Short-duration rainfall extremes in Sweden: a regional analysis

We investigate short-duration rainfall extremes in a 22-year long (1996–2017) time series of 15-min values from 128 stations in Sweden. Two different approaches are used to fit extreme value distributions (EVDs) to the data: generalized extreme value distribution with annual maximum values and generalized Pareto distribution with peak-over-threshold values. First, EVDs are fitted to the data from each station, and the extreme value parameters are used together with the station coordinates in a cluster analysis. Based on the clustering results, a division of Sweden into four regions is proposed. Following the station-year method, the time series from all stations in each cluster are pooled into regional series after which depth–duration–frequency (DDF) statistics are calculated. Averaged over durations and return periods, the highest DDF values are found in region south-west, being ∼25% higher than the lowest ones found in the northern region. A six-parameter DDF function is proposed which, together with a simple empirical expression for the associated confidence interval, allows for easy DDF estimation at arbitrary durations and return periods, including uncertainty. No tendencies of change in magnitude or frequency of the short-duration extremes are found during the period.