Search

Three type of uncertainties related to rainfall hazard study are presented. The first one concerns the measurement of extreme rainfall using a sparse rain gage network: it has been shown that before 1945, the probability that a rainfall event less than 200 km2 is not measured by the network, is up to three times more important comparing to the current period of observation. The second point is about the shape of the probability distribution of frequency (PDF) of the daily maxima annual rainfall: does the tail of the distribution correspond to a Gumbel PDF or does it show an ?hyper exponential? behaviour? According to the chronics duration, one can difficulty argue while analysing a single rainfall chronic because of sampling uncertainties. It has been shown that a regional PDF has an hyper exponential tail. The quantiles are less subject to sampling variability comparing to the one obtain with a local PDF fitting. In the last point the mean area rainfall PDF is studied according to the area. A scale relation between the PDF corresponding to different areas is presented. The sampling variability of the PDF shape parameter is presented. Conclusion concerning the asymptotic behaviour of the mean areal rainfall PDF needs further analysis.