Terrestrial water storage changes over 25 global river basins extracted by local mean decomposition from GRACE Monthly Gravity Field solutions

Abstract

The strong striping and high-frequency noise existed in Gravity Recovery and Climate Experiment (GRACE) solutions drowned the real geophysical signals, which need other signal extraction methods. Considering the advantages of local mean decomposition (LMD) in extracting geophysical signals from noisy time series, we adopt it to filter the noise and estimate the terrestrial water storage (TWS) changes over 25 global main river basins from the time series of 14-year (2002.04~2016.08) Release 06 (RL06) monthly gravity field models provided by Center for Space Research (CSR), together with the empirical mode decomposition (EMD) as a comparison. To evaluate the efficiency of eliminating noise by LMD and EMD, the ratios of the latitude weighted RMS over the land and ocean signals are adopted. The results show that all RMS ratios of land relative to ocean signals derived by LMD are higher than EMD with the mean values 3.4458 and 3.3302, respectively. Moreover, relative to the Global Land Data Assimilation System (GLDAS) Noah model, the extracted TWS changes by LMD approach have smaller root mean squared errors than EMD over 25 global river basins. Therefore, it is reasonable to conclude that LMD approach outperforms EMD in extracting TWS changes and filtering out the strong noise existed in GRACE monthly gravity field solutions.