An integrated postprocessing approach for extracting time variable signals from GRACE monthly gravity field models

Abstract

Since significant north–south striped noise exists in the GRACE (Gravity Recovery and Climate Experiment) monthly gravity field models, various filtering approaches are employed before extracting time variable signals with a harmonic fitting model that includes trends, annual and semiannual terms, leading to signal attenuation and leakage. This study develops an Integrated Postprocessing Approach (IPA) to iteratively estimate time variable signals and filter noise using a harmonic model plus a non-seasonal signal term. The stochastic characteristics of non-seasonal signals are described using an exponential function. To improve computational efficiency, the non-seasonal signals are represented by a Gauss–Markov process and estimated by Kalman filtering. The Terrestrial Water Storage Anomaly (TWSA) signals across 22 global basins from April 2002 to December 2023 estimated by IPA exhibit high spatial resolution, closely matching Combined Mascon (CM) solutions. Relative to CM solutions, IPA reduces the latitude-weighted RMSEs of spatial TWSA signals by 8.0% while increases the Nash-Sutcliffe Efficiency (NSE) of TWSA series by 2.1% compared to Improved Parameter Filtering (IPF; Zhang et al. 2024) approach. Additionally, IPA estimates significantly enhanced trends and strong annual amplitudes of TWSA signals in the Amazon and Parana basins, along with the volume variances of Three-Gorges Reservoir and Poyang Lake, as well as 2.1 and 4.2-year interannual volume variance signals in Victoria Lake. The Water Storage Drought Index (WSDI) constructed by IPA aligns well with SPEI-6 in detecting the duration and intensity of severe droughts and floods in typical basins. Further experiments indicate the IPA outperforms the other eight common spectral filters.