ALBERT

Description: The C-band synthetic aperture radar (SAR) satellites of Sentinel-1 with capability to obtain acquisition in Terrain Observation with Progressive Scan (TOPS) mode have brought new opportunities for large-scale monitoring of the ground surface deformation using interferometric SAR (InSAR) technique. However, despite the potential to generate large-scale interferograms, the highly spatiotemporal variability in troposphere, limits InSAR measurements accuracy. In addition, the measurement accuracy might be degraded by the signals due to the ionosphere, which is not negligible even at C-band data. One way for the atmospheric delay correction in interferograms is using external sources such as the global navigation satellite system (GNSS). The total electron content (TEC) and the zenith total delay (ZTD) values measured from a dense network of the GNSS receivers can be used to retrieve the ionospheric and tropospheric contributions to the interferometric phase, respectively. However, interpolation of the data is a big challenge, as we need to find a suitable function to predict the delay for the whole interferogram, which is challenging for large-scale Sentine-1 interferograms. In this study, we propose a new technique based on machine learning (ML) regression approach using the combination of small-baseline interferograms and the GNSS derived TEC and ZTD values to mitigate the atmospheric contributions. The technique produces the differential atmospheric (using the TEC and ZTD values) map for short-interval intergerograms based on the phase-atmosphere relation by this assumption that the deformation contribution to the interferometric phase is negligible in the short intervals. It then estimates the differential atmospheric maps for the longer-interval interferograms using the atmospheric maps with short intervals. The technique facilitates the corrections, as we do not need to deal with finding a suitable function for interpolation of distributed external observations. We implement our method on 12 concatenated frames of Sentinel-1 images acquired between May-October 2016 along a track over Norway to correct the interferograms from the atmospheric effects. Then, we apply the small baseline subset (SBAS) approach on the atmospherically corrected interferograms. The results on the stack of large-scale Sentinel-1 interferograms show that the ML-based method largely removes the ionospheric and tropospheric effects and thus improves the InSAR time series analysis results. To validate the results we compare the displacement time-series derived by small-baseline interferograms corrected by our method with the displacement time-series observed by GNSS receivers.