Hacker et al. (2026) Multidecadal reconstruction of terrestrial water storage changes by combining pre-GRACE satellite observations and climate data

Identification

• Journal: Earth system science data
• Year: 2026
• Date: 2026-03-09
• Authors: Charlotte Hacker, Benjamin D. Gutknecht, Anno Löcher, Jürgen Kusche
• DOI: 10.5194/essd-18-1747-2026

Research Groups

• Institute for Geodesy and Geoinformation, University of Bonn, Bonn, Germany

Short Summary

This study reconstructs multidecadal terrestrial water storage anomalies (TWSA) for global land from 1984 to 2020 by optimally combining pre-GRACE geodetic satellite observations (SLR and DORIS) with climate data-driven regression models, providing a long-term consistent dataset (TWSTORE) for climate change attribution and hydrological studies.

Objective

• To generate a multidecadal (1984-2020) reconstruction of GRACE-like terrestrial water storage anomalies (TWSA) for global land (excluding Greenland and Antarctica) by optimally combining pre-GRACE geodetic satellite observations (SLR and DORIS) with climate data-driven regression models, aiming for improved long-term consistency and resolution compared to existing reconstructions.

Study Configuration

• Spatial Scale: Global land (excluding Greenland and Antarctica), reconstructed on a 0.5° grid with an effective resolution of approximately 330 km, utilizing a basin-based Empirical Orthogonal Functions (EOF) approach and aggregated "super" polygons for data combination.
• Temporal Scale: Monthly resolution from 1984 to 2020.

Methodology and Data

• Models used:
  • Empirical Orthogonal Functions (EOFs) for dimensionality reduction and spatial pattern representation.
  • Autoregressive process with exogenous variables (ARX), Multiple Linear Regression (MLR), and Random Forest (RF) for climate data-driven TWSA reconstruction.
  • Variance Component Estimation (VCE) for optimal combination of geodetic and climate-driven TWSA.
  • Generalized Extreme Value (GEV) distribution for extreme event analysis.
  • Terrestrial water budget equation for deriving evapotranspiration.
• Data sources:
  • Satellite Gravity: GRACE and GRACE-FO Level-2 products (ITSG-Grace2018, COST-G RL01) and mascon products (NASA GSFC), Satellite Laser Ranging (SLR), and Doppler Orbitography by Radiopositioning Integrated on Satellite (DORIS) observations.
  • Climate Data (Predictors): Sea Surface Temperature (HadISST1, COBE-SST 2), Precipitation (GPCC, CPC Global Precipitation V1.0), Soil Moisture (CPC Soil Moisture v2, GLEAM Soil Moisture v4.1a), Air Temperature (GHCN CAMS Monthly Temperature), Leaf Area Index (Globmap LAI v3).
  • Validation/Comparison Data: Global Mean Sea Level (GMSL) products (Frederikse et al., 2020b; Horwath et al., 2022), Global Runoff Data Center (GRDC) river discharge, GLEAM Actual Evaporation v4.2a, ERA5 reanalysis (total precipitation and evaporation).
  • Ancillary Data: HydroBASINS for defining river catchments.

Main Results

• The new combined reconstruction (TWSTORE) effectively integrates information from geodetic methods, particularly enhancing long-term signals and showing significant differences from other reconstructions at multi-decadal timescales.
• The study confirms that GRACE-derived trends should not be directly extrapolated to the past, as significant changes in storage trends are evident in the pre-GRACE period.
• TWSTORE exhibits strong accelerations in terrestrial water storage, often inherited from the SLR-DORIS data, and generally indicates a decrease in TWSA across most continents in the pre-GRACE era.
• Comparison with GRACE/-FO (2002–2020) reveals lower predictive skills (Root Mean Square Deviation up to 200 mm) in Alpine/highland regions and areas with substantial anthropogenic water use, while showing high correlations (up to 0.9) in humid climates.
• Extreme value analysis indicates that one-in-five-year storage deficits/surpluses reached 20–30 cm, and one-in-ten-year events up to 60 cm (e.g., in the Amazon basin), identifying the Amazon as a global hotspot for TWSA extremes.
• A sliding-window analysis of return levels shows only minor temporal changes, with slightly higher storage deficits in the 1984–2007 period compared to 1997–2020, suggesting no clear intensification of water storage extremes at the reconstruction's scale.
• Reconstructions generally underestimate water-mass loss from glacier melting when compared to GMSL-derived land water storage including glaciers, although TWSTORE shows improved agreement due to the influence of SLR-DORIS data.
• Budget-derived evapotranspiration (ET) for 11 major river basins correlates well with GLEAM and ERA5 (median Pearson correlation coefficients of 0.73 and 0.77, respectively), with exceptions in dynamic regions like the Amazon (CC=0.1) and upper Zambezi (CC=0.5).
• Reduced ET flux was observed in the Amazon basin, potentially linked to deforestation, while increased ET rates in the Niger basin compensated for increased rainfall.

Contributions

• Presents the first multidecadal (1984-2020) reconstruction of GRACE-like terrestrial water storage anomalies (TWSA) that optimally combines low-resolution geodetic satellite observations (SLR and DORIS) with climate data-driven regression models.
• Introduces TWSTORE, a new synthesis dataset designed for enhanced long-term consistency and improved resolution by leveraging a modular approach based on GRACE-derived EOFs and variance component estimation.
• Provides publicly available reconstructed TWSA fields and associated uncertainty information (via Zenodo: https://doi.org/10.5281/zenodo.15827789).
• Offers a new dataset of multi-decadal area-averaged evapotranspiration time series for 11 major river basins, derived from the water balance equation using the TWSTORE reconstruction (via Zenodo: https://doi.org/10.5281/zenodo.16643628).
• Confirms the non-stationarity of terrestrial water storage trends, demonstrating that GRACE-derived trends cannot be simply extrapolated to the past.
• Delivers an extended observational record valuable for analyzing terrestrial water budgets, evaluating pre-GRACE modeled water storage, and validating climate model (e.g., CMIP) runs.

Funding

• Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – SFB 1502/1-2022 – Project No. 450058266
• Open Access Publication Fund of the University of Bonn

Citation

@article{Hacker2026Multidecadal,
  author = {Hacker, Charlotte and Gutknecht, Benjamin D. and Löcher, Anno and Kusche, Jürgen},
  title = {Multidecadal reconstruction of terrestrial water storage changes by combining pre-GRACE satellite observations and climate data},
  journal = {Earth system science data},
  year = {2026},
  doi = {10.5194/essd-18-1747-2026},
  url = {https://doi.org/10.5194/essd-18-1747-2026}
}