(Article of periodic en Anglais - 1985)

Document title

The effect of seasonal variation and serial correlation on the extreme value distribution of rainfall data

Authors(s) and Affiliation(s)



The classical theory of extreme values is based on the assumption that data are independent and identically distributed. Even when the data are correlated the classical theory may give the correct asymptotic distribution of maxima. However, the theory should be extended if rare events occur in clusters. Then the distribution of maxima is not directly related to the average number of exceedances of some high-level r as in the independent case, but to the number of clusters with large values. This result holds also for sequences with a seasonal component. For 5-min rainfall data from De Bilt, quantiles of the annuals maxima are overestimated by about 10 mm/h if the effect of serial correlation is ignored. (JPB).


Article of periodic

published at : Journal of climate and applied meteorology Boston, Mass.

Editor :

Millesime : 1985, vol. 24, no2 [pp. 154-160]

Bibliographic references : 14 réf.



Refdoc record number (ud4) : 12183448 : Permanent link - XML version
Powered by Pxxo