Comparative Analysis of Topside Ionospheric Scale Height Modelling Based on Three Machine Learning Algorithms

The top ionosphere refers to the ionospheric region above the height of the maximum electron density of F2 layer, and the topside ionosphere scale height is an important parameter to characterize the ionospheric electron density profile above the top. The variation rule of the topside ionosphere scale height model is analyzed, and the relevant model is established to improve the accuracy of the empirical model of the top ionosphere. It plays a vital role in space science research and related engineering applications. In this paper, three kinds of topside ionosphere scale height models are established based on COSMIC observation data from 2007 to 2020, combined with three machine learning algorithms (BP, RF and SVM) and peak parameters (peak height and peak density) are introduced as the main control parameters. BP model only has 0.12GB memory and 0.27s estimated time, which has obvious advantages. The estimated value of the RF model has the most significant correlation with the observed value and the slightest error. The correlation coefficient is 0.84, and the root-mean-square error is 25.97km, significantly improving accuracy.