Automatic Seismic Phase Picking and Earthquake Location Using 2006 Yogyakarta Aftershock Datasets Based on A Machine Learning Algorithm

This study examines the application of automatic P and S wave picking using EQTransformer. The subsequent step involves determining earthquake hypocenters using GaMMA (Gaussian Mixture Associator) by associating the picking results based on a Gaussian distribution. This research uses seismic waveform data from June to August 2006, recorded by 16 seismic stations around the Special Region of Yogyakarta. The goal is to improve efficiency and accuracy in both the picking process and hypocenter determination. The problem with manual picking is time-consuming and dependent on human effort, while previous automated methods, using rule-based algorithms or simple statistical models, are ineffective with high-noise or complex data. This study proposes EQTransformer as a solution, offering more efficient and accurate pattern recognition in seismic data, even at a large scale. EQTransformer is a deep learning model based on Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) designed for the automatic picking of P and S wave phases. After automatic picking, earthquake hypocenters are determined using the Gaussian Mixture Associator (GaMMA), associating results based on Gaussian distribution. The findings indicate that this method produces picking results comparable to manual picking but with significantly faster processing times. The resulting hypocenter locations are accurate, particularly around the Opak fault, the area impacted by the major earthquake in May 2006. The significance of this study lies in its enhanced efficiency and consistency in seismic data processing, which is crucial for early earthquake detection and disaster mitigation.