Stochastic modeling of the auroral electrojet index

Substorms are often identified by bursts of activities in the magnetosphere-ionosphere system characterized by the auroral electrojet (AE) index. The highly complex nature of substorm-related bursts suggests that a stochastic approach would be needed. Stochastic models including fractional Brownian motion, linear fractional stable motion, Fokker-Planck equation and Ito-type stochastic differential equation have been suggested to model the AE index. This paper provides a stochastic model for the AE in the form of fractional stochastic differential equation. The long memory of the AE time series is represented by a fractional derivative, while its bursty behavior is modeled by a Levy noise with inverse Gaussian marginal distribution. The equation has the form of the classical Stokes-Boussinesq-Basset equation of motion for a spherical particle in a fluid with retarded viscosity. Parameter estimation and approximation schemes are detailed for the simulation of the equation. The fractional order of the equation conforms with the previous finding that the fluctuations of the magnetosphere-ionosphere system as seen in the AE reflect the fluctuations in the solar wind: they both possess the same extent of long-range dependence. The introduction of a fractional derivative term into the equation to capture the extent of long-range dependence together with an inverse Gaussian noise input describe the right amount of intermittency inherent in the AE data.