Constructing an Earthquake Catalog Based on Deep Learning

Abstract

In this study, we employed deep learning-based seismic signal detection method (PhaseNet) to detect the seismic sequences of the Weishan region in Shandong, China. By combining seismic association techniques and NLLoc seismic location technology, we constructed a seismic catalog. The deep learning-based catalog was compared with the manual catalog, and its performance, completeness, and accuracy were tested and evaluated for medium and small seismic sequences. The results demonstrated that the deep learning-based seismic catalog exhibited higher efficiency. Compared to the manual catalog, it included 1.44 times more seismic events. Additionally, the deep learning-based catalog showed better seismic association, with the majority of missed events concentrated in cases of smaller magnitudes, while capturing fewer phases. In summary, the deep learning-based seismic catalog demonstrated advantages in terms of efficiency, seismic quantity, seismic association, and the number of picked phases. This study provides a more accurate and comprehensive seismic catalog for the detection and location of medium and small seismic sequences, offering valuable support to seismic monitoring and related research fields.