A Convolutional Neural Network Based Teleseismic Shear-wave Splitting Measurements Auto-picking Method

A Convolutional Neural Network Based Teleseismic Shear-wave Splitting Measurements Auto-picking Method Yanwei Zhang & Stephen S. Gao Abstract: Shear-wave splitting (SWS) analysis has been widely used over the past several decades to provide critical constraints on crustal and mantle structure and dynamics. Numerous studies demonstrate that in order to obtain reliable splitting measurements, an essential step is to visually verify all the measurements to reject problematic measurements, a task that is increasingly time consuming due to the exponential increase in the amount of data. Additionally, the criteria for an acceptable measurement are operator-dependent and thus are subjective. In this study, we utilize a convolutional neural network (CNN) based method to automatically select reliable SWS measurements. The CNN was trained and validated using 86,903 human-verified and published XKS (including SKS, PKS, and SKKS) splitting measurements from 1108 stations mostly from North America. The XKS seismograms are firstly processed using a set of automatic procedures based on the minimization of transverse energy approach to compute the splitting parameters, which are then used to calculate the corrected radial and transverse components. The input data of the CNN consist of 4 seismic traces, including the original and corrected radial and transverse components. The trained CNN is tested using several sets of 1000 synthetic seismograms with different levels of random noise. When the signal to noise ratio is greater than 4.0, CNN is able to identify 99.8% of the non-null measurements as reliable measurements. Additional testing using real data recorded in Alaska and central China reveals that the agreement between CNN and human selected measurements is 87.96% and 88.93%, respectively, for the two areas. The ongoing study demonstrates the potential of CNN based approaches for greatly improving the efficiency and objectiveness in shear wave splitting and other structural seismological investigations.