The effects of non-stationarity on the evaluation of critical design storms

The critical storm is generally carried out to design urban drainage systems and other flood management works starting from the available historical information. Its evaluation associated with a fixed return period is usually obtained by fitting the annual maxims of the rainfall depth with an extreme value distribution. This statistical procedure, however, leads to dubious results when the data present a non-stationarity, induced for example, by a long-term variability. To assess the effects of non-stationarity, four daily rainfall series observed in Italy, with at least 90 years of continuous data, are analysed here. For each record and each year of the observation period, critical design storms are estimated fitting the annual maxima collected in the past, so allowing us to assess the progress of the design storm along time. Four different extreme value distributions are used. The results show that an analysis of non-stationarity is required when urban drainage systems and other hydraulic engineering works are designed. (C) 1998 Published by Elsevier Science Ltd. All rights reserved.