Why Does a Shallow Land Model Underestimate the Dryland Drying Trends?

Abstract

Abstract Most land surface models (LSMs) and macroscale hydrological models with shallow soil profiles underestimate observed terrestrial water storage (TWS) declines across global drylands. Here, we incorporated a recently developed global data set of depth to bedrock with spatially varying deeper soils into Noah‐MP with advanced soil hydrology and plant hydraulics. Results show that the more realistic deeper soil representation better reproduces both the magnitude and spatial patterns of GRACE‐observed drying trends across global drylands, whereas the default shallow‐soil configuration of 2 m, representative of models commonly used in GRACE‐LSM residual methods (e.g., GRACE‐GLDAS), substantially underestimates these declines. Deeper soils enhance evapotranspiration supported by capillary rise and greater deep root water uptake of stored antecedent precipitation, particularly during dry seasons, thereby amplifying long‐term TWS declines. Simulated dryland drying trends are dominated by groundwater storage changes rather than soil moisture changes. These findings suggest that the commonly used approach of estimating groundwater storage anomalies as the residual between GRACE TWSA and modeled TWSA from shallow LSMs may lead to biased or misleading inferences in GRACE‐based groundwater assessments.