Springer et al. (2026) A review of current best practices and future directions in assimilating GRACE/-FO terrestrial water storage data into numerical models

Identification

• Journal: Hydrology and earth system sciences
• Year: 2026
• Date: 2026-02-19
• Authors: Anne Springer, Gabriëlle De Lannoy, Matthew Rodell, Yorck Ewerdwalbesloh, Helena Gerdener, Mehdi Khaki, Bailing Li, F. Li, Maike Schumacher, Natthachet Tangdamrongsub, Mohammad J. Tourian, Wanshu Nie, Oliver Baur
• DOI: 10.5194/hess-30-985-2026

Research Groups

• Institute of Geodesy and Geoinformation, University of Bonn, Germany
• Department of Earth and Environmental Sciences, Catholic University Leuven, Belgium
• Earth Sciences Division, NASA Goddard Space Flight Center, USA
• School of Engineering, University of Newcastle, Australia
• Hydrological Sciences Laboratory, NASA Goddard Space Flight Center, USA
• ESSIC University of Maryland, USA
• Department of Sustainability and Planning, Aalborg University, Denmark
• Water Engineering and Management, Asian Institute of Technology, Thailand
• Institute of Geodesy, University of Stuttgart, Germany
• Science Applications International Corporation, USA

Short Summary

This review synthesizes insights from approximately 60 GRACE/-FO data assimilation studies to identify best practices and future directions for integrating terrestrial water storage anomaly data into numerical models, revealing that effective strategies leverage ensemble Kalman filters, localization, and explicit accounting for correlated observation errors.

Objective

• To synthesize insights from existing GRACE/-FO data assimilation studies to identify best practices and future directions for integrating terrestrial water storage anomaly (TWSA) data into numerical models.
• To systematically analyze common settings within current GRACE/-FO data assimilation (DA) frameworks and evaluate the present lack of consensus regarding DA strategies.
• To outline directions that may support convergence in DA methodologies and open up perspectives on new research avenues, including low-latency products, multi-sensor integration, and machine learning.

Study Configuration

• Spatial Scale: Review covers studies ranging from sub-regional to global scales. GRACE/-FO observations typically have a coarse spatial resolution of approximately 300 kilometers, while hydrological and land surface models operate at finer resolutions, often from a few kilometers to 0.25-4 degrees.
• Temporal Scale: GRACE/-FO observations are primarily monthly, with some studies exploring sub-monthly (weekly or daily) products. Models typically simulate water cycle components at daily or sub-daily time steps, with reanalyses covering multi-decadal periods.

Methodology and Data

• Models used:
  • Global Hydrological Models (GHMs): WaterGAP Global Hydrology Model (WGHM), PCR-GLOBWB, W3RA, MGB, LISFLOOD, HBV-SIMREG, SWBM, WBM.
  • Land Surface Models (LSMs): Catchment Land Surface Model (CLSM), Community Land Model (CLM3.5, CLM4, CLM5), Noah, Noah-MP, ORCHIDEE, CABLE, HTESSEL, SURFEX-TRIP, VIC, AWRA-L.
  • Hybrid/Coupled Models: MESH, ParFlow-CLM.
  • Data Assimilation Algorithms: Ensemble Kalman Filter (EnKF), Ensemble Kalman Smoother (EnKS), Local Ensemble Transform Kalman Filter (LETKF), Square Root Analysis (SQRA), Kalman Takens, Ensemble Square Root Filter (EnSRF), Unsupervised Weak Constrained Ensemble Kalman Filter (UWCenKF), Adaptive Ensemble Kalman Filter (AEnKF), Ensemble Adjustment Kalman Filter (EAKF), Support Vector Machine (SVM), Localized Ensemble Kalman Filter (LEnKF), Particle Filters.
• Data sources:
  • Satellite Gravimetry: GRACE and GRACE Follow-On (GRACE/-FO) terrestrial water storage anomaly (TWSA) data (Level-1b along-orbit line-of-sight gravity differences, Level-2 spherical harmonics, Level-3 mascon solutions).
  • Other Satellite Observations (for multi-sensor DA): SMOS, SMAP, Leaf Area Index, water surface elevation, flood extent.
  • In-situ Observations: Groundwater wells, streamflow gauging stations, snow depth, soil moisture.
  • Reanalysis Data: Global atmospheric reanalyses (for meteorological forcing).
  • Geophysical Corrections: Glacial Isostatic Adjustment (GIA) models, earthquake models, Atmosphere and Ocean De-Aliasing (AOD1B) background model.
  • Validation Data: Global Navigation Satellite System (GNSS) vertical land motion, various drought/flood indices, evapotranspiration products.

Main Results

• GRACE/-FO data assimilation significantly improves water cycle reanalyses by uniquely constraining total water storage variability across terrestrial compartments, leading to better representation of trends in hydrological variables, climate-driven changes, and anthropogenic influences like irrigation-induced groundwater depletion.
• The most effective assimilation strategies involve robust modifications of the classical Ensemble Kalman Filter (EnKF) and localization techniques, explicitly accounting for correlated observation errors, and addressing biases in both observations and model perturbations.
• Unmodeled processes (e.g., reservoirs, glaciers, human water use) must be carefully handled through signal separation, multi-source assimilation, or removal prior to assimilation to ensure realistic mass change distribution.
• GRACE/-FO DA has demonstrated improvements in groundwater, soil moisture, and snow estimates, as well as in drought and flood monitoring, but its impact on streamflow and evapotranspiration can be mixed, especially in regions with significant human water management not explicitly represented in models.
• Key challenges include resolving spatial and temporal resolution mismatches between GRACE/-FO data and high-resolution models, handling non-Gaussian behaviors of terrestrial water storage components, and accurately quantifying GRACE/-FO observation errors (especially spatial correlations and aliasing errors).
• Future directions emphasize developing low-latency products for near-real-time assimilation, integrating enhanced and combined satellite observations (e.g., from upcoming Next Generation Gravity Missions like GRACE-C and NGGM), and employing machine learning approaches for downscaling and hybrid assimilation.

Contributions

• Provides a comprehensive review and synthesis of approximately 60 GRACE/-FO data assimilation studies, detailing their methodologies, applications, strengths, and limitations.
• Systematically analyzes common settings and methodological choices within current GRACE/-FO DA frameworks, highlighting the diversity of approaches and the lack of community consensus.
• Identifies and discusses critical challenges unique to GRACE/-FO DA, such as unmodeled terrestrial water storage processes, issues in innovation computation at different resolutions, and non-Gaussian behaviors of hydrological variables.
• Offers best practice recommendations for GRACE/-FO DA, covering model and process description, GRACE/-FO post-processing, and DA strategy (algorithms and tuning).
• Outlines future research directions, including the development of low-latency TWSA products, direct assimilation of line-of-sight gravity measurements, and the integration of machine learning into DA systems.
• Proposes concrete steps for the community to foster standardization, comparability, and reproducibility of GRACE/-FO DA experiments, such as defining benchmark experiments, agreeing on core performance metrics, and establishing intercomparison initiatives.

Funding

• Deutsche Forschungsgemeinschaft (DFG) within the CRC DETECT (grant no. DFG – SFB 1502/1-2022 – Projektnummer: 450058266)
• Deutsche Forschungsgemeinschaft (DFG) within the research unit GlobalCDA (KU1207/26-1)
• European Space Agency (ESA) for the "NGGM and MAGIC Science and Applications Impact Study" (Contract No. 4000145265/24/NL/SC)
• Belspo EO4Peat (grant no. SR/00/414)
• NASA’s GRACE-FO Science Team
• VILLUM FONDEN (research grant number VIL60779)

Citation

@article{Springer2026review,
  author = {Springer, Anne and Lannoy, Gabriëlle De and Rodell, Matthew and Ewerdwalbesloh, Yorck and Gerdener, Helena and Khaki, Mehdi and Li, Bailing and Li, F. and Schumacher, Maike and Tangdamrongsub, Natthachet and Tourian, Mohammad J. and Nie, Wanshu and Baur, Oliver},
  title = {A review of current best practices and future directions in assimilating GRACE/-FO terrestrial water storage data into numerical models},
  journal = {Hydrology and earth system sciences},
  year = {2026},
  doi = {10.5194/hess-30-985-2026},
  url = {https://doi.org/10.5194/hess-30-985-2026}
}