An Enhanced Parameter Filtering Approach for Postprocessing GRACE Monthly Gravity Field Models

An effective filtering approach is essential for accurately interpreting GRACE (Gravity Recovery and Climate Experiment) monthly gravity field models. The Improved Parameter Filtering (IPF; Zhang et al. [1]) simultaneously estimates signal components with deterministic harmonic parameters and time-variable irregular signals through Kalman Filtering (KF), followed by signal denoising based on their signal and noise covariance matrices. However, it has two critical limitations: (a) excessive computational burden due to redundant dynamic calculations of deterministic parameters in KF, and (b) signal attenuation resulting from a suboptimal two-step estimation framework. For this reason, this letter proposes an Enhanced Parameter Filtering (EPF) based on a more rigorous parameter estimation criterion, which independently resolves deterministic parameters and irregular signals, thereby avoiding dynamic estimations of deterministic parameters while effectively integrating the two-step procedures of IPF. Here, we employ EPF to denoise the ITSG-Grace2018 model at degree 96 from April 2002 to December 2023, comparing its performance with IPF. Results demonstrate that the computational efficiency of EPF is improved by 92.3%, with fitting errors reduced by 62.4% and signal-to-noise ratios enhanced by 4.4%. Spatial analysis of filtered global Terrestrial Water Storage Anomalies (TWSAs) shows EPF better matches CSR Mascon (CSRM) RL06, JPL Mascon (JPLM) RL06, and NOAH products, with average Nash-Sutcliffe Coefficients increased by 1.7%, 2.9%, and 8.3%, respectively. Further comparisons of TWSAs across 30 global basins, mass changes in Greenland and Antarctica, and co-seismic gravity signals of the 2004 Sumatra-Andaman and 2010 Chile earthquakes, reveal the superior performance of EPF over IPF and four recently proposed filters.