A Generalised Machine-Learning Ambient-Adaptive GNSS Ionospheric Correction Model with Geomagnetic Field Components as Predictors

The Global Navigation Satellite System (GNSS) has emerged as a backbone of modern civilisation, industry, and society. Degradations and disruptions of the GNSS Positioning, Navigation, and Timing (PNT) service performance are caused by natural and adversarial sources. The ionospheric effects form the principal single class of the GNSS PNT performance degradation causes. Traditional GNSS ionospheric correction models appear unable to resolve the problem for their global nature, and the intrinsic lack of agility and flexibility. Here we contribute to the case with the proposal of concept and methodology for tailored GNSS ionospheric correction model development in support of GNSS resilience development, based on: (i) a massive dataset of long-term (annual) GNSS-derived total electron content TEC observations, as target variable (ii) a massive dataset of geomagnetic field density components, as predictors, and (iii) utilisation of statistical/machine learning predictive model development methods. The proposed approach emerges as a component of the previously introduced architecture-agnostic Ambient-Aware Application-Aligned (AA2) GNSS PNT concept, introducing the GNSS positioning environment situation awareness. Proposed concept and methodology is successfully demonstrated in the case of tailored GNSS ionospheric correction model development using the R environment for statistical computing in the case-scenario of mid-latitude single-frequency commercial-grade GNSS rover.