flood risk assessment in the netherlands: a case study for dike … faculteit/afdelingen... · for...

Risk Analysis, Vol. 28, No. 5, 2008 DOI: 10.1111/j.1539-6924.2008.01103.x

Flood Risk Assessment in the Netherlands: A Case Studyfor Dike Ring South Holland

Sebastiaan N. Jonkman,1,2∗ Matthijs Kok,1,3 and Johannes K. Vrijling1

Large parts of the Netherlands are below sea level. Therefore, it is important to have insightinto the possible consequences and risks of flooding. In this article, an analysis of the risks dueto flooding of the dike ring area South Holland in the Netherlands is presented. For differentflood scenarios the potential number of fatalities is estimated. Results indicate that a floodevent in this area can expose large and densely populated areas and result in hundreds tothousands of fatalities. Evacuation of South Holland before a coastal flood will be difficultdue to the large amount of time required for evacuation and the limited time available. Bycombination with available information regarding the probability of occurrence of differentflood scenarios, the flood risks have been quantified. The probability of death for a personin South Holland due to flooding, the so-called individual risk, is small. The probability of aflood disaster with many fatalities, the so-called societal risk, is relatively large in comparisonwith the societal risks in other sectors in the Netherlands, such as the chemical sector andaviation. The societal risk of flooding appears to be unacceptable according to some of theexisting risk limits that have been proposed in literature. These results indicate the necessityof a further societal discussion on the acceptable level of flood risk in the Netherlands andthe need for additional risk reducing measures.

KEY WORDS: Flood defense; flood risk; loss of life; quantitative risk analysis; risk evaluation

1. INTRODUCTION

1.1. General

The catastrophic flooding of New Orleans dueto Hurricane Katrina in the year 2005 showedthe world the catastrophic consequences of large-scale floods. More than 1,100 people were killed inthe State of Louisiana and the majority of fatali-

1 Delft University, Faculty of Civil Engineering and Geosciences,Delft, the Netherlands.

2 Royal Haskoning, Coastal & Rivers Division, Rotterdam, theNetherlands.

3 HKV Consultants, Lelystad, the Netherlands.∗ Address correspondence to Sebastiaan Jonkman, Delft Uni-

versity, Faculty of Civil Engineering and Geosciences, Stevin-weg 1, 2628 CN, Delft, the Netherlands; tel: +31615943962;[email protected].

ties occurred in the flooded areas of New Orleans.A study by the Interagency Performance EvaluationTeam(1) provides estimates of the direct damages toresidential and nonresidential buildings (US $21 bil-lion) and damages to public structures and utility in-frastructure (US $ 7 billion). Almost two years afterthe disaster, the City of New Orleans has got merelya half of its population back, and business activity isreviving slowly.(2) The flooding after Hurricane Ka-trina illustrates the vulnerability of modern societiesthat are situated in low-lying areas in deltas. This isparticularly the case in the Netherlands, where morethan 60% of the country is below sea level or the highwater levels of the rivers (see also next section). Theanticipated effects of climate change, the expectedsea level rise, and a growth of the population andeconomy in flood-prone areas are expected to resultin an increase of flood risk levels in the future. It is

1357 0272-4332/08/0100-1357$22.00/1 C© 2008 Society for Risk Analysis

1358 Jonkman, Kok, and Vrijling

important to have insight into the possible conse-quences and risks of flooding to support decisionmaking regarding flood defense strategies and pro-tection standards.

The objective of this article is to show howthe flood risks for such flood-prone areas in theNetherlands can be determined and evaluated. Theanalyses in this article focus on the hazards to peopleand potential loss of life, as this is one of the most im-portant types of consequences. The presented analy-ses are based on a method that has been developedfor the estimation of loss of life due to floods(3) andinformation from the project “Flood Risk and Safetyin the Netherlands” (FLORIS).(4–6) That project hasaimed to give insight into flood risk levels in theNetherlands by determining the probabilities andconsequences of flooding due to failure of primaryflood defenses (dikes, dunes) in the Netherlands. Re-sults are presented for the area South Holland, asthis is the largest and most densely populated areain the Netherlands (see Section 1.3 for a further de-scription).

The article is structured as follows. The remain-der of Section 1 gives general background informa-tion regarding flood defense in the Netherlands andthe case study area. The methods used for risk quan-tification are summarized in Section 2. Results arepresented in Section 3. A discussion of the results isprovided in Section 4, in which the calculated floodrisk levels are compared with the risk levels in othersectors and existing criteria for risk evaluation. Con-cluding remarks are given in Section 5.

1.2. Flood Defense in the Netherlands

Large parts of the Netherlands are below sealevel or the high water levels of rivers and lakes.Without the protection of dikes, dunes, and hydraulicstructures (e.g., storm surge barriers) approximately60% of the country would be flooded regularly. Dueto this situation, the Netherlands has a long historyof flood disasters. The last disastrous flood occurredin 1953. A storm surge from the North Sea floodedlarge parts of the southwest of the country. Apartfrom enormous economic damage, more than 1,800people drowned during this disaster. After this event,the Delta Committee was installed to investigate thepossibilities for a new approach to flood defense. Thecommittee proposed new safety standards for flooddefenses. These standards were based on an econo-metric analysis in which the incremental investmentsin more safety were balanced with the reduction ofthe risk.(7) The flood-prone areas in the Netherlands

are divided into so-called dike ring areas, i.e., areasprotected against floods by a series of water defenses(dikes, dunes, hydraulic structures) and high ground.The safety standards for the various dike rings areshown in Fig. 1. The height of these standards de-pends on the (economic) value of the area and thesource of flooding (coast or river). For coastal ar-eas design water levels have been chosen with ex-ceedance frequencies of 1/4,000 per year and 1/10,000per year. For the Dutch river area, the safety stan-dards were set at 1/1,250 per year and 1/2,000 peryear. Some smaller dike ring areas bordering theriver Meuse in the south of the country have a safetystandard of 1/250 per year. For every dike ring area,design water levels are determined with a probabilityof exceedance that corresponds with the safety stan-dard. The current design criteria for flood defensesand the process for safety evaluation are based onthese design water levels.

These safety standards were mostly derived inthe 1960s. Since then, the population and economicvalues in these dike ring areas have grown drastically.A recent investigation(8) concluded that these stan-dards are no longer in proportion to the economicand societal values that are meant to be protectedby the flood defense systems. In the last decade, theDutch Ministry of Transport, Public Works, and Wa-ter Management has initiated the FLORIS project(4)

to reinvestigate the level of flood risk in the Nether-lands. The outcomes of this study will be used to as-sess and evaluate the level of flood risk in the Nether-lands and the need for alteration of the currentpolicies and standards.

1.3. Study Area: Dike Ring South Holland

South Holland (dike ring number 14 in Fig. 1)is the largest dike ring in the Netherlands. It is themost densely populated area in the country and it in-cludes major cities such as Amsterdam, Rotterdam,and Den Haag. The area has 3.6 million inhabitantsand the total potential direct economic damage is es-timated at €290 billion.(5) Fig. 2 gives an overviewof the area and the main cities. The area is threat-ened by floods from the North Sea and the river sys-tem in the south (the Nieuwe Waterweg and Hol-landsche IJssel). The flood defense system consists ofsand dunes along the coast and earthen dikes alongthe rivers. As part of the delta works, storm surgebarriers have been constructed in the river system(e.g., the Maeslant barrier near Hoek van Hollandand a barrier in the Hollandsche IJssel) to preventstorm surges in the North Sea leading to flooding in

Flood Risk Assessment in the Netherlands 1359

Fig. 1. Dike ring areas in the Netherlandsand safety standards (source:Rijkswaterstaat). Smaller dike ring areasalong the river Meuse are not shown.

the lower river system. Depending on the location ofa breach, substantial parts of this dike ring can beflooded, as the area includes some of the lowest partsof the Netherlands. Some of these areas are almost6 meters below mean sea level.

Assessment of the risk for this area is of partic-ular interest as: (1) it is the largest dike ring area inthe Netherlands with the highest potential damage;(2) the first safety standards and design water level(with a 1/10,000 year probability of exceedance) weredetermined for this area by the Delta Committee inthe 1960s;(7) (3) it can be flooded from the coast andrivers, leading to different damage patterns and par-ticular challenges with respect to evacuation of thepopulation.

2. METHOD FOR FLOOD RISK ANALYSIS

2.1. General Approach

Methods for the analysis of flood risk, see, e.g.,References 9 and 10, generally include three main

steps: (1) determination of the probability of flood-ing; (2) simulation of flood characteristics; and (3) as-sessment of the consequences. In theory, the risk esti-mate should be based on a fully probabilistic analysisin which all the possible loads on the flood defensesystem, the resistance of the system, and possiblebreaches, flood patterns, and their consequences areincluded. Such an approach would require a numer-ical elaboration and a very large number of simu-lations. Due to limitations in time and resources asimplified approach has been chosen in this study.A limited number of so-called flood scenarios hasbeen selected and elaborated. Each scenario refersto a breach at a certain location in the flood defensesystem (or a set of multiple breaches) and the result-ing pattern of flooding. For the selected flood sce-narios the course of flooding and consequences havebeen analyzed by means of deterministic methods.By combination of information on the probabilityof the scenarios and their consequences the risk canbe estimated. The selection of flood scenarios shouldideally be based on their contribution to the overall


Fig. 2. Overview of dike ring SouthHolland. Breach locations that are usedfor analysis of flood scenarios areindicated in the figure with dots.

risk level,(11) but this would require numerical elab-oration of all the flood scenarios. In this study, theflood scenarios have been selected with the largestcontribution to the overall flooding probability. It isexpected that these scenarios are also the most rele-vant in terms of risk. For the studied case, the prob-ability of scenarios with very extreme consequencesis very small so that their contribution to the riskis relatively small (see also Section 3.1 for furtherdiscussion). This approach for scenario selection hasits limitations and several refinements and improve-ments are possible. However, this approach is ex-pected to provide realistic and indicative estimatesof the actual risk level. Below, the methods that havebeen used in the FLORIS project for the analysis ofthe probability of flooding, the flood event, and theconsequences are further described.

2.2. Determination of the Probability of Flooding

In assessing the probability of failure of a flooddefense system it is necessary to take into accountthat failure of different elements in the system and

(for each element) different failure mechanisms canlead to flooding.(12) The elements in the studied sys-tem include dune and dike sections and hydraulicstructures. In the assessment of the failure probabil-ity the hydraulic load conditions and the resistance(or strength) of the flood defense are taken into ac-count. The most important load characteristics arethe water level and waves. The resistance proper-ties include the geometry of the flood defense andthe soil characteristics. For every type of element inthe system, the relevant mechanisms that can lead tofailure are taken into account. For example, relevantfailure mechanisms for a river dike include overflow-ing, instability, and seepage/piping. For sand dunesalong the coast erosion of the dune is the most im-portant mechanism. A limit state function is usedfor each failure mechanism and it describes whichcombination of load and resistance leads to failure.For example, a dike section will fail due to over-flow/overtopping if the outside water level exceedsthe height of the dike.

The resistance and load characteristics are char-acterized by means of stochastic variables. Both in-herent uncertainties (e.g., related to the occurrence


of water levels in time) and knowledge uncertainties(e.g., limited knowledge of the soil characteristics oruncertainties in a model) are described by means ofstochastic distributions. The two types of uncertaintyare integrated into one numerical estimate of theprobability of failure by means of Bayesian proba-bility theory (see also Section 2.6 in Reference 13 forfurther background). This means that statistical in-formation and expert opinions are combined to pro-vide quantitative estimate of the probability.(14) Theresulting probability values are expressed as proba-bilities per unit time, i.e., per year.

The above approach has been implemented inan advanced program for reliability analysis of ringdike systems: PC-RING. It considers all principaldike failure modes for the elements in a system andtakes into account dependencies between failures ofthe elements (see References 15 and 16 for a furtherdescription). With this model, the probability of oc-currence can be determined for a so-called flood sce-nario. A flood scenario refers to one breach or a setof multiple simultaneous breaches in the dike ringand the resulting pattern of flooding, including theflood characteristics. To determine a set of possibleflood scenarios, the dike ring system is first dividedinto different stretches. The stretches are chosen insuch a way so that the flooding pattern and dam-age will be very similar for different breaches withinone stretch. Especially in the Netherlands where theterrain behind the flood defenses is relatively low-lying and flat such homogeneous stretches can beidentified.(9) The probability of a breach in a stretchis determined with the PC-RING model by consid-ering the contribution of different elements withinone stretch to the flooding probability. Also, possibleflood scenarios with multiple simultaneous4 breachesin different stretches are considered in the assess-ment. As output, the probabilistic PC-RING modelprovides insight into the set of the (physically) pos-sible flood scenarios and their probabilities. The sumof the probabilities of all the flood scenarios equalsthe flooding probability of the whole system.

2.3. Simulation of Flood Characteristics

To assess the damage of a flood it is necessaryto have an understanding of its hydraulic character-istics, such as depth, velocity, rise rate, and arrival

4 With simultaneous breaches we refer to the occurrence of mul-tiple breaches during the same flood event (e.g., during one highriver discharge or one storm surge).

time. These are determined for different flood sce-narios. For South Holland, several flood scenarioshave been analyzed to account for the differences inflood patterns and resulting consequences. For eachflood scenario a representative location of breach-ing has been assumed and the outside hydraulic loadconditions (water level, waves) and breach growthrate have been determined. Given the simplified ap-proach adopted in this study (see Section 2.1), de-terministic estimates of the hydraulic boundary con-ditions and breach growth have been used as input.These have been chosen somewhat more conserva-tively than the design water level that follows fromthe probabilistic analysis to account for the possibil-ity of higher boundary conditions. However, in the-ory, the full range of loading variables and the result-ing flood patterns would have to be analyzed.(11)

The development of the flood flow in the areahas been simulated with a two-dimensional hydro-dynamic model (Sobek 1D2D) that has been de-veloped by WL|Delft Hydraulics. The model givesinsight into the development of the inundation thatresults from a breach in the flood defense. As output,the model gives values of the depth and flow veloc-ity for different time steps. The value of the rise rateof the water has been determined based on these re-sults. An example of the results of a flood simulationfor South Holland is given in Fig. 3. The presence ofline elements in the area, such as roads, railways, anddikes, could influence the flood flow as they may actas compartment dikes that block the flood flow.

2.4. Assessment of the Consequences

The consequences for different flood scenarioscan be estimated based on the outputs of flood sim-ulations and information regarding population den-sity and spatial distribution of economic assets. Thedirect financial economic damages have been calcu-lated with existing damage functions.(17,18) These re-late the damage level (as fraction of total value) tothe occurring water depth. Below, the proposed ap-proach for estimation of loss of life is briefly summa-rized. Other damage categories, such as the numberof injuries and losses of ecological and historical val-ues, have not been analyzed.

2.5. Estimation of Loss of Life

Loss of life has been determined with the methodproposed in References 3 and 19. The method en-compasses two main steps: (1) estimation of the


Fig. 3. Example of output of a floodsimulation showing the maximum waterdepth for a flood scenario with breachesat Den Haag (Scheveningen) and TerHeijde (breach locations are indicatedwith dots).

exposed population; and (2) estimation of mortalityamong the exposed population.

First, the exposed population for each flood sce-nario is estimated based on the time available be-fore flooding and the time required for evacuation.The time available before flooding is determined bythe possibilities to predict the flood, leading to dif-ferent available times for coastal and river flooding.The time required is determined with an evacuationmodel.(20) This is a macro-scale traffic model thattakes into account the road network and the road ca-pacities, the locations and capacities of exits in theconsidered area, and the effects of traffic manage-ment. The model has been calibrated on historicaltraffic data in the Netherlands. As input the numberof evacuees and the temporal distribution of depar-ture of the evacuees have to be given. Delays due todecision making, warning, and response of the popu-lation have to be taken into account. As output, themodel provides an estimate of the evacuated frac-

tion of the population as a function of time. If thetime available is known, the evacuated fraction ofthe population in the exposed area can be estimated.The reduction of the number of exposed due to shel-ter is found by assuming that inhabitants of high-rise buildings find shelter within the exposed area. Inthe Netherlands, this fraction of the population canrange between 0 (rural areas) and 0.2 (urban areas).

The number of people exposed has been deter-mined for four so-called evacuation situations thatdiffer with respect to the type of flooding and thelevel of organization of the evacuation (see the eventtree in Fig. 4). The type of flood mainly influencesthe time available, while the level of organizationof the evacuation influences the time required. Eachsituation results in a different number of peopleexposed.

For risk quantification, the (conditional) proba-bilities for these different evacuation situations thatcould occur for one flood scenario need to be known.


Fig. 4. Event tree showing different evacuation situations thathave been used to estimate the number of people exposed.

Conditional probabilities for these situations havebeen estimated by means of discussion with a groupof experts.(6) They jointly gave estimates of the prob-abilities of different situations based on the type ofthreat (coast vs. river), the likelihood of unexpectedoccurrence of failure of the flood defense, and thelevel of evacuation preparedness. As an example,estimates of (conditional) probabilities for differentevacuation situations for a coastal flood scenario inSouth Holland are indicated in Fig. 4.

Consequently, the loss of life is calculated foreach situation by means of mortality functions de-rived from Reference 3. These functions relate themortality (= number of killed people/number of peo-ple exposed) to the local flood characteristics, such aswater depth, rise rate, and flow velocity. For a flooddue to breaching of a flood defense three zones aredistinguished to account for different flood charac-teristics and mortality patterns (see Fig. 5). For ex-ample, the breach zone will be characterized by manycollapsed buildings and a high mortality due to highflow velocities just behind the breach. The zone withrapidly rising waters is also characterized by a rela-tively high mortality because people are less likely

Fig. 5. Proposed hazard zones for loss oflife estimation.(3)

Fig. 6. Mortality function and observations for the zone withrapidly rising waters.(3)

to reach shelters, higher ground, or higher floors ofbuildings. For each zone a mortality function hasbeen derived based on observations for historicaldisasters, such as the floods in 1953 in the Nether-lands and floods in Japan after Typhoon Isewan in1959. For these historical events, data regarding themortality and flood characteristics (depth, velocity,and rise rate) have been collected. It has been inves-tigated whether a statistical relationship (a so-calledmortality function) can be derived to relate the mor-tality to the flood characteristics. An example of amortality function for the zone with rapidly risingwater is shown in Fig. 6. It is noted that these func-tions will be associated with many uncertainties, suchas the behavior of people affected. Model uncertain-ties in the mortality functions have been derived inReference 3 (Fig. 7–13, p. 207) and can be presentedby means of the 95% confidence interval around thebest-fit mortality function. One other issue concernsthe possibility of temporal differences in mortalitytrends, for example, due to changes in building qual-ity and improvement of warning systems. This mightreduce the validity of the derived functions that aremainly based on event data from the 1950s. However,comparison of the outcomes of the proposed methodwith information from relatively recent flood events


shows that it gives an accurate approximation5 ofthe number of fatalities.(3) A more detailed descrip-tion of the method, including a further discussion ofthe main uncertainties and sensitivities, is given inReference 3.

3. RESULTS OF RISK QUANTIFICATION

In this section, the results of risk quantificationare presented. First, probability and consequence es-timates for different flood scenarios in South Hol-land are presented (Section 3.1). These are used inSection 3.2 to calculate the individual and societalrisk.

3.1. Probability and Consequence Estimates

Flooding of South Holland can occur due tobreaches at different locations and various combina-tions of (multiple) breaches. A set of 427 physicallypossible flood scenarios has been identified with theavailable probabilistic model (see Sections 2.1 and2.2). The overall probability of flooding for dike ringSouth Holland is found by summing the probabili-ties of all these scenarios and this results in a valueof 3.99 × 10−4 per year, or approximately once in2,500 years.(4) Table I shows the probabilities for the10 flood scenarios with the highest probabilities indescending order.(5) It is noted that the table includessome flood scenarios with breaches at multiple lo-cations. Each scenario is indicated by means of itsbreach location(s) and these locations are shown inFig. 2. For these 10 scenarios, flood simulations havebeen made and the financial economic damages havebeen assessed (see results in Table I).

The first 10 scenarios result in a total probabilityof 3.94 × 10−4 per year and thereby they determineapproximately 99% of the total flooding probability.Based on the estimate of maximum contribution tothe risk of the remaining 417 flood scenarios(5) it isfound that the selected scenarios provide a reason-able approximation of the risk level.6 The above se-lection implies that flood scenarios with probabilities

5 For most of the considered validation cases, the predicted lossof life deviates less than a factor of two from the observedmortality.(3)

6 The possible contribution of the remaining flood scenarios hadbeen conservatively estimated as follows. The total probabilityof the 417 remaining flood scenarios (approximately 5 × 10−6)has been multiplied with the largest consequence of the first10 scenarios. In that conservative case, the remaining scenariosdetermine 10% of the expected economic damage.(5)

Table I. Probabilities and Economic Consequences for FloodScenarios for Dike Ring South Holland(5)

Probability EconomicBreach Location(s) (1/Year) Damage (€109)

Rotterdam (Kralingen) 1.36 × 10−4 6.8Den Haag (Boulevard) 1.19 × 10−4 1.9Den Haag (Scheveningen) 7.63 × 10−5 3.6Katwijk 2.44 × 10−5 11.3Hoek van Holland 1.15 × 10−5 2.0Katwijk and Den Haag 8.36 × 10−6 6.0Den Haag (Scheveningen)

and Ter Heijde7.23 × 10−6 22.8

Rotterdam west 4.89 × 10−6 2.5Rotterdam east 3.65 × 10−6 5.7Katwijk, Den Haag, and

Ter Heijde2.23 × 10−6 37.2

that are smaller than approximately 10−6 per yearhave not been included in the analysis.

It is noted that the presented economic damagesfor the flood scenarios are substantially smaller thanthe absolute maximum possible economic damagefor dike ring South Holland. That value, 290 billionEuros, would occur if the whole area of the dike ringSouth Holland were completely flooded. This indi-cates that the flood scenarios in South Holland areonly expected to flood a certain part of the wholearea.

For each flood scenario the loss of life has beenestimated for different levels of evacuation effective-ness, ranging from no evacuation to a fully organizedevacuation (see also Fig. 4). Results are presentedin Table II, which also shows the number of peopleexposed in the flooded area for a situation withoutevacuation.

The effect of evacuation on the loss of life hasbeen estimated with the method presented in Sec-tion 2.3, which requires insight into the time availableand the time required for evacuation. Most of theelaborated flood scenarios are associated with coastalstorm surges. For these events, the time available forevacuation (i.e., expected time between the final pre-diction and dike breach) is generally limited, and esti-mated to be between 10 and 20 hours. Analyses withan evacuation model show that the required time forcomplete evacuation of (parts of) South Holland isoften more than 24 hours and sometimes more than50 hours.(21) Depending on the considered flood sce-nario and the type of evacuation (organized or dis-organized), the estimated evacuated fraction of thepopulation ranges between 0.2 (for a predicted flood


Table II. Number of People Exposed and Number of Fatalities by Flood Scenario (Rounded by Decimals); A Distinction Is MadeBetween Different Evacuation Situations

Fatalities

Unexpected Flood Predicted FloodPeople

Exposed in No Disorganized Disorganized OrganizedFlood Scenario Flooded Area Evacuation Evacuation Evacuation Evacuation

Rotterdam (Kralingen) 180,880 1070 1060 900 860Den Haag (Boulevard) 112,140 110 100 100 100Den Haag (Scheveningen) 179,270 230 220 210 210Katwijk 205,960 400 380 340 330Hoek van Holland 102,690 110 100 100 100Katwijk and Den Haag 299,280 550 530 470 460Den Haag (Scheveningen)

and Ter Heijde706,650 3460 3290 3210 3170

Rotterdam west 107,440 190 180 170 170Rotterdam east 187,840 600 600 510 480Katwijk, Den Haag, and

Ter Heijde1,016,560 5090 4850 4720 4670

and an organized evacuation) and 0.01 (for an un-expected flood) (see Reference 21 for more details).These findings imply that only a very limited fractionof the population of South Holland can be evacuatedin the case of a (threatening) coastal flood. There-fore, there are small differences between the fatalitynumbers for different evacuation situations for oneflood scenario in Table II.

As an example, the output for the scenariowith breaches at Den Haag (Scheveningen) and TerHeijde is considered. For a situation without evacua-tion, the estimated number of fatalities is more than3,400 and more than 700,000 people are exposed.Fig. 7 shows the spatial distribution of the number offatalities. The majority of fatalities, nearly 1,900, oc-cur in areas with rapidly rising waters and large flooddepths, for example, South of Den Haag.

Discussion of Results

Results indicate that a flood of dike ring SouthHolland can cause hundreds to thousands of fatali-ties. In the case of flooding of South Holland, it isexpected that more than 100,000 people will be af-fected. In the analyzed set of scenarios, the largestnumbers for the exposed population (1 million—without evacuation) and loss of life (around 5,000)are found for a flood scenario with multiple breachesalong the coast, i.e., at Katwijk, Den Haag, and TerHeijde. This flood scenario affects a large part ofthe western part of South Holland, yet with a smallprobability (approximately 2 × 10−6 per year). For

this scenario (and one other flood scenario) the esti-mated number of fatalities is higher than the loss oflife caused by the flood disaster in New Orleans dueto Hurricane Katrina in the year 2005. That event ledto more than 1,100 fatalities in the State of Louisianaand most of these fatalities occurred in the floodedareas.(3) This illustrates the catastrophic potential oflarge-scale flooding of low-lying areas in the Dutchdelta.

The results from Table II reflect the differencesin outcomes between flood scenarios. Mortality byscenario (i.e., the number of fatalities divided bythe number of people exposed) varies between 0.1%and 0.6% with an average of 0.3% for the consid-ered set of scenarios. The mortality values are rela-tively constant because the flood conditions (depth,velocity, rise rate) for the various flood scenarios arefairly similar. However, larger variations in outcomescould emerge when model uncertainty in the loss oflife model is taken into account. It is possible to as-sess the effect of the model uncertainty that is asso-ciated with the variation in and shortage of empiri-cal mortality data that have been used to derive themortality function. Application of the mortality func-tions for the 95% confidence interval from Reference3 shows that the upper and lower bounds approxi-mately differ by a factor of two from the values pre-sented in Table II.

Other uncertainties can be associated with theselection and modeling of flood scenarios. The con-sequences for a single scenario are strongly in-fluenced by the choice of outside hydraulic load


Fig. 7. Fatalities by neighborhood andflooded area for the scenario withbreaches at Den Haag (Scheveningen)and Ter Heijde. Locations of thebreaches are shown in Fig. 3.

conditions (storm surge height and duration), and themodeling of breach growth. Another issue concernsthe assumptions with respect to the number of occur-ring breaches, as the number of breaches will affectthe extent of the flooded area and the consequences.In the dataset, the scenario with the largest conse-quences has three breaches. However, documenta-tion of historical coastal floods shows that these havebeen always characterized by even more breaches.Examples are the 1916 floods in the Netherlands(22 breaches),(22) 1953 floods in the Netherlands (ap-proximately 140 breaches), the 1966 floods in Ham-burg (more than 10),(23) and the flooding of NewOrleans after Hurricane Katrina (approximately 25breaches). In the remaining 417 flood scenarios thathave not been elaborated for consequence analyses(see above) there are some scenarios with four orfive breaches. It is expected that the consequencesfor these scenarios could be larger than the highestvalue reported in Table II. It is important to note thatthe probabilities of such extreme scenarios are smalland in the order of magnitude of 10−7 to 10−8 peryear or even smaller. It is thereby expected that theseextreme scenarios have a relatively limited contribu-tion to the expected damage, but it is recommendedto take them into account in a more detailed riskassessment.

3.2. Risk Quantification

Based on the above information regarding prob-abilities and consequences, the individual risk and

the societal risk are quantified. The individual risk in-dicates the probability of death for a person at a cer-tain location in South Holland due to flooding. Thesocietal risk expresses the probability of a disasterwith many fatalities. These two so-called risk mea-sures are used in the Netherlands and other countriesto display and limit the risks in different sectors, suchas the transport and storage of hazardous materialsand the risks of airports.(24) In a similar way it wouldalso be possible to determine the economic risk, e.g.,in the format of an expected damage or a so-calledfrequency-damage or FD curve. However, given thescope of this article the presented analysis is limitedto the risks of loss of life.

3.2.1. Individual Risk

First, individual risk (IR) is determined for SouthHolland with the following formula:

I R(x, y) =∑

i

Pf,i FD|i (x, y) (1)

where IR(x,y)—individual risk at location (x,y)[yr−1]; Pf ,i—probability of occurrence of flood sce-nario i [yr−1]; FD|i(x,y)—mortality at location (x,y)given flood scenario i [–].

As input the probabilities for different flood sce-narios have been used (see Table I). The mortality ata location is calculated for each flood scenario withthe mortality functions described in Section 2.4 andusing the results of the flood simulations as input. Inthe elaboration of individual risk, permanent and un-protected presence of people in the area is assumed.


The effects of evacuation are neglected and the indi-vidual risk becomes of the characteristic of a certainlocation. This concept is thereby consistent with thedefinitions used in the Netherlands in the so-calledexternal safety domain (see Section 4.1) where theindividual risk is used for spatial planning and riskzoning.(25) It is possible to take into account the ef-fects of evacuation in the determination of individualrisk by means of an evacuation factor (see also Ref-erence 3). However, this would lead to a very smallreduction of individual risk as possibilities for evacu-ation in South Holland are very limited (see above).

Fig. 8 shows the individual risk for South Hol-land. The individual risk is relatively high (higher10−5 per year) for deep areas exposed to flood sce-narios with (relatively) high probabilities, e.g., inthe areas northeast of Rotterdam and south of DenHaag. The highest individual risk is found north-east of Rotterdam and is around 10−4 per year. Formost of the areas in South Holland the individual

Fig. 8. Individual risk for South Holland.

risk is relatively low (below 10−6 per year); see alsoSection 4.2 for a further discussion.

3.2.2. Societal Risk

Next, the societal risk is determined by means ofa FN curve and the expected number of fatalities. Inthe analysis of societal risk, the effects of evacuationand probabilities of different evacuation types havebeen taken into account. Fig. 9 shows the FN curvefor South Holland. It shows the probability of ex-ceedance in one year of a certain number of fatalitiesdue to one event. Both axes are generally displayedin logarithmic scale. The FN curve gives informationon the probability of a flood disaster with a certainmagnitude of consequences and is used to displayand limit the risks in different sectors.(24)

The intersection with the vertical axis equals theflooding probability of South Holland (i.e., 3.99 ×10−4 yr−1). Given the selection of scenarios, the FN


Fig. 9. FN curve for flooding of South Holland.

curve is shown for values of the probability of ex-ceedance that are higher than approximately 10−6

per year (see also Section 3.1). The FN curve showsthat a flooding of South Holland is expected to leadto hundreds to thousands of fatalities.

Based on the probabilities and fatality numbersfor the selected scenarios, the expected number of fa-talities can be determined. The expected value canalso be found by integrating the area under the FNcurve.(26) This yields: E(N) = 0.21 fat/yr. The stan-dard deviation equals: σ (N) = 16.1 fat/yr. For thistype of small probability, large consequence eventthe expected number of fatalities per year is gener-ally relatively small. However, for this type of eventthe number of fatalities in one single event can belarge, resulting in a large standard deviation of thenumber of fatalities.

The selection of a relatively small number of sce-narios in this analysis could likely affect the spatial

Fig. 10. FN curve for the external safetydomain for the Netherlands and floodrisk for South Holland.a

aSource of risk estimates for external safety: Milieu en Natuur Compendium (http://www.mnp.nl/mnc/i-nl-0303.html (accessed May 31, 2006)).

distribution of the individual risk (Fig. 8) and the FNcurve (Fig. 9). To obtain a representative estimateof the spatial distribution of individual risk, sufficientscenarios should be analyzed that affect a certain lo-cation. To fully represent the FN curve sufficient sce-narios should be analyzed for a range of probabilityand consequence levels. The presented analyses arelimited to events with a probability larger than 10−6

year and the results are expected to give indicative,but realistic, first estimates of the risk level.

4. DISCUSSION

4.1. Comparison of the Societal Risk for Floodingwith Other Sectors

In this section, the calculated societal flood risksfor South Holland are compared with those in othersectors in the Netherlands. Fig. 10 depicts the FNcurve for the external safety domain in the Nether-lands and the calculated FN curve for flooding ofSouth Holland. The external safety domain is con-cerned with the risks of transport and storage of dan-gerous goods, and airport safety in the Netherlands.The risks for this domain are assessed by means ofstandardized risk analyses methods and the risks fordifferent installations are aggregated to the nationallevel.(25) Within the standardized methodology therisk estimates for the external safety domain canbe considered as best estimates. The uncertaintieshave been addressed in a similar way as for flood-ing (see also Section 2.2) and thereby the outcomesof the curves can be compared.

The societal risk for flooding of dike ring SouthHolland is larger than the risk for the external safety


domain for events with more than 100 fatalities. TheFN curves for other dike ring areas have to be addedto obtain the societal risk for flooding at a nationalscale. The FN curves of different dike rings can beadded if the risks are independent. This means thatthe FN curve for flooding in Fig. 10 will rise in thevertical direction as the probability of a flood disasterwith certain consequences at a national scale will be-come larger if multiple dike rings are assessed. Basedon the above results, it is expected that the flood risksin the Netherlands are higher than the risks for exter-nal safety. In general, the presented results confirmthe outcomes of previous studies, e.g., Reference 8,that also concluded that the societal risk of floodingat a national scale is higher than the societal risk forthe external safety domain.

The above comparison gives insight into the rel-ative magnitude of the risks in different domains. It isnoted that other risk measures, such as the expectednumber of fatalities and individual risk, can also beused to compare different risk domains. These typesof information can be of interest for policymakersand could be used for a discussion about the neces-sary expenditures on risk reduction in different sec-tors. However, when comparing different risk sectorsin decision making it is necessary to take into accountthe characteristics of the activities. For example, theperception of the activity and the benefits associatedwith the activity are important.(8,27,28)

4.2. Evaluation of the Flood Risk with ExistingRisk Limits

In this section, the calculated flood risk levels arecompared to existing risk limits that have been pro-posed in the literature. In the Netherlands, the con-cepts of individual and societal risk have been usedto determine and limit the risk in the external safetydomain. A more general framework for these risklimits has been proposed to make them applicable toother domains such as flooding. Further backgroundon this framework is given in References 29–31.

First, the limitation of individual risk for all cit-izens ensures that no one will be disproportionallyexposed to the risk and thus ensures equity. The fol-lowing criterion for the limitation of individual risk isapplied:

I R < β10−4 (2)

In this expression, the value of the policy factorβ[–] varies according to the degree to which partic-ipation in the activity is voluntary and with the per-

ceived benefit. Available statistics regarding the risksof different types of activities have been evaluated(31)

and this showed that purely voluntary activities (e.g.,mountain climbing) entailed a much higher risk thaninvoluntary activities (e.g., exposure to the risks of ahazardous installation). Based on the available statis-tics β values have been proposed ranging from β =0.01 for involuntary activities to β = 10 for a volun-tary activity for personal benefit.(31)

Second, a criterion for the judgment of soci-etal risk is needed. The aggregated level of risk ona national scale could still be considered unaccept-able even when the individual risks are consideredacceptable. The acceptable societal risk at a nationalscale can be limited as follows:

1 − FN(n) < CN/nα (3)

where FN(n)—cumulative distribution function ofthe number of fatalities; CN—constant that deter-mines the vertical position of the FN limit line at a na-tional scale [yr−1 fat−α]; α—risk aversion coefficientthat determines the steepness of the FN curve.

The coefficient α reflects risk aversion towardlarge accidents. A standard with a value of α = 1 iscalled risk neutral. If α = 2, the standard is calledrisk averse as larger accidents with many fatalities areaccepted with a relatively smaller probability thansmaller accidents. In the further analysis, we assumea steepness of the limit line α = 2, as this value isalso used in other sectors in the Netherlands.(25) Thevalue of the constant that determines the vertical po-sition of the limit line at a national scale (CN) canbe derived from the following formula that has beenproposed by Vrijling et al.:(31)

CN =(

β100k

)2

(4)

where k—constant, with a proposed value ofk = 3 (see Reference 31 for further background).

The proposed approach takes into account char-acteristics of the activity by means of the policy fac-tor β (see above). To obtain a risk limit for one in-stallation, the nationally acceptable risk has to bedistributed over the different installations/objects inone country. For a single installation a risk limit canbe applied that has the same format as Equation(3), but the constant CI is used instead of CN toindicate the vertical position of the limit line for oneinstallation, and CI ≤ CN . In deriving the value ofCI , it seems reasonable to distribute the nationallyacceptable risk over objects according to the relativesize of an object at a national scale.


Fig. 11. Regions in dike ring SouthHolland that exceed the proposed limitvalues for individual risk due to flooding.

Finally, for a more complete risk evaluation alsothe use of economic optimization for this area is rec-ommended; see, e.g., References 7 and 32. In suchan approach, the sum of the investments in floodprotection and the expected economic damages areminimized.

Comparison of the Flood Risk with ExistingRisk Limits

First, the individual risk is analyzed with Equa-tion (2). For flood risk, a value of β between 0.01(characteristic for an involuntary activity with lit-tle benefit: IR ≤ 10−6 yr−1) and 0.1 (characteris-tic for an involuntary activity with some benefit: IR≤ 10−5 yr−1) seems reasonable. These values corre-spond to IR limits used for chemical installations inthe Netherlands.(25) Fig. 11 shows the areas of SouthHolland where the risks are unacceptable according

Fig. 12. FN curve for dike ring South Holland and two limit linesor different values of the CI that determines the vertical positionof the limit line.

to these two proposed values. Results show that theindividual risk level of 10−5 per year is exceeded onlyin a few areas. The 10−6 per year individual risk levelis exceeded in low-lying areas south of Den Haag andeast of Rotterdam.

The societal flood risk for dike ring South Hol-land is compared with the criteria for risk acceptancethat have been discussed above. Fig. 12 compares theFN curve for flooding of South Holland with limitlines for different values of the constant that deter-mines the vertical position of the limit line for thedike ring (CI). According to Equation (4), a valueof CN = 11 for the limit line at a national scale isobtained for a value of β = 0.1. In the Netherlands,approximately 10 million people live in flood-proneareas and approximately 3.6 million of these peoplelive in dike ring South Holland. Taking into accountthe relative size of South Holland would imply thatthe value of the constant that determines the verticalheight of the limit line for the dike ring (CI) wouldbe approximately 36% of the constant that has beenderived for the national scale (CN) (see Reference3 for more details). This leads to CI = 4 and resultsshow that the flood risks for South Holland would beunacceptable for this value of CI . The actual floodrisks would be considered acceptable for a limit linethat corresponds to CI ≈ 100. It is thus found that thecurrent societal risk for South Holland is higher thanwould be considered acceptable according to existinglimits proposed in literature.(29–31)

4.3. Effectiveness of Measures to Reducethe Flood Risk

If the determined risk levels are considered to beunacceptably high, it can be decided to reduce the


Fig. 13. FN curve indicating the effects of two types of measures.

risk. A distinction can be made between measuresthat reduce the flooding probability and measuresthat reduce the consequences. The effects of thesetwo types are indicated in the schematic FN curvein Fig. 13. Measures to reduce the flooding probabil-ity could be dike strengthening or space for rivers.Measures that aim at a reduction of consequencescan include the construction of internal compartmentdikes, the improvement of evacuation plans, or mea-sures in the field of spatial planning.

In general, an analysis of effectiveness of mea-sures requires insight into the necessary investmentsand the reduction of (expected) damages to econ-omy, environment, and population. A cost-benefitanalysis can be used to examine which (combinationsof) measures are favorable. The cost effectiveness ofmeasures can be specifically related to the reductionof loss of life by means of the evaluation of the costof saving an extra statistical life (CSX) or the cost ofsaving an extra life year (CSXY).(33,34) This approachrelates the investments in safety measures to the re-duction of the expected number of fatalities, andthus the results of the presented risk calculations areneeded as input to make these assessments. Such re-sults can be presented to decisionmakers to supportdecisions regarding risk reduction measures.

5. CONCLUSIONS

The risks due to flooding of the dike ring areaSouth Holland in the Netherlands have been ana-lyzed in a case study. The results indicate that a floodevent in this area can expose large and densely popu-lated areas and result in hundreds to thousands of fa-talities. Evacuation of South Holland before a coastalflood will be difficult due to the large amount of time

required for evacuation and the limited time avail-able. Based on a quantitative analysis of the floodrisk, it is shown that the probability of death for a per-son in South Holland due to flooding, the so-calledindividual risk, is small. The probability of a flooddisaster with many fatalities, the so-called societalrisk, is relatively large in comparison with the soci-etal risks in other sectors in the Netherlands, suchas the chemical sector and aviation. The societal riskof flooding appears to be unacceptable according tosome of the existing risk limits that have been pro-posed in literature.

Presented results are based on the elaborationof a limited number of flood scenarios. For a morecomplete evaluation it is recommended to take intoaccount a more complete set of scenarios that givesa better representation of the possible load situa-tions, breach combinations, and flooding conditions.The flood characteristics are highly dependent on thechoice of the flood scenario, especially the assumednumber of breaches and their locations. In futureanalyses, it is important to give special attention tothe possibility of more extreme flood scenarios as-sociated with multiple breaches in the flood defensesystem. Although the current calculation indicatesthat such extreme scenarios are unlikely, documen-tation of historical coastal floods (e.g., in the Nether-lands in 1953 and in New Orleans in the 2005) showsthat these were characterized by multiple breaches.A further analysis of uncertainties in loss of life esti-mates is also recommended. Given the limitations inthe existing analyses, the presented results must beconsidered as indicative, but realistic, first estimatesof the risk level.

Based on these results, further investigation ofpossibilities for improvement of evacuation of SouthHolland, and development of alternative strategies,e.g., for shelter in place, are strongly recommended.Information regarding the elaborated flood scenariosand the number of people exposed can be used forthe development of strategies for evacuation, shelter,and rescue operations. It is necessary to have emer-gency plans prepared and practiced before a seriousflood occurs.

For a more complete evaluation of the flood haz-ards in the Netherlands the flood risks for other areashave to be determined and evaluated as well. This isdone in the project “Flood Risks and Safety in theNetherlands” (FLORIS). Some of the issues foundfor South Holland, e.g., the lack of time for evac-uation, are characteristic for coastal areas. In areasthreatened by river floods, there will be more time


for evacuation, as high discharges in the river can bepredicted longer in advance.

The current standards for flood protection inthe Netherlands are mainly based on an economet-ric analysis and not on potential risks to people. Itis proposed to investigate the possibilities to developadditional limits for flood risks to people.(8,35) In ad-dition to the assessment of individual and societalrisks it is also important to reevaluate the economicfoundation of the current risk limits; see, e.g., Refer-ences 8 and 32. In these evaluations, it is also impor-tant to take into account the possible effects of cli-mate change and the future land-use developmentsin flood-prone areas.

Overall, the results indicate the necessity of afurther societal discussion on the acceptable level offlood risk in the Netherlands. The decision has tobe made whether the current risks are acceptable orwhether additional risk reducing measures are neces-sary. The results presented in this article provide theinput information to make these decisions.

ACKNOWLEDGMENTS

This study was supported by Rijkswaterstaat (theDirectorate General of Public Works and WaterManagement in the Netherlands). The authors ac-knowledge Plony Cappendijk (Rijkswaterstaat) forher help in the preparation of the figures in thisarticle.

REFERENCES

1. IPET (Interagency Performance Evaluation Team). (2007).Performance evaluation of the New Orleans and SoutheastLouisiana hurricane protection system, Volume VII—The con-sequences, Final report 26 March 2007.

2. Liu, A., Fellowes, M., & Mabanta, M., (2006). Special editionof the Katrina index: A one-year review of key indicators of re-covery in post-storm New Orleans. Metropolitan Policy Pro-gram, Washington, DC: The Brookings Institution.

3. Jonkman, S. N. (2007). Loss of life estimation in flood risk as-sessment. Ph.d. Thesis, Delft University.

4. Rijkswaterstaat. (2006). Flood risks and safety in theNetherlands—Full report. Rijkswaterstaat report DWW 2006-014.

5. Melisie, E. J. (2006). Veiligheid Nederland in Kaart—Risicocase dijkring 14 Zuid Holland, berekening van het over-stromingsrisico, Rijkswaterstaat DWW report.

6. Jonkman, S. N., & Cappendijk, P. (2006). Veiligheid Nederlandin kaart—inschatting van het aantal slachtoffers ten gevolge vanoverstroming, DWW report DWW-2006-012.

7. Van Dantzig, D. (1956). Economic decision problems for floodprevention. Econometrica, 24, 276–287.

8. RIVM (National Institute for Public Health and Environ-ment). (2004). Dutch dikes, and risk hikes—A thematic policyevaluation of risks of flooding in the Netherlands (summary inEnglish, report in Dutch). RIVM report 500799002.

9. Van Manen, S. E., & Brinkhuis, M. (2005). Quantitative floodrisk assessment for polders. Reliability Engineering and SystemSafety, 90(2–3), 229–237.

10. Apel, H., Thieken, A. H., Merz, B., & Bloschl, G. (2006). Aprobabilistic modelling system for assessing flood risk. NaturalHazards, 38(1–2), 79–100.

11. Dawson, R., & Hall, J. (2006). Adaptive importance samplingfor risk analysis of complex infrastructure systems. Proceed-ings of the Royal Society A: Mathematical, Physical and Engi-neering Sciences, 462(2075), 3343–3362.

12. Vrijling, J. K. (2001). Probabilistic design of flood defence sys-tems in the Netherlands. Reliability Engineering and SystemSafety, 74(3), 337–344.

13. Van Gelder, P. H. A. J. M. (2000). Statistical methods for therisk-based design of civil structures. Ph.D. thesis, Delft Uni-versity.

14. Apostolakis, G. (1990). The concept of probability in safetyassessment of technological systems. Science, 250, 1359–1364.

15. Lassing, B. L., Vrouwenvelder, A. C. W. M., & Waarts, P.H. (2003). Reliability analysis of flood defence systems in theNetherlands. In P. H. A. J. M. van Gelder & T. Bedford (Eds.),Safety and reliability, proc. of the ESREL 2003 conference,Maastricht (pp. 1005–1014). Lisse, the Netherlands: BalkemaPublishers.

16. Steenbergen, H. M. G.M., Lassing, B. L., Vrouwenvelder, A.C. W.M., & Waarts, P. H. (2004). Reliability analysis of flooddefence systems. Heron 49(1), 51–73.

17. Kok M., Huizinga, H. J., Vrouwenvelder, A. C. W.M., & VanDen Braak, W. E. W. (2005). Standaardmethode2005 schadeen slachtoffers als gevolg van overstromingen. HKV reportPR999.10.

18. Jonkman, S. N., Bockarjova, M., Kok, M., & Bernardini, P.(2008). Integrated hydrodynamic and economic flood damagemodelling in the Netherlands. Ecological Economics, 66, 77–90.

19. Jonkman, S. N, Vrijling, J. K., & Vrouwenvelder, A. C. W. M.(2008). Methods for the estimation of loss of life due to floods:A literature review and a proposal for a new method. NaturalHazards, 46(3), 355–389.

20. Van Zuilekom, K. M., van Maarsseveen, M. F. A. M.,& van der Doef, M. R. (2005). A decision support sys-tem for preventive evacuation of people. In Proceedingsof the first international symposium on geo-informationfor disaster management. Delft, the Netherlands: SpringerVerlag.

21. Van Der Doef, M., & Cappendijk, P. (2006). Veiligheid Ned-erland in Kaart – modellering en analyse van evacuatie. ReportDWW-2006-007. Rijkswaterstaat DWW.

22. Rijkswaterstaat. (1916). Verslag over den stormvloed van 13/14januari 1916. Gebr. J. & H. van Langenhuysen, ’s Gravenhage.

23. Kolb, A. (1962). Sturmflut 17. Februar 1962—Morpholgieder Deich- und Flurbeschadigungen zwischen Moorbrugund Cranz. Hamburger Geographischen Studien, Hamburg1962.

24. Jonkman, S. N., van Gelder, P. H. A. J. M., & Vrijling, J. K.(2003). An overview of quantitative risk measures for loss oflife and economic damage. Journal of Hazardous Materials,A99, 1–30.

25. Bottelberghs, P. H. (2000). Risk analysis and safety policy de-velopments in the Netherlands. Journal of Hazardous Materi-als, 71(1–3), 59–84.

26. Vrijling, J. K., & van Gelder, P. H. A. J. M. (1997). Societal riskand the concept of risk aversion. In C. Guedes Soares (Ed.),Advances in safety and reliability. proceedings of ESREL 1997Lissabon, 1 (pp. 45–52). Oxford, UK: Pergamon.

27. Jongejan, R. B. (2006). A new framework for risk appraisal—Regulating industrial and flood risks. Interim Report VIII,Ph.D. research, Delft University.


28. Slovic, P. (1987). Perception of risk. Science, 236, 280–285.29. Vrijling, J. K., Hengel, W. van, & Houben, R. J. (1995). A

framework for risk evaluation. Journal of Hazardous Materi-als, 43(3), 245–261.

30. Stallen, P. J. M., Geerts, R., & Vrijling, J. K. (1996). Three con-ceptions of quantified societal risk. Risk Analysis, 16(5), 635–644.

31. Vrijling, J. K., Hengel, W. van, & Houben, R. J. (1998). Ac-ceptable risk as a basis for design. Reliability Engineering andSystem Safety, 59, 141–150.

32. Eijgenraam, C. J. J. (2006). Optimal safety standards for dike-ring areas. CPB discussion paper no 62, March 2006.

33. Vrijling, J. K., & Gelder, P. H. A. J. M. van (2000). An analysisof the valuation of a human life. In M. P. Cottam, D. W. Har-vey, R. P. Pape, & J. Tait (Eds.), ESREL 2000—“Foresight andprecaution,” Vol. 1 (pp. 197–200). Edinburgh, Scotland, UK.

34. Tengs, T. O., Adams, M. E., Pliskin, J. S., Gelb Safran D.,Siegel, J. E., Weinstein, M. C., & Graham, J. D. (1995). Five-hundred life saving interventions and their cost effectiveness.Risk Analysis, 15(3), 369–390.

35. TAW (Technical Advisory Committee on Water Defences).(1985). Some considerations of an acceptable level of risk inthe Netherlands. Report by TAW workgroup 10 “Probabilisticmethods.”

flood risk assessment in the netherlands: a case study for dike … faculteit/afdelingen... · for...

Documents