Skills for forecasting space weather

Abstract

Coronal mass ejections (CMEs) from the Sun can have severe impacts on the Earth environment in the form of geomagnetic storms. These storms pose a risk to the global technological infrastructure, making the prediction of these events imperative. In this paper, we have broadened the statistical verification of the Bz4Cast tool, the first empirically-driven model to forecast solar wind magnetic vectors inside a CME prior to their Earth arrival. Twenty five CME events (between 2012 and 2016) have been tested with the Bz4Cast model, and the skills have been compared to the heuristic approach of NOAA's Space Weather Prediction Center (SWPC) G-scale for 3-day geomagnetic storm forecasts. For a broad range of scores, and within uncertainty, the Bz4Cast architecture provided the same skill as the experienced on-duty forecasters at SWPC. The most prominent difference is that the Bz4Cast architecture provides a slightly higher false alarm ratio than the SWPC 3-day forecast.