Comparative Analysis of Global Terrestrial Water Storage Simulations: Assessing CABLE, Noah-MP, PCR-GLOBWB, and GLDAS Performances during the GRACE and GRACE-FO Era

Abstract

Hydrology and land surface and models (HM and LSM) are essential tools for estimating global terrestrial water storage (TWS), an important component of the global water budget for assessing the accessibility and long-term variability of water supplies. With the expansion of open-source and open-data policies, the community can now perform model TWS simulation from source codes as well as directly exploit end-user hydrologic products for water resource applications. Regardless of the model effectiveness and usability, an accuracy assessment is necessary to quantify the model’s global and regional strengths, weaknesses, and reliability. This paper compares the most recent global TWS estimates from six models, namely the PCRaster Global Water Balance (PCR-GLOBWB), Noah, Noah-Multiparameterization (Noah-MP), Catchment LSM, and Variable Infiltration Capacity (VIC), and Community Atmosphere Biosphere Land Exchange (CABLE)—the latter of which is cross validated for the first time. TWS observations from the Gravity Recovery And Climate Experiment (GRACE) and GRACE Follow-On (GRACE-FO) satellite missions between 2002 and 2021 are used to validate the model. The analyses show that Noah-MP outperforms other models in terms of global average correlations and root mean square errors. PCR-GLOBWB performance is superior in irrigated regions because of the inclusion of human intervention components in the model. CABLE, a core LSM of the Australian climate model, significantly outperforms all others in Australia. CLSM performs reasonably well, but the TWS long-term trend appears to be incorrect due to an overestimated groundwater component. Noah performs similarly (but inferiorly) to Noah-MP, most likely due to model physics sharing. VIC has the least agreement with GRACE and GRACE-FO. The evaluation also sheds some light on the role of forcing data in model performance, particularly for ready-to-use products such as GLDAS, where incorporating MERRA-2 or ERA5 data into GLDAS Noah simulations may potentially improve its TWS accuracy, which has previously been overlooked due to limited modeling capacity. Despite each model’s unique strength, the ensemble mean TWS, particularly when Noah-MP and PCR-GLOBWB are included, yields better TWS estimates than an individual model result. The findings of this study could serve as a benchmark for future model development and the data published in this paper could aid in the scientific advancement and discoveries of the hydrology community.