Risk Assessment for Floods Due to Precipitation Exceeding Drainage Capacity

Transcription

1 Risk Assessment for Floods Due to Precipitation Exceeding Drainage Capacity November 2006 Umut Karamahmut Faculty of Civil Engineering and Geosciences i

2 i. Abstract Studies on flood risk modeling were concentrated on floods caused by breaches of dunes and levees. Another kind of flood which was not considered in risk calculations was floods due to precipitation exceeding drainage capacity of low lands. As a result of the increase in the extreme precipitation events due to climate change and increased land value, the risk due to this kind of floods increased considerably, and must be calculated. This study aims to investigate and improve current situation in risk assessment of floods due to rainfall exceeding capacity of the drainage system of polders. In order to achieve this, commercially available models were investigated to find out if any of them are capable of calculating risk for these floods. Research on existing models showed that none of these models were applicable for this problem. Calculation of risk for this kind of floods comes along with massive work load. In order to able to carry on these calculations the problem must be simplified by eliminating one o the parameters. In order to validate this simplification, correlation between two flood parameters namely, flood depth and flood duration were proved. Finally applicability of a risk analysis tool for this problem was investigated with a case study on Polder Berkel. Results showed that risk analysis methods were applicable to the case but some improvements were necessary. ii

3 ii. Acknowledgments I would like to express my thanks to Elgard van Leeuwen and Olivier Hoes for their constant supervision and valuable comments through out my studies. I also would like to thank to Nick van de Giesen, Elgard van Leeuwen and Olivier Hoes for taking part in my graduation committee. I appreciate contributions of Colin Green and Edmund Penning-Rowsell from Flood Hazard Research Centre, Middlesex University, United Kingdom, Roy Leigh from Natural Hazards Research Center, Macquarie University, Australia and Duncan Faulkner from JBA Consulting Engineers & Scientists and all WL Delft Hydraulics employees who were always there to answer my questions and support me. The last but not the least I would like to thank to my family and friends, without their support none of this would be possible. Especially to my mother, for holding up to life. iii

4 Table of Contents 1. Introduction: Flooding Problem Objectives Report structure Research on existing flood risk models Introduction Basics of flood risk estimation Existing Flood Loss Estimation Models Evaluation Conclusion Correlation of flood depth and duration for different soil types Introduction Methodology Model Schematization Model Data Post processing of simulation results Results Evaluation & Conclusion Case Study: Polder Berkel Introduction Polder Berkel WB21 Method Risk Model Case Discussion & Comparison Conclusions Conclusions & Recommendations Conclusions Recommendations References Appendix iv

5 List of Figures Figure 1-1 Inundation map of Netherlands without dikes, dunes and pumping stations... 1 Figure 2-1 Water Surface Profiles Plot Figure 2-2Depth-Percent Damage Functions For Apartments Figure 2-3 Scale levels of damage evaluation Figure 2-4Property Damages Output of MDSF Figure 2-5Components of FloodAUS Figure D representation of flood extend Figure 3-1SOBEK Model Schematization Figure 3-2Ernst Drainage Calculation Parameters Figure 3-3Drainage Coefficients Input Screen Figure 3-4Visual Basic script for determination of events Figure 3-5Output file view of the script Figure 3-6Depth Duration graph for Sand Average (average) Figure 3-7Depth Duration graph for Sand Average (maximum) Figure 3-8Trend lines of different soil types (average) Figure 3-9Trend lines of different soil types (maximum) Figure 4-1Sub-polders and target elevations Figure 4-2Satellite Image of Polder Berkel Figure 4-3Damage Function for Greenhouses and Urban Areas Figure 4-4SOBEK Model for Polder Berkel Figure 4-5WB21 Script Output File View Figure 4-6Digital Elevation Map Figure 4-7Water Compartments Figure 4-8 Land Use Map Figure 4-9Risk Model Damage Functions Figure 4-10Hymstat Output Figure 4-11Risk Map v

6 List of Tables Table 2-1 Damage categories... 8 Table 2-2 Inundation parameters... 9 Table 2-3Stage-Damage relations for residential properties Table 2-4Damage categories for commercial properties Table 2-5Stage-Damage relations for commercial properties Table 2-6Coverage of existing flood loss estimation models Table 3-1Unpaved node parameters Table 3-2Ernst coefficients for different soil types Table 3-3Coefficient of Determination for different soil types Table 3-4Correlation Coefficient for different soil types Table 4-1Maximum damage per hectare for different land use Table 4-2Workability coefficients for seasons and soil types for grassland Table 4-3Workability coefficients for seasons and soil types for agriculture Table 4-4Workability coefficients for high quality agriculture and horticulture Table 4-5Drowning coefficients Table 4-6Risk calculated by Risk Model and WB21 method Table 4-7Monetary Difference and Ratio between Risk Model and WB21 method vi

7 1. Introduction: 1.1. Flooding The Netherlands, being located in delta of The Rhine, The Meuse and The Scheldt, has a long history in coping with floods. As a result of past water management practices, land reclamation and subsidence, higher percentage of The Netherlands lies on large flat plains under mean sea level. Thus they require both protection from sea and constant drainage of the excess water out of the polders. (See Figure 1.1) Figure 1-1 Inundation map of The Netherlands without dikes, dunes and pumping stations Source: Hoes,

8 Studies on flood protection and flood damage modeling were mostly concentrated on the floods caused by breaches of dunes and levees since a flood resulting from these would be sudden and extensive and combined effects may be catastrophic but recently attention was also given to floods due to precipitation exceeding capacity of the drainage canals and pumping stations of polders. This kind of flood is neither life threatening nor as catastrophic as the floods due to breaches of dunes and levees but they might occur rather frequently resulting in substantial losses. (Hoes, 2005) Both total annual precipitation and extreme precipitation events are following an increasing trend especially in the last two decades. It is believed that this trend will continue due to climate change and further more floods events will be more frequent because of sea level rise and subsidence.(ipcc, 2001) Increase in frequency and magnitude of these events once again showed that regional rainfall induced floods can not always be prevented. On the other hand possible losses due to these events are also escalating because of increasing value of land and on going urbanization. In order to avoid these losses many water systems must be upgraded. Risk of flooding must be calculated in order to asses the feasibility of the measures taken to upgrade these systems. 2

9 1.2. Problem Commercially available models did not focus on this kind of floods but with current increase in risk these floods must also be covered. On the other hand risk estimation for this kind of floods is rather difficult. For risk assessments of river and sea floods in low lands, structures are assigned a failure probability then the risk can be determined by multiplying this probability with the possible damage that failure of this structure will cause. Total flood risk is the summation of risk values of all sections and structures. Not like river and sea floods, for a flood caused by precipitation exceeding drainage capacity, failure is not limited to one section or structure and also there is not only one failure probability for a section. Failure probability differs from frequent floods with small damages to low frequency floods with a higher damage and this probability distribution is dependent on elevation of each pixel. In other words both probability and damage are spatially distributed. This makes risk assessment much more difficult.(hoes, 2005) Total risk for this kind of flood is summation of all multiplications of probability and damage. Calculation of rainfall induced flood risk has a vast workload due to the fact that probability of occurrence and damage in case of occurrence is spatially distributed. In order to be able to estimate the risk, this work load has to be reduced. In order to achieve this one of the parameters used in calculations can be excluded but this can be done only if the excluded parameter will be represented inclusively by the other parameters (ie. If there exists a correlation between them). 3

10 1.3. Objectives This study aims to investigate and improve current situation in determination of risk of floods in low lands due to rainfall exceeding drainage capacity. In order to achieve this, following objectives will be studied through out the study. - To figure out if any of the commercially available models are capable of solving this problem considering the different nature of rainfall induced floods in low lands. - To prove the correlation between flood depth and flood duration. This correlation is rather important because proof of such a correlation will allow us to eliminate one of these parameters, reducing the vast workload and enabling us to calculate risk. - To investigate the applicability of a new risk analysis tool for calculation of risk for rainfall induced regional floods in low lands. 4

11 1.4. Report structure The above mentioned objectives were addressed in different chapters as described below. In chapter two current practices and models in three countries namely as United Kingdom, United States and Australia were investigated in order to figure out if any of the commercially available models were capable of solving this problem. Existing models were not capable of carrying out this calculation. In the third chapter correlation between flood depth and duration was proved by simulating water levels for a long enough period for 12 most common soil types in Netherlands. In the fourth chapter a case study was carried out in order to investigate the applicability of a new risk analysis tool for calculation of risk for rainfall induced regional floods. Annual average risk calculated by this tool was compared with the risk value calculated by a traditional damage assessment method. Rationales behind risk calculation were investigated in order to reach a balance between workload and accuracy. The study was concluded and recommendations for further studies were given in chapter five. 5

12 2. Research on existing flood risk models 2.1. Introduction In this chapter applicability of existing flood risk models to the case of floods in low lands due to precipitation exceeding drainage capacity will be investigated by studying working principles of 6 models used in 3 countries. These models and countries are as follows. United States : - HAZUS - MH - HEC-FDA United Kingdom : - ESTDAM - MDSF Australia : - ANUFLOOD - FloodAUS Studies showed that current flood risk modeling practices in different countries are not applicable for modeling of flood risk due to high precipitation which exceeds the capacity of the drainage system. Reasons why they are not applicable will be mentioned further on. In following section basics of flood risk analysis and common practices will be mentioned. In third section an overview of the flood risk estimation methods used in different countries will be given. Evaluation of applicability of these methods to the case of concern will be carried out in forth section and conclusions will be given in fifth section. 6

13 2.2. Basics of flood risk estimation In this section, common practices used in different models thus different countries will be mentioned. In all of the models flood risk is defined as the sum of multiplication of damage in case of occurrence of events and probabilities that those events will occur. In order to calculate risk inundation maps with known occurrence probabilities were used. Flood damage and risk were categorized in different ways. These are as follows Type of flooding Damage and risk can be categorized according to the source of flooding. A flood might be caused by sea, river or precipitation. The source of flooding effects damage in various ways. For example a sea flood will have extra damage on agricultural fields due to salination. Also source of flooding changes calculation method for risk of flooding Categorization according to consequences In general there exists there different criteria to classify damages caused by natural disasters. First division is between tangible and intangible damages. Tangible damages are those which can be described in monetary units, thus they can be evaluated and compared. Damages to buildings or contents of buildings can be an example to tangible damages. Intangible damages are the ones which is difficult to describe in economic terms, for example physical and psychological traumas. Recently more studies are being carried out for quantification of intangible damages. An exemplary is Anxiety- Productivity and Income Interrelationship Approach (API). This approach is explained in detail elsewhere (Lekuthai, Vongvisessomjai, 2001) 7

14 Another division is direct and indirect losses. Direct losses are caused by physical contact of flood water while indirect losses are caused through interruption and disruption of economic and social activities as a consequence of direct flood damages. (Dutta et al.,2001) Destruction of buildings is a direct damage while production loss is an indirect damage. Direct, indirect and tangible, intangible damages can further be divided as primary and secondary damages. The table below can be an example for the division of damages according to the above mentioned criteria. Table 2-1 Damage categories Category Tangible Intangible Primary Capital Loss (houses, crops, Victims, ecosystems, pollution, Direct cars, factory buildings) monuments, culture loss Social disruption, emotional Indirect Production losses, income loss damage Secondary Production losses outside the flood area, unemployment, migration, inflation Emotional damage, damage to ecosystem outside the flood area Induced Costs for relief aid Evacuation stress Source: K.M. de Bruijn, 2005, pg. 41. Ideally all these kinds of damages should be considered in estimations but in practice this is impossible. In most of the studies damages are restricted to primary tangible damages and part of the production losses by companies and agriculture. This is due to the complexity of calculating secondary tangible or intangible damages. These damages are introduced in bulk form by multiplying the primary tangible damage by a factor which is dependent to the properties of the region. 8

15 Effective parameters Adequate determination of flood parameters is also crucial for loss estimation. A list of inundation parameters is given in Table 2.2. Most important parameters are flow velocity, duration and depth of flow. For most cases only parameter used in models is the flow depth. This is an acceptable simplification since flood depth and duration are closely related to each other. In other words if the flood depth is high then it will take more time to drain the flood plain thus the flood duration will be longer. Table 2-2 Inundation parameters Inundation parameter Relevance Area Determines which elements at risk will be affected Depth Has the strongest influence on damage Duration Influence on damages on building fabric Velocity Only high velocities will lead to increase in damage Risk rate Influence on damage reducing effects of warning and evacuation Time of occurrence Important on agricultural products Contaminations Contaminations and loads may increase damage significantly Salt / Sweetwater Salt water can increase damages in coastal areas Source: Penning-Rowsell et al., 2005 Most of the models follow the unit loss approach for estimates. Unit loss model is based on unit by unit assessment of potential damage and summation of these possible damages gives the total expected damage. Success of loss estimation models mostly depends on the establishment of the relation of the damage with flood parameters. This is done with so called stage-damage functions which define the possible damage percentage for a given value of flood parameter. These functions are derived according to historical loss data, questionnaire or results of experiments. Potential damage for a given stage is found by multiplying the percentage corresponding to that stage value with the value of the structure. In common practice above mentioned principles are used in models but there exists some differences between methods and models developed in different countries. These methods, models and differences between them will be mentioned in next chapter. 9

16 2.3. Existing Flood Loss Estimation Models Different damage assessment models were developed in different countries. These models are mainly built for cost efficiency studies of flood mitigation measures or assessment of risk for insurance purposes. In this section different models used in United States, Australia and United Kingdom will be mentioned United States: In United States a variety of organizations are involved in damage assessment and prevention. As a result no standard method has been developed. (K. de Bruijn, 2001) There are two commonly used models, HEC-FDA and HAZUS-MH HAZUS-MH: The name of the model stands for Hazard United States Multiple Hazards. HAZUS was initially developed for assessment of earthquake damages by Federal Emergency Management Agency (FEMA). Later FEMA released a newer version by which a variety of hazards, including floods, and their risk assessments may be investigated. HAZUS is a flexible program that allows performing the analysis on different levels depending on resources and analysis needs. Level 1 uses available hazard and inventory data provided by HAZUS-MH, limited additional data is required in this level. Level 2 analyses require local data which is readily available for most of the cases or can be converted to model requirements easily by Flood Information Tool (FIT, a built in function of the model for conversion of data). Level 3 involves adjustment of built in loss estimation models. 10

17 Loss estimate analysis can be run for three different analysis options. These options are; (1) multiple return periods of 10, 50, 100, 200 and 500 years, (2) a user defined single frequency or (3) annualized loss. For comparison of flood mitigation measures third option will be most adequate. (FEMA, 2004) Although the model gives a quick estimate of the possible damage, results will not be accurate enough unless the model is run on third level, which requires aggregation of detailed local data and adjustment of loss estimate models HEC-FDA: In United States, US Army Corps of Engineers (USACE) has nationwide responsibilities on water resources planning and management. (Dutta et al.,2001) Thus for flood mitigation measures USACE produced its own guidelines namely as the National Economic Development Procedures (USACE,1988) and The Hydrologic Engineering Center (HEC) designed the Hydrologic Engineering Center s Flood Damage Analysis (HEC-FDA) program in order to assist risk-based analysis methods for flood damage reduction studies as required by USACE. HEC-FDA uses Monte Carlo simulation, a numerical model that computes the expected value of damage while explicitly accounting uncertainties in basic functions. It can quantify the uncertainty in discharge frequency, stage discharge, geotechnical levee failure and stage damage functions and incorporate these into economic and performance analysis of alternative flood damage reduction plans. Evaluations are carried on in terms of expected annual damage equivalent annual damage or project performance. (USACE, 1998) 11

18 Model uses water surface profiles and depth damage functions for calculating damage and risk. Water surface profiles can be discharge or stage based. A data set must contain eight profiles. These are defined as 0.50, 0.20, 0.10, 0.04, 0.02, 0.01, and exceeding probability flood events. Profiles can be used for developing with or without project condition functions. They are also used to from stage-damage functions. An example plot of water surface profiles was given in Figure 2.1. (Burnham, 1997) Figure 2-1 Water Surface Profiles Plot Depth-percent damage functions can be assigned for each occupancy type. Program allows user to define three types of depth-damage functions namely as Structure, Content and Other. These functions can be calculated according to historical loss data, questionnaire or experimental results. Some depth-percentage damage functions used in a case is given below. (See Figure 2.2) The methodology adopted is very comprehensive for estimation of damage to urban buildings and to agriculture. However no specific methods have been developed for estimation of damage to lifeline systems and indirect losses such as interruption losses. (Dutta et al., 2001) 12

19 Figure 2-2Depth-Percent Damage Functions For Apartments (Left: structure, Right: Content) United Kingdom: In United Kingdom it is mandatory to use a standard approach for flood damage assessment for local authorities which want the assistance of central government with flood mitigation measures. Flood Hazard Research Center (FHRC) in Middlesex University had been leading the studies for development of flood damage estimation methodologies on UK. (Dutta et al., 2001). FHRC published 4 manuals presenting results of their studies. The Blue Manuel (Penning-Rowsell and Chatterton, 1977) covers assessment techniques and provides a range of depth-damage data. The Red Manuel (Parker et al., 1987) provides depth-damage data and assessment methods for common indirect losses and direct losses except the residential losses were also covered in this manual. The Yellow Manual (Penning-Rowsell et al., 1992) covers the effects of coastal erosion and assessment of environmental effects of floods. Finally FHRC 13

20 published the Multi-Coloured Manual (Penning-Rowsell et al., 2003). This manual is called Multi-Coloured since it combines the techniques mentioned in previous manuals. It covers flood alleviation benefits, indirect benefits and coast protection and sea defense benefits in an improved and updated manner. In UK an object oriented hierarchical method is used for flood damage estimation. A methodology is selected according to size of the area under investigation and precision required from the study. Three different approaches were recommended according to size of area and precision namely as; macro scale, meso scale and micro scale damage evaluation (See Figure 2.3). Each method recommended for respective scale differs in terms of data requirements, damage categories considered, inundation characteristics needed, land use data, value assets, damage functions, damage calculation and presentation. (Penning-Rowsell et al.,2005) Accuracy Size of Area under Investigation micro scale local meso scale regional macro scale (inter-)national Effort, Costs/ Unit of Area Source: Meyer 2001, p. 30; Reese 2003, p. 54 Figure 2-3 Scale levels of damage evaluation 14

21 It can be observed that in United Kingdom damage functions published in the Multi- Coloured Manuel from FHRC build the basis of damage evaluation studies. For small scale project appraisals the full detail of the database is used. For meso and macro scales more aggregated damage functions are used. (Penning-Rowsell et al.,2005) This set provides synthetically derived depth-damage functions for 100 residential and more then 10 non-residential property types. For residential flats, first a definition and inventory of this standard property type is done. Secondly, for each of its typical building fabric and inventory components the monetary value is determined. Thirdly, expert assessors estimate the susceptibility of each item to inundation depth so depth-damage functions can be constructed. For non-residential properties surveys are carried out, in which responsible persons in each firm are asked about the value of assets at risk and susceptibility of these assets to inundation depth. From survey results average depth-damage curves per square meter of property are derived for different economic branches. (Penning Rowsell et al., 2003) These damage functions not only consider the inundation depth but also they consider duration of flooding (i.e. more or less than 12 hours), coastal flood or not (i.e. salt or fresh water), if a warning more than two hours is received. Two models used in United Kingdom will be mentioned briefly. 15

22 ESTDAM: ESTDAM non-gis based model developed by FHRC. It is mostly used in micro scale studies for project appraisals. It applies a property by property approach and it is matched with the standard depth-damage data. It first calculates the depth of flooding in each individual property from the output of flood extent model. For each individual property it has the details of land use classification data. So once the depth of flooding in the property is determined it looks up the depth-damage function for relevant land use class and can calculate the flood damage at that individual property. (Penning-Rowsell E.C. et al., 1987) Depth-damage functions published in the Multi-Coloured Manuel are used to the full extend in this program. It also calculates the loss-probability curve and hence calculates the risk and present value of benefits. But it must be kept in mind that ESTDAM was developed in midseventies. Since the economic functions are not up to date, nowadays tendency is taking the event losses from ESTDAM output and calculate these values with more sophisticated, dedicated programs MDSF: MDSF stands for Modelling and Decision Support Framework. It was developed in 2001 to support Catchment Flood Management Planning by a consortium of organizations which was founded by Department for Environment Food and Rural Affairs (DEFRA) and the Environmental Agency, led by H R Wallingford and including Halcrow, the Centre for Ecology and Hydrology at Wallingford and the FHRC at Middlesex University. (Defra, 2003) 16

23 MDSF was designed as customized GIS tool to work with ArcView. MDSF is not a decision making tool and it does not contain a hydraulic model. It was designed as a decision support framework, providing common approaches and tools for assisting determination of flood management options at broad scale. It is particularly strong in assessment of the economic and social effects of flood management policies (Defra, 2003). As common practice in UK, it uses the depth-damage functions provided by the Multi- Coloured Manual. On the catchment level it uses only one sector average function for residential properties and ten for non-residential properties. Functionalities provided by the software can be listed as flows (Defra, 2004), - Facilitates for managing and viewing spatial data. - Assessment of flood extend and depth. - Calculation of economic damages due to flooding. - Calculation of social impacts due to flooding including the population in flood risk area and their social vulnerability. - Economic assessment of erosion losses. - Presentation of results for a range of Cases to assist the user in the selection of the preferred policy. Each case is a combination of climate scenario, land use scenario and flood management option. - Procedure for estimating uncertainty in the results. - Framework for comparing flood damages and social impacts as an aid to policy evaluation. - Archiving of cases. 17

24 Powerful visualization of results in GIS environment is a major advantage of the software since it makes the communication and comparison of the results much easier and more understandable for policy makers. A property damage map and tabulation is shown in figure 2.4. Figure 2-4Property Damages Output of MDSF Australia: A recent research in Australia suggests that there is no standard approach for flood damage assessment in Australia. (Dutta et al., 2001) Nevertheless, Department of Natural Resources and Mines (NR&M) published Guidance on the Assessment of Tangible Flood Damages in September This guidance will be explained in the remainder of this section. 18

25 NR&M recommends adopting the stage-damage curves developed for ANUFLOOD. The curves for this flood damage model were developed for a range of building types and sizes. They cover residential buildings for a range of property size and commercial buildings for a range of contents and size. Flood damages can be estimated in 5 steps according to the guidance (NR&M, 2002). 1. Identify flood-affected properties and the likely height of inundation. Flood extend maps provides information about the locations of properties that might possibly be effected from a flood. In order to be able to use stage-damage curves an inundation depth must be estimated. This is done by simply subtracting ground height (site survey or existing maps) and floor level (building approval record) from the flood height (predicted by flood model). 2. Select appropriate stage-damage curves for determining potential direct damages. In this guidance there exist 3 curves for residential properties classified according to their sizes. Commercial properties are divided according to their size and branch of commerce. Details of these curves were given in Table 2.3, Table 2.4 and Table 2.5. Table 2-3Stage-Damage relations for residential properties Source: CRES, 1992, ANUFLOOD: A Field Guide, prepared by D.I. Smith and M.A. Greenaway. 3. Apply stage-damage curves to estimate potential direct damages from flooding. Application of stage-damage curves is simply finding the relevant stage-damage curve and interpolating the respective damage according to the inundation depth. 19

26 4. Estimate indirect losses. In common practice indirect losses are estimated as a percentage of direct losses. ANUFLOOD model uses 15% of direct losses for residential properties and 55% for commercial properties. 5. Calculate total (direct and indirect) damages. Total damage is summation of direct and indirect damages. Table 2-4Damage categories for commercial properties Source: CRES, 1992, ANUFLOOD: A Field Guide, prepared by D.I. Smith and M.A. Greenaway. 20

27 Table 2-5Stage-Damage relations for commercial properties Source: CRES, 1992, ANUFLOOD: A Field Guide, prepared by D.I. Smith and M.A. Greenaway. 21

28 For economic assessment of flood mitigation projects results must be given in terms of average annual damages (AAD). Calculation of AAD requires potential damage bills of a number of flood sizes with different occurrence intervals. AAD can be calculated in 4 steps. 1. Estimate potential damage costs from a range of flood sizes. 2. Plot graph of potential damages versus annual exceedance probability. 3. Calculate annual average damage costs from flooding. (i.e. the area under the damage vs. probability graph) 4. Calculate potential reduction in annual average damage from flood mitigation activities. Two models are distinguished in Australia. First was is ANUFLOOD, developed by Center for Resource and Environmental Studies (CRES) at Australian Natural University (ANU). Macquire Researc Ltd. purchuased the intellectual rights of ANUFLOOD on behalve of Natural Hazards Research Centre (NHRC) in order to modify it for insurance purposes and they release FloodAUS. Both ANUFLOOD and FloodAUS performs the above mentioned procedures. Both models will be mentioned briefly ANUFLOOD: ANUFLOOD was developed during 1980 s and early 1990 s by David Ingle Smith and Mark Greenaway. It is an interactive program designed to assess tangible urban flood damage. (Penning-Rowsell E.C. et al., 1987) 22

29 Input information includes building-by-building description of location, ground and floor heights, construction material, value, house size number of storey and so on. Flood frequency input to ANUFLOOD uses a listing of flood stages expressed as probabilities. Stage damage curves are provided for three residential properties with a further set of commercial property subdivided by size and susceptibility of contents to flood damage. Program also allows the user to input stage-damage curves. Inputs and processes of ANUFLOOD can be listed as follows. (Penning-Rowsell E.C. et al., 1987) FloodAUS: FloodAUS is a GIS based risk rating tool developed by Risk Frontiers to estimate mainstream flood risk in urban areas on a per address basis. Model uses the following information to estimate flood risk: - Digital terrain models - Flood surface elevation information - Property street address databases Source: Risk Frontiers, 2002 Figure 2-5Components of FloodAUS 23

30 Information about extend and depth of flooding is achieved by combining the DTM and flood surface. Figure FFF shows inundated areas for a 100-year flood in New South Wales. Dark blue represents deep water and light blue shallow water. Source: Risk Frontiers, 2002 Figure D representation of flood extend The main output is a database of street addresses, each with a flood risk rating. FloodAUS provides estimates of Average Recurrence Interval (ARI) of inundation at ground level, 1 meter above ground level and 2 meters above ground level (Risk Frontiers, 2002). 24

31 2.4. Evaluation In the section above flood risk assessment methods and models in different countries were investigated. It was observed that these methodologies vary largely in different countries. For example determination of infrastructure damages is covered in detail in the Multi-Coloured Manual in United Kingdom but in Australia infrastructure damage assessment seems to be limited while in United States it is not covered at all. In other terms in these countries depth-damage curves for rural areas are not considered. Components covered in different methodologies where tabulated in Table 2.6. Table 2-6Coverage of existing flood loss estimation models Damage Categories United States United Kingdom Australia Urban Damage Residential detail detail detail Non-residential detail detail detail Rural Damage Crop damage rough rough rough Farmland detail detail none Fishery none detail detail Infrastructure System damage none detail rough Service loss none detail rough Business Loss detail detail detail Environmental Damage none detail none When risk estimation models were observed it was noted that all models use unit loss model. In other words they calculate the possible damage on a property-by-property basis. Risk is calculated in all models by finding the possible damage for different flood magnitudes and then weighting them with occurrence probability of respective floods. Possible damages were found either by using absolute depth-damage curves or using relative depth-damage curves and multiplying the damage percentages with the value of assets. 25

32 It was also observed that all of the damage models were mainly developed for urban damages. Rural damage functions were not considered in so much detail. In cases where crop damage was considered, damage functions did not consider effects of high groundwater levels. In other words depth-damage functions were plotted starting from ground level. But in real life effects of high groundwater levels on crop damage are known and must not be neglected. All above mentioned models were developed for river and sea floods. Calculation of risk for these kinds of floods differs from calculation of risk for floods occurring due to precipitation exceeding drainage capacity. While assessing risk for sea and river floods, structures are assigned a failure probability and risk is calculated as the product of this probability and possible damage that will be caused if the structure fails. In floods due to precipitation exceeding drainage capacity failure is not limited with one structure and failure probability is not constant. Failure might occur frequently with small damage and with a high damage but with lower frequency. Thus flood damage must be calculated over Probability-Density function. Also probability depends on elevation of each pixel and it is spatially distributed. Above mentioned models are not capable of assessing risk when both probability and possible damage are spatially distributed. 26

33 2.5. Conclusion Common practices in flood risk assessment in different countries and different flood risk models were investigated in order to find if any of these existent models are applicable to the problem of assessment of flood damage due to precipitation exceeding drainage capacity. The following conclusions were drawn. 1- All of the models studied were developed for floods caused by breaches of dunes and levees and were not able to calculate risk for floods caused by precipitation exceeding drainage capacity due to flowing reasons. - These models calculate damage for several inundation maps with known probabilities. But such a match of probability and inundation map for rainfall induced floods in low lands is not possible. - For this kind of floods failure is not limited to one section. Meaning, failure probability differs from frequent floods with small damages to low frequency floods with higher damage. - Probability is also dependent on the elevation of the pixel. As a result probability will be spatially distributed. Current models are not capable of calculating risk for spatially distributed probability functions. - Existing models do not cover effects of high ground water levels. 27

34 2- A new model must be developed that will be capable of handling the calculations due to the spatially distributed nature of probability and damage data. A GIS based model would be appropriate for this case. Probability and damage functions can be modeled by two separate grid layers. This way risk can be calculated by unit loss approach in terms of grid-by-grid consideration of risk. 3- Effects of high groundwater levels must be included in damage functions. As current models were developed mostly for urban damage, these effects were ignored. But for this kind of floods rural damage has a higher importance and effects of high groundwater levels can not be ignored. 28

35 3. Correlation of flood depth and duration for different soil types 3.1. Introduction As mentioned earlier, risk calculation for floods due to rainfall exceeding drainage capacity differs from river and sea floods. In the second one failure probability is constant but in the first case failure might occur frequently with small damage or less frequently but with a higher damage. As a result of this risk must be calculated for all the points on the probability distribution function of water levels. This means an enormous work load for calculation of risk. Thus any simplifications that will decrease this work load have great importance for such a risk model to work efficiently. Most important parameters for damage calculations in risk models are flood depth and flood duration. If these parameters can be replaced by one parameter the work load will reduce significantly making it possible to calculate the risk. In this chapter, correlation between flood depth and duration will be proved. As a result of this correlation flood depth can be used solely, while effect of duration will be covered inclusively. Relation between these parameters is dependent on drainage properties of the soil. In order to include effects of soil properties, 12 soil types were investigated. 29

36 3.2. Methodology Flood depth can be used as an indicative parameter. This is an acceptable assumption since flood duration is closely related to flood depth. In other words, if the flood depth is high then it will take more time to drain the flood plain, thus flood duration will be longer. At this section of the study validity of this assumption was investigated. In order to verify this assumption groundwater levels were simulated for a long enough time period that would enable the researcher to comment statistically on the results. These simulations were carried out with SOBEK Rainfall Runoff Module for 12 most common soil types in Netherlands. Results were investigated statistically in means of R- Square and coefficient of correlation. 30

37 3.3. Model Schematization A simple model was built in SOBEK which will be capable of simulating the groundwater levels. This model consists of one unpaved node connected to an open water node and two pumps combining this node to boundary nodes in a way that will model the drainage system. Figure 3-1SOBEK Model Schematization In this model, precipitation falling on unpaved node is transferred to the open water node and drained further by downstream pump. Drainage capacity of the polder system was modeled by the capacity of the downstream pump. On the other hand upstream pump and upstream boundary node assures that open water elevation is kept on target level. (i.e. In case of drought water level is brought back to target level by pumping water in to the open water node) In the model unsaturated zone was simulated by using CAPSIM, which means that the storage coefficient used is calculated according to the actual groundwater level through out the simulation period. An hourly rainfall series of 333 years and a daily evaporation series for the same time period were used. Such a long simulation time gives enough events to judge on statistically. 31

38 3.4. Model Data While setting up the model attention was given to input data in a way that the model will be able to reflect the real world situation in the best way possible. In order to achieve this, input data was determined by using previous studies, values used in common practice and expert advice. In this part, input data used in the simulations were given for every node Unpaved Node: An unpaved node of 100 ha was used as a representative land. Vegetation was chosen as grass in order to avoid interference of vegetation in groundwater calculations. Parameters used are explained in detail below and listed in TABLE 3.1. Table 3-1Unpaved node parameters Parameter Area Ground Water Thickness Surface Level On Land Storage (max) Infiltration Capacity Value 100 ha 5 m 0 m 5 mm * area 20 mm/hr Storage coefficient determines the capacity of soil to store water before surface runoff occurs. Surface runoff starts when the precipitation is greater than the sum of maximum storage and infiltration capacity of the soil. In this simulation storage coefficient was chosen as 5 (mm * area). This value was determined by investigating previously used models. Infiltration capacity is the amount of water that can be infiltrated per unit area in unit time. In case that infiltration capacity is exceeded, water will be stored on land. In this simulation infiltration capacity was used as 20 mm/hr. This value was determined by investigating previously used models. 32

39 Drainage resistance is one of the most important parameters in groundwater level modeling. Groundwater outflow is calculated by using groundwater level, drainage resistance values, soil storage coefficients and downstream water level. d q Figure 3-2Ernst Drainage Calculation Parameters Ernst formula was chosen among the drainage calculation formulas since it is more convenient to use Ernst when the calculations in the unsaturated zone are carried out by CAPSIM. Ernst equation follows as; q = dh/ γ.f Where: q = drainage [m/d] dh = difference between groundwater level and drainage basis [m] γ = drainage resistance [d] f = factor depending on the shape of the groundwater table [-] (Ernst, 1978) 33

40 Figure 3-3Drainage Coefficients Input Screen Ernst values used in the model were determined with help of expert advice on subject for different soil types. Values used are given in TABLE

41 Table 3-2Ernst coefficients for different soil types Soil Type Ernst Coefficient Sand Maximum 50 Peat Maximum 10 Clay Maximum 20 Peat Average 20 Sand Average 20 Silt Maximum 50 Peat Minimum 10 Clay Average 20 Sand Minimum 50 Silt Average 20 Clay Minimum 20 Silt Minimum Open Water Node: A constant area of 5 ha was used as open water node. This area was again determined according to regulations and previously carried out studies. Bottom level was determined as datum 2m. In this case bottom level does not have any importance because the upstream pump will avoid an extensive decrease in the open water level by pumping in water from the upstream boundary node. Target level of the open water node was set to datum 1m Pumps: Upstream pump station functions in a way that will keep the open water level at target value at times when rainfall is not encountered for a long period. This reflects the real world situation, since in periods with out rainfall, decrease of groundwater level in agricultural areas are prevented by controlling the open water level in the area by pumping in water. On the other hand it does not have a direct impact on the aim of this study. The study aims to model the drainage properties of soils under floods. If the 35

42 groundwater level is brought back to target level in case of drought, this will only increase the number of events during the simulation period, which will make the results statistically sounder. The upstream pump works as an inlet and checks downstream water levels for operation. Downstream pump station models the drainage system. It functions as a normal pump and checks upstream water levels. If the target value is exceed it starts operating. In order to avoid any lag, operation rules of the pump was set in a way that it would start operation if the deviation from target level is 1cm. This is not the case in real world operations due to the fact that such a management will increase operation costs. But since the aim is modeling of the soil, this is an acceptable application in the model. Pump capacity used in the model was 6.94m 2 /min. This value was determined as the mean value of pumps that were used in previous studies Boundary nodes: Boundary nodes were set in order to isolate the model. In other words with the help of boundary nodes it was made sure that there will always be enough water in the upstream to be used in case of drought and the downstream pump will always be able to pump out the maximum capacity of the pump. 36

43 3.5. Post processing of simulation results The aim of this study was obtaining a series of flood depth and flood duration parameters and observing them statistically in order to prove the correlation between these parameters. In order to obtain these series following processes were carried out. In order to begin analysis parameters had to be defined first. Definitions used were as follows. An event was defined as water level exceeding a given threshold. In this study the threshold was defined as datum -.70m, in other words 30cm above the target level. Flood duration was defined as the time between the first time that the water level exceeds the threshold and the time when the water level goes below the threshold. Two parameters were defined for flood depth, namely as average depth and maximum depth. Maximum depth was defined as the flood depth at the time when the water level reaches its highest value within an event while average depth was defined as the mean value of flood depth through out the entire event duration. Once the model was run, results were recorded to a history (.his) file. This history file included hourly values of unpaved node parameters for 333 years. Since the simulation period was excessively large, it was not possible to work further on these history files due to large file sizes up to 1.5 gigabyte. In order to be able to process, groundwater depth data were exported to tab separated text (.txt) files. These files were containing water level values for almost 3 million time steps. A script was written in visual basic in order to pick events within this large text file. The script used hourly water levels as input and recorded another text file which involves event duration, maximum depth and average depth parameters for every single event and a summary of entire simulation period at the end of the file. (I.e. Number of events, total duration above threshold, total simulation period) The script used is given in figure 3.4 and an exemplary output file view is given in figure

44 Figure 3-4Visual Basic script for determination of events 38

45 Figure 3-5Output file view of the script Results in this file were plotted as two series, namely as maximum and average. Series Maximum indicates the duration and corresponding maximum groundwater level for each event. While series Average indicates the duration of the event and the mean value of groundwater level within that event. Further on statistical operations were carried out on these data sets in order to observe the correlation between these two parameters. First a trend line was calculated for each series. In order to be able to observe the correlation coefficient, trend line was chosen to be linear, which can be represented by the equation: y = (m*x) + b, and can be calculated by least squares fit method. Then coefficient of determination (i.e. R square) and correlation coefficient was calculated for each series. 39

46 Coefficient of determination (R 2 ) is the proportion of a sample variance of a response variable that is "explained" by the predictor variables when a linear regression is done. In other words it is the proportion of the variability in one series, it is a measure of the quality of fit. 100% R-square means perfect predictability. The formula for R 2 is where, E SS = explained sum of squares, R SS = residual sum of squares, and T SS = total sum of squares. Correlation coefficient (r), indicates the strength and direction of a linear relationship between two random variables. In general statistical usage, correlation refers to the departure of two variables from independence. The correlation coefficient will vary from -1.0 to indicates perfect negative correlation, and 1.0 indicates perfect positive correlation. If there is only one predictor variable than correlation coefficient can be calculated as the square root of coefficient of determination. r = 2 R 40

47 3.6. Results Above mentioned operations were carried out for 12 different soil types. After the post process of simulations flood depth flood duration graphs were plotted. Examples of these graphs for average and maximum values can ve observed in figure 3.6 and figure 3.7 respectively. Sand Average - Average Depth (m) y = x R 2 = Average Linear (Average) Duration (h) Figure 3-6Depth Duration graph for Sand Average (average) 41

48 Sand Average - Maximum y = x R 2 = Depth (m) Maximum Linear (Maximum) Duration (h) Figure 3-7Depth Duration graph for Sand Average (maximum) Linear trend lines were calculated by least square fit method for each soil type. Trend lines for average and maximum flood height values for different soil types are shown in Figure 3.8 and Figure

49 Depth - Duration Average Silt Average Clay Minimum Clay Maximum Clay Average Peat Average Depth (m) Peat Minimum600 Peat Maximum -0.2 Silt Maximum Sand Average -0.4 Sand Maximum Silt Minimum Sand Minimum Duration (h) Figure 3-8Trend lines of different soil types (average) Depth - Duration Maximum 3 Depth (m) Silt Average Clay Minimum Clay Average Clay Maximum Peat Average Peat Minimum Peat Maximum Sand Average Silt Maximum Sand Minimum Sand Maximum Silt Minimum Duration (h) Figure 3-9Trend lines of different soil types (maximum) 43

50 Then coefficient of determination was determined for every different soil type. Resulting R-square values are given in the table 3.3 for average and maximum flood depth cases. Table 3-3Coefficient of Determination for different soil types Soil Type R^2 (Average) R^2 (Maximum) Peat Minimum Silt Average Clay Minimum Clay Average Clay Maximum Peat Average Silt Maximum Sand Maximum Silt Minimum Sand Minimum Peat Maximum Sand Average Correlation coefficient was calculated as square root of coefficient of determination. Resulting r values are given in the table 3.4. Table 3-4Correlation Coefficient for different soil types Soil Type Correlation Coefficient (Average) Correlation Coefficient (Maximum) Peat Minimum Silt Average Clay Minimum Clay Average Clay Maximum Peat Average Silt Maximum Sand Maximum Silt Minimum Sand Minimum Peat Maximum Sand Average

51 3.7. Evaluation & Conclusion In this section drainage characteristics of 12 most common soil types in Netherlands were simulated. Main aim of this simulation was to see the correlation between flood depth and flood duration. Following conclusions were drawn from the simulations. 1- Correlation between flood depth and flood duration was proved. Thus it will be an acceptable assumption to disregard flood duration and use flood depth as an indicative parameter which will cover both coefficients. This will decrease the computational workload significantly. Correlation coefficient was noted to have an average of The lowest value was 0.60 for peat minimum and silt average, while the highest value reaches to 0.90 for sand average. This value represents a strong positive correlation between flood depth and flood duration. 2- Trend lines for depth duration relation was created from simulation results. It was observed that trendlines for clay was steeper than the ones for sand. This was an expected result due to the differences in permeability and storage coefficient. Flood depth tends to increase faster, reaches higher values, and remains high for a longer period in clay. While in sand, increse in flood depth is rather slowly and drainage is faster compared to clay. This explains the differences in slopes of trend lines. 3- Coefficient of determination was noted to decrease with increasing trend line slope. This is due to two main reasons. A lower trend line slope means faster and rather simple drainage. But in a steeper trend line, drainage is rather slow and other parameters like horizontal flow or effects of consequent rainfall events make it harder to be modeled linearly. Second reason is statistical. With the increasing slope of trend line number of events will increase. Thus with more events, number of deviations from the trend line also increases. 45

52 4. Case Study: Polder Berkel 4.1. Introduction A case study was carried out in order to investigate the applicability of a new risk analysis tool to floods due to precipitation exceeding drainage capacity. For this purpose the risk analysis tool that works on GIS basis using land use data, digital elevation model and probability density function of water levels was used to calculate the risk in the case study area. The risk calculated by the above mentioned risk analysis tool was compared with risk calculated by the risk calculated by another method namely as WB21. (Abbreviation for Waterbeheer 21 st ) As a matter of fact WB21 is not a risk model. It is rather a method for damage calculation for single events. But a risk value was obtained by summing up damage for every single event in a long enough period and dividing this damage sum to the simulation period. It must be noted that this case study does not aim to show the correlation between flood depth and duration which was proved in previous chapter since there exist many other parameters that effect the calculations within each model. But it must be also noted that the risk analysis tool used calculates the risk according to the fact that these two parameters are correlated. Further detail on both models will be given below. This chapter starts with brief introduction about the study area: Polder Berkel, follows with descriptions of both methods, comparison of results obtained and concludes with comparison and discussion of case study outcomes. 46

53 4.2. Polder Berkel The case study was carried out in Polder Berkel. The main reason for selecting this area as the case study area was the accessibility of meteorological data and rainfall-runoff model. In this section general information about the case study area will be provided. Polder Berkel lies between Rotterdam and Zoetermeer and covers an area of ha. It lies within the borders of Berkel and Rodenrijs and Pijnacker-Nootdorp municipalities. The polder is divided into 12 sub-polders with 7 different target levels. The area is drained by three main drains to Binnenboezem. Sub-polders within the polder are named as follows, 1. Bergboezem 2. Meerpolder 3. Nieuwe droogmaking 4. Nieuwe Rodenrijsche droogmakerij 5. Noordpolder 6. Oostmeerpolder 7. Oudeland 8. Oude Leede 9. Voorafsche polder 10. Westpolder 11. Zuidpolder 12. Zuidpolder Rodenrijs Water level within the polder varies from -1.5m + NAP (abbreviation for Normaal Amsterdams Peil, i.e. Normal Amsterdam Water Level) in the middle parts of the Oudeland and -5.85m + NAP in the Zuidpolder. Sub-polders Oudeland and Voorafsche Polder are relatively higher within the main polder with an average level of -2.7m + NAP. On the other hand the Bergboezem, Westpolder and Zuidpolder are the lowest areas in the case area with an average elevation of -5.1m +NAP. A general view of the sub-polders and target elevations for summer and winter season is given in Figure

54 Source: TAUW, 2002 Figure 4-1Sub-polders and target elevations 48

55 Prevailing soil type in the polder is clay. The polder is mostly covered with grass and agriculture. Percentage of grass and agricultural areas reaches to 64% while green houses cover 14% of the total area. 14% of the polder is used as urban. This distribution can be observed from the satellite image in Figure 4.2. Figure 4-2Satellite Image of Polder Berkel 49

56 4.3. WB21 Method This method had been presented for modeling damage due to high groundwater levels. Main reason of this damage is the fact that high ground water levels cause anaerobic conditions in the root zone and this leads to drowning of the crop. Other effects of high groundwater level on crop productivity are as follows: - Growing season for crops is shortened due to decreased yield and low temperature. - Fine soil particles form a crust layer. - Due to denitrification, nutrients that feed the crops are lost. - With high groundwater brackish or salt water can reach to root zone. The proposed model uses the general formula given below for damage calculations. D = Where; D f ( h, t) * D max = damage per hectare f ( h, t) = damage function dependent of depth and duration D max = maximum damage Damage functions and maximum damage amounts are depended on land use. This method uses 5 different land use form. Maximum damage amounts for different land use forms are given in Table 4.1. These amounts were calculated by Agriculture Economic Institute (Landbouw Economisch Instituut, LEI) as the average gross real turnover by hectare. 50

57 Table 4-1Maximum damage per hectare for different land use Land Use Damage Function Parameters Maximum Damage Grass Season 900 /ha Agriculture Season, workability, drowning /ha High quality agriculture and Season, workability, drowning /ha horticulture Greenhouses Depth or duration /ha Urban Depth /ha Since agricultural damage is highly dependent on the time when the flood took place, different coefficients were defined for all four seasons. Damage due to workability is taken into account by a coefficient called sum of overshootings (SOW) of the critical groundwater elevation. Unit of this coefficient is cm*day. It is calculated by multiplying the overshooting with the duration. For calculating damage due to drowning, quadratic sum of overshootings (SKOW) is used. This is expressed as cm 2 *day. This is calculated by multiplying square of the overshooting with duration. The reason for using quadratic sum is to model the non-linear character of drowning damage. Four different damage functions are given for grass, agriculture, greenhouse, and urban. These functions are given in detail below. 51

58 Grass Damage Damage in grass land is mainly due to workability condition thus it is dependent on soil type and season. In this method, this differentiation is made by a coefficient called workability coefficient for grass land. Values of this coefficient are given in Table 4.2. Damage function for grass land is given as follows. f ( h, t) = c0 ( t) * SOW cd Where; c ( ) = workability coefficient for grass land (cm -1 *d -1 ) 0 t SOW cd = sum of overshootings for given critical depth (cm*d) Table 4-2Workability coefficients for seasons and soil types for grassland Season (cm -1 *d -1 ) (cm -1 *d -1 ) (cm -1 *d -1 ) Spring 20* *10-5 5*10-5 Summer 26* * *10-5 Autumn 6*10-5 8*10-5 8*10-5 Winter Critical Depth (cm) Agriculture and Horticulture Damage Damage on agriculture is mostly dependent on the drowning of plants. Damage function is defined in a way that water level exceeding the root zone for three days causes complete loss of the crop. (Bolt et.al., 2000). Further on damage function is defined as the maximum of the damage due to workability and damage due to drowning. And it must be kept in mind that it can not be greater then 1 since damage function equal to 1 means 100% damage and a higher damage is not possible. 52

59 f ( h, t) = max( c ( t) * SOWcd, c2 ( t) * SKOW Where; 1 rz c ( ) = workability coefficient (cm-1 *d -1 ) 1 t SOW cd = sum of overshootings for given critical depth (cm*d) c ( ) = drowning coefficient (cm-2 *d -1 ) 2 t SKOW rz = sum of overshootings for given root zone (cm 2 *d) ) Coefficients to be used in the above formula are given in tables below. Differentiation for this coefficient was made with respect to seasons and soil types. Table 4.3 gives the workability coefficient for agriculture. Workability coefficient for high quality agriculture and horticulture is given in Table 4.4 and drowning coefficients are given in Table 4.5. Table 4-3Workability coefficients for seasons and soil types for agriculture Season Sand (cm - 1 *d -1 ) (cm -1 *d -1 ) (cm -1 *d -1 ) (cm -1 *d -1 ) (cm -1 *d -1 ) Spring 8*10-5 9*10-5 8*10-5 9* *10-5 Summer 14* * * * *10-5 Autumn 7*10-5 7*10-5 7*10-5 8*10-5 8*10-5 Winter Critical Depth (cm)

60 Table 4-4Workability coefficients for high quality agriculture and horticulture Season Sand (cm - 1 *d -1 ) (cm -1 *d -1 ) (cm -1 *d -1 ) (cm -1 *d -1 ) (cm -1 *d -1 ) Spring 14* * * * *10-5 Summer 24* * * * *10-5 Autumn 11* * * * *10-5 Winter Critical Depth (cm) Table 4-5Drowning coefficients Season (cm -2 *d -1 ) (cm -2 *d -1 ) (cm -2 *d -1 ) (cm -2 *d -1 ) (cm -2 *d -1 ) Spring, Summer, 21* * *10-5 7*10-5 5*10-5 Autumn Winter Root Zone Depth Greenhouses Damage In greenhouses high groundwater levels do not causes damage. In this case damage is caused by inundation. Two different damage functions were defined for greenhouses. First function is dependent on inundation depth while the second one is dependent on inundation duration. Damage functions for greenhouses are as follows. f ( h) = min( * h,1) f ( t) = min( * t 2,1) Where; h t = inundation depth (m) = inundation duration (d) 54

61 Urban Damage In this method urban damage is modeled according to the depth damage function given below. (See Figure 4.3) f ( h) = min(0.01+ h,1) Where; h = inundation depth (m) Damage Functions Damage percentage (%) Urban Greenhouses Inundation depth (m) Figure 4-3Damage Function for Greenhouses and Urban Areas Simulation and Risk Calculation For calculation of risk water levels within the polder had to be simulated for both methods. This simulation was carried out in SOBEK Rainfall-Runoff module. Model used for this simulation can be seen in Figure

62 Figure 4-4SOBEK Model for Polder Berkel This simulation was run for 100 year long time period and groundwater and open water elevations were recorded for this period. Since the file size of these records were extensively large, these files were converted to tab separated text files, making it possible to further process this data. As it was mentioned earlier WB21 method is capable of calculating damage per event. In order to be able to calculate a risk value with this method, damage caused by all events within a period had to be calculated one by one and the yearly average of these damages would be the annual average risk value calculated with this method. Time period must be chosen long enough that this definition of annual average risk will be valid. In this study a 100-years period was chosen. 56

63 In order to carry on above mentioned calculations a script was compiled. This script takes groundwater levels as input and calculates annual average risk for each water compartment. This script takes land use, season, duration and depth into account as it was mentioned in the definition of WB21 method. Further on this script provides statistical information which will enable the user to judge on and compare outcomes. Following information was provided in the output file of the script; number of events in each season, duration of events in each season, number and damage function sum of events due to workability, number and damage function sum of events due to drowning, total duration above target level, total simulation period, total damage function and damage for grass and agriculture, total damage and annual risk. A view of output file can be seen in Figure 4.5. Figure 4-5WB21 Script Output File View 57

64 4.4. Risk Model The second model used in this study was a GIS based risk model. This model takes various data as input and calculates the annual average risk accordingly. Input files for this model can be categorized into two as, GIS data and other data. - GIS Data (maps in asci file format) - Digital Elevation Map (DEM) - Water compartments - Land use map - Other Input Data (data in comma separated value (csv) file format) - Target levels for water compartments - Damage functions and maximum damage values - Probability density functions for open water levels In order to run this model a digital elevation map of 25m X 25m was used. This map can be seen in Figure 4.6. In order to assess damage for sub-polders with different target elevations, water compartments were defined. These compartments were defined according to the unpaved nodes in the SOBEK model. A detailed presentation of these water compartments can be observed in Figure

65 Figure 4-6Digital Elevation Map Figure 4-7Water Compartments Land use data was provided by a 25m X 25m land use file which consists of 16 major land use types. This land use map can be seen in Figure

66 Figure 4-8 Land Use Map 60