Forecasting of global data-binning parameters for high-resolution empirical geomagnetic field models

摘要

A recently developed empirical model, TS07D (http://geomag_field.jhuapl.edu/model/), has provided for the first time a realistic description of the storm-scale geomagnetic field, free from a priori assumptions about the shape of the main magnetospheric current systems. The model uses information about the global state of the magnetosphere and its solar wind driving in the form of the Sym-H index averaged over substorm scales, its time derivative, and a similarly averaged solar wind electric field. The set of global parameters is used to bin a large number of points in a historical magnetic field database and to fit the model using only a part of the database composed of points taken at times when the global data-binning parameters were close to those at the time of interest. Transition from modeling to forecasting in such models requires a modification of their data binning procedure to use only the information about the state of the solar wind and the magnetosphere prior to the time of interest. This can be done using another (forecasting) set of binning parameters, which is optimized to provide the best prediction of the original (modeling) binning set. Several sets of such parameters are investigated. It is shown that the original (modeling) set contains spurious information, which appears because of the averaging of the solar wind electric field and Sym-H index over future moments in time. As a result, the problem of the optimization of the forecasting set becomes ill-posed, and the prediction efficiency of the original binning parameters, especially the averaged time derivative of the Sym-H index, is strongly limited. The predictions are shown to be substantially improved when the forecasting set includes an exponential decay function of the solar wind electric field proposed originally by Burton et al. (1975) with the e-folding time of the order of one hour.