Bayesian Spatially Smooth Regional Estimation of flood quantiles: Case study in Northern Italy

Regional flood frequency analysis methods make use of hydrological information from multiple sites to reliably estimate flood quantiles in ungauged or poorly gauged catchments. The Spatially Smooth Regional Estimation method, which is based on multiregressive estimation of L-moments without requiring the definition of homogeneous regions (Laio et al., 2011) is the standard method used in north-western Italy. The regional regression models assume homoscedasticity of the uncertainty associated with the estimation of variation (L-CV) and asymmetry (L-CA) of the flood frequency curves. This assumption could be criticised since one would expect larger uncertainty in the estimation of large values of high order L-moments. In this work, we apply the model of Laio et al. (2011) to a large database in northern Italy. The novelties of our application are: (1) the use of Bayesian inference for parameter estimation, and (2) the comparison of results obtained assuming homoscedasticity or a linear dependence between the regionally estimated L-CV (and L-CA) and their uncertainty. The results are compared in terms of similarity between the estimated flood frequency curves in gauged and ungauged basins. Laio, F., D. Ganora, P. Claps, and G. Galeati (2011) Spatially smooth regional estimation of the flood frequency curve (with uncertainty), Journal of Hydrology, 408, 67-77. doi:10.1016/j.jhydrol.2011.07.022