Zhang et al. (2026) Scale-dependent model-observation inconsistencies in global terrestrial water storage models

Identification

• Journal: Communications Earth & Environment
• Year: 2026
• Date: 2026-02-23
• Authors: Gangqiang Zhang, Tongren Xu, Shaomin Liu, L. Zhang, Sayed Mohammadreza Bateni, Jianzhi Dong, Wenjie Yin, Zhijie Li
• DOI: 10.1038/s43247-026-03327-z

Research Groups

• State Key Laboratory of Earth Surface Processes and Disaster Risk Reduction, Faculty of Geographical Science, Beijing Normal University, Beijing, China
• State Key Laboratory of Remote Sensing and Digital Earth, Faculty of Geographical Science, Beijing Normal University, Beijing, China
• Department of Civil, Environmental and Construction Engineering, and Water Resources Research Center, University of Hawaii at Manoa, Honolulu, HI, USA
• UNESCO-UNISA Africa Chair in Nanoscience and Nanotechnology, College of Graduate Studies, University of South Africa, Pretoria, South Africa
• Institute of Surface-Earth System Science, School of Earth System Science, Tianjin University, Tianjin, China
• Satellite Application Center for Ecology and Environment, Ministry of Ecology and Environment, Beijing, China
• Jiangsu Province Surveying & Mapping Engineering Institute, Nanjing, China

Short Summary

This study evaluates scale-dependent model-observation inconsistencies in seven global terrestrial water storage models against GRACE observations across global, climate-zone, and basin scales, finding that satellite observation-constrained assimilation significantly reduces these inconsistencies compared to model-driven products.

Objective

• To systematically assess scale-dependent model-observation inconsistencies in seven global terrestrial water storage models against GRACE observations across global, climate-zone, and multiple basin scales.
• To evaluate the effectiveness of data assimilation techniques in improving water storage simulations and capturing terrestrial water storage anomaly (TWSA) responses to major climate drivers like ENSO and precipitation variability across different spatial scales.

Study Configuration

• Spatial Scale: Global, five Köppen climate zones (Tropical, Arid, Temperate, Cold, Polar), and 310 global river basins (categorized into Large, Medium, and Small basins, each with an area exceeding 50,000 km²). GRACE data downscaled to 0.05° resolution. ERA5-Land at 0.1° resolution.
• Temporal Scale:
  • Core global and climate-zone analyses: January 2004–December 2014.
  • Basin-scale analyses: February 2003–December 2019.
  • Global seasonal variations: 2002–2022.
  • Temporal resolution: Monthly.

Methodology and Data

• Models used:
  • Land Surface Models (LSMs): Noah v2.1, VIC v2.1, CLM 5.0
  • Global Hydrological Models (GHMs): WGHM (WaterGAP Global Hydrology Model), PCR-GLOBWB (PCRaster Global Water Balance)
  • Reanalysis Product: ERA5-Land
  • Data Assimilation System: CLSM v2.2 (from GLDAS, assimilates GRACE CSR product)
• Data sources:
  • Satellite observations: GRACE and GRACE-FO missions (CSR RL06, JPL RL06, GFZ RL06, GSFC RL06 mascon products).
  • Merged/Downscaled GRACE product: HWSA v1.0 (0.05° resolution, monthly TWSA from April 2002 through December 2022).
  • Climate data: El Niño-Southern Oscillation (ENSO) indices (ENSO 3.4, ENSO 4, Multivariate ENSO Index (MEI V2), and Oceanic Niño Index (ONI)) from NOAA.
  • Precipitation data: Climate Research Unit (CRU).
  • Auxiliary data: Köppen-Geiger climate classification (KGCC) maps (1991-2020).
• Evaluation Metrics: Spearman rank correlation coefficient (CC), Root Mean Square Error (RMSE), Normalized Root Mean Square Error (NRMSE), Structural Similarity Index Measure (SSIM), Relative Entropy (RE), Normalized Relative Entropy (NRE), Mann–Kendall test for trend significance.

Main Results

• Global Temporal Variations: Hydrological models (e.g., WGHM) show the highest temporal correlation (0.94) with GRACE, while the assimilation system (CLSM v2.2) exhibits a comparatively lower correlation (0.80). All products capture the global declining trend of –1.71 mm/yr and dominant seasonal fluctuations.
• Global Spatial Patterns: The assimilation system (CLSM v2.2) demonstrates the highest spatial agreement with GRACE (CC: 0.73, SSIM: 0.75). Hydrological models (WGHM, PCR-GLOBWB) capture broad patterns but with lower spatial correlations (CCs: 0.16–0.32). Structurally incomplete models (LSMs, ERA5-Land) show distinct trend characteristics.
• Performance Across Climate Zones: CLSM v2.2 maintains robust performance across all five climate zones (CC: 0.69–0.84, NRMSE: 0.55–0.76). Physical models (LSMs, GHMs) show notable degradation in performance in arid and cold climate zones, and particularly in polar regions (e.g., PCR-GLOBWB CC: -0.40).
• Performance Over Multiple Basin Scales: CLSM v2.2 consistently performs best across all basin scales (CC: 0.68–0.78; RMSE: 3.55–3.90 mm/yr), with higher directional accuracy (80.65%) compared to GRACE. Hydrological model performance (e.g., WGHM) systematically deteriorates from large to small basins (CC: 0.29–0.69). Normalized Relative Entropy (NRE) values generally increase with decreasing basin scale, indicating reduced model accuracy in smaller basins.
• Response to Climate Events: CLSM v2.2 best reproduces the observed TWSA-ENSO correlation patterns from GRACE, especially in tropical and subtropical regions. Discrepancies between model simulations and GRACE observations are mainly concentrated in basins with strong ENSO signals, with no notable dependence on basin area.

Contributions

• Establishes a comprehensive multi-source, multi-scale evaluation framework for terrestrial water storage anomaly (TWSA) model-observation inconsistencies.
• Systematically assesses seven diverse global terrestrial water storage modeling approaches against GRACE observations across global, climate-zone, and multiple basin scales.
• Reveals scale-dependent performance patterns and identifies specific model-observation inconsistencies in physically-based models, particularly in polar regions and smaller basins.
• Quantifies the substantial reduction in inconsistencies achieved by satellite observation-constrained assimilation systems (CLSM v2.2) compared to purely model-driven products.
• Provides essential scientific guidance for targeted model improvement, optimal model selection for diverse hydrological applications, and climate adaptation/water management strategies.
• Examines how well these models capture TWSA responses to major climate drivers (ENSO and precipitation variability) across different spatial scales.

Funding

• National Natural Science Foundation of China (42571391).

Citation

@article{Zhang2026Scaledependent,
  author = {Zhang, Gangqiang and Xu, Tongren and Liu, Shaomin and Zhang, L. and Bateni, Sayed Mohammadreza and Dong, Jianzhi and Yin, Wenjie and Li, Zhijie},
  title = {Scale-dependent model-observation inconsistencies in global terrestrial water storage models},
  journal = {Communications Earth & Environment},
  year = {2026},
  doi = {10.1038/s43247-026-03327-z},
  url = {https://doi.org/10.1038/s43247-026-03327-z}
}