Sharifi et al. (2025) Technical note: GRACE-compatible filtering of water storage data sets via spatial autocorrelation analysis

Identification

• Journal: Hydrology and earth system sciences
• Year: 2025
• Date: 2025-12-03
• Authors: Ehsan Sharifi, Julian Haas, Eva Boergens, Henryk Dobslaw, Andreas Güntner
• DOI: 10.5194/hess-29-6985-2025

Research Groups

• GFZ Helmholtz Centre for Geosciences, Potsdam, Germany
• Institute of Meteorology and Climate Research-Troposphere Research (IMK-TRO), Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany
• Institute of Environmental Science and Geography, University of Potsdam, Potsdam, Germany

Short Summary

This study develops a methodology to determine an optimal spatial filtering approach for water storage compartment (WSC) datasets to ensure spatial compatibility with GRACE/GRACE-FO terrestrial water storage anomaly (TWSA) products. It identifies an isotropic Gaussian filter with a 250 km width as optimal for combined WSCs, enabling consistent subtraction from GRACE-TWSA for groundwater storage estimation.

Objective

• To identify an adequate spatial filtering approach for global water storage compartment (WSC) datasets to make them spatially consistent with terrestrial water storage anomaly (TWSA) data derived from GRACE/GRACE-FO satellite gravimetry.
• To determine the optimal filter type and width by minimizing the differences between the empirical spatial autocorrelation functions of aggregated WSCs and the spatial correlation function of GRACE-based TWSA.

Study Configuration

• Spatial Scale: Global, with data harmonized to a 0.5° spatial resolution. Spatial autocorrelation analysis was performed for distances up to 2000 km.
• Temporal Scale: Monthly data from April 2002 to September 2023. Storage anomalies were calculated relative to a long-term mean from April 2002 to December 2020.

Methodology and Data

• Models used:
  • LISFLOOD (version 4.0) for surface water storage (SWS).
  • Empirical stretched exponential model (Weibull model) for fitting autocorrelation decay.
• Data sources:
  • GRACE/GRACE-FO Terrestrial Water Storage Anomaly (TWSA): International Combination Service for Time-variable Gravity Fields (COST-G) Level-2 spherical harmonic solutions, filtered with VDK5 and VDK3. Also compared with GFZ, ITSG, and UB-ITSG products.
  • Snow Water Equivalent (SWE): Copernicus Global Land service (CGLS) daily data (0.05° spatial resolution), merged with ERA5-land model.
  • Root Zone Soil Moisture (RZSM): Copernicus Climate Change Service (C3S) satellite-based product (0.25° spatial resolution), derived from active and passive microwave products.
  • Glacier Mass (GM): C3S glacier mass product (yearly, 0.5° spatial resolution), derived from annual glacier mass change results.
  • Filtering methods tested: Isotropic Gaussian filter (with widths ranging from 50 km to 600 km in 50 km increments) and anisotropic DDK filter.
  • Analysis: Spatial autocorrelation analysis was performed on de-trended and de-seasonalized data. The Root Mean Square Differences (RMSD) metric was used to quantify the similarity between WSC and TWSA autocorrelation functions.

Main Results

• The anisotropic DDK filter, commonly used for GRACE data, was found to be unsuitable for filtering WSC datasets as it introduced spurious striping artifacts and Gibbs effects.
• An isotropic Gaussian filter was identified as an appropriate filtering approach for WSC data.
• The optimal Gaussian filter width for the combined 4WSC (Root Zone Soil Moisture, Snow Water Equivalent, Glacier Mass, and Surface Water Storage) to achieve spatial autocorrelation consistent with GRACE-based TWSA was determined to be 250 km, resulting in a minimum RMSD of 0.02.
• Unfiltered WSCs exhibited distinct spatial autocorrelation lengths: RZSM (481 km), SWE (205 km), GM (32 km), and SWS (5 km). The combined 4WSC had a correlation length of 306 km, while GRACE-based TWSA had a significantly larger correlation length of 636 km.
• Optimal Gaussian filter widths for individual WSCs varied: RZSM (150 km), SWE (200 km), GM (300 km), and SWS (350 km).
• Resampling WSC data to a common 0.5° spatial resolution prior to filtering significantly reduced computational costs (approximately 26 times faster and 14 times less memory) without affecting the scientific results.
• Filtering the combined 4WSC dataset once yielded practically identical results to filtering each WSC individually and then combining them, offering a more computationally efficient approach.

Contributions

• Presents a novel and systematic methodology for determining the optimal spatial filtering approach for water storage compartment (WSC) datasets to ensure spatial compatibility with GRACE/GRACE-FO terrestrial water storage anomaly (TWSA) products.
• Demonstrates the unsuitability of the DDK filter for WSC data and advocates for the use of an isotropic Gaussian filter, providing quantitative guidance on optimal filter widths.
• Quantifies the distinct spatial autocorrelation characteristics of various WSCs and GRACE-based TWSA, enhancing understanding of their inherent spatial scales.
• Offers practical recommendations for computationally efficient processing chains in GRACE-hydrology studies, such as pre-aggregation of WSC data and single-pass filtering of combined WSCs.
• Contributes to improving the accuracy and consistency of groundwater storage anomaly derivations from GRACE data by ensuring compatible spatial resolutions of all subtracted components.

Funding

• European Union’s Horizon 2020 research and innovation programme for G3P (Global Gravity-based Groundwater Product) under grant agreement no. 870353.
• GFZ Helmholtz Centre for Geosciences (covered article processing charges).

Citation

@article{Sharifi2025Technical,
  author = {Sharifi, Ehsan and Haas, Julian and Boergens, Eva and Dobslaw, Henryk and Güntner, Andreas},
  title = {Technical note: GRACE-compatible filtering of water storage data sets via spatial autocorrelation analysis},
  journal = {Hydrology and earth system sciences},
  year = {2025},
  doi = {10.5194/hess-29-6985-2025},
  url = {https://doi.org/10.5194/hess-29-6985-2025}
}