Neural network prediction of relativistic electrons at geosynchronous orbit during the storm recovery phase: effects of recurring substorms

Abstract. During the recovery phase of geomagnetic storms, the flux of relativistic (>2 MeV) electrons at geosynchronous orbits is enhanced. This enhancement reaches a level that can cause devastating damage to instruments on satellites. To predict these temporal variations, we have developed neural network models that predict the flux for the period 1–12 h ahead. The electron-flux data obtained during storms, from the Space Environment Monitor on board a Geostationary Meteorological Satellite, were used to construct the model. Various combinations of the input parameters AL, SAL, Dst and SDst were tested (where S denotes the summation from the time of the minimum Dst). It was found that the model, including SAL as one of the input parameters, can provide some measure of relativistic electron-flux prediction at geosynchronous orbit during the recovery phase. We suggest from this result that the relativistic electron-flux enhancement during the recovery phase is associated with recurring substorms after Dst minimum and their accumulation effect.

Key words. Magnetospheric physics (energetic particles, trapped; magnetospheric configuration and dynamics; storms and substorms)