Strategies in an uncertain world:
A Systems Dynamics analysis of different flood
protection strategies
Raphael Klein, Sjoerd Mecuwsen* and Jill Slinger
July 2, 2016
Abstract
Flood defences are a key issue in low lying coastal areas. These de-
can protect the economy of hinterland regions along with countless
A ly simple abstraction of the complexity of a flood de-
fence system is formulated in system dynamic
a ring dike concept as used in The Netherlands. The model is composed
of three sub-models: the levee life cycle, the average height of the levees
and the safety of the inhabi Floods are introduced as an exter-
nal event. There are four key inputs which are meant to represent the
diversity of policies that can be adapted in dealing with flooding with dif-
ferent countries. They are the investment level, the public perception of
a government's action, the expertise of a certain country and the resource
allocation. An exploratory study across different flood regimes
expected results, with higher safety for higher investment, higher expertise
and a higher level of public perception. A policy analysis study also de-
tails different policies for imaginary countries along with their associated
results, The outcomes indicate that the model can be useful for policy
selection and insight but should not be used to judge a specific country’s
policies.
fen
lives.
This model is based on
Keywords: Floods, strategic flood defense planning, infrastructure, natural
hazards, policy-making, public perception.
1 Introduction
Levees are a crucial part of the flood protection system in low lying plains around
the world. In some countries, the amount of capital that is being protected from
an be counted in billions of dollars,
In others, countless lives of the people living in the middle of Hood
Plains are at stake. The flood protection approaches taken in different countries
flooding by levee
*The first two authors have contributed equally to this work.
as was shown for the countries surrounding the North Sea
This can be due to the different policies being adopted
governments, or because of socio-cultural differences in the
relationship between citizens and their states as was shown in
Due to these different approaches, and the complexity of the system
sing considered, it can sometimes be difficult for policy-makers to oversee the
full consequences of their decisio
There have been many efforts at using large and detailed models to represent
strategic planning for flood defences. However these models are very cumber-
> and inconvenient for simple analyses [Duong et al]
Previous System Dynamics research has beon performed in the field of
ion and flood protection. An example is provided by
, who simulated natural hazards, the reaction to these haz-
of different poli
hazards. The models presented are abstract and present a general approach to
dealing with natural hazards. In applied to an example
case of the New York basin flooding.
Following the model development and initial application, this paper presents
a further exploration of the problem. The model presented is a simple abstract
flood protection model that considers the life cycle of levees and the impact
that the occurrence of floods have on these levees and the
inhabitants. This model secks to conceptualise of flood protecti
such a way that the influence of different poli an be ecolained
This paper is structured in six main se
described first overall and then detailed in ‘
describe the results and analysis. This analysis is furthered for several case
study in Section 5] by looking at different countrie s and looking at
the results of the policies from these countries. This paper is concluded by an
overall d sion and conclusion provided in [Section 6]
s that can be used to be mitigate these
2 Model description
The model presented in this paper is a conceptually simple abstraction of the
complexity of primary flood defe1 The model is inspired by and based
upon the c iption and conceptualisation for the course Advanced System
Dynami This model is based on a ring dike or levee approach.
The ring co is applied in The Netherlands to protect a large part
of its vulnerable territory from floods (Oost and Hockstral [Figure T]
presents the ring dike concept. In thi e, the ring protects a large part of the
economic heartland of the Netherlands. This model simulates the life cycle of
tion through maintenance to breakage, and the effect of the
oce ‘ocial aspects that can impact up on flood defence projects
and policies are also included. This section explains the overall model through
describing the boundaries, the country specific parameters and the conceptual
model.
levees from constrv
rrence of flood:
Figure 1: Dike ring area 14 in the Netherlands, reproduced from [Oost_and]
Hoek
2.1 System elements and system boundaries
Considering that this model is an abstract model, it is important to understand
that it does not treat a specific case study. This has several repercussions on the
modeling approach. The first is related to the level of details considered with
respect to the levee failure modes. Levees can fail in any number of ways as
shown in[ATlsop et al.] However, to attain the requisite level of simplicity
all failure modes of levees are not considered. Instead, a primary mode of
failure is incorporated within the model, namely: failure through overtopping
of the levee. Note that this does not mean that the levee is necessarily broken,
only that overtopping and flooding of the hinterland occurs. A second failure
mechanism is the breaking of a levee. This means that the levee will have to be
replaced as it is not effective anymore in any condition.
This abstraction has a further impact on the model. The model considers
the equivalent of a territory enclosed by levees. The interior of this territory is
at the sea level height or lower meaning that if a levee breaches or the waves
are higher than the levees, then floods are expected to occur throughout the
territory. No elevated terrain is modelled within this system. This simplifies
the analysis but prevents the modelling of evacuation as a method of flood
risk management. It also means that human safety cannot be modelled and is
considered beyond the scope of the model.
The boundaries are therefore set to exclude the study of human resilience
in the face of floods, or different risk a ment methods that are being used
in Germany as shown by The model built focuses on the
extent of flooded area and the safety and perceived safety of inhabitants within
the area. The recovery aspect is therefore also not considered.
2.2 The important model factors
The model focuses on the interaction between policies on flood defence public
perception regarding safety and the construction and maintenance of levees.
The main end s factors are represented in grey. These model the relation
between the levee standards/norms and length, the safety of inhabitants in the
face of floods, sea level rise over time. These are external factors (in green).
The country specific parameters are then shown in magenta as inputs. They
are presented in later sections.
Figure 2: System diagram of the model delimiting the boundaries of the model
with thee exogenous factors (in green), the endogenous factors (in grey) and the
potential policy measures (in magenta).
shows the relation between the inputs ans
the main factors considered within the model. In effect, larger investments and
a larger expertise will lead to a higher amount of levees of a sufficient height.
These levees are then able to prevent flooding that might occur from external
events. The percentage of the ring closed also has a large imp:
of land that can be flooded as when the ring is not closed, the entire area can
be flooded. The norms are set by policy makers and also affect the robustness
on the amount
of the model with respect to the incoming floods. Time has an impact on the
entire system as the levees age and deteriorate or break with age.
2.3. The country specific parameters
The inputs in magenta ( an be considered to be country specific
parameters. Some of these parameters can be influenced by policy makers,
while others are considered to be socio-culturally determined. Four of these
parameters are used in the model.
2.3.1 Investment level
The first country specific parameter considered is the investment level. This
corresponds to the level of investment that is attributed to flood defenses by the
government. It is a policy driven parameter. This investment level contributes
‘tting of the safety norms for which the levees are being built. It therefore
the new levees but also the maintained levees (or standard levees as they
led within the model). It also has a small impact on the time it takes for
to deteriorate. This parameter is graded on a scale from 0 to 10 where 0
signifies the lowest amount of investment and therefore the lowest safety norms.
This will also mean that the levees will deteriorate slightly faster.
2.3.2 Public perception
The second country specific parameter is also the only socially driven parameter
and is referred to as the public perception of the government's action. This
parameter attempts to quantify the expectation that the public has of their
government. In some countries, the public exp the government to protect
them fully from any natural hazard (Van Raak| while in other countries,
the public does not expect the government to play a large role in their safety. In
the latter countries, the public takes an itself more individualistic and self-reliant
stance. For this parameter, the willingness to build and maintain is affected by
the perception of the public. A population that expects more governmental
aid will be assigned a higher grade on a scale from 1 to 10 and have a larger
willingness to build and maintain. The opposite is also true. This has an i
on the slope of the function that is used to represent this willingne:
The plotted function is shown in
2.3.3 Resource allocation
The third country specific parameter is the government resource allocation prior-
ity. This parameter is influenced by policy makers and it represents the priority
given to the construction of new levees rather than to the maintenance of ex-
isting levees. It is aimed at simulating the limited workforce capacity problem
ent in any country. There are only a number of workers that are
qualified enough to work on a flood defense project. This parameter is graded
from 0 to 10 where 0 means that the government is focusing all its resources,
and therefore workforce, on the renovation of levees. Opposite to this, a grade
of 10 means that the government places all of its resources on the building of
new levees. A grade of 5 therefore means that the resources are split. Note that
this parameter does not say anything about the level of investment committed
by a government to flood defenses an be very low or very
high, it has no impact on the priority of resource allocation.
2.3.4 Expertise
The fourth country specific parameter is the expertise. This parameter is meant
to portray the knowledge that any government might have in levee construction
and flood defense strategy. This parameter is tied to the time it takes for a
levee to deteriorate or to break. Again, this parameter is estimated from 1
to 10. For a grade of 1, the expert
therefore of lower quality and will degrade or break faster. The opposite is true
for a score of 10. The exper a small impact on the norms set by
the government. A high grade will mean higher norms. Note that this expertise
is considered low, the levees built are
also has
is not only government based as it can be increased by government through
the use of consultancies. It therefore does not only address the knowledge of a
govertiment but also its potential knowledge if there ig a use of external expertise:
2.4 Conceptual model
Now that the model has been appropriately delimited and that the important
inputs and factors have been de: it is interesting to examine the con-
ceptual model. An aggregate causal loop diagram is used. It is presented in
cribed, it
‘ levees that should
‘be maintained
? <
ance ofhxees D.)
me ipo
* tg seine
flooded
cg £
build ra
Me
perceived ~~
fe
Figure 3: An aggregated Causal Loop Diagram (CLD) of the flood defence
system
The causal loop diagram contains four main loops. The first loop (1) is the
Maintenance Repair loop. This loop describe the relation between the amount
of leve
the system, the more levees will ultimately deteriorate and the more levees there
will need to be repaired. This is therefore a self reinforcing loop.
The second loop (2) and the third loop (3) are similar loops but act in
opposite ways. The third loop or Sufficient Safety loop, is a feedback loop in
s and the repaired levees. In effect, the more levees are present within
which the increase of levees leads to an increase in perceived safety and therefore
a lack of willingne cond loop acts with a delay
and is opposite to this loop. This is because the larger the amount of levees, the
larger number of levees to be repaired. This occurs with a certain deterioration
delay. This in turn means that there will be more flooding, less perceived safety
and an increasing willingness to build new levees.
The fourth and final loop (4) is the broken levee loop. This is a negative
delayed loop where an increase in the levees that should be maintained leads,
with a large delay, to an increase in broken levees. This in turns means that
there are less levees that are up to standards and therefore less levees that should
loops describes what happens to the levees that should be
maintained, but that are not repaired in time and that fall in disrepair and are
ultimately considered to be broken.
ss to build more levees. The set
be maintained. This
3 Detailed model
The model is split into three main parts: the levee life cycle sub-model, the
levee height calculation sub-model and the perceived safety sub-model. These
sub-models have different sizes and complexity. They are all detailed within
3.1 Levee life cycle sub-model
b-
ance anc
The first and perhaps most important sub-model is the levee life cycl
model. This is the model that represents the construction, mainte
destruction of all levees considered in the model. Its Vensim representation is
shown in This sub-model can be further split into three main parts:
the construction of levees (the left part of the figure), the maintenance of levees
(the central part of the figure) and the destruction of levees (the central bottom
part of the figure).
The construction of new levees is based on se
veral parameters. It is related
s to build, to the building capacity and to the total ring dike
length. Considering the model assumes a levee ring around a specific land mass,
the aim of the model is to close this ring so as to best protect the inhabitants
within it. This is the driving factor for the construction of new levees. Once the
ring has been closed, there is no need to build more levees. The building capacity
is an arbitrary constant that can be affected by the resource allocation with
respect to the construction of new levees. The constru i
performed to a certain
level. The new levees
ion of new le also
ed by the in
ign time and construction time, their
d. After this delay, the
new levees then take a certain
andard or norm level which is affe« tment
have a certain des
construction is therefore not i
antaneous, but del
levees appear within the “new levee’ stock. Th
time to deteriorate and flow to the ‘levees that should be maintained’ stock.
This deterioration time is also a constant that is impacted by the experti
certain country has and its investment level. A country with more knowledge
a
Levee creation and deterioration submodel
Figure 4: Sub-model showing the levee life cycle model.
8
and expertise or consultancy assistance on the construction of levees will see its
deterioration time increase.
The second part of the sub-model consis
It therefore starts at the ‘levees that should be maintained’ stock placed
trally and in red in[Figure J] The levees that need maintenance are constantly
being repaired. This rate of repair is dependent on the willingness to repair and
also related to a certain percentage of levees that can be repaired at any time
representing a limited workforce. This percentage can be impacted by the re-
source allocation. These repairs also take a certain time. They are not repaired
at the same standard as the ‘new levees’ but are repaired to a slightly lower
safety level. Once repaired, these levees are considered to be standard levees.
s selves then deteriorate after a certain amount of time
dependent on the countries’ exper dete
can be accelerated by floods. Once deteriorated, these levees flow back into the
‘levees that should be maintained’ and the cycle continues.
Finally, the third part of this sub-model reflects the destruction of levees.
Once levees have to be maintained, some will fail and be destroyed. Thes
fully removed from the system and will need to be rebuilt. The time
s of the maintenance of the I
cen-
> and investment level. This oration
for a levee to be destroyed or broken is a constant that can also be affected by
the expertise which a country might have and its investment level.
3.2. Flood impact determination sub-model
‘A separate sub-model is used to show the impact that a flood has on the system.
For a given height the percentage of the ring length that will be flooded is
calculated. Considering that the height of every levee is not precisely tracke
rough distinction and generalisation is made between levees. First, it is assumed
that a levee that is in the ‘levees that should be maintained stock will be flooded
when there is a flood. This is because these levees are considered not to be up
to the norms. Second, the height of the other levees which are located in the
’new levees’ and ‘standard levees’ stocks is calculated. These calculations take
ount the sea level rise and the different safety norms that apply for both
stocks. If it is found that the flood will be higher than this calculated minimum
into ac
levee height, all levees contained in these ste are considered breached and
the entire area will flood.
3.3 Official and perceived safety sub-model
The third sub-model is the official and perceived safety sub-model. This is an
important part of the overall model as it is the main link between the floods
and their impact with the levee life cycle sub-model. As hinted, this sub-model
deals with two types of safety. The first one is the official safety. This safety
parameter is a technical factor which is determined by two main conditions: the
s which are up to standard. The first
part of the safety is fully dependent on whether the levee ring has been fully
closed or not. If the ring is not closed, then the official safety is c
closed levee ring and the amount: of leve
idered to
be zero. If it is closed, the official safety is allowed to climb above 0. The rest
of the official safety is dependent on the overall average height of the levees.
If all these levees are up to standards, that is either in the ‘new levees’ or the
‘standard levees’ stocks, then the official safety is 100%.
Official and perceived safety dynamics
Figure 5: Submodel detailing the official and perceived safety models.
The second safety component is the perceived safety. This safety is psycho-
ically and socially driven. It is constantly building up. As long as no flood
's, the inhabitants feel confident and therefore feel s However, once a
flood hitsand dependitiy’oni:itstoverall chagnituda:.the inhabitants areveminded
of the flood threat and the perceived safety drops instantly. It then builds up
gradually again as no floods lead to an increase of confidence.
This perceived safety represents a very important link to the construction
and maintenance of levees. It directly affects the willingness to build and main-
tain. The higher the perceived safety, the lower the willingness of the inhabitants
to build or renovate it is not important. However, after a
flood, the per and the inhabitants are very willing to renovate
and maintain their leve y are aware of the damages of the last flood.
This link is an important one as it affects the model behaviour.
4 Results and analysis
Now that the model has been explained, it is possible to examine some of the
results from the model. This section presents the reference case simulation
to gain insight on the normal behaviour of the model. This is followed by a
thorough uncertainty analysis to better study the overall impact of the different
potential policies on the model outcomes.
4.1 Reference case
The reference ‘imulated over a period of 1000 months. A flood of height
8.5 meters over a duration of one month occurs within the simulation at month
10
200. This is done to si
results are provided in
late
> the response of the model t
for the levee stocks and in
indicators.
a flood event. The
for the safety
related key performan
Selected Variables
20,000
15,000
£10,000
5000
0
0 100 200 300 400 500 600 700 800 900 1000
Time (month)
Sanda eves
Figure 6: Results for the reference run displaying the total amount of levees
(in grey), the standard levees (in green), the levees that should be maintained
(in blue) and the new levees (in red).
Selected Variables
675
0 100 200 «300 ©6400 ©6500 §=600 §=6700 §=800 §=— 900 =—:1000
‘Time (month)
Official current safety
Perceived safety : Refe
Figure 7: Results for the reference run displaying the official current safety (in
blue) and the perceived safety (in red).
The flood has an impact on the levee st
slight increase in the ‘levees that should be maintained a dented reduction
of ‘new levees’. The amount of ‘standard levees’ is also reduced by this flood.
scribed to the damages caused by the relatively small flood. The flood
as indicated by a
am
11
does not have an impact on the total amount of levees as the floods within
the model do not destroy levees directly but instead damage them. The entire
system recovers quickly after the flood.
The second behaviour that can be ob
is more evident after 500 months. Thi
to build and maintain, which decres
rved occurs throughout the model but
behaviour is related to the willingness
linearly after a flood has occurred and
depends on the size of the flood as it is directly and linearly dependent on the
perceived safety. The results for the perceived safety are shown in
The flood has a small impact on the ‘perceived safety’ as can be seen within the
figure. As the ‘perceived safety’ increases, the willingness to build decreases.
This reflects clearly within the levee results where the amount of ‘levees that
should be maintained’ quickly and steadily increases. Note the ‘official safety’
which jumps from 0 to a value of about 30% around time 200. Th
to the flood, but instead due to the closing of the levee ring leading to a high
official safety level. This official safety level is strongly influence hereafter by
the levee figure and the redirection in the willingness to build and maintain at
time 500. As less levees are being maintained and built, the overall height of the
levees within the model decreases and this is illustrated by the slow reduction
not due
in official safety.
4.2 Uncertainty analysis
To analyse the different potential behaviours of the model, an uncertainty anal-
ysis is performed on the model. This analysis aims to randomly vary different
parameters presented in model along with a possible range of sea level rise.
The full range of variation covered in the multivariate sensitivity analysis is
presented in [Table I] These are combined to form a large range of scenarios.
The combinations are chosen using a latin hypercube sampling method
This ensures the entire constraint space is explored equally. Note that
uublic perception parameter is not present in[Table 1] This is because this
particular parameter affects the willingness to build whic
function is presented in [Appendix B]
For th arios, different flooding regimes are considered. They are
These flood regimes are considered to represent
different approached to flood risk management that could occur in different
countries around the world. Some countries are more likely to experience a
large flood event every few decades while others are more likely to experience
regular smaller floods. An exceptional also included in which two
large floods occur within one decade so as to observe the resulting behaviour of
the model.
Several key performance indicators (KPIs) are recorded for each run. Re-
lated to the amount of levees, the safety or the amount of land flooded, for
conciseness, the results presented within this section only contain results for the
standard levees KPI. The behaviour observed for the other KPIs is compleme
tary. Within thi flood regime are presented,
using an analysis of four different country specific parameters. For the three
is a function. This
nario is
ction, the results for the fir:
12
Table 1: Range of the parameters that are varied for the uncertainty analys
{Parameters Minimum | Reference [ Maximum
[ Yearly sea level rise 0.01 0.03 0.06
Investment level 0 5 10
Renovation safety level 0.01 0.06 0.11
Destruction levees
New safety ley
time multiplier 0.50 1.00 1.50
1 0.00 0.03 0.03
Resources allocation 0 5 10
Building capacity T0000 6000 2000
Renovation safety percentages 05 03 0.1
Expertise 0 5 10
Destruction levees time 15 30 rn
New safety level multiplier 0.50 ai 1.50
Time before renovation 5 15 25
Figure 8: The different flood regimes being considered.
additional flood regimes, the results focus on the public perception parameter
only.
The interpretation of the results ysis of the graphi
in combination with insights derived from the model structure. The results of
z is for the first flood regime are pre
Figure 10] (oem T}and [Figure T2] The statistical analysis present on the Te
side of the figures are histogram representations of the results at the specified
times.
The first figure, ‘ q
applied on the resource allocation parameter which decides whether the govern-
ment renovates or builds more levees. The results are split into three categories.
based on anal: esults
re
In red are the results corresponding to the tactic values which are above 7 on
a scale from 0 to 10. The blue curves correspond to the tactic values between
4 and 7 while the green ones are for values below 4. On the right side of the
figure, one can see the distribution of the results at the last time step (time step
Figure 9: Standard levees - Resources allocation results - Flood 1.
a similar figure but presents the results with a post-processing
filter applied on the expertise. The results presented on the right of the figure
are very different in this case. The results are shown at time 100 months and at
the final time of 1000 months. One can see that the distribution changes over
time with the majority of the low grade expertise scene umulating at the
bottom which means there are less standard leve em. This can be
explained by the fe
This behaviour is
t that levees deteriorate faster due to the lack of expertise.
s present for the where the expertise is considered
is even less present for the scenarios with grades higher
than 7 where the standard levee stocks start to decrease due to the decreasing
willingness to build and not really due to the deterioration.
moves on to the public perception parameter. As in previous
figures, the graphs on the right display the scenario distribution, but this time
before and after the flood along with the final time. The results seem to indicate
that this tactic has little impact on the results considering the model behaviour.
One can observe a large reduction in standard levees when the public perception
parameter is very low. Although this behaviour is already attributed partly to
the expertise, it would also seem to be related to the public perception parameter
particularly considering the large peak at time 1000. The public perception does
have an impact on the willingness to build leading to a decrease in the willingness
to build, which would explain the large decrease in standard levees quickly after
the flood occurs at time 200.
[Figure T2] displays the results regarding the investment level of the govern-
marios
to be average. Thi
4
Figure 11: Standard levees - Public perception results - Flood 1.
ment. A government investing less, has a large number of scenarios with an
almost depleted standard levee stock. This also results from the expertise and
public perception as shown previously.
[Figure 13] is the first of the figure that displays the response to a second
flood. This flood regime comprises of a large flood event followed by a smaller
one only a decade later. The results are processed for the public perception
parameter. The distribution graphs are shown before the first flood in between
the two floods, and at the final time. The results show a net difference from the
previous results for the first flood regime. The willingness to build remains fairly
high as the floods keep the perceived safety fairly low throughout the simulation.
Figure 12: Standard levees - Investment level results - Flood 1.
Towards the end, there is still a net reduction in the standard levees for most
of the However, as shown for the previous flood regime, the lowest
graded scenarios end up quickly at the bottom of the distribution graphs. This
is the case throughout whether it be before, or after the floods. This shows that
the public perception parameter has a large impact on the level of the leve
on the protection of the inhabitants. It is important to consider that, similarly
to previous flood regimes, this is also impacted by the low priority and low
investment grades.
arios
and
sini
Figure 13: Standard levees - Public perception results - Flood 2.
presents the results for the third flood regime with recurrent small
floods. The results clearly display the relentlessness of the floods which keep
16
the perceived safety very low and hence the willingness to build very high. This
translates to a public perception parameter in which most of the distributions
are fairly similar due to this high perceived safety. This shows that regard-
less of the policy being applied, a constant low perceived safety will lead to a
need to build more levees. A small impact of the policy can be observed for
the distribution as the low graded curves are slightly lower, but this is almost
negligible.
Figure 15: Standard levees - Public perception results - Flood 4.
Finally, Figure 15|displays the final flood regime which consists of two large
floods happening within a decade. The results displayed here are similar to
the results obtained for the third flood regime. The main difference relates to
17
the magnitude of the results which are affected by the large second flood. This
affects the distribution displayed on the right of the main figure. They remain
similar except for the scenarios with low grades.
5 Policy analysis
After the uncertainty analysis was performed for a range of randomly selected
scenarios, it is interesting to focus on more specific scenario that could represent
real world countries and their internal poli
To look at different eral stand-in
their potential policies are mapped out as shown in[Table 2] The values chosen
represent coherent policies that could be followed by real life countries. The
main aim from this analysis is to show the outcomes of policies on the state of
the flood defense system within a certain country.
nario, ‘ountries are selected and
Table 2: Policy approaches considered for five imaginary countries.
‘ rn Resources Public Investment
Countries | Expertise ¢ .
allocation | perception level
A 10 4 10 10
B 7 10 2 3
Cc 7 5 3 2
D 7 4 8 4
E 8 4 8 5
sults of the policy analyses are shown through the main KPI used
within this paper: the amount of standard levees. This is shown in [Figure 16|
Note that this analysis is only run for the first flood regime. This is done for
ness. The results obtained can lead to some observation on the results
obtained for the different policies. It is important to mention that the model
used in this paper can in no way be used to judge the policies of different
is is because it only addresses investments and policies related to
additional
conci
the construction and maintenance of new levees. It does not addr
sts that need to be considered in the event of a flood such as evacuation or
According[Figure 16] it is clear that country A is the country with the highest
amount of standard levees. It could therefore be considered as the safest country
from a flood defense perspective is the highest and it is
less affected by the flood than the other countries portrayed within the figure.
However, the inevitable perceived safety leads to a reduction of the standard
levees as there is no longer any experience of flooding. The investment level,
along with the expertise and the public perception are fairly high. This leads
to robust flood defen
. The amount of leve
The results obtained for the other countries are very different. There are
18
Standard levees
7000
5250
1750
0 100 200 300 400 500 600 700 800 900 1000
Time (month)
Standard levees : A Standard levee:
Standard levees : B_ Standard levee:
Standard levees : C
Figure 16: Policy anal;
flood regime.
sis results plotted for the standard levee stock for first
two main groups that can be considered. The first one is formed by country D
and country E. For this group, the flood defenses appear fairly robust as the
standard levee count is fairly high. This is due to a relatively high investment
level overall and a high level of expertise. Note that the decrease due to a
decreasing willingness to build also occurs late, as it did for country A. This is
due to a predominantly high score for the public perception parameter.
The second group consists of countries B and C. These are the countries
that exhibit the lowest amount of standard levees. This can also be explained
by the policies that are adopted by policy makers as well as country specific
attributes. Country B shows the poorest behaviour due in part to its very low
public perception grade. This means that the country is less likely to finance
flood protection and this is translated to a lower willingness to build in the
model. The low amount of investment compared to the other countries is the
main reason why the amount of standard levees is much lower. As a positive
factor, the large amount of expertise does allow the amount of standard levees
to not decrease at such a higher rate.
6 Discussion and conclusion
The model presented in this paper exhibits several behaviours of interest. These
behaviours are closely related to the country specific parameters.
The first point of note is that the results obtained were, for the most part,
19
very predictable. The model has however highlighted important points with
respect to the willingness to build and the perceived safety. It has shown that
after a number of years, the public will forget past floods and being to feel
safe. This has been observed in real life too, where some countries will slowly
lose the urgency to build or maintain levees when they do not experience a
catastrophic flood in the recent past. Additionally, the model has shown that
after a catastrophe there will be an urgency to build defences in a country. An
» of such real life behaviour is the response to hurricane Katrina and the
construction of one of the largest flood defence systems in the United States
in New Orleans, (Jha et all] . The model also highlighted a great deal of
investment and expertise will lead to higher amount of levees which are also of
higher quality.
This model do
1. The policy analysis suggested that the situation in certain
countries might lie with the expertise, priorities on maintenance, the public per-
ception and the level of investment. It would however be incorrect to provide
advice based on the results obtained. This is because all significant differences
are not represented in the model used within this report.
For example, a country like The Netherlands will emphasise and invest in flood
protection programs. This is justified by limited evacuation and recovery plan-
ning and the inability of inhabitants to flee to higher land in the case of a major
flood. This justifies a very high level of investment. Other countries will have
a much lower investment level, but this is not necessarily bad. Such
may have higher land close to floodplains which would allow the evac
inhabitants and reduce the potential loss of life. Th
cus more on the recovery from flooding. These aspect
the model.
Another aspect that was mentioned within the policy analysis
relation between flood defenses and flood regimes. It is evident that different
countries are likely to face different types of natural hazard. For example,
the Gulf of Mexico region is more likely to face occasional flooding due to
hurricanes while the North Sea region is more likely to be battered by more
frequent but lower magnitude storms. This
is presented. Based on the expected natural conditions, policy makers
can choose on policies that are more adequate for their region. In this way,
some flood defense protection will be designed to withstand regular low flooding
events while other will be built to withstand larger storms that generally occur
once a decade, or simply not be constructed in favor of evacuation plans.
To conclude, this model has enabled a comparison between high level flood
risk policy making and its complications for system behaviour. The model
should not be used to design such a system. It is merely intended to provide
an clegant basis for indicating the complexity of flood risk management policy
decision making.
however, have a number of limitations that should be
between the countrie:
yuntries
ation of
ountries tend to also fo-
are not presented within
is not taken into account in the
20
References
Allsop, W., Kortenhaus, A., Morris, M., Buijs, F., Hassan, R., Young, M.,
Doorn, N., van der Meer, J., van Gelder, P., Dyer, M., et al. (2007). Failure
tructures. FLOODsite Report. T04-06-01.
mechanisms for flood defenc
Apel, H., Thicken, A. H., Merz, B., and Bléschl, G. (2004). Flood risk
ment and associated uncertainty. Natural Hazards and Earth System Science,
4(2):295-308.
Deegan, M. A. (2005). Extreme event policy design: A conceptual model to
analyze policies and the policy process for natural hazards. Proceedings of the
23rd International System Dynamics Society, Jul, pages 17-21.
Deegan, M. A. (2006). Defining the policy space for disaster management: A
system dynamics approach to us flood policy analysis. Policy, 1009:1.
Duong, T. M., Ranasinghe, R., Walstra, D., and Roelvink, D. (2015). /
climate change impacts on the stability of small tidal inlet systems:
how? Barth-Science Reviews.
Jha, A. K., Bloch, R., and Lamond, J. (2012). Cities and flooding: a guide
to integrated urban flood risk management for the 21st century. World Bank
Lesser, G., Roelvink, J., Van Kester, J., and Stelling, G. (2004). Development
and validation of a three-dimensional morphological model. Coastal engineer-
ing, 51(8):883-915.
Oost, J. and Hoekstra, A. (2009). Flood damage reduction by compartmental-
ization of a dike ring: Comparing the effectiveness of three strategies. Journal
of flood risk management, 2(4):315-321.
Slinger, J. (2015). Case description, advanced system dynamics. Delft University
of Technology, Delft.
M., and Marchand, M. (2008). The policy preferenc
and policy makers. In Samuels, P., Huntington, S., Allsop,
W,cand Harrop; J.. The. European conferenceon'flood risk-managements' Fe:
search into practice (FLOODRISK 2008). Boca Raton, London, New York,
Leiden, Taylor € Francis Group.
Slinger, J., Cuppe
Stein, M. (1987). Large sample properties of simulations using latin hypercube
sampling. Technometrics, 29(2):143-151.
Van Raak, R. (2004). Facing the threat from the North Sea. PhD thesis, Delft
University of Technology, Delft, The Netherlands.
21
A Model equations
The following table presents the model equations that are used throughout the
mode
their respective unites
It presents the variables that are cons
dered within the model along with
and formulas.
of levees
Variable Unit Formula
‘Average time that it will | month
take a new levee to go to —_
should be maintained Fea Tever'delay new + time betore Fenovation
‘Average time that it will | month
take that a standard . a, -
levee will go to main- Time before renovation + Sea Tevel delay standard
tained
Behaviour of fist food | Dial PULSE(moment Ist Hoodprofile2, duration of flood 1 month
floodprofile2
Behaviour of fist food | Dial PULSE TRAIN(60, 1, 60, 1000)-PULSE(300, 1)
floodprofiles
Behaviour of Hoodpro- | Dil PULSE(moment Ist Hoodprofilel, duration of flood 1 month)
Behaviour of frist food- | Dil PULSE(moment Ist foodprofile 4, duration of flood 3 month)
profiled
Behaviour of second | Dial PULSE(moment 2nd Hoodprofile2, duration of food 1 month)
flood floodprofile2
Behaviour of second | Dial PULSE(Moment 13 meter flood, Hood profile 3°, duration of
flood floodprofiles flood 3 month)
Behaviour of second | Dial PULSE(moment Ind Aoodprofile 4, duration of flood 3 month)
flood floodprofiled
Building capacity mn (2000-+800* Government allocation priority)
capacity to repair im total realised levees*percentage of total realised that can be re
paired* willingness to build
decrease in perceived | I/month | (Perceived safety*Percentage fooded)/Unit delay
safety
Design and construction | km/month | MIN (max((Desired km of Teves + 2750"-total veal
ces)/Design and construction tim, 0), used building capac-
ity/Design and construction time)
Design and construction | Month 90
time
desired kim of lev Kan/month | 12000
are
Desired km of lev
2750
mn
desired kin of levee + 2750
Destruction time multi- | Dmnl 0.5+Governmental priority towards robust levees/10
plier
deterioration of levees kim/month | MIN(Standard levees/time before renovation +Standard lev-
ees/sea level delay standar + fraction SL deterioration when
flood *Standard levees,Standard levees)
deterioration of new Te
MIN((New levees/time before renovation) -+(New levees/sea level
)Percentage NL deterioration when flooded “New lev-
delay ne
1300 centimeter
ees
ces, New levees)
duration of flood 1 | Month T
month
duration of flood 3 | Month 3
month
extra height for flood of |-m/km TS
flood mm/km IF THEN ELSE( flood profile selector = 0, 0, IF THEN
ELSE(flood profile select = 1, floodprofilel, IF THEN
profile selector = floodprofile, IF THEN
profile selector = floodprofiled, IF THEN
2, 0)))))
Flood of 1400 centimeter | m/km 4
Flood of 800 centimeter | m/km 8
Flood of 850 centimeter | m/km 85
flood profile selector Dmnl 0
Hooded kin of ring Dan IF THEN ELSE(flood >= 7, kin of levee that are missing, 0)
Hloodprofilel m/km (Flood of 850 centimeter+sea rise constant correct for
months*Time)*behaviour of floodprofilel
Hoodprofile2 im/kim (Flood of 1400 centimeter -- sea rise constant correct for
months*Time) * Behaviour of first flood floodprofile2 + (Flood
of 800 centimeter +sea rise constant correct for months * Time)
* Behaviour of second flood floodprofile2
Hoodprol m/kin (Flood of 850 centimetertsea rise constant correct for
months*Time)*Behaviour of first flood floodprofile3+(extra
height for flood of 1300 centimeter-+sea rise constant correct for
months*Time)*Behaviour of second flood floodprofile3
Hoodprofiled tm] km (Flood of 1400 centimetertsea rise constant correct for
months*Time)*Behaviour of frist floodprofiled. + (Flood
of 1400 centimeter+sea rise constant correct for months
“Time)*behaviour of second flood floodprofiled
Traction SL deteriation | Dimnl max(0,MIN(I,(flood - ((I+renovation safety level)*7+sea rise
when flood constant correct for months*Time)) / ((1+renovation safety
level)*7+sea rise constant correct for months*Time)))
Jovernment allocation | Dimml 5
priority
Sovernmental _ priority | Dmnl 5
towards robust levees
Tndividualism Dmnl 5
initial standard safety | m/km 7
level
of levee that are | km IF THEN ELSE desired km of levee-total realised leveesj=0 , 0,
ing desired km of levee-total realised levees )
kin of new levee flooded [km New levees*percentage of new levees flooded
standard levee | km Standard levees*percentage of standard levees flooded
hhould be maintamed | km Levees that should be maintained™percentage of maintained lev-
levee flooded ees flooded
Levee construction ex- | Dmnl 5
pertise
levees broken kim/month | (Levees that should be maintamed/time till destruction of levees)
Levees that should be | km i of levees +d jon of new levees-levees
maintained broken-repairing of levees - Initial value: 4500
iinimum height levee in| m/km safety height of a repaired levee™(I+new safety level)-safety
stock new levee height increase due to time for new levees
minimum height levee in| m/km safety height of a repaired levee*(I+renovation safety level)-
stock standard levee safety height increase due to time for standard levees
Moment 13 meter flood, | month 360
flood profile 3
moment Ist foodprofile | month 360
moment Ist loodprofilel_[ month 200
moment Ist foodprofile2 | month 360
Tmoment 2nd floodprofile | month 480
moment 2nd floodpro- | month 480
file2
months in a year month 2
New levee mn Design and construction of levees-deterioration of new leve
Initial value: 0
new Safety level Dmnl (0.01-+ Governmental priority towards robust levees*0.01) "Safety
level multiplier
Official current safety Dmnl IF THEN ELSE(percentage ring closedj=I, percentage of sufi
cient levees,0)
Perceived safety Dan total realised levees2-decrease in perceived safety,0)
Percentage flooded Dail total kin of levees flooded/desired km of levee
Percentage NL deteriora~ | Dimnl max(0,MIN(I, (flood - ((1-tnew safety level) *7-+sea rise constant
tion when flood correct for months*Time)) / ((1+new safety level)*7+sea rise
constant correct for months*Time)))
Percentage of levees | Dimnl (km of new levee flooded-+km of standard levee fooded+km
flooded should be maintained levee flooded) /total realised levees
percentage of maintained | Dimnl IF THEN ELSE(flood>=7, 1, 0)
levees flooded
percentage of new levees | Dimnl IF THEN ELSE(food>=minimum height levee in stock new
flooded levee , IF THEN ELSE(flood<=(minimum height levee in stock
new levee+safety height increase due to time for new levees),
((flood-minimum height levee in stock new levee)/safety height
increase due to), 1), 0) time for new levees
percentage of standard | Dmnl IF THEN ELSE(food<=(minimum height levee in stock stan-
levees flooded dard levee+safety height increase due to time for stan-
dard levees), ((flood-minimum height levee in stock standard
levee)/safety height increase due to time for standard levees),
1) , 0)
percentage of sufficient | Dmnl (Standard levees-+New levees) /desired km of levee
levees
percentage of total re | Dmnl (0.5-Government allocation priority*0.04)
alised that can be r
paired
percentage ring closed Dal total realised levees/desired kin of levee
renovation safety level Dmnt (0.01-+(Governmental priority towards robust __lev-
ees/100))*Safety level multiplier
Tepairing of levees
(MIN(Levees that should be maintained/time needed to repair
and construct, capacity to repair/time needed to repair and con-
struct))
Safety height increase | m/month | sea rise constant correct for months*average time that it will take
due to time for new a new levee to go to should be maintained
levees
safety height increase | m/month | sea rise constant correct for months*average time that it will take
due to time for standard that a standard levee will go to maintained
levee
safety height of a re- | m/month | sea rise constant correct for months*Time+initial standard safety
paired levee level
Safety level multiplier Dan 0.54 Levee construction expertise/10
Safety perception time | month 480
sea level delay new month initial standard safety level"new safety level/sea rise constant
correct for months
sea level delay standard | month initial standard safety level"renovation salety level/sea rise con-
stant correct for months
Sea rise constant tm/kim/year| 0.03
Sea tise constant correct | m/km/monph(sea rise constant / months in a year)
for months
Standard levees km repairing of levees-deterioration of levees
time before renovation month (G42*Levee construction expertise)" Destruction time multiplier
time needed to repair | month ZU]
and construct
time till destruction of [month (5+3* Levee construction expertise) "Destruction time mult
levees plier*12
total km of levees | km IF THEN BLSE( (flooded km of ring+km of new
flooded flooded+km of standard levee flooded-+km should be maintained
tained levee flooded)
levee flooded)>=12000, 12000, flooded km of ring+km of new
levee flooded+km of standard levee flooded+km should be main-
total realised levees ian Lever
that should be maintained+Standard levees+New levees
Increase of perceived | Dmnl/montfh (max(I-Perceived safety,percontage ring closed-Perceived
safety safety))/Safety perception time
Unit delay month T
used building capacity kn building capacity*willingness to build
willmgness to build Dmnl (EPerceived safety)* (Individualism)
B_ Willingness to build lookup
The lookup function that is used for the calculation of the willingness to build is
ed in [Figure I7] As is shown in the figure, different grades for the public
perception will lead to a different lookup representing a country where citizens
fool like the government should take care of their protection or countries where
izens feel responsible for their own safety.
present
pt
tow pale poreepion
rf
oi 02 08 O04 05 06 O7 08 09 1
Perceived safety
Figure 17: Public perception lookup function.
