Machine learning model of the plasmasphere to forecast satellite charging caused by solar storms.

Solar storms are hazardous events consisting of a high emission of particles and radiation from the sun that can have adverse effect both in space and on Earth. In particular, the satellites can be damaged by energetic particles through surface and deep dielectric charging. The Prediction of Adverse effects of Geomagnetic storms and Energetic Radiation (PAGER) is an EU Horizon 2020 project, which aims to provide a forecast of satellite charging through a pipeline of algorithms connecting the solar activity with the satellite charging. The plasmasphere modeling is an essential component of this pipeline, as plasma density is a crucial parameter for evaluating surface charging. Moreover, plasma density in the plasmasphere has very significant scientific applications, as it controls the growth of waves and how waves interact with particles. Successful plasmasphere machine learning models have been already developed, using as input several geomagnetic indices. However, in the context of the PAGER project one is constrained to use solar wind features and Kp index, whose forecasts are provided by other components of the pipeline. Here, we develop a machine learning model of the plasma density using solar wind features and the Kp geomagnetic index. We validate and test the model by measuring its performance in particular during geomagnetic storms on independent datasets withheld from the training set and by comparing the model predictions with global images of He+ distribution in the Earth’s plasmasphere from the IMAGE Extreme UltraViolet (EUV) instrument. Finally, we present the results of both local and global plasma density reconstruction.