Deep Learning Based Early Earthquake Detection through Terrestrial Optical Networks

We demonstrate the use of existing terrestrial optical networks as a smart sensing grid for early earthquake detection we integrate real ground displacement data from seven earthquakes, magnitudes ranging from four to six, to simulate the strains within fiber cables and collect large set of light’s polarization evolution data to train the model to detect Primary Waves (P waves) arrivals that precede earthquakes’ destructive Surface waves. . The main idea of our approach is to deploy a fast, accurate and reliable trained deep learning model. We evaluated the performance of LSTM and GRU models on experimentally emulated data collected from a 38 km deployed fiber link in Turin, Italy. Our results demonstrate that the GRU model consistently outperformed the LSTM model with 99% recall for P-wave detection.