The role of modeling scheme and model input factors uncertainty in the analysis and mitigation of backwater induced urban flood-risk

Recent flood-related disasters, which affect many areas of the world due to a combination of climate change and increasing urbanization, have prompted researchers to study catastrophic phenomena for which large-scale mitigation strategies are needed. However, lower impact flood events occur more frequently, and although these floods are relatively less destructive, they can be effectively mitigated with local protection measures. This work focuses on understanding the factors that influence the reliable modeling of these lower impact - and less studied - flood events. We propose a procedure that combines the use of a simple but effective numerical model of the last segment of Ticino River with a Global Sensitivity Analysis for an in-depth evaluation of: i) the backwater-induced flooding of an urban area in the historical city of Pavia (Northern Italy) and ii) the potential flood damage due to such events. A 1D, unsteady HEC-RAS model (termed the 3R Model) was set up and validated against historical floods. Then, the 3R Model output was compared with a hybrid 1D/2D model of the same area. Both model schemes provide reliable results and similar predictions of the flood depth and its spatial extent. The 3R Model is demonstrated to be a successful modeling approach, providing a computational cost that is always less than one-tenth of the cost of the hybrid 1D/2D for the simulated floods. The Morris method of sensitivity analysis was performed to identify the most influential factors affecting the results of the 3R Model to help inform local policies aimed at flood risk reduction. The Morris results highlight that the maximum water level of the Po river has a dominant role in determining the flood magnitude for the whole area, with an influence at the control points in the floodplain that is more than triple that of the Ticino maximum flow rate. A Damage Index quantifying the housing loss due to the urban flood extent was introduced and analyzed using the Morris method. The results show that the predicted flood damage is strongly influenced by the uncertainty in the roughness coefficient of the floodplains. The contemporaneity of the floods on the Ticino and Po Rivers was also influential, playing a key role in flood severity within the urban area. A 5-hour delay of the Ticino flood wave (maximum flow rate of 2000 m3/s) with respect to that of the Po River (maximum stage level of 59.8 m a.s.l.) reduces the water level in the flooded area by 8 cm. These results will help the development of flood risk reduction strategies.