Applying Fuzzy Delphi Method to Select the Variables of a
Sustainable Urban System Dynamics Model
Yu-Feng Ho
Graduate School of Architecture and Urban Design,
Chaoyang University of Technology
Email:hyfarch@ms32.hinet.net
Hsiao-Lin Wang
Department of Landscape Architecture,
Tunghai University, Taichung
Email:whl1435@hotmail.com
Abstract
A system dynamics model is composed of many variables. These
variables simplify complex phenomena and provide a description of a system’s
current state or problems. Basic variables that describe the real-world urban
development can be established from the elements that make up a city’s
different dimensions such as industry product, population growth and vacancy
rate. The urban development framework takes a system-based approach by
systemizing the city’s internal elements. The systemic variables then provide
not only a clear reflection of the interactions between all of the sub-systems but
also how they relate to the overall system. It is therefore very important to
select the appropriate variables. Most variables of system dynamics models
are, however, set up by the designer, served as a subjective and unscientific
approach. This study therefore applies the Fuzzy Delphi Method to the
selection process of system variables to increase the confidence of the model.
This was accomplished by first examining the system relationships as well as
the intent and meaning of the sub-system variables to be created. After
establishing the criteria for variable selection, an empirical case study was
used to devise the evaluation variables for each sub-system.
1. Introduction
Over the past 20 years, an acute demand or national environmental
information has emerged due to the increase in international awareness of
environmental issues. Many environmental variable programs have therefore
been set up in developed countries to provide better channels for
communication. In 1989, the G7 summit of industrialized nations asked the
OECD (Organization for Economic Co-operation and Development) to propose
a suitable solution. This eventually led to the creation of sustainability variables
that take both environmental and socio-economic development into account.
The call for sustainability variables reached its peak at the 1992 Earth Summit
and laid the foundations for the sustainable development variables promoted
by the UN (United Nations) CSD (Commission on Sustainable Development).
In 1991, the IUCN (World Conservation Union), UNEP (United Nations
Environment Programme) and WWF(World Wildlife Fund) joined together to
draft the “Caring for the Earth” guidelines. This specified two areas of coverage
for any sustainable development variables: quality of life and ecological
sustainability. Here quality of life included longevity, knowledge and income;
ecological sustainability was measured by:
1. Conserving life-support systems and biodiversity;
2. Ensuring the use of renewable resources are sustainable and minimizing the
depletion of non-renewable resources;
3. Keeping within the carrying capacity of supporting eco systems.
In 1992 the city of Seattle in Washington State, USA passed a growth
management plan. Four principles for selecting “good variable” were
developed by the working group: “vision”, “ease of understanding and
acceptance”, “benefits and attractiveness” and “measurability’. These were
used to choose 40 variables out of 100 to establish a system to guide Seattle’s
transformation into a sustainable city. The system was then used as the basis
for an empirical study titled “Sustainable Seattle 1993”. The collection and
analysis of data has now begun on 20 of those 40 variables. Seattle also
launched a 20-year Comprehensive Plan 91994~2014) in 1994 with the goal of
developing Seattle into a sustainable city. The UNCHS (United Nations Center
for Human Settlements) and the World Bank partnered to propose the
Variables Programme (1997) for comparing the level of sustainable
development between the cities of the world (Table 1).
Table 1 The UNCHS (Habitat) Variables Programme (1997)
Land Use
City Population
Population Growth
Environmental
Management
Percentage of Wastewater
Treated
Solid Waste Generated
Disposal methods for Solid
Waste
Background |Woman Headed Regular Collection and
Data Households Transportation of Solid Waste
|Average Household Size Housing Destroyed
Household Formation Rate Major Sources of Income
Income Distribution Per-Capita Capital Expenditure
City Product per Person [Debt Services Charge
Tenure Type Local Government Employees
Households Below Poverty Wages in the Budget
Line Local
Informal or Unreported Government |Contracted Recurrent
Socio- Employment Expenditure Ratio
E Hospital Beds Government Level Providing
conomic :
Development = — Services 5
Child Mortality ‘Control by Higher Levels of
Government
School Classrooms Ratio of House Price to Income
Crime Rates Ratio of Renting Cost to Income)
Household Connection Housing Average Floor Area per Person
Levels Affordability
lnfrastrieture: [Access to Potable Water Durable Assets _
Consumption of Water Housing Commitment
Median Prices of Water, Planning Permission Multiplier
Scarce Season (Increase in Value)
Modal Split Housing Public Infrastructure Expenses
[Travel Time Supply Total Bond Value
Transportation /Expenditure on Road Housing Supply
Infrastructure
Automobile Ownership Ratio
Housing Investment
Source: UNCHS (Habitat), 1997
Sustainable development is
encompasses
the environmental,
a system-integration concept that
economic
and social dimensions.
Sustainability variables must therefore possess the following characteristics:
1.Support clear policies, targets and action plans for sustainable development;
2.Expresses the balance between environment, economy, technology,
industry and society;
3.Has a PRS (Pressure-State-Response) structure to allow the evaluation of
interactions between human activities and the environment;
4.Variables’ values must be measurable or at least observable. The data must
exist or be available;
5.The methodology for establishing the variables must be clear and
cost-effective;
6.The variables framework must be politically acceptable and has the ability to
promote or influence decision-making;
7.The variables must achieve widespread acceptance in society in order to
serve as an effective tool for exchange and communication between
sustainable development and society.
Sustainable development basically starts from ecological conservation
and the sustainable use of natural resources. Its essence is to balance
environmental protection with economic development. Sustainability variables
based on these concepts can not only serve as a tool in decision-making but
can also be used to evaluate overall progress in sustainable development. For
the general public, sustainability variables offer the dual advantages of being
“qualitative variables” and “quantitative variables” as well. The use of
sustainability variables facilitates the building of understanding and consensus.
The devising of sustainability variables therefore makes a worthwhile
contribution to sustainable development.
To create a comprehensive and practical sustainability variables system,
involves more than just clearly defining the variables and implementation
framework. The system must also respond appropriately to policy needs as
well. A sound variables system therefore not only reflects the current state of
development but must also taken into account the time factor and be
integrated with policy tools for it to be truly effective. Sustainability variables
therefore serve a range of purposes: a decision-making tool for sustainable
development; evaluate progress and trends in efforts at promoting sustainable
development; study the relationship between goal and target as well as
implementation performance; compare the effects of changes in time and
space; and provide warning information on environmental change. In this
study, a set of sustainable variables for urban development are proposed
based on the principles of sustainable development and the Taiwanese
experience in urban development. These are listed in Table 2.
Table 2 Sustainable urban development variables
Subsystem variable Subsystem variable
/ Number of companies Rate of unoccupied houses
g Productivity of IA staff BB § | Average housing price
z Industry value 3 2 2 | Rate of land development in IA
Zz Net revenue of capital E83 | Landarea per A staff
‘Amount of imports House rental rate in IA
3 ‘Amount of exports ‘Amount of dust fallen
R & D expenditure
Total population
Population growth rate
Natural increase rate
Total amount of suspended particulates
Daily sew age disposal per capita
Daily refuse production per capita
Amount of refuse collected per day
No. of motorcycles per 1000 persons
Social increase rate
Environmental pollution
subsystem
€
ra Average size of household No. of vehicles per 1000 persons
rd
z Urban-to-total population ratio Number of factories registered
a Population density Number of environmental pollution law suits
8 Age structure Amount of saving per household
s Education level Total regular income per family
a
é Water consumption per capita Housing-to-total family expenditure ratio
Power consumption per capita A Rate of self-owned houses
Population of IA staff g No. of automobiles per 1000 persons
Age structure of IA staff 3 Rate of unemployment
Education level of IA staff 2 Low income-to-total population ratio
= Area of agricultural land 8 No. of industrial units
2
§ ¢_ |_Urban-to-total area ratio a Indus trial population
2 a Urban area per capita Indus trial-to-total population ratio
3 g & [Population served by piped water Indus try value
= > ® | Residential floor area per capita Area of Indus trial land
2. The Fuzzy Delphi Method
The definition of detailed variables provides the basic elements and
references for understanding sustainable urban development. The relevant
literature was studied to compile a list of detailed variables in accordance with
the guidelines for variable selection. Urban characteristics unique to cities
were also taken into account in the interactions between the sub-systems
within the sustainable development system to arrive at a suite of variables for
urban dynamics model.
Large numbers of variables however make building the model more
complex and difficult. There are also some semantic uncertainties in how some
sustainability variables should be evaluated, making a clear answer difficult to
give. This study looked at the multitude of fuzzy theory derived analytical
methods before finally settling on the Fuzzy Delphi Method. This will be used
to establish a basis and method for evaluating an urban sustainability variables
system. A general outline of the Fuzzy Delphi Method's characteristics is
followed below.
The Fuzzy Delphi Method is an analytical method based on the Delphi
Method that draws on the ideas of the Fuzzy Theory. The Delphi Method is a
type of collective decision-making method (Linstone & Turoff, 2002), with
several rounds of anonymous written questionnaire surveys conducted to ask
for experts’ opinion. As a direct prediction method based on the expert
judgment and expert meeting investigation method, it possesses the following
properties:
1. Anonymity: The experts involved with the prediction process do not see
each other, remain anonymous and don’t know how many experts are
involved. This helps to prevent them from influencing and encourages
objectivity.
2. Feedback: The survey feedback gives the participants an idea about the
main ideas in the group. They can then draw from it information relevant to
them, make a new judgment, and then submit it to the group again.
3. Statistical: The expert opinions are processed statistically and a splines
graph produced with the expert opinion frequencies arrayed chronologically.
The top is the majority consensus (50% experts) representing the
prediction team’s opinion. The top and bottom quarter percentile (each
representing 25% of the experts) represent the prediction deviation.
4. Convergence: Through multiple reverse feedback make the final prediction
results converge.
The purpose of the Delphi Method is to achieve a consensus among the
experts on the subject being evaluated. When used with one-to-many
objectives, multi-principle, multi-proposal and multi-participant decision-making
problems, the method not only serves to draw on a large body of opinion but
also meets the requirement for independence in the experts’ judgment.
The Delphi Method requires multiple repetitions when asking experts for
their opinion. This must continue until the experts arrive at a consensus. As a
result, it generally has the following weaknesses: (Ho and Chen, 2007)
(1) Repeatedly surveying experts and collecting their opinions is very time
consuming.
(2) Experts must be surveyed and the collated results analyzed multiple times,
increasing costs.
(3) Expert cooperation is required before a consensus is reached, needlessly
increasing the difficulty of coordination and communication.
(4) Consensus of expert opinion occurs during a certain part of the analytical
process. The fuzziness of this part is however not taken into consideration.
This makes it easy to misinterpret the expert’s opinion.
(5) The analytical process has problems with some opinions being
systematically weakened or suppressed.
To solve the problem of fuzziness in expert consensus in group decision
making, researchers from around the world came up with new methods:
Murray, Pipino & Gigch (1985) proposed the application of Fuzzy Theory to the
7
Delphi Method, with semantic variables used to solve the problem with
fuzziness in the Delphi Method. Kir and Folger (1988) proposed a mean
normalization mode. Ishikawa ef al. (1993) used the Maximum-Minimum
Method together with cumulative frequency distribution and fuzzy scoring to
compile the expert opinions into fuzzy numbers. The expert prediction interval
value was then used to derive the fuzzy numbers, resulting in the Fuzzy Delphi
Method. Hsu and Chen (1996) proposed the fuzzy similarity aggregation
method. Using the similarity function, similarities between experts were
collated and fuzzy numbers assigned directly to each expert to determine the
agreement degree between them. The consensus coefficient was then used to
aggragate all experts’ fuzzy evaluation values. If the agreement degree
between experts is too low however the survey must be conducted again.
Acomparison of the strengths and weaknesses between the Fuzzy Delphi
Method and the traditional Delphi Method is provided below in Table 3.
Table 3 Comparison of the strengths and weaknesses between the Fuzzy
Delphi Method and the Delphi Method
Method Description Strengths and Weaknesses
Goal is to achieve consensus in [Takes more time to collate expert opinions.
lexpert opinion. Draws on a wide range |Higher cost.
lof opinions while providing quality of |Survey must be repeated multiple times.
Traditional|independent expert opinion. |The survey recovery rate is low.
Delphi The expert survey is repeated and|In pushing for a consensus it’ easy to
Method experts asked to revise their own misinterpret expert opinion.
opinions based on the results from the |Consensus of expert opinions only applies
previous survey until the opinions ito a certain rage. The fuzziness of that
converge. range is not taken into account.
As Delphi Method surveys have [Saves survey time.
some semantic fuzziness in both the Lower cost.
questions and the answers, cumulative/Reduces number of surveys, increases
frequency distribution and fuzzy questionnaire recovery rate.
scoring were therefore used to collate /Experts can fully express their opinions,
Fuzzy |the expert opinions into fuzzy lensuring the completeness and
Delphi |numbers. consistency of the group opinion.
Method Here similarity function is used to . .
[Takes into account the fuzziness that can’t
evaluate the agreement degree .
be avoided during the survey process.
between two experts. The consensus _. “6
_ Does not misinterpret experts’ original
coefficient for each expert was then _. e .
; lopinions and provides a true reflection of
used to derive the fuzzy evaluation :
their response.
value from all experts.
3. Questionnaires and Analysis Method
To establish a set of general variables for urban sustainable system
dynamics model, a general perspective of the urban development system must
be gained first. This study therefore carried out a questionnaire survey of
experts and academics in different fields. After collating the evaluating a wide
range of data, a total of 20 experts in the industry, academic, profession and
government organizations were selected for the survey. The goal of the expert
questionnaire survey was to gain an understanding of the variables that must
be taken into account within a sustainable urban system dynamics model. After
reviewing the relevant literature and developing a theoretical urban system,
this study began screening the variables. These could be sorted into three
parts according to content:
(1) Basic Information
Gender, age, specialization and level of professional experience.
(2) Questionnaire instructions
Provides instructions on how to answer the survey with samples provided.
This gave the respondents a better idea of the survey format, reducing
the time they need to spend and speeding up the survey.
(3) Variable definitions & answers
The survey focused on asking the respondents to rate the importance and
range of variables. A precise definition of each variable was also given for
the respondents to refer to when answering. The explanations were
provided as follows:
A. “Optimal” level of importance: Please evaluate the importance of this
and write down what you personally think is the optimal value.
B. Importance scope: Please evaluate the acceptable range for the
importance of this variable, and also write down what you think is the
maximum and minimum acceptable value for this variable.
9
Weighting:
0 1 3 5 7 9 10
Very Quite Quite Very
Unimportant Unimportant Important Important
Unimportant Neutral Important
Table 4 Importance of environmental variable to the influence of price of land development
Importance (0-10)
Unimportant
Neutral
Important
Quite Important
Very Important
Export evaluation
Variable
Very Unimportant
Quite Unimportant
ssibility of school
lity of cultural
Accessibility of park
The statistics from the survey showed that the experts’ average age was
44. Experts surveyed included those involved in the technology industry,
economics, environmental engineering, land administration, urban planning,
architecture and landscape. The most numerous were urban planning followed
by environmental engineering. The results of this expert survey were
representative of the perspectives from a range of different disciplines and
satisfied the requirement for a general perspective in this study.
Based on the collated information, this study conducted an expert
questionnaire survey on the 52 variables proposed in Table 2. The evaluation
method adopted was the evaluation fuzzy number defuzzification analysis
used by Tzeng (1993). In accordance with the Delphi Method, the 25%
percentile above and the 25% percentile below the median was used to
calculate the expert value. The magnitude of the expert value was then used
for screening the variables. In evaluation fuzzy number defuzzification analysis,
10
the fuzzy semantic level is converted into the variables and a more colloquial
format used to help experts perform their evaluation more precisely. The
function relationships were then subjected to defuzzification analysis to derive
the value for that evaluation factor. This method avoids the problem of experts
being constrained by their own maximum and minimum values. It also converts
the expert’s own fuzziness into an overall fuzzy evaluation of the variables.
This produces an evaluation result more in keeping with the overall expert
opinion. The calculation principles are described below:
Tzeng used a semantic approach to collect information on the
respondents’ preferences. That is, the “importance” semantic variable was
used as the interface to evenly divide the semantic scale into triangular fuzzy
numbers. This quantification was then used to calculate the membership as
shown in Fig. 1. For the fuzzy semantic scale, Tzeng proposed the concept
and method used by Chen and Hwang (1992) to convert the terms of semantic
expression into fuzzy numbers. The fuzzy set was then converted into crisp
score.
The solution for this analytical process is as follows:
1. Mean of the sustainable variables
First, use function (1) to calculate the mean of the triangular fuzzy
numbers for each variable. This gives a value for each variable that can be
used for defuzzification analysis.
Vim=(1/N)[nmi(0,0,1/(L-1))+11m2(0, 1/(L-1),2/(L-1)) +...
+ nm2((k-2)/(L-1),(k-D)/(L-1),K/(L- 1) +... Emm (K-2)(L-D), LD] see (1)
In function (1), Vm is the mean of the m-th variable; Nm is the number of
times that the k-th semantic scale was selected for the m-th variable. N is the
sum of nm; L is the number of partitions along the semantic scale; K is k-th
semantic scale.
Very Unimportant
Quite Unimportant
Unimportant
Neutral
Important
Quite Important
Very Important
Membership
>
0 14 2/6 3/6 4/6 5/6 1
Importance
Figure 1. Importance semantic variables
(2) Fuzzy Number Defuzzification Analysis
The goal of defuzzification analysis is to convert the triangular fuzzy
numbers into an exact value so the factors can be analyzed and ranked. The
steps are as follows: (1) Define the “maximum set” and “minimum set” of the
semantic scale set, where
Hmax (x) = x,0SxSl
{ O,otherwise —...... C2Y
Hmin (x) = foes
O,otherwise —...... (3)
Umax aNd Umin each intersects with the right and left boundary of Vs as
shown in Fig. 2. Vs=(a, b, c) is known and represents the three coordinates
(a,0), (b,1) and (c,0). Its triangular fuzzy numbers then form a fuzzy linear
equation:
y= (x-a)/(b-a) and (x-c) / (b-c)
The relationship between the maximum membership equation of Umax (x)
and the Vs fuzzy equation is as shown in function (4):
He (m) =sup min[[tmax (x) Vs (x) J]... (4)
In function (4), the two focal coordinates on the right boundary are:
(a/(1+a-b),a/(1+a-b)) and (c/(1+e-b),c/(1+e-b))
Take the y value of the larger y coordinate (membership) to represent u, (x).
Ho max (X) =X
Membership
o
‘©
0 a b c 1.0
Average of Triangular Fuzzy Number
Figure 2. Defuzzification Chart
The same also applies to the relationship between the minimum membership
equation of Umin (x) and the Vs fuzzy equation is as shown in function (5):
be Cm) =sup ming[fmin (x) wVs (x) J wee (5)
In function (5), the two focal coordinates on the left boundary are:
(b/(1+b-a),(1-a)/(1+b-a)) and (b/(1+b-c),(1-c)/(1+b-c))
Take the y value of the larger y coordinate (membership) to represent ut (x).
Finally, use defuzzification to calculate the fuzzy set and derive the
defuzzified point:
ur (m) =0.5[HR (m) +1-p, (m) ]
From this, it is possible to see that as the triangular fuzzy numbers tended
towards the right (importance is hence higher), its defuzzification value u: (m)
becomes greater. By using this method to establish the fuzzy semantic scale
allows the traditional method to be bypassed for fuzzy evaluation.
Respondents can then express their preferences in a colloquial everyday
13
manner.
4. Results
To ensure the objectivity of the screening process while screening the
variables according to the magnitude of their scores, this study chose to use
the method of Ho, Wang, & Lu (2002) and set separate thresholds for each
sub-system as it best matched the principles of this study. First, the arithmetic
mean of the internal variables in each sub-system was used as the baseline.
The standard deviation was then set as the cut-off point and variables
identified in expert feedback as being of relatively low importance were
removed. The arithmetic mean for the retained variables was then calculated
for use as the threshold value for that sub-system. This process avoided any
subjectivity in the setting of the threshold value while preserving the experts’
feedback on the importance of each sub-system. The relatively important
variables in each sub-system were thus retained.
Basically, after the evaluation fuzzy number defuzzification analysis the
variables with an expert score higher than the threshold value were retained
and the rest discarded. Table 5 lists the urban sustainability variables retained
by the screening process.
The screening results showed that 22 variables were chosen based on
their importance to the variable system. Of the 3 variables from the industry
sub-system, the industry output was the highest at 0.77. Of the 6 variables for
the population sub-system, population density was the highest at 0.76. Of the 6
variables in the residential land sub-system, the average amount of urban land
per capita was the highest at 0.72. For the 3 variables in the environmental
pollution sub-system, the wastewater generated per capita was the highest at
0.83. For the economy sub-system, industry output was the highest at 0.82.
Table 5 Sustainable variables and their evaluation value
Subsystem variable Value | Subsystem variable Value
Number of companies 0.66 Za Rate of unoccupied houses 0.70*
5 Productivity of IA staff one | 38 ‘Average housing price 0.70*
ra Industry value 0.77 | @ 2 | Rate ofland development in 1A 0.69*
4 Net revenue of capital 0.69 a3 Land area per IA staff 0.70*
2 Amount of imports 0.64 nike House rental rate in JA 0.58*
2 ‘Amount of exports 0.65 - ‘Amount of dust fallen 0.69
a R & D expenditure 0.70* & Total amount of suspended | 0.72
2 particulates
Total population 0.63 i Daily sew age disposal per capita | 0.83*
Population growth rate 0.74" 2 Daily refuse production per capita _| 0.81*
Natural increase rate 0.56 —_|_Amount of refuse collected per day _| 0.75
Social increase rate 0.70" | No. of motoreycles per 1000 persons | 0.64
Average size of household 0.63 2 No. of vehicles per 1000 persons 0.70
. Urban-to-total population ratio 0.65 g Number of factories registered 0.58
a Population density 0.76" & Number of environmental pollution [| 0.82*
4 law suits
g Age structure o71* ‘Amount of saving per household 0.68*
3 Education level om Total regular income per family o7i*
E Water consumption per capita 0.56 Housing-to-total family expenditure | 0.67
ratio
Power consumption per capita 057 5 Rate of self-owned houses 0.61
Population of IA staff 0.59 % _ [No. of automobiles per 1000 persons | 0.55
Age structure of IA staff 0.60 3 Rate of unemployment 0.70"
Education level of IA staff O71 B | Low income-to-total population | 0.57
z ratio
7 Area of agricultural land 0.46 a No. of industrial units 0.61
a2 Urban-to-total area ratio 0.58 Indus trial population 0.60
Fy 2 Urban area per capita 0.72" Indus trial-to-total population ratio _| 0.65
Ze Population served by piped water_| 0.65 Indus try value 0.82*
aie Residential floor area per capita | 0.70* Area of Indus trial land 0.64
Note: * A dopted by higher values than the threshold:0.695 (industry subsystem), 0.675 (Population subsystem), 0.669
(Housing/Landuse subsystem), 0.760 (Environmental pollution subsystem), 0.674 (Urban economy subsystem)
Drawing on the expert survey results and the system interaction
relationships, the author provides the following description of the
cause-and-effect relationships for the variables within the five sub-systems of
sustainable city development:
1. Industry Sub-System
In the industry sub-system, the city's secondary and tertiary industry
output is the primary stock while the number of employees is the secondary
stock. Here labor demand and labor productivity are the main factors that
influence changes in the sub-system. As the number of industry employees
increases, total industry output increases as well, creating a positive feedback
loop within the system. However, if industry development stagnates as the
number of employees increases, there is then a corresponding decrease in
unit productivity. This lowers the overall industry output, resulting in a negative
feedback loop within the system. Research and development is the basis for
industrial progress as the technical innovations it brings stimulate business
and industrial growth. The higher the industrial output, the more money is
spent on R&D. This generates more R&D results and boosts unit productivity.
The result is an increase in overall output and a positive influence on the
system as a whole. The industry sub-system therefore has 3 feedback loops,
with 2 positive and 1 negative.
2. Population Sub-System
In the city population sub-system the total population is the primary stock
in the cause-and-effect loop. The primary factors that influence the total
population of the Science City are natural population change and social
population change. Analysis of the expert surveys indicates that the social
growth rate is far more important than natural growth for population changes in
the Science City. Natural change is based on the permanent population while
social change is based on the transient population. Of these two key factors,
the cause-and-effect loop for natural change is influenced by the positive
contribution from the number of births and the negative contribution from the
number of deaths. As for social change, the number of immigrants makes a
positive contribution to the Science City population while the number of
emigrants makes a negative contribution. In terms of impact, as the number of
births increase, the total population increases as well. This forms a positive
feedback loop. As the total population increases, the number of deaths
increases as well. This leads to a decrease in the total population, forming a
negative feedback loop. With social change, if a region enjoys higher personal
income, better quality of life and has more employment opportunities, the
number of people who move to the region will increase. If the situation is
reversed, the contribution becomes negative as the number of people moving
16
out of the region increases instead. When the Park's industrial output is high,
its manpower demand will be high as well, attracting workers from other
regions; conversely, if the Park's high-tech industry output is decreasing, its
manpower demand will decrease as well, leading to a loss of population as
workers move away. The population sub-system therefore has 3 feedback
loops, including 2 positive and 1 negative.
3. Housing Sub-System
The housing sub-systems primary stocks are the number of households
and the number of housing units. City population growth generates a demand
for housing and stimulates the housing supply. This drives the development of
residential land and construction of housing, forming a positive feedback loop.
As the city's industrial output grows, this in turn drives the demand for labor in
the Park and brings in more workers. This has the indirect effect of increasing
builders' willingness to construct housing. The total number of housing
therefore increases, forming another positive feedback loop that affects
housing stock. Excessive residential development however leads to
over-supply and high vacancy rates. This impacts the willingness of
consumers and builders to buy or construct housing, resulting in a control loop
based on negative feedback. This sub-system therefore has 4 feedback loops
in total, with 2 positive and 2 negative.
4. Environmental Pollution Sub-System
The key issue in the environmental pollution sub-system is the level of
impact from environmental pollution. The main considerations during analysis
are the interactions between the Science City and the different types of
pollution sources. This study initially divided environmental pollution into three
types: water, air and waste. The analysis of the expert survey results indicated
that all of the air pollution related variables scored lower than the threshold
value. Air pollution is therefore not a significant issue for the Science City.
Waste is a byproduct of the industrial manufacturing processes. At Science
City, the main development focus is the high-technology industry. For the
high-technology industry, the source of main pollution is water pollution rather
than air pollution. Water and waste pollution is also produced by the population.
This study therefore chose to discount air pollution and base the
cause-and-effect loop for the Science City's environmental pollution
sub-system on water and waste pollution. In terms of impact, water pollution
can be reduced through the negative control of sewage treatment; waste
pollution can also be reduced through the negative control of waste collection
and processing. Pollution treatment therefore forms a negative feedback loop
that regulates this sub-system from the inside and prevents it from expanding
unchecked. This sub-system therefore has 4 feedback loops in total, with 2
positive and 2 negative.
5. Economy Sub-System
Once the expert survey opinions were collated, they showed that the
emphasis for this sub-system is on evaluating the change in total industrial
output. The system's cause-and-effect loop can be derived through the
number of industry workers, demand for labor and labor productivity. The
supply and demand of labor is the main factor that influences the total
industrial output. Generally speaking, if labor productivity is a constant, then
increasing the industrial output will increase the demand for labor. This results
in more industry workers, creating a positive feedback loop. As the number of
workers increases however, average productivity decreases for industrial
output that are fixed against time. If achieving the same industrial output is
used as the evaluation criteria, lower average productivity will consume more
resources and increase basic expenditure. This impacts indirectly on
willingness to invest in the industry and eventually slows the rate of growth in
industrial output. The result is a negative feedback loop and this is the most
important control loop within the city economy sub-system. As the city output
increases, the demand for labor from the service industry increases as well.
This forms another positive feedback loop within the system. This sub-system
therefore has 5 feedback loops in total, with 3 positive and 2 negative.
Based on the targets, implications and structure of the sub-systems
described above, this study looks at the relationships between the variables in
the sustainable urban development system and considers the City's
development characteristics. By linking the sub-systems together through
variables such as industrial output, population and pollution volume, a causal
loop for the sustainable urban development system was constructed (Fig 3). In
the diagram, there are 22 feedback loops in total, with 14 positive and 8
negative.
rate of unoccupied ——~w rate of land
birch rate __ dead rate a yo
erase
demand v housing
total ene “SB house unit price
ppsiaton —™
+ R&D
social . . expenditure
increase rate industry, Ee i.
industrial productivity
° +\ water 3 area A of 7 staff
Tefise potting
oo = total industry Ss aad of
( ‘ 7 S +4 value sta’
+ dipose of 7 \
refuse sewage dew of
disposal industr ___ Service labour
NCUSITY. industry value
amount LO value \ vs : =~ .
+ 7 >
demand of industry labor demand of
industry labors ~_ periodicity __ service
Sy wa industry labor
+
[EE industry \ 7
labors * | service
” labors
Fig 3 Causal loop diagram for sustainable urban development
5. Conclusion
To construct a sustainable urban development system, this study set the
development of the industry as the main axis while balancing the needs of the
urban ecological environment and social dimensions. The system's variable
sets were made up of 5 sub-systems: industry, population, housing/landuse,
environmental pollution and economy, By conducting an expert survey and
analyzing the results with the Fuzzy Delphi Method, an urban sustainability
variables system was established. This framework will serve as the foundation
for the construction of a system model in the future. After using the Fuzzy
Delphi Method to analyze the results, the ranking of the sub-systems’ screened
value from the highest to the lowest were: industry, environmental pollution,
population, economy then housing/landuse; the ranking of the threshold values
from the highest to the lowest were in order: environmental pollution, industry,
population, economy and housing & land. Examination of the variable values
showed that the 20 experts generally had assigned higher ratings to the
variables related to the industry sub-systems. This showed that the industry
was the most important for the city’s development. This result agreed with the
hypothesis of this study that the industry forms the main axis of development.
Apart from the industry, the variables related to the environmental pollution
sub-system also received higher ratings as well. This showed that the
environmental dimension still plays a very important role in sustainable
development.
References:
Chen, S. J., & Hwang, C. L., 1992, Fuzzy multiple attribute decision making method
and application, New-York:Springer- Verlag.
Hsu, H.M., & Chen, C. T., 1996, Aggregation of fuzzy opinions under group decision
making, Fuzzy Sets and Systems 79, pp.279-285.
Ho, Y.F., Wang, H. L. and Lu, C. H., 2002, The Dynamic Simulation Model for the
Sustainable Development of Taichung City, Journal of Architecture, No.41,
pp.107-128, Taipei.
Ho, Y.F., & Wang, H. L. 2005, Ssytem Dynamics Model for the Sustainable
Development of Science City, The 23" International Conference of System Dynamics
20
Society, Sloan School of Management, MIT, USA.
Ho, Y.F. and Chen, H.L., 2007, Healthy Housing Rating System, Journal of
Architecture, No. 60, pp.115-136, Taipei.
IUCN, UNEP & WWE, 1991, Caring for the earth, INCH, Switzerland.
Ishikawa, A., Amagasa, M., Shiga, T., Tomizawa, G., Tatsuta, R., & Mieno, H., 1993,
The Max-Min Delphi method and Fuzzy Delphi method via fuzzy integration, Fuzz
Sets and Systems, Vol.55, Issue 3, pp.241-253.
Klir, G J., & Folger, T. A., 1988, Fuzzy Sets, uncertainty, and information,
Prentic-Hall International.
Linstone, H. A., & Turoff, M., ed, 2002, The Delphi Method: Techniques and
Applications, ISBN 0-201-04294-0.
Murray, T. J., Pipino, L. L. and Gigch, J. P., 1985, A Pilot Study of Fuzzy Set
Modification of Delphi, Human Systems Management, 5, pp.76-80.
UNCHS, 1995, Using indicators in policy, Indicators Newsletter, 3 pp.1-8, United
Nations Center for Human Settlement, Habitat.
Tseng, G. H., 1993, Transport planning and Fuzzy multiple-objectives planning
method, Journal of Transport, No.3, pp.135-150, Taipei.
21
