Retraining neural networks for the prediction of Dst in the Rice magnetospheric specification and forecast model
Costello, Kirt Allen
Master of Science
Artificial Neural Networks have been developed at Rice University for the forecasting of the Dst index from solar wind and Dst parameters. The one hour Dst index is an Earth based measurement of variations in the H-component of the magnetic field that is indicative of the strength of the ring current, and thus magnetic storms. Comparison of the neural networks' outputs to the OMNI dataset values of Dst will be presented. These results verify the success of the neural networks in predicting Dst. Network performance when predicting Dst two or more hours into the future and testing of MSFM output based on neural net Dst input for the August 1990 storm will be presented. Comparisons between MSFM equatorial particle fluxes and CRRES satellite observations show the MSFM 10 keV proton equatorial fluxes raise interesting questions about the MSFM's use of the Dst input parameter.