Topside Ionospheric Scale Height Prediction: Random Forest Modeling and Performance Evaluation

The characteristics of the topside ionospheric scale height are influenced by factors such as solar activity, geographical location, and season. Establishing an accurate and reliable model for topside ionospheric scale height features is of paramount importance for ionospheric climatology research, as it subsequently enhances the accuracy of empirical models for the topside ionosphere. In this study, we propose the application of the Random Forest method in modeling the topside ionospheric scale height features. The data used for this study encompass ionospheric data from COSMIC (Constellation Observing System for Meteorology, Ionosphere, and Climate) occultation observations between 2007 and 2020. The input parameters of the model include Day of Year (DOY), Local Time (LT), Latitude, Longitude, peak parameters (peak height hmF2 and peak density NmF2), solar activity indices, and geomagnetic activity indices. The results indicate that established model effectively captures the spatiotemporal variations in the ionospheric scale height features, with a root mean square error (RMSE) of 25.87 km when applied to sample data outside the COSMIC occultation training dataset, and a correlation coefficient of 0.84.