Regional variation of climatological cloudburst frequency estimated from historical observations of daily precipitation sums

Cloudburst are geographically localized extreme rainfall events where a large amount of rain falls within a few hours. The combination of small spatial scale, short duration and scarceness makes it difficult to reveal any systematic regional differences in occurrence. Here we estimate climatological cloudburst frequencies from the daily precipitation sums for a dense network of 161 historical Danish stations covering the period 1914-2010. We do this using supplementary sub-hourly precipitation observations from a modern network and relate the daily probability of cloudburst occurrence to the corresponding daily precipitation sum using binary regression. This allows a subsequent estimation of the cloudburst frequency from the daily sums from the historical observations. To validate the method, we use stations from the modern network that have been operating for 30 years or longer. For these stations, we demonstrate significant skill by comparing observed and estimated cloudburst frequencies with a jackknife procedure. We then apply the binary regression model using the 161 historical series as input and estimate climatological cloudburst frequencies throughout Denmark. We find large and systematic regional variations across Denmark. The methodology also allows determining temporal changes of cloudburst frequency and we find large differences across Denmark.