Magnetic Cloud Prediction Algorithm using Bayesian Inference for Real-Time Forecasts of Geomagnetic Storms

We present details of a space weather forecasting tool which attempts to accurately predict the occurrence and severity of large geomagnetic storms caused by prolonged periods of south directed magnetic field components associated with magnetic clouds. The algorithm updates and modifies the work of Chen et al. (1996, 1997) to run in a real-time operational environment, with input solar wind data from the Deep Space Climate Observatory (DSCOVR) spacecraft at L1. From the real-time magnetic field measurements, the algorithm identifies the initial magnetic field rotation signature assuming it represents the initial phase of a magnetic cloud. From the field rotation, estimates of the solar wind profile upstream of the spacecraft are determined, in particular the expected event duration and maximum Bz field strength. Using Bayesian inference, the tool returns the probability of a large geomagnetic storm occurring and a measure of its geoeffectiveness, updating the forecast as the event evolves. An expected warning time of several hours, to potentially more than 10 hours (Arge et al. 2002), is possible for certain magnetic field orientations. We discuss the current algorithm performance as well the limitations of the model.