<p dir="ltr">We collected a total of 1,853 paired samples of 30 m Landsat-8 and 10 m Sentinel-2 images, along with corresponding 10 m DEM from the USGS 3D Elevation Program, 10 m land cover types from Dynamic World, and 10 m Sentinel-1 SAR GRD. Data preprocessing was conducted on Google Earth Engine platform, including general filtering, cloud and shadow screening, inter-sensor band adjustment, and atmospheric correction. For model training, the dataset was split into 1,377 training pairs and 476 testing pairs.</p>