Published works

Natural disaster prevention activities are attracting greater priority since prevention is more cost-effective and less uncertain than response, and aligned with the vision and mission of sustainable development. Increasing the resilience of communities and businesses is dependent on the extension of structural and non-structural risk mitigation activities. Hence, the nation-wide frameworks of natural disaster risk management are promoting a global movement from reactive activities (response and recovery) to proactive actions (prevention and mitigation). In Australia, flood risk management is of high priority since flood is a frequent natural hazard with significant financial consequences. Flood risk assessment and flood damage estimation are the primary steps in the flood risk management process because they are essential for the identification and prioritisation of top priority areas, cost-benefit analysis, checking the feasibility of risk mitigation options, selecting best practices in risk reduction and land use planning. This research aims to develop a validated flood damage assessment framework for the geographical area of Australia using historical data collected in several disaster events to inform disaster management policy in support of the development of risk reduction measures. In Australia, due to a lack of empirical data, most damage models are not calibrated with real damage data, and few studies have been conducted on the validation of results. In addition, most approaches are absolute, which is quite rigid and does not easily transfer across time and space. All approaches are of the traditional type, which relies on a deterministic relationship between type or use of the properties at risk and the depth of water. Thus, the interaction of most damage-influencing parameters and the uncertainty of data are neglected. This study has attempted to address these issues and the knowledge gaps.

Firstly, a comprehensive empirical data set including information on damage extent, flood impact variables and resistance factors was collected, and data mining, data preparation and data transformation were conducted. Since the function approach is a common and internationally accepted methodology for estimating the value of flood losses, some new relative multi-parameter flood damage assessment functions were derived, calibrated and validated for the most common residential and commercial building types in Australia. The functions were developed using the bootstrapping approach and considered the inherent uncertainty in the data sample. The performance of the new flood loss functions, in comparison to the empirical data, was contrasted with that of well-known flood damage assessment models from overseas and Australia. The new model was then transferred to a study area in Italy to check the ease of using local empirical data, evaluating the accuracy of the outcome, and assessing the ability to change parameters based on building practices across the world. Flood damage assessment is a complicated process and can be dependent on a variety of parameters which are not considered in stage-damage functions. Accordingly, a treebased model was developed for exploring the interaction, importance and influence of other damage-influencing parameters on the extent of losses. Finally, the candidate has explored the predictive performance of the new approaches (i.e. flood loss functions and tree-based flood loss models) in assessing the extent of physical damages after temporal and spatial transfer. The predictive power of these models was tested for precision, variation and reliability, and was also checked for some sub-classes of water depth and some groups of building type. The advantages of the newly derived stage-damage functions compared to the existing Australian models include: calibration with empirical data, greater accuracy in results, a better level of transferability in time and space, consideration of the epistemic uncertainty of data, transparency of the logic behind the model and the ability to change parameters based on building practices across the world. Furthermore, results of the tree-based analysis showed that while water depth is the most significant damage predictor in the area of study, floor space, private precautionary measures, building value and building quality also correlate with the extent of flood losses.

Also, the tree based models are shown to be more accurate than the stage-damage function. Thus, considering more parameters and taking advantage of tree-based models are recommended. Finally, it has been shown that considering more details of the damaging process can be useful for enhancing the level of transferability of damage models in time and/or space. Overall, this thesis presents a significant contribution to the flood damage assessment process by offering a calibrated and validated flood loss estimation framework. The results provide the input data for subsequent damage reduction, vulnerability mitigation and disaster risk reduction.