Data-Adaptive Detection of Transient Deformation in GNSS Networks

Friday, 19 December 2014: 4:30 PM
Eric Calais1, Damian Walwer1 and Michael Ghil1,2, (1)Ecole Normale Supérieure Paris, Paris, France, (2)University of California Los Angeles, Los Angeles, CA, United States
Dense Global Navigation Satellite System (GNSS) networks have recently been developed in actively deforming regions and elsewhere. Their operation is leading to a rapidly increasing amount of data, and position time series are now routinely provided by several high-quality services. These networks often capture transient-deformation features of geophysical origin that are difficult to separate from the background noise or from seasonal residuals in the time series. In addition, because of the very large number of stations now available, it has become impossible to systematically inspect each time series and visually compare them at all neighboring sites. In order to overcome these limitations, we adapt Multichannel Singular Spectrum Analysis (M-SSA), a method derived from the analysis of dynamical systems, to the spatial and temporal analysis of GNSS position time series in dense networks. We show that this data-adaptive method — previously applied to climate, bio-medical and macro-economic indicators — allows us to extract spatio-temporal features of geophysical interest from GPS time series without a priori knowledge of the system's dynamics or of the data set’s noise characteristics. We illustrate our results with examples from seasonal signals in Alaska and from micro-inflation/deflation episodes at an Aleutian-arc volcano.