Search

Precipitation extremes form the basis of many engineering design decisions. Extremes are rare events which may differ strongly from “normal” observations. Unfortunately, some of the observed extremes may be inaccurate or false. The purpose of this investigation is to present a quality check of observed extremes using space-time statistical methods. As a first step, the biggest values for each observation location and event duration are selected. For each of these the observed values of all other stations corresponding to the same time steps are collected and transformed using a Box-Cox transformation factor derived from a fitted truncated normal distribution. The value at the extreme location is estimated using the surrounding stations and the calculated spatial variogram, and this estimated value is compared to the observed extreme. If the difference exceeds the critical value of the test, the extreme is flagged as possible outlier. The same procedure is repeated for different aggregations in order to avoid singularities caused by convection. The flagged extremes are then compared to the extremes of the surrounding stations using the same procedure – interpolation and subsequent comparison of the interpolated and the observed values. Flagged extremes are subsequently compared to the corresponding radar and discharge observations and finally, implausible extremes are removed. The procedure is demonstrated using observations of sub-daily and daily temporal resolution in Germany. [ABSTRACT FROM AUTHOR]