NATIONWIDE DEM GENERATION USING REPEAT-PASS SAR INTERFEROMETRY

Abstract

The Sentinel-1 mission was launched by the European Space Agency (ESA) in 2014. It carries a C-band (5.405 GHz; 5.5 cm) Synthetic Aperture Radar (SAR) sensor. Sentinel-1 is extensively used for earth observation as the radar data can be acquired in both day and night without being affected by cloud cover. The data is openly available throughout the globe. Repeat-pass Sentinel-1 SLC data can be used for DEM generation if the spatial (150-400 m) and temporal baselines (e.g., few days) are kept appropriate. Here, we have generated a nationwide DEM from Sentinel-1 repeat-pass SLC data. Our aim was to fully exploit Sentinel-1 SLC radar data and generate timestamp DEMs that cover the entire country. The DEM was generated using 140 SLC image pairs. The data was processed in GAMMA remote sensing software and the process was automated using bash scripting. It takes around 4 hours to generate a DEM using a pair of Sentinel-1 SLC images. In addition, we also estimated the accuracy of the generated Sentinel-1 DEMs using CORS network, ICESat-2 altimetry data, and GPS data. The RMSE value of DEM using CORS and ICESat-2 data came out to be 38.152 m and 38.117 m respectively. The RMSE of generated Sentinel-1 DEM and SRTM DEM using GPS data came out to be 11.142 m and 11.369 m respectively. We also highlighted the challenges encountered during the DEM generation process. The main challenge was to find the suitable image pairs for DEM generation. The other challenges were spatial and temporal decorrelation. The accuracy of the DEMs can be improved considering the vegetation growth in the area. For better DEM generation vegetation growth should be minimum. We can also use longer wavelength L-Band SAR on board NISAR to be launched in coming months, for better coherence and subsequent better DEM accuracy.