Neural network prediction of solar energetic proton events with long lead times
Abstract
The high-energy particles that are emitted by the sun during Solar Proton Events (SPEs) pose a serious hazard to interplanetary and near-earth spacecraft causing damage to electronics, spurious signals and a general disruption of spacecraft operations. An analysis of GOES X-rays and full disk solar radio flux prior to proton events has revealed significant differences compared to times at which no proton events occurred. Selecting predictor variables using the F-statistic, neural network models have been generated which use data from tens of days prior to SPEs to predict their occurrence 48 hours in advance with a 65% success rate. This is an improvement in lead time of an order of magnitude over current SPE prediction models.
- Publication:
-
Solspa 2001, Proceedings of the Second Solar Cycle and Space Weather Euroconference
- Pub Date:
- March 2002
- Bibcode:
- 2002ESASP.477..517P
- Keywords:
-
- Solar Energetic Protons