Probabilistic risk assessment (PRA) has been used in various engineering and technological fields to assist regulatory agencies, and decision-makers to assess and reduce the risks inherent in complex systems. PRA allows decision-makers to make risk-informed choices rather than simply relying on traditional deterministic flood analyses (e.g., a Probable Maximum Flood) and therefore supports good engineering design practice. Type and quantity of available data is often a key factor in PRA at an early stage for determining the best methodology. However, implementation of PRA becomes difficult and challenging since probability distributions need to be derived to describe the variable states. Flood protection is one of the rare fields in civil engineering where probability is extensively used to describe uncertainty and where the concept of failure risk is explicitly part of the design. The concept of return period is taught in all civil engineering classes throughout the world, and most cities in the developed world have developed flood risk maps where the limits of the 50-year or 100-year flood are shown. While this approach is useful, it has several limitations:
• It is based on a single flow value while all flow ranges contribute to the risk;
• It is not linked to the actual economic damage of floods;
• So far, flood risk maps only account for river water levels. It has been demonstrated that intense rainfall causes significant property damages in West Africa.
This study aimed to explore the possibility of developing and implementing a probabilistic flood risk estimation framework where all flow ranges are accounted for: 1) The probability of flood occurrence and the probabilistic distribution of hydraulic parameters, and 2) The probability of damages are spatially calculated in order for the decision-makers to take optimal adaptation decisions (e.g., flood protection dike design, recommendations for new buildings, etc.). In this study the challenges of inferring the probability distribution of different physical flood parameters in a context of sparse data, of linking their parameters to flood damages, and finally the translation of the estimation risk into decision were explored. The effect of the choice of the one-dimensional (1-D) or two-dimensional (2-D) hydraulic models on the estimated flood risk and ultimately on the adaptation decisions was investigated.
A first case study on the city of Niamey (Niger, West Africa), was performed using readily available data and 1-D and 2-D HEC-RAS (Hydrologic Engineering Center-River Analysis System) models. Adaptation options to flood risk in Niamey area were examined by looking at two main variables: a) Buildings’ material (CAS: Informal constructions – a mixture of sundried clay and straw, also known as Banco, BAN: Mud walls, DUR: Concrete walls, and SDU: Mud walls covered by mortar); and b) Dike height within a scenario-based framework, where numerical modelling was undertaken to quantify the inundated area. The 1-D and 2-D hydraulic models, HEC-RAS, were tested on a 160 km reach of the Niger River. Using the numerical modelling, water levels within the inundated areas have been identified. The extent of residential areas as well as exposed assets (polygons and building material) associated with each scenario have been evaluated. 1000 probabilistic flood maps were generated and considered in the estimation of the total loss. Benefits and costs of different adaptation options were then compared for residential land-use class in order to implement flood risk maps in the city of Niamey. Results show the individual as well as the combined impact of the two abovementioned variables in flood risk estimation in Niamey region. Dike heights ranged from 180.5 m to 184 m, at a 0.5 m interval, and buildings’ material were considered to be of 0% to 100% of each type, respectively. The results enable decision-makers as well as the regulators to have a quantitative tool for choosing the best preventive measures to alleviate the adverse impacts arising from flood. Also, because of the lack of detailed information on the exposed infrastructure elements in the study area, a feasible yet fast and precise method of extracting buildings from high-resolution aerial images was investigated using an Artificial Intelligence (AI) method – Deep Learning (DL). The applied deep learning method showed promising results with high accuracy rate for the area of interest in this study and was able to successfully identify two introduced classes of Building and Background (non-building). The findings contend that although the proposed structural adaptation options, as a resisting to environment approach, are applied to the area of interest and considered to be technically feasible, other non-structural measures, which have long-term effect of risk mitigation, should be taken into consideration, especially for highly hazard-prone areas. The results of this study would significantly help in loss estimation to the buildings due to the yearly floods in the region of interest, Niamey, Niger. However, since the buildings are of various type of material, having an accurate building database has a great importance in assessing the expected level of damage in the inundated areas, especially to the critical buildings (hospitals, schools, research labs, etc.) in the area.
| Identifer | oai:union.ndltd.org:uottawa.ca/oai:ruor.uottawa.ca:10393/43015 |
| Date | 13 December 2021 |
| Creators | Kheradmand, Saeideh |
| Contributors | Seidou, Ousmane |
| Publisher | Université d'Ottawa / University of Ottawa |
| Source Sets | Université d’Ottawa |
| Language | English |
| Detected Language | English |
| Type | Thesis |
| Format | application/pdf |
Page generated in 0.0258 seconds