Variational Bayesian Independent Component Analysis applied to geodetic time series

Geodetic time series data are usually studied through classical statistical techniques, that is decomposing them into different deterministic signals. Recently, multivariate statistical techniques have been applied to geodetic data, in order to extract as much information as possible from them. An example is the Principal Component Analysis (PCA), used both to detect network errors in GNSS data and to identify geophysical signals common to a certain region. The latter approach is particularly useful for understanding geophysical processes. Nonetheless, a strong limitation of the PCA is that it is not able to separate multiple mixed sources. In other words, the PCA technique is not effective in treating the so-called Blind Source Separation (BSS) problem. For this goal, it reveals to be an efficient technique the Independent Component Analysis (ICA). I will introduce an ICA technique based on the Variational Bayesian approach, and the results concerning both tests on synthetic data and application to real data. In particular, I have characterised the spatio-temporal evolution of ground displacements related to the 2009 L'Aquila (central Italy) earthquake, the 2012 Emilia (northern Italy) seismic sequence, and the 2006 Guerrero (Mexico) Slow Slip Event (SSE).