A comprehensive study on application of LSTM neural networks for radon-based earthquake anomaly detection in Istanbul

Abstract

In this study, soil gas radon concentrations were measured in the vicinity of the North Anatolian Fault Zone, one of the most seismically active regions in Turkey. The collected 41,550 radon measurement data points, and 82 earthquakes were comprehensively analyzed and utilized to develop a predictive model based on Long Short-Term Memory (LSTM), a deep learning approach well-suited for time-series analysis. The results indicate that the LSTM model effectively captures temporal variations in radon concentrations and reveals their potential correlation with seismic activity in the Istanbul region. This study underscores the potential of artificial intelligence techniques in advancing earthquake and radon research and provides valuable insights for future earthquake monitoring and prediction strategies.