A novel flexible model to calibrate single-layer height for ionospheric modelling using neural network model

Abstract

In the study of 2-D ionospheric modelling, the single-layer hypothesis is one of the most widely used definitions for construction and usage of ionospheric models, on which most ionospheric model and products exploited are based. Ionospheric effective height (IEH), as a key factor to affect ionospheric modelling accuracies by dominating the mapping errors, is defined as the single-layer height. Of previous studies, the fixed IEH model for a global or local area is unreasonable with respect of the dynamic ionosphere obviously. Bearing this problem in mind, we try to find a flexible-IEH solution based on neural network models, namely backpropagation neural network optimized by genetic algorithm (BP-NN-GA) and radial basis function neural network (RBF-NN). In this method, variables calibrating a given ray path and reflecting geomagnetic and solar activities are taken as inputs, and the outputs of IEHs are derived from mapping function converting slant total electron content (STEC) to vertical total electron content (VTEC) obtained from International GNSS Service (IGS) final global ionospheric map (GIM). Data of Hongkong Continuously Operating Reference Stations (CORS) on DOY 102, 2021 are chosen to construct and validate the flexible IEH model. First, some efforts are conducted to confirm variability of IEH and give some support to our contributions. Subsequently, minimum root mean square (RMS) values of VTECs from final GIMs and raw ionospheric measurements are calculated to determine IEH and RMS value varies from 1.73 to 5.81 TECU. Likewise, the RMS values are calculated based on the flexible IEH model in different periods and the results show the two kinds of trained NN models have slightly different performances, but most of RMS values are below 2 TECU a lot smaller than previous results. Finally, to validate the reliability of the flexible model, taking VTEC calculated by IGS final GIM as reference, we calculate the RMS values of VTECs estimated by GIM and spherical harmonics functions (SHFs) based on the fixed and flexible IEH model. Compare with values of around 2.2 TECU obtained by SHF based the fixed IEH model, the RMS values of SHFs based on the flexible IEH models of RBF-NN and BP-NN-GA are around 1.8 and 1.9 TECU with improving accuracies of approximately 18% and 13%, respectively. Gathering all findings above together verifies the reliability and feasibility of the flexible IEH model constructed by NN model.