Determining new representations of “Geoeffectiveness” using deep learning

Abstract:

Disruptions to our space environment, i.e., 'space weather,' originate at the Sun and are manifested in complex structures in the solar wind. One of the great unresolved and oft-asked questions in space weather is how these structures lead to (or fail to lead to) space weather impacts such as Geomagnetic storms, which can potentially cause economic and technological disasters by imposing the geomagnetic induced currents (GICs) through our infrastructures like power grid, railways, telecommunication systems, and pipelines. Here, we outline a new approach to quantify the connection using ground-based magnetic perturbations that lead to GICs. The new approach that includes deep learning has potential to more capably represent the complex relationships between solar wind input, BH, and dB/dt. Ultimately, we create a new forecast model of ground magnetic field perturbations (as determined from global networks of ground magnetometers from the Super Magnetometer Initiative) given past history of solar wind properties measured at L1. We will describe new findings on the information content of solar wind and time history of magnetic perturbations, which may guide future space weather prediction efforts, and detail the capabilities of the new predictive model.