SAR-based Elevation Models - from Global to Local

Abstract:
The on-going TanDEM-X mission (2010-2015) has produced a global, interferometric Digital Elevation Model almost 10 years ago, which has been followed up by its successor layer WorldDEM Neo (2017-2021). Both datasets constitute a homogenous and highly accurate Digital Surface Model (DSM) with 12m respectively 5m grid spacing. The accuracy of the WorldDEM Neo layer has been recently evaluated using ICESat-2 laser altimetry data. By exploiting deep learning techniques, local updates of the DSM are now possible over urban areas, using single very high-resolution radar images from the same sensors. The resulting elevation data can be used to upgrade the global models over cities where the SARinherent DEM characteristics often show their weakness due to the side-looking image acquisition geometry.