Floods are one of the most destructive natural hazards and have severe social and economic impacts. Therefore, improvements are urgently required in the operational flood risk management, in particular, the qualitative assessment of existing flood forecasting and early warning systems. With the objective of improving flood forecasting, I have divided this dissertation into four research topics; (i) inundation forecast validation, (ii) real-time flood inundation forecasting, (iii) uncertainty quantification of forecasting models, and (iv) communication of uncertainties.
In operational flood forecasting, hydrological models are used to forecast only discharges at specific gauges. However, to determine the distributed predictions of flood hazards, hydrodynamic models should be used to generate high-resolution spatial-temporal flood inundation maps. These maps depict inundated areas for floods above certain exceedance levels, which improves flood risk assessment by enhancing civil security and urban development. Despite the importance of flood inundation maps, flood management agencies do not generate them in real-time since unsteady hydrodynamic models are data-rich and computationally too expensive to run. To tackle this issue, I have developed a framework for real-time flood forecasting based on a pre-recorded scenario database. The framework overcomes the high computational time required by hydrodynamic models to provide dynamic inundation maps. It consists of 180 scenarios and uses real-time discharge forecasts as an input to generate inundation maps for a lead-time of 12 hours.
To evaluate the accuracy and predictive capabilities of the inundation forecasts, their validation is essential, and it is important to build trust by reducing false alarms with the help of validation data. However, spatial and temporal flood validation data is scarce in urban areas. Fortunately, recent technological developments have led to new genres of data sources, such as images and videos from smartphones and CCTV cameras. I have presented a new methodology that employs this validation data in a flood forecasting framework in order to improve the forecasting and to establish a communication from crowd-source back to the inundation forecasts. The results show that with the use of validation data, the number of false alarms, as well as the equifinality in the model parameters, can be reduced significantly.
Furthermore, incorporating uncertainties into flood risk assessment has received increasing attention over the last two decades. However, the uncertainties are often not reported due to the lack of best practices and too wide uncertainty bounds. In this dissertation, I have reviewed and quantified major sources of uncertainty with a focus on flood inundation forecasting. I developed a method to constrain the hydrodynamic model roughness based on measured water levels and to reduce the uncertainty bounds of a two-dimensional hydrodynamic model. The results show that the uncertainties in the flood forecasting models are significant and can have a major impact on the prediction of the extent of inundation. This information is vital for decision-makers in order to optimise early warning and evacuation planning.
In addition, the effective communication of the quantified uncertainties to decision-makers is a challenging issue. In operational flood risk management, the impact of flooding is assessed using a single best-model, which might misrepresent uncertainties in the modelling process. I have developed a novel methodology, which assesses the impact of flooding using a multi-model combination by incorporating buildings to develop hazard maps based on exceedance probability scenarios. These maps take into account underlying uncertainties and are ready-to-use for decision-makers for a variety of purposes, such as flood impact assessment, spatial planning, early warning and emergency planning.
This dissertation presents a prototype framework for flood inundation forecasting by combining the four research topics. The framework incorporates underlying uncertainties and communicates them to the decision-makers in the field of flood risk management. The major advantage of the framework is that it is independent from the choice of forecasting models, thus, making it suitable for use in other study areas, regardless of their size.
«
Floods are one of the most destructive natural hazards and have severe social and economic impacts. Therefore, improvements are urgently required in the operational flood risk management, in particular, the qualitative assessment of existing flood forecasting and early warning systems. With the objective of improving flood forecasting, I have divided this dissertation into four research topics; (i) inundation forecast validation, (ii) real-time flood inundation forecasting, (iii) uncertainty qu...
»