Pre-earthquake Ionospheric Perturbation Analysis Using Deep Learning Techniques
Abstract views: 305 | PDF downloads: 97
https://doi.org/10.5281/zenodo.10202326
Abstract
It has been observed in many studies that ionosphere create a significant perturbation before major earthquakes. Therefore, forecasting of earthquakes on the basis of the ionosperic anomalies become trending. However, we still need more accurate and reliable advance technique to predict. In this study, we have analysed the Awaran, Pakistan (Mw 7.7) earthquake and associated Total Electron Content (TEC) anomalies which was happended on 24 Sep, 2013 using long short-term memory (LSTM) network model. We have recorded data from Global Navigation Satellite System (GNSS) for the proposed study with a 2-h temporal of 45 days. Afterwards, we set the hyperparameters, train the model on the dataset and maintain a prediction with high accuracy (0.07 TECU Loss). We performed the forecasting model on two different regions to differentiate the observed anomalies as pure seismic precursors without any external influence in the ionosphere. In this regard, we constructed one TEC time series right above the Awaran earthquake epicenter and one for a point located outside the earthquake preparation area (EPA). The results demonstrated that strong positive anomalies found on Sep 21, 3 days before the Awaran earthquake within EPA, which is consistent with findings of the previous studies. The occurrence of these irregularities can be attributed to the Awaran earthquake, which can be attributed to the calm space weather conditions during those days. On the other hand, none of the similar anomalies exist outside the EPA. Our study presents new insight into the AI techniques in seismo-ionospheric earthquake forecasting.
Keywords:
Anomaly Detection, Earthquake Forecast, Long Short Term Memory, Time Series Analysis, Total Electron ContentDownloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2023 Advances in Geomatics
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.