Overcoming single model bias through GRACE and multi model data reveals Iran water storage depletion drivers

Rapid changes in terrestrial water storage (TWS) pose serious threats to water security in arid and semi-arid regions such as Iran. However, the inherent uncertainties of individual hydrological models hinder robust assessments of the respective impacts of natural variability and anthropogenic influence on water storage dynamics in these areas. To address this issue, our study integrates GRACE/GRACE-FO satellite gravity data, five mainstream hydrological models (GLDAS-Noah, GLDAS-VIC, GLDAS-CLSM, ERA5, and WGHM), and GPM global precipitation data. Four observational datasets related to precipitation, runoff, and evapotranspiration were derived, and 64 different hydrological model combinations were constructed. These combinations were comprehensively evaluated against Mascon products as a benchmark. Ultimately, the model combination with the best fitting performance was selected for spatial and temporal variation analysis and attribution analysis. The findings reveal that: (1) The model combination constructed using ERA5-derived evapotranspiration and runoff data, combined with precipitation data from VIC/CLSM, exhibits the highest consistency with Mascon data. (2) In densely populated northern and southwestern regions, the natural water flux shows significant upward trends. Nevertheless, TWSA declines there because anthropogenic extraction (captured in the net residual) outweighs the natural increase, causing groundwater discharge to surface systems and amplifying evapotranspiration and runoff losses. (3) In sparsely populated arid central regions, TWSA remains relatively stable, with an average annual natural water anomaly change rate of approximately + 0.01 cm/yr, primarily due to the offsetting effects of precipitation and evapotranspiration. This study systematically evaluates the applicability of a multi-model approach in arid regions. The integrated hydrological modeling framework developed herein provides a methodology for selecting model configurations and quantifying associated uncertainties in water storage assessment, thereby offering a scientific basis for sustainable water resource management in arid environments.