Title: "Geodetic imaging of tectonic deformation: implications for earthquakes predictability"
Abstract
I will discuss how observed geodetic deformation relates to seismic activity and the predictability of stick-slip systems. I will first briefly describe multivariate statistical techniques and how they can be used to extract information from geodetic times series. I will present in particular a variational Bayesian Independent Component Analysis (vbICA). This technique is capable of separating tectonic sources and non-tectonic sources of deformation (e.g., deformation due to hydrology). I will present a study of the post-seismic deformation following the Mw 7.2, 2010, El Mayor-Cucapah (EMC, Mexico) earthquake and of Slow Slip Events (SSEs) in Cascadia from 2007 to 2017. For the EMC event, I separate in a natural way afterslip and viscoelastic relaxation. I further compare them with the seismicity and find that not only afterslip drove clustered seismicity after the earthquake, but also that long-range earthquake interactions were modulated by viscoelastic relaxation at large scales in space (>5 times the fault rupture length) and time (>7 years). Concerning the SSEs in Cascadia, the derived slip history offers the possibility to study the stick-slip behavior over multiple cycles. The results show a clear segmentation with a few major patches interacting with one another, a behavior that recalls that of a discrete body system. I will use tools from dynamical system analysis and extreme value theory to prove that we are dealing with a low-dimensional chaotic system rather than a stochastic non-linear system. It is possible to evaluate the dimension of the attractor and the entropy of the system, corresponding to the sum of all positive Lyapunov exponents. This allows to determine at what rate two trajectories in the phase space diverge, and thus set a predictability horizon. As SSEs might be regarded as earthquakes in slow motion, regular earthquakes might be similarly chaotic and predictable.