Spatiotemporal monitoring of water storage in the North China Plain from 2002 to 2022 based on an improved GRACE downscaling method

North China Plain (NCP), a major agricultural region in China. The coupling effects of key drivers on water storage dynamics were quantitatively analyzed, integrating frequency-domain correlation analysis to identify lag effects, which were incorporated into a Random Forest (RF) downscaling method. GRACE data were refined through this approach, enhancing spatial resolution while maintaining accuracy, with the aim of precisely characterizing water storage dynamics and examining its interactions with climatic and anthropogenic factors, particularly the long-term impact of groundwater fluctuations on surface deformation. Terrestrial Water Storage Anomaly (TWSA), Shallow Water Storage Anomaly (SWSA), and Groundwater Storage Anomaly (GWSA) were derived for the NCP over the past two decades at a 0.25° resolution. The enhanced downscaling model demonstrated improved performance, with a higher Nash-Sutcliffe efficiency (+0.05), an increased correlation coefficient (+0.03), and a reduced root-mean-square error (-0.32 cm). From 2014–2022, interannual water storage fluctuations intensified, with divergent trends for TWSA (0.30 cm/year), SWSA (1.47 cm/year), and GWSA (-0.97 cm/year). Major influencing factors include water diversion projects, increased precipitation, and reduced societal water consumption. Surface deformation lags behind GWSA by 5–6 months, with a long-term lag of 10 months and a correlation of 0.81. These findings deepen the understanding of water storage dynamics and their impact on surface deformation in the NCP. • Quantitative analysis of key drivers to enhance downscaling interpretability. • The improved downscaling method reduces model error (RMSE) by 0.32 cm. • Water storage dynamics in the NCP over the past two decades are extensively analyzed. • Drivers-water storage dynamics-surface deformation response are fully revealed.