probabilistic design of coastal flood defences in vietnam

Probabilistic Design
of
Coastal Flood Defences in Vietnam

Probabilistic Design
of
Coastal Flood Defences in Vietnam







Proefschrift







ter verkrijging van de graad van doctor
aan de Technische Universiteit Delft,
op gezag van de Rector Magnificus professor ir. K.C.A.M. Luyben,
voorzitter van het College voor Promoties,
in het openbaar te verdedigen op dinsdag 21 september 2010 om 10.00 uur





door

MAI VAN CONG
Master of Science in Hydraulic Engineering
geboren te Ninh Binh, Vietnam



Dit proefschrift is goedgekeurd door de promotor:
Prof. drs. ir. J.K. Vrijling
Copromotor:
Dr. ir. P.H.A.J.M. van Gelder


Samenstelling promotiecommissie:
Rector Magnificus,
Prof. drs. ir. J.K. Vrijling
Dr. ir. P.H.A.J.M. van Gelder
Prof. dr. ir. M.J.F. Stive
Prof. dr. H. Hanson
Prof. dr. J.W. Hall
Ir. K.W. Pilarczyk
Dr. ir. S.N. Jonkman
Prof. dr. W.S.J. Uijttewaal
voorzitter

Technical University of Delft, promotor
Technical University of Delft, copromotor
Technical University of Delft
Lund University, Sweden
Newcastle University, UK
Formerly Rijkswaterstaat
Technical University of Delft
Technical University of Delft, reservelid










Copyright © 2010 by Mai Van Cong

All rights reserved. No part of this book may be reproduced in any form or by any
means including photocopy, without written permission from the copyright holder.

ISBN 978-90-9025648-1

Printed by: Sieca Repro, Delft, the Netherlands

Cover layout: Pham Quang Dieu
Cover image: Violent ocean wave (source: Ocean Waves Wallpapers)

to my family



i
SUMMARY
Probabilistic design of coastal flood defences in Vietnam
The year 2005 brought hurricane Katrina to New Orleans (U.S), it was also a historical
year for Vietnam because of the disastrous impact of typhoons on sea flood defences and
coastal regions. In total, eight typhoons hit the Vietnamese coast that year resulting in
human casualties and enormous economic damage. In response to that, both the central
and local governments of Vietnam have paid the highest-ever attention to the rehabilita-
tion of sea dike systems and the improvement of the safety of its coastal regions.
The improvement of the flood defence systems of Vietnam calls for the use of the latest
available knowledge at all levels. Vietnam has a profound practical experience in the field
of flood protection; however, there is a lack of up-to-date knowledge in the field of dike
design and flood risk management, especially regarding to coastal flood defences. This
study focuses on coastal flood defences in Vietnam and its objectives are to evaluate the
current level of safety and reliability, to develop safety standards and design practices
and to formulate a proposal for the improvement of the flood defences.
To meet the study objectives, the research presented in this thesis has focused on firstly

reviewing, developing and widening the application of probabilistic design and state of
the art of risk assessment and risk evaluation in the field of flood defences (theory part).
Secondly, as a demonstration of knowledge transfer, the proposed approach has been ap-
plied to the coastal flood defence system of Nam Dinh province in Vietnam which is con-
sidered as a representative coastal system in the country (application part).
The main contents covered in this thesis are summarized below.
Chapter 2 gives an overview of the coastal area, existing approaches for flood defence in
Vietnam and a description of the case study area. The coastal zone in the Red River
Delta of Vietnam, where the case study was carried out, is under a potential threat of
fluvial flooding, coastal flooding and coastal erosion. In the region, sea dikes have been
used as a predominant countermeasure with two intended functions, i.e. protecting low-
lying areas from coastal flooding and reducing risk for the hinterland caused by coastal
erosion. However, as was also experienced during recent typhoons in the last years, the
sea dikes do not seem to function very well and their strength is insufficient to withstand
the hydraulic loads with a sufficiently low frequency of occurrence. This is due to the
underestimation of the hydraulic boundary conditions and imperfections in the design,
construction and maintenance of the dike systems. A qualitative multi-criteria analysis
has been developed for proposing a suitable coastal protection strategy for the study
area. It is proposed to use a combination of sea dike systems to provide flood protection,
Summary
ii
and beach erosion control measures, such as groins and/or sand nourishments to preserve
a shallow foreshore.
To have a better understanding of the actual hydraulic load conditions in the coastal
area and to derive input information for the reliability analysis, methods for the statisti-
cal analysis of extreme hydraulic load conditions have been developed and applied
(Chapter 3). These include data management techniques (i.e. trend analysis, stationarity
tests, seasonality analysis and peaks over threshold analysis); and statistical methods to
model the occurrence probabilities of extreme values (the value which is exceeded by the
random environmental variable with very small probability p). Hydraulic observations

for the coast Vietnam are available for a period of 35 years and on that basis an estimate
needs to be derived for a design sea load, i.e. waves and storm surge levels, corresponding
to a 100 to 10,000 years return period. The methods of Regional Frequency Analysis
(RFA) and Mixed Distribution Fitting (MDF) are used in this study for such a sea load.
In order to ensure that the statistically predicted values of the interested sea loads will
be as close as possible to the true values, recommendations are made to compare these
values with physical hindcast models which are based on historical information of ty-
phoon characteristics such as wind speed.
Inspection of monthly sea water level data at Nam Dinh coast reveals that there exists a
quadratic rising trend, with periodicity of 6 months, 12 months and 224 months in the
data. The 6-month periodicity reflects well the influence of two monsoons per year to the
fluctuation of sea water level; the 12-month periodicity is explained by the influence of
yearly typhoons in the region. The 224-month periodicity, which is approximately 18.6
years, is almost exactly equal to a tidal circle.
Although sea level rise (SLR) is not a focus of this study, based on a detected linear
trend of the observed water level data the rate of SLR can be estimated approximately
at 0.68 cm per year at Nam Dinh coast. In addition, this study recommends that estima-
tion of SLR should be based on not only mathematical extrapolations, but also the long-
term nature of sea level fluctuation.
The theoretical background of probabilistic design and reliability analysis of flood de-
fences is reviewed in Chapter 4. Models for safety assessment and reliability based design
are developed for the specific application to coastal flood defence systems. In the safety
assessment model, present probabilistic calculation techniques are applied to quantify the
failure probability of the system and the system components. The effect of the length of
the dike system on the system reliability was deployed in the model as well. The reliabil-
ity-based design model is used to search for an optimal geometry of the dike system
given an admissible failure probability.
Chapter 5 demonstrates the application of the models proposed in the previous chapter
to Vietnam. The case study area is the Nam Dinh coastal flood defence system in the
Red River delta of Vietnam. It is found that the existing coastal flood defence system is

less safe than its required safety standards. According to the standards the dikes should
be designed to provide a target safety of 1/20 per year. However, the total failure prob-
ability of the dike system is estimated to be much higher, about 0.15 per year. This cor-
responds to the occurrence of a failure once in 7 years instead of once in 20 years as in-
tended. Excessive wave overtopping induced dike failure is the dominant failure mode,
Summary
iii
which likely contributes about 44% to the total failure of the dike system. Instability of
armour units of the revetments gives an influence of 25%. Overflowing and geotechnical
instability of inner slope have smaller, but still a considerable contribution (10 and 20 %,
respectively). Piping, geotechnical instability of the outer slope and toe foot instability
have a negligible influence on the safety. A sensitivity analysis shows that wave height
and water level are the most important loading parameters. The findings are in line with
the practical experiences and observations from dike failures in recent years that oc-
curred in the Nam Dinh dike system.
The application of the reliability based design model shows that if the existing safety
standard (1/20 per year) would be applied, the geometry of the current dikes would be
insufficient. For instance, the crest level of the dike required from the reliability based
design is 6.80 meters, whereas the existing crest level of the dike is 5.25 to 5.50 meters.
An extensive overview of approaches for risk assessment and risk evaluation is given in
Chapter 6. In general, risk is defined as the product of probability of occurrence and con-
sequence of disasters to be considered. Regarding flood risk in this study, the conse-
quences are considered in terms of loss of life (individual risk
†
and societal risk
‡
) and
economic damage (economic risk). A central question in risk management concerns the
acceptance of risk by the public and the decision makers. Presently, the acceptable flood-
ing risk is judged from two points of view which relate to the acceptance of inundation

consequences. The first is acceptable risk regarding loss of life, which is under judgement
of society and individuals. The second is viewed under an economic cost benefit balance.
Chapter 7 and Chapter 8 present applications of the theory given in Chapter 6 to the
Vietnamese situation. In Chapter 7, an approach for estimation of the level of acceptable
flood risk for Vietnam is proposed. The approach is based on earlier framework of Vri-
jling (2000) in which acceptable risk level is based on characteristics of the activity
(benefit, voluntariness) and baseline mortality statistics in a country. In this assessment,
first the mortality due to natural causes and daily activities such as traffic is involved.
Based on Vietnamese statistics, baseline mortality has been determined including the
risks of traffic and living in a flood-prone area in Vietnam.
An FN-curve for flooding in Vietnam is constructed showing the probability of accidents
with fatalities. Adapting the risk evaluation framework of Vrijling (2000), the national
level of acceptable risk of Vietnam is established. Evaluation of the present floods on the
national scale shows that acceptable societal risk of Vietnam is five to ten times higher
than in the Netherlands.
To obtain a risk limit for the Nam Dinh coastal region, the nationally acceptable risk has
to be distributed over all the regions in one country. A societal risk limit for Nam Dinh
is obtained and it is found that the current level of flood risk exceeds this limit by far.
According to the limit the flood risk is acceptable if the low lying coastal region of Nam

†
Individual risk: The probability (per year) of being killed at a certain location assuming perma-
nent presence of the population.
‡
Societal risk: The probability of exceedance (per year) of an accident with a certain number of
fatalities. Societal risk is often shown by means of an FN curve. It displays the probability per
year of accidents with N or more fatalities.
Summary
iv
Dinh is protected to a safety standard with exceedance probabilities lower than 1/150

per year.
The optimal safety is evaluated based on an economic approach for risk based design in
Chapter 8. Historical data on economic loss due to floods in Vietnam is available since
1970. The distribution of these losses can best be approximated by a lognormal distribu-
tion and the yearly expected value of the damage equals approximately $US 589x10
6
per
year. This value is comparable with the actual flood damage of Vietnam during the last
years (over the the last 10 years the total flood damages on average are about 1.0% of
the GDP of Vietnam per year).
Economic risk analysis of the Nam Dinh dike ring shows that the safety standard of 1/50
years should be applied to the present economic situation (Situation 1). Because of the
fast economic growth and accumulated wealth in the future (Situation 2), a safety stan-
dard of 1/100 is proposed.
The societal risk of flooding in Nam Dinh as well as in Vietnam appears to be unaccept-
able according to some existing risk limits that have been referred to literature. These
results indicate the necessity of more societal discussion and further studies on the ac-
ceptable level of flood risk in Vietnam. The decision has to be made if the current risks
are acceptable or additional risk reducing measures are necessary. The methods and re-
sults presented in this thesis provide an important input background for making these
decisions.

Delft, 2010 July 29

Mai Van Cong
Summary
v

SAMENVATTING
♣

Probabilistisch ontwerp van waterkeringen langs de kust van Vietnam
In het jaar 2005 werd New Orleans (Verenigde Staten) getroffen door orkaan Katrina.
Het was ook een historisch jaar voor Vietnam vanwege de catastrofale impact van ty-
foons op de kustverdediging en kustgebieden. In totaal hebben acht tyfoons de Vietna-
mese kust bereikt. Deze resulteerden in menselijke slachtoffers en enorme economische
schade. Als repsons op deze gebeurtenissen hebben de lokale en centrale overheden van
Vietnam zeer veel aandacht besteed aan het herstel van de waterkeringen langs de kust
en de verbetering van de veiligheid van de kustgebieden.
Voor de verbetering van het waterkeringsysteem van Vietnam is de meest recente kennis
op verschillende niveaus benodigd. Vietnam heeft uitgebreide praktijkervaring op het ge-
bied van hoogwaterbescherming. Echter, er is een gebrek aan up-to-date kennis over
dijkontwerp en de beheersing van overstromingsrisico’s, specifiek voor kustgebieden. Deze
studie richt zich op de waterkeringen in de kustgebieden van Vietnam. De doelen betref-
fen het evalueren van het huidige veiligheidsniveau en de betrouwbaarheid van de water-
keringen, het ontwikkelen van veiligheidsnormen en ontwerprichtlijnen en het formuleren
van een voorstel voor de verbetering van de waterkeringen.
Om deze doelen te bereiken heeft het onderzoek zich ten eerste gericht op het evalueren,
ontwikkelen en verbreden van de toepassing van probabilistisch ontwerpen en risicoanaly-
se voor waterkeringen (het theorie deel).Vervolgens is, mede in het kader van kennisover-
dracht, de voorgestelde aanpak toegepast op het waterkeringssysteem langs de kust van
de provincie Nam Dinh in Vietnam. Dit gebied is representatief voor het kustsysteem in
Vietnam (het toepassings deel).
De belangrijkste onderwerpen die in dit proefschrift zijn behandeld zijn hieronder sa-
mengevat.
Hoofdstuk 2 geeft een overzicht van het kustgebied, de bestaande wijze van hoogwater-
bescherming in Vietnam en een overzicht van het case studie gebied. De kustzone in de
delta van de Rode Rivier in Vietnam, waarvoor de case studie is uitgevoerd, wordt be-
dreigd door rivieroverstromingen, kustoverstromingen en kusterosie. In het gebied hebben

waterkeringen langs de kust twee functies, namelijk het beschermen van laaggelegen ge-
bieden tegen overstroming en het beperken van de risico’s voor het achterland door kus-
terosie. Echter, tijdens recente tyfoons is gebleken dat de waterkeringen langs de kust
niet goed functioneren. De sterkte van de keringen is zelfs onvoldoende om hydraulische
belastingen die relatief vaak voorkomen te weerstaan. Dit komt (mede) door de onder-
schatting van de hydraulische belastingen in de huidige praktijk en imperfecties in het
ontwerp, de constructie en het beheer en onderhoud van dijksystemen. Een kwalitatieve
multi-criteria analyse is ontwikkeld om geschikte strategieën voor kustverdediging van
het gebied te bepalen. Er wordt voorgesteld om een combinatie van zeedijken voor be-

♣
This “Samenvatting” is translated by Bas Jonkman and Wim Kanning (TUDelft) from its Eng-
lish version.
Summary
vi
scherming tegen overstromingen, en kusterosie beperkende maatregelen, zoals strand-
hoofden en zandsuppleties toe te passen om zo een flauwe vooroever the handhaven.
Methoden voor de statistische analyse van extreme hydraulische belastingen in het kust-
gebied zijn ontwikkeld en toegepast (hoofdstuk 3). Hiermee ontstaat een beter begrip van
de daadwerkelijke hydraulische belastingen en komt informatie beschikbaar waarmee een
betrouwbaarheidsanalyse kan worden uitgevoerd. De beschouwde methoden betreffen
technieken voor data beheer (trend analyse, stationariteits testen, seizoenscyclus analyse
en peaks-over-threshold analyse) en statistische methoden om de kans op het voorkomen
van een extreme waarde (de waarde die wordt overschreden door de willekeurige omge-
vingsvariabele met een zeer kleine kans) te bepalen. Hydraulische observaties voor de
kust van Vietnam zijn beschikbaar over een periode van 35 jaar. Op basis daarvan dient
een schatting gemaakt te worden van een ontwerp belasting, bestaande uit golven en
stormopzet, die overeen komt met een herhalingstijd van 100 tot 10.000 jaar. De metho-
den Regionale Frequentie Analyse (RFA) en het fitten met Gemengde Verdelingen zijn
gebruikt voor het afleiden van de hydraulische belastingen voor het kustgebied. Om zeker

te stellen dat de statistisch voorspelde waarden van de relevante belastingen dicht bij de
daadwerkelijke waarden liggen, zijn aanbevelingen gedaan om deze waarden te vergelij-
ken met de resultaten van fysische modelevaluaties van historische gebeurtenissen. Deze
zijn gebaseerd op historische kenmerken van tyfoons, zoals de windsnelheid.
Hoewel de analyse van zeespiegelstijging niet de focus van deze studie betreft, blijkt dat
de zeespiegelstijging lineair verloopt. Op basis van de beschikbare gegevens is de absolute
zeespiegelstijging ingeschat op 0,68cm per jaar voor de kust van Nam Dinh. Daarnaast is
aanbevolen om de inschatting van zeespiegelstijging niet alleen te baseren op wiskundige
extrapolatie van gegevens, maar ook op de langjarige cycli die het zeespiegelniveau bepa-
len.
De theoretische achtergrond van het probabilistisch ontwerp van waterkeringen is be-
schouwd in hoofdstuk 4. Modellen voor de veiligheidstoetsing en op betrouwbaarheid ge-
baseerd ontwerp zijn ontwikkeld voor de specifieke toepassing voor waterkeringssystemen
langs de kust. In het model voor veiligheidstoetsing zijn probabilistische rekentechnieken
gebruikt om de kans op het falen van het systeem en systeemcomponenten te kwantifice-
ren. Het effect van de lengte van het dijkensysteem op de betrouwbaarheid van het sys-
teem is in het model ook meegenomen. Het model voor op betrouwbaarheid gebaseerd
ontwerp is gebruikt om een optimale geometrie van het dijksysteem af te leiden uitgaan-
de van een toelaatbare faalkans.
Hoofdstuk 5 demonstreert de toepassing voor Vietnam van de modellen die in het vorige
hoofdstuk zijn behandeld. Dit is gedaan in een case studie voor de Rode rivier delta in
de provincie Nam Dinh in Vietnam. Het blijkt dat het waterkeringsysteem langs de kust
minder veilig is dan wordt beoogd met de bestaande ontwerpnormen. Volgens deze nor-
men zouden de dijken ontworpen moeten worden om een veiligheid van 1/20 per jaar te
bieden. Echter, de totale faalkans van het dijken system wordt hoger geschat, op onge-
veer 0,15 per jaar. Dit komt overeen met het optreden van een faalgeval gemiddeld eens
per 7 jaar, in plaats van de beoogde faalkans van eens per 20 jaar. Falen door excessieve
golfoverslag is het dominante faalmechanisme en dit draagt voor ongeveer 44% bij aan de
faalkans van het totale systeem. Instabiliteit van steenbekledingen draagt 25% bij. Over-
loop en geotechnische instabiliteit van het binnentalud hebben een kleinere, maar nog

Summary
vii
steeds aanzienlijke bijdrage aan de faalkans (10%, respectievelijk 20%). Piping en geo-
technische instabiliteit van het buitenland talud en de dijkteen hebben een verwaarloos-
bare bijdrage. Uit een gevoeligheidsanalyse blijkt dat de golfhoogte en de waterstand de
belangrijkste belasting variabelen zijn.
De bevindingen liggen in lijn met praktijkervaringen en observaties betreffende recente
dijkdoorbraken in het dijkensysteem van Nam Dinh. De toepassing van het model voor
op betrouwbaarheid gebaseerd ontwerp laat zien dat de afmetingen van de bestaande
dijken onvoldoende zijn om te voldoen aan de huidige veiligheidsnorm (1/20 per jaar).
Het blijkt bijvoorbeeld uit de betrouwbaarheidsgebaseerd ontwerp dat de benodigde kru-
inhoogte 6.80 meter moet zijn, terwijl de bestaande kruinhoogte 5.25 tot 5.50 meter be-
draagt.
Een uitgebreid overzicht van methoden voor risicoanalyse en risico evaluatie is gegeven in
hoofdstuk 6. Over het algemeen wordt risico gedefinieerd als het product van de kans op
het optreden van rampen en de gevolgen daarvan. In deze studie is het overstromingsrisi-
co beschouwd in termen van slachtofferrisico (individueel
**
en maatschappelijk risico
††
) en
economische schade (economisch risico). Een centrale vraag in risico management betreft
de aanvaardbaarheid van risico’s voor het publiek en beleidsmakers. De aanvaardbaar-
heid van overstromingsrisico’s wordt over het algemeen vanuit twee gezichtspunten be-
oordeeld. Het eerste betreft de aanvaardbaarheid van slachtofferrisico’s, zowel voor indi-
viduen als grote groepen. Het tweede gezichtspunt heeft betrekking op een economische
balans van kosten en baten.
In hoofdstuk 7 en 8 is de in hoofdstuk 6 gepresenteerde theorie toegepast op Vietnam. In
hoofdstuk 7 is een aanpak voor de bepaling van het aanvaardbare niveau van overstro-
mingsrisico’s voor Vietnam voorgesteld. De aanpak is gebaseerd op een eerder raamwerk

van Vrijling (2000) waarin het acceptabele risico niveau wordt bepaald op basis van de
kenmerken van de activiteit (baten, vrijwilligheid) en statistieken voor sterfte door diver-
se oorzaken in een land. Hierbij wordt allereerst de sterfte door natuurlijke oorzaken en
reguliere activiteiten, zoals deelname aan het verkeer, beschouwd. Op basis van beschik-
bare statistieken, is de basis sterftekans voor Vietnam bepaald inclusief de risico’s die
samenhangen met het verkeer en het leven in overstromingsgevoelige gebieden in Viet-
nam.
Een FN curve voor overstromingen in Vietnam is afgeleid en deze toont de kans op ge-
beurtenissen met bijbehorende aantallen slachtoffers. Het raamwerk voor risico evaluatie
van Vrijling (2000) is aangepast en het nationale acceptabele risiconiveau voor Vietnam
is afgeleid. Evaluatie van het overstromingsrisico op nationale schaal laat zien dat het
acceptabele maatschappelijk risiconiveau voor Vietnam vijf tot tien maal hoger is dan
voor Nederland.
Om een risiconorm vast te stellen voor het kustgebied van Nam Dinh, is het nationaal
aanvaardbare risico verdeeld over alle regio’s in het land. Hiermee wordt een risiconorm

**
Individueel risico: de kans (per jaar) om om het leven te komen op een zekere locatie uitgaande
van permanente aanwezigheid van de bevolking.
††
Maatschappelijk risico: de overschrijdingskans (per jaar) van een ongeval met meerdere slacht-
offers. Maatschappelijk risico wordt vaak afgebeeld in een FN curve. Deze toont de kans per jaar
op ongevallen met N of meer slachtoffers.
Summary
viii
voor Nam Dinh verkregen en het blijkt dat het huidige overstromingsrisico ver boven de
norm ligt. Volgens de norm is het overstromingsrisico aanvaardbaar als laaggelegen kust-
gebieden een veiligheidsnorm krijgen van ten minste 1/150 per jaar.
Het optimale veiligheidsniveau is geëvalueerd in hoofdstuk 8 op basis van een economi-
sche benadering voor risico-gebaseerd ontwerpen. Historische gegevens over economische

schade door overstromingen in Vietnam zijn beschikbaar sinds 1970. De statistische ver-
deling van deze schades kan het best benaderd worden met een lognormale verdeling. De
karakteristieke waarde van de jaarlijkse schade is ongeveer gelijk aan $US 589x10
6
per
jaar. Deze waarde is vergelijkbaar met de daadwerkelijk opgetreden schade door over-
stromingen in Vietnam in recente jaren. Gedurende de laatste tien jaren bedroegen de
schades door overstroming gemiddeld 1,5% van het bruto nationaal product van Viet-
nam.
Een economische risicoanalyse (optimalisatie) voor het dijkensysteem van Nam Dinh laat
zien dat een veiligheidsnorm van 1/50 per jaar gekozen zou moeten worden uitgaande
van de huidige economische situatie. Rekening houdend met snelle economische ontwikke-
ling in de toekomst wordt een veiligheidsnorm van 1/100 per jaar voorgesteld.
Het maatschappelijk risico door overstromingen in Nam Dinh en de rest van Vietnam
lijkt niet aanvaardbaar volgens risiconormen die zijn afgeleid op basis van de literatuur.
Deze resultaten wijzen op de noodzaak van meer maatschappelijke discussie over en ver-
der onderzoek naar het aanvaardbare niveau van overstromingsrisico’s in Vietnam. Be-
sluitvorming is nodig over de aanvaardbaarheid van de bestaande risico’s en de noodzaak
van aanvullende maatregelen. De methoden en resultaten die zijn gepresenteerd in dit
proefschrift leveren belangrijke achtergrond informatie aan om deze besluiten te nemen.
Delft, 29 juli 2010

Mai Van Cong

9
CONTENTS
SUMMARY I
CONTENTS 9
CHAPTER 1 INTRODUCTION 15
1.1 General background on flood defences 15

1.2 Developments in the design and safety assessment of flood defences 16
1.3 Catalysts in new policy towards flood risk 18
1.4 Additional stimulus of this research 19
1.5 Probabilistic vs. deterministic approach 19
1.6 Fundamentals of this study 21
1.6.1 Problem definition 21
1.6.2 Study objectives and approach 21
1.7 Scientific and social relevance 23
1.7.1 Scientific relevance 23
1.7.2 Social relevance 24
1.8 Structure of this thesis 24
CHAPTER 2 COASTAL MANAGEMENT AND FLOOD DEFENCES IN VIETNAM 27
2.1 Flood defences in Vietnam: current situation and design practice 27
2.2 Coastal protection in the Red River Delta: current strategies 29
2.2.1 Description of study area 29
2.2.2 Causes of coastal erosion 32
2.2.3 Historical information on coastline changes 34
2.2.4 The year 2005 of typhoons and its consequences 36
2.3 Hydrodynamics in the gulf of Tonkin 37
2.4 Longshore sediment transport and coastline changes 38
2.5 Actual safety of the coastal protections 40
2.6 Discussion 42
2.7 Conclusions 45
Contents
10
CHAPTER 3 ESTABLISHING THE COASTAL HYDRAULIC CONDITIONS 47
3.1 Introduction 47
3.2 Regional frequency analysis 47
3.3 Data management methods in extreme value analysis 49
3.3.1 The Annual Maxima (Gumbel) Method 49

3.3.2 Peak-over-threshold in modelling extreme value data 50
3.3.3 Trend analysis, stationarity test and detection of seasonality and natural fluctuation 51
3.4 Statistical background use in RFA 52
3.4.1 Definitions, steps in RFA and site discordance detection 52
3.4.2 L-Moment statistics 53
3.4.3 Discordance measure 55
3.4.4 Robust distances 55
3.5 RFA of extreme sea water level along the Vietnamese East sea coasts 57
3.5.1 Extreme sea water level data 57
3.5.2 Regional frequency analysis of the sea data 58
3.5.3 Discussion on the application of RFA 61
3.6 Statistical model of extreme sea water level at Nam Dinh coast 62
3.6.1 Introduction 62
3.6.2 Data inspections: rising trend, periodicity and sea level rise 63
3.6.3 Annual maxima Gumbel to model the extreme 65
3.6.4 Mixed distribution to model the extreme 66
3.6.5 Summary 69
3.7 Combined statistics and physics 69
3.7.1 Conceptual model and steps 69
3.7.2 Example of implementation 70
3.8 Discussion 73
CHAPTER 4 RELIABILITY-BASED DESIGN: THEORY 75
4.1 Introduction 75
4.2 Definition of an engineering system 75
4.3 Reliability of a system component 77
4.3.1 Non linear Z-function and design values 81
4.3.2 Non normally distributed basic variables 83
4.4 Analysis of a system 83
4.4.1 Basic systems 84
4.5 Reliability of a system 86

4.6 Length effects and system reliability 89
4.6.1 Hypothetical case: a dike system comprises one-type cross section 91
4.6.2 Dike system comprises multi dike sections 96
4.6.3 Correlation distance/scale of fluctuation 98
4.7 Discussion 99
Contents
11
4.7.1 Application model 1: Reliability safety assessment 99
4.7.2 Application model 2: Reliability based optimal design 100
4.7.3 On the probabilistic models and reliability calculation methods 101
CHAPTER 5 RELIABILITY-BASED DESIGN OF COASTAL FLOOD DEFENCES IN VIETNAM . 109
5.1 Introduction 109
5.2 Description of Nam Dinh coastal flood defences 110
5.3 Possible failure mechanisms of the sea dike systems 112
5.4 Physical process-based limit state equations of failure mechanisms 114
5.5 Relibility analysis of the Nam Dinh sea dike system 119
5.5.1 Safety assessment 119
5.5.2 Fault tree analysis and system failure probability 120
5.5.3 Sensitivity analysis of stochastic variables 122
5.5.4 Section length effects 124
5.5.5 System length effects 129
5.5.6 Total failure probability of Hai Hau dike system 131
5.5.7 Translation of system safety to design frequency of cross sections 132
5.5.8 Total failure probability of Nam Dinh coastal flood defence system 134
5.5.9 Reliability-based design 135
5.6 Discussion 137
CHAPTER 6 RISK ANALYSIS CONCEPTS AND RISK-BASED DESIGN: THEORY REVIEW 141
6.1 Historical development of risk analysis and risk management 141
6.2 Recent developments in risk analysis 143
6.2.1 Overview 143

6.2.2 What is risk? Risk definition and presentation 144
6.2.3 Discussion of risk measures for loss of life 146
6.2.4 Damage (economic) risk 149
6.2.5 Flood risk 149
6.2.6 Risk perception 150
6.2.7 Risk acceptance 151
6.3 Development of flood risk policy in the Dutch situation 152
6.3.1 Overview 152
6.3.2 Framework for determination of acceptable risk due to floods 153
6.3.3 Individual risk 154
6.3.4 Societal risk 155
6.3.5 Estimation of potential loss of life due to floods 158
6.3.6 Economic optimisation 160
6.4 Discussion 162
6.4.1 Risk related to loss of life 162
6.4.2 Economic risk based approach 163
Contents
12
6.4.3 An adapted economic risk-based framework for flood defences in Vietnam 164
CHAPTER 7 RISK EVALUATION AND ACCEPTABLE RISK IN THE VIETNAMESE SITUATION
167
7.1 Introduction 167
7.2 Assessment of actual risks and present safety standard 167
7.2.1 Present risks due to (all possible) natural causes 168
7.2.2 National acceptable risk levels: Safety standard for Vietnam 168
7.2.3 Present risk due to traffic accidents 169
7.2.4 Historical loss of life due to storms and floods in Vietnam 170
7.2.5 Discussion 172
7.3 Loss of life due to flooding in Vietnam 173
7.3.1 Individual risk for coastal flooding in Vietnam 173

7.3.2 Societal risk due to flood on a country level 174
7.4 Case study: analysis of flood risk in Nam Dinh coastal zone 179
7.4.1 Loss of life estimates for Nam Dinh coastal zone 179
7.4.2 Acceptable flood risk for Nam Dinh coastal zone 184
7.4.3 Discussion 186
CHAPTER 8 APPLICATION OF ECONOMIC RISK-BASED DESIGN 187
8.1 Economic losses of Vietnam due to flooding 187
8.2 Boundary conditions 188
8.2.1 Cost estimates 188
8.2.2 Sea water levels 189
8.2.3 Dike height in relation to exceedance frequency of water level 190
8.3 Economic optimisation of Nam Dinh sea flood defences 191
8.3.1 Cost optimal function 191
8.3.2 General assumptions 192
8.3.3 Determination of potential economic risk 193
8.3.4 Summary of the results 193
8.3.5 Sensitivity analysis 195
8.4 Extension of the van Dantzig model 197
8.4.1 Basic case: van Dantzig model (1956) 197
8.4.2 Protected area with inland slope 200
8.4.3 Sensitivity analysis 204
8.5 Discussion 205
CHAPTER 9 CONCLUSIONS AND RECOMMENDATIONS 207
9.1 Conclusions 207
9.1.1 General 207
9.1.2 Establishment of hydraulic boundary conditions for coastal areas 208
Contents
13
9.1.3 Investigation and development of safety assessment and reliability based design 209
9.1.4 Risk based design with a special application to the Vietnamese situation 210

9.1.5 Application to the Vietnamese coastal flood defences 210
9.2 Recommendations 212
9.2.1 Reliability and safety assessments 212
9.2.2 Acceptable risk 213
9.2.3 Establishment of sea boundary conditions 213
9.2.4 The Nam Dinh sea defences and optimal safety of Nam Dinh coastal zone 214
9.2.5 Coastal protection strategies for the Red River delta of Vietnam 215
9.2.6 Vietnam flood risk and flood defence strategies 216
R
EFERENCES 217
LIST OF FIGURES 227
LIST OF TABLES 231
APPENDICES 233
Appendix 1 List of scientific papers 233
Appendix 2 Statistical description of the random variables used in reliability analyses 235
Appendix 3 Overview of population data of Nam Dinh 236
LIST OF NOTATIONS 239
ACKNOWLEDGEMENTS 243
ABOUT THE AUTHOR 245


15
Chapter 1
I
NTRODUCTION
1.1 GENERAL BACKGROUND ON FLOOD DEFENCES
Coastal and fluvial areas bring enormous benefits for human life, however both areas are
liable to potential risk the sea and rivers, i.e. storm waves and surges, tsunamis, high
river discharges and consequently floods. To avoid flooding risks while living in these ar-
eas our ancestors adapted to nature recognisably. In coastal areas human settlements

were in the higher natural areas or on man made hills, named "mounds". In the river ar-
eas people lived on the natural higher ground with lower area were being used for grow-
ing food. As a result, the regular floods had little negative effects and even deposited fer-
tile silt on the land, which approximately enabled the land to keep pace with the natu-
rally rising sea level.
The rise in population meant that increasing numbers of lower lying areas were taken
into use. At the same time, provisions were put into place in relation to agriculture to
drain the land; wells were made for drinking water and peat was dug up in many places
for fuel and salt production. The consequence was a fall in the level of the ground sur-
face, as a result of which flooding became a greater problem.
In addition, a rise in population is often accompanied by a rise in human actives in the
flood-prone areas, which often results in an increase of socio-, cultural- and economic-
values. This means more value and life may be at risk due to flooding and need to be
protected.
In response, dikes were constructed. Initially, they only supplemented the natural
heights; in the rivers area for instance, perpendicular to the bank walls to redirect river
water flowing outside the banks along populated areas to lower-lying back lands. In the
coastal area the influence of the sea increased steadily due to sea level rise and land sub-
sidence, which means that, more and more lowlands had to be protected by dikes (Ton-
neijck & Weijers 2008).
Generally, almost all low-lying coastal and/or fluvial land around the world is densely
populated. Many large-scale urban expansions are concentrated in low-lying regions
where below mean sea level (MSL). In time, not only the protected areas increase in size,
Introduction
16
but also the level of required protection increases. The value of property logically in-
creases and the number of human lives threatened by flooding increases as well. The con-
sequences of a flood will, therefore, be much more serious than before as the inundation
depth and the protected values of the area increased during the centuries.
In summary, structural problems are: a densely populated, highly developed, but low-

lying areas, where flooding could lead to loss of human life, economic, cultural and less
tangible values and disruption of a society. There is no other way than ensuring that we
secure our residential areas prudently. Therefore, flood defences are needed.
1.2 DEVELOPMENTS IN THE DESIGN AND SAFETY ASSESSMENT OF
FLOOD DEFENCES

In the past, one of the first solutions to overcome flooding problems was to live with wa-
ter. There was no protection against floods coming from the sea, the lakes or the rivers.
In case of a high water level people simply moved up if necessary and stayed on higher
grounds where possible.
In the beginning of the Middle Ages, more proactive action was taken in order to cope
with flooding. At first small high ground were constructed and offered the population a
relatively safe place. People and their livestock could survive a period of flood while their
land around would get covered with a fresh and fertile layer of clay.
Later, when the population and the need for space increased the development of dikes
began. Dikes covered protected areas alongside rives and seas. The water flowing down
the river would be diverted to a place where it would cause less damage. In time, as the
need for more space increased, the amount of protected area grew.
The first dikes were designed based on experience and skill of the builder. The experience
and skill was gathered in practice. Practice was to build the dike at the highest known
water level, which was the so-called design water level, plus a certain additional free-
board, a half or one meter was applied worldwide. This design water level and the con-
struction techniques used had no scientific base. This high water level in the river is
mainly caused by heavy rain upstream, while along the sea coasts, this is often caused by
surges during storms.
In the last century, mathematical and statistical knowledge improved. Combined with
the introduction in practice of fluid and soil mechanics the approach for dike design be-
came more and more scientific. The hydraulic load on a flood defence could be predicted
more accurate and the strength of the flood defences could be calculated, always based
on extreme value extrapolation of the existing situation.

In the Netherlands, after the disaster in 1953 a statistical approach to the storm surge
levels was chosen and an extrapolated storm surge level formed the basis for dike design.
It was also described how the Delta Committee optimised these safety levels, which were
expressed in terms of the return period of high water levels which must be withstood by
the primary dike system (van Dantzig, 1956).
1.2 Developments in the design and safety assessment of flood defences
17
Since 1980, the development and the application of reliability theory made it possible to
assess the flooding risks taking into account the multiple failure mechanisms of a dike or
a structure. Dutch hydraulic engineers were among the first to apply this theory in the
practical design of structures. Reliability models were first used during the design and
the construction of a storm surge barrier in the Easternscheldt in 1976 and later in the
design of a storm surge barrier in the Nieuwe Waterweg.
In approximately 1979 a project was started to apply the probabilistic methods to the
design of dikes in general (Vrijling, 2001). The development of a complete approach to
water defense systems took a considerable time (Tonneijck & Weijers, 2008). Recently,
the approach was tested applied on many Dutch polders or dike rings. The probability of
flooding of these polders was calculated and insight in the weak spots was gained. Re-
sults can be viewed in detail e.g. in Vrijling et al. (2001), Jonkman et al. (2008) and
FLORIS reports
6
.
Actually, the results of the calculations stimulated political debate about whether the
present safety level of the water defense system is still sufficient in the Netherlands. This
question should be posed because the economic activity in the protected areas has grown
considerably since the defence systems were constructed. Moreover, ambitious private
and public investments, particularly in infrastructure, are planned. The national econ-
omy has and will become more vulnerable to flooding. In addition, the safety of a grow-
ing number of inhabitants is at stake. These types of situations and discussions are not
only occurring in the Netherlands, but also in many other flood prone countries. Gener-

ally, the number of discussions on how safe is safe enough has been of increasing atten-
tion nowadays worldwide.
Probabilistic design with risk based design concepts are considered the most modern ap-
proaches in the field of flood defences. Advantages of the methods are that allows de-
signer to take into account the uncertainties of the input parameters as random vari-
ables; and to describe the flood defence as a system, including various structural compo-
nents and its protected area. Moreover, for each system component various possible fail-
ure modes can be considered. These all help to determine the true probability of flooding
of a protected area and to judge its acceptability in a view of the consequences of the
protected area. Thus, probabilistic approach is apparently an essential tool to analyse
the flood defence systems.
Although the probabilistic approach has been increasingly proposed and applied in the
last years in the West (see e.g. the concept, method and application in Bakker & Vrijling
1980; Vrijling et al. 1998; Hall et al. 2000; Oumeraci et al. 2001 and Voortman 2002;),
application of these methods in developing countries is still limited.
In Vietnam, the approach has only recently been introduced as a new modern design
methodology. Still very few applications of the approach have been done (Mai Van et al.
2007b). Thus, widening the applications of the method to the Vietnamese situation of
flood defences and coastal protection is therefore necessary. The methods can also be ap-
plied to cases of other developing countries.

6
FLORIS: Floodrisk and Safety assessment of primary flood defences in the Netherlands