Abstract
People across the world have been fascinated by solar eclipses for thousands of years. Solar eclipses are not only fascinating to observe but also provide opportunities for scientific research. During a solar eclipse, the quantity of solar energy reaching the Earth’s surface is reduced as the Moon passes in front of the Sun. This reduction in solar energy can have an effect on the Total Electron Content of the Earth’s ionosphere. In this paper, prediction and analysis of TEC variations in the Ionosphere during the solar eclipses happened on 26.12.2019 between 04:51 to 7:34 hours (UTC) and 09.03.2016 between 12:18 to 1:02 hours (UTC) over the Indonesia region were done by using two models: Ordinary Kriging based Surrogate Model (OKSM) and Feed-Forward Neural Network (FFNN). During the eclipse period, the TEC values were predicted by the OKSM and FFNN models and it is validated using literature. For this study, the GPS data belonging to the BAKO station situated in Indonesia were collected from IONOLAB servers and the input parameters were collected from the OMNIWEB servers. Forty days prior TEC data and input parameters were used to predict the TEC values. The credibility of the predicted results is assessed using statistical factors such as RMSE, CC, MAE, MAPE, sMAPE and R-Square. The statistical results show OKSM has performed well when compared to the FFNN model over the annular and total solar eclipse period. The study suggests that combining multiple modelling methods, such as OKSM and FFNN can improve our understanding of ionospheric variability during solar eclipses and provide more accurate predictions of TEC variations. This has important implications for satellite communications and navigation systems that rely on accurate TEC measurements for positioning and timing.