A comprehensive water bodies dataset of high-mountain Asia

High-Mountain Asia (HMA) hosts numerous water bodies that are highly sensitive to climate change. However, many of them, especially small ones, remain understudied due to the region's complex terrain and extreme environmental conditions. In this study, we propose a multi-stage segmentation-classification framework that combines spectral-textural feature extraction with object-based refinement. This method demonstrates enhanced capabilities for detecting water bodies as small as 0.001 km², even under complex imaging conditions. Validation shows a unit-level misclassification rate of 0.1% and an 88% reduction in false positives, particularly from shadows and non-water patches, compared to existing satellite-derived datasets. Consequently, we generated a comprehensive inventory of 451,777 water bodies across the HMA, more than doubling previous records. The dataset includes morphological and hydrological attributes (e.g., area, perimeter, connectivity), with 83,246 water bodies in glacial regions (2-6 times more than previous estimates) and 274,039 in permafrost regions, 16.6% of which are newly identified small water bodies (<3 km²). These results provide a critical resource for understanding water dynamics in this climatically vulnerable region.