A spatiotemporal reconstruction framework for GRACE/GRACE-FO-derived groundwater storage anomalies based on historical phase-constrained gap filling and physics-guided global–local downscaling

Observations from the Gravity Recovery and Climate Experiment (GRACE) and GRACE Follow-On (GRACE-FO) provide essential information on terrestrial water storage anomalies (TWSA), from which groundwater storage anomalies (GWSA) can be derived. However, the approximately 11-month mission gap and the coarse spatial resolution of the original products constrain their application in regional groundwater monitoring and water-resource assessment. This study developed an integrated spatiotemporal reconstruction framework for the Three-North Region of China by combining historical phase-constrained gap filling with physics-guided global–local downscaling. First, a Kling-Gupta Efficiency-based singular spectrum analysis method (KGC-SSA) was developed to reconstruct the mission gap without external forcing data. Second, a physics-guided global–local-residual geographically weighted extreme gradient boosting framework (GLR-GWXGB) was developed to downscale GWSA to 1 km under grid-wise coarse-scale consistency constraints. KGC-SSA yielded a mean ensemble reconstruction spread of 0.06 cm during the mission gap and performed broadly comparably to benchmark methods. Relative to the global benchmark model, GLR-GWXGB improved internal validation performance, preserved regional-scale consistency with the original GRACE-derived GWSA, and enhanced local spatial detail. Correlations with groundwater observations improved at more than 70% of monitoring sites. The reconstructed 1 km product further captured persistent groundwater-storage decline during 2002–2022, with intensified depletion after approximately 2012, while retaining the dominant long-term and seasonal characteristics of the original series. These results demonstrate that the proposed framework provides a robust approach for generating temporally continuous and spatially refined groundwater-storage information from GRACE/GRACE-FO observations.