Evaluating Digital Elevation Model generation from Sentinel-1 SAR data in challenging tropical environments

Digital Elevation Models (DEMs) are crucial for various geospatial applications, but generating accurate DEMs in tropical regions using Sentinel-1 Synthetic Aperture Radar (SAR) data poses significant challenges due to high vegetation cover and temporal decorrelation. This study investigates the feasibility of producing reliable DEMs from Sentinel-1 data across six test sites within tropical zones. We employed a single-pair repeat-pass interferometric approach, leveraging 71 Sentinel-1 SAR pairs and incorporating global seasonal coherence maps to optimize site selection. DEM accuracy was assessed using Copernicus DEM (30 m) as a visual reference for horizontal shifts and ICESat-2 ATL08 data for vertical accuracy evaluation. Our results indicate that 28 DEMs were successfully generated across five test sites, with RMSE values as low as 22.52 m, demonstrating the potential of using Sentinel-1 data for DEM generation in tropical environments. DEM horizontal shift analysis revealed minimal displacements, with most shifts averaging less than one pixel (14 m), confirming the geometric accuracy of the produced DEMs. However, temporal decorrelation due to high vegetation and local climatic conditions remains a challenge, particularly in the most densely vegetated test site. This study proposes practical strategies, such as selecting pairs with shorter temporal baselines and using coherence maps for optimal site selection, to mitigate these challenges. The findings contribute to the advancement of DEM extraction methodologies and provide new insights for geospatial research in tropical regions.