Machine Learning-Based Tomography of the Ridgecrest Region

Abstract:

We perform ambient noise tomography using data recorded from Nodal arrays within a ~50 km by 50 km area including the 2019 M7.1 and M6.4 Ridgecrest, CA, earthquakes. An important goal of our study is to image properties of the fault zones, such as low velocities. Therefore, we adopt the locally sparse tomography (LST) approach because of its ability to resolve sharp contrasts in velocity maps, a challenging task for most conventional tomography methods. LST is an unsupervised dictionary learning approach with least-squares regularization, and the method directly learns from local patches without requiring a large volume of training data. Our preliminary imaging shows a 1-9 km wide low velocity zone around the fault traces for the M7.1 and M6.4 events, well correlated with recorded aftershocks, with a reduction in Rayleigh wave velocity of 25%.