Prediction of GPS-TEC on Mw > 5 Earthquake Days Using Bayesian Regularization Backpropagation Algorithm
Seçil Karatay, Saide Eda Gul
Abstract
The detection of earthquake precursor signals a few days before the earthquake day is one of the most studied subjects today. In recent years, a strong correlation is observed between earthquakes and ionospheric parameters. In this study, a feed-forward back propagation artificial neural network (ANN) Bayesian regularization (BR) algorithm is applied to detect the seismic disturbances and anomalies by predicting global positioning system (GPS)-total electron content (TEC) on earthquake days with magnitude greater than 5. It is observed that TEC is predicted with greater error margins for the stations at a maximum distance of 50 km from the epicenters. The errors for earthquakes less than Mw 7 are smaller than those for greater than 7.