Regional frequency analysis of short duration rainfall extremes using gridded daily rainfall data as co-variate

Abstract

A regional partial duration series (PDS) model is applied for estimation of intensity duration frequency relationships of extreme rainfalls in Denmark. The model uses generalised least squares regression to relate the PDS parameters to gridded rainfall statistics from a dense network of rain gauges with daily measurements. The Poisson rate is positively correlated to the mean annual precipitation for all durations considered (1 min to 48 hours). The mean intensity can be assumed constant over Denmark for durations up to 1 hour. For durations larger than 1 hour, the mean intensity is significantly correlated to the mean extreme daily precipitation. A Generalised Pareto distribution with a regional constant shape parameter is adopted. Compared to previous regional studies in Denmark, a general increase in extreme rainfall intensity for durations up to 1 hour is found, whereas for larger durations both increases and decreases are seen. A subsample analysis is conducted to evaluate the impacts of non-stationarities in the rainfall data. The regional model includes the non-stationarities as an additional source of uncertainty, together with sampling uncertainty and uncertainty caused by spatial variability.