EVALUATING DEEP LEARNING MODELS FOR AUTOMATICPHASE PICKING IN LARGE EARTHQUAKES (M > ۴.۵):ANALYSIS OF IIEES SEISMIC NETWORK

As the volume of data from seismic networks grows, manually analyzing and identifyingearthquake phases becomes impractical. This has fueled the rise of automated methods, in recent yearsdeep learning models, for accurate and efficient phase identification. This study compared theperformance of four popular deep learning models and an energy detector against expert analysis,using nearly ۹,۰۰۰ seismic phases from earthquakes exceeding magnitude ۴.۵, reported by the IIEESofficial website. The results revealed PhaseNet to be the most effective model for determining P-typephases in our analysis and can be used in automatic procedures of the National Center of BroadbandSeismic Network of Iran by IIEES. Although the EQTransformer shows the most effective model fordetecting S-type phases. However, this requires manual checking by an expert and cannot be used inautomatic procedures.

کلیدواژه ها:

Detection ، Phase-Picking ، Deep Learning ، Signal Processing ، Seismic Network