We present the Global Supraglacial Debris Dataset (GSDD), developed to support the training, validation, and testing of deep learning models for supraglacial debris mapping. (test_v1.1) is refined test dataset. Baseline results using standard CNNs and a geo-foundational model are reported in our article:“Improving supraglacial debris mapping using newly annotated multisource remote sensing data and a geo-foundational model” by Kaushik Saurabh¹, Maurya Lalit, Tellman Elizabeth, and Zhang Guoqing. The dataset comprises Sentinel-2 spectral bands (Blue, Green, Red, NIR, SWIR1, SWIR2), a derived Normalized Difference Debris Index (NDDI), topographic layers (slope and elevation), and glacier velocity data. These layers help distinguish supraglacial debris (on-glacier) from proglacial debris (off-glacier), with minimal manual annotation effort. GSDD was systematically generated to assist the machine learning and geoscience communities in improving glacier mapping workflows and evaluating the applicability of newly proposed geo-foundational models for downstream cryosphere-related tasks—an area that remains underexplored. Please cite:Kaushik Saurabh¹, Maurya Lalit, Tellman Elizabeth, Zhang Guoqing. “Improving supraglacial debris mapping using newly annotated multisource remote sensing data and a geo-foundational model.”