Regional Ionosphere Delay Models Based on CORS Data and Machine Learning

Abstract

The ionospheric refraction of GNSS signals can have an impact on positioning accuracy, especially in cases of single-frequency observations. Ionosphere models that are broadcasted by the satellite systems (e.g., Klobuchar, NeQuick-G) do not include enough details to permit them to correct single-frequency observations with sufficient accuracy. To address this issue, regional ionosphere models (RIMs) have been developed in several countries in the western Balkans based on dense Continuous Operating Reference Stations (CORS) observations. Subsequently, a RIM for the western Balkans was built using an artificial neural network that combined regional ionosphere parameters estimated from the CORS data with spatiotemporal (latitude, longitude, hour of day), solar (F10.7) and geomagnetic (Kp, Dst) parameters. The RIMs were tested at the solar maximum (March 2014), a geomagnetic storm (March 2015), and the solar minimum (March 2018). The new RIMs mimic the integrated electron density much more effectively than the Klobuchar model. Furthermore, RIMs significantly reduce the ionospheric effects on single-frequency positioning, indicating their necessity for use in positioning applications.

Uncertainty Quantification for Machine Learning‐Based Ionosphere and Space Weather Forecasting: Ensemble, Bayesian Neural Network, and Quantile Gradient Boosting

“…The Bayesian neural network (BNN) represents a modification of an ANN in which the deterministic network parameters or weights are replaced by probability distributions of those weights (Abdar et al, 2021;Blundell et al, 2015;Kendall & Gal, 2017); for more details on the architecture and computation of an ANN, see Natras, et al (2023a). The probability distributions are used to model the uncertainty in the weights and consequently can be used to estimate the uncertainty due to the model parameter uncertainty based on Bayes' theorem.…”

Ionospheric Error Models for Satellite-Based Navigation—Paving the Road towards LEO-PNT Solutions

“…Mathematical models rely on pure mathematical or numerical solutions to provide theoretical values of ionospheric parameters [29,30]. Empirical models represent the ionosphere using equations derived from observational data [31]. Physical models are based on equations that simulate the chemical content and processes governing the ionosphere [31,32].…”

“…Empirical models represent the ionosphere using equations derived from observational data [31]. Physical models are based on equations that simulate the chemical content and processes governing the ionosphere [31,32]. Data-driven models use multiple methods (e.g., linear regression, autocorrelation, machine learning, neural networks, etc.)…”

“…In [44,45], processing algorithms and measurements from reference stations were used to compute the ionospheric delay at a regional level with high accuracy but with a certain latency. Other research, such as [31], used reference stations along with Machine Learning (ML) algorithms and neural networks to establish accurate values of the ionospheric delay at a regional level. Since there are many neural network-based models in the literature, it has been hard to find a relevant benchmark among them, and therefore, they are not included in our comparisons.…”

MaxEnt SeismoSense Model: Ionospheric Earthquake Anomaly Detection Based on the Maximum Entropy Principle

“…The accuracy of such models has become inadequate to meet the growing demand as human exploration of ionospheric seismic anomalies deepens. Because these models, such as IRI and Nequick, are typical global models, they are good at predicting long-term ionospheric changes, but they cannot be expected to be sensitive to phenomena occurring on shorter time scales, such as rapid changes in the ionosphere caused by magnetic storms or earthquakes [24,25].…”