Rainfall extremes exhibit temporal clustering at multi-decadal time scales, most probably as a result of persistence in large scale atmospheric circulation over such time scales. Analysis of a 107-year time series of 10-minute rainfall intensities since 1898 at Uccle, Brussels, has shown that the 1960s and the 1990s-2000s had a higher frequency and amplitude of high rainfall intensities at various time scales in the range between 10 minutes and 1 month. These periods are alternated with periods of lower rainfall quantiles, e.g. in the 1970s-1980s.
The climate oscillations have to be accounted for when calculating extreme rainfall statistics, e.g. IDF relationships and synthetic storms commonly applied on the basis of urban drainage systems design. The importance of this and how this climate oscillation accounting can be done is demonstrated in this paper based on the Uccle rainfall data. Old and new IDF statistics, based on, respectively, shorter and longer rainfall series have been compared. It is shown that recent increases in rainfall statistics should not necessarily be attributed to climate change but may also be due to a different positioning of the periods with available rainfall data in comparison with the climate oscillation high and low periods. Comparison of old IDF statistics based on the period 1967-1993 versus new statistics based on the full period 1898-2007 or the period 1970-2007 covering one climate oscillation cycle, shows 7.5% difference in extreme rainfall quantiles for return periods higher than 1 year. Adjustment with +7.5% is required to remove the bias in the old rainfall design values in comparison with the long-term statistics.