Least square support vector machine for detection of tec-seismo-ionospheric anomalies associated with the powerful Nepal Earthquake (Mw=7.5) OF 25 April 2015

Abstract

Due to the irrepalable devastations of strong earthquakes, accurate anomaly detection in time series of different precursors for creating a trustworthy early warning system has brought new challenges. In this paper the predictability of Least Square Support Vector Machine (LSSVM) has been investigated by forecasting the GPS-TEC (Total Electron Content) variations around the time and location of Nepal earthquake. In 77 km NW of Kathmandu in Nepal (28.147° N, 84.708° E, depth=15.0 km) a powerful earthquake of Mw=7.8 took place at 06:11:26 UTC on April 25, 2015. For comparing purpose, other two methods including Median and ANN (Artificial Neural Network) have been implemented. All implemented algorithms indicate on striking TEC anomalies 2 days prior to the main shock. Results reveal that LSSVM method is promising for TEC sesimo- ionospheric anomalies detection.