Presentation #106.31 in the session Solar Eruptive Events: Posters.
Solar Energetic Particle (SEP) event is a sudden enhancement of protons and heavier ions in space that is often associated with solar eruptions such as flares and coronal mass ejections (CMEs). SEPs directed towards Earth determine the dosage exposure on astronauts and spacecraft equipment outside the magnetosphere, while protons >100MeV can penetrate the Earth’s upper atmosphere. Hence, efficient predictions of SEPs are vital to mitigate space weather hazards for space travel.
In this work, we address the prediction of >10 MeV SEP events from a multivariate time series perspective using supervised machine learning (ML) models. Between 1986 and 2020, we identified and developed a benchmark data set of over 400 SEP events. We consider the X-ray and proton fluxes from the Geostationary Operational Environmental Satellites (GOES). We examine the correlations of these fluxes and extract statical features to understand important parameters that help in the predictability of SEP events. This strategy offers a new perspective in establishing predictions for geo-effective SEP events arising from a large flare.
In this work, we develop an ensemble of time series classifiers to forecast SEP events under a binary classification problem domain. The results from the models will be ensembled using probability-based learners to obtain efficient forecasts with anticipation of reduced false alarm rates and improved warning time. The motivation behind this work to use ML is that the models can learn and make decisions based on observational data and issue quicker forecasts to improve upon the results of the existing statistical models.