Eh predictions w.r.t the persistence model and SWMF model (results)

In this study, we proposed a multi-fidelity boosting approach, i.e., ProBoost, to forecast geoelectric fields during geomagnetic storms. The developed model was based on a Gated Recurrent Unit (GRU) neural network and a uncertainty quantification method based on ACCRUE.

The results for all 50 selected storm events, including comparisons with the persistence model and, if available, the SWMF model, across various forecast intervals, ranging from 10 minutes to 60 minutes ahead.