CONSEQUENCES FOR ESTABLISHED DESIGN PRACTICE FROM GEOGRAPHICAL VARIATION OF HISTORICAL RAINFALL DATA

The Danish measuring network for high-resolution rainfall data was ini tiated in 1979 and consists of approximately 50 tipping bucket rain ga uges separated by one to 300 km, covering an area of 43,000 square kil ometres. T-year design events and the associated sampling error varian ces were estimated at each site using the bootstrap method and the par tial duration series method and a methodology was developed for quanti fying the inter-site correlation structure due to spatial coverage of rain storms. The data reveals a dramatic geographical(regional) variat ion that may be divided into true regional variation and variation due to (correlated) sampling errors. Further analyses indicate that the o bserved variation can be explained only partially by correlation with regional climatological variables and that a significant residual vari ation remains, especially for large return periods. The new perception s question the value of local rain data for design and call for an inc reased use of statistical concepts in engineering design practice. (C) 1997 IAWQ. Published by Elsevier Science Ltd.