Randomness in Geomagnetic Storms

Abstract

Although geomagnetic forecasting is extremely important, geomagnetic predictions usually rely on forecasters' experience in distinguishing events expected to be geoeffective. Initial investigations with powerful forecasting techniques such as neural networks have recently been applied to geomagnetic prediction, but these are based on the assumption that the phenomenon to be forecasted has only several degrees of freedom. Thus a basic question related to geomagnetic forecasting is whether the randomness associated with geomagnetic evolution is produced by linear Gaussian noise or by nonlinear chaotic dynamics with only a few degrees of freedom. In order to try to answer this important question, we took a geomagnetic index time series and applied a recently developed technique which distinguishes nonlinear deterministic systems from linear ones.