Machine Learning–Enhanced Downscaling of GRACE Terrestrial Water Storage Reveals Accelerating Water Depletion Across Egypt

Terrestrial Water Storage (TWS) monitoring is critical for understanding hydrological dynamics and ensuring sustainable water resource management in arid countries such as Egypt, where data scarcity and climate variability pose significant challenges. The Gravity Recovery and Climate Experiment (GRACE) data provides valuable insights into TWS variability by tracking temporal changes in Earth’s gravitational field. However, GRACE’s coarse spatial resolution (~0.5°) restricts its direct application for regional-scale hydrological assessments. This study addresses this limitation by developing a statistical downscaling framework to enhance GRACE-derived Terrestrial Water Storage Anomalies (TWSA) from 0.5° to 0.1° resolution over Egypt for the period April 2002 to October 2024. The framework integrates machine learning models, including Random Forest (RF), Boosted Regression Trees (BRT), and Extreme Gradient Boosting (XGBoost). ERA5 reanalysis climate data, comprising 13 variables, were evaluated, and the 10 most influential predictors, including soil moisture, total evaporation, and net solar radiation, were selected based on their correlation with monthly GRACE TWSA. Results demonstrate that the downscaling approach significantly improves spatial detail and captures localized TWS variations. Declining storage trends were observed in western and central Egypt, while relatively stable or increasing trends were identified in southeastern regions, including parts of Sinai. The models achieved strong predictive performance (R up to 0.91; RMSE up to 1.43), with residual kriging further enhancing spatial continuity and reducing local biases. The validated high-resolution TWSA dataset provides a valuable resource for groundwater monitoring, drought risk assessment, and sustainable water management. Future studies should integrate anthropogenic influences, such as groundwater extraction and land-use change, to further refine TWS assessments.