Statistical evaluation of extreme ice loads

The problem of how to estimate the probability of extreme icing events by historical data is crucial in the optimal design of structures in cold regions. For performing the extreme value analysis, numerous different methods are widely used, and several software packages available. There is no consensus on which method should be preferred. Furthermore, there are different criteria in use for the goodness of a method that estimates the cumulative distribution function and the extremes.By using the probabilistically correct goodness criterion, we present a method that provides estimates for the extremes that are considerably better than obtained by the conventional methods. This new method does not require subjective decisions by the user, and is particularly useful for small data sets, such as icing events observed over a short time-period.