Downscaled GRACE data reveals anthropogenic dominance in groundwater storage decline across China’s oases

Abstract

Groundwater is vital for sustaining China’s oases, yet climate change, population growth, urbanization, and oasis expansion threaten these fragile ecosystems. To address limited understanding of groundwater dynamics, we combined GRACE/GRACE-FO satellite data with machine learning to generate high-resolution (0.1°) maps of total water storage (TWS) and groundwater storage (GWS) in China’s oases. Our spatiotemporal random forest model (RFst) produced monthly TWS anomalies (TWSA) and GWS anomalies (GWSA) consistent with original GRACE data (0.5° resolution; cross-validation R 2 = 0.97, RMSE = 0.59). Downscaled data improved agreement with groundwater levels from 103 monitoring wells, increasing wells with strong correlations (>0.6) by 20 %. From 2002 to 2023, both TWS and GWS declined significantly (−5.29 mm/year and − 5.49 mm/year, respectively), with GWS accounting for 84 % of TWS loss. A two-stage generalized additive model (GAM) revealed that 78–90 % of GWS decline resulted from human activities—primarily population growth, cropland expansion, and urbanization. In contrast, oases near humid regions (e.g., Inner Mongolia, Ningxia) showed fluctuating GWS trends, where climate and human factors explained less variability (R 2 = 0.62–0.64), suggesting localized complexities. Spatial heterogeneity in groundwater depletion highlights varying pressures: arid-region oases (Xinjiang, Qinghai) faced severe declines driven by human demand, while transitional zones exhibited mixed responses. These findings underscore the need for region-specific water management strategies to address climate-human interactions. By enhancing resolution and validation, our study clarifies groundwater trends and provides a scientific foundation for sustainable oasis development in Northwest China’s arid regions.

Topics & Concepts