Long Short Term Memory Networks (LSTM)-Monte-Carlo Simulation of Surface Ionization using Radon

Abstract

Earthquake events are usually associated with the atmospheric processes modification. Some of the important events in seismic periods include the atmospheric electricity/conductivity modification, which is often related to the ion population in the earthquake preparation area vicinity. Radon, together with cosmic radiation is the major ionization sources in the lower troposphere. The ion-pair production rate is estimated as caused by radon in Erzincan, a city along the North Anatolian Fault Zone, Turkey. Long short term memory and Monte Carlo methods are proposed to account for uncertainty in the estimations and also in the radon data to simulate the ion production rate. The advantage of the LSTM model is taken to study radon anomalies during the M=5 Girlevik earthquake, Erzincan. Radon concentration in Erzincan is found to be very high when compared to other regions, and this might relate to the geological settings of the region. Radon concentration is found to increase prior to the 5.0 Girlevik earthquake. According to estimations, 23×109 ion-pairs m-3s-1 were generated during the earthquake. The ion production estimation rate due to radon and its progeny in Erzincan is at the order of 109 ion-pairs m-3s-1. When added to other ionization sources more pronounced conditions could favor ionospheric perturbations. Ionospheric disturbance simulations would be significant in regions with high radon concentration for understanding atmospheric processes.