Micro-zonation-based Flood Risk Assessment in Urbanized Floodplain

Proceedings of Second annual IIASA-DPRI forum on Integrated Disaster Risk Management June 31- August 4 Laxenburg, Austria Micro-zonation-based Flood Risk Assessment in Urbanized Floodplain Tomoharu HORI Department of Civil Engineering Systems, Kyoto University Ji-quan Zhang Foreign Collaborative Researcher of DPRI, Kyoto University Hirokazu TATANO Disaster Prevention Research Institute, Kyoto University Norio OKADA Disaster Prevention Research Institute, Kyoto University Shuichi IIKEBUCHI 1. Introduction Risk assessment of natural disasters is defined as the assessment on both the probability of natural disaster occurrence and the degree of danger caused by natural disasters. It can be said that natural disasters result from the interaction between physical impact (hazard) and human and environmental vulnerability. The risk associated with flood disaster for any region is a product of both the region s exposure to the hazard (natural event) and the vulnerability of objects (society) to the hazard. It suggests that three main factors contribute to a region s flood disaster risk: hazard, exposure, and vulnerability. Traditional flood protection scheme in Japan focused on how to control the hazardous natural event: the main interest was to keep flood flow in river channels. Only a little attention has been paid to the case of greater flood than a design scale and failure of flood protection facilities. Off course a great deal of effort devoted to the construction work associated with this scheme has reduced the flood risk significantly. There still remains not a little flood risk in highly urbanized lowland, which was shown by flooding in Nagoya in 2000 and by flooding in Fukuoka underground shopping area in 1999. While flood control facilities such as dams and dikes reduce the frequency of hazard, they call in more properties to potentially dangerous floodplain. Reduction of the frequency of hazardous event also brings people s overconfidence of safety. They may increase the vulnerability of mega-cities to the flood. It is very important, especially in mega-cities, to take measures to avoid the situation where a large-scale flood event does not lead to catastrophic loss directly. Since it is impossible to keep all the possible sizes of floods into river channel, countermeasures taken in floodplain to control inundation water should be considered. From this viewpoint, this study aims to develop a designing scheme of countermeasures taken in floodplain to avoid the catastrophic damage. Taking up the floodplain as a target field instead of a river channel, micro-zonation concept should be incorporated where an entire floodplain considered is divided into numerous unit areas and several risk measures associated with flood loss are computed through two-dimensional inundation analyses. Inundation depth, flow velocity, duration, and time to inundation can be the risk measures at each zone set on a target floodplain. Adopting these measures makes it possible to express the condition required in protection plan making. In case of human loss reduction, the constraint can be expressed by combination of allowable threshold values of the measures. Cost minimization problem under the condition that values of the risk measures at all the zones in floodplain must be lower than the allowable levels can be defined to avoid the catastrophic human loss. The solution will guarantee the chance of evacuation. In terms of direct economic loss, maximum depth of inundation water, inundation duration, 120

and maximum flow velocity mainly contribute the vulnerability of each zone. In this present work, depth-damage functions are used to estimate the vulnerability of each zone. 2. Simulation-based Stochastic Optimization of Integrated Flood Risk Management Measures Flood risk management measures can be classified into two categories: one is risk control and the other is risk finance. Risk control comprises of risk avoidance such as land-use regulation and mitigation mainly by the construction of facilities. One of the typical risk transfer is the installation of flood insurance. Historically in Japan, mitigation by the construction of dams, dikes and retarding ponds has been emphasized as the main countermeasures against flood disasters and actually a lot of effort has been devoted to design and construct those facilities. Mitigation by facilities has great effect to reduce the frequency of inundation in lowland area but cannot be perfect for managing flood risk. There still remains the possibility of larger floods than the design scale. We cannot entirely neglect the possibility of wrong operation of dam reservoirs or failure of facilities. Going too much only by mitigation measures bring the concentration of properties and population to protected lowlands, which leads to the increase of vulnerability of urbanized area. In order to cope with the increasing vulnerability especially in urbanized area, it is important to take other types of countermeasure into account: mitigation measures taken in flood plain and risk finance measures. Since it costs too much to keep large floods with low-frequency into river channels, mitigation measures should be considered not only in river channels but also in protected lowland. In highly urbanized and thus vulnerable area, mitigation measures will be taken mainly for assuring the possibility of emergency work such as evacuation and rescue in catastrophic floods, while in relatively rural areas they may be taken to reduce the loss also. Risk finance is necessary to give the people to recover from the damage. Both of the floodplain-oriented mitigation and flood insurance designs need the detailed risk assessment in floodplain. And taking the inherent uncertainties in flood phenomena into account is indispensable. Uncertainties in the occurrence of severe rainfall, temporal and spatial distribution of the rainfall, field conditions of catchments and channels bring the significant Scenario Generator Structural & non-structual alternative of counter measures Stochastic Optimization: Stochastic Quasi Gradient Method Simulation Flood Disaster Performance criteria safety of human lives, sustainability of insurance systems,welfare of citizens, Figure 1 Simulation-based stochastic optimization of flood management measures 121

differences in flood phenomena. In order to estimate the loss brought by floods and effect of mitigation measures, it is necessary to simulate the dynamics of floodwater in floodplains. Therefore we introduce the simulation-based stochastic optimization framework for the design of countermeasures as shown in Figure 1 Scenario generator produces the external force which causes the flood loss. With an alternative of countermeasures performance measures or risk measures are estimated by inundation simulation. Inundation analysis model enables to assess the risk and/or performance measures at each point in floodplain. Since huge amount of computation is needed to estimate the performance measures under various scenarios, stochastic quasi-gradient method can be introduced to reduce the considerable computational burden. Therefore in this study, some basic simulation models based on micro-zonation of lowland area are developed in the following sections. 3. Inundation Risk Assessment based on Micro-zonation of Lowland Areas To estimate the flood risk at each point of lowland area is crucial to the design of risk finance framework and mitigation measures taken in a floodplain. Depth of floodwater is a key information about the flood risk and velocity of flood flow is also important to consider the possibility of evacuation and building loss. Dynamics of inundation flow on lowland area can be modeled as a shallow flow where the velocity distribution are integrated over the vertical direction. The governing equations can be expressed as follows: h M N + + = 0 t x y M ( um) ( vm) H τ bx + + = gh t x y x ρ N ( un) ( vn) H τ by + + = gh t x y y ρ where h is the water depth, M, N are fluxes in X and Y direction, uv, mean flow velocities, g : acceleration of gravity, τ b : sheer stress, H = h+ z, z : ground elevation. Since the calculus of finite differences is necessary to integrate these equations numerically, micro-zonation of flood area is introduced. The mesh sizes should be determined based on the availability of land elevation data and property distribution data. The availability of digital data is important since the use of too fine grids requires additional work for making digital data. In Japan, digital elevation map with 50m grids is now available. So in this preliminary computation, the grid size of 50m is adopted. Figure 2 shows the simulated flood flow after the dike break in Shinkawa-river flood plain in Nagoya, Japan, which was struck by severe flooding in 1999. Once the flow dynamics is simulated, risk measures can be computed like Fig. 3. Since only the digital elevation data is used for producing grids covering the floodplain, the effect of street network and large buildings is not well expressed in this preliminary application. It is preferable to incorporate non-homogenous grids to express the effect of structures on floodwater dynamics, which will be done in the near future. 4. Micro Model Simulation of Flood Evacuation One of the important things in the design of countermeasures to save human life is to know people s reaction to flooding. Evacuation should be considered to be one of the final countermeasures but is hard to be well managed. Actually it is often reported by many field 122

investigations that not a few people loses the chance to evacuate because of their late decision. On the other hand, in some cases people decided to evacuate without any information from the authorities, which proved to be a correct decision. In order to assess directly the human loss, it is necessary to simulate the people s behavior with mental decision process in flood situation. Since there have been many field and questionnaire researches of evacuation behavior and related mental aspect. it is important to take peoples behavior pattern and attitude in flood disasters into a simulation model. From this viewpoint, a microscopic flood refuge action model developed by the authors can be used to estimate the human behavior and the resulting loss. The model incorporates three classes of determining factors of people s decision and action process in flood situation. They are: 1. Initial factors of evacuation such as their flood experiences, attitude to floods, and daily life pattern, 2. External factors such as rainfall, inundation conditions and information provided, and 3. Mental factors which determine the way of decision making and response to the external factors. Initial factors related to flood experience is expresses by one numerical parameter, vi, which we call out look on disasters. The parameter vi takes a number from 0 to 1, and the value of 0 means that family i feels no danger on the flood while 1 means extremely dangerous. Each family s recognition about current situation during flood is expressed by the parameter, di (t ), (a) 5 minutes after dike break (b) 15 minutes after dike break (c) 30 minutes after dike break (d) 60 minutes after dike break Figure 2 Example of flood inundation simulation 123

which we call danger recognition rate, and it is calculated at each time step as follows: d () t = v p() t i i i where pi () t is the i -th family s subjective probability that the event expressed by vi. The value of pi () t varies at every time step according to the inundation conditions and information provided, v i while remains constant during one flood. Thus the danger recognition level of each family changes according to its outlook on flood disasters, i.e. its usual attitude on flood disasters. refuge oreder is issued at 8:30 60 50 40 on the way 30 complete 20 10 0 time 19:00 19:30 20:00 20:30 21:00 21:30 number of househ number of househo refuge order is issued at 8:00 60 50 40 30 on the way 20 complete 10 0 time 19:00 19:30 20:00 20:30 21:00 21:30 (a) Order is issued at 8:30 (b) Order is issued at 8:00 re fu ge o rde r is issued at 7:30 number of household time 60 50 40 30 20 10 0 on the way com plete 19:00 19:30 20:00 20:30 21:00 21:30 (c) Order is issued at 8:00 Figure 3 A result of micro model simulation of flood evacuation The decision process on flood evacuation of the i -th family s at time t is as follows: 1. Change of pi () t is computed with the fuzzy inferenc e rules, conditional rule of which contains rainfall, inundation level at time t, and the increase of inundation depth within a time step; 2. The change of pi () t computed above is modified based on the information and reliance of the information; 3. Danger recognition rate at time is computed by the above equation; then t 4. An action at time t is determined by the combination of the value of danger recognition rate and triggering information. Figure 2 shows one of the results of evacuation simulation in the flood which struck Nagasaki city in Japan in 1982. Three figures show the difference of people s action caused by the difference of issuing time of evacuation order. 5. Concluding Remarks 124

In this study, we have shown the framework of a simulation-based stochastic optimization of integrated flood risk management measures. The framework is based on distributed risk assessment by micro-zonation of lowland area. The flood inundation model developed here enables to assess inundation water depth, velocity in each zone in protected lowland and combined use of flood inundation model and evacuation model will help us to estimate human loss. Economic loss will be computed by the combined use of flood inundation model and loss function, which is now being developed. The application of optimization will be studied in near future. 125