Methods Sentinel-1

The TOPSAR acquisition technique is implemented on the Sentinel-1A IWS and EWS modes, providing medium-resolution wide-coverage images. Like any other bursted mode, the TOPSAR needs synchronization when used for repeat-pass interferometry. In order to maximize the common Doppler bandwidth and then the quality of the interferograms, the scanning sequence between different passes is synchronized with precision down to milliseconds. The same synchronization approach, applied to the forthcoming twin satellite, Sentinel-1B, will allow reducing the interferometric repeat pass delay from the current twelve days (with one satellite) to six days (two satellites).

In this paper we analyze the possibility to “trade interferometric revisit time with resolution” in TOPSAR mode.  

First, we recall that interferograms quality analyses results (e.g. coherence), carried out on stacks of Sentinel-1A data, show a quite high coherence at a repeat pass of twelve days, even on vegetated areas. Secondly, we notice that two non-synchronized interferometric TOPSAR acquisitions (that we call “Stag-TOPS” images), can be conveniently processed in order to have the total Doppler bandwidth increased, and consequently the azimuth resolution of the coherent targets enhanced. For example, in the case of a perfectly contiguous acquisition (i.e. two proximal bursts) the resolution is halved.

In May 2015, an anomaly caused few S-1A data-takes to be non-synchronized. We used these as an opportunity for testing the Stag-TOPS concept, obtaining 10-meters azimuth-resolution IWS and 20-m resolution EWS images.

In the paper, we describe the Stag-TOPS concept, the performed processing steps and the results obtained by processing the available mis-synchronized S-1A data-takes.

Terrain Observation by Progressive Scan (TOPS) is the default imaging mode for the Sentinel-1A satellite.  By electrical steering from backward to forward with a constant rotation rate, TOPS achieves wide swath coverage without usual ScanSAR problems such as scalloping and azimuth-varying signal to ambiguity ratio. However, to avoid phase ramps in azimuth direction, the resulting large Doppler centroid variation within the bursts of TOPS acquisitions requires unprecedented accuracy (better than 0.001 pixels) to coregister Single Look Complex (SLC) data.

Such high accuracy in the azimuth direction can be achieved by a combination of geometrical coregistration (using precise orbits and a Digital Elevation Model (DEM)) and spectral diversity approaches to estimate a constant azimuth misregistration using differential Interferometric phases in the overlapping regions of the bursts. However, the accuracy of the estimated azimuth misregistration depends highly on the quality of the phase and coherence of the overlapping interferograms. This dependence implies that the estimated misregistration from spectral diversity can be inaccurate for decorrelated interferograms, which can lead to undesired phase ramps in azimuth direction. This issue is particularly evident when all SLCs are coregistered to a single master or when long temporal baseline interferograms are used in PS-like approaches.

To overcome these limitations, we explore an approach that uses a network of small temporal baseline pairs of SLCs to estimate the azimuth misregistration. In this approach, we first coregister all slave SLCs to the burst overlaps of a single stack master using the precise orbits and a DEM. After resampling the overlap bursts to the master date we use spectral diversity of a network of small temporal baseline pairs and estimate the azimuth misregistration using only coherent pixels with a given threshold (usually higher than 0.85). Small temporal baseline pairs ensure high accuracy for the estimated misregistrations. Afterwards, we invert the differential misregistrations of the fully connected network of the pairs to estimate the unbiased misregistrations of all SLCs with respect to the master date. We also apply a similar network-based refinement to estimate the range misregistration, which has a less important effect than the azimuth misregistration. The obtained azimuth and range misregistrations are used to coregister the full scenes of the burst SLCs precisely to the master date. We validate this approach using ~1.5 years of Sentinel-1A IW acquisitions in western US across the San Andreas fault and evaluate several tectonic and hydrologic ground displacement signals in this region. 

The number of Sentinel-1 products suitable for Interferometry has been steadily increasing since the Sentinel-1 Science Hub opened for online access in October 2014 and with the start of routine operations mid-2015, all IW (Interferometric Wideswath mode – the standard TOPS mode over land for Sentinel-1) data acquired over land are systematically processed to SLCs. Large stacks of 20 or more images are now becoming available and the important step of stacking, where all SLCs are coregistered to a single master, will be of interest to InSAR users who exploit stack-based techniques such as PSI [1], SBAS [2] or SqueeSAR™ [3] type methods.

It is now well known that TOPS places very stringent requirements on the azimuth coregistration accuracy. While for stripmap mode, coregistration accuracies of approximately 0.1 pixels were sufficient, TOPS requires azimuth coregistration accuracies of roughly 1 milli pixel due to the large Doppler centroid ramp along bursts of several thousand Hertz [4].

Traditional approaches to coregistration include coherent and incoherent crosscorrelation [5], while for PSI, the stable scatterer based approach has provided reliable coregistration of ERS-1/2 stacks spanning a decade [6]. Although correlation techniques do reach the required 1 milli pixel accuracy, they achieve this for temporal baselines up to approximately 10 12-day Sentinel-1 repeat-pass cycles as illustrated in Figure 1. This is based on Gaussian distributed scatterers that follow an exponential temporal decorrelation model with a time constant of 35 days at C-Band [7]. An example of temporal decorrelation is shown in Figure 2 for a stack of 6 Sentinel-1 IW images over Mexico city. The coherence over the city itself remains fairly stable over temporal baselines of 1 to 5 cycles, while the coherence in rural areas decreases noticeably. Use of stable scatterers is not expected to prove possible because past experience shows that their density in rural areas is simply too low to reliably deliver 1 milli pixel accuracy, although this may be possible given large urban areas.

A significantly higher azimuth coregistration accuracy than that provided by the methods mentioned above is attained by exploiting the large spectral diversity between bursts using the ESD (Enhanced Spectral Diversity) approach suggested in [8]. While ESD performs significantly better than crosscorrelation techniques, it still relies on coherence and therefore also suffers from temporal decorrelation effects, albeit only after significantly more than 10 cycles as illustrated in Figure 1.

One way to achieve high coregistration accuracy over long time spans is to use joint techniques that exploit the shifts estimated between all or a subset of images pairs within the stack [9]. For example, one could estimate the shifts between temporally adjacent scenes and concatenate this information to coregister non-adjacent scenes.

In the paper we will investigate various approaches to achieving the high coregistration accuracy required over large stacks for TOPS mode and provide examples of their use in practice on Sentinel-1 IW stacks.

References

[1]    Ferretti, C. Prati, F. Rocca, “Permanent Scatterers in SAR Interferometry,” IEEE Trans. Geoscience and Remote Sensing, vol. 39, issue 1, pp. 8-20, 2001.

[2]    Berardino, P., Fornaro, G., Lanari, R., Sansosti, E., “A New Algorithm for Surface Deformation Monitoring Based on Small Baseline Differential SAR Interferograms,” IEEE Trans. Geoscience and Remote Sensing, vol. 40, issue 11, pp. 2375-2383, Nov. 2002.

[3]    Ferretti, A., Fumagalli, A., Novali, F., Prati, C., Rocca, F., Rucci, A., “A New Algorithm for Processing Interferometric Data-Stacks: SqueeSAR,” IEEE Trans. Geoscience and Remote Sensing, vol. 49, issue 9, pp. 3460-3470, 2011.

[4]    Scheiber, R., Moreira. A., “Coregistration of Interferometric SAR Images Using Spectral Diversity,” IEEE Trans. on Geoscience and Remote Sensing, vol. 38, issue 5. September 2000.

[5]    Bamler, R., Eineder, M., “Accuracy of Differential Shift Estimation by Correlation and Split-Bandwidth Interferometry for Wideband and Delta-k SAR Systems,” IEEE Geoscience and Remote Sensing Letters, vol. 2, issue 2, pp. 151-155, 2005.

[6]    Adam, N., Liebhart, W., Parizzi, A., Rodriguez-Gonzalez, F., Brcic, R., “Persistent Scatter Interferometry Wide Area Product Methodology and Final Characteristics,” Terrafirma Stage 3, DLR-IMF – Remote Sensing Technology Institute, 2012.

[7]    Parizzi, A., “Speckle statistics and long-term coherent SAR interferograms,” Proceedings IGARSS 2012, pp. 5590-5593, 2012.

[8]    Prats-Iraola, P., Scheiber, R., Marotti, L., Wollstadt, S. Reigber, A., “TOPS Interferometry With TerraSAR-X,” IEEE Trans. Geoscience and Remote Sensing, vol. 50, issue 8, pp. 3179-3188, August 2012.

[9]    De Zan, F. “Coherent Shift Estimation for Stacks of SAR Images,” IEEE Geoscience and Remote Sensing Letters, vol. 8, issue 6, pp. 1095-1099, November 2011.

The Sentinel-1 (S-1) mission recently completed its first year in operational mode. The default acquisition mode of S-1 over land is the innovative TOPS mode. Initially, the InSAR community was faced with many processing challenges due to the introduction of this new mode, in particular accurate and reliable coregistration. These initial challenges are by now mostly addressed by the community, and it can be freely stated that how to perform reliable basic InSAR processing of S-1 data is well understood and demonstrated within the community.

While we can partially close the TOPS mode specific issues in InSAR processing, other characteristics of the S-1 mission still provide new challenges and opportunities. First, the manifold increased spatial coverage lead to increased complexity and the need to revisit assumptions applied in further processing of basic interferograms. In particular, phase unwrapping and estimation of atmospheric phase contributions over large spatial scales are challenging. Second, the regular data acquisition plan, coupled with free and open data policy and "rapid" access to data, opens new possibilities for operational scientific and commercial applications.

Our contribution obviously boils down to aspects of InSAR time series analysis. This contribution will focus on how to accommodate and adapt current state-of-the-art for S-1 TOPS InSAR stack processing. Specifically, a basic S-1 TOPS interferometric product is really a stacks of stacks, rather than a single interferogram stack. Existing processing flows developed for stripmap InSAR had to be adapted, extended, and sometimes completely modified to this new situation. We will critically review the "conventional" processing chain from the S-1 TOPS perspective, and present our adapted approach. In addition, we will summarize and discuss remaining open questions and opportunities that we expect will emerge as more data become available, in particular with the launch of S1B.

We will motivate discussion with results obtained from specific pilot sites within the ESA SEOM InSARap project that cover a large part of the InSAR application domain - "from millimeter to decimeter".

Acknowledgment: This work was funded by European Space Agency's SEOM programme, under the contract 4000110680/14/I-BG: "SEOM – Sentinel-1 InSAR Performance Study With TOPS Data".

The proposed contribution will focus on the results of the ESA study entitled “INSARAP: Sentinel-1 InSAR Performance study with TOPS Data”. The study investigates the performance of the interferometric wide swath (IW) mode of Sentinel-1, which is implemented using the terrain observation by progressive scans (TOPS) mode [1]. The key aspects of the TOPS mode that need to be considered for accurate interferometric processing, the conclusions achieved, the cross-comparison with other sensors (ENVISAT, TerraSAR-X and COSMO-SkyMed), the validation with in situ measurements, different interferometric experiments, etc. will be shown.
The interferometric processing of TOPS data requires extremely accurate image co-registration and appropriate image resampling methods in order to handle the huge Doppler variation characteristic of this image scanning mechanism. The identified solution relies on the computation of the local geometric offsets based on the geometry, namely, by exploiting the state vectors and an external DEM [2]. The global offset is estimated by exploiting the spectral diversity in the overlap area between adjacent burst [3]. Thanks to the proposed method, no phase jumps are observable in scenes without azimuthal motion.
The evaluation of differential interferometric and persistent scatterer interferometry (PSI) results considered several aspects. On the one hand, the interferometric performance of Sentinel-1 in terms of burst synchronization and interferometric performance (e.g., NESZ, coherence over different land types, phase unwrapping) will be presented.
On the other hand, the assessment of the PSI performance for surface deformation with respect to in-situ geodetic measurements, as well as with the cross-comparison between different PSI processing chains and different input SAR data will be shown. Four different pilot sites have been considered: Campi Flegrei/Vesuvius area and Mount Etna, both located in Italy, Istanbul city in Turkey, and Mexico City. The evaluation of the results was performed using in-situ geodetic measurements based on continuous GPS stations located on the different sites.
Finally, different other interferometric results based on Sentinel-1 data will be presented, related to the 2015 Nepal earthquake, to glaciers motion estimation, etc.
REFERENCES
[1] F. De Zan and A. Monti Guarnieri, “TOPSAR: Terrain Observation by Progressive Scans,” IEEE Trans. Geosci. Remote Sens., vol. 44, no. 9, pp. 2352–2360, Sep. 2006.
[2] E. Sansosti, P. Berardino, M. Manunta, F. Serafino, and G. Fornaro, “Geometrical SAR image registration,” IEEE Trans. Geosci. Remote Sens., vol. 44, no. 10, pp. 2861–2870, Oct. 2006.
X,” IEEE Trans. Geosci. Remote Sens., vol. 50, no. 8, pp. 3179–3188, Aug. 2012.
[3] R. Scheiber and A. Moreira, “Coregistration of interferometric SAR images using spectral diversity,” Geoscience