Quantifying Vegetation-Associated Variability in Terrestrial Water Storage Across the Loess Plateau

Large-scale afforestation has been carried out across the Loess Plateau since 1999, yet its impact on terrestrial water storage (TWS) remains uncertain. Here, we integrated GRACE-derived TWS data, leaf area index (LAI), and multiple ancillary datasets, and applied a multivariate regression framework incorporating precipitation, LAI, and temperature to quantify their conditional associations with annual TWS change (TWSC) in the substantially greening region of the Loess Plateau. The signals in the Loess Plateau observed by GRACE were separated into coal mining, soil erosion, leakage of water storage changes from North China, and local TWS using multiple datasets and their variation characteristics were analyzed. In the substantially greening area, annual TWSC was negatively associated with LAI after accounting for precipitation and temperature. Multiplying the fitted LAI coefficient by the observed LAI trend yielded an estimated trend of −2.1 ± 0.5 mm/yr in the LAI-associated component of annual TWSC during the study period. Observed runoff increased by only ~0.6 mm/yr, compared with ~1.7 mm/yr expected from the historical precipitation–runoff relationship, indicating a runoff reduction of ~1.1 ± 0.8 mm/yr relative to that expected from precipitation alone. This departure cannot be uniquely attributed to vegetation because concurrent changes in human activities were not explicitly resolved. Analyses using multiple evapotranspiration products further associated increasing LAI with an ET trend component of approximately 2.0–7.5 mm/yr across the analyzed products. Vegetation greening was associated with regional TWS changes, with enhanced evapotranspiration emerging as an important hydrological pathway. These findings underscore the need to account for non-target mass signals and vegetation–water interactions when evaluating water-cycle responses to ecological restoration.