الفهرس 
SUMMARYOnly 14 pages are availabe for public view
 |  | from | 170 |  |  |
| | | from | 170 | | |
Abstract
”This thesis contains four main components. The first one addresses the combination of building information modeling with the assessment of environmental flood hazards to evaluate the susceptibility of structures to flash floods. The next segment includes a comparative investigation involving various machine learning techniques utilized to generate the flood susceptibility map (FSM). The third component aims to determine the flash flood susceptibility by utilizing a combination of hydrodynamic modeling and machine learning algorithm. The final element aims to develop a proactive measure to mitigate the consequence of flash flood in urban region. In the first part of this study, the main goal is to develop new method to measure the building vulnerability index (BVI), which is used to figure out how buildings are prone to flash floods (Ibaraki prefecture, Japan was chosen as a case study). There are two parts to the BVI: intrinsic vulnerability (IV) and environmental flood hazard (EFH). This study uses machine learning methods, especially an artificial neural network (ANN) with a multilayer perceptron architecture that works as both classifier and regressor. The EFH classification divides 40,521 buildings into five levels of hazard, with 90% of them being in classes 2 and 3. The next part of the study uses four different machine learning (ML) methods for generating flood susceptibility maps (FSMs) in Ibaraki prefecture, Japan. It compares and evaluates these methods. The selected machine learning methods are ANN-MLP regression, support vector regression (SVR), gradient boosting regression (GBR), and least absolute shrinkage and selection operator (LASSO). The models were trained using 70% of the data, and the other 30% was kept for validation using the receiver operating characteristics (ROC) curve. The results show that the performance of ANN-MLP regression and SVR models are higher than the other adopted models.Utilizing a machine learning approach, the third part focuses on finding areas in New Cairo, Egypt that are more likely to experience flash floods. A digital elevation model (DEM) has been utilized to calculate and show the flow accumulation, basins, and subbasins. Moreover, twelve environmental factors were used, including simulated runoff. Flood Susceptibility Map (FSM) is generated using the Genetic Algorithm for Rule-set Prediction (GARP) method. According to the FSM, about 20% of the basin area is classified as a very high-hazard category. In the last part of the thesis, the effects of flash floods in Fifth District, Egypt are investigated to create a mitigation measure. The evaluation is conducted at both the community and building levels, with the goal of finding the levels of hazard that come with the expected hydrographs. Ultimately, the effects of flash floods have been reduced after using 32 underground tanks. This research demonstrates a variety of original contributions. Firstly, it combines environmental flood hazard with building attributes to assess the vulnerability of structures to flash floods. In addition, the study presents a unique method that considers simulated runoff depth as a contributing factor in the creation of flood susceptibility maps. This approach enhances the overall evaluation of flood risk by providing a more comprehensive assessment. Finally, the research implemented mitigation measures to overcome the adverse impacts of flash floods using underground tanks.”