Digital Elevation Model Super-Resolution Incorporating Optical Image2Gradient Information: A Case Study of the Catalonia Region in Spain

This dataset provides a paired low- and high-resolution DEM dataset designed for super-resolution reconstruction tasks in digital elevation modeling. Unlike traditional SR datasets that rely on simple interpolation-based downscaling, our dataset is constructed from real-world DEM sources to ensure realistic low-to-high resolution mappings.The high-resolution DEM data is sourced from the Institut Cartogràfic i Geològic de Catalunya (ICGC) with a spatial resolution of 2 meters, covering the Catalonia region in Spain. The low-resolution counterpart is derived from ALOS PALSAR DEM at 12.5 meters resolution, which offers extensive land coverage but lower spatial detail. Additionally, paired RGB optical remote sensing images from ICGC's aerial orthophotos (0.5m resolution) are included to facilitate multi-modal learning approaches.This dataset contains approximately 7,000 training pairs and around 600 test pairs, providing a robust benchmark for DEM super-resolution, deep learning-based terrain modeling, and geospatial data fusion research.