A global dataset of reservoir in-situ water levels for hydrological and remote sensing applications

Reservoirs are fundamental to global water security, and their monitoring increasingly relies on satellite remote sensing. However, the absence of a unified, large-scale in-situ dataset has hindered robust validation and inter-comparison of remote sensing algorithms. To address this gap, we developed the Global Reservoir Observed Water Levels (GROWL) dataset by systematically compiling publicly available in-situ water level and storage time series from global, national, and regional sources. All records were processed through a harmonized workflow involving unit standardization and multi-stage quality control. To enhance spatial coverage, altimeter-based water level data were incorporated and cross-validated against in-situ observations. The final GROWL dataset comprises 4,134 long-term time series, including 3,154 in-situ station records, 973 satellite altimetry-derived records, and 7 literature-based records. Among these, 77% are provided at daily temporal resolution and 23% at monthly resolution. Most reservoirs have record lengths of 5-40 years (with a mean of 28 years). GROWL provides a crucial benchmark for calibrating and validating satellite-based reservoir monitoring algorithms and hydrological models, as well as for supporting reservoir-related deep learning algorithms, thereby fostering reproducible science and accelerating progress in the Earth observation and water resources communities.