72-Hours Ahead Prediction of Ionospheric TEC using Radial Basis Function Neural Networks

Abstract

Prediction of Indonesia's local and regional ionosphere TEC for the next 72 hours is required for space weather services at PUSSAINSA through the Space Weather Information and Forecast Services SWIFTS website, especially during ionosphere predictions on Friday which requires predicting the ionosphere condition from Saturday to Monday according to user needs. To this date, a global modeling the form of the W index, has been used for the prediction. Therefore, we developed a local ionosphere TEC prediction model as a starting point in the development of a regional ionosphere prediction model for Indonesia. The prediction model is built using a Radial Basis Function Neural Network (RBFNN). The input of the RBFNN model is the ionospheric TEC data for the previous 72 hours and the minimum value of the geomagnetic disturbance index (Dst) for the last3 days. The output isa prediction of the TEC 72 hours ahead. In the testing phase, the RBFNN model was able to predict local TEC with a daily standard deviation of between 2.75 and 4.9 Total Electron Content Unit (TECU).