Flood Damage Model Case Study Results

Transcription

1 Work Package: Document Name: WP3 Impact Assessment Flood Damage Model Case Study Results Date: 11 August 2014 Report Number: D3.4 Revision Number: 2 Deliverable Number: D3.4 Due date for deliverable: 31 December 2011 Actual submission date: 15 August 2014 Author: University of Exeter ************************* CORFU is co-funded by the European Community Seventh Framework Programme. CORFU is a Collaborative Project in the FP7 Environment Programme Start date April 2010, duration 4 Years. ****************************** Document Dissemination: PU (Publically disseminated) Co-ordinator: University of Exeter, United Kingdom Project Contract No: Project website: ACKNOWLEDGEMENTS The work described in this publication was supported by the European Community s Seventh Framework Programme through the grant to the budget of CORFU Collaborative Research on Flood Resilience in Urban Areas, Contract DISCLAIMER This document reflects only the authors views and not those of the European Community. The work may rely on data from sources external to the CORFU consortium. Members of the Consortium do not accept liability for loss or damage suffered by any third party as a result of errors or inaccuracies in such data. The information in this document is provided as is and no guarantee or warranty is given that the information is fit for any particular purpose. The user thereof uses the information at its sole risk and neither the European Commission nor any member of the CORFU Consortium is liable for any use that may be made of the information. CORFU Consortium

2 DOCUMENT INFORMATION: Title Lead Authors Contributors Distribution Flood Damage Model Case Study Results Michael Hammond, Albert Chen, Slobodan Djordjevic David Butler, Marc Velasco, Beniamino Russo, David Khan, Jelena Batica, Kapil Gupta, Yuwen Zhou, Natasa Manojlovic, Ming-Hsi Hsu PU - Public Document Reference WP3/D3.4 DOCUMENT HISTORY: Date Revision Prepared by Organisation Approved by Notes 01 July MJH UoE SDj 11 August MJH UoE SDj 2

3 SUMMARY This deliverable presents the flood damage model case study results for the current / baseline situation. The flood damage model developed and described in D3.3 is used to estimate the expected annual damage (EAD) using a range of simulated flood events in the case study cities. The differences in geographical, climate, cultural and socio-economic conditions will provide a wide range for model parameter settings. Historic data is used to calibrate and verify the extended flood damage assessment model. The influences of climate change and urban growth scenarios on flood damages for the case studies are examined to identify hot-spots that are vulnerable to flooding for given scenarios. The analysis of results will provide feedback to WP1 and WP2 for adjusting urban growth prediction and setting up flood resilience measures. This deliverable requires inputs from: Related deliverables Deliverable 2.4 flood hazard modelling results for present situation Deliverable 3.3 framework for flood damage assessment This deliverable provides inputs for: Deliverable 4.5 flood damage results in present case 3

4 Contents 1 Introduction Case study Barcelona Case study area overview Scenarios Hydraulic modelling Damage / impact modelling Direct tangible impacts Indirect tangible impacts Intangible impacts Discussion and conclusions References for the Barcelona case study Case study Beijing Case study area overview Scenarios Hydraulic modelling Damage / impact modelling Traffic impact modelling Discussion and conclusions Case study Dhaka Case study area overview Scenarios Hydraulic modelling

5 4.4 Damage / impact modelling Direct tangible impacts Development of Damage Functions for Dhaka city Results from the damage or impact assessment Discussion and conclusions Case study Hamburg Case study area overview Scenarios Case study focus Case study area Wilhelmsburg Case study area Wandse Hydraulic modelling - Wilhelmsburg Hydraulic modelling approach and tool Hydraulic modelling runs and results Damage / impact modelling - Wilhemsburg Direct tangible impacts Results from the damage or impact assessment - Wilhelmsburg Hydraulic modelling - Wandse Hydraulic modelling approach and tool Hydraulic modelling calibration and verification Damage / impact modelling - Wandse Direct tangible impacts Results from the damage or impact assessment - Wandse Discussion and conclusions References Case study -Mumbai Case study area overview Scenarios Hydraulic modelling Damage/ impact modelling Direct tangible impacts Indirect tangible impacts Intangible impacts

6 6.4.1 Results from the damage or impact assessment Discussion and conclusions References for the Mumbai case study Case study Nice Case study area overview Flood related problems in the case study Scenarios Hydraulic modelling Hydraulic modelling approach and tool Hydraulic modelling data Hydraulic modelling calibration and verification Hydraulic modelling runs and results Damage / impact modelling Direct tangible impacts Indirect tangible impacts Intangible impacts Results from the damage or impact assessment Discussion and conclusions References Taipei case study Case study area overview Hydraulic modelling Damage / impact modelling Tangible damage assessment Intangible damage assessment Damage assessment results Tangible damage assessment Intangible damage assessment Discussion and conclusions Concluding remarks

7 1 Introduction This report will present the results of the damage or impact assessments undertaken in each of the case study cities. These assessments reflect the current state or existing situation, and these results serve as a benchmark from which future scenarios and the corresponding flood impacts can be compared. In this report, authors from each of the case study cities will describe the main steps followed in the flood damage assessment, using the flood damage model. This will include details of the hydraulic modelling, the driving conditions leading to flooding that have been simulated, the methodology of the damage assessment, and the main results, describing any particular damage hotspots. The authors will also comment on the key assumptions that have been made, and the limitations of the assessment. The report will be updated as more work is undertaken and the methodologies are developed further. 7

8 2 Case study Barcelona 2.1 Case study area overview Barcelona with a population of 1,621,537 inhabitants within its administrative limits on a land area of km2 (15,980 inhab./km2) is located on the Northeast coast of Spain, facing the Mediterranean Sea, on a plateau limited by the mountain range of Collserola, the Llobregat river to the south-west and the Besòs river to the north east. The city benefits from a classic Mediterranean climate and occasionally suffers heavy rainfalls of great intensities generating flash flood events. The yearly average rainfall is 600 mm, but the maximum intensity in 5 min, corresponding to a return period of 10 years, is mm/h. Consequently, it is not rare that 50 % of the annual precipitation occurs during two or three rainfall events. The morphology of Barcelona presents areas close to the Collserola Mountain with high gradients (with an average of 4%) and other flat areas near to the Mediterranean Sea with lower slopes (with an average of 1%) (Figure 2-1). There are 31 catchments in the city. This morphology produces flash floods in the bottom part of the city in case of heavy storm events. Figure 2-1 Typical gradients and morphology in Barcelona. The red circle represents the location of Raval District. A specific area of Barcelona, the Raval District, was selected as case study, considering, for the whole area, a flood risk assessment based on the hazard and vulnerability evaluation. The Raval District, with almost 50,000 inhabitants in an area of 1.09 km 2 is one of the most densely populated areas in Europe (approx. 44,000 inh./km 2 ). Figure 2-2 shows the boundaries of the Raval District and Barelona. This district, located in a hollow area of the city, suffers from flooding problems when heavy storm events occur. These problems are caused by the excess of surface runoff and the poor capacity of the sewer system. Then, stormwater not conveyed into the sewer network and overflows from sewer manholes generate urban floods with low depths and high velocities. Moreover the hydrological response time of Raval District catchment is very short (less than 30 minutes). Such events produce significant hazard to the population, as well as economic damages to the buildings. 2.2 Scenarios The impact assessment described in this deliverable is calculated for a single scenario: the presentday situation. Future deliverables will provide the impact assessment results that have been undertaken assuming different future scenarios (D3.8) and different measures and strategies (D3.9). 8

9 Figure 2-2. Case study area: the Raval district of Barcelona. 2.3 Hydraulic modelling A detailed 1D/2D coupled model, simulating surface and sewer flows was developed using Infoworks ICM version 3.0 by Innovyze (2013). ICM solves the complete 2D Saint Venant equations in a finite volume semi-implicit scheme (Godunov, 1959) with a Riemann solver (Alcrudo and Mulet-Martí, 2005). The estimation of flood depth in a very accurate way is crucial for a micro scale assessment as this one. Therefore, there was a need for a coupled 1D/2D approach in order to take into account surface flows coming from upstream catchments and the interactions between the two drainage layers (known, respectively, as major system formed by streets, sidewalks, squares, etc. and minor system formed by the sewer network), as shown in Figure 2-3. Special attention was paid to the hydraulic characterization of the inlet systems (representing the interface between surface and underground flows) using experimental expressions achieved in the Technical University of Catalonia (Gómez and Russo, 2011). In order to consider surface and sewer flows coming into the Raval District from upstream catchments, an extended area was considered in the study. Only main sewers were considered for these catchments, while main and secondary networks were taken into account for Raval District. The final model considered a total area of 44 km 2 with 3874 nodes, 241 km of total pipe length and 6 major storage facilities with a total capacity of 170,000 m 3. A 2D mesh covered the whole analyzed domain with 403,822 triangles. Parks and other green areas were represented in the same 2D mesh, trough 2D infiltration zones characterized by their specific hydrological, physical and geometric parameters, while buildings were represented as void areas. 9

10 Runoff produced in the building areas was estimated considering an approximation of single nonlinear reservoir (whose routing coefficient depends on surface roughness, surface area, ground slope and catchment width) and directly conveyed into the sewer network. This goes in accordance with local practice in Barcelona, where roofs and terraces (approximately corresponding to 50% of the whole analyzed domain) are directly connected to the underground sewers. Sewer model was calibrated and validated using data regarding 4 critical rainfall events occurred in 2011 and provided by CLABSA Control Center. These data concerned 11 rain gages, 29 limnimeters and several time series related to real time control devices. Moreover, other data collected in the post events emergency reports (elaborated by policemen and firemen), and amateur videos recorded during the selected storm events were used to calibrate surface flow. Detailed information about the features of the model and, above all, the interactions between 1D and 2D layers is available in other deliverables (D2.2 and D2.4) and Russo et al. (2012 and 2013). Figure 2-3. Interaction of surface and sewer flow (dual drainage concept) (Schmitt et al. 2004) 2.4 Damage / impact modelling The flooding problems that occur in the Raval produce significant hazard for traffic and pedestrians, as well as economic damages in terms of goods and properties. Since the goal of this study is to determine the cost-effectiveness of several adaptation strategies, an accurate economic appraisal of the damages is crucial. Then, comparing the baseline scenario with the future ones (taking into account both global changes and adaptation measures), the benefits of adaptation can be quantified and the different measures prioritized. In the following sections, different types of impacts should be assessed. However, due to the area studied, only direct damages have been included. First of all, the direct tangible damages related to the buildings are assessed. Then, in an intangible way and giving several risk levels, the impacts to people and vehicles are calculated. Risk is defined as the probability or threat of a hazard occurring in a vulnerable area and that may be avoided or minimised through preventive actions. For the three different impact categories, flood 10

11 Vulnerability Project Report risk is assessed in the same way. Flood risk maps related to each specific scenarios and return period are obtained by combining hazard maps and vulnerability maps, as shown in Figure 2-4. Figure 2-4. Combination of hazard and vulnerability maps to produce a flood risk map. In the case of direct tangible damages, risk is expressed in terms of monetary values thanks to the depths damage curves. For the other two categories, risk maps are created multiplying the vulnerability index (1, 2 or 3, corresponding to low, moderate and high vulnerability) by the hazard index (1, 2 or 3, corresponding to low, moderate and high hazard). Finally the total risk varies from 1 to 9 where higher levels indicate higher risk. This methodology is summarized in the following matrix (Figure 2-5). Risk Matrix Hazard Figure 2-5. Risk matrix obtained by multiplying the vulnerability index by the hazard index Direct tangible impacts Regarding the scale of the study, Messner et al. (2007) proposed the following classification: macro, meso, and micro scales. Here, as the case study area is a city district of approximately 1 km 2, the development of a micro-scale study is needed. Hence, as opposed to other scales, data requirements are high, and some properties must be considered on an individual basis. As the flood damage assessment is an issue that has not been deeply studied in Spain, experiences from other countries (Penning Rowsell et al. 2005, Nascimiento et al. 2007, Kok et al. 2004, Reese et al. 2003) were reviewed. In general, the estimation of the damage caused by flooding focuses on the 11

12 flood depth. Therefore, depth-damage or stage-damage curves have been adopted in multiple locations around the world as the most commonly used technique to assess flood vulnerability. A vulnerability assessment of direct tangible damages at a micro-scale level will require the following three elements: stage-damage curves; flood depth maps; and land-use maps. In addition, in order to ease the calculation of the final vulnerability maps, the GIS-based toolbox developed in D3.3 will be used. Data regarding flood depths were already described in section 2.3. Following, the descriptions of depth damage curves and land-use maps used are presented Hazard levels for direct tangible impacts Figure 2-6 shows the depths generated by the three rainfall events simulated. It is worth noting that, the model outputs (i.e. water depth in the streets) have been converted into water depth inside the buildings in order to ease the calculation of the damages. Consequently, Figure 2-6 shows water only inside the buildings and not in the streets. In addition, the reader should note that water depths of less than 15 cm will cause no damage, due to the depth damage curves developed. Therefore, although the light blue colour is sometimes widespread, only the values over 15 cm will be considered for the damage model. For an event of 1 year return period, the flood problems are small and since most of them are small depths, they will not cause a lot of damages. For an event of 10 years of return period, there will be some localised problems on a rather small area of the district. When increasing the severity of the rain event, this area enlarges and the water depths also increase. Consequently, for a 100 year return period event, the model presents a generalised flood in the district, which is especially intense in the southern part. This situation can be explained considering that the drainage system of Barcelona is generally designed for a return period of 10 years. Of course, the situation presented by the flood maps in Figure 2-6 is an upper bound of the actual situation. This is because both the synthetic rains used, and the assumption of having the same depth inside the buildings than in the streets leads to an overestimation of the depths. Nevertheless, since the final goal of this study is to compare the current situation to the future ones defined by the different scenarios that will be developed, using an upper bound will imply selecting the most robust strategies to cope with flood impacts. 12

13 Figure 2-6. Flood depths in the Raval district for a rain event of return period of 1 year (left), 10 years (center) and 100 years (right) Vulnerability levels for direct tangible impacts In this case, flood vulnerability is expressed as the combination of two different things: the stagedamage curves and the land-use maps. The two of them are explained in the following sections Stage-damage curves Depending on the information available and the goals of the assessment, there are several types of depth-damage curves (Merz et al. 2010). As in Barcelona there is a lack of historical damage data, and the building typology may be remarkably different from other regions previously studied, following the recommendations of D3.3, it was decided to develop specific synthetic relative depthdamage curves for the city. The curves are used to obtain damage costs for a certain water depth relative to the extent flooded. Then, multiplying the obtained value by the affected area of the building, the damage cost at a block scale is acquired. These curves are calculated for different types of land-use. Taking into account the main uses identified in the case study area, six different categories have been defined (Figure 2-9). Then, via a what-if analysis and using flood expertise acquired from past flood events, a stage-damage curve for each type of building and economic activity has been obtained. These final stage-damage functions are composed by two independent curves. The first one, related to the re-conditioning of the building (cleaning, painting the walls, changing the floors or doors, etc.), has been obtained through the expertise gained in past local floods (with the collaboration of an expert in flooding damages appraisals for insurance companies). The second component is related to the damages to the contents, which have been developed using the FloReTo tool (Manojlovic & Pasche 2010). The several curves developed can be seen in Figure

14 Figure 2-7. Depth damage curves for the buildings (left) and content (right) taking into account the local conditions of the Raval district. The relationship between building and contents damages strongly depends on the type of land-use considered. Whereas in households the building damages tend to be higher than the contents (Thieken et al. 2005), this trend is not so clear in other land-uses. In the case of commercial use, flood losses are highly variable due to the differences depending on the kind of business considered (Gissing & Blong 2004). Since the Raval district mainly has small and simple retail shops with many goods, the curves define the content damages up to five times greater than the building ones. In order to assess the ability of the curves developed to represent the Raval case study, a survey was undertaken in the area. Taking advantage of the flood event that occurred on 31 st July 2011, a series of interviews were carried out (Velasco & Cabello 2012). Although the number of affected people was small, the answers provided very interesting information which was very useful to validate the synthetic curves previously developed. In addition, actual damage data from the Consorcio de Compensación de Seguros (CCS), the reassurance that covers the catastrophic and extreme situations in Spain, were obtained for the same flood event in the Raval district. Altogether, a two-step validation was done: a qualitative one, using the general trends and behaviours extrapolated from the surveys; and a semi-quantitative one, comparing the flood damage maps and total figures for the studied area. Some of the general outcomes of the surveys applied in the curves are: in the Raval district, damages to contents will be more important than the damages to the building itself consequently, the differences between the several pairs of curves from Figure 2-7 are justified; flood depths smaller than 10 cm do not cause any damages and hence, the damages expressed by the curves are zero until this value is reached. When comparing the actual reported damages with the simulated ones, a few differences were found. The simulated damages were larger, as well as the affected area was. That is why the threshold was increased to 15 cm, meaning that water depths smaller than this value would cause no damages. 14

15 Using the curves with this final correction (Figure 2-7), the maps from Figure 2-8 were obtained. Here, the simulated damages (left) and the actual damages reported to the CCS (right), show very similar spatial distributions. In terms of global damages, the model seems to overestimate the effective damages. Although the values are not coincident, they are within the same order of magnitude and the differences can be explained by the following reasons: Due to the methodology followed (presented later in section ), the damages per each block will be obtained by multiplying the relative cost by its area. When there are very big blocks (as some of the affected ones in Figure 2-8), which may mean that there is only one owner or that it is a cultural or public building, the damages are always overestimated. Some of the flooded buildings may have not reported their damages to the CCS because they were small, the property was not insured or they were not aware that they could be compensated. The simulated damages are assuming that no flood risk reduction strategies are being used. However, since some of the local population have suffered from previous flood events, it seems reasonable to think that some of them might have protected their assets, by placing wood gates in their doors, or moving the goods to higher areas. Figure 2-8. Spatial pattern of damages from the 31/07/2011 flood. Simulated damages (left) present very similar pattern than the actual reported damages od the CCS (right). Although the curves seem to represent accurately (despite the deviations just mentioned) the flood damage processes in the Raval district, they could still be improved. Consequently, whenever more 15

16 data is acquired, updates of the curves will be carried out to improve their capacity to represent the real situation of the area. For example, if long series of damage records are obtained, instead of validating the curves for a given event, the damages could be compared in terms of mean values. Therefore, the mean annual damage recorded could be compared to the simulated EAD Land-use information As stated previously, good quality and precision of data is crucial when carrying out a micro-scale study. Consequently, regarding land-use, a GIS map has been developed using data from the local land registry at block level (Figure 2-9 left). As the flood typology in the case study area consists in flash floods producing low water depths and high speeds, only land-uses of the ground floor and basements have been added to the dataset. For each block, more than one land-use type is possible, so the area related to the several land-use types is given. Multiplying these values by the relative damages obtained from the stage-damage curves, the total damages of the block can be obtained. Additionally, it possible to calculate the vulnerability map, which shows the maximum potential damage in monetary units (Figure 2-9 right). As it was previously mentioned, for this vulnerability assessment only six land-use types have been considered. Since the Raval district is a densely populated urban area, there are almost no green or industrial areas (and no agricultural areas at all). Therefore, the uses that have been considered are: (1) warehouses and parkings; (2) commercial; (3) residential; (4) hotels and leisure; (5) public and cultural buildings; and (6) sites of interest. This last category has been introduced because of the special importance of some of the buildings in this specific area, such as museums, churches or historical buildings with inestimable value. Since the water depth is going to be different depending on the floor level, the land-use maps present this variable into two separate maps. Moreover, from the datasets it is observed that, whereas in the ground level the land-use types are evenly distributed, in the basements there is a predominance of the warehouses and parkings category. 16

17 Figure 2-9. Land-use classes in the Raval district, the main land-use class of the ground floor is shown at a block scale (left), and vulnerability map, presenting the potential damage of the assets at risk (right) Risk levels for direct tangible impacts To integrate the damage modelling with the hydraulic model results, the GIS-based toolbox developed in the frame of WP3 (and described in D3.3) has been particularized for the case of the Raval District. This version of the toolbox, enables to automatize the three following steps, increasing the speed of the post processing of data and so, easing the simulation of several scenarios: 1. Assign a water depth to each building. 2. Interpolate this value in the stage-damage curve to obtain the relative cost. 3. Multiply the relative cost by the area, obtaining the total damage value per each block. Finally, a shape file with the total damages is obtained, being able to calculate the total costs that have been caused by the extreme rainfall event. This process can be easily repeated for several flood-driven events. Using this toolbox and the several data described previously, synthetic rain events of 1, 10 and 100 years of return period have been simulated. Then, using the EAD previously defined, an estimate of the current damages in the area has been given. Using the flood maps presented in Figure 2-6 and the methodology described earlier, flood damage maps are obtained for the studied area (Figure 2-10). As it can be seen in the right map of this figure, some buildings present extremely high damages (more than 200,000 ). 17

18 As damage is represented in each block, larger blocks will accordingly present larger damages (as the area flooded will be multiplied by the relative damage to obtain the total figure). This is something which should always be taken into account, because in general, the blocks presenting the highest damage values are also the ones with the largest areas. From Figure 2-10, the conclusion extracted from the hazard maps (Figure 2-6) is reinforced: the southern and south-western parts of the district are the most vulnerable to floods. Due to the historic characteristic of this region, some particular buildings of inestimable value are located in flood prone areas. Finally, the EAD of the whole area has been calculated. As it has been stressed before, this value is of high interest as it will be crucial to assess the cost-effectiveness of the adaptation measures proposed. Although this part of the project is still on-going, the baseline scenario to which the rest will be compared has already been determined. The authors would like to stress that the EAD calculated does not intend to determine the actual damages of the area, but expressing the benefits that could be obtained if some adaptation measures are implemented. In addition, as it was stated previously, the flood maps are an upper bound of the actual situation. Consequently, this methodology will also lead to an overestimation of the damage values. Although this may seem inappropriate, it has been selected as the best hypothesis to allow the identification of the most robust strategies to deal with the impacts of floods. Figure Flood damages in the Raval district for a rain event of return period of 1 year (left), 10 years (centre) and 100 years (right). In Figure 2-11, the damage probability curve for the Raval District is presented (for the detailed values, see Table 2-1). In it, the aggregated damages of the entire district are plotted against its probability. As explained before, the area below this curve is the EAD, which has a final value of 1,697, As mentioned, this figure is an overestimation of the annual damage that may be caused by floods in the Raval District, but it provides an estimate of the order of magnitude in which this value ranges. 18

19 Figure Damage probability curve for the whole Raval district. The area under this curve expresses the EAD of the region. The probabilities 1, 0.1 and 0.01 represent the events of 1, 10 and 100 years of return period, respectively. Table Damages and probabilities for the three synthetic rain events simulated. Return period (years) Probability Damage ( ) 78,846 1,615,738 19,156, Indirect tangible impacts The indirect tangible impacts are the ones that can be economically assessed, but which have not been created due to the direct contact with water. Such impacts are the disruption of businesses activities and transport networks, amongst others. In the Raval District, the flood typology could be defined as flash flooding, with low depths and high velocities. The impacts induced by such floods are localized, and the retention time of the water is generally very short. On the other hand, the Raval District is not crossed by any important communication route, and there are no big commercial or industrial uses within the district. Consequently, in the area studied the indirect damages will tend to be small compared to the direct ones. Therefore, such damages are not going to be included in this study Intangible impacts As it has been mentioned before, in this case impacts to pedestrians and traffic will be assessed. Since these values will be determined in terms of different risk levels, they are considered in the intangible section Impacts to pedestrian circulation Hazard levels for pedestrian circulation In order to define specific hazard criteria related to runoff in urban areas, Technical University of Catalonia (Spain) promoted a new research line based on the study of the stability of pedestrians circulating in flooded streets (Russo et al., 2013). The results of the experimental campaign, presented in terms of flow velocities able to produce critical situations for pedestrian circulation are summarized in the Table

20 Table 2-2. Hazard levels according to flow parameters in flooded streets. Hazard level Flow conditions (for flow depths between 9 and 16 cm) High v 1.88 m/s Moderate 1.51 v < 1.88 m/s Low v < 1.51 On the basis of this study, in order to elaborate the hazard maps related to pedestrian circulation for the Barcelona case study, the following high hazard criteria were defined (Figure 22): - Maximum flow depth y max = 0.1 m (corresponding to the minimum depth of the kerb of the sidewalks in Barcelona) - Maximum flow velocity v max = 1.9 m/s (corresponding to the high hazard level threshold flow velocity for flow depths between 9 and 16 cm). For the moderate hazard levels the following hazard criteria were adopted: - Maximum flow depth y max = 0.06 m (corresponding to a flooded lane 3 m wide with a transverse slope of 2%) - Maximum flow velocity v max = 1.5 m/s (corresponding to the moderate hazard level threshold flow velocity for flow depths between 9 and 16 cm). A GIS post process procedure was applied to adapt the model outputs to the pedestrian flood hazard criteria previously defined. According to it, the hazard map from Figure 2-13 was elaborated for the baseline scenario. Figure Barcelona pedestrian hazard criteria (HH: high hazard; MH: moderate hazard; LH: low hazard). 20

21 Figure Pedestrian hazard maps of the Raval District for the Baseline Scenario (year 2010). In red high hazard conditions are shown, while in yellow and green colours moderate and low hazard conditions are represented Vulnerability levels for pedestrian circulation In order to assess the human vulnerability of the Raval District, statistical data of current and forecasted population in 21 different census areas were used. These data (updated to 2012 and summarized in Table 2-3) were provided by Barcelona municipality and concern: C: Density of people with critical age: less than 15 years old and more than 65. F: Density of foreign people D: General people density B: Presence of critical buildings (such as hospitals, schools, etc.) Once data were available for the baseline, the following thresholds were defined in order to assess the vulnerability of each census area. Specifically for the human vulnerability related to the people density, thresholds were deduced from the medium density of Barcelona (16000 inhabitants per Km 2 ) and the definition of the National Institute of Statistics of urban area defined as a group of minimum 10 houses in a distance less than 200 m (equivalent to 1273 inhabitants per Km 2 ). The other defined thresholds are shown in Table 2-4. Three vulnerability indexes were defined according to the 3 first data types and for the final vulnerability index, the average value between C, D and E was computed and in case there was any critical building in the census area, a 0.5 value was added. The final vulnerability level was achieved according to the formulations proposed in the Table

22 Census area ID Census area (m 2 ) Table 2-3. Population data for the baseline scenario. Total inhabitants (2012) People density People with age less than 15 years old People with age more than 65 years old Table 2-4. Thresholds to assess human vulnerability according to different criteria. Vulnerability index C % people age < 15 F % of foreign D People density Foreign people or > 65 years old people 1 (low) 33% 33% (medium) 33% < X 50% 33% < X 50% 1273<X (high) > 50% > 50% > Table 2-5. Formulation to compute the total vulnerability index. Vulnerability level Formulation * Low (D+C+F)/3 < 1.5 Medium 1.5 < (D+C+F)/3 < 2.5 High D+C+F)/3 > 2.5 *In case there is a critical building is in the subdistrict area, 0.5 must be added to the average value of (D+C+F)/3. Applying the described methodology, human vulnerability map was obtained for the baseline scenario. In this map, the different vulnerability levels (high, moderate and low vulnerability) of census areas were represented using, respectively red, yellow and green colours (Figure 2-15). Moreover the presence of critical buildings (schools, hospitals, etc.) was also shown in the same map. 22

23 Figure Human vulnerability for the baseline scenario Risk levels for pedestrian circulation For the flood risk assessment related to pedestrian circulation in the urban areas analysed, the general methodology based on the risk matrix was implemented obtaining the risk for the baseline scenario and the selected return periods. The risk was defined for each census area and represented with the same range of colours previously described. In order to define the risk level of each census area a statistic treatment of the risk of the cells was carried out. The risk level of each census area was assumed equal to the maximum risk level of the cells, as long as they represented at least a 15% of the total cells located in the same census area. In Figure 2-15 the risk map for the baseline scenario is shown. Obviously, the risk level increases along with the return period. 23

24 Figure Risk maps for pedestrian related to Baseline scenario and Business as usual scenario for return periods of T = 1, 10 and 100 years Impacts on vehicles Hazard levels for vehicles A similar procedure to the one defined for pedestrian circulation has been suggested. Stability criteria for stationary vehicles have been defined to create vehicle hazard maps. The conclusions come from a specific report developed by Engineers Australia contracted by Water Research Laboratory: Appropriate safety criteria for vehicles (Shand et al. 2010). This report reviews and discusses previous experimental and analytical investigations of vehicle stability for stationary vehicles (Hydroplaning is therefore not considered further within this study). The two recognized hydrodynamic mechanisms by which stability is lost include buoyancy or floating and friction instability or sliding (Figure 2-16). Authors of this report finally propose several stability criteria for stationary vehicles that have been summarized in Table

25 Figure Hydrodynamic mechanism by which vehicular stability is lost. Table 2-6. Hazard stability criteria for stationary vehicles. On the basis of this study, in order to elaborate the hazard maps related to vehicles for the Barcelona case study, the following criteria were adopted (Figure 2-17): - High hazard criteria were referred to critical flow conditions concerning a vehicle class Large passenger. In this case, practically all type of vehicles loss their stability. - Moderate hazard criteria were referred to critical conditionsconcerning a vehicle class Small passenger. In this case, only small cars could loss their stability. Using these hazard criteria and the results of the models for the baseline scenario, the hazard map for vehicles was elaborated (Figure 2-18). As it can be seen, several streets (above all located in the southern part of the district) could be affected by high flood risk for the return period of 100 years. However, many of these critical streets have little traffic flow as shown in the following vulnerability section concerning traffic flow (Figure 2-19). 25

26 Figure Hazard criteria for vehicle types in Barcelona case study (HH: high hazard; MH: moderate hazard; LH: low hazard). Figure Vehicle hazard maps for the baseline scenario considering return periods of T = 1, 10 and 100 years. In red high hazard conditions are shown, while in yellow and green colours moderate and low hazard conditions are represented. 26

27 Vulnerability levels for traffic Traffic vulnerability was obtained through assessing the traffic flow data in the Raval district provided by the Traffic Department of Barcelona Municipality. Thresholds were arbitrarily decided based on the average values of the traffic flow in Barcelona streets (Table 2-7). The vulnerability map for traffic is presented in Figure Table 2-7. Vulnerability levels of vehicular circulation. Vulnerability level Vehicular flow intensity (VFI) (vehicles in 24h) Low VFI < 5000 Medium 5000 VFI High VFI > Figure Vehicular vulnerability map for the baseline scenario based on traffic intensity (shown values express vehicular flow intensities x 1000 in 24 hours) Risk levels for traffic As applied with the human risk, a matrix combining hazard and vulnerability data for vehicles was implemented for the assessment of vehicular traffic flood risk. 27

28 The objective of this step was to determine the risk level of the street on the basis on the hazard levels of each cell and the traffic flow intensity of each traffic lane. Crossing these data and using the risk matrix shown in Figure 2-5, risk maps for vehicular circulation were obtained for the baseline scenario considering the selected return periods (T = 1, 10 and 100 years). They are presented in Figure The streets with high risk for a high return period (100 years) are Ronda de San Pau Street, Parallel Avenue, Rambla del Raval and Pelayo Street (Figure 2-20). Figure Risk maps for pedestrian related to Baseline scenario and Business as usual scenario for return periods of T = 1, 10 and 100 years. 2.5 Discussion and conclusions In order to improve the capacity to represent urban floods in the Raval District, a 1D/2D coupled model has been developed. The interface between the two drainage layers has been characterized through empirical expressions related to hydraulic performance of surface drainage systems. The 2D domain covers 44 km 2 of the city land involving 235 km of sewers, while 2D mesh counts 403,925 triangles. Calibration and validation of the model is based on the data (rain gauge data, time series of flow depths recorded by water level gauges, reports and videos concerning flooded areas) related to 4 heavy storm events occurred in The obtained results show that it is possible to reproduce the effects of urban floods in the Raval District in a more realistic way than traditional 1D sewer flow simulations. 28

29 With the development of synthetic depth-damage curves regionalized for the case study, an exhaustive economic damage assessment can be carried out when heavy storm events occur. Implementing the described methodology with the GIS-based toolbox, the EAD of the Raval district can be calculated. This enables the determination of the critical points of the district in terms of flooding impacts. In addition, impacts to pedestrian and vehicular circulation have been assessed, using qualitative methodologies defining several hazard, vulnerability and risk levels. This will allow to be able to assess the benefits of the several adaptation strategies, in non-economic terms. In the coming months, following the same methodology, several adaptation measures (such as improvements of the sewer network, construction of SUDS or green-roofs, local flood mitigation strategies, etc.) will be simulated. This will allow determining the effectiveness of these strategies in terms of vulnerability reduction, so the ones presenting the best performance can be prioritized over the others. Currently, the presented data and methodology is ready to be applied, both for the present situation and for the potential future ones, taking into account socio-economic and climatic changes and the implementation of adaptation measures. Even though, this task has not yet been concluded and so, only the baseline scenario has been modelled. It is worth noting that the damages calculated are an upper bound of the actual damages of the district, because several of the assumptions that have been done are conservative. The aim is to allow determining high robust adaptation strategies that can cope with damages larger than the current ones. 2.6 References for the Barcelona case study Alcrudo F. and Mulet-Marti J. (2005). Urban inundation models based upon the Shallow Water equations. Numerical and practical issues. Proceedings of Finite Volumes for Complex Applications IV. Problems and Perspectives. Hermes Science publishing. pp 3-1. ISBN Gissing, A. & Blong, R Accounting for variability in commercial flood damage estimation. Australian Geographer, 35, (2): Godunov S. K. (1959). A Difference Scheme for Numerical Solution of Discontinuous Solution of Hydrodynamic Equations. Math. Sbornik, 47, , translated US Joint Publ. Res. Service, JPRS 7226, Gómez M. and Russo B. (2011). Methodology to estimate hydraulic efficiency of drain inlets. Proceedings of the ICE - Water Management. Institution of Civil Engineers, 164(1), Innovyze (2012). InfoWorks ICM (Integrated Catchment Modeling) v.2.5. User manual references. Kok, M., Huizinga, H.J., Vrouwenfelder, A.C.W.M. & Barendregt, A Standard Method Damage and casualties caused by flooding. Client: Highway and Hydraulic Engineering Department. 29

30 Manojlovic N., & Pasche, E Theory and Technology to Improve Stakeholder Participation in the development of Flood Resilient Cities, Proc. Int. 21st IAPS Conference on Vulnerability, Risk and Complexity: Impacts of Global Change on Human Habitats, Leipzig, Germany. Merz, B., Kreibich, H., Schwarze, R. & Thieken, A Review article: assessment of economic flood damage. Natural Hazards and Earth System Science, 10, 8, Messner, F., Penning-Rowsell, E., Green, C., Meyer, V., Tunstall, S. & Van der Veen, A Evaluating flood damages: guidance and recommendations on principles and methods. FLOODsite Project. Nascimento, N., Machado, M. L., Baptista, M. & Silva, A. D. P The assessment of damage caused by floods in the Brazilian context. Urban water journal, 4, Penning-Rowsell, E., Johnson, C., Tunstall, S., Tapsell, S., Morris, J., Chatterton, J. B. & Green, C The benefits of flood and coastal risk management: a manual of assessment techniques, Middlesex University Press. Reese, S., Markau, H.J. & Sterr, H MERK Mikroskalige Evaluation der Risiken in überflutungsgefährdeten Küstenniederungen. Abschlussbericht. Kiel. Russo B., Suñer D., Velasco M. and Djordjević S Flood hazard assessment in the Raval District of Barcelona using a 1D/2D coupled model.9th International Conference on Urban Drainage Modelling. Belgrado, Serbia. ISBN Russo B., Gómez M. And Macchione F Pedestrian hazard criteria for flooded urban areas. Natural Hazards. Springer. 63(11), DOI: /s Shand T. D., Cox R. C., Blacka M. J., Smith G. B Appropriate Safety Criteria for Vehicles. Literature Review. Stage 2 Report, Australian rainfall and runoff Project 10. Engineering Australia, Water Engineering. Thieken, A.H., Müller, M., Kreibich, H. & Merz, B Flood damage and influencing factors: New insights from the August 2002 flood in Germany. Water Resources Research, 41, (12). Velasco, M. & Cabello, A Review of the July 2011 urban flood in Barcelona. In preparation. 30

31 3 Case study Beijing 3.1 Case study area overview The Beijing case study looks at the flooding impacts at various scales. Although a city-wide hydraulic model in coarse resolution has been developed (See D2.4) for the central Beijing city, the detailed land use information are not available for the central Beijing area such that the flood damage cannot be evaluated at the city-wide scale. Instead, the new urban development area Yizhuang was used for damage assessment. Yizhuang is a satellite town in the south of Beijing. It has been quickly developed in the last two decades with population grew from 104,000 in 2001 to 700,000 in Yizhuang has attracted huge investment and the Beijing city government has provided detailed information, including terrain model, land uses and sewer network, that we can use in the CORFU project. These data are utilised in the hydraulic model, together with the current rainfall patterns, to simulate the flooding in Yizhuang. The results are combined with the land uses estimate the flood damage for the baseline scenario. For the central Beijing city, we have learnt that the traffic disruption during flooding is more critical than the tangible damage to buildings. Most of the disruptions were caused by flooding at the under bridge tunnels. The terrain and the poor drainage in those areas resulted in flooding that affected the traffic significantly when heavy rainfall occurred. Many of the under bridge tunnels are located at the junctions of trunk roads such that the ripple effect often spreads widely and stalled the traffic in the main road networks. To investigate the influence of flooding on the traffic, we built a detailed hydraulic model for the under bridge tunnels (See D2.4) to simulate the dynamic of flooding at such locations. For the impact assessment, we collected the related traffic information for Beijing and tried to develop a traffic model using the software called SUMO. Nevertheless, the SUMO software requires some essential car journeys information that were not yet available. Hence, we gathered other related information on population, car journeys, and the road network from public websites. In February 2012, the number of vehicles in Beijing reached 5 million ( and 70% of those are private vehicles. The average traffic in central Beijing (inner 6th ring road) is 30 million passengers/day in 2012 (8.3 million via buses or trams, 5.1million via subways, 9.9million by private cars, 2 million by taxis, and 4.2million by bicycles) (Beijing Transportation Research Centre, 2013, Year book of traffic development in Beijing city, 2013). With hydraulic modelling results for the central Beijing area, the flood hotspots of those under bridge tunnels will be identified. With the detailed hydraulic model of the bridge areas, the flood extents, depth and duration will be simulated such that the triggers for traffic disruption can be properly established in the SUMO model for traffic modelling. The results will be compared to the baseline model with normal traffic condition such the influence of flooding on the traffic at the citywide scale can be evaluated (in terms of extra journey time, consumption of fuel, or loss of working hours, etc.). 31

32 3.2 Scenarios The impact assessment described in this deliverable is calculated for a single scenario: the presentday situation. Future deliverables will provide the impact assessment results that have been undertaken assuming different future scenarios (D3.8) and different measures and strategies (D3.9). 3.3 Hydraulic modelling Pipe network data, ground elevation data and rainfall information are the three essential components for the hydraulic model. The drainage network data includes topology, structure of the pipes, pumping stations, relevant hydraulic facilities and other hydrological and hydraulic parameters. The surface information contains catchment processes parameters and a ground digital elevation model, for which a 10m*10m regular grid model is generated for the case study. The rainfall information is the input for 1D network modelling, combined with Beijing annual maximum storm intensity formula and 24-h hydrological hyetograph to generate design rainfall for various return periods. The hydraulic modelling results have been provided in D Damage / impact modelling Building content flood damage evaluation typically encompasses four steps: hazard analysis, bearing body exposure assessment, vulnerability analysis and loss quantification. Firstly, hazard analysis is the process of acquiring information such as indictor intensity, frequency and scope, and hydraulic model to execute scenario analysis becomes the mainstream direction. The premise for constructing urban flooding model is to gather detailed information of drainage network, river, drainage structures, ground surface and rainfall, which could be achieved by remote sensing analysis and site survey. Based on a series of hydrology and hydraulic principles, this numerical model simulates physical process of flood occurrence and evolution to obtain hazard indicators variation over time and space. Secondly, through socio-economic survey and statistics and geospatial information database, the exposure assessment takes advantage of area weighting method to derive spatial attributes of socio-economic condition, thereby reflecting spatial distribution difference in the economic indicator of bearing body. Lastly, vulnerability analysis, usually represented by depthdamage curve, relies on the typical sampling survey to establish statistical relationship between hazard and economic losses factors. In view of above statement, the flood damage assessment procedure is demonstrated in Figure

33 Remote Sensing Analysis Site Survey Hydraulic Model Socio-economic Survey Socio-economic Statistic Geospatial information Hazard Indicator (Intensity,Frequency,Scope) Exposure Assessment (Land use,value,distribution) Typical losses survey vulnerability analysis (Loss rate,absolute loss) Loss Quantification (Value,Spatial Distribution) Figure 3-1 Flood damage assessment procedure The vulnerability expression could be identified as relationships between hazard and loss rate or absolute loss value. The depth-absolute loss vulnerability curves of various types of assets have been set up in large amount of developed countries, but the loss rate is more preferred in China. In fact, these two approaches just differ slightly in expression manner but with same essence. Due to the association between the building land use type and loss rate, the following formula is generated to calculate flood damage. L L i, j i j Where L is total flood loss, is loss value of property category at depth j, is the loss rate of property category at depth. ij Deriving water depth-damage curve of buildings which fits local economic development situation is basic for implementing loss and risk evaluation. Based on synthesis assumption analysis of indoor property value, part of building depth-damage curves of Beijing has been constructed, which is based on 10 types of land use according to the flood damage database of UK (Penning Rowsell et al, 2010) as shown in Figure

34 Flood Damage(GDP per m 2 ) Project Report Residential Manufacturing and Processing Commercial Service activities Education and research Government services Recreational facilities Mixed use Transport services Parking lots Flood Depth(m) Figure 3-2 UK Depth-damage curves Building vector data were collected for the case study. Each building has a unique index value is introduced to distinguish building vector data, which could be overlain with land use maps to generate the building land use spatial distribution (Figure 3-3). Residential Manufacturing and Processing Commercial Service activities Education and research Government services Recreational facilities Mixed use Transport services Parking lots Green spaces Water body Cropland Unused area Figure 3-3 Building land use spatial distribution To deal with direct flood damage evaluation, WP 3 has developed a tool using Python script and the ArcObjects, the geoprocessing functions within the ESRI ArcGIS software, which can evaluate the damage directly from the hydraulic model results. This tool is capable of exerting flood damage calculation and risk analysis with spatial properties at single building scale (Figure 3-4 Building damage results of (a)10, (b)20, (c)50, and (d)100 year return period of rainfall event). The flood damage statistics is summarized in Table 3-1. Table 3-1 Flood damage and rainfall statistics for different return periods Return period (year) Rainfall(mm) Total Loss(hundred million RMB)

35 hundred million RMB Project Report Figure 3-4 Building damage results of (a)10, (b)20, (c)50, and (d)100 year return period of rainfall events The relationship of flood damage versus design storm frequency is shown in Figure mm Total Loss (hundred million RMB) Rainfall(mm) Frequency Figure 3-5 Relationship of flood damage versus design storm frequency Indirect tangible impacts Traffic impact modelling An assessment of indirect tangible impacts will be made by an analysis of the impacts of flooding on traffic through disruption. Traffic disruption results from reduced vehicle speeds and blockage of traffic at critical spots. This problem occurs frequently in Beijing. A serious storm hit Beijing in June 35

36 23, It disrupted the whole traffic system in BeijingFigure 3-6 presents data on the traffic condition from that event. Green represents good traffic condition, and black in completely blocked. The other colours represent different conditions, depending on the road type. The classification is presented in Table 3-2. Another serious storm hit Beijing in July It again disrupted the whole traffic system in Beijing and more than 70 people were killed by flooding. Data from that storm are presented in Figure 3-7. Impacts on traffic are represented in Figure 3-8Figure 3-8 Impacts on traffic in Beijing, July 23, 2012 Figure 3-6 Impacts on traffic and real time traffic situation in Beijing, June 23, 2011 Table 3-2 Classification for the traffic map colours Classification Grade Design Speed Traffic speed Colour (km/h) (km/h) Ⅰ 100 Green >65 Fast road Ⅱ 80 Yellow (50,65] Ⅲ 60 Red 50 Ⅰ 60 Green >45 Trunk road Ⅱ 50 Yellow (25,45] Ⅲ 40 Red 25 Ⅰ 50 Green >35 Secondary road Ⅱ 40 Yellow (15,35] Ⅲ 30 Red 15 Ⅰ 40 Green Branch Ⅱ 30 Yellow Ⅲ 20 Red 36

37 Figure 3-7 Storm with spatiotemporal distribution in Beijing Figure 3-8 Impacts on traffic in Beijing, July 23, 2012 After July , Beijing News reported that the city's water authority admitted the shortcomings of its infrastructure, but experts said it would take time to upgrade Beijing's flood protection systems. Figure 3-9 show the road maps of Beijing and Figure 3-10 shows the example of the detail information for the road database. The traffic models, such as VlSSIM (Germany), CORSIM, Synchro/SimTraffic and TransModeler (USA), Paramics (UK), AIMSUN (Spain) and DYNAMEQ 37

38 (Canada) are used in China. We adopted SUMO to simulate the traffic in Beijing. The data were converted to create a SUMO traffic model for the Beijing city, as shown in Figure Figure 3-9. Road map Figure Road information database 38

39 Figure SUMO model for the traffic in Beijing city The impacts of rainfall on traffic have been identified in two aspects. One is the significant negative changes of road condition and visibility during the rain events process, another is adverse effects of the residue accumulated on the road resulted from the extreme rainfall. In the operation process of urban transportation, the velocity of vehicles is reduced. Especially for the locations where traffic intensive and water accumulated, traffic congestion and paralysis possibly occur, which is one of the important indicators to justify the impacts of flood on traffic. Since it is difficult to express the influence by mathematical model, this research takes into account the effects of water inundation to road and focuses on the derivation of mathematical model. The following model analysis regards the normal operation status of urban traffic as reference. The attenuation model of traffic velocity with water depth is illustrated in the following formula: v0 x a v0 v tanh( ) 2 b 2 Where v is the velocity of vehicles (km/h), v0 is design velocity of vehicles which varies according to the road grade (km/h), x is the accumulated water depth (cm), a is critical water depth that results vehicles stopped (cm), b is the coefficient of attenuation which identifies the decreasing rate of velocity versus depth, the range 3-5 (the smaller of b, the faster of declining). Table 3-3 Traffic loss categories due to excessive rainfall Aspects of impacts Types of loss Categories of loss Economy Politics and Society Direct loss Indirect loss Indirect lossz5 39 Loss of fuel consumption Z 1 Loss of time consuming Z 2 Loss of personal working hours Z 3 Loss of infrastructure destruction Z 4 The political effects of major events and the dissatisfying of residence to municipal facilities

40 Total traffic loss Z=Z 1 +Z 2 +Z 3 +Z 4 +Z 5 The economic loss due to extra fuel consumption could be calculated by the relationship between velocity of vehicles and fuel usage. The velocity of traffic is decreased by water inundation, which is less than the critical velocity with minimum fuel consumption. Based on mechanical theory, the losses raised by fuel consumption could be acquired by the following formula: Z 1 n Pe be 1 1 ( ) 70 v v 1 Where Pe is the power of engine (KW), be is the efficiency of fuel consumption (g/kw/h), v b is critical vehicles velocity with minimum fuel consumption (km/h), is the unit price of fuel (RMB/L), n is the influenced traffic by flood inundation. The economic loss of time consuming could be get according to the following formula: Z M M M1 is economic loss of time consuming of private car: b n1 M t m 1 1 i1 M2 is economic loss of time consuming of motor coach: n2 M t m 2 2 i2 Where: t is the duration of damage (h) which could be get through site survey or estimated by average duration of public travelling, is evaluation indicator of time value (RMB/h), m1 is the average seating numbers of private car 1.5-2, m2is the average seating numbers of motor coach 40-80,n1 and n2 are traffic numbers of private cars and motor coach respectively in the survey area. The traffic simulation model could be constructed after collection and rearrangement of basic dada for study area, while the parameters for calculating traffic loss is able to derive from the model by appropriate assumption input (e.g. the number of influenced vehicles). However, due to the deficiency of basic data, this part of work is still on-going. 3.5 Discussion and conclusions For Beijing case study, currently only the depth-damage curves regarding residential building have been synthesized, depth-damage curves of other building types are still on the process. Therefore, the information extracted from UK flood damage database is adopted for the case study of Yizhuang. Due to the diversification of regional economic circumstances, the outcome of flood loss in this case is relatively magnified compared to actual status. The approaches of traffic loss evaluation have 40

41 been generated. Owning to the deficiency of data required for traffic simulation model, for instance traffic load and road condition information, the process of constructing traffic model is not implemented at the moment. However, the construction and initial calibration of hydraulic model for 20 bridges in Beijing has been completed. The next step will place emphasis on collecting basic traffic data and implementing traffic evaluation of specific Beijing case study based on urban local flood model. 41

42 4 Case study Dhaka 4.1 Case study area overview Dhaka City is surrounded by Tongi Khal on the north, Turag Buriganga River system on the west, BaluRiver on the east and Sitalakhya River on the south. These rivers are the distributaries of Brahmaputra River system and during monsoon spills from Jamuna flows through this system and create flooding. Ground elevation of the city varies from 0.5m to 12m (PWD). About 60% to 70% of the city area includes low lands, abandoned channels and depressions and the elevation of these areas vary from 0.5 to 5m (PWD) (Hossain, 2004). The drainage system of the City is operated by two agencies DCC and DWASA. The surface drains are maintained by DCC and DWASA maintains the sub surface storm drains. There are also several open channels in the City. The City is protected by embankments in the East and the West from river flooding. The water is pumped outside of the embankments during monsoon. So the capacities of pumps are very important for the drainage performance of the City. The Dhaka Case Study focuses on central part of the City which drains towards east and covers 45 km 2. There are several reservoirs within this system and the drainage network consists of underground pipes and box culverts. The study area also includes part of the eastern catchment which is low lying area beside Balu River and covers almost 78 km 2. The drainage system in the eastern catchments contains natural channels. Figure 4-1 shows the study area drainage system. The drainage in study area is dependent on two aspects: 1) Operation of storm-water drainage system including pumps and regulators; 2) Water levels on the peripheral rivers. Thus, flooding/drainage congestion in Dhaka may occur due to any of the following three reasons or combination of all of them at a time: Flooding due to congestion of storm-water/wastewater drainage systems inside the City area; Flooding due to the effect of high water level in the peripheral rivers under which circumstance drainage is only possible through pumping; Intrusion of floodwater from the peripheral rivers to the city area through the drainage routes. 4.2 Scenarios The impact assessment described in this deliverable is calculated for a single scenario: the presentday situation. Future deliverables will provide the impact assessment results that have been undertaken assuming different future scenarios (D3.8) and different measures and strategies (D3.9). 4.3 Hydraulic modelling The drainage system of Dhaka Case City has two distinct components. As mentioned earlier the western side is protected, contains planned development area and as a result has urbanized drainage system consisting of pipes, box culverts, pumps and retention ponds. On the other hand the eastern side is unprotected from flooding, the area is underdeveloped and the drainage system contains natural channels. So the hydraulic modelling for Dhaka Case study was done using two 42

43 different models to simulate the drainage condition of these two areas with different hydrologic and hydraulic characteristics (see Figure 4-1). See Chapter D2.4 for more details. In monsoon the gates of the regulators of Central Dhaka remains closed when the water level of Balu River is high. Rainfall event of high intensity at that period causes drainage congestion and urban flood. To drain out the excess water several pumps are used and the drainage system no longer works on gravity and becomes dependent on the pump capacity. On the other hand the water level of Balu River controls the drainage system of Eastern Dhaka. When the water level of Balu River is high, the low areas of Eastern Dhaka get flooded and the flood recedes when the water level of Balu River decreases. Therefore, not only the inundation depth but also the duration of flood of Eastern Dhaka is dependent on water level of Balu River. More details on hydraulic modelling is discussed in chapter D2.4. Figure 4-1: Drainage system of Dhaka Case Study 4.4 Damage / impact modelling Damage is the negative result of the spatial and temporal impact of an event on societal elements (people, buildings, etc.), societal processes (interruption of production, services, etc.) and the environment (Vetere Arellano et al., 2003). To understand the damage or losses that floods can cause, the CORFU damage model comprises three damage model components: Direct tangible damage 43

44 Indirect tangible damage Intangible damage Details for Each of these components related to Dhaka Case study are described in the flowing sections Direct tangible impacts Direct tangible damage refers to the physical damage caused to property and contents in both residential and non-residential (industrial and public sector) sectors by direct contact with flood waters. Under the CORFU project damage tool has been developed which uses data like flood depth, flood duration, land use maps, building footprints and depth-damage-curves (DDC) to quantify the amount of flood damage from an event. Availability of appropriate and large information to assess floods damage is a big challenge for the Dhaka case study. In order to account for all the information (input) required for the tools a significant amount of effort was given to collect and prepare data for the Dhaka case study. For damage assessment, aggregated land-use or object based land use could be utilized. Objectoriented data records individual properties and buildings, whereas aggregated data, by its nature, records aggregated information of more or less homogeneous areas. Object oriented data was also collected and prepared for the study area. No central agency is responsible for keeping record of object oriented data. So data were collated from various sources. These data were then updated so that they can be used together. Object oriented data contains actual building footprints and associated land-use information. The associated land use information are: Building use 1 st tier Structure type of building Building use 2 nd tier No of stories Building use 3 rd tier Construction year The flood depth information was collected from the flood models described in the previous section. As mentioned earlier 1D-2D coupled model were used to generate the flood maps which can replicate overland flow processes and can generate flood map with better accuracy. The damage tool has been used to determine building damages in both central Dhaka and eastern Dhaka. Land use data collected from RAJUK that was used in damage tool is shown in Figure 4-2. Damage calculated for rainfall event of different return period for 2004 using building data and land classes are presented in Figure

45 Figure 4-2: Land use data used in damage tool 45

46 Figure 4-3: Damage Calculated for different return periods 46

47 4.4.2 Development of Damage Functions for Dhaka city A survey was conducted to develop an understanding of estimating damage function related to flooding in Dhaka city. The damage function is a function that explains the relationship between damages incurred by a household (or a business) due to water logging in their area in one year based on duration of water logging days in a years and depth of water in their location. The survey was conducted using primary survey from households and business organizations to capture this information. Survey was conducted using a systematic stratified random sampling method based on the land elevation (flood depth) and structure of the building/premise. It was found from the survey that every year business activities have to experience many types of flood events including floor damage, wall damage, plinth damage etc. 47

48 Percentage of cases Percentage of cases Project Report Figure 4-4: Experienced events around the business place in last 10 years Around the business premise, mostly common experienced event is floor damage (56.30%), wall damage (51.26%), roof damage (31.09%) and compound damage (25.53%) ( 60% 50% 40% 30% 20% 10% 0% 60% 50% 40% 30% 20% 10% 0% Experienced events due to Water logging Floor damage Wall damage Window damage Roof damage Door damage Figure 4-4). There is also a significant effect on public properties like as roads, drainage, electricity and water supply network damage. About 80% respondents stated about roads damage in their locality and 61% stated about drains damage; 93% stated about clogging of drains in their locality due to flood events. Now if only asset damage in response to flood depth is considered, study estimated a damage depth function. A damage depth function is a mathematical relationship between the flood water level above from the plinth of road (in front of the structure) and the amount of damage per square feet area that can be attributed to that water in per flood event. Here damage is calculated in response to unit change of flood depth (inch). Symbolically, In econometric form, (4) Compound damage (3) Other damage Cases (%) 56.30% 51.26% 5.88% 31.09% 8.40% 23.53% 9.24% Experienced events due to Water logging Floor damage Wall damage Window damage Roof damage Door damage Compound damage Other damage Cases (%) 56.30% 51.26% 5.88% 31.09% 8.40% 23.53% 9.24% 48

49 Damage per square feet per event (BDT) Project Report Here, ; AD= Asset Damage, TA= Area (square feet). So it indicates damage per square feet area for jth event of ith business organization, business organization. = Flood depth (inch) for jth flood event of ith By estimating this function, study found that for one inch increase of flood depth leads about BDT damage cost keeping other factors constant in per square feet area. In base condition (mean depth 16.4 inch), without any change, damage is BDT 14. In zero (0) depth, there is also some damage (BDT 0.34). It indicates that the area is already states in low elevation where flood incurs damage in spite of no change of depth. Study has estimated the following damage curve for per square feet area with changes of depth unit (in inch) for per event (Figure 4-5) Damage Depth Curve (Commercial) Average depth (in inch) per event Figure 4-5: Per event business asset damage curve (square feet) in response to depth Survey found that, at 10 inches flood depth, per square feet damage cost is BDT 8.7, which is almost half in amount. When flood depth increases to 25 inches (from the base situation) without considering duration, damage also increases remarkably (BDT 21). It further increases with the increase of flood depth, so there is a positive relationship with depth and damage. Survey also found that every household faced about 3.5 flood events in last 10 years and a large amount of damage cost every year. About 70 percent household lives in rented house where is limited capacity in taking precautions and mitigation strategies for minimizing damage. According to household survey and Figure 4-6 shows that, about 70 percent household faces wall damage and about 46 percent household faces floor damage during the flood event or after the events. These are mostly common experienced events. Besides, other important events are door damage (33%), roof damage (29%) and window damage (23%). 49

50 Others 2.14% Cases (%) Water reservoir damage 4.29% Floor damage 45.71% Compound damage 15% Door damage 32.86% Roof damage 28.57% Washed away yard's soil 7.14% Window damage 22.86% Wall damage 69.29% Land slide 4.29% Figure 4-6: Experienced damage events by household. Households also reported about flood effect on their locality. According to survey results, most of the respondents reported about clogging of drains (87%), 67% reported about damage of roads and 64% reported about damage of drains in their locality during and after water logging events. Household s has to lose a large amount of money regarding asset damage every year and 50 percent household reported about damage due to flood events. Mostly reported case in this study is damage of furniture (61%) and wall damage (33%) in terms of money. Table 4-1 shows different household asset damage category considered for this study. Table 4-1: Household asset damage category Asset Damage category Cases (%) Roof damage 20 Plinth damage 9 Wall damage 33 Loss of tree 2 Damage of utensils 6 Cost of relocation 5 Vehicle damage 2 Damage of furniture 61 Loss of livestock 1 Income loss 13 Business loss (income) 1 Business loss (instruments) 5 Others 13 50

51 Damage per square feet per event (BDT) Project Report The estimated damage curve ( Figure 4-7) in response to depth shows that asset damage starts at 4 inches depth without considering flood duration. Up to 4 inches of flood depth, there is no damage. With the increase of flood depth at 40 inches, damage becomes BDT 28 in zero (0) day duration and Flood Depth (inch) then it increases with depth. Flood Damage with Depth Figure 4-7: Household s asset damage (per square feet) in response to flood depth per event To assess the damage due to flood, damage curves for different Building Classes were developed using these surveys. For Central Dhaka the Residential Building is 66% of total area (see Table 4-2). Therefore, the damage for residential structures governs. For the future scenarios Land Classes are used instead of Building Type as the results of Urban Growth Model (UGM) provide land classes. To deduce damage curves for Land Classes, the ratio between land classes and building types was considered to be the same for the future. The damage curve for the land classes were generated using the damage calculation tool from the damage curves of the buildings. Table 4-2: Percentage of total area for different land classes Building Type % of total area Commercial Activity 7.6 Community Service 1.4 Education & Research 2.8 Governmental Service 1.7 Manufacturing & Proc 2.6 Mixed Use 9.8 Residential 66.1 Service Activity 6.2 Transport & Communication

52 4.5 Results from the damage or impact assessment Using these damage curves the damage due to flood was calculated for different scenarios. The calculated damages are for rainfall event of different return periods in 2004 are presented in Figure 4-8. Figure 4-8: Damage comparison for different events in Dhaka Case Study The damage calculation for an actual event in 2004 and also a 30-yr design storms was calculated and it was found that in the larger the event the damage impact becomes more pronounced. EAD was calculated for the scenario and it was also found that annual damage for the study areas is around 0.4 million USD. Eastern Dhaka is unprotected. But due to less urban areas the asset damage was relatively small. Figure 4-9 shows the damage maps for Central Dhaka and Eastern Dhaka for the actual event of

53 Figure 4-9: Damage map using damage tool 4.6 Discussion and conclusions The damage due to flood is done using land use classification for existing scenario. The damage for 1 in 100 year event is greater than 1 in 30 year event as the total amount of rainfall as well as rainfall intensity increases with return period of the event. An important aspect has to be noted is the damage function developed from survey for Dhaka. It only considers asset damage and does not consider loss of income or any other related damage. Also the total damage was for only the study area which is one third of the City and also contains large undeveloped area. So in that sense the total damage may appear low. For the future scenarios the results from urban growth model will be used to relate damage with inundation extent and depth. Assessment of Flood damage for future scenarios is explained in chapter D3.8 in details. 53

54 5 Case study Hamburg 5.1 Case study area overview In order to address the main flood problems in Hamburg, being the storm surges and the riverine floods, two areas have been selected and are given as - the island of Wilhelmsburg (storm surges) - the catchment of the urban water course Wandse (riverine floods), also depicted in Figure 5-1. Due to the specific nature and typologies of the flood related issues in both areas, they are given separately in the following text. The Elbe River Figure 5-1 Case study areas in Hamburg (red- the island of Wilhelmsburg prone to storm surges; yellowthe Wandse catchment prone to riverine floods) 1 Case study area Wilhelmsburg In the city of Hamburg/Germany the Elbe estuary bifurcates into a northern and a southern branch, forming an island in the middle called Wilhelmsburg. It is part of the urban area, but still with a high potential of spatial development. This has been discovered by the policy which launched a new development program for Wilhelmsburg, Leap across the Elbe, which should create new homes for to citizens on this island. Although more than 100 km away from the North Sea, the Elbe estuary has still a tidal range of 3.5 m at the gauge in Hamburg. Every year, Hamburg is threatened by storm surges which led to a catastrophic flooding of wide parts of the city and 1 For the island of Wilhelmsburg, the results gained in the research project XtremRisK ( ) have been used as a basis. For the river Wandse, the research work has been aligned with the KLIMZUG-Nord Project ( funded by the German Government. 54

55 especially the island Wilhelmsburg in 1962, where more than 300 people were drowned. This led to a dike enforcement program, in which a dike ring around the island has been raised twice in the last 40 years. Today the design high water stage for this dike system is 7.30 m ASL, which is about 85 cm higher than the highest recorded flood. Due to a free board that varies in dependence on the expected wave height the crest height of the dikes and flood protection walls varies between 7.70 m ASL and 8.35 m ASL. Thus at present the dikes seem to provide sufficient security to overtopping (Pasche et al., 2008). It is however, required to explore the potential futures and their impacts to the flood situation in the Wilhlemsburg area. Case study area Wandse Wandse is a small urban catchment of about 87 Km 2, whereby 60 Km 2 is located in the Hamburg area. In terms of its topographic characteristics it is considered as a low-lying area (0-80 m asl), spreading from the NW to SE. The upper catchment is close to the natural state dominated by farmland and nature protection area. Main urban area, located in the mid and lower catchment, is a high density residential area, dominated by detached buildings (23,85% out of all landuse types in the Wandse catchment). Industrial area is mostly located in the mid and lower catchment area, partly directly at the river (e.g. Yeast factory at the Km or a commercial centre encroaching the river Wandse at the Km ). 66,6 % of the catchment drains in the separate system, the lower catchment part to the combined sewerage system. The main characteristics of the Wandse catchment are summarised in Figure 5-2. Topography: - predominantly lowland m a s l Soil type: - dominated by medium to light clayey sand Urbanization type: - differentiated; upstream and middle part areas are closer to the natural state Sewerage system: - partly combined, two third of sewer system is of separate nature Figure 5-2 Summary of the main parameters characterizing the Wandse cathcment area The Wandse River caused flood problems in the past. The recorded history of recent floods at the river Wandse dates back to Since October 1998, four significant flood events have been recorded, when the emergency services were put in action. It is expected that the flood problem will become more severe in the future. 55

56 5.2 Scenarios For the scope of this report, the present state is considered i.e. the reference year is set to This will be used as a reference state and compared to the potential damage for the future scenarios. 5.3 Case study focus Case study area Wilhelmsburg The major source of flooding in Wilhelmsburg is from extreme storm surges that exceed the current level of protection. Storm surge scenarios have previously been developed in the German XtremRisK project based on the physically possible largest extreme storm surges for the pilot site. They are based on the analysis of field data and numerical modeling by analyzing all storm surge constituents (e.g. wind surge, tide and external surge) and their non-linear interactions as compared to their linear superposition (Oumeraci et al. 2012). We have selected the storm surge HH_XR2010A for use in CORFU. This artificial storm surge s peak water level is based on the storm surge that occurred in January 1976, with a wind surge of 370cm at low tide. This backwater curve was overlaid with a maximum external surge and a maximum spring tide based on non-linear factors. Furthermore, the highest observed upstream runoff of Q = 3600 m³/s is considered Case study area Wandse Based on the discussions with the key end users being the responsible authorities for flood risk management in the city of Hamburg, NGOs, spatial planners and ecologists, the 3 focus areas in the Wandse catchment has been selected which indicates the highest damage potential being Wandsbeker Chaussee, Ostend pond and Rahlstedt. They represent a typical urban, urban-suburban and suburban areas resp. They are given in Figure 5-3. The hydrologic and hydraulic modelling has been performed and presented for those areas. 56

57 Figure 5-3 Focus areas for the damage assessment in the Wandse catchment Wandsbeker Chaussee, Ostend pond and Rahlstedt (indicated by the red line) 5.4 Hydraulic modelling - Wilhelmsburg Hydraulic modelling approach and tool In order to determine the hydraulic input parameters for the damage calculation, an inundation model was established to simulate the propagation. The modelling of the hinterland flooding due to the extreme storm surge has been performed using the DHI software MIKE 21 FM HD (for hydrodynamic 2D simulation using flexible mesh). Buildings were not included in the digital terrain model and neglected for the mesh generation. The boundary conditions for the inundation simulation of the hinterland were determined by wave overtopping and overflow calculations. For this purpose, the unsteady conditions of water level and wave parameters at the toe of the defence structures were used as input data. The time series were gained by processing a 2d hydrodynamic model using RMA Kalypso which had been coupled with the wave model SWAN (simulating waves nearshore) providing all necessary wave parameters. The inflow has been set as time-dependent boundary conditions defining the discharge at 94 homogeneous sections of the ring dike Hydraulic modelling approach and tool The modelling of the hinterland flooding due to the extreme storm surge has been performed using the DHI software MIKE 21 FM HD (for hydrodynamic 2D simulation using flexible mesh). Buildings were not included in the digital terrain model and neglected for the mesh generation. The boundary conditions for the inundation simulation of the hinterland were determined by wave overtopping and overflow calculations. For this purpose, the unsteady conditions of water level and wave parameters at the toe of the defence structures were used as input data. The time series were 57

58 gained by processing a 2d hydrodynamic model using RMA Kalypso which had been coupled with the wave model SWAN (simulating waves nearshore) providing all necessary wave parameters. The inflow has been set as time-dependent boundary conditions defining the discharge at 94 homogeneous sections of the ring dike Hydraulic modelling data For the model set-up a digital terrain model (DTM) was generated based on the combination of laser scanning point data, breaklines and polygons of the channel beds describing the waterways. On this basis the DTM contains the sufficient accuracy for a realistically representation of the modelled environment and the simulated inundation of the water within the area. Roughness modelling was considered based on landuse categories and breaklines were derived from the datasets. The combination of breaklines from both the roughness polygons and the DTM is the basis for the mesh generation (Figure 5-4) using GAJA3D, a MATLAB based library for flexible mesh creation (Rath, 2007). Figure 5-4: Mesh creation for inundation model based on breaklines from DTM and landuse Hydraulic modelling calibration and verification For the calibration of the model no data of historical events had been available. The verification had been performed by balancing the inflow volume with regards to the time-series of the boundary conditions. The total inflow volume has later been compared with the calculated water of the inundated area after the last time step. For the calculation the default values for model parameters had been chosen Hydraulic modelling runs and results In order to describe the modelling runs undertaken Table 5-1 gives an overview of the storm surge for current climate change scenarios ( year 2010 ). The according peak water levels and exceedance probabilities are presented. It can be seen that extreme events with high peak water levels and corresponding low exceedance probabilities were selected. The exceedance probabilities P e are determined based on multivariate statistics (Wahl et al. 2012) considering the peak water level and the fullness as well as the discharge describing the intensity of the storm surge curve. 58

59 Table 5-1: Extreme Storm surge events for Hamburg storm surge events peak water level [m NN] exceedance probability [1/year] HH-XR2010A , HH-XR2010B , HH-XR2010C , The results of the inundation modelling have differently strong impacts due to varying storm surge heights and the resulting inundated areas according to Figure 5-5 and Table 5-2. In the flood event XR2010A almost 75% of the area is affected by the inundation, while in the flood event XR2010C almost the whole area is submerged with almost 98% of the area resulting in water depth up to 8m. Solely the waste deposit site (in the northeast part of Wilhelmsburg) which is higher elevated is excluded from the inundation. In contrast, flood event XR2010B shows a remote inundation resulting in almost 1% of the area. Figure 5-5: The results from the inundation modelling for the flood events XR2010A, XR2010B and XR2010C Table 5-2: Resulting inundated areas from Storm surge events 5.5 Damage / impact modelling - Wilhemsburg For the spatial modelling of the tangible damages a Cell-based Risk Assessment (CRA) approach has been used that allows an integrated and robust framework for the integration of tangible damages 59

60 in the following risk analysis (Burzel & Oumeraci 2012). The polygon based concept for spatial analysis allows the intersection of risk elements and the developed damage functions on the basis of sample-type buildings with the results from the inundation model. Therefore, the maximum water depths are assigned to a compartmentation of regular cell sizes (100 m, 50 m, 10 m) and intersected with the objects of each damage category. The resulting damages for residential objects on a 50m grid are exemplarily presented in Figure 5-6 Figure 5-6 Tangible losses on residential buildings based on a 50m grid Direct tangible impacts For the assessment, direct damages are classified into damages to residential buildings, commercial objects, infrastructure and agriculture. In the present study, the focus is on the first two categories. Basically, the spatial distribution of elements at risk is carried out for the determination of direct damages in the site under investigation. Furthermore, a classification is required in order to assess the inventory for the development of depth-damage functions. For the residential buildings the predominant building types are determined based on site inspections. Sample-type buildings are classified based on the type of building, occupancy of the ground floor and the type of facing of the building. For the sample-type buildings the damage potential is determined based on recovery and replacement costs of inventory and building construction. Object based damage-functions are created based on building inspections, photos and floor plans of the buildings. The commercial objects are classified based on the branch of economic activity with respect to the firm s industry classification and the business sizes measured by the number of employees. The assets are downscaled to Wilhelmsburg from the national accounts defined as fixed assets, divided into buildings and equipment. Relative damage-functions are derived for buildings and equipment separately based on direct inquiries and visits with local businesses (Ujeyl et al. 2012). 60

61 The damages to infrastructure and agriculture are determined based on meso-scale approaches. Indirect impacts are caused by disruption of economic and social activities as a consequence of direct flood damages. The applied economic model simulates the indirect impacts in the affected region which are due to business interruptions, production losses during the reconstruction period and service losses in the housing sector. They are represented by the reduction of the total value added in the economy because of the disaster. The Adaptive Regional Input-Output (ARIO) model was developed by Hallegatte (2012). The ARIO model aims to explain the development of a regional economy in the aftermath of a natural disaster. It is based on a regional input-output table. This provides a highly disaggregated sector structure and allows to account for inter-industry supply and demand relationships of intermediate products. The idea is that destroyed productive capital reduces the production capacity in the affected industries and, at the same time, raises demand for reconstruction in the construction and manufacturing sector. It is generally assumed that each industry produces according to (i) observed demand depending on orders it receives from households and other sectors, (ii) the availability of inputs depending on supplier production and the level of inventories, and (iii) its own production capacity. Thereby, industries are able to increase their production capacity as response to production shortages, e.g. by increasing working hours or temporary hiring workers from outside the region. By taking into account inventories and production dynamics the model is able to represent production bottlenecks that can propagate into the regional economy. Thus, firms which are not directly affected by the disaster might suffer from supply bottlenecks of intermediate products or profit from increasing demand for reconstruction. 5.6 Results from the damage or impact assessment - Wilhelmsburg Direct damages Based on the storm surge event XR2010A the damages are calculated with respect to the maximum water stages occurring at each location over the timespan of propagation. The damages are calculated separately for the residential buildings and the commercial objects and in each case separately for the buildings and the inventory. According to the inundated areas the damages result in different magnitudes (Table 5-3). In Scenario XR2010A the damages to residential objects (buildings and inventory) dominate the overall damages with 61% and 7,135 of 9,462 buildings are affected. The damages to the commercial objects sum up to 38% of the overall damages with 830 of 1,459 businesses being affected. In case of the flood event XR2010C in which the whole Island and thus all objects are affected is equivalent to the total loss and the overall damage potential of the island. Here the damages to the commercial objects make up 50% of the overall damage and the residential 46%. The flood event XR2010B causes a low damage to 18 commercial objects and 8 residential objects due to the relatively small inundated area. Thereby the damages to the commercial objects are dominant with 90% of the total damage, thus the residential objects make up 8% of the total damages. 61

62 Table 5-3: Calculated damages based on the flood events Indirect damages In the time after the event, production capacities in Wilhelmsburg are reduced, residential houses are damaged and firms and household demand for reconstruction and reparation. In the dynamic effects, which can be observed in the economy in each event are illustrated. In the event XR2010A with medium damages value added in Hamburg decreases immediately after the storm surge by 0.3 per cent due to destruction of productive capital (Figure 5-7(left)). Rapidly, reconstruction and reparation begin so that production capacities increase again. After about four months, value added overshoots the pre-disaster level due to the combined effect of higher demand and overproduction capacity, which has been build up because of the high reconstruction demand. As reconstruction demand decreases, overproduction capacities are cut back as well. The economy returns to the predisaster situation after about 1.5 years. In this scenario, output losses are solely due to the reduction in production capacities. Forward or backward ripple effects are not observed. This is due to the relatively low direct damages and the large amount of remaining production capacity in the parts of Hamburg which are not affected. In total, the city even benefits from the event due to the high reconstruction demand. Figure 5-7 Total value added losses for XR2010A (left), XR2010B (middle), XR2010C (right) However, impacts substantially differ between industries. Furthermore, it can be assumed that output losses occur to firms directly affected by the flood event, i.e. located in Wilhelmsburg. The highest losses are observed for the Real Estate sector in which the damages to residential buildings occur and losses of housing services are captured. Furthermore, Wholesale and retail trade and 62

63 Transport, storage and communication experience high value added losses amounting to 4.2 and 5.8 million euros, respectively. In contrast, firms for the most part located outside of the affected district benefit from the reconstruction demand either directly, such as Construction (49.8 million euros) and Manufacture of transport equipment (7.4 million euros), or indirectly, such as Financial intermediation (3.3 million euros) (Table 5-4). Table 5-4: Direct damages and total value added losses XR2010A direct damages indirect losses XR2010B direct damages indirect losses XR2010C direct damages indirect losses Sector million Agriculture, forestry, fishing, mining Manufacture of food products and beverage, manufacture of tobacco products Manufacture of textiles and textile products; leather and leather products Manufacture of wood and wood products (except furniture) Manufacture of pulp, paper and paper products; publishing, printing Manufacture of coke, refined petroleum products and nuclear fuel Manufacture of chemicals, chemical products and man-made fibres Manufacture of rubber and plastic products Manufacture of other non-metalic mineral products Manufacture of basic metals and fabricated metal products Manufacture of machinery and equipment Manufacture of electrical and optical equipment Manufacture of transport equipment Manufacture n.e.c., Recycling Electricity, gas and water supply Construction Wholesale and retail trade; repair of motor vehicles, motorcycles and personal and household goods Hotels and restaurant Transport, storage and communication Financial intermediation Real estate, renting and business activities , Public administration and defence; compulsory social security Education Health and social work Other community, social and personal service activities SUM 1, , In the event of XR2010B, where direct damages are very low, aggregated value added increases by 90,000 euros (Table 5-1, Figure 5-7(middle)). Losses are only observed in the Manufacture of basic metals and metal products (70,000 euros), the Manufacture of machinery equipment (30,000 euros) and Other service activities (120,000 euros). These sectors also suffer from direct damages. The construction sector benefits from the reconstruction demand and value added gains are 170,000 euros. The event XR2010C shows that large scale disasters can have significant impact on a regional economy. In this scenario, the total value added losses amount to 548 million euros. This result adds to earlier findings that the relationship between direct and indirect damages is not a linear one (e.g. Hallegatte 2008, Cavallo & Noy 2009). Above a specific threshold indirect losses increase more rapidly than direct losses. This is due to two effects. Firstly, the reduction of reconstruction capacities is higher after a large-scale event than after a small-scale event because of the destruction of productive capital. At the same time, reconstruction needs are higher after large than after smallscale events. Furthermore, exceeding a certain amount of damages, forward and backward ripple effects occur leading to a stronger increase in indirect damages (see also Figure 5-7(right)). Thus, firms which are not directly affected by the storm surge suffer from bottlenecks and also have to cut back their production. 63

64 5.7 Hydraulic modelling - Wandse Hydraulic modelling approach and tool In order to assess potential damage in the selected focus areas in the Wandse catchment for the present state, the open source modelling platform Kalypso ( has been deployed. KALYPSO is based on a set of generic functions for managing spatial data called KalypsoBASE. It handles and stores all model data in the OGC Geography Markup Language (GML), a well-known structured text format, and implements a number of OGC web service standards. Five advanced modules for specific simulation purposes are founded upon KalypsoBASE (Kalypso Hydrology, Kalypso WSPM, Kalypso 1D2D, Kalypso Flood and Kalypso Risk). For the hydrologic and hydraulic and risk modelling purposes in the Wandse catchment, the Kalypso Hydrology, Kalypso WSPM, Kalypso 1D2D, Kalypso Flood have been chained as shown in Figure 5-8. Figure 5-8 Kalypso modules for the process of flood risk management planning. The scope of Kalypso Hydrology is to simulate the rainfall-runoff part of the hydrologic cycle. Kalypso WSPM computes one-dimensional, steady-state water surface profiles and supports standard approaches for hydraulic computations in rivers. In addition to the steady-state module, Kalypso 1D2D simulates unsteady,1d-2d coupled and 2D surface flow. The Kalypso Flood module determines and displays the inundated areas and flow depths based on high-resolution digital terrain data and 1D or 2D model results Hydraulic modelling data The data requirement for the model set up are: topographic data, urban drainage network data, climate data (daily to sub hourly precipitation, daily mean temperature, daily potential evapotranspiration), land use and land cover data, soil data, geological data, and stream flow data. The climate data input required by KalypsoHydrology model are daily precipitation time series, daily mean temperature time series and daily average potential evapo-transpiration. And the output from the hydrological model are discharge time series for selected subcatchments and subcatchment nodes, precipitation time series, evapo-transpiration time series and average groundwater level time series for each subcatchments. 64

65 5.7.2 Hydraulic modelling calibration and verification Once the KalypsoHydrology-based model has been set up and calibrated for the long and short term periods, it was used for the simulation of climate and land use change scenarios. In order to simulate impact of climate change on urban hydrology, three impact indicators are selected. For the calibration, the event of July 2005 (gauging station Wandsbeker Allee peak discharge Q= 6,4 m3/s ari 2) has been selected due to the availability and reliability of the measured data. The calibration has been performed for all main gauging stations, using the peak discharge criteria (short term simulation) for the calibration procedure. 5.8 Damage / impact modelling - Wandse Direct tangible impacts The damage assessment has been performed for different elements of focus areas; the focus has been put on the assessment of the damages to buildings. The building stock is analysed leading to the development of the typologies which represent the buildings with the similar characteristics following the defined parameters. Table 5-5 outlines the main parameters decisive for the definition of the building typologies. Table 5-5 Parameters for the definition of the building typologies (B-basement; GF- ground floor) BUILDING PARAMETERISATION Dominant building function Residential, commercial, industrial, mixed, etc. Number of basement floors B 1 or 2 (from ground level) Construction year type Old, after-war, new construction B Brickwork, wood-beamed ceilings, cap ceilings Reinforced concrete, concrete ceilings (solid) Dominant construction materials from Precast exterior walls and ceilings (Surfaces Timber frame exposed to flood) GF Clay construction Masonry (solid) Reinforced concrete (solid) Building condition and functionality B/GF Flood-proofed Functional, in operation Poor conditions, not functional For the analysis, both residential and industrial buildings have been considered. The main types of buildings classified according to their dominant function are given in 65

66 Industry/Commercial Residential Project Report Table 5-6. Table 5-6 The considered building types classified according to their dominant function Function Residential adjoining buildings Residential main building Trade and services public and private, state Business and industry Utilities Transport/Storage facilities Agriculture Building Representatives Garages, garden sheds Multi-storey buildings, terraced house, semi-detached, apartment house - residential block, single family, cottage, garden house (solid), dorm, living and trade mix Restaurant, entertainment venue, insurance, management, office, medical practice, community pool, sports facilities, court, police, schools (general, vocational), Education and Research, kindergarten, other public administration, community center, leisure center, mixed-use (commercial and service within the ground floor) Business-industry-residential mix, production/factory, craft workshop, factory building, warehouse building, Operations building, gas, water services, electricity, waste treatment, waste storage, other utilities Railway equipment, operations building, parking deck, underground parking Rural residential building, greenhouse, barn, stables, other farm buildings For the residential buildings, the main typologies considered for the Wandse catchment and parameters relevant for the damage assessment are given in Table 5-7 (Adapted from Office of Environmental Health Hamburg (2002)). In addition to the building typologies, the typical inventory for different types of the building has been assessed based on the INFAS data (INFAS, 2013) or in a form of a site survey. They have been further processed by FLORETO, a web based tool for damage assessment (Manojlovic, 2014) and the typical potential damages depending on the type of inventory (both commercial and residential) and lifestyle of the dwellers (residential) and the and have been derived. An example is given in Figure

67 Figure 5-9 An example of the derivation of damage curves for inventory utilising FLORETO ( ) 67

68 Table 5-7 Building typologies considered for the Wandse catchment and parameters relevant for classification of the damage degree and the damage degree function after Maiwald&Schwarz, 2010 (Adapted from Office of Environmental Health Hamburg (2002)); Construction type: Exterior walls: solid, Solid brick (masonry-/framework); Floors: wood beams; Masonry: UG: HW-B, A= 0,055, B= -0,495; EG: HW-B-a, A= 0,381, B= -0,495; The parameters for the damage degree function (A /B): Timber frame: UG, HW-B, A= 0,055, B= -0,495; EG: HW-A, A= 0,683, B= -0,495 Construction type: Exterior walls: solid, solid brick, double-glazed masonry; Floors: solid brickwork, vaulted ceiling, concrete, reinforced concrete The parameters for the damage degree function (A /B) Masonry: UG: HW-B, A= 0,122, B= -0,495; EG: HW-B-a, A= 0,281, B= -0,495 Construction type: The parameters for the damage degree function (A /B): Exterior walls: solid, perforated brick, sandstone-limestone, sandstone-solid brick (double-glazed), solid brick, crushed concrete brick (double-glazed); Floors: solid, concrete, reinforced concrete Masonry: UG: HW-B, A= 0,122, B= -0,495; EG: HW-B-a, A= 0,281, B= -0,495 68

69 Construction type: The parameters for the damage degree function (A /B): : Exterior walls: solid, perforated brick (double-glazed), sandstonelimestone, sandstone-solid brick (double-glazed), crushed concrete brick (double-glazed), insulated concrete forms; Floors: solid, concrete, reinforced concrete Masonry: UG: HW-C, A= 0,143, B= -0,465; EG: HW-C, A= 0,130, B= -0,465 Construction type: The parameters for the damage degree function (A /B): Exterior walls: solid, perforated brick (double-glazed), sandstonelimestone (curtain wall), aerated concrete (double-glazed); Floors: solid, concrete, reinforced concrete Masonry: UG: HW-C, A= 0,143, B= -0,465; EG: HW-C, A= 0,130, B= - 0,465 Construction type: The parameters for the damage degree function (A /B): Exterior walls: solid, see (poor data avilability); Floors: solid, see (poor data availability) Masonry: UG: HW-C, A= 0,143, B= -0,465; EG: HW-C, A= 0,130, B= - 0,465 69

70 Construction type: The parameters for the damage degree function (A /B): Exterior walls: solid, see (poor data status); Floors: solid, see (poor data status) Masonry: UG: HW-C, A= 0,143, B= -0,465; EG: HW-C, A= 0,130, B= -0,465 For the assessment of the structural damage to buildings, the concept of damage degrees has been introduced, indicating the type and the extent of damages to buildings, based on their characteristics and type of flood factors acting on a building. The damage degrees (after Maiwald & Schwarz 2010) are summarized into 5 classes indicating the level of the structural damage to the building (from 1- moistening of the walls to 5- total damage). The damage degree can be computed utilizing Equation 1. Equation 1: Quantification of damage degree to the structure of the building (Vulnerability function) (Maiwald & Schwarz 2010) D m= average damage degree [-] A= to be selected building parameter (basement/ground floor) [-] B= to be selected building parameter (basement/ground floor) [-] H GF= water depth a.s.l. [m] Parameters A and B, are derived based on the ex post analysis of the flood events in Germany and Central Europe of 2002, depending on the construction types, characteristics of the building elements, and construction year and are available in Maiwald & Schwarz, The obtained damage degree indicates the type of damage to a building for the given flood factor (water depth) and the building characteristics enabling the definition of the relevant damage curves, which represent the level of damage to a building (in %) depending on the water depth h GF. The assessment of the damage degree after Maiwald & Schwarz (2010) is illustrated in Figure

71 Figure 5-10 Definition of damage degrees after Maiwald & Schwarz (2010). The colours indicate the damage degree ranging from blue (moistening of the building damage) to red (severe structural damage) For the analysed buildings, the damage degree has been assessed applying Equation 1. This led to the derivation of the damage curves and the associated refurbishment costs to rebuild the property after being exposed to a flood hazard. An example of a damage curve is given in Figure Figure 5-11 Damage degrees to the building stock in the Wandse catchment 71

72 The existing buildings have been divided into classes as given in Table 5-8. This led to the derivation of the damage curves and the associated refurbishment costs to rebuild the property after being exposed to a flood hazard. An example of a damage curve is given in Figure Table 5-8 Classification of the buildings in the Wandse catchment and the associated damage degree class Use Damage degree class Building levels considered Public-private services (hospital, kindergarten) Residential buildings_i- (Underground parking) Residential buildingsdetached and terraced buildings Trade, hotels, and restaurants-(leisure) Public civil works- (surfaces) HW-C- (normal resistance) HW-D- (increased resistance) HW-A-(very vulnerable Bausubstanz) HW-B- (sensitive buildings) I- (high quality, normal susceptibility) GF- (Damage starting at ground floor) B (Damage starting at basement) B (Damage starting at basement) GF-(Damage starting at ground floor) Streets-Plazas (no basement) Damage curve to the residential buildings (HW-C) Figure 5-12 An example of a damage curve: multi-, duplex-, detached, medium resistance (1948) from the ground floor (residential, and public buildings) 72

73 Damage degree Figure 5-13 Damage curve relating the damage degree and the associated costs to refurbish a building (here given for the building type ) The relative potential impacts obtained through the damage degree are given the monetary value and are expressed as the refurbishment costs including the costs for the manpower, energy, material as well as the costs associated with drying and cleaning. An example of such function is given in Figure This leads to derivation of the absolute damage curves for the defined building typologies, which relate the flood factor (in this case water depth) to the refurbishment costs for the defined building typologies. For the Wandse catchment, the German cost index system (BKI) has been applied. Further, the annual expected damage (AED) for the direct tangible damage is calculated based on the Equation 2. It combines the probability of occurrence and the corresponding potential impacts. Equation 2: Annual expected damage (AED) calculation With = damage at a flood event of a certain probability p, p i = the difference of probability between the flood event i-1 and I, and max = the maximum number of flood events to be considered for damage assessment which is due to the design flood chosen for the area under consideration. In order to calculate the potential damage for the focus areas, the GIS based open source modelling platform KALYPSO has been deployed following the model chain scheme as given in Figure 5-8. For the modelling purposes, the study area is represented as a raster grid, where the size and the 73

74 distribution of the grid cells is determined by the water depth raster. The damage within each grid cell of the inundated area can be calculated with the following equation: Equation 3 Specific damage for grid cell With D i,j = specific damage for grid cell [i,j] in ( /m²), C i,j = percentage of damage in dependence of land use category Ln or the building typology class and the water depth h i,j in grid cell [i,j] and V Ln = specific asset/capital stock for each land use type LU or building typology class BT in [ /m²]. On the basis of the damage functions the damage within each grid cell can be calculated by applying Equation 3 and utilizing the obtained asset values and refurbishment costs in the KalypsoRisk module. An example of the calculation procedure applying the KalypsoRisk tool for the focus area Ostender Pond is given in Figure Figure 5-14 Screenshot from land use properties in KALYPSORisk, assigned assets and relative damage functions 5.9 Results from the damage or impact assessment - Wandse For the flood risk assessment the damage of the flood events with a return period (Ti) of once in 100 years, once in 30 years and once in 5 years have been considered for the present state. The annual expected damage is than calculated applying the Equation 2). The results of the damages for the specific flood events with the probability pi and for the annual expected damage are summarised in Table

75 (1) (2) (3) 75

76 (4) (5) 76

77 (6) (7) 77 (8)

78 (9) Figure 5-15 Spatial distribution of specific damage ( /m²) for a 5, 30 and 100 year flood event on the basis of the present scenario (year 2010); The results are illustrated for the focus areas: a highly urbanised (1-3), an urban-suburban (4-6) and a suburban area (7-9) Table 5-9 Damage at a flood event of a certain probability p and annual expected damage (AED) for the Wandse catchment on the basis of the reference land use- present state (Present state) Return Period T i Probability p i S (Damage at flood event p i ) [a] [1/a] [1000 /a] 5 0,20 500,5 30 0, , , ,7 Annual Expected Damage: AED [1000 /a] 172,2 2 78

79 5.10 Discussion and conclusions With the help of hydrological and hydraulic models and detailed data about residential and commercial assets, direct damages where estimated in the flooded area. The separated assessment of inventory and buildings enables the evaluation of the distribution of damages in the damage categories for both residential and commercial objects. In the case of the residential objects the highest damage potential lies in the buildings due to high asset values. The inventory makes up 10-12% of the buildings damage potential. In the case of commercial objects it is contrary for the events XR2010A and XR2010C in which vast inundation is caused. Here, due to high asset values of the inventory it is the predominant damage category with damages to buildings summing up to only 28-43% of the damage to inventory. The high damages to the inventory can be traced back to the relative damage functions of the inventory which show significant damage grades for relatively low inundation depths. The calculated direct damages are the input parameter for the calculation of indirect damages. The following economic model delivered estimates about the total value added losses which result from production interruption and service losses in the housing sector. However, it has to be mentioned that the uncertainty in the estimation of economic costs from natural disasters is large. The presented results give an idea about the dimension of indirect losses and the relation between direct and indirect losses. Future research needs to further include sensitivity test of the model parameters. They will provide evidence on the key factors which influence the amount of indirect losses and, thus, the resilience of the economic system. For the Wandse catchment, the present conditions the resilient or adaptive capacity of the area is likely to be exceeded for the events between 5 and 30 years. This has to be considered for the definition and implementation of the adaptive measures References Burzel, A. & Oumeraci, H Development of a Framework for the Spatial Modelling of Extreme Risks and the Consideration of Risk Acceptance: Progress Report 1: Cellbased Risk Assessment (CRA) approach. State of the Art Report. Leichtweiß-Institute for Hydraulic Engineering and Water Resources, Technische Universität Braunschweig, 36 p., unpublished. Cavallo E. & Noy I The Economics of Natural Disasters A survey. Inter-American Development Bank Working Paper 124. Hallegatte S An adaptive regional input-output model and its application to the assessment of the economic cost of Katrina. Risk Analysis 25(3): Hallegatte, S. (2012): Modeling the roles of heterogeneity, substitution, and inventories in the assessment of natural disaster economic costs. World Bank Policy Research Working Paper

80 Hellmers S. (2010). Hydrological Impacts of Climate Change on Flood Probability in Small Urban Catchments and Possibilities of Flood Risk Mitigation. Wasserbauschrift Band 13, TUHH, Hamburg, Germany. Hellmers S, Hüffmeyer N. (2014). Folgen für Kanalnetz und Gewässer. In: Kruse E., Zimmermann T., Kittel A., Dickhaut W. Knieling J., Sörensen C. Stadtentwicklung und Klimaanpassung beispielhaft dargestellt am Einzugsgebiet der Wandse, Hamburg, TuTech Verlag, Hamburg, Germany. Maiwald, H. and Schwarz, J. (2010) Ermittlung von Hochwasserschäden unter Berücksichtigung der Bauwerksverletzbarkeit EDAC-Hochwasserschadensmodell, Verlag der Bauhaus-Universität Weimar, Germany. Manojlovic N (2014) Improving dwellers participation in the developemnt of flood resilient cities, Doctoral Thesis at the Institute of River&Coastal Engineering, TUHH (submitted) Pasche, E.; Ujeyl, G.; Goltermann, D.; Meng, J.H.; Nehlsen, E.; Wilke, M. (2008). Cascading flood compartments with adaptive response, Proc. First International Conference on Flood Recovery, Innovation and Response (FRIAR), London, UK, Juli 2008, ISBN , Publisher: D. Proverbs, C.A. Brebbia und E. Penning-Rouwsell, WIT Press, Southampton, S Rath, S Model Discretisation in 2D Hydroinformatics based on High Resolution Remote Sensing Data and the Feasibility of Automated model Parameterization, Phd Thesis, Hamburg- Harburg University of Technology, Hamburg, Germany Ujeyl et al. (2012). Evaluating direct damages of residential and commercial assets on a micro scale Results of the XtremRisK Project. Proc, 2nd European Conf. on FLOODrisk Management Rotterdam Wahl, T.;Mudersbach, C. & Jensen, J. (2012). Statistical assessment of storm surge scenarios within integrated risk analyses Results of the XtremRisK project. Proc, 2nd European Conf. on FLOODrisk Management Rotterdam 80

81 6 Case study -Mumbai 6.1 Case study area overview Mumbai, an island city having a population of million (2011 census) in an area of km 2 (average population density is 27,209 persons per km 2 ) is located on the Western coast of India. The city has been formed due to merger of eight islands and reclamation from the sea. The city elevation at some locations is just one meter above mean sea level (MCGM, 2011). Mumbai receives an average annual rainfall of 2430 mm, of which 95% falls during the monsoon months from June to September. 60 % of the average annual rainfall occurs in July and August and 50 % of this occurs in just two or three events. The maximum annual rainfall ever recorded was 3452 mm in 1952 (MCGM, 2011). Mumbai was severely flooded on 26 July, 2005 due to the 944 mm rainfall in Mithi River catchment in 24 hours (recorded at Santacruz rain gauge) resulting in the overflow of the Mithi River. However, Mumbai city rain gauge at Colaba recorded only 72 mm in 24 hours. (Figure 6-1). Mithi River catchment (study area) Mumbai India Mithi River Figure Location of Mumbai, Mithi River and study area Following the July 26 th 2005 flooding, various measures, both structural and non-structural, have been implemented by the authorities. Major steps were initiated to restore the capacity of Mithi 81

82 River by widening and deepening the sections. Mithi River has been widened from 7 m to 20 m in upstream, 20 m to 35 m in mid stream and 35 m to m in downstream reaches. A flood wall has been constructed for km on both banks in the upstream reaches of the river. Two weirs have been constructed on the Mithi River to detain flows during high intensity rainfall events. 6.2 Scenarios The impact assessment described in this deliverable is calculated for a single scenario: the presentday situation. Future deliverables will provide the impact assessment results that have been undertaken assuming different future scenarios (D3.8) and different measures and strategies (D3.9). 6.3 Hydraulic modelling The hydraulic modelling of flows in the Mithi River has been carried out using MIKE 11 (described in Section D2.4). The rainfall runoff model and 1-D hydrodynamic model have been used to compute the water levels at various locations along the Mithi River for various rainfall intensities. SRTM DEM has been analyzed and compared with the elevations derived from available topographical maps of Mumbai and available levels in the critical areas from MCGM. These have been supplemented by ground verification by the research team at critical locations. It is assumed that the drains in the catchment will adequately drain the runoff generated from rainfall intensity of 25 mm/h for which the existing system was designed. The overflows from the drains will contribute to flooding along with overbank flow from the river. Rainfall records for were obtained from the MCGM automatic weather station located at Powai in the Mithi River catchment and water levels at one hour interval from Krantinagar MCGM gauging site. Three storm events were selected for calibration, while another one was selected for verification. Three events corresponding to return periods of 1, 10 and 100 years have been simulated in the river. The flooding that occurs in the downstream reach of Mithi River of about 7.50 km (from the mouth) is due to high rainfall coinciding with high tide. Subsequently, flood spread and depth corresponding to these events and scenarios due to backflow from river have been estimated using expert judgement and knowledge gained from site visits. 6.4 Damage/ impact modelling As already mentioned the flooding that occurs in the downstream reach of Mithi River of about 7.50 km (from the mouth) is due to high rainfall coinciding with high tide. This results in damages to goods and properties located in the flood plains. The direct tangible damages have been computed using the flood levels simulated earlier. 82

83 Depth (m) Project Report Direct tangible impacts The vulnerability assessment of direct tangible damages has been carried out using depth-damage curves, flood depth maps and land-use maps. Depth-damage curves (DDC) The depth-damage curves (DDC) for the flood prone buildings in the Mithi River catchment have been developed based on site visits and expert knowledge. Site information has been compiled to develop a data base characterising the contents and structure of residences representative of those located in the flood plains of Mithi River. The DDC have been developed for different types of landuse and building occupancy types. Four different categories of land use/ building types have been identified: (i) slum at ground level, (ii) slum with elevated plinth levels, (iii) building at ground level and (iv) building without elevated plinth. The DDC developed in this study are shown in Figure 6-2. These curves can then be used to obtain damage costs for predefined increments of water depth. Then, multiplying the obtained value by the affected area of the building, the damage cost in the area can be obtained Damage (Rs. per sq.m.) Building at ground level Slum at ground Building with elevated plinth Slum with elevated plinth levels Figure Developed depth damage curves Land-use information The land-use map has been developed using site visits, internet resources and available maps. As the flood typology in the case study area consists of floods due to high tides coinciding with heavy rain, 83

84 land-uses at the ground level in the flood prone area only have been considered as shown in Figure 6-3. Building area Slum area Mithi River Figure Land-use classes For each sub area, there are many land-use types. Multiplying these values by the relative damages obtained from the DDC, the total damages of the areas has been obtained. The vulnerability map has been developed based on the depth damage curve (Figure 6-2) and this shows the maximum potential damage in monetary units and these are shown in Figure 6-4. Indirect tangible impacts The indirect tangible impacts are the ones which have not been created due to the direct contact with water but can be economically assessed. Such impacts are the disruption of businesses activities and transport networks, etc. The impacts induced by floods caused by high tide coinciding with high intensity rainfall are localized, and the retention time of the water is short. In the area studied there is no large business activity or any transport network. Therefore, such damages are not included in this study. Intangible impacts Intangible impacts are not included in this study. 84

85 T= 1 in 1 yr I = 36 mm/h Vulnerability 50 INR/m INR/m INR/m INR/m 2 T= 1 in 10 yr I = 76mm/h Vulnerability 50 INR/m INR/m INR/m INR/m 2 T= 1 in 100 yr I = 111 mm/h Vulnerability 50 INR/m INR/m INR/m INR/m 2 Figure 6-4- Vulnerability map (INR/m 2 ), showing the potential damage of the areas at risk for rainfall events of return period of 1, 10 and 100 years. 85

86 6.4.1 Results from the damage or impact assessment Using the flood spread maps, flood damage maps have been obtained and are shown in Figure 6-5. T= 1 in 1 yr I = 36 mm/h Damage < 1 m INR 1 2 m INR 2 3 m INR 3 4 m INR 4 5 m INR 5 6 m INR 6 7 m INR 7 8 m INR 8 9 m INR 9 10 m INR m INR m INR T= 1 in 10 yr I = 76mm/h Damage < 1 m INR 1 2 m INR 2 3 m INR 3 4 m INR 4 5 m INR 5 6 m INR 6 7 m INR 7 8 m INR 8 9 m INR 9 10 m INR m INR m INR T= 1 in 100 yr I = 111 mm/h Damage < 1 m INR 1 2 m INR 2 3 m INR 3 4 m INR 4 5 m INR 5 6 m INR 6 7 m INR 7 8 m INR 8 9 m INR 9 10 m INR m INR m INR Figure Flood damages (in M INR) in the downstream portion of Mithi River for rainfall events of return period of 1, 10 and 100 years. 86

87 Damage (M INR) Project Report The detailed analysis carried out in this study has enabled the formulation of a methodology to determine the water levels, flood spread and depth, depth damage curve, vulnerability and flood damage. The expected damage estimates will enable the decision makers to decide on the adaptation measures to be implemented. The damage return period curve for the downstream portion of Mithi River is presented in Figure 6-6 and Table 6-1. It provides an estimate of the order of magnitude of the annual damage that may be caused by floods Return period (year) Figure Damage return period curve for Mithi River catchment. Table Damages for the three rain events simulated. S. No. Return period (years) Damage (M INR) Discussion and conclusions A methodology to characterize flood susceptibility using a 1D model has been developed taking the Mithi River catchment in Mumbai, India as a case study. Calibration and validation of the model has been carried out using the rain gauge data and flow depths recorded by water level gauge and reports of flooding in the river basin. 87

88 The depth-damage curves have been developed for the case study area and damage assessment has been carried out. This will enable the determination of the critical areas in the catchment in terms of flooding impacts and help the decision makers to decide on adaptation strategies. This methodology can be extended for determining water levels, flood spread and depth, depth damage curve, vulnerability and flood damage of other flood prone areas in other cities with similar conditions. 6.6 References for the Mumbai case study Gupta, K. (2007). Urban flood resilience planning and management and lessons for the future: a case study of Mumbai, India, Urban Water Journal, 4, 3, Government of Maharashtra, Report of the Fact Finding Committee (FFC) on Mumbai floods, McBean E. A., Gorrie, J., Fortin, M., Ding, J. and Moulton, R. (1988). Flood depth-damage curves by interview survey. Journal of Water Resources Panning and Management, 114, 6, Merz, B., Kreibich, H., Schwarze, R. and Thieken, A. (2010). Review article: assessment of economic flood damage. Natural Hazards and Earth System Science, 10, 8, Nascimento, N., Machado, M. L., Baptista, M. and Silva, A. D. P. (2007). The assessment of damage by floods in Bazilian context. Urban Water Journal, 4, 3,

89 7 Case study Nice 7.1 Case study area overview The Urban Community of Nice Cote d Azur is the inter-communal structure gathering the City of Nice and its suburbs. The Urban community of Nice has four main directions of competences: Development through support of new business and driving economic development Spatial planning based on urban planning and transportation monitoring and improvement (car parks, roads, equipments, etc.) City management managing among waste management, energy management, water management etc. Housing public housing sector The 27 communities belong to Nice cote d Azure urban community with inhabitants covering area of 450km 2. The city of Nice is spread on over 72km 2 hectares with hills, flat parts including the seafront, the central basin and the valleys. Diversity of Nice terrain comes through: plane parts in about 18km 2. They are occupied by the dense urban patterns. The hilly areas (about 47km 2) are occupied by medium density areas mostly residential (individual and collective), agricultural activities and forested areas (Figure 7-1). Figure 7-1: The city of Nice The rain patterns in Nice are characterized with two main periods, during the autumn and early spring. The dry period is during the hot weather from June to October. The average annual temperature of 15 and a mean annual rainfall of 826mm (with recorded maximum rainfall is in November) conceal an uneven distribution of temperature and precipitation during the seasonal cycle Flood related problems in the case study There are four types of flooding likely to happen in Nice. River flooding, Pluvial flooding, 89

90 Flash floods. Coastal flooding Pluvial flooding in this area can occur due to a high precipitation where huge amount of water is placed in a localized area is. All water cannot be stored in drainage system. In this case, rain is the source of the flood: not water coming from a river, but the water on its way to the river. In the city of Nice due to high and fast urbanization the existing drainage system has no capacity to accept rain water from upstream areas. Caused by downstream concept the problem is becoming more dominant nowadays. Additional drainage pipes in order to increase drainage capacity Figure 7-2: Runoff paths in Nice along with the additional drainage pipes In addition to that the additional pipes are constructed while ago with the purpose to increase drainage capacity and minimize the pressure that is created with urbanization on existing pipes. The additional pipes are accepting the rain water form upstream part of the urban catchment and directly transporting to the sea relaxing the drainage pipes in downstream part. Flash floods are likely to happen due to high runoff. Steep slope of hilly areas contributes to very short concentration time. Formed runoff is characterized with high velocities and inability for water to enter the urban drainage system. The coastal line is fully artificial and its function is protecting highly urbanized part of the city along the coast from waves. Actual risk from coastal flooding exists in the sense that only part of a walking path will be affected. Within CORFU project the pluvial flooding is analyzed as well as measures that will increase flood resilience. 7.2 Scenarios The impact assessment described in this deliverable is calculated for a single scenario: the presentday situation. Future deliverables will provide the impact assessment results that have been undertaken assuming different future scenarios (D3.8) and different measures and strategies (D3.9). 7.3 Hydraulic modelling In order to describe flooding in Nice case study are the 2D hydraulic model is used. The existing drainage system in Nice doesn t have the capacity to catch whole runoff. Hence, the surface runoff is 90

91 the main focus for the hydraulic modelling for the Nice case study. The 2D model is developed using MIKE Hydraulic modelling approach and tool The main objective for hydraulic modelling was to simulate surface runoff and estimate flood impact of selected area. One of the main obstacles was terrain configuration. This Mediterranean area is characterized by flat and very steep areas. In addition, the two types of mesh is applied: regular and flexible. This was done in order to test both approaches and estimate accuracy of results. Also, the two different mesh type was used to examine the model stability and behaviour on very steep areas. Both meshes are created with buildings included. The urban bathymetry is created using regular mesh the most interested areas (e.g. Old city area) were more detailed as it shown on the Figure 7-3. For the regular mesh the buildings are included into the topography. Using this approach the accuracy of results with respect to buildings is a bit reduced. Figure 7-3: Urban bathymetry regular grid, Nice case study Software used for this model is MIKE by DHI, for regular mesh the numerical scheme MIKE21. At the end the model with regular grid is used. This was chosen due to the instability that occurred using flexible mesh Hydraulic modelling data Available data for building hydraulic model is accurate topographic data. Since the model focuses on surface runoff the different return periods were used (T10, T50 and T100). In addition, for verification the mapped places with water levels are available. The modelling data are explained in detail in the paragraphs below. Additional data are provided from historical event (river flooding). Three runoff scenarios matching with the hydrological analysis results for runoff assessment from the Nice municipality have been elaborated and tested. These scenarios have in common the fact that they consider the sewer system as non-available and as the area is fully urbanized, infiltration was consider as null over such impervious area. Scenarios are presented in the table below. First scenario (S1) is a three hours long with 30 minute long rainfall event with constant 65,5 mm/h intensity which corresponds to event of 100 year return period. Second scenario (S2) is an a three hours long with 30 minute long rainfall event with constant 59,0 mm/h intensity which corresponds 91

92 to event of 50 year return period. The third scenario (S3) is a three hours long with 30 minute long rainfall event with constant 43,6 mm/h intensity which corresponds to event of 10 year return period. Althrough scenarios are over the same domain they will not be directly compared in this paper. Table 7-1: Nice station, return periods used for models Return period Precipitation intensity mm/h S1=T10 43,6 S2=T50 59,0 S3=T100 65, Historical flood The flood of November 5, 1994 is perhaps the most spectacular event recorded in the lower valley of the Var. The flood came after a rainfall event in its intensity but its spatial extension it affects more than 2/3 of the watershed (Gourbesville 2009). Precipitation is generated by a low pressure front that reached the Var basin from 2 nd November and until 5 th November, the average precipitation volume was over 350 mm within 72 hours. The reason that this flood is significant is the location of airport in the flood zone. The flood map for this event will be used for evaluation of the vulnerability of airport and new constructions in the low Var valley Topographic data In order to present urban system the accurate topographic data are needed. This is in addition of presentation of all the buildings, streets, banks and sidewalks in the case study areas. Stormwater runoff is characterized with relatively small water depths which also need accurate topography. In order to present urban system the accurate topographic data are needed. This is in addition of presentation of all the buildings, streets, banks and sidewalks in the case study areas. Stormwater runoff is characterized with relatively small water depths which also need accurate topography. The LiDAR data set has been gathered for Nice Municipality (DIGNCA) in 2005 during a specific flight with an average flight altitude of 1300 m. The produced airborne LiDAR mapping covers 350 km 2. Average density of laser point is 1 per 1.25 m 2. There over thirty georeferenced markers located over the domain and used for the georeferencing of the dataset. Above ground features such as bridges, elevated roads (overpasses) and tunnels are included in the LiDAR information. The vegetation has been removed from the raw LiDAR signal. This result with DSM a 2 m per 2 m resolution grid with an average horizontal accuracy of 0.3 m and a vertical accuracy of 0.15 m. Figure 7-4 illustrates a part of the LiDAR born DSM whith buildings included. The number of class of elements created as vectorialized features is about 50. Over the area used for this study there is over buidings introduced introduced under vector form information (polygons). The buildings are here very important for the post modelling actions. 92

93 Figure 7-4: High resolution topographic data, Nice, FRANCE DSM and integration in standard hydraulic models DSM may incorporate 3D data taken from photogrametry with more ground features that DSMs based just on LiDAR information. In this report the analysis is done using available LiDAR data. In our case, topology configuration within DSM includes roads and especially overpasses that are creating non realistic simulation results. According to that the necessary modification of DSM is done within the MIKE Zero module. The irregularity reflects on simulation results on areas where elevated road (overpass) exists. In this case the overpass is acting as dam and prevents water to pass through. This results in high water depths upstream of overpasses (Figure 7-6). The extensive investigation of mapping the areas where the overpass has done. For the whole area of 72km 2 the areas are edited. The procedure takes into account surrounding elevation cells and adapting the elevated cells located where overpass is located. As mention above modification of topography is done within the MIKE Zero module. Figure 7-7 shows the edited topography file on the left. Marked areas present edited parts where the elevation cells are adjusted to fit the real elevation values. At this point the building were not included into DSM yet. The insertion of buildings is done later in order to have more realistic presentation of background for hydraulic modelling (Figure 7-5). The building shape file has been transferend to raster. All buildings are set to have the same hight with the resolution set to 4m. At the end the building raster is added to existing DSM with all roads included. This now represents the realistic presentation for hydraulic modeling (Figure 7-7). Introduced in MIKE 21 the adjusted DSM is represented with points with resolution of 4m. It is important to say that DSMs use as computational grid in this form doesn t cause troubles with regular mesh based modelling (Abily at al., 2013). The time investement in this case in mesh creation comes with benefits. 93

94 Figure 7-5: Insertation of buildings into existing DEM Figure 7-6: Located overpasses in DTM file, location SNCF station Nice, France Figure 7-7: Solving the problem of overpasses, location SNCF station Nice, France 94

95 Sensitivity analysis The representation of runoff flowpaths that are influenced by above ground structures (sidewalks, buildings, street gutters, etc) is not equall in DSM generated based on LiDAR data. This is challenging when standard 2D modelling tools are used in terms of data feasibility and integration with modelling tools. The surface inclusion in highly detailed runoff 2D models and its possibilities and challenges ask for specific consideration. On one side the high resolution data is available and on the other the software and hardware limitation has influence on results. Adaptation of resolution with respect to exisiting hardwer and softwer is done. The very accurate topogfraphic data is therefore aggregated from original resolution of 2m to 4m. As mentioned this was needed in order ot set a hydraulic model for 72km 2 using Mike21 regular mesh. a b c Figure 7-8: The example of different density of streamlines for resolutions of a) 2m, b) 4m and c) 10m Further to that, the test is done using resolutions respectevly of 2m, 4m and 10m. The density of streamlines is observed and it is concluded that density varies with respect to different resolution of DEM is available. On figure 8 the three different cases of streamlines are presented. It is visible that density of streamlines changes with respect to resolution. From original 2m resolution the DEM is set to 4m resolution in order to set the hydraulic model using available hardware and software. The differences in number of stream lines is easy to notice and also to present. The number of stream lines when 2m resolution is used is 6812, for the resolution of 4m is 2584 and for the resolution of 10m the number of stream lines is 829. In the table below the presented values in percentenge shows how the resolution is influencing the density of stream lines in chosen part of DEM. Table 7-2: Number of stream lines with different resolutions (2m, 4m and 10m) Resolution Number of stream lines Difference 2m m % 10m % Hydraulic modelling calibration and verification The existing information s regarding flooding in Nice is for river flooding. The existing flood maps consider only river flooding. Also available flood risk management planning considers just river flooding. Available data for validation is the observed flooded areas. The locations are in the old city of Nice and in areas around the railway station (Figure 7-9). 95

96 Var Valley NORTH Priority zones Priority zones Priority zones Figure 7-9: The priority zones for flooding Hydraulic modelling runs and results The two models are created. The one with with regular mesh and another with flexible mesh. After preliminary testing the model with regular mesh is created with MIKE21. This was done due to the instability that model with flexible mesh shown. Following the description of scenarios in Deliverable 1.2 the return periods with 10, 50 and 100 years are simulated. Hydraulic modelling results are shown in figure below. The maps with different return periods are presented. 96

97 Figure 7-10: Flood map for 10 year return period, Nice, FRANCE Figure 7-11: Flood map for 50 year return period, Nice, FRANCE 97

98 Figure 7-12: Flood map for 100 year return period, Nice, FRANCE 7.4 Damage / impact modelling The models are in the calibration phase so the damage assessment will be added later following further development. The expected flood impacts are on infrastructure and residential property. The use of CORFU damage tool is planned for evaluating direct flood impact. The vulnerability assessment of direct direct damages demands depth/damage curves and corresponding land use maps. Following the development methodology for mapping the urban systems (Batica et al., 2012) the land use is presented with urban functions. This methodology is presented with Deliverable 4.3. The Figure 7-13 presents the central city area with adapted land use (Batica et al., 2012). 98

99 Figure 7-13: Mapping of urban system, Nice, FRANCE Figure 7-14: Central city area with adapted land use 99

100 Damage ( /m 2 ) Project Report Depth damage curves (Figure 7-15) are also adapted to the rearranged land use. As mentioned the land use is represented through different urban functions, therefore the created depth-damage functions are following the same procedure Water depth (m) DF 1 education DF 2 food DF 3 governance DF 4 Health DF 5 housing DF 6 leisure DF 7 religion DF 8 Transport DF 9 Work DF 11 Mixed Figure 7-15: Depth damage curves for buildings The depth damage curves created and adapted to fit the primary mapping of urban system on different urban functions (Figure 7-13). Adaptation is done with the data obtained from the report on flood damage functions in EU member states (Huizinga, 2007). The damage functions from France and Germany are from (i) building, (ii) commerce, (iii) industry and (iv) roads. The damage functions for transportation do not applied to buildings that are located on railway and bus stations and buildings located on airport. The analysis of flood damages for Nice case study is presented with a total number of flooded urban functions in analyzed area and the amount of damage calculated using CORFU damage tool. The results are presented in the table 3 below. Flood damage expressed in monetary terms is presented graphically (Figure 7-16) and in Table

101 Table 7-3: Statistics of flood damages Urban Function (UF) Number of flooded buildings 10year return period 50year return period 100year return period Education Food Governance Health Housing Leisure Religion Transport Working Mixed Sum , , , , , , , , , Flood damage per building 10year 50year 100year Education Food Governance Health Housing Leisure Religion Working Mixed Figure 7-16: Direct flood damages per urban function 101

102 Table 7-4: Flood damage in Euro ( ) Urban Function (UF) Total Damage (Euro) 10year 50year 100year Education 192, , , Food 1,123, ,822, ,907, Governance 6,246, ,279, ,165, Health 4,544, ,200, ,310, Housing 21,035, ,902, ,688, Leisure 195, , , Religion 117, , , Transport 10,336, ,455, ,851, Working 16,942, ,219, ,300, Mixed 10,545, ,755, ,853, Sum 71,279, ,328, ,789, Direct tangible impacts Direct tangable losses are estimated using developed damage curves presented on Figure Phisical damage to property is calculated using damage curves Indirect tangible impacts Intangible direct losses are not considered for the Nice case study Intangible impacts Intangible direct losses are not considered for the Nice case study. 7.5 Results from the damage or impact assessment The results of damage assessment are graphically presented in the figure 16 below. 102

103 Figure 7-17: Calculated flood damage for Nice case study for 10, 50 and 100 year return period As presented in Table 7-3 and Table 7-4 the within analysis the flood damage calculation is done for the whole case study area (72 km 2 ). The damages functions are used on mapped urban functions and flood damage is calculated based on flood maps (Figure 7-17) for three return periods (i) 1 year, (ii) 50 year and (iv) 100 years. Expected annual damages (EAD) Defined as the average damage determined from floods of different annual exceedance probabilities over a long period. Within Nice case study the flood damages are determined for three different return periods (i) 10, (ii) 50 and (iii) 100 year. The study period is 100 year covered with three rain events. It is important to stress that only runoff generation was modeled over the whole case study area (72km 2 ). The importance of this section in the report comes from that fact. Since the capacity of the drainage system is considered as not able to receive any part of runoff an assumption has been made for 5 year return period (probability 0.2) there are no direct flood damages. This is presented in Figure 7-18 where the damages are plotted against flood probability for three considered return periods. The calculation of EAD, then follows the equation: 103

104 Damage ( ) 90,000,000 60,000,000 30,000, P (1/a) Figure 7-18: Damage probability curve for the Nice case study The calculated surface represents an EAD and its value is 10,313, This value should be considered as overestimation of flood damages that can occur in Nice, but on the other side it is providing an estimation of a range of values for this area. 7.6 Discussion and conclusions In order to estimate the flood impact in the case study area the hydraulic model is developed taking in account three return periods (i) 1 year, (ii) 50 year and (iv) 100 years. Based on the methodology in Deliverable 4.3 the case study area is mapped according to different urban functions (Figure 7-14). This new land use is coupled with depth damage functions. The depth damage curves are also developed for the mapped urban functions. The flood damage impact is finalized using the developed CORFU damage tool. The flood damage results for corresponding return periods are presented in Figure The damage values with respect to different urban functions are further presented in Table 7-3 and Table References Abily M., Duluc C.M., Faes, J.B and Gourbesville P., Performance assessment of modeling tools for high resolution runoff simulation over an industrial site. Journal of Hydroinformatics, IWA publishing doi : /hydro , p.16. BATICA, GOURBESVILLE A resilience measures towards assessed urban flood management CORFU project, Urban Drainage Modeling conference, Belgrade, Serbia 2012 Batica, Hu 2012 Flood Resilience and Urban Systems: Nice and Taipei Case Studies, FLOODRISK 2012 CONFERENCE, DELFTH, THE NETHERLAND, 2012 Gourbesville 2009 Le Bassin Versant du Var la Crue 1994, Atelier-Table Ronde du GIR Maralpin Fleuves, territoires et infrastructures Regards croisés sur la Plaine du Var, Nice, le 10 novembre 2009 Huizinga, H. J. "Flood damage functions for EU member states." HKV Lijn in water, Lelystad, the Netherlands (2007). INSEE Natinal Institute of Statistics and Economic Studies, FRANCE 104

105 8 Taipei case study This document describes the flood damage assessment of the baseline scenario of Taipei case study. The hydraulic modelling results have been reported in Deliverable 2.4. We considered the flood situations of four return period events in the Central Taipei Area (CTA) to evaluate the tangible and intangible damage. 8.1 Case study area overview Taipei, the capital of Taiwan, is a highly developed city with an average population density 9,881 inhabitants/km 2. The original Taipei City had 10 districts and 1.20 million populations in 1967 and now it has 2.68 million populations in 12 districts. The boundaries of districts have been changed as the city grew. The CTA is the area covered by the original districts except the north part of Zhongshan District which is divided by the Keelung River. The CTA occupies only 23.0% land area but the population density is over than 20,000 persons/km 2 in all CTA districts, and the average density 22,772 inhabitants/km 2 is approximate 3.78 times the average 6,025 inhabitants/km 2 of the non-cta districts. Figure 8-1 shows the land value per square meter in Taipei city. Most of the CTA, the south west part of the city, has a land value higher than 1,650USD/m 2. Figure 8-1 The land value per square meter in Taipei City. 105

106 8.2 Hydraulic modelling The city is located at the downstream floodplain of the Danshuei River and its tributaries such that it often suffers flooding problems. In the CORFU project, we simulated the flooding of 10, 25, 100 and 200 year return period events in the CTA using MIKE Urban model. Details of hydraulic modelling results can be found in the Deliverable 2.4. Figure 8-2 shows the simulated maximum flood extents and depths of different events in the CTA. In general, the Xinyi, Songshan, Datong and Zhongshan districts have more areas that are susceptible to pluvial flooding. In this report, we estimated the tangible damage in the CTA based on the hydraulic modelling results. T=10 T=25 T=100 T=200 Figure 8-2 Flood modelling results of the baseline scenario in the CTA 106

107 8.3 Damage / impact modelling Tangible damage assessment We adopted the CORFU damage assessment tools to evaluate the tangible flood damage in the CTA, which require the hazard information, the building use types and the relationship between the hazard and damage for different land uses (i.e. depth-damage curves). The hazard information, as shown in Figure 8-2, was obtained via hydraulic modelling that is described in Deliverable 2.4. The land uses data were obtained from the government database as shown in Figure 8-3. Figure 8-3 Land zonings in Taipei City. For the depth damage relationships, we considered that human activities, which are related to the types of land use, are the major factor that affects the level of loss once a flooding is occurring. Wang (2003) collected the data of a field survey from the flooded areas in Taipei city, and of flood loss claims for tax relief from the government revenue office after a major typhoon event in Wang associated the damage information with the land use types, which were classified into 107

108 Damages ( $US /m 2 ) Project Report residential, commercial (retailer, service), industrial (manufacturing, wholesaler) and cultural zones, and developed the DDCs as shown in Figure Flood depth(m) Residential Zone Industrial Zone Cultural, School Retail Business Wholesale Business Commerce Figure 8-4 Depth-damage curves for residential, commercial, industrial and cultural zones in Taipei City (Wang 2003) Intangible damage assessment There are three risk analysis types: (1) Qualitative Risk Analysis; (2) Quantitative Risk Analysis; (3) Semi-quantitative Analysis. The qualitative assessment describes the consequences of possibility and severity in a narrative manner. Quantitative risk analysis uses more specific data to provide numerical estimates of the frequency of events, the number of deaths and injuries or the financial loss. Semi-quantitative assessment is based on the value of qualitative analysis, but the data do not directly equate to the actual impact of the extent and frequency of risk (Ye 2009). In this study, the risk to human life is described using the Quantitative Risk Analysis, instead of monetary terms, at the village scale in the CTA. There are two important factors in Quantitative Risk Analysis. One is the hazard value of the village and the other is the social vulnerability index value of the village. Hence, we defined and quantify these two factors and used the risk matrices to calculate risk. Risk matrices are used in risk analysis to express the level of risk. Setting up the risk matrix begins with defining the classification clearly, which is decided by the user. Anbalagan and Singh (1996) argued that the aim of risk assessment is to estimate the extent of damage; the losses can be categorised as the loss of life and the property damage. A risk matrix is a flexible tool to identify the level of risk and simplify complex risk problems. To formulate the risk matrix the user must explicitly define and demarcate the ranks. We applied the Eq. (1) of Quantitative Risk Analysis to evaluated the flood risk, including social vulnerability index (SVI) and hazard that value have to full scale from 0 to

109 (1) Where, Risk j is the risk value of the village, Haz j is the hazard value of the village, and I SVI j is the social vulnerability index value of the village Hazard value of the village (Haz j ) We defined a positive relationship between the hazard value and the depth of flood in the study. Pan et al. (2012) converted the depth of flood to the hazard value for analyzing the flood risk in Tainan City, Taiwan. Since this study focuses on the risk of human life instead of finance damage, the definitions of depth level ranges are modified, as shown in Table 8-1, from the study of Pan et al. (2012). We selected 1.5m as the highest level of flood depth due to that the most people s noses and mouths are swamped when flood depth is more than 1.5m. And in each level the depth is linearly interpolated from Haz j. For example, if the depth is 0.4m, it s means Haz j is 0.6. To utilise the hazard information from the flood maps, we aggregated the flood depth within a village into a single value. However, the simple average depth was found to underestimate hazard if only a minor part of area within a village was flooded. On the other hand, the maximum flood depth tended to overestimate the hazard situation for the whole village. Therefore, we used the average of the top 10% of flood depth data in each village to estimate the Haz j in this study. Table 8-1 Relationship between the depth of flood and the hazard value (Haz j ) Depth (m) Haz j 0.0 ~ ~ ~ ~ ~ ~ ~ ~ Social vulnerability index (SVI) To estimate the SVI, Wu and Chiang (2008) gathered the opinions from experts to select the factors that affected the social vulnerability most, and estimated their weights using the Fuzzy Delphi Method (FDM) and Fuzzy Analytic Hierarchy Process (FAHP). The FDM is a systemic program, improved from the Delphi Method, to express the views of an expert group. The Delphi Method has some shortcomings, as the collection of expert cognition is resource intensive, and the cognition imply an indefinite range, the result will distort the view of the experts. We adopted the FDM and FAHP to determine eight factors, as shown in Table 8-2, that are critical for calculating the SVI in the CTA. However, these factors have different dimensions and units. To aid flood risk management, decision makers require integrated information to highlight the most vulnerable communities that need more resources to mitigate the impact of a hazard. The eight social vulnerability factors have different units so we standardised them as the (average standard score) using Eq. (2) such that the factors can be combined in one equation to determine the overall. ( ),, (2) 109

110 Where, is the value of social vulnerability factor of the village, is the mean of all values, is the standard deviation, and m is number of village (j). Because the effect of each vulnerability factors is different, we set the weights of these factors by using the AHP. The SVI is summed by the and their 8 weighting factors, expressing as:, (3) where, is the weight of factor. For comparing the SVI of the different cities, we normalised into the range of 0 to 1: ( ) (4) ( ) where, is the normalised value, is the SVI of village j, Min (SVI j ) is the smallest value of all, and D is the range of the values. Table 8-2 Weighting factors for calculating the social vulnerability index in the CTA Factors Weight Elderly people (above 65 ages) Children (0~14 ages) Alone living elderly Disabled population Rescue equipment Fireman Social shelters Nursing homes for the elderly Risk level index Once the Haz j and the I SVI j values determined, we can calculate the risk value of Risk j of a village using the Quantitative risk analysis matrix as shown in Table 8-3. The Risk j values are further classified into Risk level using the definition in Table 8-4. When a village has a Risk j value between 0 and 0.1, its Haz j would be less than 0.5 (depths of flood lower than 0.3m), in the case, the village will have very low risk level and the flooding is unlikely to cause causality. For a village with more vulnerable groups (I SVI j > 0.2), Risk j will increase rapidly when Haz j rises, which means the disadvantaged people are less capable to cope with the flooding. 110

111 Table 8-3 Quantitative risk analysis matrix for determining the risk value Haz j I SVI j Table 8-4 A statement of quantitative risk analysis level Risk level Risk j Description of risk 0 0 No risk ~ 0.1 Very low ~ 0.2 Low ~ 0.4 Medium ~ 0.6 High ~ 1.0 Very high 8.4 Damage assessment results Tangible damage assessment We evaluated the pluvial flood damage in the CTA for the baseline scenario using the DDCs and the flood maps as mentioned in the previous section. Table 8-5 and Figure 8-5 show the flood damage of different return period events. The total flood damage for 10 year event is 21 million USD and most of them occur in Xinyi district. The damage increases to 42 million USD for 25 year event. For 100 year event, more districts including Datong, Zhongshan, Songshan and Xinyi are serious flooded and the total damage in the CTA is 87 million USD. For 200 year event, Datong, Zhongshan, Songshan and Xinyi districts have the worst damage and the total is 116 million USD. Table 8-5 Average expected annual damage assessment in CTA Return period (year) Flood damage (million USD) Average expected annual damage

112 T=10 T=25 T=100 T=200 Figure 8-5 Pluvial flood damage in the CTA of different return period under the baseline scenario Intangible damage assessment Social vulnerability index map To assess the intangible damage, the spatial information including land-use, zoning, administrative boundary, and digital terrain were gathered. The related census data, including data on demography, vehicles, commercial and industrial activities were also collected for exposure analysis. These collected census data were aggregated by the census tract, the basic unit of city administration. These aggregated data were disaggregated into the 40 m 40 m cells such that they can be overlaid with the flood hazard maps for analysing. Land-use and zoning information were used to assist the disaggregation process. Figure 8-6 shows the SVI map of the CTA for the baseline scenario. Wanhua and Wenshan Districts have higher SVI than the other districts due to high ratios of the elderly people. Then highest value of occurs in Wanhua District ( = 0.98), where has not only has high ratios of the elderly people but also more alone living elderly and the disabled population. And the lowest SVI occurs in in Zhongzheng District, where has more firefighters, rescue equipment and rescue resources than other districts. 112

113 The SVI is a relatively relationship. A region with higher value of vulnerability than other regions indicates that the region has more disadvantaged groups, less equipment or resources (such as firefighters and shelters, etc.) to cope with flooding. Once the information is analysed together with the hazard information, the areas with higher flood risk would be highlighted and more flood mitigation efforts should be applied to these vulnerable areas. SVI Kilometers Figure 8-6 Social Vulnerability Index map of baseline scenario in the CTA Results of risk map assessment Figure 8-7 shows the risk level maps of the baseline scenario. In order to prevent the seriously disaster that must be establishing protection and implementing the mitigation measures to safe residents in advance, thus to calculate how many the population live in the high-risk (level 4 and level 5) regions for each scenario in this case study, as shown in Table 8-6. That method would appear the result if the urban has implement the adaptation measures or not, then it would propose the better measures further to safe those residents. Obviously Wenshan and Xinyi District are higher than other districts, due to the high ratios of both SVI and Haz j are higher than 0.7 (the depth of flood is above 1.0m at some villages). And the seriously risk region is in Wenshan District, no matter what the scenarios of the villages are at very high risk regions (level 5), due to the high ratios of the SVI and Haz j are higher than 0.9 (the depth of flood is above 1.0m). Although the hazard did not happen in some high-vulnerability regions, it can t indicate that the hazard will not happen in future. If the storm events will happen in the future or the poor drainage system will cause the flood occurring in some high-vulnerability regions, which also shows the risk in the regions. Table 8-6 Populations living in high flood risk regions of the CTA Return period Total population 113

114 10-yr 25-yr 100-yr 200-yr in Taipei City *unit: thousand 10 year 25 year 100 year 200 year Figure 8-7 Risk level map of the baseline scenario RISK Discussion and conclusions In the study, we adopted hydraulic modelling to obtain the flood potential information for various rainfall conditions. For the flood damage, we reviewed the literatures regarding to the flood damage assessment and collected information from related researches in Taiwan to analyse the relationships among direct flood damage, flood frequency, flood depth and land use. The tangible flood damage was evaluated using the hydraulic modelling results and the CORFU flood damage assessment tool. We also developed the methodology to evaluate the intangible flood damage at village level by considering the disadvantage groups and the available resources. The risk level maps were produced to identify the areas with higher flood. The following development of flood resilience measures should consider those vulnerable areas as the priority. 114