Design rainfall estimates are usually obtained through various kinds of sampling: annual maximum values, peak-over-threshold values or daily series. The aim of this paper was to compare scale parameters of maximum rainfall distribution while taking into account the effect of seasonality and spatial reduction, as well as the use of different sampling methods.
Two sets data from the Moselle (192 rain gauges) and Rhône (213 rain gauges) regions of France were used. The seasonal effect is significant and can induce an under-estimation of rainfall distribution by about 10 to 20% for areal rainfall to a catchment, and 10 to 40% for local rainfall to a specific location. The areal reduction effect was about 10 to 15%, when comparing local and areal statistics.
The comparison of areal rainfall distributions was carried out for 63 and 11 catchments, in the Meuse and Rhône areas respectively. Because more information was available, the peak-over-threshold sampling method and the use of daily series provided more accurate estimates than the annual maximum sampling method. In using the peak-over-threshold approach, one should carry out a sensitivity analysis to choose the threshold, and use probabilistic distributions in accordance with null rainfall values. The two alternative sampling approaches were in agreement and are recommended for extreme rainfall analysis.
Maximum rainfall, sampling, gradient of extreme values (GRADEX), areal reduction factor, seasonal effect.
Unité de Recherche Hydrologie-Hydraulique,
3 bis quai Chauveau,
69336 Lyon Cedex 09,