Probability distributions of daily rainfall extremes in Lazio and Sicily, Italy, and design rainfall inferences

•Daily rainfall data modelling is fundamental to define hydraulic project’s forcing.•Light-tailed distributions may substantial under- or over-estimate extreme events.•We analyse daily rainfall maxima of two Italian regions with different climate.•Heavy-tailed distributions describe more accurately maxima daily rainfall values. We investigate samples from two Italian regions, i.e. Lazio and Sicily, located in central and south Italy, respectively, and characterized by two diverse climates. In engineering practice, the study of maxima daily rainfall values is commonly dealt with light-tailed probability distribution functions, such as the Gumbel. The choice of a distribution rather than another may cause estimation errors of rainfall values associated to specific return periods. Recently, several studies demonstrate that heavy-tailed distributions are preferable for extreme events modelling. Here, we opt for six theoretical probability distribution functions and evaluate their performance in fitting extreme precipitation samples. We select the samples with two common methods, i.e. the Peak-Over-Threshold and the Annual Maxima. We assess the best fitting distribution to the empirical samples of extreme values through the Ratio Mean Square Error Method and the Kolmogorov-Smirnov test. The assessment of the best fitting distribution to daily rainfall of the two different areas investigated here leads to interesting remarks. Despite the diversity of their climate, results suggest that heavy-tailed distributions describe more accurately empirical data rather than light-tailed ones. Therefore, extreme events may have been largely underestimated in the past in both areas. The proposed investigation can prompt the choice of the best fitting probability distribution to evaluate the design hydrological quantities supporting common engineering practice.