Investigation of new approaches for the determination of the annual maximum rainfall depths of different durations

A careful assessment of extreme rainfalls is needed in the hydrological practice. Annual maximum rainfall depths, Hd, of different durations, d, are commonly the basic elements for determining a representative hyetograph over a watershed as an input to lumped rainfall-runoff models finalized to the design of hydraulic structures. Some related issues require to be examined more in depth and properly addressed. Three factors of wide interest over a given watershed are investigated here: the homogenization of historical series associated to each rain gauge station and including data, Hda, characterized by different rainfall aggregation times, the analysis of the temporal distribution of the heaviest extreme rainfalls for each duration and the identification of a probability density function, PDF, that for each value of d overall well represents the dataset deduced from the available rain gauge network. New methodological approaches are proposed and applied to a very large inland area in the Mediterranean zone. First, although the shape of the rainfall temporal pattern significantly affects the adjusting sampling factor to be applied to Hda to obtain a reliable estimate of Hd, Hde, our results show that it cannot be practically used for data homogenization. Indeed, for each value of d a variety of shapes were observed and the Hda series can be better dealt with relationships involving sampling factor and duration. As to the temporal analysis for each d, the 100 heaviest Hda values through the whole study area, Top100d, have been considered. They are strictly linked with the development of alluvial events and landslides of catastrophic type. It has been found that the temporal trend of the frequencies normalized by the functioning rain gauge network density doesn’t indicate significant climate change effects, in terms of storm magnitude too. On the other hand, an incorrect analysis based on the standard frequencies for short and moderate durations would lead to a heavy misinterpretation of the temporal trend in the last 30 years because of the role of convective storms. As to the determination of an areal-average PDF for each duration, a well-founded approach which relies upon the Regional Frequency Analysis is first adopted and, for a geographic area equipped with a large number of homogeneous stations, a simplified approach is also proposed. For each d it relies upon the grouping of all the Hde values associated to the whole rain gauge network in a sole series of large dimension. Finally, results in terms of rainfall depth-durationfrequency curves for different return periods are compared with the ones obtained by a classical approach widely used in the hydrological practice.