A model for the Am (Km) planetary geomagnetic activity index and application to prediction

We studied the existence of dynamical stochastic relations in the evolution of the am index. A first analysis of the autocorrelation functions showed evidence of several seasonalities. We first used linear (ARMA) models, and it was found that these do not account for the whole internal dynamics of the data series. We then used various non-linear models to provide a better fit to reality. The forecast performances of the non-linear models are not significantly different from those of the linear model. We give a tentative explanation for the failure of the non-linear predictions. Finally, ARCH models were used in order to take into account the fact that the confidence interval for the predicted value depends on past observations.