C.M. Fontanazza

Bayesian inference analysis of the uncertainty linked to the evaluation of potential flood damage in urban areas.

Flood damage in urbanized watersheds may be assessed by combining the flood depth-damage curves and the outputs of urban flood models. The complexity of the physical processes that must be simulated and the limited amount of data available for model calibration may lead to high uncertainty in the model results and consequently in damage estimation. Moreover depth-damage functions are usually affected by significant uncertainty related to the collected data and to the simplified structure of the regression law that is used. The present paper carries out the analysis of the uncertainty connected to the flood damage estimate obtained combining the use of hydraulic models and depth-damage curves. A Bayesian inference analysis was proposed along with a probabilistic approach for the parameters estimating. The analysis demonstrated that the Bayesian approach is very effective considering that the available databases are usually short.

Analysis Of Intermittent Supply Systems InWater Scarcity Conditions And Evaluation Of TheResource Distribution Equity Indices

Generally, urban water distribution shortage situations are solved by introducing discontinuous service and rationing the available water resources. This approach is widely adopted, not only in developing countries but also in developed ones, for solving short term scarcity conditions which can be caused by unpredicted drought periods. Intermittent distribution has the advantage of requiring small financial efforts but it leads to network operating conditions that are very far from the usual design ones. With the aim to analyse and describe the water supply network behaviour in intermittent conditions, a network hydraulic model has been set up in which both user and manager dependent regulation structures have been schematised (pumps, private reservoirs, etc.). The analysis allowed evaluating network performance introducing several control strategies so suggesting operational plans for reducing the impact of water scarcity events on population and improving resources distribution equity. The presented model has been applied to the water distribution network of Palermo (Italy).

Identification of the best flood retrofitting scenario in an urban watershed by means of a Bayesian Decision Network

Urban resilience to floods can be defined as a city’s capacity to avoid damage through the implementation of structural and non-structural measures, to reduce damage in the case of a flood that exceeds a desired threshold, to recover quickly to the same or an equivalent state, and to adapt to an uncertain future. To build flood resilience, planners need to identify and analyse risk, to understand the impacts of flooding, and how they cope with these impacts by means of innovative and adaptable strategies and measures. The number of possible retrofitting scenarios to cope with flooding problems in an urban watershed could be greatly increased by the combination of several stormwater management practices. Therefore, the present study aims to develop an expert system in the form of a Bayesian Decision Network (BDN) able to evaluate the efficiency of some possible urban flood retrofitting scenario by examining all significant water management variables and their inherent uncertainty. The methodology was applied to an urbanized area of the city of Palermo (Italy).

Flood frequency analysis for an urban watershed: comparison between several statistical methodologies simulating synthetic rainfall events

To obtain the flooding frequency distribution for an urban watershed, different methods based on simulations of synthetic rainfall events were compared with an empirical analysis of the flooding data and with the results of long-term simulations. A copula-based multivariate statistical analysis of the main hydrological variables was proposed to generate synthetic hyetographs. Two different approaches were adopted to assess a temporal pattern to the synthetic rainfall: one analyses all available historical rainfall patterns, and another adopts the cluster analysis in three different variants to reduce the computational effort of the analysis. To test the methodology reliability, the analysis was carried out for a real urban watershed. To carry out the flooding frequency analysis, all generated synthetic hyetographs were used as input of a dual drainage mathematical model of the analysed drainage system. Results showed that the method based on the analysis of all historical rainfall patterns was efficient in the estimation of flooding frequency, especially for higher return periods and approximately halved the computational costs of ordinary long-term analysis. Regarding clustering approaches, although attractive for their computational efficiency, their adoption must be carefully evaluated because it could neglect relevant information and result in a less accurate flood frequency analysis.