Tecniche per la Difesa del Suolo e dall'Inquinamento

This paper presents a comparison between the relevant statistics of annual maximum daily rainfall during a historical period, evaluated using data simulated by regional climate models developed within the EURO-CORDEX initiative, and those derived from the same models in future projection periods. The daily data of 17 combinations of general (GCM) and regional (RCM) climate models under two emission scenarios (RCP4.5 and RCP8.5) were considered. The data, available on a grid with resolution of 0.11° (ap-proximately 12.5 km), underwent preliminarily bias correction using the distribution mapping method. This correction is based on historical data from the daily gridded observational dataset E-OBS-0.1 (Cornes et al., 2018) at a resolution of approximately 11 km, covering the whole Europe. The investigated area covers the entire territory under the jurisdiction of the Po River District Basin Author-ity (86861 km2) including areas in French, Switzerland and Austria, and some Adriatic river basins. For each location with available precipitation data, both in the simulation (historical) and in the projection (future) period, probability distributions that best fit annual maximum daily rainfall data were identified, referring to the methodology proposed by Hosking and Wallis (1997), based on dimensionless L-moments. The selection favoured the Generalized Extreme Value (GEV), Generalized Logistic (GLO), Generalized Pa-reto (GPA) and three-parameter lognormal (LN3) distributions. Once the parameters of the four probabil-ity distributions were defined, daily rainfall heights corresponding to specific return periods were evaluated as the arithmetic mean of the quantiles associated with those return period, according to the procedure known as model averaging (Burnham and Anderson, 2002). Then, the percentage change between the pre-cipitation height of a given return period in the projection period and in the historical simulation period was computed for each model. The results presented in this paper refer to the median values derived from the set of 17 models, showing, in the majority of the study area, an increase of the extreme values, more remarkable for the long-term period (2071-2100) and the RCP8.5 scenario. Finally, the comparison between the quantiles derived from the application of climate models in the histori-cal period and those obtained from observed historical data revealed significant discrepancies, suggesting the need for further investigation into the bias correction procedure.