Deep learning enhances earthquake monitoring by directly mining seismic waveforms. Existing neural networks, however, struggle with generalization across diverse regions. This study uses a data recombination method to create generalized earthquakes for training, enabling universal application with varying monitoring setups for real-time Earthquake Early Warning (EEW). Applied to Japan and California datasets, models reliably report earthquake locations and magnitudes within 4 seconds of the initial P-wave arrival, with mean errors of 2.6–7.3 km and 0.05–0.32, respectively. This approach simplifies real-time EEW, eliminating complex empirical configurations of traditional methods.

Xiong Zhang, Miao Zhang