Time-series Forecasting of Ionospheric Space Weather using Ensemble Machine Learning

Space weather describes varying conditions in the space environment between the Sun and Earth that can affect satellites and technologies on the Earth such as navigation systems, power grids, radio and satellite communications. In order to model and predict the space weather, a complex chain of physical processes between the Sun, the interplanetary space, the Earth’s magnetic field and the ionosphere have to be taken into account. Often, however, we do not have physical and/or mathematical relations that describe these coupled processes. On the other hand, there are data from satellites and observatories that monitor space weather processes between Sun and Earth. The approach of learning directly from the data via machine learning algorithms can lead to discovering the hidden knowledge and finding the functions, which can describe space weather processes. In this study, machine learning is applied to develop a forecast model of the space weather manifestation in the Earth’s ionosphere that can be used to estimate corrections for navigation applications as well as an early-warning system. Nonlinear relationships from the data are approximated utilizing machine learning algorithms based on ensemble learning such as bagging and boosting. The machine learning models performed well during severe space weather, with the RMSE (Root Mean Square Error) of the 24-hour forecast horizon being approximately twice the RMSE of the 1-hour horizon.     «

Space weather describes varying conditions in the space environment between the Sun and Earth that can affect satellites and technologies on the Earth such as navigation systems, power grids, radio and satellite communications. In order to model and predict the space weather, a complex chain of physical processes between the Sun, the interplanetary space, the Earth’s magnetic field and the ionosphere have to be taken into account. Often, however, we do not have physical and/or mathematical relat...     »