Mosaicking Andean morphostructure and seismic cycle crustal deformation patterns using GNSS velocities and machine learning

We use unsupervised machine learning techniques to analyze continental-scale crustal motions in areas affected by the seismic cycle of large subduction earthquakes along Chilean Trench. Specifically, we agglomerative clustering algorithm as an exploratory tool investigate spatial patterns GNSS regional velocities without complexity modeling a physical source. present velocity field including all available data for two-time windows (pre-2014, 2018–2021) that represents two periods with different deformation cycle. test pre-processing methodologies design features from GNSS-derived velocities. The first method uses direction and magnitude secular rates input algorithm. These results show spatially related deformation, separating latitudinal segments fore-arc back-arc, well regions postseismic relaxation. Thus, highlighting effectiveness this mapping first-order active zone, are particularly variations on interplate coupling transient deformation. In more sophisticated approach, surface strain rotational second methodology. Here, develop novel methodology estimate rotation accounting heterogeneity GNSS-network. determine scale at which these estimated least squares inversions, using Bayesian model class selection method. distribution stations allows identify heterogeneities scales larger than 50 km, being notorious main > 100 km. Interestingly, correlation between segmentation geologic structural domains arc back-arc. Our demonstrate ability combination inverse methods efficiently their relationship regional-scale geological structures. Furthermore, our analysis suggests Andean structures influence observed field.