Deep Learning of Ionosphere Single-Layer Model and Tomography

Abstract

The ionosphere layer modeling and determining the parameter of total electronic content (TEC) has an important role in recognizing this layer and controlling its effects on human activities. Ionospheric variations, followed by atmospheric disturbances, can have detrimental effects on mapping and navigation systems such as global positioning system (GPS), communication systems, and security systems. For this purpose, a single-layer ionospheric model and tomography were considered using GPS data in northwest and southeast Iran. Generalized singular value decomposition (GSVD) is used in tomography to solve the ill-posed least squares problem. We analyze 2 data set from the Iran GPS measurements and produce deep learning (DL) TEC daily variation map and tomography at northwest and southeast Iran sites. The novelty of this research is the study of 2D and 3D ionosphere based on DL. This method has been used to model the ionosphere with 100 hidden layers. The average DL accuracy is 81% and with IRI-2016 it has a 93% correlation.