Prediction of Northern Polar ionospheric scintillation Using the Random Forest and LSTM Machine Learning Methods

Ionospheric scintillation is one of the most important space weather phenomena in the polar ionosphere, which is clo sely related to people's lives, especially to the widely used satellite navigation and positioning system with unavoidable impacts, so the prediction of ionospheric scintillation needs to be addressed. Nowadays, a network of multiple Global Navigation Satellite System (GNSS) receivers has been established in the Arctic, and the sc intillation index data provided by the GNSS receivers make the prediction of northern polar ionospheric scintillation possible. To predict the polar scintillation events, we developed a new deep le arning method. Based on the 2011–2016 dataset from the Eureka station (79.99°N,274.09°E), by using a random forest approach to regressively analyze the physical covariates that characterize the solar-terrestrial environment, we evaluate the correlation of 24 different parameters with the phase scintillation index (σ<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">φ</inf>), filter out the five covariates with the highest correlation as inputs to the long-short time memory (LSTM) algorithm, and make a prediction of σ<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">φ</inf> one hour in advance. The results show that the model can accurately predict polar ionospheric scintillation, demonstrating that the deep learning method has a wide range of prospects in the field of predicting polar scintillation.