Predicting Equatorial Ionospheric Convective Instability Using Machine Learning

Abstract

The numerical forecast methods used to predict ionospheric convective plasma instabilities associated with Equatorial Spread-F (ESF) have limited accuracy and are often computationally expensive. We test whether it is possible to bypass first-principle numeric simulations and forecast irregularities using machine learning models. The data are obtained from the incoherent scatter radar at the Jicamarca Radio Observatory located in Lima, Peru. Our models map vertical plasma drifts, time, and solar activity to the occurrence and location of clusters of echoes telltale of ionospheric irregularities. Our results show that these models are capable of identifying the predictive power of the tested inputs, obtaining accuracies around 75%.