Extreme event distribution in Space Weather: Characterization of heavy tail distribution using Hurst exponents

Characterization of extreme conditions of space weather is essential for potential mitigation strategies. The non-equilibrium nature of magnetosphere makes such efforts complicated and new techniques to understand its extreme event distribution are required. The heavy tail distribution in such systems can be a modeled using Stable distribution whose stability parameter is a measure of scaling in the cumulative distribution and is related to the Hurst exponent. This exponent can be readily measured in stationary time series using several techniques and detrended fluctuation analysis (DFA) is widely used in the presence of non-stationarities. However DFA has severe limitations in cases with non-linear and atypical trends. We propose a new technique that utilizes nonlinear dynamical predictions as a measure of trends and estimates the Hurst exponents. Furthermore, such a measure provides us with a new way to characterize predictability, as perfectly detrended data have no long term memory akin to Gaussian noise Ab initio calculation of weekly Hurst exponents using the auroral electrojet index AL over a span of few decades shows that these exponents are time varying and so is its fractal structure. Such time series data with time varying Hurst exponents are modeled well using multifractional Brownian motion and it is shown that DFA estimates a single time averaged value for Hurst exponent in such data. Our results show that using time varying Hurst exponent structure, we can (a) Estimate stability parameter, -a measure of scaling in heavy tails, (b) Define and identify epochs when the magnetosphere switches between regimes with and without extreme events, and, (c) Study the dependence of the Hurst exponents on the solar activity.