Abstract
Geomagnetic storms, which are governed by the plasma
magnetohydrodynamics of the solar-interplanetary-magnetosphere system,
entail a formidable challenge for physical forward modeling. Yet, the
abundance of high quality observational data has been amenable for the
application of data-hungry neural networks to geomagnetic storm
forecasting. Previous applications of neural networks to storm
forecasting have utilized solar wind observations from the Earth-Sun
first Lagrangian point (L1) or closer and have all generated
deterministic output without uncertainty estimates. Furthermore,
forecasting work has focused on indices that are also sensitive to
induced internal magnetic fields, complicating the forecasting problem
with another layer of non-linearity. We address these points, presenting
neural networks trained on observations from both the solar disk and the
L1 point. Our architecture generates reliable probabilistic forecasts
over Est, the external component of the disturbance storm time index,
showing that neural networks can gauge confidence in their output.