Iteration Empirical Mode Decomposition to reduce the north-south striping noise in GRACE post-processing

Abstract

Considering that the traditional EMD method cannot be used to sufficiently extract true geophysical signals during GRACE time-varying gravity field postprocessing, a novel iterative empirical mode decomposition (EMD) processing strategy was proposed. The proposed iterative EMD method fully considers the inadequacy of the traditional EMD method in extracting geophysical signals and filtering noise. In this research, a 14-yr spherical harmonic (SH) time series of GRACE CSR RL06 data truncated to degree and order 60 was analysed by the iterative EMD method, and the results were compared with those of the traditional EMD method. To gain insight into the extracted signals, we analysed them from two perspectives: the spectral domain and the spatial domain. The results showed that the correlation coefficients between the filtered SH coefficients obtained by the iterative EMD method and the original SH coefficients are lower than those obtained by the traditional EMD method; for example, the correlation coefficients of C 12,12 and C 60,60 were 0.95 and 0.88 and 0.72 and 0.58, respectively. The spatial domain results indicated that compared with the traditional EMD method, the iterative EMD method could effectively retain the signal intensity while filtering north ‒ south stripe noise. Moreover, to evaluate the noise filtering efficiency, the ratios of the latitude-weighted RMS over land and ocean were adopted, and the mean RMS ratios for all available months obtained via the iterative and traditional EMD methods were 3.54 and 3.34, respectively, representing a relative improvement of 6.17 %. Finally, to verify the accuracy of the iterative EMD method in extracting geophysical signals, two river basins were analysed via comprehensive comparison with GRACE mascon data, and the results showed that the estimated regional mean mass change series obtained via the iterative EMD method is closer to that obtained from the GRACE mascon data. Thus, it can be concluded that the iterative EMD method can be employed to effectively suppress noise and more accurately extract real geophysical signals relative to the traditional EMD method.