Estimating land surface subsidence rates in Tehran using satellite-derived topographic differencing

Depletion of non-renewable groundwater drives land surface subsidence across Iran. In south-west Tehran, Sentinel-1 InSAR estimated land subsidence rates reach ~130 mm/yr, resulting in ~1 m cumulative subsidence in eight years. Despite its application in mapping land subsidence at regional scales, Sentinel-1 InSAR deformation estimates can incur noise in vegetation rich environments with varying soil moisture. Alternatively, we test the utility of satellite-derived topographic data in estimating land subsidence magnitudes in the urban-agricultural region of south-west Tehran. We construct elevation point clouds and Digital Elevation Models (DEMs) using very high-resolution stereo-satellite imagery, 2012-2023. We remove buildings from point clouds to construct Digital Surface Models (DSM) and Digital Terrain Models (DTM) before differencing temporally separated DSM and DTM pairs. We additionally difference DEMs and satellite LiDAR returns, which yields strong correlation (R 2 =0.83±0.11) between subsidence magnitude and temporal baseline length. DTM-LiDAR subsidence magnitude estimates are consistently ~0.2-0.3 m larger than InSAR displacement estimates, which we attribute to 0.2-1 m vertical errors in Pléiades imagery geolocation, as well as horizontal and vertical inaccuracies in satellite LiDAR returns. We find the use of ICESat-2 returns as Ground Control Points during point cloud construction reduces DEM vertical errors in some instances by 3 m. Combining altimetry and stereo-optical DEMs during differencing as well as DEM construction therefore offers strong opportunities to meet gaps in the topographic differencing field, including monitoring coastal land subsidence. However, current satellite LiDAR spatial and temporal resolution, and stereo-imagery vertical errors, currently prevent consistent retrieval of less than 1 m deformation.