Machine Learning for Solar Energetic Particle Forecasting

Abstract:

Radiation storms consisting of solar energetic particles (SEPs) are a major component of space weather. SEP events can result in spacecraft malfunction; pose a radiation risk for passengers and flight crew on polar flight routes; and significantly increase radiation dose for astronauts going beyond low Earth orbit. Physics-based numerical models are not yet at the level where they can provide the real-time forecasts required in an operational setting. Alternatively, there exist a number of empirical models which use real-time observations of SEP-associated signatures and phenomena, including the PROTONS prediction model currently in operations at NOAA’s Space Weather Prediction Center (SWPC). We have applied machine learning classification techniques to a historical dataset of SEP events to assess the applicability of a new machine learning approach for SEP forecasting. Performance of the machine learning model is measured against the SWPC PROTONS prediction model and compared to other empirical models capable of operating in real-time.