Prediction of the AL index using solar wind parameters

摘要

Two empirical models, one simple, one more complex, are introduced to predict the AL index, a measure of the Earth’s auroral activity derived from magnetometers in the Northern hemisphere. Both models are based on solar wind and magnetometer data from the year 1995. The simple model predicts the 10-min averaged AL index for 1995 using only solar wind measurements with a prediction efficiency of 0.644 and a linear correlation of 0.803. The more complex model has a prediction efficiency of 0.723 and a linear correlation of 0.850. The simple model forward integrates with a 10-min time step the model AL term using a driver term on the basis of the solar wind magnetic field magnitude and direction, solar wind velocity and density, and a nonlinear decay term whose amplitude depends on the model AL amplitude. The more complex model adds annual, semiannual, and diurnal variations, saturation, and some time delays and an additional term that is a direct function of the solar wind magnetic field and velocity. Though both models were optimized to predict the 10-min averaged AL index, both models predict a longer-averaged AL index substantially better (prediction efficiency of 0.736 and 0.825, respectively, for the 2-hour averaged AL using the simple and complex models) while predicting the difference between 2-hour averaged AL and the 10-min AL poorly (prediction efficiency of 0.083 and 0.136, respectively) implying much less predictability for shorter timescale variations. The models show that the AL index is strongly dependent on the solar wind magnetic field and velocity but is practically independent of the solar wind density.