Non-parametric statistics on extreme rainfall

Abstract

A monitoring program is examined consisting of 56 gauges in Denmark, which have observation periods ranging from 2 to 14 years, and which sample rain in a quantity of 0.2 mm with a resolution in time of 1 minute. Two variables have been investigated: peak intensity and depth. Design values for return periods in the range from 0.1 to 2 years have been estimated for each gauge separately by means of the bootstrap resampling method. The estimating includes expected value, standard deviation and confidence intervals of the design value. For large return periods the uncertainty of the estimates prevents a distinction by gauge between different statistical populations. However, for small return periods a test shows significant variation between gauges, i.e. the uncertainty of the estimates may not be assumed to be due to sampling variability only. -from Authors