EXTREME RAINFALL ESTIMATION BY COMBINING DATA FROM SEVERAL SITES

Nowadays it is well recognized that inference about the form of the upper tail of the distribution of precipitation amounts has to be based on a joint evaluation of several records in a region. Three methods of regional estimation are reviewed: (a) the station-year method; (b) regional averaging of statistics of the data; and (c) maximization of a joint likelihood function. Methods (b) and (c) are illustrated with the regional estimation of parameters of the generalized extreme value (GEV) distribution for 1 day annual maximum precipitation amounts in The Netherlands. The adequacy of regional models can be tested in several ways. Special attention is paid to the number of exceedances of regional estimates of certain large quantiles (return periods 250, 500 and 1000 years) in a catalogue of all daily precipitation amounts greater-than-or-equal-to 80 mm at official KNMI stations during the period 1866-1989. The consistency of frequency curves for different durations is also considered. Other methods to evaluate the suitability of regional models are briefly mentioned.