Application of a neural network model to GPS ionosphere error correction

This paper presents the use of neural network modeling to predict electron concentration in the altitudes from 140 to 660 km as well as total electron content (TEC) to reduce GPS signal propagation errors. In training the neural network we have used incoherent scatter radar (ISR) data from the Arecibo Observatory, solar flux data from National Oceanic and Atmospheric Administration (NOAA), and simulated data from the International Reference Ionosphere (IRI). The ISR data covers almost two solar cycles, which allows the network to make accurate predictions based on local time, seasonal, and solar cycle variations above Arecibo, Puerto Rico (18.21 N, 66.45 W). We demonstrate that neural network models are not only accurate predictors of dynamic systems, but also perform better than the commonly referenced IRI model.