To study the spatiotemporal characteristics of groundwater storage changes in the hilly areas of Jiangxi Province, based on the gravity recovery and climate experiment (GRACE)/GRACE-FO satellite inversion data from 2003 to 2023, a spatiotemporal evolution analysis framework of groundwater integrating downscaling reconstruction and trend identification was constructed, revealing the pattern of changes in groundwater storage anomalies (GWSA). The high-resolution reconstruction of GWSA was carried out through a deep learning model. The long-term trend component was extracted using seasonal-trend decomposition using LOESS (STL). The groundwater change trend and its significance were identified by combining the Mann-Kendall test and the Theil-Sen slope estimation method. The results show that the reconstructed high-resolution GWSA data clearly reveal the spatiotemporal evolution characteristics of groundwater storage in the hilly areas of Jiangxi Province; after STL, the sequence shows a significant long-term trend and stable seasonal fluctuations. From 2003 to 2023, the groundwater level in the entire province has risen slowly, and the overall long-term trend shows an evolution pattern of “recovery in northeastern and central regions but continuous depletion in southern regions”. Notably, Ganzhou City in southern Jiangxi exhibits a significant decline in groundwater storage, making it a representative groundwater-deficit zone. Groundwater variations exhibit compound characteristics, including regional disparities, phased evolution, and intensified high-frequency anomalies, indicating that the system remains unstable and is in a critical state vulnerable to further depletion.