CORREL: Automated Onset Estimation for Controlled-Source Seismic Experiment and 3D Preliminary Seismic Velocity Models for Northern Chile.

Reyes-Wagner, Valentina; Comte, Diana; Calle-Gardella, Daniela; Roecker, Steven; Rietbrock, Andreas

Abstract

Automated onset estimation is a powerful tool to create a large seismic dataset in a short time. The development of these techniques have shown great progress, but mainly focused on natural seismicity. In this work we focus on the analysis of a controlled-source seismic experiment off-shore northern Chile (PICTURES experiment) which was recorded by 50 short period stations deployed on-shore. We tested the Regressive EStimator autopicker package (REST) developed for the automated onset estimation of P-and S-seismic waves. In the process, we learned that this approach was both too restrictive and not considering the a priori information and the fundamental characteristics of the air-gun shots that could help improve the detection. Thus, a new algorithm named CORREL was designed and developed with the purpose of estimating the first onset of controlled-source seismic events. The main obstacles presented by the dataset come from the event's narrow bandwidth, low SNR and highly variable response at the stations. CORREL complements the difficulties of REST with a cross-correlation approach, based on the assumption that the distortion in response is gradual from one event to the next. Also, a polynomial predictor is used based on the assumption that the moveout of arrival times is a relatively smooth function of distance. With this approach, we were able to improve the detection and quality of the picks. By comparing the results obtained with REST and CORREL we found that, although using REST, a greater number of picks were made (34% for REST vs 29% for CORREL of the total theoretic picks), the number of false picks was greater and therefore the quality and quantity of true onsets was superior with CORREL. This work also shows the preliminary 3D Vp and Vp/Vs models obtained with local earthquake tomography based on the onsets resulting from the CORREL run. Along with this data set, we also incorporate the natural seismicity that occurred in the instrumentation period and other events recorded by different networks installed after the 2014 Iquique earthquake, all processed with the usual REST algorithm.

Más información

Fecha de publicación: 2021
URL: https://ui.adsabs.harvard.edu/abs/2021AGUFM.S15D0285R/abstract