Parametrisation of a Regional Hydrologic Model for Piemonte: simulation of large scale floods

The aim of this study is the regionalization of the parameters of a hydrological model (TUWmodel) for the Piedmont region, through an innovative procedure involving machine learning techniques (HydroPASS) for estimation of runoff at ungauged locations. The regionalization is performed by the determination of a regional functional relationship between model parameters and a selected set of descriptors (CDs) over a reference grid (resolution around 12 km) without any a priori assumptions about the function itself. The first part of the study involved the determination of the needed input data to then move to the actual application of the algorithm. The climate data (Precipitation, mean temperature, and PET) over the grid (from 1961 to 2021) have been obtained from ARPA Piemonte and simple calculations while the discharges observations at 127 gauged stations have been obtained from open databases. This information has been used to perform the local lumped calibrations of the TUWmodel parameters for each catchment (DE algorithm), resulting in an average ME of 0.86. The 52 selected descriptors, evaluated for both the catchments and the reference grid, are instead divided into to five groups (climate, morphology, land use, soil, and curve numbers). After the application of the regionalization algorithm, the searched distributed parameters have been obtained resulting in a distributed ME of 0.7. Further analysis shown the flexibility of the regionalization procedure when provided with more limited number of CDs, resulting in similar, not degraded, efficiencies. The newfound distributed parameters have then been applied for the reconstruction of five historical flooding events (1994, 2000, 2008, 2016, 2020) over the Piedmont region. The model, despite presenting some limits in the quality of the observed data, well performed resulting in coherent simulations, both spatially and temporary, at ungauged locations, in agreement with the descriptions contained in the official reports redacted by the local authorities.