Reza Mansoori

eismic phase detection, phase picking, and distinguishing between true seismic signals and noise are critical for earthquake monitoring and the development of accurate earthquake catalogs. Traditional manual methods for identifying P-wave and S-wave arrivals and differentiating them from noise have proven complex, especially given the continuous data recorded by seismic monitoring stations. The emergence of Artificial Intelligence (AI) and Machine Learning (ML) has revolutionized this process, enabling automatic phase detection and phase picking with increased accuracy. EQTransformer, a deep neural network model, provides a comprehensive solution for seismic signal analysis through its multi-task architecture. Incorporating 1D convolutions, LSTM units, and attention mechanisms, the model processes seismic time-series data, creating high-level representations that facilitate precise detection of earthquake signals and differentiation of P- and S-phases. This study applies EQTransformer to continuous earthquake data recorded by Iranian seismological stations, specifically examining an earthquake in the Malard region of Tehran Province. Results indicate that the model robustly detects seismic signals, achieving a total of 49302 detected signals, with 56661 P picks and 52774 S picks across 52 stations. The attention-based architecture of EQTransformer ensures efficient phase picking and demonstrates significant potential for advancing automated seismic data analysis. This approach optimizes data collection, enhances accuracy, and supports the development of reliable earthquake catalogs essential for geophysical research and seismic monitoring