An optimization study of the prediction performance for the substorm model WINDMI is presented. The model is based on the Earth's magnetospheric dynamics and provides a low order description of the nightside energy loading and unloading that takes place during the substorm process. Previous studies of this model on isolated substorms have indicated that it can be a good predictor of solar wind driven substorm activity as measured by fluctuations in the AL index for selected substorms. Because the model is based on a set of VB s driven nonlinear ordinary differential equations which can exhibit bifurcation and catastrophe like behavior, an optimization of the model using conventional minimization techniques over a large data set does not work well. For such systems the genetic algorithm method of optimization is more efficient at exploring the parameter space. We present the results of a genetic algorithm optimization of WINDMI using the Blanchard-McPherron and the Bargatze data set and test statistically alternative forms of the model which include the effects of ionospheric conductivity enhancements and region 2 coupling. A key result from the large scale computations used to search for a uniform convergence of the prediction over the 117 substorm database, is the finding that there are three distinct types of VB s-AL wave forms characterizing the substorms in the Blanchard-McPherron database. Two types are given by the internally triggered WINDMI model and the third type requires an external trigger such as the northward turning of the IMF model of Lyons 1995.
分享
引用
