Technical note: GRACE-compatible filtering of water storage data sets via spatial autocorrelation analysis

Abstract. Groundwater storage anomaly (GWSA) can be derived on a global scale by subtracting various water storage compartments (WSCs) such as soil moisture, snow, surface water bodies, and glaciers from terrestrial water storage anomaly (TWSA) variations based on the GRACE/GRACE-FO satellite missions. Due to the nature of data acquisition by GRACE and GRACE-FO, filtering is essential to minimize North-South-oriented striping errors, thus resulting in a spatially smoothed TWSA signal. Nowadays, specific anisotropic decorrelation filters, such as DDK or VDK (time-variable DDK) filters, are applied. For a consistent subtraction of the individual storage compartments from GRACE-based TWSA, they need to be filtered in a similar way. This study utilized WSCs from observation-based data products (glaciers, soil moisture, and snow) and the global hydrological model LISFLOOD (surface water storage) to determine a suitable filter type over the analysis period 2002–2023. Analysis revealed that the routinely used decorrelation filter, e.g. the DDK filter, introduced striping artefacts into the smoothed data and has consequently been deemed inappropriate for filtering datasets lacking GRACE-like correlated error patterns. As an alternative, an isotropic Gaussian filter was chosen for further analysis. To determine the optimal filter width, an empirical correlation function was employed. By minimizing differences between the empirical spatial correlation functions of aggregated WSCs and the spatial correlation function of GRACE-based TWSA to a minimum RMSD of 0.02, an optimal filter width of 250 km was identified. This filter width could be applied to the aggregated WSCs to achieve a spatial structure similar to GRACE-TWSA, ensuring compatibility for the subtraction of WSCs from GRACE-TWSA to isolate groundwater storage.