Consequences for established design practice from geographical variation of historical rainfall data

The Danish measuring network for high-resolution rainfall data was initiated in 1979 and consists of approximately 50 tipping bucket rain gauges separated by one to 300 km, covering an area of 43,000 square kilometres. T-year design events and the associated sampling error variances were estimated at each site using the bootstrap method and the partial duration series method and a methodology was developed for quantifying the inter-site correlation structure due to spatial coverage of rain storms. The data reveals a dramatic geographical(regional) variation that may be divided into true regional variation and variation due to (correlated) sampling errors. Further analyses indicate that the observed variation can be explained only partially by correlation with regional climatological variables and that a significant residual variation remains, especially for large return periods. The new perceptions question the value of local rain data for design and call for an increased use of statistical concepts in engineering design practice. (C) 1997 IAWQ. Published by Elsevier Science Ltd.