An LS-MARS method for modeling regional 3D ionospheric electron density based on GPS data and IRI

The methods of developing an accurate and effective ionospheric electron density (IED) model have greatly interested ionosphere researchers. Numerous scholars have proposed many effective and reliable models and methods of global positioning system (GPS)-based computerized ionospheric tomography (CIT) in the past decades. This study introduced a new function-based CIT method, namely the LS-MARS (Least Squares method-Multivariate Adaptive Regression Splines), combining MARS with IEDs calculated by International Reference Ionosphere (IRI) to automatically choose the best representing basis functions for the three-dimensional (3D) electron density inside that modeling area. This selected basis functions was substituted into the observation equation of the GPS total electron content (TEC) to calculate the design matrix. Finally, the weighted damped least squares (WDLS) were adopted to reestimate the IED model coefficients. In contrast to common function-based CIT methods, the LS-MARS can be used to attain optimal 3D model automatically, flexibly, adaptively based on the IRI without a priori knowledge of the IED distribution mathematical function. The findings indicated that the LS-MARS model had a smaller recovery TEC error than did the MARS_IRI2012 model, and the VTEC calculated using the LS-MARS model was closer to the VTEC obtained from International GNSS Service (IGS) final IONEX files than was the VTEC calculated using the MARS_IRI2012 and IRI2012. Therefore, this method exhibits strong modeling effectiveness and reliability, and can be an efficient alternative method for estimating regional 3D IED models.