.27 % Vol. and 0.89 respectively.

However, the most accurate soil moisture retrieval is obtained with the synergy of SAR, surface reflectance at 2130 nm, terrain features and LAI, and the global daily solar radiation. In this case the RMSE and r are equal to 4.14 % Vol. and 0.93, respectively.

As the next step, the benefit of higher resolution (250 m) LAI maps [6] will be investigated. The main advantage of using the proposed product for soil moisture retrieval, instead of the standard MODIS LAI product, is the improved spatial resolution (from 1 km to 250 m), which is necessary to handle the heterogeneity of Alpine areas. This is an ongoing research and a detailed analysis of these results will be presented during the symposium.

Future work will focus on the increase of the existing dataset with images acquired by different sensors, such as RADARSAT-2, Sentinel 1 and Sentinel 2. Ground measurements from further fixed stations are planned to be included. The possibility of including knowledge of soil moisture obtained from other models, e.g. from hydrological modeling, will be considered as well.

 

[1] S.-B. Kim, M. Moghaddam, L. Tsang, M. Burgin, X. Xu, and E. Njoku, “Models of L-band radar backscattering coefficients over global terrain for soil moisture retrieval”, IEEE Transactions on Geoscience and Remote Sensing, vol. 52, no. 2, pp. 1381–1396, 2014.

[2] S. Paloscia, P. Pampaloni, S. Pettinato, and E. Santi, “A comparison of algorithms for retrieving soil moisture from ENVISAT/ASAR images”, IEEE Transactions on Geoscience and Remote Sensing, vol. 46, no. 10, pp. 3274–3284, 2008.

[3] H. Lievens and N. Verhoest, “On the retrieval of soil moisture in wheat fields from L-band SAR based on water cloud modeling, the IEM, and effective roughness parameters”, IEEE Geoscience and Remote Sensing Letters, vol. 8, no. 4, pp. 740–744, 2011.

[4] C. K. Williams and C. E. Rasmussen, Gaussian Processes for machine learning. the MIT Press, 2006.

[5] R. Mueller, T. Behrendt, A. Hammer, and A. Kemper. A new algorithm for the satellite based retrieval of solar surface irradiance in spectral bands. Remote Sensing, 4(3), pp. 622-647, 2012

 

[6] L. Pasolli, S. Asam, M. Castelli, L. Bruzzone, G. Wohlfahrt, M.Zebisch, C. Notarnicola, “Retrieval of leaf area index in mountain grasslands in the Alps from MODIS satellite imagery”, Remote Sensing Of Environment, vol. 165, pp. 159 - 174, 2015

For a sustainable and efficient land management in agriculture, spatially distributed and current information on the land surface is of central importance. A continuous flow of land management information is the basis to improve decisions on use, cultivation intensity and allocation of resources (e.g. water for irrigation). Remote sensing is in a unique position to contribute to this task as it is globally available and provides specific information about current crop status. Since 2015, the Sentinel sensor family started to continuously monitor the land surface at different spatial scales (10m – 300m) and with different systems (optical, microwave). For the generation of land management information out of this data stream of different resolutions and wavelength ranges, an integrated analysis of all the available image data is required.

In the frame of the M⁴Land project (Model based, Multi-temporal, Multi scale and Multi sensoral retrieval of continuous land management information), a method to derive products for a sustainable management of the land surface is being developed. The method is designed to combine the full bandwidth of the spatial information provided by the Sentinel series as well as EO data from various other sensors (e.g. Landsat/RapidEye) within a model environment, consisting of the land surface process model PROMET (Processes of Radiation, Mass and Energy Transfer) and the surface reflectance model SLC (Soil-Leaf-Canopy). In PROMET, the development of crops is simulated dynamically depending on the environmental conditions (mainly temperature, radiation and moisture conditions), while standard farming practices (e.g. seeding and harvest dates) are taken into account. Ensembles of model runs are used to represent different crop types, fertilization states, soil colors and soil moistures. Resulting surface state parameters like leaf area index or phenology of different crops are then transferred to SLC. SLC simulates spatially distributed surface reflectance spectra considering sensor specific configurations like spectral bands and solar/observation geometries. By pixel-wise and multi-temporal comparisons of simulated spectra and spectra from real satellite images, the land cover/crop type can be classified dynamically. The classification is model-supervised and without in-situ based training data. Products on crop cycle and intensity of agricultural productivity are generated automatically. The intensity of agricultural production determines both agricultural productivity and ecological damage and therefore supports management decisions concerning the ecological intensification or extensification.

The approach and its transferability were validated in recent studies using Landsat and RapidEye imagery for different test sites in Germany showing overall classification accuracies over 80%. The M⁴Land analysis system is designed to integrate multi-mission data and is well suited for the use of Sentinel-2s spatially and temporally continuous data stream. Its 10-20m spatial resolution, its 5-day revisit frequency as well as the availability of three bands in the red edge, which provide key information on the state of vegetation, offer new opportunities for regional to global agriculture monitoring. Within this study the newly available Sentinel-2 data is going to be implemented in the M4Land system at a high resolution scale suitable for agricultural areas. In the presentation, the method and results, including the generation of products on dynamic land use and cultivation intensity by using Sentinel-2 data for a test site in Germany and the year 2015 will be shown and discussed.

The Copernicus Global Land Service provides operationally a series of bio-geophysical products on the status and the evolution of land surface at global scale. The products are used to monitor the vegetation, the energy budget and the water cycle. A large number of these products were initially derived from the SPOT-VEGETATION sensor until end of 2013, and have since been based on similar 1 km product from PROBA-V. Because PROBA-V provides also daily global data at a spatial resolution of 300 m, an evolution for a number of products was initiated to calculate the variables also at 300 m resolution. It is likely that in the future, the Service will focus entirely on the 300 m spatial resolution, especially because Sentinel-3 will provide data at this resolution.

 

However, combining data from different sensors and at different resolutions to create homogeneous long time series requires several issues to be addressed. These are (1) the similarity of the products among the different sensors, (2) the consistency of the products among the different resolutions, and (3) the ability to create a 300 m resolution archive derived from the VGT 1km products. The last issue is particularly important to calculate anomalies, which express the performance of the present conditions compared to all known conditions in the past for the same day and location, i.e. the Long Term Statistic (LTS). Currently, the GL Service provides the Vegetation Condition Index (VCI) and the Vegetation Productivity Index (VPI) products which are based on the NDVI.

 

The first issue deals with the combination of different sensors at the spectral level. Calculating the anomaly products using PROBA-V data is only possible if a long-term time series is used to calculate LTS. We discuss the similarity between the time series of VGT and PROBA-V NDVI and the performance of the spectral correction that takes into account the differences in spectral response functions of the sensors.

The second issue deals with the consistency between the PROBA-V 1 km and 300 m NDVI products. Although both products are derived from the same input data, their compositing is done differently, which creates an inconsistency between the two products. For the 300 m composites, there is a preferential selection rule of observations within a viewing geometry of 40°, whereas this rule is not applied for PROBA-V 1 km and VGT data. For VGT however, it is known that off-nadir pixels are predominantly selected. Because NDVI is still sensitive to the viewing geometry, this results in less reliable anomaly estimates at 300 m. 

Thirdly, it is necessary to create the VGT-based LTS at 300 m. In order to do so, two options were investigated. These are (1) a simple expansion of the 1 km pixels into 300 m and (2) a data fusion method that uses the relationship between the PROBA-V 300 m and 1 km NDVI to derive the LTS at 300 m. The current results indicate that both methods are performing similarly, probably because the time series to base the fusion method on is currently too short (only one year used).The same analysis is done for fAPAR that is based on bidirectional normalized reflectances.

 

In this presentation, we discuss the three issues with a specific focus on VCI and VPI derived from NDVI and fAPAR with the objective to provide anomaly products at 300 m resolution. At last, we look forward to the challenges ahead when moving from PROBA-V to Sentinel-3 data. 

Land Surface Models (LSM) have become indispensable tools to quantify the most important physical, chemical and biological processes to determine water and nutrient fluxes in support of land management strategies or the prediction of climate change impacts. However, the utilization of LSM requires numerous soil and vegetation parameters which are seldom available in spatial distribution or in an appropriate temporal frequency. As shown in recent studies, the quality of these model input parameters, especially the spatial heterogeneity and temporal variability of soil parameters, has a strong effect on land surface model simulations.

The presented study tries to assess the potential of microwave remote sensing data for the retrieval of soil physical properties. Microwave remote sensing is able to penetrate in an imaged media (soil, vegetation), thus being capable to retrieve information beneath such a surface. In this study, microwave data acquired with DLRs E-SAR system at 1.3 GHz and at different polarizations during the ESA funded AgriSAR 2006 campaign is utilized for the retrieval of information about soil physical properties such as porosity and the dielectric characteristics of the soil. The dielectric properties are used for the retrieval of soil texture information in conjunction with geostatistics. In a regression kriging approach, the retrieved dielectric properties of the imaged fields are used as co-variables to predict soil texture.

Finally, maps are being created showing the content of sand, silt and clay, as well as porosity of the upper soil layer. The developed approach is validated with in-situ data from different field campaigns carried out over the TERENO test-site “North-Eastern German Lowland Observatorium” as well as during the AgriSAR 2006 campaign.

The presentation shows results of this novel technique and highlights the potential of the proposed approach for the derivation and regionalization of spatial distributed soil physical properties at field scale.

Since the launch of the SMOS mission in 2009, two other satellites carrying L-band radiometers joined it on orbit. Aquarius was launched in June 2011 and SMAP in January 2015. Unfortunately, Aquarius ceased operation later that year. All 3 instruments have been operating simultaneously between April and June 2015. Although this golden age of L-band on orbit radiometry was short lived, it allowed for sound comparison of the performances of these 3 radiometers. Moreover, its untimely termination emphasizes the need for reliable inter calibration to build long term consistent archives of brightness temperature and higher level products.

Still, since all these instruments do not share the same technology and even principle of acquisitions, direct comparison and synergistic use of their measurements is not straightforward.

The objective of this paper is to demonstrate a set of methods to make them inter-comparable, down to a common reference. Instead of a ground reference, we use here SMOS as a transfer radiometer.

This method can be applied over different types of surfaces: i) making use of a stable target to assess the consistency and stability of both data sets. This is done over the area surrounding Dome Concordia in Antarctica. After careful selection and filtering, statistics of the comparison are retrieved along with long term trends in both data sets. ii) Once every so often, satellites overpass the same area within a very short time period. Due to different orbit inclinations these alignments occur essentially along the equator, but over different surfaces, giving access to wide dynamic range in brightness temperature. iii) At last, all radiometers are aimed at the deep sky for calibration purposes. But despite this use of a common reference, it can be shown that the retrieved brightness temperature exhibit some differences, traceable to the differences in calibration strategies and acquisition principles.

This presentation will briefly describe the already established methods, along with a full record of results over the lifetimes of SMOS, SMAP and Aquarius. A more general conclusion on common use of these data sets will also be given.

Biomass is an important player in the area of renewable energy sources. While large-scaled features like forests and agricultural fields have been plentifully assessed and quantified, the focus now shifts to small-scaled biomass features like vegetation alongside roads or hedges between fields. Since such features are oftentimes too small to be detected with the regular spatial resolution of satellite images, spectral unmixing needs to be applied. We are presenting an approach using different classification techniques (e.g. spectral unmixing with endmember detection based on the method SMACC [Sequential Maximum Angle Convex Cone] as well as other pixel-based supervised and unsupervised algorithms) to detect and differentiate various types of vegetation including grass, forest, agriculture, and hedges. The derived endmembers and classes are also evaluated based on their possibility of post-processing to further enhance the outcome. Examples of this step are combinations of endmembers or histogram modifications to focus solely on a specific part of the grey value range. The data sets for our research originate from Landsat-8 and will work as a guideline for the application of the algorithms on upcoming Sentinel-2 data.

With the delineated small-scaled features of vegetation, we are able to determine the quantity of the potential annual biomass harvest. Additional inputs are GIS data, height information (e.g. LiDAR or aerial stereo images) and the re-growth pattern of graminaceous, herbaceous and ligneous plants. The GIS data, consisting of road networks, communities and nature reserves, helps eliminating small-scaled vegetation features that are not accessible via the road networks. With the 3D vegetation structures, it is possible to estimate the amount of solid biomass (as opposed to total 3D volume) by applying a species-specific ratio to the total volume. The re-growth rates, therefore, can be applied to the remaining volume and be taken as annual, accessible harvest in the defined region. With the location of communities and nature reserves, we can outline an area where processing facilities are ideally located taking into account distance, local transport and construction restrictions.

Our study concludes that the detection of sub-pixel-scale vegetation is possible using different spectral unmixing and classification techniques. We present a combination of algorithms that can be applied to determine the type of plant as well as an estimated annual harvest. This information is an important addition to the forest inventories. It helps assess the regional and national vegetation characteristics and can be used to estimate energy potentials and to enhance the independency of fossil fuels. With the finer spatial and spectral resolution of the upcoming Sentinel-2 data, this approach can be implemented (semi-)automatically and form an important part of the application of ESA’s Copernicus Programme.

Forest disturbance due to fire and subsequent regeneration are important processes particularly in South East Asia. Forests damaged by fire are still valuable in terms of conservation due to their ability to recover, sequester carbon and restore ecosystem services and should therefore be monitored. Sungai Wain Protected Forest (SWPF) located near the city of Balikpapan, East Kalimantan was affected by fires driven by El Nino Southern Oscillation (ENSO) in 1998 which led to the degradation of a previously undisturbed Dipterocarpacea forest. With the frequency and magnitude of ENSO events likely to increase in the future it is important to monitor forest regeneration after fire in the long term with Remote Sensing being a particularly valuable option. Spaceborne Interferometric Synthetic Aperture Radar (InSAR) offers an alternative to optical sensors in tropical areas where cloud cover and haze is frequent and hampers the retrieval of meaningful information. In particular, the TanDEM-X missions offers the opportunity to acquire data in interferometric mode thanks to two satellites closely orbiting in a double helix formation with minimization of temporal baseline and at high spatial resolution (approximately 2 m ground resolution at 41° incidence angle in StripMap single polarisation) compared to previous satellite missions. Typically temporal decorrelation has severely limited the ability to use coherence to monitor forests. The unprecedented provision of TanDEM-X coherence through simultaneous capture, thus without the limitation of temporal decorrelation, solves this problem. The aim is to assess the sensitivity of TanDEM-X coherence to horizontal vegetation structure (canopy cover) and vertical structure (vegetation height) these being derived from a high resolution airborne LiDAR (1 m resolution). Tests were done on 150 plots (35 x 35 m) located in secondary forest and agriculture, with mean canopy cover of 73% and height between 0.4 to 36.4 m (mean 14.7 m). Results indicate a negative linear relationship between TanDEM-X coherence and canopy cover (R²=0.64) (n= 150). This suggests that coherence is sensitive to the spatial distribution of scattering volume which causes decorrelation proportional to canopy cover. However, it was found that for fully regenerated secondary Dipterocarp forest with high canopy cover coherence ranges between 0.4 and 0.8 resulting in high scatter for high canopy cover. This was further investigated by assessing whether topography played a role in lowering coherence. No correlation between slope and coherence was found. Another possible cause for low coherence could be due to layover. Hoekman and Varekamp (2001) suggest that a high height difference between the forest components (presence of tall emergent trees) results in lowered coherence due to geometric decorrelation. Moreover, relationship between the vertical structure component (vegetation height derived from a LiDAR Canopy Height Model- CHM) was also tested resulting in a good correlation with coherence (R²=0.68) while, a weaker relationship was found with LiDAR CHM mode (R²=0.63) and, lower still with the distribution of standard deviation (R²=0.5): but some of these independent correlations offer possibilities for better mapping of canopy cover. These results offer a promising potential for the use of coherence for detecting forest structural parameters such canopy cover and forest height.

The overall goal of the ESA DUE Innovator III project “SAR for REDD” is to provide cloud-penetrating satellite synthetic aperture radar (SAR) pre-processing and analyzing capabilities and tools to support operational tropical forest monitoring in REDD countries and primarily in Africa. The project’s end-user is the Observatoir Satellitale des Forêts d’Afrique Centrale (OSFAC).

C-band Sentinel-1A CSAR (S1A) and L-band ALOS-2 PALSAR-2 (A2P) have both been launched in 2014 and are the main sensors that the development of remote sensing forest products, i.e. forest land cover (FLC) and forest change detection (FCD) maps will be based on in this project.

The pilot focus region is the Mai-Ndombe district of about 12.9 Million ha in the west of Democratic Republic of Congo and full coverage with historical Envisat ASAR, ALOS Palsar until 2011 data have been collected and processed in addition to recent full coverage from S1A and A2P for the years 2014 and 2015.

Image mosaics based on multi-temporal acquisition and forest land cover products have been produced for each sensor individually and are compared. For forest applications, L-band SAR is generally better suited than C- (or X-band) since its longer wavelength signal penetrates deeper into the forest canopy and thus, also provides more information on biomass. However, the denser time series of more frequent revisit time of S1A seem to balance up to a strong degree the disadvantage of a shorter radar wavelength. The different seasonal backscatter evolution of different land covers can be used to classify the land covers in much more detail than just a forest/non-forest map. Inundated forest and seasonally flooded areas along rivers are clearly detectable. High-resolution optical data from the SPOT5/TAKE5 campaign collected from April to September 2015 are used for validating the SAR based FLC products. Interoperability and complementarity of S1A and A2P data will also be investigated.

Forest change maps covering the 5-year period between the ASAR/Palsar era around 2010 and the start of the  S1A/A2P era (2015) will also be presented and are compared to Global Forest Watch results based on optical Landsat data.

Changes in land use and land cover (LULCC) are the main cause of the current biodiversity crisis and protected areas are a global mean to protect wildlife habitats and populations against direct and indirect impacts of human land use. Large mammals are particularly challenging to protect in human-dominated landscapes, in part because they often critically depend on habitat surrounding protected areas. Russia harbors exceptional biodiversity and a well-established network of protected areas to safeguard it. The breakdown of the Soviet Union in 1991 triggered drastic socio-economic shocks and institutional upheavals resulting in widespread LULCC, while at the same time funding for nature conservation was reduced. It therefore remains unclear, how effective protected areas were in times of rapid changes and how the breakdown of the Soviet Union affected large mammals’ habitats within and outside of protected areas and species population dynamics over time. Our objectives were first, to assess rates and spatial patterns of post-Soviet LULCC by analyzing a long time series of Landsat TM/ETM+ images (1984-2010) in our study region in temperate European Russia. Second, we evaluated the effectiveness of two protected areas, Oksky and Mordovsky State Nature Reserves, in preventing LULCC. Third, we studied the impact of post-Soviet LULCC on habitats and the population dynamics of large mammal species within and outside protected areas by evaluating long-term data of species occurrences and abundances from winter track count data. We applied state-of-the-art methods in remote sensing (e.g., support vector machines classification) and effectiveness evaluation (e.g., matching statistics). We used a time-calibrated habitat model to link post-Soviet LULCC and changes in important landscape metrics to the habitats of large mammals, i.e., wild boar (Sus scrofa), moose (Alces alces), and wolf (Canis lupus). Finally, we applied a panel data regression model to evaluate the relative impact of LULCC and other environmental and human-impact variables on large mammals’ population dynamics. Our results showed that LULCC was widespread in our study region: 5% of the forests were disturbed, 40% of the 1988 farmland abandoned until 2010, while concurrently 46% of the forest disturbance sites and 9% of the abandoned area were reforested by 2010. These massive changes also triggered prominent changes in forest fragmentation. At the same time, the protected areas were effective in preventing forest disturbances and Oksky State Nature Reserve showed a constantly high share of suitable large mammals’ habitat within its strictly protected core zone. Nevertheless, post-Soviet LULCC, particularly farmland abandonment, substantially increased potential wildlife habitat, for example, moose habitat surrounding Oksky State Nature Reserve increased by 27% between 1987 and 2007. Furthermore, our results indicated that variables related to human impact, for example, hunting or the size of human population as well as environmental variables such as predation had a significant effect on large mammals’ population dynamics after the breakdown of the Soviet Union. Finally, our analyses contribute to the mapping and monitoring of structural habitat characteristics and support models of wildlife distributions and abundances in European Russia. They highlight that remote sensing-based analyses of land cover and land use provide valuable insights into the complex interrelations between LULCC and biodiversity. This is particularly important in regions that experience drastic shifts in socio-economic and institutional conditions because these shocks may provide challenges and opportunities for biodiversity conservation.

Earth observation was looked at as magic and the solution for all water management data related problems - in the seventies. Expectations came down to Earth in the nineties. This was the time, when the first generation of African EO scientists started to seriously contribute to the advance of the technology and the African water managers showed readiness for using it. In the beginning of the new millennium, the European Space Agency initiated and launched the TIGER Initiative to support the African water sector with direct cooperation with a focus on capacity building and data provision. This programme has been implemented throughout several phases, and for these days a broad network and community grew out from it.

The first research themes were mostly related to the implementation and application of established methods developed and applied elsewhere. These first efforts were strongly affected by the limited access to proprietary software and expensive hardware. Furthermore, many participants of the programme needed basic training in remote sensing.  Recently, in a cooperation between African researchers and stakeholders and ESA-led European partners, sophisticated EO application methods have been developed and implemented in the Water Observation Information System (WOIS). This open source software combo incorporates several research results of the African partners and enables the user to implement new algorithms as new workflows.

The presentation will review the development of the African Earth observation research and application arena through the experiences of the TIGER Initiative by following the development of the topics, methods and the analysed data of dozens of research and application projects. The role of free access to satellite data and open software in the special African environment will be discussed, assessing the possibilities of developing tailored products. Increasing data volumes and heterogeneous data sources represent a new challenge that needs careful assessment and a new, globalized level of cooperation for meeting the information needs of Africa.

Tropical forests play an important role in the global carbon cycle. Thereby, deforestation is not only responsible for direct carbon emissions but also alters the forest structure and extends the forest edge area in which trees suffer increased mortality due to altered microclimatic conditions. Our aim is to quantify the global amount of anthropogenically created forest edge area and the resulting additional CO2-emissions by combining remote sensing data (Landsat, ENVISAT and others) with previous empirical and modelling results.

We found that 1,106 million ha and thereby 10% of the global tropical forested area lies within the forest edge area and that 84% of this area is anthropogenically created. From this area, a total amount of 8 Gt C is emitted due to tropical forest fragmentation, which accounts for an annual loss of 0.25 Gt C equaling 17% of the annual carbon losses due to deforestation. Fragmentation in the tropics hence augments carbon loss from deforestation substantially and should be taken into account both when analyzing the role of vegetation in the global carbon balance and when adopting new management strategies in tropical forests.

Robinia pseudoacacia (black locust) cover the 24 percentage of Hungary's forest stands. As an invasive species, besides its economic significance, it is important to monitor its spatial spread.

To evaluate such task, remotely sensed multispectral data is essential. Landsat imagery is the most frequently used remotely sensed data in many fields related to monitoring of the Earth's surface. In the case of this study Landsat data is the best source because of its spatial resolution and its longevity which makes it possible to study separate forest stands in long time interval. As Landsat satellites have been gathering data since 1972, lots of valuable information has been stored and can be derived from imagery provided by United States Geological Surveys Landsat archive, which is open since 2008. In addition we study Sentinel-2 data and its possible integration to our research and synchronized usage with Landsat data.

The key in mapping the spatial spread of a certain plant species is the ability to distinguish it from the other species, which can be done by their spectral properties. However, the spectral difference between certain broadleaf species is not as obvious as the difference between broadleaf and coniferous species, hence broadleaf species are barely or non-distinguishable from each other by their spectral properties at the spectral resolution of TM, ETM+ and OLI sensors. Preliminary research has shown that the desired result is not achievable by the usage of an individual Landsat image as input therefore the available data have to be studied not just in geographic space but in time as well. To store the data into spatiotemporal database makes it possible to examine a species specific occurrence of phenological events during the growing season and phenological curves can be built up for each and every pixel. Whit this method Robinia pseudoacacia can be distinguished from other broadleaf species. Of course, there are quite a few common time-series related factors to be taken into account and Landsat data has to be pre-processed before add it to the database. During the pre-processing, atmospheric correction and topographic correction was applied to have the data comparable on an absolute scale. For atmospheric correction the 6s algorithm is applied inside GRASS (Geographic Resource Analysis Support System) GIS and the whole pre-processing chain is composed of GRASS algorithms. Vegetation indices and tasselled cap greenness derived from the pre-processed data and stored into PostGis database in favour of faster data analysis and development of the final mapping algorithm which is recognizes Robinia pseudoacacia. In this study I present the concept of time-series based plant species recognition, the build-up of the required pre-processing chain, the final mapping algorithm and the result maps which show the spatial spread of Robinia pseudoacacia in the Carpathian Basin on a yearly basis.

The FLEX/S3 (FLuorescence EXplorer/Sentinel-3) tandem mission, selected by ESA to become the Earth Explorer 8, is the first mission that will allow the study of vegetation photosynthesis, through the retrieval of Sun-Induced Chlorophyll Fluorescence (SICF). Although primarily designed for land, its unique combination of very high spectral resolution, high sensitivity and moderate spatial resolution, opens the opportunity for the study of phytoplankton in coastal and inland waters.

FLEX consist on a platform with a hyperspectral instrument FLORIS (FLuORescence Imaging Spectrometer), acquiring in the spectral range between 500 nm and 800 nm, with 300 m of pixel size and 150 km of swath, flying in tandem with S3. The synergy between FLORIS, OLCI (Ocean and Land Color Instrument), and SLSTR (Sea and Land Surface Temperature Radiometer) on board of S3, enables to characterize the atmospheric state and determine the vegetation biophysical parameters for a reliable retrieval of SICF.

FLEX/S3 tandem mission, will provide images from a global coverage for all land surfaces including major islands and coastal waters within 50 km of any land as well as ocean waters with a depth of less than 10 m. If the Ocean Colour community express its interest, some experimental areas of the open ocean could also be included in the mission coverage.

In order to evaluate the capabilities of FLEX/S3 mission for the detection of phytoplankton SICF from water (with an emission peak on 685 nm with a half-height width of around 25 nm), we performed a simulation using HidroLigth Radiative Transfer Model (RTM) combining a wide range of Optically Active Constituents (OAC) for producing realistic water-leaving radiances (including chlorophyll fluorescence emission) and doing the water-atmosphere coupling using MODTRAN RTM Look Up Tables at the FLORIS spectral resolution, for several atmospheric scenarios. We compared the fluorescence signal with the instrument noise-equivalent radiances and obtained the Signal-to-Noise Ratio (SNR) over water targets.

The results showed that, if a mild spectral binning is applied to FLORIS images, its SNR over water will be similar to that of MERIS and close to the expected SNR of OLCI, at the same spatial resolution. After that binning, FLORIS will still have enough spectral resolution (10 nm) in the 500-800 nm range, as well as in the SICF emission band (1 nm), for the retrieval of biophysical products. The analysis of the simulated dataset indicated also that the combination of SNR and spectral resolution could allow the detection of the subtle changes induced by pigment composition and SICF in the water-leaving spectra.

With this purpose, we tested several inversion procedures for the simultaneous retrieval of OAC (i.e. chlorophyll-a and other phytoplankton pigments and suspended minerals) and SICF, with good results in the simulated dataset. Several field campaigns, as well as overflights with the FLORIS airborne simulator HyPlant, will be used for the validation of the inversion models in the forthcoming years.

The products of FLEX over waters will complement those of OLCI, thanks to the expected quasi-simultaneous image acquisition and co-registration in the common swath. FLEX will benefit from the wider spectral range of OLCI and SLSTR images, especially for the radiometric calibration, the atmospheric correction and for some water products (such as CDOM) based on spectral information outside the FLEX band set. On the other hand, FLEX will improve the accuracy of the S3 water products, especially for those indices (such as Fluorescence Line Height - FLH or Maximum Peak Height - MPH) based on bands in the red – near infrared spectral region. It will also add information on other phytoplankton pigments, besides chlorophyll-a, and it will permit the retrieval of SICF with an unprecedented accuracy.

One of the key ecological paradigms is on the relationship between biodiversity and productivity. Higher biodiversity leads to more productive ecosystems, and more productive environments support more biodiversity. At the population level, more productive ecosystems often provide a higher habitat quality in which species persist. Ongoing global changes are affecting ecosystem productivity, being expected to affect productivity-biodiversity and productivity-habitat quality relationships. Mediterranean-type ecosystems are hotspots of biodiversity threatened by ongoing changes in land use and climate, and these changes likely affect ecosystem productivity and the inherent habitat quality. Here we test to which extent are changes in land use and climate affecting habitat productivity, and whether productivity variation affect habitat quality and use by medium-sized mammals. To do this we analyzed the Enhanced Vegetation Index (EVI) for the Landsat time-series over an oak woodland ecosystem in southern Portugal and linked it to meteorological station data and to radio-tracking data of genets (Genetta genetta), stone martens (Martes foina) and badgers (Meles meles). We found that over the last 15 years: (1) on average EVI did not change while there was a significant decrease in max EVI and increase in min EVI, (2) EVI was strongly correlated with relative humidity and negatively correlated with temperature, (3) stone marten and badger presence was positively associated with productivity (EVI and NDVI) while genet habitat model was not significant, and (4) stone marten habitat preferences before and after a drought period showed that the species still preferred oak woodlands over other land cover types; cork oak woodlands are still the land cover type with higher EVI despite the significant reduction in EVI from 1997-1998 to 2005-2006 (EVI_97-98=0.32, EV I_05-06=0.25). These results suggest that despite changes in productivity, cork oak woodlands are still the most selected ecosystem. Cork oak woodland capacity to provide habitat is associated with a diversity of resources embedded within this ecosystem, i.e. due to its high biodiversity.

Heather burning is a widely used tool for upland management in England.  Regulations and guidance are in place to control the size, frequency and timing of burning and are enforced by Natural England; there is a clear need for burn monitoring and mapping services to ensure that they have the required evidence to determine whether policies, management agreements and regulations on heather burning are leading to desired environmental outcomes. Satellite Earth observation, and the Sentinel programme in particular, offers great potential to fulfil this need.  The purpose of this project was to evaluate whether Sentinel-1 SAR can provides a realistic and achievable method for routine upland heather burn monitoring in England. 

 

The project explored methods for using a time series of satellite SAR images to deliver a heather burn mapping and monitoring system for the North York Moors.  A proxy Sentinel-1 time series was constructed using 32 ERS-2 images from 2006-2009, providing data for three complete burn seasons.  Candidate heather burns were identified using a combination of multi-temporal features and SAR intensity.  A rule-based classification was subsequently used to generate burn maps for each season.  Comparison against contemporaneous optical imagery and historical burn maps demonstrated that the system was capable of reliably identifying burns within mature heather that were larger than regulations permitted.  However, areas of the moor that experience frequent re-burning showed a greatly reduced reliability of detection; it is suspected that this is because the difference between burnt areas and young regenerating heather is not as distinct as where mature heather has been burnt. 

 

The study demonstrated that C-band SAR is capable of providing a burn detection service for large burns within areas of mature heather that are less heavily managed than the North York Moors.  Looking forwards, Sentinel-1 is capable of offering an improved burn detection capability given its improved radiometric performance, spatial resolution and systematic dual-polarised capture over England.

Soil moisture is an essential parameter from both a scientific and economic point of view. On one side, soil moisture is related to the water cycle, controlling the exchange of water and heat between the land surface and the atmosphere through evaporation and plant transpiration. Additionally, its knowledge is of high interest for different applications in fields suchas agriculture (e.g. optimization of fresh water use to improve the production quality and quantity),natural hazard prevention and control (predict when excess rainfall can cause water logging and/or flooding), etc.

Different approaches are used to measure the soil moisture content (SMC), including direct and indirect measurements, which can be destructive (in the case of some direct measurements) or non-destructive, as in the case of ground based or remotely sensed indirect measurements.

Recently, the use of the reflected GNSS signals from the soil for measuring soil moisture content (SMC) has been proposed (GNSS reflectometry, GNSS-R). The approach is based on the variability of the soil’s dielectric properties as a function of the soil moisture, which is translated into changes in the electromagnetic. Reflectivity can be estimated from the power ratio between direct and soil-reflected GNSS signals, and related to soil characteristics, as demonstrated in [1].

Here a new concept based on GNSS signalsis proposed, which relies on their capability to penetrate the soil. This technique is similar to the GNSS-R approach, but the direct (clear-sky) signal is compared to the signal transmitted through the soil, instead of the signal reflected by the surface, in order to derive the transmissivity and attenuation coefficient, which are a function of the soil characteristics (i.e. soil moisture content, soil type, temperature, roughness, etc). A preliminary experimentwas performed to demonstrate the validity of this technique, where the signal received by a GNSS-R L1/E1 RHCP antenna buried at 10 cm below the surface, was compared to the one received by a GNSS-R L1/E1 RHCP antenna with clear sky visibility. Results showed significant power variations as a function of the soil’s humidity on the signal acquired by the buried antennas.

A new experiment is proposed,based on the use of 4 GNSS antennas, one with clear sky visibility, which will be used as reference, and three deployed at different depths. This approach will allow us to measure the attenuation suffered by the navigation signal through the soil, comparing the GNSS signal strength received at the different depths with the reference one. Ground truth data acquired during the experimental campaigns will be used to validate the GNSS measurements.

Details related to the GNSS soil moisture modeling, instrument preparation, measurement campaign, data processing and validation will be presented at the conference.

FLEX & SENTINEL 3 : A TANDEM TO MONITOR VEGETATION

E. Jauffraud¹, P. Bassaler¹, P. Coppo²,A. Taiti², E.Battistelli², M. Rossi²

¹Thales Alenia Space, France. ²Selex ES, Italy

Keywords:  Future mission concepts, ESA Programmes: Future Missions (Sentinel, EE and Meteorological Programmes)

1.            Introduction

The FLuorescence EXplorer (FLEX) mission is selected as the ESA’s 8th Earth Explorer opportunity mission.

The ESA-funded phase A-B1 study has been devoted to demonstrating the system feasibility and a breadboard activity on the FLORIS payload is on-going.

The FLEX flight segment consists of a single satellite carrying the FLORIS payload and flying in formation with Sentinel 3 mission in a Sun synchronous orbit at a height of about 815 km, making use of data synergy with other visible reflectance (from OLCI ) and surface temperature data (from SLSTR).

The satellite is based on a recurrent platform. This ensures that the required levels of performances will be met and allows the satellite development to achieve the industrial cost target while minimizing development risks.

The physical satellite configuration is driven by the accommodation of the payload to be mounted on top of the platform guaranteeing an unobstructed view of the Earth.

FLORIS, a pushbroom hyperspectral imager,  will observe the vegetation fluorescence and reflectance within a spectral range between 500 and 780 nm at medium spatial resolution (300 m). Multi-frames acquisitions on matrix detectors during the satellite movement will allow the production of 2D Earth scene images in two different spectral channels, called High Resolution (HR) and Low Resolution (LR) with spectral resolution of 0.3 and 2 nm respectively and spectral oversampling of a factor 3.

The main FLEX/FLORIS requirements are displayed in chapter 2. The satellite configuration is described in subsection 3.1, followed by the FLORIS payload in subsections 3.2 and complemented by the description of the satellite subsystems in subsection 3.3. The performance and budgets are displayed in chapter 4, followed by the programmatics information in chapter 5.

2.            Main requirements for FLEX/FLORIS

3.           Overall design

3.1                Satellite Configuration

3.1.1     Overview

3.1.2     Instrument accommodation

3.1.3     Launcher compatibility

3.2                FLORIS Instrument

3.2.1     Optical Concept

3.2.1.1                Telescope

3.2.1.2                High Resolution Spectrometer

3.2.1.3                Low Resolution Spectrometer

3.2.2     Detection Chain

3.2.2.1                Detector & Focal Plane Architecture

3.2.2.2                Main Electronics

3.2.3     Mechanical and Thermal Architecture

3.2.4     Calibration

3.3                Platform

3.3.1     Mechanical and thermal architecture

3.3.2     Avionics architecture

3.3.3     Propulsion

4.            Performances and Budgets

4.1                Mass Budgets

4.2                Power budgets

4.3                Delta V and propellant budgets

4.4                Data rate and volume

5.            Programmatics

5.1                Overall Design and Development Approach

5.2                Schedule

6.            Conclusions

Ecosystem processes, such as photosynthesis, evapotranspiration, and carbon accumulation are responsive to variability in hydrometeorological and stress conditions induced by changes in land use and climate. Vegetation structure and composition, and hence the characteristic plant trait constellation, partly explains differences in feedbacks between ecosystems and atmosphere. Our current understanding of these processes is mainly based on relatively few in-situ observations from which to explore the linkage of vegetation properties and ecosystem functioning such as their efficiency to absorb light for photosynthesis, use nutrients and water for carbon accumulation, among others. The existing space-borne archive of remote sensing data collected over the past forty years is insufficient to further our understanding of the biogeochemical and biogeophysical feedbacks between vegetation and the atmosphere.

Consequently, there is an urgent need to increase observations of terrestrial ecosystem structure, function and composition. NASA recently has begun development of three new missions on the International Space Station (ISS) that will advance our ability to monitor and model terrestrial ecosystems: the Global Ecosystem Dynamics Investigation (GEDI) LiDAR; the ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS), and the Orbiting Carbon Observatory 3 (OCO-3). Scheduled for deployment between 2017 and 2019, these three missions provide a rare opportunity for coordinated observation of ecosystem structure and function. Furthermore, the potential addition of imaging spectroscopy, through airborne sensors such as AVIRIS, or spaceborne missions such as EnMap and Hyspiri, provides pathways for the characterization of composition and other functional traits.

GEDI is a multi-beam, waveform lidar and will measure ecosystem structure, including canopy heights and foliar vertical profile. Characterizing this structure is essential for quantifying aboveground carbon, species richness and diversity, habitat quality and other ecosystem services.  ECOSTRESS is a multi-band thermal infrared radiometer that can accurately measure the temperature of plants and by virtue of the ISS orbit, at different hours throughout the day over the course of a year, thus providing observations of plant response to temperature and moisture over the diurnal cycle. OCO-3 uses three high spectral resolution grating spectrometers to measure atmospheric column CO2, with high precision and solar induced fluorescence (SIF), a proxy for photosynthesis. OCO-3 uses a sampling design to provide observations needed to assess the spatial and temporal variability of CO2 over an annual cycle.

In this talk we describe the observations to be obtained from these instruments with a particular focus on how the data products produced from each mission can be used synergistically to provide a more complete picture of ecosystem structure, function and composition. The spatial and temporal coincidence of the novel measurements from GEDI, OCO-3 and ECOSTRESS will enable the scientific community to address questions of ecosystem dynamics and interactions that cannot be answered from any one instrument alone, and ushers in new era of ecosystem characterization and modeling.

Remote sensing data are a valuable source for the monitoring and understanding of mountain ecosystems since they provide repetitive and spatially explicit information even for remote areas. Specifically, Sentinel 1 and 2 data provide a unique information to monitor mountain dynamics thanks to the high ground resolution and temporal frequency. However, monitoring land surface processes in mountain environments is particularly challenging due to the complex terrain morphology.

We designed methods to both derive important biophysical parameters such as soil moisture and vegetation parameters and to monitor the changes in the forest and snow status over the Alps (43-48.5° N / 4.6-17°E ) taking into account the peculiarity of the mountain areas. Our methods rely on the synergetic use of radar and optical images (to simulate Sentinel 2 data, we mainly used LANDSAT and RapidEye images) and account for illumination differences.

In detail, the derivation of biophysical parameters is based on modeling approaches. Biophysical vegetation parameters such as leaf area index (LAI), chlorophyll content, or canopy water content are derived through the inversion of a canopy radiative transfer model. For soil moisture the regression procedure uses an advanced machine learning procedure such as Support Vector Regression (SVR) which is capable to handlecomplex and non-linear problems, by managing different kinds of inputs. In factespecially in mountainous terrains, other parameters like topography or vegetation strongly influence the measured SAR signal and need to be included in the retrieval process.To train the algorithm for alpine wide soil moisture retrieval, modelled soil moisture from the global climate model ERA-Land was used while ground data are used for validation. Therefore, the algorithm could be easily adapted for any region in the world.

For the monitoring of changes in land cover/land use (snow and forest), a framework has been implemented which consists of: (i) a multi-sensor training library of change signatures trapped by the Sentinels in a set-up phase and caused by different events such as landslides, floods, snow cover, deforestation and agricultural operations, and (ii) classifiers that use the elements of the library to automatically recognize, identify, and map new changes. The library represents a knowledge system which takes advantages from the combination of Sentinel 1 and Sentinel 2 data and that can be extended to other classes of changes.

It is finally worth noting that, this work is carried out in the framework of the Sentinel Alpine Observatory. The Sentinel Alpine Observatory is a new initiative by EURAC to foster Sentinel based remote sensing applications for mountains and, specifically, for the European Alps. It comprises a long-term archive for Sentinel data over the Alps, processing approaches adapted for mountains and the development of tailored applications for users from mountain regions on snow, vegetation, soil moisture and other relevant topics. Data will be processed through federated infrastructures and distributed following an “open” policy.

Biomass values are deeply embedded in estimates of land-use change fluxes, which are derived from estimates of areas of forest disturbance, together with values for the mean biomass in these areas before and after change. The urgent need to reduce the uncertainty in these estimates has motivated major efforts to estimate the spatial distribution of above ground biomass (AGB) from a range of data sources, including remote sensing instruments. Since these were mainly designed for other applications, they still yield high uncertainty and disagreement, especially in dense AGB tropical forests.

          It was in this context that Biomass was selected by the European Space Agency (ESA) in 2013. Biomass will deliver three primary geophysical products: maps of forest AGB and height at 200 m scale every six months, and annual maps of severe forest disturbances at 50 m scale. To achieve the mission objectives, the sensor will consist of a single satellite with a P-band (432-438 MHz) Synthetic Aperture Radar payload. However, International Telecommunication Union Radiocommunication Sector regulations impose restrictions regarding the use of this frequency band by active space sensors, namely, operation within view of registered ground stations is not allowed. The registered stations, which together form the network of Space Object Tracking Radars (SOTRs), were taken into account at the time of mission selection and are located in North America and the UK. Here we evaluate the impact of SOTRs in terms of the associated loss of forest coverage and AGB representativeness, using the most up-to-date global datasets about the distribution of tree cover and forest AGB. Additionally, a different perspective on the effect of loss of coverage due to SOTR operations, particularly relevant to large scale carbon cycle and the modelling community, is given by examining how biomass is distributed in climate space (precipitation versus temperature) and how this distribution changes when the area affected by SOTR operations is removed.

          Dense forests are of particular interest for Biomass, because of its unique sensitivity to the biomass of forests with high carbon density (here defined as greater than 200 Mg ha^-1, which is beyond the saturation level of L-band SAR). High carbon tropical forest regions represent approximately 50% of the global forest area and 65% in terms of AGB carbon stocks. The loss of coverage due to SOTRs operations in these regions is responsible for a loss of only 4% and 3% in terms of global tropical forest area and carbon stocks respectively. The impact over subtropical forests will be much higher, with losses of 35% and 28% of global subtropical forest area and carbon stocks respectively; however, these forests represent only 8% in terms of global forested areas and carbon stocks. All tropical forests in South America, Africa and Southeast Asia will be unaffected. Mexico, which has extensive forest resources and is currently engaged in important forest conservation programmes, will be severely affected by SOTR operations, but is extensively covered by ground networks of national forest inventory plots.

We propose the approach for estimation biophysical parameters, in particular LAItrue and FAPAR, based on in-situ observations and satellite measurements with use of regression-based models (single-factor and multiple-factor). In-situ data were collected during 2013–2015 vegetation seasons within several field campaigns at the multi-scaled JECAM test site in Ukraine. We have estimated stability of relationships between NDVI index (with use of Landsat-8 imagery) and ground-based biophysical parameters values as well as relationship of ground parameters to reflectance in separate satellite image bands (with use of SPOT-5 data).

In our presentation we will discuss the robustness of selected models for LAItrue, FAPAR crop-specific mapping with results validation on independent dataset.

For our case study have been obtained the next results.

1. Relations of LAI True and FAPAR are the same for major crops over JECAM test site.

2. Relations of LAI True and FAPAR are the same for SPOT-5 and Landsat-8 data.

3. Use of multiple-factor regression model on available dataset and 4 bands of SPOT-5 imagery leads to adding statistically non-significant parameters.

4. The most important feature derives from multispectral satellite imagery is NIR band.

5. Use of models calibrated with data from previous growing seasons provide good enough (in terms of RMSE-error) results for operational biophysical parameters mapping.

The current results are important for the further development of operational agriculture services based on satellite imagery (crop state assessment, yield forecasting with high-resolution maps of biophysical parameters). Such services are currently under the development within collaboration activities between Space Research Institute (Ukraine) and European Commission Joint Research Center (EC-JRC), and several international and European research projects.

In further investigations we also plan to use data from Proba-V, Sentinel 1/2 systems and other perspective satellite systems Sentinel family.

The construction of the controversial Gibe III hydropower dam has recently been completed on the Omo River, the main tributary of Lake Turkana, Kenya. This study has used ENVISAT satellite products to investigate the potential impacts of this dam on ecology of Lake Turkana, the world’s largest desert lake and the largest alkaline lake. The lake is of international importance due to its unique biological and cultural diversity, and its three National Parks form a World Heritage Site. It is critically threatened by the Gibe III dam, and associated irrigation developments, which will drastically reduce the amount of water reaching the lake and completely alter the hydrological regime. These hydrological changes have the potential to disrupt the lake’s ecology, including the productive fisheries on which indigenous communities depend. 

In this study, remotely-sensed products, produced as part of the ESA DUE Diversity II program, were used to investigate the natural spatio-temporal variability of Lake Turkana, and to define an ecological baseline against which future changes can be assessed. The Diversity II products, derived from MERIS imagery, provide information on lake water quality parameters including, chlorophyll-a, CDOM and TSM. These data were used to produce time series and maps of water quality in Lake Turkana, which showed a strong seasonal cycle and large spatial gradients in the lake water quality. A bloom in chlorophyll-a was observed at the north end of the lake, where the Omo River enters, which peaked in magnitude in Aug-Sep each year. Similar patterns were observed for CDOM and TSM. This water quality information was combined with hydrological variables, including lake levels and river inflow, and the results showed that the seasonal peak in lake productivity closely matched the annual flooding of the Omo River. The Gibe III dam will greatly modify hydrological cycles, including the seasonal flooding of the Omo River, and hence it poses a great risk to Lake Turkana’s productive fisheries.

The Diversity II products used in this study have provided unique insights into the phenology and the spatio-temporal dynamics of water quality in Lake Turkana. These insights would not be possible without remote sensing, given the vast scale, high degree of spatial heterogeneity and extreme remoteness of the lake, which make in situ studies very challenging. Work is ongoing to document the changes taking place in the lake now that the Gibe III dam is complete. This will be done using data from new satellites, including Sentinel-2 and Sentinel-3 OLCI imagery. Field campaigns are planned for Dec 2015 and April 2016 to validate the Diversity II products, and other satellite-based water quality products, for Lake Turkana. Initial results from these field studies will also be presented.

We present the results for fAPAR and LAI from the QA4ECV (Quality Assurance fo(u)r Essential Climate Variables) project, using white sky albedos in the visible domain (VIS) and the near infrared (NIR). The Two-Stream-Inversion-Package (TIP) relates white sky albedos from two spectral bands to effective fapar and LAI of the canopy layer, along with the full variance-covariance matrix of their uncertainties.
Building on work of Pinty et al. (2006) and Voßbeck et al. (2010), the method was extended to make use of the full uncertainty information coming with the albedos in the form of a variance-covariance matrix and to propagate this to the results. TIP finds optimal values of fAPAR and LAI by minimizing a cost function quantifying the misfit between the measured and simulated albedos.
The gradient is efficiently computed by the adjoint of Two-Stream which is generated by the automatic differentiation tool TAF (Transformation of Algorithm in Fortran). The error covariance of fAPAR and LAI is the inverse of the Hessian matrix at the minimum of the cost function. Again, this second order derivative code is generated by TAF. For performace, the results for all combinations of VIS and NIR albedos are tabulated and consolidated (Clerici et al. 2010) to Look-Up-Tables
(LUT). Compared to earlier versions of the algorithm, which were using a joint uncertainty on VIS and NIR albedos, the results show a general reduction of posterior uncertainty on fAPAR and LAI while largely reproducing their values. A comparison of the results using albedos from QA4ECV with those from the GlobAlbedo project is presented.

The most distinctive feature of wetland ecosystems is high water content at or near the soil surface. The preponderant role of humidity in the definition of wetlands implies that the hydrological dynamics in the system are a major determinant of the ecological processes occurring in wetlands and that a large fraction of the energy entering the system will translate into evapotranspiration (ET). In consideration of these wetland attributes, the Belgian Science Policy (BELSPO) supports the HiWET project where the suitability of ET estimates as indicator of wetland ecosystems health is investigated.

The present study is part of the HiWET project and focuses specifically on the utilization of remote sensing (RS) and other data sources in the derivation of ET by means of the computation of an energy balance. The envisaged ET estimates have spatial and temporal scales that allow, respectively, the discrimination of wetlands from adjacent ecosystems and the detection of seasonal anomalies. In this respect, this study considers the integration of RS products derived from the Meteosat Second Generation (MSG) satellite, in virtue of their high temporal resolution, and PROBA-V imagery, supplying moderate spatial resolution data. The study sites are three wetlands sites located in Europe where eddy covariance measurements are available for validation. The sites are: Anklam (Germany), Spreewald (Germany) and Stordalen (Sweden).

The methods are based on the ET product designed and set in operation in the framework of the LSA-SAF initiative (www.landsaf.meteo.pt). The LSA-SAF ET product is an operational product that exploit RS to generate ET estimates in near-real time at continental Europe/Africa. It is largely based on MSG observations and generates estimates at high temporal resolution (time step=0.5 hours). However, the spatial detail may be too coarse for monitoring small patches as it adopts the MSG grid (4-5 km pixel size in continental Europe). The latter is an important constraint in ecosystem monitoring in highly fragmented landscapes, as is the case in several regions of Europe. In this study we intend to improve the spatial resolution of ET estimates by taking the information on seasonal patterns of leaf area index (LAI) and albedo from PROBA-V observations and by considering the CORINE land cover map (pixel size ~100 m) in the derivation of land cover-related parameters.

At the moment of conducting this work, the PROBA-V LAI and albedo products are still under development by the Copernicus Programme (http://land.copernicus.eu). For testing purposes, LAI was derived from empirical relations based on PROBA-V NDVI (pixel size ~333 m and ~100m) and MODIS albedo (MCD43A3) was projected onto the PROBA-V spatial grid. The results revealed the potential of the upcoming PROBA-V LAI and albedo products to be part of the forcing in ET modelling algorithms. These results also open the gate for extending the application to the soon available Sentinel-3 observations.

Soil moisture observations are expected to play an important role in monitoring global climate trends. However, measuring soil moisture remains challenging because of its high spatial and temporal variability. Point-scale in situ measurements are scarce and remote sensing is the only practical means for obtaining regional- and global-scale soil moisture estimates. The ESA SMOS mission, launched in 2009, is measuring the Earth’s surface soil moisture at daily time scales with an unprecedented level of accuracy. Yet, incorporating these soil moisture observations to land surface models is complex since it is usually represented as an index of water content in the soil using different hydrological schemes and different physical approaches of the soil-water interaction. In this context, it has become important to characterize the climatological differences between models, satellite estimates, and ground-based measurements of soil moisture.

This study investigates the temporal dynamics of soil moisture and its anomalies based on three main data sources: SMOS, in situ observations, and the land surface reanalysis ERA-Interim. The analysis includes the determination of mean annual conditions, trends and anomalies for modeled and satellite-based soil moisture estimates. Their spatial coherence and relationships through precipitation and evaporation are also analyzed to improve understanding of global water cycle dynamics. Ground-based estimates are used as a benchmark in four target regions representative of arid, semi-arid, sub-humid and humid climates across global land biomes.

Results show that, despite still being a short data set, SMOS data provide coherent and reliable variability patterns at both seasonal and inter-annual scales. This work is an important support to all land data assimilation projects that are using remote sensing products to initialize forecast systems. 

Wetlands are important ecosystems providing numerous ecosystem services. They are of particular importance to communities in East Africa where small and large scale agriculture is the most important economic sector and where the availability of food to households is still an important factor of vulnerability. Their uncontrolled use, accompanied by use conflicts among farmers, cattle herders, wildlife and other users, poses a threat to the biodiversity of the vulnerable ecosystems. A sustainable use of wetlands where different human uses and the preservation of environmental functions is envisaged within the GlobE East African Wetland research project in Kenya, Rwanda, Tanzania and Uganda.

During an intensive field campaign to four wetland test sites within the East African region in the dry season of 2013 were Fractional Vegetation Cover (FVC) measurements, botanical vegetation cover and vegetation structure estimates acquired. In a stratified random sampling was each wetland area sampled with 250 m x 250 m tiles. Within each tile were different land use classes mapped on ground. FVC cover data were acquired for each land use unit in three strata: ground layer, midstorey and overstorey (woody vegetation greater than 2m).

Time series of FVC have been computed for Landsat and MODIS satellite data based on regionally developed algorithms in Australian Savannah environments. Fractional cover estimates for the green and no-green vegetative components for both satellite imagery and field work are evaluated for their usability in the East African region.

Aim of this exercise is i) validation of the existing FVC algorithm for the East African Wetlands field sites, ii) to delineate wetlands from their surrounding uplands, and iii) to utilize FVC estimates to characterize wetlands and wetland use to contribute to wetlands management decision making within the region. A five year time series of FVC from sequential MODIS satellite images is used  to differentiate between different, seasonally influences wetland types.

Operational Quality Control (QC) of ESA’s Soil Moisture and Ocean Salinity (SMOS) mission is currently carried out under the Image Data quality Evaluation and Analysis Service (IDEAS+). This paper focuses on the Quality Control (QC) operations for the v62x baseline, which includes reprocessing and operational v62x datasets.

QC checks are routinely applied over telemetry, calibration and scientific SMOS products with the aim of monitoring the quality of the data and detecting instrument or processing anomalies. Assessment of v62x data and plots of QC results are included.

Finally, the paper also presents details of where the users can obtain information related to the v62x baseline dataset.

The Sahel is a semiarid region south of the Sahara Desert that stretches from the Atlantic Ocean to the Red Sea. Mainly based on rainfed agriculture, the Sahelian region is known to be quite vulnerable to climate change and precipitation variability. In this context, national and regional institutions are interested in getting soil wetness conditions at the regional scale in order to derive/forecast crop yield estimations. To fulfill this objective, classical methods are based on daily precipitation measurements from the raingauge network, however, they suffer from the poor density raingauge network in Africa.

An alternative strategy is to use Land Surface Models (LSM) designed to determine multi-layer soil moisture conditions and related fluxes in response to atmospheric forcing. In this study, we performed soil moisture maps over the Sahelian region using the SURFEX LSM from Meteo-France, at the 0.25° resolution for the 2010-2013 period. First simulations showed that LSM outputs exhibited some large discrepancies with in-situ soil moisture measurements and associated fluxes when satellite precipitation products are used instead of in situ precipitation measurements. Moreover, LSM outputs were also found to be rather different according to the precipitation product forcing used in the LSM model (CMORPH, PERSIANN or TRMM-3B42).

In this study, it was proposed to constrained LSM soil moisture simulations with satellite soil moisture measurements provided by the SMOS (Soil Moisture and Ocean Salinity) mission. SMOS is a L-band radiometer using aperture synthesis and was successfully launched by the European Space Agency (ESA) on November 2, 2009. SMOS soil moisture product (L3SM) was assimilated into the SURFEX model using a Simplified Extended Kalman Filter (SEKF) scheme. Outputs of LSM+SMOS simulations were found to be in much better agreement with in situ soil moisture measurements. Another significant result deals with the soil moisture spatial distribution of the LSM+SMOS simulations which was found to be closely linked to the SMOS soil moisture variability.

These two characteristics (accuracy and spatial distribution of soil moisture) are closely related to the crop production since the soil water content is the main limitation of the plant growth in this region. Crop yield measurements were obtained over 10 farms located in the Niamey region in Niger from 2005 to 2012. The mean production (8 years) is 690 kg/ha, and ranges from 381 to 872 kg/ha during this period. Various statistical relationships based on soil moisture estimates were tested and the most promising one rely soil moisture anomalies from mid-August to mid-September in the root zone with crop yield anomalies. The statistical relationship provides an estimation of the crop yield with a correlation coefficient of R=0.98 using in situ soil moisture measurements. This result will be extended to the whole Sahelian region and compare to other ground-based crop yield measurements in Burkina Faso.

The northern forests act as carbon sinks and it is estimated that global warming and elevated CO₂ concentrations will enhance these forest capabilities to store carbon. However, a changing climate might also increases the frequency and severity of forest disturbance events resulting in reduced capacity of the forests to take up carbon; if disturbances increase more than forest growth there is a risk that forests will turn into sources rather than sinks. Hence, it is important to develop efficient methods that enable early detection as well as monitoring of forest damage, and to estimate the extent of these forest disturbances and the impact they have on the carbon cycle.

 

One cause of forest disturbance is insects. Insects’ response to a changing climate is an area of research with insufficient knowledge. This lacking knowledge makes future prediction of insect attacks uncertain and the importance of efficient monitoring systems crucial. Several studies have concluded that satellite based remote sensing can be used to detect insect damage in forests with high accuracy and various change detection techniques have been tested. Some of these studies are based on time-series analyses.

 

In this study we present a remote sensing method for near real-time monitoring of insect disturbances in forests that was developed and tested in the mountain birch forests around Abisko in northern Sweden. Furthermore, we utilize GPP derived from eddy-covariance fluxes to estimate the impact on GPP of an insect outbreak in the study area. The disturbance monitoring method utilizes MODIS data with 250 m spatial and 8-days temporal resolution and a Kalman filter is applied to handle noise in the data. A seasonal trajectory of NDVI for birch forest with no disturbances is identified, where healthy conditions are defined as the n years with highest summer NDVI. The method is then applied per-pixel and disturbances are identified as deviations from the healthy seasonal trajectory. Cumulative sums are used to identify when deviations will be sufficiently large to be classified as damage depending on a threshold value. The threshold can be set to favor damage detection or to avoid misclassification of healthy pixels. Evaluation showed that the method detected 66% of the defoliation with 40% misclassification of healthy pixels; an adjusted threshold resulted in 93% of the damage being detected with 60% misclassification of healthy pixels. The impact from the insect outbreak on GPP is estimated by establishing a model for GPP for years without disturbances and comparing modelled GPP with GPP from a year with an insect outbreak.

The Earth system is increasingly impacted by human activities. The high demand for natural resources, human interference in biogeochemical cycles, or the conversion of functional land surfaces and ecosystems cause notable changes within the Earth system. The changing climate system will affect natural water balances and water availability, thus impacting biological processes, energy cycling, or human wellbeing. The growing human demand for water will further increase the pressure on global water resources.

New data collection strategies including remote sensing are urgently requested to broaden understanding of the water cycle and face upcoming challenges related to natural water resources. In this context, vegetation ecosystems play a crucial role for comprehensive water cycle research. Vegetation and vegetated ecosystems are an important water infrastructure and a critical component of the water cycle: roughly 70% of the land surface is covered by vegetation and deep root systems integrate a relevant part of the soil layer. Vegetation ecosystems largely mediate water exchanges between the soil and the atmosphere through plant transpiration, the dominating but variable water flux contributing to evapotranspiration rates of vegetated ecosystem. Further, plant transpiration is essential to facilitate the concept of water use efficiency, representing the rate ratio of carbon assimilation and transpiration, and having significant implications for food security or ecosystem functioning. Although various remote sensing based strategies exist to obtain estimates of plant transpiration, approaches are still less reliable than required.

The new Earth Explorer 8 “Fluorescence Explorer (FLEX)” will provide critical measurements of vegetation ecosystems that, in combination with process models, can facilitate new strategies to advanced water cycle research. We outline a conceptual framework based on combined FLEX observations and process models to derive plant transpiration and water use efficiency. We discuss implications of such products for various Earth science applications.

In recent decades, importance of data collection of SAR satellites increased in earth observation.

Results achieved in land cover surveys as well as the broadening and uninterrupted availability of SAR data gives a reason for investigation of the possibilities of its usage regarded to land cover type classification.

 Injection of SAR imagery based information in production of Copernicus High Resolution Layers (HRL)s can help to refine information served by optical satellite imagery. Together with a-priori knowledge it may overcome the gaps caused by the cloud cover issue. However, it requires methodological adaptation, given the reasonably different nature of SAR compared to optical data. The methodological adaptation shall allow operational implementation, and shall help reducing the elapsed time between available satellite imagery and derived information services. It requires the analysis of the potential usage of SAR based imagery within the COPERNICUS land context, supported by case studies.

In this presentation the contribution of radar polarimetry for distinguishing land cover categories is evaluated, as these categories show unique spectral as well as geometrical structural characteristics (e.g.: forest, grassland, wetland, imperviousness).

For this purpose two individual study areas were selected in Hungary (vicinity of Lake Tisza and Lake Balaton), having complex characteristics and fairly rich in different types of land cover categories. Comparative analysis was carried out, focused on the efficiency in distinguishing land cover categories by using different type of optical and radar satellite images for classification.

The following optical HR and VHR satellite images were taken into our analysis: Landsat TM8, SPOT5, Pleiades, Worldview2, Sentinel2.

Polarimetric radar satellite images involved in our analysis are listed below: RADARSAT2 fine with different polarisations such as quad, dual (HH+HV), single (HH), TERRASAR-X stripmap dual polarisation (HH+HV) data, and freely available time series of Sentinel1 dual polarisation (VV+VH) satellite images.

Different types of spectral indices were derived from optical satellite images. Polarimetric descriptors were generated based on three different decompositions (H/A/Alpha, Yamaguchi, Touzi) of coherency matrix in case of quad-pol images, and H/A/Alpha decomposition of covariance matrix for dual-pol satellite images.

 Based on the preliminary results of our case studies, it can be declared that SAR images have notable relevance for each layer of Copernicus HRLs. Synergistic usage of radar and optical images can improve the quality of results in identification of HRLs. The abundance of time series of Sentinel1 images provides improvement of HRLs identification. The contribution of Sentinel1 time series in HRLs mapping is significant, especially in case of grassland and imperviousness.

Evapotranspiration is the water exchanged between the surface (soil, rivers and vegetation) and the atmosphere. Monitoring evapotranspiration is important from primary to tertiary sectors of human activities, as it is an observation of soil water and temperature loss. At continental scale, it is essential for climate scientists: evapotranspiration is a major component of the water cycle strongly connected to the energy cycle of the Earth system.

We introduce one of the few satellite-based operational evapotranspiration (ET) products, generated continuously and in near real-time over Europe, Africa and part of South America. The ET products(30 minutes instantaneous and daily) are generated at the EUMETSAT’s Satellite Application Facility on Land Surface Analysis (LSA-SAF) operations centre.Since 2009, the continuous operational generation and dissemination free of charge of the LSA-SAF instantaneous and daily ET products has been ensured (http://landsaf.meteo.pt). The last product generation survey (operations report for the first semester of 2015) shows that overall product generation is above 95% providing evidence on the reliability of the operation chain. Different publications in international peer-reviewed journals attest on the expectations but also on the confidence of users of the ET products on different application domains. Following our commitments to our user’s community, we are continuously performing research, looking for new ways to improve the product. To accomplish this, the input/feedback from users and potential users of the products is also of great interest.

In this contribution, we outline the strategy we develop in the context of LSA-SAF, show examples of the products and address possible applications.

SAR Polarimetry, constitutes a major and separate field of SAR Remote Sensing

since has its own rules and theory and it appears such capabilities that can offer

notable results contributing in studying of specific phenomena on earth that no

other science could do with such accuracy, speed and detail.

Amazon Rainforest is one of the biggest and more important rainforests in Globe

and its observation and protection are of extremely high importance. SAR

Polarimetry has the advantage of achieving to study in depth the vegetation in

such areas where optical satellite instruments are just useless. Moreover it can

distinguish the different vegetated areas and cultivated parts of land.

In this study we used SAR satellite data, SAR Polarimetry and multitemporal

methods for analyzing the data and monitoring the forest vegetation of the area.

The area under investigation was an extended area of Brazilian Amazon Forest

that covers the states: Para, Tocantins and Goias. The area is covered by crops,

residential areas and rain forest land too. So it is about a very interesting natural

area suffered by the human stress.

To address the above aim we have used 2 different datasets: 10 ENVISAT

ASAR.PRI ascending scenes, VV, swath IS2, track 131 covering the period 2004-

2006 and 27 ALOS-PALSAR ascending scenes, antenna pointing right, fully

polarimetric product 1.1 covering the period 2009-2010..

Both the two data sets cover a long year period of almost 7 years with break 2007-

2008, allowing to apply a multitemporal study. Processing of the scenes was carry

out using Sentinel-1 tool box for ENVISAT data set and PolSARpro for ALOS

data.

Envisat images, spatially cover the Western part of Tocantins state where the

Tocantins River cross the area while the main residencial centers are Crixás do

Tocantins and Aliança do Tocantins and the rest is rain forest, were split up to 2

subsets. The first group of 7 scenes was used for applying a seasonal analysis of

the area since all the images are of the same date (2006). The other group of 4

images was used for multitemporal analysis since the scenes refer to different year

dates (2004-2006). Processing of ASAR.PRI data (30 m ground range x 30 m

azimuth) includes corregistration of dataset’s scenes and application of Lee (5x5

kernel) and Frost (7x7 kernel) filters for better speckle reduce.

The area that the 24 ALOS scenes cover includes the states of Pará, Amapá,

Tocantins and Goiás.

As for the ALOS dataset processing main steps include: Data extraction,

Geocoding, coherence elements production, Decomposition parameters

production (H/A/alpha), polarimetric segmentation, H/A/alpha classification,

Wishart classification as well as Freeman, Yamaguchi and Van Zyl

decompositions which showed great results.

In this image, we can see the results from the RVI application in an Amazon River

area on the state of Para. RVI is actually the Radar Vegetation Index which helped

us figure out many changes in the vegetation cover of the study areas. In these

scenes, we can distinguish some differences in the river banks and that the

vegetated and marshy areas on the North-West are presented more intense in the

date of 2009. Of course, we must consider that these dates refer to different

seasons (May and December respectively).

In the transition from fossil and nuclear fuels to renewable and sustainable energy sources, solar energy plays an important role globally, with even increasing relevance in the near future. There is a large and still mainly unused potential for the conversion of solar irradiation into electricity and heat especially in the sunny regions of the earth. The economic and ecological sustainability of installed solar systems is principally dependent on the longest possible product lifetime and maximum reliability. However, due to outdoor exposure in different climate zones of the earth, many of them with harsh environmental conditions, the various materials used in solar power systems are negatively affected by stress factors such as UV radiation or atmospheric corrosivity. Depending on the geographical location, the individual or combined impact of environmental and climatic stress factors and corresponding performance losses vary significantly. Therefore, as a strategy to reduce investment risks and operating and maintenance costs, it is necessary to adapt the materials and components of solar energy systems specifically to regional environmental conditions.

The pilot project «GloBe Solar» supports this strategy by focusing on the development of a global stress classification system for materials in solar energy applications. The aim of this classification system is to assist in the identification of the individual stress conditions for every location on the earth’s surface. GIS based global models and classification systems for different stress factors: UV irradiation, dust accumulation (soiling), atmospheric corrosion and thermo-cycling are developed within the project. Therefore satellite data and available ground-based measurements are forming the primary data basis. The backbone remains the focus on vast archives of historical meteorological and earth observation data like e.g. the ones from the ECMWF in the framework of MACC-II (the European Union’s Copernicus Service “Monitoring and atmospheric composition and climate”) or from the Climate Forecast System Reanalysis and Reforecast (CFSRR) from National Oceanic and Atmospheric Administration (NOAA). Furthermore MACC-II provides atmospheric services that are of great value to the project like UV observation or noxious gases like NO_x and SO₂ that are integrated to atmospheric corrosion models. Land use maps derived from 300m MERIS sensor on board the ENVISAT satellite mission are provided by European Space Agency (ESA) as GlobCover maps are used to obtain up-to-date high-resolution information about relevant parameters like land-cover and land-cover change, land-use, soil and water cycle, etc. and then combined with suitable meteorological data like wind conditions, precipitation and humidity resulting in a global map of soiling risk.

A new synergic opportunity opens with launch of European Earth Observation Sentinel satellites. The integration of these datasets will make it possible to improve and calibrate the models as well as to validate the underlying models and reduce uncertainty in modelling elements. This data will bring much higher spatial detail of mapped atmospheric parameters than current sources thereby allowing detailed analysis of spatial variability of stress factors and uncertainty of delivered maps arising from this variability. Referred data are not only analyzed to model various stress factors on solar energy systems, but also to develop a geospatial data infrastructure which combines the satellite data and meteorological datasets (data layers), processing tools and models into a Stress Classification System (SCS). The developed stress models and the classification systems will be embedded in an ISO/OGC compliant GDI, offering standardized interfaces to the layers of the system as well as to external users. A metadata catalogue will publish metadata so that services can find and bind to these services. The model results will be registered within the GEOSS portal as a Web Map Service.

Although MIRAS is operating in a protected band, the contamination of L1a data by man-made sources of radio-frequency interferences (RFIs) operating close to the L-band are frequently observed. Although the geolocation of such RFI sources has already been addressed, the accuracy is almost always kilometric (> 1 Km). This contribution is devoted to a major improvement brought to this problem by introducing an approach that performs the geolocation of such RFI sources with a sub-kilometric accuracy (< 1 Km).

RFI sources are modeled according to a parametric model where the location and the brightness temperature of the source at ground level are the key parameters. These parameters are optimized in a constrained iterative process where the visibilities corresponding to a given RFI source are computed with the actual modeling of the instrument and subtracted from the visibilities acquired by MIRAS. Since there is no guarantee that a given RFI source is found at the same ground location in all the snapshots, the average or the median location is retained.

The performances have been addressed with numerical simulations as well as with SMOS data contaminated by RFI. Among the many RFI sources affecting SMOS data worldwide, few of them have been geolocated. For most of them, ground coordinates have been provided by national spectrum management authorities of the corresponding country. For a very small number, these coordinates have been cross-checked with independent GPS measurements and/or accompanied with a ground truth or an evidence. This is the case of an RFI source in Ukraine as well as with another one in Germany.

In the first case, more than 1100 snapshots from 13 orbits passing close to the emitter in June 2011 have been processed in X and Y polarizations. The final accuracy on the location of this source is about 125 m.

The same processing has been performed over Germany, but over a longer period of time (6 months) because this RFI source turned out to be frequently switched off and often contaminated by other emitters. Nevertheless, the final accuracy on the location of this RFI source is about 140 m.

Distributed hydrological models of energy and mass balance usually need in input many soil and vegetation parameters, which are usually difficult to define. This paper will try to approach this problem performing a parameters calibration based on satellite land surface temperature data (LST) as a complementary method to the traditional calibration with ground data.

A pixel to pixel calibration procedure of soil hydraulic and vegetation parameters for each pixel of the domain is proposed according to the comparison between observed and simulated land surface temperature.

A distributed hydrological model, FEST-EWB, that solves the system of energy and mass balance equations as a function of the representative equilibrium temperature (RET) will be used. RET is comparable to the land surface temperature as retrieved from operational remote sensing data.

This equilibrium surface temperature, which is a critical model state variable, is compared to land surface temperature from MODIS.

A similar calibration procedure will also be applied performing the traditional calibration using only discharge measurements at different available control cross sections.

The dynamic of two big lakes in the basin will also be analyzed; FEST-EWB model is calibrated using MERIS data to infer surface area changes and altimetry data from ENVISAT, JASON, Topex/Poseidon (LEGOS) for water level.

The advantage of performing an internal validation of FEST-EWB model using satellite LST has been shown.

This approach will be tested for the Yangtze River basin (China) from 2003 to 2006, a basin of about 2’000’000 km², within the DRAGON 3 project in the framework of ESA and MOST collaboration programme.

 

Biomass is an essential indicator for monitoring the Earth’s ecosystems and climate. It is a critical input for greenhouse gas accounting, the estimation of forest losses and degradation, renewable energy assessment, the development of climate change mitigation policies such as REDD+, and many other applications. Rapid, recurrent, wall-to-wall biomass assessment is possible with remote sensing (RS) only. However, there is no RS instrument for direct biomass measurement. All available or foreseen methods require extant, diverse, up-to-date, reliable and comparable in situ data for calibration/validation.

The International Forest Biomass Network (IFBN) is an international cooperation to establish and maintain a global in-situ forest biomass database. The IFBN project is launched under the auspices of the European Space Agency to provide ground biomass data measurements for the Earth Explorer mission BIOMASS. The project aims to establish a data sharing policy and overcome the existing barrier between data providers and users. Another important task is to define the technical requirements to ensure data comparability.

IFBN will include national and international consortia (e.g. AfriTRON, RAINFOR, CTFS-ForestGEO) devoted to coordinating long-term research with permanent sampling forest plots. They have (i) established permanent sampling plots, (ii) proposed robust protocols for accurate tree mapping, and measurement, (iii) monitored existing plots repeatedly, and (iv) established databases with a special emphasis on data quality control at the tree level and valorisation of historical databases.

A minimum set of database values include: principal investigator and institution, plot coordinates, number of trees, forest type and tree species composition, wood density, canopy height, above ground biomass of trees over 10 cm in diameter. Plot size is 1 ha (preferably) or at least 0.25 ha.

This database will be essential for validating and calibrating satellite observations and various models. The focus is to provide ground support for the ESA Earth Explorer BIOMASS mission. We are looking into synergy with other ongoing projects (e.g. GlobBiomass) and other missions (e.g. NASA GEDI, NISAR and JAXA ALOS).

The accident that damaged the Fukushima Daiichi Nuclear Power Plant site in March 2011 resulted in massive atmospheric releases of radioactive substances (mainly iodine, cesium and tellurium isotopes) that were dispersed and deposited onto the continental and oceanic surfaces. The widespread deposition resulted in populations being exposed to γ radiation from external exposure or through the ingestion of contaminated foodstuffs. The most seriously contaminated terrestrial areas are within 80km of the damaged site, and comprise forests (70%), agricultural fields (including rice paddy) or meadows (20%), and inhabited areas (10%).

For the purpose of several projects related to the post-accidental management of the nuclear accident at Fukushima-daichii and the understanding of radionuclide transfers in various compartments of the environment, it is necessary to have a good land use, recent and accurate as possible (minimal mapped unit less than one ha).

A land use inventory for this area of Japan has been achieved but for the moment (ALOS, Globe land 30, Kondo, ...) none of them was satisfactory: too old, not suitable nomenclature Summary cover, large unit (25 ha) ...

Therefore, it was decided to conduct a land use by satellite images digital processing, an area of about 100 km around the Fukushima-daichii. Soil elements to be identified are those classical used and defined by the modeling tools. Although accurate, the nomenclature is nevertheless simple (about 10 classes).

For a near field (50 km area), we used SPOT5 scenes. Two diachronic images were available with a relatively low proportion of clouds (November 13, 2013 and July 22, 2014). Rice paddy fields are clearly identifiable, these crops should appear vegetated (or tilled) in November and irrigated in summer period. The deciduous/evergreen differentiation is also obtained for these two dates, the level of leaf chlorophyll being much lower in mid-November.

For a field called distant (100 km area), we used LANDSAT8 scenes. Only one scene covers almost the entire area. Three images were available with a small amount of cloud (September 17, 2013, April 13, 2014 and May 31, 2014). The scenes of 2014 allow discrimination wooded surfaces deciduous and evergreen. From its vegetated state, the rice fields are more difficult to identify, a scene from mid-September 2013 was therefore used.

The advantage of using SPOT5 scenes is the fact that the resolution is the same of Sentinel2A. The algorithms used may be resumed when Sentinels2A scenes will be available. Like LANDSAT8 scenes, one can expect a good durability and can detect changes in the landscape over several years (even decades).

The processing of SPOT5 images and LANDSAT8 images is operated by supervised classification based on SVM algorithm.The "Media Vector Machines" (SVM or French in support vector machines) are a supervised statistical learning method. Considered in the literature as the most efficient classifier, the SVM algorithm is integrated into the library Open- source Orfeo Toolbox (OTB) developed by CNES.

The challenge for discriminating classes of land use is to treat multi-temporal images.Rather than treating each data independently, multi-temporal spectral signatures can be directly analyzed by a single process. For this, it is necessary to stack the tapes of all the images into a single concatenated data on which can be carried supervised classification. This is an immediate medium and without intermediate treatment to take into account the multi-temporal spectral characteristics.

The selection of samples for learning and training of the classifier is undoubtedly one of the most critical phases of the supervised classification.Indeed sample selection techniques are numerous and complex, but there is however no "rules" absolute. The sample design requires a high level of expertise and can be tedious and time consuming for large surfaces, as in this study.

The oral presentation will show the obtained land use and will continue with some results from models that take into account this land use : marketing limit, wet and dry deposition, uncertainties, leaching watershed …

Evapotranspiration is a fundamental variable of the hydrological cycle and plays a major role in surface water and energy balances. The estimation of evapotranspiration is required for water resources management and climate studies. At the local scale evapotranspiration can be accurately determined from detailed ground observations (eddy covariance towers, lysimeters) but these measurements are too time-consuming and costly to assess the evapotranspiration variability at the regional scale. Therefore, remote sensing provides cost-effective methodologies to assess the spatial distribution of evapotranspiration at regional scale.

EVASPA (EVapotranspiration Assessment from SPAce) is a tool that has been developed to produce evapotranspiration maps at relevant spatial and time scales for hydrological or agronomical purposes. The tool includes several evapotranspiration estimation methods (S-SEBI method, the triangle approach and aerodynamic equations from SEBAL and SEBS) and various equations for estimating the required input information (e.g. albedo, net radiation and ground heat flux). Highlighted features of this tool are: (i) the possibility of integrating and comparing data from various remote sensing sensors, (ii) a flexible design which allows easily adaptation to new sensors, (iii) options to evaluate the uncertainties related to the evapotranspiration estimates obtained through the different approaches and (iv) the provision of continuous daily evapotranspiration maps including for days without available remote sensing images (by means of interpolation procedures).

Using MODIS data (freely available since 2001), EVASPA makes it possible to provide long time series of daily evapotranspiration with a 1km spatial resolution over areas of 1000 to 10000 square kilometres. In this study, EVASPA is used for evaluating the evapotranspiration term in the water budget of the Crau aquifer (543 km²) over 15 years. As an annual mean evapotranspiration represents 240 hm³, to be compared to 300 hm³ given by rain. Uncertainty in annual evapotranspiration represents +/- 20 hm³. The EVASPA estimates are evaluated against surface energy balance data acquired (using flux towers) over different ecosystems within the Crau-Camargue area, south-east France (saltmarsh scrubs, dry grassland, irrigated grassland, agricultural area).

The EVASPA tool is a software currently in the prototype phase and is developed with support of CNES through the TOSCA research calls and within the frame of the European FP7 project SIRRIMED.

This paper reports interim results of the North State project of the Framework Program 7 of the European Union (grant agreement 606962). Several other abstracts have been submitted to the Symposium on the details of project achievements. At the time of writing the abstract the project has completed its two first years and it will continue for one more year until fall 2016.

The project will demonstrate how innovative methods applied to the Sentinel data streams and other satellite data can be combined with models to monitor carbon and water fluxes for pan-boreal Europe. Incorporation of the Sentinel data in the flux estimation is expected to improve accuracy and improve the timeliness of flux estimation.

The project will provide a paradigm for the development of products for future Copernicus services that will be applicable far beyond its specific application to the boreal zone.

The North State has a south boreal and north boreal sites of 200 km by 200 km sites for intensive study in Finland, and one site in Iceland and in Russia. Additionally, the fluxes will be modelled over a pan-boreal – pan-arctic region reaching from Iceland to the Ural Mountains.

The two applied carbon flux models are a Dynamic Global Vegetation Model of the University of Sheffield and a semi-empirical forest specific model of the University of Helsinki. Basically these models are designed for different spatial extents and grain sizes, which makes their inter-comparison particularly interesting.

The identified EO based variables for the models are: land cover classification, plant functional type or main tree species, logging events and burned area, fAPAR or alternatively Leaf Area Index, tree biomass, tree height, site fertility, and snow cover and snow water equivalent. The applied pixel size in the detailed products is 10 m or 20 m and 500 m for the large north European area.

The image analysis methods include in-house Probability fuzzy estimation and Gaussian process regression methods in addition to more conventional approaches. New algorithms have been developed for the pre-processing and change detection from Sentinel-1 imagery.

A specific objective of the study is development of a self-learning automatic image analysis system that accepts data from many sources. An initial version of the system will be developed by end 2015. It is based on continuously collecting measurements from several satellites to a grid database. The initial values of the ground reference data at each grid point are improved by applying intelligent image analysis to the frequent satellite measurements. The updated grid point values are further used as training data for automatic variable prediction method using Copernicus data.

At the time of writing this abstract results from the Finnish sites with Landsat 8 data are available. The root mean square errors for the growing stock volume that is closely correlated with above-ground dry tree biomass and height were approximately 40 % of the mean at test point level when tested with independent stand based reference data. The bias was from few per cents to ten percent for an area of some hundred hectares. The details will be given in separate papers.

The first version of the large area estimation will be computed by end 2015. Also the satellite image predictions will be applied in the models by the end of the year. 

Over land, the objective of SMOS is to retrieve soil moisture (SM) and vegetation optical depth (VOD) of observed surfaces. In case of dense vegetation cover the retrieval is hampered by the high canopy attenuation, which reduces the sensitivity to variations of soil moisture, particularly when the brightness temperature data are affected by noise and/or Radio Frequency Interference. Further specific problems are present in case of mixed pixels. After several efforts the impact of these problems has been substantially reduced, at least in terms of permanent retrieval failures. Retrieval outputs of last L2 version (V620) offer almost complete coverage of the earth surface. In this framework, the objective of this work is to test the quality and the application potential of outputs collected over forests and mixed pixels.

Preliminary tests of V620 were performed with short temporal series (4-day and 8-day) globally, or 2-year series over limited surfaces. These tests showed that forest VOD is significantly correlated with forest height estimated globally or forest biomass estimated at regional level. Comparisons with similar AMSR-2 product were also performed. On its turn, SM retrieved over forests was compared against available ground measurements, particularly using SCAN/SNOTEL US network. The error was higher than with low vegetation surfaces, but tests were limited by scarce availability of sampling points in forest pixels, and the algorithm version was not yet the definitive one.

At the time of this submission, final V620 products are available from May 2015 to October 2015. A preliminary analysis of results fully confirms previous findings about correlation of VOD with forest height estimates. Moreover, monthly average of VOD is stable with time, which was expected due to the dominant role of forest branches in producing attenuation at L band. SM maps show appreciable spatial and temporal variations in Tropical and Subtropical regions, while the dynamic range is still limited in Boreal regions.

As soon as V620 data will be fully available, seasonal and inter-annual variations of VOD will be investigated in detail, and comparisons with AMSR-2 product will be completed systematically. This analysis will allow us to estimate the added value SMOS VOD product can give in monitoring forest biomass and forest carbon. Trends of retrieved SM in Tropical and Subtropical areas are being compared against rainfall data bases. Detailed comparisons with ground measurements are in progress in US areas with dense network measurements, which were not covered by orbits available during preliminary tests. Other areas covered by FLUXNET network will be also considered.

Finally, the outputs of a 3-parameter retrieval work (i.e. simultaneous retrieval of SM, VOD and albedo over 8-day temporal intervals) will be presented. It was found that retrieved albedo shows a high dispersions over mixed pixels and/or regions with low VOD, but tends to values close to 0.06 in all dense forests. These results are important in view of future improvements to be applied to SMOS L2 algorithm over land.

Sun-Induced Fluorescence (SIF) estimates from remote sensing data are becoming more and more popular, from leaf level to global scale. This signal is a promising tool for the remote estimation of plants functioning, and its accurate quantification from space in the red and far red regions is the main aim of the Florescence Explorer (FLEX) mission, candidate for European Space Agency eighth Earth Explorer (ESA, EE-8). Although it is inherently related to plant photosynthesis efficiency, remotely sensed SIF signal is also highly influenced by environmental variables and vegetation biophysical and biochemical parameters. According to the recent Global Sensitivity Analysis (GSA) of the state-of-the-art Soil-Canopy Observation of Photochemistry and Energy fluxes (SCOPE) model, the maximum rate of carboxylation (V_cmax) drives only a relatively small portion of the SIF signal. In turn, variables such as broadband incoming shortwave radiation (R_in), Leaf Chlorophyll Content (LCC) and Leaf Area Index (LAI) are major drivers of SIF intensity. Their variation in the spatial and temporal domains, if not properly accounted for, can lead to biased interpretation of the remotely sensed SIF signal in relation to the physiological behaviour of the vegetation. In this contribution we evaluate the potential of normalizing SIF by different quantities related to the incident photosynthetically active radiation (PAR), the photosynthetically active radiation absorbed by the canopy (APAR) and traditional vegetation indices linked to canopy greenness, for minimizing the influence of R_in, LAI and LCC on the SIF signal.

The analysis is carried out using a simulated dataset generated with the SCOPE model, setting all variables fixed except for R_in, LAI and LCC. For these three variables, a broad range of input values is tested . The effect of the normalization methods is evaluated individually for several SIF wavelengths in the red and far-red emission regions.

The indices related to APAR perform best among all investigated ones, and are followed by the vegetation indices linked to canopy greenness and finally by metrics related to the incoming radiation. The use of APAR performs well in normalizing the effect of LCC variations on both red and far-red SIF, while LAI influence is still affecting the red SIF even after the normalization.

The metrics proposed for SIF normalization are then tested using an experimental dataset collected in August 2015 in an experimental farm in Northern Italy, over two monospecific soybean vegetation patches (a wild variety and a mutant one). Along with high spectral resolution top of canopy measurements, the main canopy biochemical and structural characteristics of the two soybean varieties were also measured. In-situ results confirm those obtained from the analysis of canopy fluorescence emission spectra simulated with the SCOPE model.

A catalogue of Radio Frequency Interference (RFI) sources covering all the SMOS mission in the 1400–1427MHz Passive Band has been performed with a new detection, monitoring and analysis tool implemented at ESAC.

The poster presents the catalogue, the tool and the used strategies together with complementary capabilities. These are oriented to improve the catalogue analysis and its use for on-site RFI identification. For this purpose, the tool functionality provides RFI source classification and geocoding of compact sources with sub-kilometre accuracy. In order to complete all these functions, the tool uses in parallel standard SMOS L1B and L1C Full Polarisation Products.

The tool is modular and implements the following functions:

1. Detection area segmentation through the use of different RFI signatures

2. RFI area clustering and classification: It is performed with an adaptation of a standard SW used for source detection in astronomical image processing [1]. It allows to determine the source location by means of the barycenter of the clustered RFI areas and it permits to classify the RFI sources. Each detected source is classified by its extension (compact or extended), variability (continuous or variable) and persistence (transient or permanent).

3. Sub-pixel RFI geocoding: It is achieved through the modelling of the source as an elliptical gaussian emitter in all the SMOS L1C snapshots where it appears. After the average of the different solutions an accuracy of 4 km is achieved as it has been validated with a series of RFIs with known position.

4. Sub-kilometre Optimisation: Using as a 1st approximation the location obtained before, the geographical position is refined until a sub-kilometre accuracy by means of the adjustment of a model of the Image formation in the SMOS Antenna Plane with the corresponding RFI observations in the non-projected L1B Full Polarisation Products.

5. Detection merging and geographical matching: a RFI catalogue is produced merging the different detections by means of the analysis of the persistence of the sources. The catalogue includes the RFI coordinates, characteristic of the source (brightness temperature, extension, variability, etc.), activity period and geographical identification.

A first catalogue is being producted with the complete series of reprocessed SMOS L1 V620 products. Once the first version is validated,  the catalogue will be available at ESAC premises.

[1] Bertin, E. and Arnouts, S. 1996: "SExtractor: Software for source extraction", Astronomy and Astrophysics Supplement, v.117, p.393-404.

Crop extent and frequency maps are an important input to inform the debate around land value and competitive land uses, food security and long term impacts and sustainability of agricultural practices. Such spatial datasets are likely to support decisions on natural resource management, planning and policy. The complete Landsat Time Series (LTS) archive for 23 Landsat footprins in western Queensland from 1987 to 2015 was used in a multi-temporal mapping approach. Spatial, spectral and temporal information were combined in multiple crop-modelling steps, supported by on ground training data sampled across space and time for the classes Crop and No-Crop. Temporal information within summer and winter growing seasons for each year were summarised, and combined with various vegetation indices and band ratios computed from a mid-season spectral-composite image. All available temporal information was spatially aggregated to the scale of image segments in the mid-season composite for each growing season and used to train a random forest classifier for a Crop and No-Crop classification. A threshold of 0.5 applied to the predicted probabilities of the random forest enabled mapping of Crop and No-Crop for each year and growing season. Validation revealed that the predictive accuracy varied by growing season and region to be within k = 0.87 to 0.96 and are thus suitable for mapping current and historic cropping activity. Crop frequency maps were produced for all regions and different time intervals. The crop frequency maps were validated separately with a historic crop information time series. Different land use intensities and conversions e.g. from agricultural to pastures are apparent and drivers of these conversions will be discussed.

1.      INTRODUCTION

In the last few decades huge scientific and technological progress has been made in the field of remote sensing. New platforms and instrumentation have been designed and successfully launched and many more are scheduled to follow in the future, like the Sentinel constellation within the Copernicus project [1]. In general, the role of remote sensing is providing  accurate, timely and easily accessible information to improve  management of the environment, understand how the human settlements are evolving and mitigate the effects of natural distasters.

In this context, the UK Space Agency (UKSA) has funded several projects within the International Partnership Space Programme (IPSP) with the goal to enhance  knowledge in the space segment and to develop novel tools for the Earth Observation (EO) [2]. Indeed, using remote sensing data for EO has not found yet a substantial take-up among end-users because the remote images are usually hard to interpret and understand, are often expensive, and the fusion of different types of image data (i.e. optical and SAR) acquired from different sensors and platforms, is not a trivial task.

The “Earth Observation for the preservation of the Bacalar area” (IPSP-Mexico in short) is one of the projects funded within the IPSP programme, and supported by the UKSA and the Agencia Espacial Mexicana (AEM). The aim of this project is to develop strong commercial and academic links between UK and Mexico for the exploitation of EO data to adress real-world problems. In particular, this project will promote the development of end-user-tailored products on a target area and show the value added to any monitoring activity based on EO satellite data [3].

2.      Area of sudy

IPSP – Mexico is focusing on the monitoring of the Bacalar area and associated transverse coastal corridor. This is a unique and fragile region between the Bacalar Lagoon (an international relevant wetland) and the Caribbean Sea. The area is rated as the highest at-risk conservation area in the Caribbean ring with high vulnerability to human activities and a need for sustainable stewardship. In addition, the Bacalar region suffers from the fact that potential of remote sensing is not properly exploited and applied yet; so this project presents an ideal opportunity to demonstrate the real value of EO in addressing societal problems. The IPSP, for example, aims at providing the Mexican authorities with evidence, properly derived by satellite imagery, in support of scientific arguments for the international protection of the Bacalar area. 

3.      Monitoring

Near-real-time applications are currently under design to monitor the impact of human activities in the area. The last 10 years have been selected as temporal frame in which, by employing different data sources, changes in the area, due to human activities, can be appreciated. In particular, SAR images (from TerraSAR-X and Sentinel-1 missions) and optical images (from Deimos-2 and, where possible, Sentinel-2) will be processed to produce maps [4] at different frequencies of the various ecosystems present in the area (agriculture crops, forestry, water bodies, mangroves). Remote sensing maps will be validated against in-situ measurement (where available) from the AEM and ECOSUR archives.

Results will be provided, commented and compared at conference time. Research outcomes and impacts will be analyzed to support interventions by the Mexican government in the area of study. In addition, all data products created by the project will be integrated with a web-based geospatial data platform created to a bespoke design, and made available to the project consortium and end users to support the near-real time monitoring of activities in the Bacalar region.

 

4.      References

 

[1] European Space Agency (ESA), “Copernicus: observing the Earth” [online] http://www.esa.int/Our_Activities/Observing_the_Earth/Copernicus/Overview3 (Accessed: 08/10/2015).

 

[2] UK Space Agency (UKSA), “ International Partnership Space Programme” [online]

https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/383540/IPSP_information.pdf(Accessed: 08/10/2015).

 

[3] Surrey Space Centre, “International Partnership Space Programme” [online] http://www.surrey.ac.uk/ssc/IPSP/index.htm (Accessed 15/10/2015).

 

[4]  S. Khan, A. Doulgeris, S. Savastano. R. Guida, “Automatic clustering of multispectral data using a non-Gaussian statistical model”, proc. IGARSS, pp. 4276-4279, 2014.

 

 Sun-induced fluorescence (SIF) emitted by chlorophyll molecules is one of three main de-excitation mechanisms for energy captured by light harvesting pigments in plants. SIF emitted by vegetation is seen as a meaningful indicator of plant stress, instantaneous plant photosynthetic function (e.g., carbon fixation), and possibly gross primary productivity (GPP) at the ecosystem scale. Although the SIF flux emitted from the canopy is relatively small compared to reflected sunlight (about 1-5% in the NIR), it is a broadband spectrum within the 650–800 nm spectral window that contains useful information.

Currently, no missions operative in space are optimized to capture the full SIF emission spectrum. To fill this gap, ESA recommended the Fluorescence Explorer (FLEX) as the 8^th Earth Explorer mission dedicated to measurement of the broadband SIF signal. The FLEX satellite, equipped with a Fluorescence Imaging Spectrometer (FLORIS) onboard, has been proposed to operate in a tandem mission with ESA’s Sentinel-3 satellite, the latter to provide atmospheric and land surface data needed for atmospheric corrections and accurate SIF characterizations. FLORIS will measure the radiance between 500 and 780 nm with a bandwidth between 0.3 nm and 2 nm (depending on wavelength), providing images with a 150 km swath and 300 m pixel. Such finely resolved spectral sampling will allow retrieval of the full broadband fluorescence emission spectrum and related products such as F_total (i.e., integral of the fluorescence broadband spectrum).

Apart from progress in imaging spectroscopy technology, progress in data processing can enable derivation of the complete SIF emission spectrum. This opens up opportunities to fully exploit the use of the SIF spectrum as an indicator of photosynthetic activity. Simulations performed with the coupled fluorescence–photosynthesis model SCOPE were used to determine how strongly canopy-leaving SIF can be related to net photosynthesis of the canopy (NPC) for various canopy configurations.

Regression analysis between SIF retrievals and NPC values produced the following general findings: (1) individual SIF bands that were most sensitive to NPC were located around the first emission peak (SIF_red) for heterogeneous canopy configurations (i.e. varying  biochemistry, leaf, canopy variables); (2) using two SIF retrieval bands, e.g. O₂-B at 687 nm and O₂-A at 760 nm, or the red and NIR emission peaks at 685 nm and 740 nm, led to stronger correlations than using only one band; (3) using the O₂-B and the O₂-A SIF retrieval bands was at least as effective as using the two emission peaks and; (4) superior correlations were achieved by using the four main SIF retrieval bands (Hα, O₂-B, water vapor, O₂-A); (5) further improvements may be obtained by exploiting the full SIF profile and by using an adaptive, nonlinear regression algorithm such as Gaussian processes regression (GPR).

Relationships obtained can be due to variation in photosynthetic capacity (V_cmo), but also from variation in leaf optical and canopy structural variables such as chlorophyll content and leaf area index. Overall, modeling results suggest that sampling the SIF profile in at least both O₂-B and O₂-A bands enables quantification photosynthetic activity of vegetation with high accuracy.

Urbanization happening in many countries on the globe has a significant impact on the environment and thus should be studied thoroughly in order to analyze and monitor the effects. SAR data has great potential within the field of urban analysis and monitoring in part for its high availability even in areas of high cloud coverage or pollution where optical data is not always readily available, in part for its particular scattering mechanisms in urban environments full of manmade structures. The objective of this research is to evaluate multitemporal Sentinel-1A SAR data for the extraction of urban areas.  This study is part of the EO4Urban project funded by the ESA DUE INNOVATOR III program.

The Urban extraction is based on an algorithm called KTH-Pavia Urban Extractor developed jointly by KTH Royal Institute of Technology in Stockholm, Sweden and University of Pavia in Pavia, Italy. It is based on spatial indices including the Moran index, the Geary index and the Getis-Ord index and grey level co-occurrence matrix (GLCM) variance and correlation textures representing particular patterns in builtup structures, which are then utilized to extract urban areas. In this research, multitemporal SAR data from both ascending and descending orbits in dual polarization were selected during the vegetation season of 2015. The combination of data from ascending and descending orbits in dual polarization has already been proven useful with ENVISAT ASAR data.  For Sentinel-1A data, the first results have shown a clear benefit from combinations of these images as well. Now we have better multitemporal dual orbit and dual polarization data available with respect to seasonality and temporal coverage, this would be the focus of this round of investigation.  In the first phase of this project, three main study areas were selected, Stockholm in Europe, Beijing in Asia and Lagos in Africa. The three areas represent different styles of builtup environment, climate and culture, and hence show very different behavior in their corresponding urbanization patterns and pose different challenges on the accurate extraction of urban areas.

The preliminary urban extraction results showed that urban areas and small towns could be well extracted using a single-date Sentinel-1 SAR data with the KTH-Pavia urban extractor. The urban extraction results are being further improved using multitemporal dual-orbit dual-polarization Sentinel-1A SAR data. The initial urban extraction results are promising.  Rigorous accuracy assessments will be performed and the final results will be reported at the Living Planet symposium.

The Crop Watch for South Africa (CW4SA) project is carried out in partnership between Airbus Defence and Space (Airbus DS) and the South Africa National Space Agency (SANSA). The project is part-funded by the UK Space Agency under their International Partnership Space Programme (IPSP), which aims to promote societal or economic benefits from the use of UK satellite or space technology for countries with developing space capabilities.

Low investment incentives in agriculture, uncertainties within the sector, limited water availability and associated extreme weather events such as droughts are major contributors to food insecurity in Southern Africa. Furthermore, farmers in this region are disadvantaged due to constrained access to near-real-time information on prevailing field crop condition. Extension services are challenged to meet the information demands of farmers. Financial institutions are reticent to grant production loans to most farmers partially because of the inability of farmers to readily access risk and other management information.

This work proposes to implement a Crop Watch for South Africa (CW4SA) using satellite data. The main goal is to demonstrate a set of crop stress assessment tools that use satellite data and agronomical information to optimize the monitoring of field crop areas in both irrigated and dry-land production systems in South Africa. In order to achieve this CW4SA focuses on the utilisation of a time series of newly acquired high resolution DMC-2 satellite imagery. New acquisitions are planned every five days between September 2015 and May 2016 in two research areas in the Free State. Additionally, Landsat-8 and SPOT 6 & 7 imagery is used to provide further time steps as well as enhanced spectral information for agricultural crop monitoring. From the satellite imagery a number of biophysical parameters are calculated using Airbus proprietary Overland software. Overland accepts data from a wide range of optical sensors to calculate biophysical parameters including Fraction of Green Cover (FCover, analogous to NDVI), Fraction of Brown Cover (FBrown), Leaf Area Index (LAI), Fraction of Absorbed Photosynthetically Active Radiation (FAPAR), Chlorophyll (Chl), Canopy Shade Factor (CSF) and Fraction of water (FWater). Subsequently, the biophysical parameters are implemented in SANSA-developed crop condition and health models which are used to provide an early warning for crop stress in the two areas of interest.

The CW4SA project also investigates the feasibility for implementing the SANSA-developed crop monitoring workflow(s) on the Airbus DS industrial processing platform (Airbus Processing Cloud) to assess the potential for integration of the developed prototype system in order to move towards a pre-operational, wide area service capability. A major objective for the project is to establish a pre-operational processing infrastructure applicable at regional or national scales. To support a cost effective service model for the wide area monitoring ambitions, the processing chain will be geared towards future exploitation of Sentinel-2 data.

Due to the extensive urbanization in the Po Plain in northern Italy, rivers need to be managed to alleviate flooding problems while maintaining an appropriate aquifer recharge under an increasing percentage of impermeable surfaces.
During the PO PLain Experiment field campaign in July 2015 (POPLEX 2015), both active and under-construction infiltration basins have been surveyed and analyzed to identify appropriate satellite observations that can be integrated to ground based monitoring techniques. A key strategy is to have continuous data time series on water presence and level within the basin, for which ground based monitoring can be costly and difficult to be obtained consistently. One of the major and old infiltration basin in the central Po Plain has been considered as pilot area. The basin is active from 2003 with ground based monitoring available since 2009 and supporting the development of a calibrated unsaturated-saturated two-dimensional numerical model simulating the infiltration dynamics through the basin.
A procedure to use satellite data to detect surface water change is under development based on different sources of satellite data including radar backscatter data with an appropriate incidence angle and polarization combination. An advantage of satellite radar is that it can observe surface water regardless of cloud cover, which can be persistent during rainy seasons. Then, the surface water change is correlated to the reservoir water stage to determine water storage in the basin together with integrated ground data and to give quantitative estimates of variations in the local water cycle.
Results clearly show when the basin becomes progressively clogged by biofilms that can reduce the infiltration capacity of the basin by as much as 50 times compared to when it properly works under clean conditions. Data derived from satellite image processing can be also used to constrain the boundary condition of numerical modeling related to the hydraulic head changes on the basin and the amount of basin flooded area. Moreover the flooded area data support statistical analyses with monitored precipitation to develop time related cross correlation between the two parameters that should allow to determine the rate of clogging development on the infiltration basin surface. The quantitative estimate of the evolution of the infiltration rate, allows to obtain useful insights about the general recharge behavior of basins that can be used for informed design and maintenance.

This paper describes a study in which a new environmental indicator, called Annual Vegetation Stress (AVS), has been derived from a 12-year time series of Earth observation (EO) satellite data, and applied to 108 developing countries worldwide, in order to map the area affected annually by vegetation stress due to climate-related environmental problems, in particular precipitation anomalies and forest loss. The AVS indicator is derived from the global, multi-annual database of satellite-measured values of Fraction of Absorbed Photosynthetically Active Radiation (fAPAR) that has been developed by the Joint Reseach Centre (JRC) of the European Commission. fAPAR is a bio-geophysical variable that measures the fraction of the incident solar radiation that is absorbed by land vegetation for photosynthesis activity. High fAPAR values are associated with favorable conditions for vegetation growth, while low FAPAR values are associated with stress and disturbance. For the purposes of this study, for each year in the period 2003-2014, the 12 monthly global fAPAR maps from the JRC’s fAPAR databse have been averaged into a single annual fAPAR map, which is then used to compute the annual fAPAR anomalies, calculated as the difference between the annual fAPAR values and the average fAPAR values for the entire time series (i.e. January 2003 to December 2014). For any given year, therefore, an image pixel with a negative annual fAPAR anomaly (implying an annual fAPAR value lower than the pixel’s long-term mean value), indicates conditions of relative vegetation stress during that year. Conversely, an image pixel with a positive annual anomaly (implying an annual fAPAR value higher than the pixel’s long-term mean value), indicates relatively favourable vegetation growth conditions during that year. In order to enable a region-by-region analysis and comparison of climate-related vegetation stress, for each year in the time series (i.e. 2003-2014) the annual fAPAR anomaly data are aggregated by computing, for each of the 108 selected countries of interest, the percentage of fAPAR pixels with negative anomalies (i.e. vegetation stress) relative to the total vegetation area. This computed value is referred to as the area of Annual Vegetation Stress (AVS). A detailed analysis of six countries (i.e. Nicaragua, Equatorial Guinea, Samoa, Timor Leste, Cambodia, and India), shows good correspondence between high values of AVS and the occurrence of climatic extremes (droughts) and anthropogenic disturbance (deforestation). The results for Equatorial Guinea, for example, suggest that the recent trend of large-scale droughts and negative rainfall anomalies relative to the climatological normal values, which have been observed in Central and Western Africa, contribute to increased vegetation stress in the region’s tropical rain forests. In other regions (e.g. Timor Leste), there is evidence of a “biological lag” effect, whereby the main impacts of drought on the country’s tropical dry broadleaf forests, are delayed until the year following the climate event. In summary, the Annual Vegetation Stress indicator is shown to be an effective tool for quantifying the negative impacts on land vegetation of climate-related natural hazards, and for assessing and monitoring the vulnerability to climate change of major land use sectors, such as agriculture and forestry, for any geographic regions of interest.