Towards Near-Real-Time Seismic Phase Recognition, Event Detection, and Location with Deep Neural Networks in Volcanic Area of Campi Flegrei

The real-time phase picking, detection, and location of seismic events is a crucial challenge for monitoring in densely populated volcanic areas. In such contexts, low-magnitude events may escape traditional detection methods due to high levels of anthropogenic noise, which often masks weak seismic signals. This study presents the implementation of a near-real-time automatic event detector with a seismic phase recognizer, pick associator, and localiser. The system is based on PhaseNet, a well-established deep neural network recognized for its effectiveness in seismology. The main innovation introduced in this work lies in the direct application of this method to real-time data streams. This integration allows for the enhanced identification and cataloguing of low-magnitude seismic events that would otherwise remain unobserved. The adoption of the system in a real-time operational context not only increases monitoring sensitivity and responsiveness but also contributes to a more detailed and comprehensive understanding of seismic activity in critical volcanic areas, providing essential data for risk assessment and prevention.