Abstract :
[en] Measurements of the spatio-temporal variations of Earth’s gravity field recovered from the Gravity Recovery and Climate Experiment (GRACE) mission have led to unprecedented insights into large spatial mass redistribution at secular, seasonal, and sub-seasonal time scales. GRACE solutions from various processing centers, while adopting different processing strategies, result in rather coherent estimates. However, these solutions also exhibit random as well as systematic errors, with specific spatial and temporal patterns in the latter.
In order to dampen the noise and enhance the geophysical signals in the GRACE data, several methods have been proposed. Among these, methods based on filtering techniques require a priori assumptions regarding the spatio-temporal structure of the errors. Despite the large effort to improve the quality of GRACE data for always finer geophysical applications, removing noise remains a problematic question as discussed in Chapter 1.
In this thesis, we explore an alternative approach, using a spatio-temporal filter, namely the Multichannel Singular Spectrum Analysis (M-SSA) described in Chapter 2. M-SSA is a data-adaptive, multivariate, and non-parametric method that simultaneously exploits the spatial and temporal correlations of geophysical fields to extract common modes of variability. We perform an M-SSA simultaneously on 13 years of GRACE spherical harmonics solutions from five different processing centers. We show that the method allows for the extraction of common modes of variability between solutions, and removal of the solution-specific spatio-temporal errors arising from each processing strategies. In particular, the method filters out efficiently the spurious North-South stripes, most likely caused by aliasing of the imperfect geophysical correction models of known phenomena.
In Chapter 3, we compare our GRACE solution to other spherical harmonics solutions and to mass concentration (mascon) solutions which use a priori information on the spatio-temporal pattern of geophysical signals. We also compare performance of our M-SSA GRACE solution with respect to others by predicting surface displacements induced by GRACE-derived mass loading and comparing results with independent displacement data from stations of the Global Navigation Satellite System (GNSS).
Finally, in Chapter 4 we discuss the possible application of a refined GRACE solution to answer debated post-glacial rebound questions. More precisely, we focus on separating the post-glacial rebound signal related to past ice melting and the present ice melting in the region of South Georgia.
