SELF-SIMILARITY AS A PHYSICAL BASIS FOR REGIONALIZATION OF FLOOD PROBABILITIES

The self-similarity conjecture can provide an assessment of regionalization procedures for flood frequency analysis if the scaling properties of flood flows as parameterised by drainage area are considered. The question whether simple scaling can be used to discriminate among different flood probabilities in relatively small regions with high variability of climate and geomorphological features is investigated. Detailed simulations for a small basin in Thyrrhenian Liguria (North-West Italy) show that self-similarity or simple scaling properly represent the spatial variability of flood probabilities. The scaling properties of annual flood series for North-West Italy are also investigated. Although flood statistics display multiscaling properties when considering North-West Italy as an individual homogeneous region, one can classify geographically consistent regions in this area where self-similarity cannot be rejected. Accordingly, the index flood procedure can be properly applied to these regions in order to estimate flood probabilities at ungauged catchments, and the scaling exponent of flood probabilities can be interpreted as the signature of complex interactions between climate and hydrology producing extreme flows in rivers of a region. This conjecture is substantiated by the common scaling exponent of flood probabilities resulting from both basin simulations and frequency analysis of annual series for the region including this basin.