Bayesian Inference for Solar Flare Extremes

While solar flares are a frequent occurrence, extreme much rarer events and have the potential to cause disruption life on Earth. In this paper we use Extreme Value Theory model flares, with inference performed in Bayesian paradigm. The data used been provided by National Oceanic Atmospheric Organisation consist of recordings peak flux measurements. After proposing several methods for analysis selecting our preferred technique—which substantially increases precision estimates key quantities interest—we improve upon technique still further considering informative prior distributions. Doing so, estimate that Halloween-type event, Carrington-type might occur once (on average) every 49 (29, 85) 92 (50, 176) years respectively (95% credible intervals shown parentheses). These findings similar those obtained Tsiftsi De la Luz (2018), https://doi.org/10.1029/2018SW001958 Elvidge Angling https://doi.org/10.1002/2017SW001727 however, confidence both wider than found study, lending increased certainty estimated time between such magnitude work. We argue taking extremal index into account, even when measure indicates weak temporal dependence, is beneficial analysis.