High-resolution monitoring of global reservoir water levels is critical for hydrological modeling, climate adaptation, and integrated water resource management. However, existing global observation systems exhibit significant spatiotemporal gaps, particularly for small-to-medium reservoirs (<10 km2) in data-scarce regions. To bridge this gap, we present the Global Reservoir Level Inventory (GRLI), a comprehensive dataset of instantaneous water levels derived from ICESat-2 ATL08 photon-counting LiDAR data (2018–2025). By developing a scalable computational framework that integrates a unified reservoir boundary map with an optimized spatial indexing algorithm (coupling R-tree indexing with convex hull screening), we efficiently processed approximately 8 billion laser photons across 334,353 orbits. This approach enabled the retrieval of water levels for 21,842 reservoirs globally, significantly extending monitoring capabilities beyond current radar altimetry missions. Validation against in situ measurements from 158 reservoirs demonstrates high vertical accuracy, with an average Root Mean Square Error (RMSE) of 0.54 m. Furthermore, trend analysis based on a temporally continuous subset of reservoirs reveals that a majority of systems with statistically significant signals exhibit declining water levels, with smaller reservoirs showing greater sensitivity to hydro-climatic variability and anthropogenic pressures. The GRLI serves as a strictly validated, open-access resource (available at http://back.reservoirwatch.cn:4828/reservoir/points/reservoirScreen and https://doi.org/10.5281/zenodo.17567416) to support global hydrological studies and the monitoring of UN Sustainable Development Goal 6.5.
Ji et al. (Mon,) studied this question.