<p>Reliable groundwater monitoring is essential to ensure water security. Yet, many regions lack dense observational networks. High-resolution gridded groundwater datasets offer valuable alternatives for understanding groundwater dynamics. The Gravity Recovery and Climate Experiment (GRACE) provides large-scale terrestrial water storage anomaly (TWSA) data, but its coarse spatial resolution limits regional applicability. Recent studies have increasingly utilized machine learning (ML) to improve GRACE-based water storage estimates; however, most have relied on individual ML models, which enhance prediction accuracy only to a limited extent. Based on these grounds, this study proposes a stacking ensemble machine learning (SEML) framework that integrates Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and Categorical Boosting (CatBoost) for more precise statistical downscaling of GRACE-derived groundwater storage anomalies (GWSA) to finer spatial resolutions. Model performance was examined using different input feature combinations. The proposed framework was assessed through a case study in the Kumbukkan Oya Basin, Sri Lanka, where GRACE-derived GWSA was downscaled from 0.25° to 0.05° resolution using selected climatic and environmental predictors. The SEML model demonstrated superior predictive performance (Coefficient of Determination, R² = 0.84; Root Mean Square Error, RMSE = 3.04 cm) compared to individual models. Statistical validation against GRACE-derived GWSA yielded R² &gt; 0.9 across grid points, while in-situ groundwater level comparisons showed strong correlations (CC &gt; 0.7) in most wells. By combining multiple ML algorithms, the SEML framework significantly enhances the accuracy and reliability of GRACE-based downscaling in data-scarce regions, providing essential support for more informed sustainable water management, climate adaptation planning, and hydrological modeling approaches.</p>