MESCNN: Magnitude estimation system based on convolutional neural networks

Magnitude is a critical parameter in earthquake early warning systems, directly influencing alert issuance and warning levels. In this study, we introduce the Magnitude Estimation System based on Convolutional Neural Networks (MESCNN), a novel approach built upon the Python programming language and the TensorFlow deep learning framework. MESCNN automates the calculation of earthquake magnitudes using real-time seismic data, leveraging the capabilities of convolutional neural networks (CNN) to analyze seismic waveforms. The system is designed to enhance the accuracy and efficiency of magnitude estimation, thereby enabling more timely and reliable earthquake warnings to reduce the impact of seismic events. • The system features a CNN-based model for rapid magnitude estimation (3–10 s) from user-uploaded seismic data. • Trained on 170,000+ records from Japan and New Zealand, the model shows strong accuracy and robustness. • P-wave arrivals are refined via STALTA and AIC, improving input reliability. • Visual results aid emergency response and post-disaster rescue.