Dem Elevation Error Correction in a Mountainous Region of the Alps Using Mamba

Abstract

Digital elevation models (DEMs) are essential for terrain-related remote sensing applications, but their accuracy often decreases in complex mountainous terrain. This study investigates the use of the original Mamba model for DEM elevation error correction in a selected mountainous region of the Alps. The Copernicus Digital Elevation Model (COPDEM), Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) ground elevations, and ETH2020 canopy height data were integrated to construct patch-based samples for error modeling. Experimental results show that the patch-based dataset construction outperforms tabular-data-based modeling by incorporating local terrain context, indicating that neighborhood topographic information is important for DEM error correction in mountainous areas. The original Mamba model achieves the best overall correction performance among the compared methods and further improves COPDEM accuracy relative to other DEM products, demonstrating its effectiveness in modeling spatially varying elevation errors in complex mountainous terrain.