Extreme‐Event Magnetic Storm Probabilities Derived From Rank Statistics of Historical Dst Intensities for Solar Cycles 14–24

Abstract

A compilation is made of the largest and second‐largest magnetic‐storm‐maximum intensities, − Dst 1 and − Dst 2 , for solar cycles 14–24 (1902–2016) by sampling Oulu Dcx for cycles 19–24, using published − Dst m values for 4 intense storms in cycles 14, 15, and 18 (1903, 1909, 1921, 1946), and calculating 15 new storm‐maximum − Dst m values (reported here) for cycles 14–18. Three different models are fitted to the cycle‐ranked − Dst 1 and − Dst 2 values using a maximum‐likelihood algorithm: A Gumbel model, an unconstrained Generalized‐Extreme‐Value model, and a Weibull model constrained to have a physically justified maximum storm intensity of − Dst m  = 2500 nT. All three models are good descriptions of the data. Since the best model is not clearly revealed with standard statistical tests, inference is precluded of the source process giving rise to storm‐maximum − Dst m values. Of the three candidate models, the constrained Weibull gives the lowest superstorm occurrence probabilities. Using the compiled data and the constrained Weibull model, a once‐per‐century storm intensity is estimated to be − Dst 1  = 663 nT, with a bootstrap 68% confidence interval of [497, 694] nT. Similarly, the probability that a future storm will have an intensity exceeding that of the March 1989 superstorm, − Dst m > 565 nT, is 0.246 per cycle with a 68% confidence interval of [0.140, 0.311] per cycle. Noting (possibly slight) ambiguity in the rankings of storm intensities, using the same methods, but storms more intense than those identified for cycles 14–16, would yield a higher once‐per‐century intensity and a higher probability for a − Dst m  > 565 nT storm. Past and possible future magnetic storm intensities are investigated. As part of this work, a dataset is developed of the most intense and second most intense storms for each of the past 11 solar cycles (1902–2016)—augmenting a traditional dataset that only covers the past 6 solar cycles (1957–2016) with recently published intensities for several magnetic superstorms and with new storm intensity estimates, reported here and derived from historical magnetic observatory records. These data are analyzed using statistical methods that provide estimates of the probability of future magnetic superstorms. A storm as intense as that of March 1989, which caused widespread disruption of technological systems and an electricity blackout in Québec, Canada, is predicted to occur, on average, about every four solar cycles. This is twice as often as estimated using only the traditional shorter dataset. A once‐per‐century storm is estimated to be substantially more intense than that of March 1989. Magnetic‐storm‐maximum − Dst m , ranked per solar cycle, are well represented by a Weibull model Based on geomagnetic data for solar cycles 14–24 yields, a 100‐year storm has an intensity of at least − Dst m  = 663 nT Based on geomagnetic data for solar cycles 14–24, the average return rate for a storm as intense as March 1989 (− Dst m  = 565 nT) is less than 4.1 solar cycles