Intact Rock Glacier Inventory of the Central Himalaya: A Deep Learning-Based GeoPackage Dataset with Topo-Climatic Attributes

This dataset provides the first high-resolution, automated inventory of intact rock glaciers across the Central Himalaya, spanning Nepal, China (Tibet Autonomous Region), and India (Uttarakhand and Himachal Pradesh). The inventory was produced by applying the DeepLabV3+ semantic segmentation deep learning model with an Xception backbone to Sentinel-2 true-colour RGB imagery (10 m resolution) acquired between October and December 2025, followed by manual post-processing and independent spatial validation. The inventory comprises 3,798 intact rock glacier polygons covering a total area of 957 ± 3 km², distributed across elevations of 3,536–6,380 m a.s.l. between 27.52°N–31.38°N and 78.10°E–89.06°E. Rock glaciers are mapped as intact landforms based on geomorphological criteria including well-defined furrow-and-ridge topography, swollen body morphology, and steep frontal slopes. No sub-classification into active or inactive categories is applied, as reliable distinction requires supplementary InSAR velocity data beyond the scope of optical mapping. Each polygon is accompanied by 13 topo-climatic attributes including planimetric area, perimeter, median elevation, slope, median and circular mean aspect, and Potential Incoming Solar Radiation (PISR) derived from the SRTM DEM at 30 m resolution. Climatic attributes include Mean Annual Air Temperature (MAAT) and Mean Temperature of the Warmest Quarter (MTOWQ) for two temporal periods: 1970–2000 (WorldClim v2.1) and 2000–2025 (MODIS LST-derived, bias-corrected). The dataset is distributed as a GeoPackage (.gpkg) file in the WGS 84 geographic coordinate system (EPSG:4326) under a CC BY 4.0 licence. A detailed README file describing the data dictionary, mapping methodology, known limitations, and usage instructions is included.