Analysis of Seismic Ionospheric Effects and Prediction of TEC During Earthquakes Occurred in Indonesia Based on GPS Data Academic Article uri icon

abstract

  • Total electron content (TEC), which quantifies the quantity of free electrons in the Earth’s ionosphere, is a crucial parameter that experiences discrepancies during seismic events. This study investigates the potential of utilizing TEC prediction at the BAKO position in Indonesia during earthquakes. TEC data and solar parameters were collected for six preselected earthquakes, encompassing the earthquake event periods. Three prediction models, namely, ARMA, OKSM 1, and OKSM 2, were employed to predict TEC for a period spanning 8 days. The input parameters required for TEC prediction were obtained from the IONOLAB and OMNIWeb database. The OKSM 1 model is constructed with the input parameters like solar radio flux at 10.7 cm (F10.7), disturbance storm time index (Dst), solar wind (Sw), sunspot number (SSN), and TEC values, while the OKSM 2 model is developed with the parameters like geomagnetic indices (Kp and Ap) and solar indices SSN and F10.7 along with TEC data. The ARMA model is constructed with TEC data. The primary objective of this research is to assess the utility of TEC prediction based on the influence on input parameters for the kriging models and to identify the most effective model for predicting TEC variations associated with seismic events. Four evaluation metrics were systematically utilized to gauge the performance of each model. This rigorous evaluation aims to deliver perceptions into the predictive accuracy, reliability, and potential practical implications of TEC predicting during earthquakes. Upon comparison, the OKSM 2 model demonstrated superior predictive accuracy, exhibiting a notable agreement with the true TEC. The results suggest that OKSM 2 holds promise as a reliable model for earthquake‐related TEC prediction. The average RMSE values range from 4.06 to 8.06, indicating the models’ ability to predict seismic events with a reasonable magnitude of error. Similarly, the average MAE values, ranging from 3.32 to 6.71, underscore the models’ overall accuracy in predicting the absolute differences between actual and predicted TEC. The CC values, averaging between 0.97 and 0.99, highlight a strong relationship between predicted and actual TEC values. Additionally, the average sMAPE values, ranging from 0.11 to 0.21, demonstrate the models’ effectiveness in minimizing percentage‐based errors. While variations exist across different earthquakes, these average metrics collectively suggest promising predicting capabilities.

publication date

  • 2025

start page

  • 9453529

volume

  • 2025

issue

  • 1