ALBERT

Description: [1]  Some of the potentially most destructive effects of severe space weather storms are caused by the Geomagnetically Induced Currents. GICs can cause failures of electric transformers and result in wide-spread blackouts. GICs are induced by the time variability of the magnetic field, and are closely related to the time derivative of the local magnetic field perturbation. Predicting dB/dt is rather challenging, since the local magnetic perturbations and their time derivatives are both highly fluctuating quantities, especially during geomagnetic storms. The currently available first-principles based and empirical models cannot predict the detailed minute-scale or even faster time variation of the local magnetic field. On the other hand, Pulkkinen et al. [2013] demonstrated recently that several models can predict with positive skill scores whether the horizontal component of dB/dt at a given magnetometer station will exceed some threshold value in a 20-minute time interval. In this paper we investigate if one can improve the efficiency of the prediction further. We find that the Space Weather Modeling Framework, the best performing among the five models compared by Pulkkinen et al. [2013], shows significantly better skill scores in predicting the magnetic perturbation than predicting its time derivative, especially for large deviations. We also find that there is a strong correlation between the magnitude of dB/dt and the magnitude of the horizontal magnetic perturbation itself. Combining these two results one can devise an algorithm that gives better skill scores for predicting dB/dt exceeding various thresholds in 20-minute time intervals than the direct approach.