Assessing groundwater sustainability across high mountain Asia using remote sensing

Abstract Understanding long-term evolution of groundwater storage (GWS) is critical for water security in high mountain Asia (HMA), where hydrological systems are heavily dependent on glacier-fed recharges. However, the impacts of climate changes and human activities on historical and future GWS remain insufficiently quantified. Here, we integrate remote sensing, Earth system modeling, and artificial intelligence (AI) to assess historical trends and project future evolutions in GWS across HMA. Remote sensing reveals that approximately 69% of the regions experienced GWS declines during 2003–2020, with annual losses totaling −24.2 Gt, particularly in densely populated and irrigated downstream basins such as Ganges–Brahmaputra and Indus. Using a Transformer-based framework, we attribute 47% of GWS variability to direct climate drivers and an additional 15% to cryospheric processes, while human activities contribute up to 38% of declines. The identified GWS decline in downstream basins along with its intensifying pattern is further corroborated by 2554 in situ groundwater well observations. Our data-driven projections further indicate a sustained climate-induced threat to groundwater sustainability in high-risk basins through the end of the century when maintaining current water-use levels. While cryospheric melt may temporarily offset losses until the 2060s, this buffering effect will diminish over time and lead to accelerated depletion under climate warming. This study demonstrates the value of combining remote sensing with explainable AI to improve groundwater assessment and support climate-resilient water management in high mountain regions worldwide.