Discussion of “Exceptional snowfalls and the assessment of accidental loads for structural design” by Sadovsky et al. [Cold Regions Science and Technology 72 (2012) 17-22]

In Europe, an observed value is treated as exceptional snow load “if the ratio of the largest load value to the characteristic load determined without the inclusion of that value is greater than 1.5”. While these exceptional values are neglected for the specification of the characteristic value, they form the basis of separate statistics with the basic aim of identifying extremely rare events with an annual exceedance probability of 1/10,000 which then are treated as an accidental load. The exceedance probability of the limiting value depends on the type of probability distribution and the variation coefficient V. If the annual extremes follow the Gumbel distribution, the annual exceedance probability of the limiting value is about 1/1000 for V = 0.5 and 1/500 for V = 1. As a matter of fact, the probability of observing values larger than the limiting value increases with the number of observation years and the number of independent stations. If all observations are treated as independent observations - as is done in the Sadovsky paper - it can be shown that a larger number of stations will show exceptional snowfalls just randomly. Therefore, there is no justification for the introduction of a separate accidental load case. Furthermore, the identified exceptional values do not form an appropriate basis for the identification of the 1/10,000 event.