An alternative form for the statistical distribution of extreme avalanche runout distances

There are two main approaches in the avalanche literature for the statistical characterisation of extreme avalanche runout and these are known as the alpha-beta and runout ratio methods. Recent work suggests that the latter method is a more robust approach. This paper examines the statistical reasoning behind the selection of the Extreme Value Type I or Gumbel distribution used to characterise the runout distribution in the runout ratio method. On the basis that a threshold is often applied to the distribution of extreme avalanches used in analysis, it is proposed that the Generalised Pareto Distribution is an appropriate candidate distribution based on theoretical arguments. Such a proposal is tested against data on extreme avalanches in Iceland for both the largest events on specific paths and for all events exceeding a threshold. The latter is a peaks-over-threshold approach that potentially allows more robust distribution estimation due to the increase in data availability.