Enhancing water science in Earth’s second lung: AI-generated centenary hydrological insights from two decades of satellite data in the Congo Basin

This contribution showcases advanced artificial intelligence applications that transform over 20 years of terrestrial water storage anomaly (TWSA) observations from the Gravity Recovery and Climate Experiment (GRACE) and its follow-on (GRACE-FO) mission into a comprehensive 100-year dataset for the Congo Basin. We develop CM-RecNet, a climate-memory hybrid model, to reconstruct the basin’s TWSA for the period 1923–2024. CM-RecNet combines two RecNet deep learning models—one capturing climate-driven TWSA and another capturing memory effects—fused via a multilayer perceptron. The model achieves strong performance, with a correlation coefficient (CC), Nash–Sutcliffe Efficiency, and normalized root mean square error of 0.82, 0.70, and 0.20 during the testing period, respectively. Our reconstruction aligns well with observed runoff (CC>0.6 at most stations), Normalized Difference Vegetation Index (CC = 0.71), and water balance budget (CC = 0.69). In addition to its consistency with existing reconstructions, CM-RecNet exhibits a heightened capacity to capture the basin’s climate variability. This innovative approach enables access to previously unavailable data within the Congo Basin, necessary for understanding its critical water challenges associated with climate change and anthropogenic activities.