Anda di halaman 1dari 26

A multiple criteria approach for the evaluation of the rail

transit networks in Istanbul



HALUK GEREK
1
, BIRSEN KARPAK
2
& TLAY KILINASLAN
3
1

Department of Civil Engineering, Istanbul Technical University, Ayazaga 80626, Istanbul,
Turkey (E-mail: hgercek@ins.itu.edu.tr);
2
Department of Management, Youngstown State
University, Youngstown, OH 44555-3071, USA (E-mail: bkarpak@cc.ysu.edu);
3
Department
of Urban and Regional Planning, Istanbul Technical University, Taksim 80191, Istanbul,
Turkey (E-mail: kaslan@itu.edu.tr)
Key words: analytic hierarchy process, MCDM application, rail transit networks, urban trans-
portation
Abstract. The deficiencies in the Istanbul transportation system have led the local authorities
to plan several alternative transportation projects. In this paper three alternative rail transit network
proposals are evaluated by using Analytic Hierarchy Process (AHP), a multiple criteria decision
support system. The AHP facilitates decision-making by organizing perceptions, experiences,
knowledge and judgments, the forces that influence the decision, into a hierarchical framework
with a goal, scenarios, criteria and alternatives of choice. Based on this analysis, the decision
makers have developed a new alternative as a combination of the most closely competing two
alternative rail transit networks. This combination rail transit network is currently under con-
struction.
1. Introduction
Urban transport systems are structured in an extremely complex way. The level
of infrastructure provided and the way it is managed affects the demand for
travel. Demand, which can vary by transportation mode, location, and time
of day, in turn affects the performance of the transportation system. In the
longer term, transport system performance affects land use patterns and this
further influences demand. In the short term, both the system performance
and the demand patterns produce both benefits and inconveniences for users,
residents and businesses. Therefore, it is important to ensure that any infra-
structure investment will have beneficial effects on the overall transport system
and those affected by it.
Many local authorities have recently responded to this challenge by devel-
oping integrated transport strategies that identify the ways in which
infrastructure provision, management, pricing and land use measures can be
combined most effectively to achieve a range of public policy objectives
(May 1991). These strategies have largely had some common elements. The
Transportation 31: 203228, 2004
2004 Kluwer Academic Publishers. Printed in the Netherlands.
starting point is usually a statement of vision for the urban area (often provided
by politicians) indicating the type of city that they wish to see in the horizon
year. These statements start by referring to economic activity, social equity,
and quality of life; and then raise the question of the role that transportation
should have in helping to realize that vision. This in turn leads to a specifi-
cation of transport policy objectives. These typically include such issues as
efficiency in the use of existing resources, improvement of accessibility, envi-
ronmental quality, and safety.
Istanbul, the largest urban settlement area in Turkey, has enormous com-
mercial, cultural, and historical significance. Although the average annual
growth rate of the population has declined to 3.2% in the last decade, the
city has experienced a dramatic growth in population at an average rate of
4.3% annually since 1960. The transportation system has been unable to keep
pace with this rapid growth and the changing urban structure. Mobility and
accessibility has been declining, as it is in most of the developing world
(Gakenheimer 1999). Like many other metropolitan areas in the world traffic
congestion is currently a major problem in Istanbul. Twenty-four percent of
the total motor vehicles (2.3 million motor vehicles) and 31% of the total
private cars (160 cars per 1,000 inhabitants) in Turkey are registered in
Istanbul. Car ownership and car use have increased dramatically in the last
decade. Motorization has increased eight times between 1980 and 2000, while
population has increased 112% (Gerek 2002).
The deficiencies in the Istanbul transportation system have led the local and
central governments to plan and implement several major transportation
projects. Local authorities have been struggling under the pressure of urban-
ization without sufficient funds to accommodate growth. Because changes in
the transportation infrastructure require large amounts of capital investment,
sound decisions are obviously of vital importance for the local authorities
(Pucher 1999).
The municipality of Istanbul has already built and put into operation an
18 km-long light rail transit (LRT) system on the European side of the city.
A subway line is under construction (part of it is already available for public
service), but, under the pressure of increasing travel demand in the area, new
extension lines are needed.
In 1995, the Greater Istanbul Municipality (GIM) commissioned a con-
sortium to carry out a strategic network planning study to identify the best
rail transit projects to improve the public transportation system in the European
side of the city (Tekfen & Delcan 1996). Eight rail transit networks were
proposed for further analysis. Later eight alternative rail transit networks
were separated into three groups, with one alternative from each group being
selected to be in the final evaluation.
In evaluating the three alternative rail transit networks, the decision makers
204
were faced with multiple conflicting criteria. Under these circumstances the
conventional approach is to use a scoring methodology that requires direct
estimation of the importance of each criterion. However, recent research has
shown that direct estimation of the importance rating of each criterion is not
the best way to get at the decision makers real preference for one criterion
over the other (Millet 1997). In addition, the scoring methodology works
best with criteria that can be quantified; intangible criteria such as Staging
flexibility of the rail transit network could not be included.
In this article, we show how to evaluate the final three alternatives by
using the Analytic Hierarchy Process (AHP), a multicriteria decision support
system. The AHP overcomes most of the disadvantages of scoring methods.
It is a powerful and flexible decision making process to help decision makers
set priorities and make the best decision when both qualitative and quantita-
tive aspects of a decision need to be considered. It facilitates decision-making
by providing a framework for organizing perceptions, feelings, experiences,
knowledge and judgments (Saaty 1990b).
The article consists of six sections. After the introduction, future land use
and travel demand in Istanbul are described. The AHPs methodology is
explained briefly in Section 3, and an AHP model of alternative rail networks
is presented in the fourth section. Section 5, displays the results and sensitivity
analysis. The paper ends with a discussion of the findings in the conclusion
section.
2. Future land use and travel demand in Istanbul
In 2000, the metropolitan population of Istanbul was around 10 million.
Sixty-five percent of the population lived on the European side. About one
third of the total population was employed with seventy-four percent of those
having jobs on the European side of the city (Gerek 2002).
Public transportation needs in the city are served by several public trans-
port modes: buses, minibuses, commuter rail and light rail transit lines, with
a limited number of ferries and sea buses connecting the old city with the
villages of the Bosphorus and the Asiatic coast of the Marmara Sea and the
Golden Horn (Figure 1).
Two land use scenarios had been developed by the Consortium in order
to predict the future travel demand in the horizon year of 2010. The first
scenario (The Trend Scenario) assumed that the city will continue to grow with
the present trends, and presented an uncontrolled growth approach rather
than the land use decisions in the Master Plan. In this scenario, the popula-
tion and employment in 2010 were estimated at 16.6 million and 6.3 million,
respectively. Sixty-four percent of the population and 69 percent of the employ-
205
ment were expected to be located on the European side (Tekfen & Delcan
1996).
The second scenario (The Master Plan Scenario) assumed that the city would
develop according to the land use decisions established by the GIM in the
Istanbul Metropolitan Area Sub-Region Master Plan (1995). The main strate-
gies of this master plan were as follows:
The physical growth of the metropolitan area must be controlled and a
balanced distribution of population and employment between two sides of
the city should be ensured.
The historical and natural structure of the peninsula must be preserved
and its development as a historical trade and tourism center should be
ensured.
Encouraging the development of sub-centers located in and beyond the
city boundary should help decentralize the population from the highly
populated urban areas.
Transportation demand should be reduced by better allocating the way
land is used in the metropolitan area.
206
Source: IETT Toplu Tasma Rehberi (Public Transportation Guide), Istanbul Municipality,
Istanbul.
Figure 1. General city map of Istanbul metropolitan area and public transport routes.
Based on these strategies, the Master Plan envisaged a population of 15 million
in 2010, of which 68 percent will be living in the European side of the city.
The number of working people is estimated to be about 6 million in 2010,
of which 68 percent will be working on the European side.
Based on the land use data, the travel demand in the horizon year for each
scenario was then predicted by using a conventional four-stage transporta-
tion model (ITU 1999). Four major demand corridors were identified in the
study area by using the transport model (Tekfen & Delcan 1996).
Rail transit networks were developed for further analysis based on the desire
lines between the major generation and attraction centers. The assumptions
made for all rail transit networks were as follows (Tekfen & Delcan 1996):
The north-south subway line, of which the first phase is already under
construction, will be completed.
The commuter railway lines on two sides of the city will be upgraded and
connected with a railway tube tunnel under the Bosphorus. The north-
south subway line and the existing LRT line will be connected to the
commuter railway on the European side.
3. Methodology
For this study, the Analytic Hierarchy Process (AHP) was selected as the
decision analysis tool. The AHP is based on the innate human ability to make
sound judgments about problems. Through trade-offs it clarifies the advantages
and disadvantages of policy options under circumstances of risk and uncer-
tainty. The AHP is about breaking down a problem and aggregating the
solutions of all the subproblems into a conclusion (Saaty 1990b).
In the AHP, a problem is structured as a hierarchy. A useful way to proceed
in structuring a decision is to come down from the goal as far as one can,
then start with the decision alternatives and work up until the top-down and
bottom-up structures can be connected. Once the hierarchy has been con-
structed, the decision-maker begins the prioritization procedure to determine
the relative importance of the elements in each level. Prioritization involves
eliciting judgments in response to questions about the dominance of one
element over another with respect to a property. Elements in each level are
compared pairwise in terms of their importance with respect to an element
in the next higher level.
The scale used for comparisons in AHP enables the decision-maker to
incorporate experience and knowledge intuitively (Harper & Vargas 1990). The
decision-maker can express his preference between every two elements
verbally. These descriptive preferences would then be translated into numer-
207
ical ratings 1, 3, 5, 7 and 9, respectively with 2, 4, 6 and 8 as intermediate
values for compromises between two successive qualitative judgments (Saaty
1990a). This scale is insensitive to small changes in decision makers
preferences, thereby minimizing the effect of uncertainty in evaluations (see
Table 1).
In eliciting judgments from decision-makers, one makes redundant com-
parisons to improve the validity of the answer, given that respondents may
be uncertain or may make poor judgments. Redundancy gives rise to multiple
comparisons of an element with other elements and hence to numerical incon-
sistencies (Saaty 1994). The amount of inconsistency present is expressed
by the inconsistency ratio (IR) (Saaty 1995).
Table 1. Nine point scale for pairwise comparisons.
Intensity of Definition Explanation
importance
001 Equal importance Two activities contribute equally to the objective
003 Moderate importance Experience and judgment slightly favor one
activity over another
005 Strong importance Experience and judgment strongly favor one
activity over another
007 Very strong importance An activity is favored very strongly over another
009 Extreme importance The evidence favoring one activity over another
is of the highest possible order of affirmation
There are easy-to-use packages that implement this approach and simplify
complex decision-making process considerably. Expert Choice

is the name of
one such software. (A free trial version is available from www.expertchoice.
com.) Expert Choice produces a measure of inconsistency (Forman & Saaty
1995). This measure is useful in identifying possible errors in expressing
judgments as well as actual inconsistencies in the judgments themselves.
Suggestions can be requested from the evaluators for improving consistency.
After forming the preference matrices, the process moves to deriving relative
weights for the various elements (see Appendix 1). Finally, all the compar-
isons are synthesized to rank the alternatives. The result is a set of priorities
for the alternatives. Sensitivity analysis is used to investigate the sensitivity
of the alternatives when the priorities of the criteria are changed.
Expert Choice provides various options for comparing criteria/subcriteria/
alternatives: the first is the numerical option where the decision making team
can enter numerical values between 1 and 9. The second is the verbal option
where semantic terms such as moderately more important, strongly more
important can be used. The third option is the graphical option where the
208
decision making team can make comparisons by contrasting graphical bars.
Graphical multiple bars and graphical pairwise comparisons are the two sub-
options under graphical option. As the name implies graphical multiple bars
evaluate all criteria/sub-criteria/alternatives simultaneously whereas in graph-
ical pairwise comparison option, decision-makers compare criteria/sub-criteria/
alternatives two at a time. The fourth option is the direct estimation where
the user simply produces a set of values reflecting the relative preference
for the compared elements. Millet (1997) evaluated different comparison modes
and found that direct estimation, and numerical pairwise comparison methods
are dominated by verbal comparison, graphical pairwise comparison and graph-
ical multiple bars based on perceived ease of use and accuracy criteria.
4. AHP model of alternative rail transit networks
There were a decision-making team and an expert team that carried out this
evaluation jointly. The expert team consisted of a transport expert, an expert
in quantitative decision-making, and a city planner. The decision-making
team consisted of the Mayor of Istanbul, The City Council Members, and
various bureaucrats. In this study, the

Goal was to prioritize the rail transit
network alternatives for the European side of the city.
Public transport, in general, has three main functionalities: commercial,
substitution, and social. The commercial functionality is primarily important
to the public transport systems operator trying to maximize revenues. In general
this objective can be expressed as maximize the number of passengers
served.
The substitution functionality is important for the policy makers. Motor
vehicle traffic is now considered as the major factor causing air pollution.
For a sustainable urban transportation system, the decision makers argue that
the growth of private car use in congested urban areas has to be reduced and
a shift from car to public transport is desirable. Thus, the objective of the policy
makers can be expressed as minimize the amount of car-kilometers driven in
the city.
The social functionality is less clear than the former two. In general, it
can be stated that it is a social goal in society that everyone must be able to
travel. Thus a societal goal may be expressed as everyone must have access
to the public transport system.
4.1. The criteria
The criteria of this study were developed based on these considerations. A long
list of possible measures was created based on the expert teams experience
209
and suggestions from the decision-making team. The main criteria are defined
as financial, economic, system planning, and policy.
Financial criterion: Three sub-criteria represent financial criterion:
Operating and maintenance costs (OPERATIN) of the rail networks
Network construction costs (CONSTRUC) of the alternative rail transit
networks
Rail vehicle purchasing costs (VEHICLE).
Rail transit operating and maintenance costs were calculated on the basis of
passenger-kilometers carried on each rail transit network.
Vehicle purchasing costs were calculated on the basis of the number of
rail transit vehicles required to meet future traffic demand.
Economic (ECONOMIC) criterion consisted of the following sub criteria:
Road vehicle operating costs (ROADVOC)
Road vehicle purchasing costs (ROADVIN)
Road accident costs (ACCIDENT)
Total trip time (TIME)
Environmental cost (ENVIRON)
Improving the existing rail transit system is expected to cause some changes
in the travel pattern of the commuters such as shifting from road to rail, and
this, in turn, will result in a decrease in urban road transport costs. The
transport model that was used to predict the impacts of the rail transit networks
on future travel patterns was multi-modal, i.e.; it included all transportation
modes (rail, road and sea) in the city as a whole. It was calibrated and vali-
dated with land-use and traffic data collected from 209 planning zones in
the metropolitan area in 1997. The model predicts the traffic flows by modes
of transport in respond to the changes in the land use pattern and the trans-
portation system in the city. For each rail transit network option, road transport
costs were calculated on the basis of the road vehicle-kms (private cars, buses
and minibuses) predicted to be driven on the roads of the metropolitan area.
Total trip time was the sum of the trip times of all commuters traveling
by all transport modes in the city. It was considered among the economic
criteria because the time spent in traveling is assumed to have an economic
value.
Road vehicle purchasing costs, cost of road accidents, and environmental
cost of each rail network were also calculated based on the number of road
vehicle-kms (private cars, buses and minibuses) traveled on the roads of the
metropolitan area. In the field of road traffic, there have been improvements
such as use of lead free gasoline and LPG (liquid petroleum gas) that have
210
decreased the emissions caused by road traffic. However, the rapid increase
in private cars and the longer travel lengths have limited the positive impacts
of these improvements. Total emissions caused by road traffic in Istanbul
have increased about 50% between 1990 and 2000 (Gerek 2002). Therefore,
it was assumed that the increase in the number of motor vehicles would
increase the road vehicle purchasing cost, cost of road accidents, and envi-
ronmental pollution.
System Planning (SYSPLAN) criteria consisted of the following items:
Capacity provided (CAPACITY) by the proposed rail transit network
Accessibility to major traffic generation and attraction centers (ACCESS)
Integration with other public transport modes (INTEGRAT)
Conformity of the alternative network system to the Master Plan 2010
(CONFMPL).
The capacity provided by each of the rail transit networks was calculated as
the total demand that could be met by the scheduled train services in the
peak hour over the entire rail network. Accessibility was measured as the
total number of major traffic generation and attraction centers in the city that
could be covered by the proposed rail transit network. The integration of the
rail transit system with other public transport modes criterion was further
subdivided into the connection to the main inter-city bus terminal of the city
(BUSTERM), the airport connection (AIRPORT), and the railway tube tunnel
connection to the future Bosphorus rail crossing (BOSCROSS).
Policy (POLICY) criteria consisted of the following sub criteria:
Staging flexibility of the rail transit network (STAGFLX)
Expropriation
1
difficulty (EXPROPRI)
Construction period (CONSPER)
Effective use of capacity of the existing LRT network (EXLRTUSE)
Staging flexibility represented the possibility of having a number of options
for extending the existing rail network under different circumstances in the
future with less negative impacts on the operation of the existing rail transit
services. Expropriation difficulty was measured by the total land expropriation
(land acquisition) cost of each network option. Construction period of each
alternative was estimated by considering not only the construction time of
the whole network but also the construction time of the underground sections
which would take much longer in tunneling as compared to the line sections
to be built at grade. Effective use of the capacity of the existing LRT network
was an important criterion from the operators point of view and it should
be considered together with the staging flexibility. In other words, each new
211
line to be added to the current network was expected to increase the rider-
ship on the existing lines, which are partly being operated under capacity.
Each alternative network represented the final configuration of the rail transit
network in the European side of the city in 2010. They are networks 1, 5
and 6 (Figures 2a, b and c).
An AHP model of alternative rail transit networks is given in Figure 3.
4.2. Scenarios, judgments and comparisons
It is extremely difficult to predict the future development of a fast growing
city like Istanbul. The historical development of the city has shown that the
master city plans that aimed at controlling the growth of the city in the past
have hardly been implemented for social, economical and political reasons,
and the city has generally grown beyond the limits of the development plans.
Therefore, it can be easily argued that the likelihood that the city will develop
according to the Master Plan (the second scenario) is not very high. At the
outset, it is estimated that this likelihood will be 33 percent and therefore
the expected growth of the city according to trend, the first scenario, will be
67 percent.
212
1218: Peak hour travel volumes are forecast for 2010 as 1,000 passengers per hour per
direction. Demand range covers both scenarios.
Figure 2a. Rail Transit Network 1.
213
1218: Peak hour travel volumes are forecast for 2010 as 1,000 passengers per hour per
direction. Demand range covers both scenarios.
Figure 2c. Rail Transit Network 6.
1218: Peak hour travel volumes are forecast for 2010 as 1,000 passengers per hour per
direction. Demand range covers both scenarios.
Figure 2b. Rail Transit Network 5.
The importance of the criteria was determined by our decision-making team
with the assistance of the expert team. The expert team facilitated the process
by explaining the relatively accurate and easy to use comparison methods
available in Expert Choice to the decision-making team. The decision-makers
themselves rejected the verbal comparison mode because of perceived diffi-
culty and the graphical multiple bars because of inaccuracy. Our decision-
making team found the graphical pairwise comparison mode to be the most
appealing and used it in all the following evaluations. Expert Choice trans-
lates these graphical preferences into numerical ratings to form pairwise
comparison matrix. These numerical ratings can be fractional numbers rather
than the whole numbers of the basic fundamental AHP scale since the graph-
ical mode is continuous.
4.2.1. Pairwise comparisons under trend scenario
Pairwise comparisons of main criteria. In the first matrix of comparisons of
the four main criteria with respect to the goal of choosing the best rail transit
network alternative, Financial Criterion was rated between equally to mod-
erately (closer to equal) more important than System Planning Criterion by
using graphical pairwise comparison mode and numerical translation of this
214
Figure 3. AHP model of rail transit networks.
preference given by the software was 1.2 and it was assigned in the first
row, third column (1, 3) position of the matrix (Table 2).
Similarly the score of 1.3 corresponding between equal to moderate (closer
to equal) dominance or importance, was assigned to Economic Criterion
over System Planning Criterion in the (2, 3) position with corresponding
reciprocals in the transpose positions of the matrix. Priorities derived based
on these comparisons were given in the right-most column of the matrix.
Financial criterion was the most important criterion (30 percent importance)
followed by economic, system planning, and policy criteria.
Table 2. Main criteria matrix.
Financial Economic System planning Policy Priority
Financial 1 1.4 1.2 1.3 0.301
Economic 1/1.4 1 1.3 1.3 0.262
System planning 1/1.2 1/1.3 1 1.1 0.227
Policy 1/1.3 1/1.3 1/1.1 1 0.210
Pairwise comparisons of secondary (or sub) criteria. Financial Criterion is
divided into three sub-criteria: Network Construction Costs, Rail Vehicle
Purchasing Costs, and Operating and Maintenance Cost. The importance of
these sub-criteria was considered to be proportional to the net present values
of the cost streams to be accrued within an evaluation period of 30 years
($ US 8,253 million; $ US 7,398 million, and $ US 2,143 respectively) and
determined accordingly. Therefore, their importance rating was 0.458, 0.416,
and 0.126, respectively. (Table 3)
Economic criterion is divided into five sub-criteria. Pairwise comparison
matrix of these sub-criteria and priorities as well priorities under system
planning and policy were given in Table 3.
Pairwise comparisons of tertiary (or sub-sub) criteria. Only integration with
other public transport modes was further subdivided into the fourth level,
tertiary sub-criteria. Priorities of the tertiary sub-criteria, connection to the
Main Inter-city Bus Terminal of the city (BUSTERM), the Airport connec-
tion (AIRPORT), and the connection to the future Bosphorus Rail Tunnel
crossing (BOSCROSS) are given in the priority column of Table 4.
Pairwise comparisons of the networks. Under all of the lowest level criteria
given above the decision-makers gave judgments on their preference for rail
transit network 1, 5 and 6, comparing them in pairs. The data given in
Appendix 2 were helpful in deriving these judgments. Results are given in
Table 5.
215
4.2.2. Pairwise comparisons under master plan scenario
Most of the pairwise comparisons made earlier under the first scenario
remained the same under the Master Plan Scenario. So in this section we
give only those comparisons, which changed under Master Plan scenario.
216
Table 3. Matrices for evaluating the sub-criteria under financial, economic, system planning, and
policy criteria.
Financial (1) (2) (3) Priority
(1) Network construction costs 1 1 4 0.458
(2) Rail vehicle purchasing costs 1 1 3 0.416
(3) Operating and maintenance costs 1/4 1/3 1 0.126
IR = 0.01.
Economic (1) (2) (3) (4) (5) Priority
(1) Road vehicle operating costs 1 1.6 2 3.2 1.6 0.328
(2) Road vehicle purchasing costs 1/1.6 1 1.7 1.7 1.5 0.228
(3) Road accident costs 1/2 1/1.7 1 1/1.2 1/1.9 0.126
(4) Total trip time 1/3.2 1/1.7 1.2 1 1.3 0.131
(5) Environmental costs 1/1.6 1/1.5 1.9 1/1.3 1 0.186
IR = 0.01.
System planning (1) (2) (3) (4) Priority
(1) Capacity 1 1.4 1.5 3.7 0.394
(2) Accessibility to major traffic
generation and attraction centers 1/1.4 1 1 1.6 0.238
(3) Integration with other public
transport modes 1/1.5 1 1 1.6 0.233
(4) Conformity to the master plan 2010 1/3.7 1/1.6 1/1.6 1 0.134
IR = 0.01.
Policy (1) (2) (3) (4) Priority
(1) Staging flexibility of the rail
transit network 1 1.3 1.2 3.4 0.300
(2) Expropriation difficulty 1/1.3 1 1.1 3.1 0.280
(3) Construction period 1/1.2 1/1.1 1 5.5 0.241
(4) Effective use of capacity of the
existing LRT network 1/3.4 1/3.1 1/5.5 1 0.079
IR = 0.02.
For example system planning priorities under the Master Plan were different
from their priorities under the Trend Scenario. The importance of confor-
mity to the Master Plan with respect to the capacity provided by the rail
transit network went up. Priorities of the other two sub-criteria remained almost
the same
Integration with other public transportation modes was another example
where the priorities were different. The priorities of the sub-criteria of the
policy criterion also changed. The Importance of expropriation difficulty
declined versus the importance of construction period.
5. Results and sensitivity analysis
In the final step all the comparisons are combined, or synthesized, to rank
the alternatives. The combined priorities of each alternative under the different
scenarios are given in Table 5. They are obtained as the sum of the products
of criteria priorities column times the priorities under each alternative
network. Under trend Network 6 is the best at 0.344 whereas it is second
best under master plan scenario. Assuming that the likelihood of Trend scenario
is 67 percent and implementing the Master plan is only 33 percent, overall,
the Rail Transit Network 6 received the highest ranking (0.67

0.344 +
0.33 0.335 = 0.341) followed by Network 1 (0.338) (Figure 4). Network 1
was better than Network 6 under the Master Plan Scenario, whereas Network
6 outranked Network 1 under the Trend Scenario. Networks 6 and 1 domi-
nated Network 5, under both scenarios.
A sensitivity analysis was performed and presented to the decision-makers.
The results of the sensitivity analysis showed that the Master Plan and the
Trend Scenarios are not significantly different from each other. Under both
scenarios, Network 6 performs considerably better than Network 1 under the
system-planning criterion, whereas under the policy criterion Network 1 is
the best rail transit network followed by Network 5 (Figure 5). If the system-
planning criterion gains more importance, e.g. its importance rating increases
217
Table 4. Priorities of tertiary sub-criteria with respect to the Integration with other Public
Transport Modes sub-criterion under Master Plan Scenario.
Priority
(1) Connection to the inter-city bus terminal 0.194
(2) Airport connection 0.063
(3) Connection to the Bosphorus rail crossing 0.743
IR = 0.01.
218
T
a
b
l
e

5
.

P
r
i
o
r
i
t
i
e
s

o
f

t
h
e

t
h
r
e
e

r
a
i
l

t
r
a
n
s
i
t

n
e
t
w
o
r
k
s

f
o
r

b
o
t
h

s
c
e
n
a
r
i
o
s
.
C
r
i
t
e
r
a
P
r
i
o
r
i
t
i
e
s
T
r
e
n
d
M
a
s
t
e
r

P
l
a
n
C
r
i
t
e
r
i
a

R
T
N
1
R
T
N
5
R
T
N
6
C
r
i
t
e
r
i
a

R
T
N
1
R
T
N
5
R
T
N
6
p
r
i
o
r
i
t
i
e
s
p
r
i
o
r
i
t
i
e
s
O
p
e
r
a
t
i
n
g

a
n
d

m
a
i
n
t
e
n
a
n
c
e

c
o
s
t
s
0
.
0
3
6
0
.
3
4
7
0
.
3
2
6
0
.
3
2
8
0
.
0
3
6
0
.
3
4
5
0
.
3
2
9
0
.
3
2
6
N
e
t
w
o
r
k

c
o
n
s
t
r
u
c
t
i
o
n

c
o
s
t
0
.
1
4
0
0
.
3
7
0
0
.
3
6
5
0
.
2
6
6
0
.
1
3
9
0
.
3
7
0
0
.
3
6
5
0
.
2
6
6
R
a
i
l

v
e
h
i
c
l
e

p
u
r
c
h
a
s
i
n
g

c
o
s
t
s
0
.
1
2
5
0
.
3
2
4
0
.
3
1
3
0
.
3
6
3
0
.
1
2
4
0
.
3
7
2
0
.
3
1
6
0
.
3
1
2
R
o
a
d

v
e
h
i
c
l
e

o
p
e
r
a
t
i
n
g

c
o
s
t
s
0
.
0
8
7
0
.
3
3
5
0
.
3
3
4
0
.
3
3
1
0
.
0
8
5
0
.
3
3
4
0
.
3
3
4
0
.
3
3
2
R
o
a
d

v
e
h
i
c
l
e

p
u
r
c
h
a
s
i
n
g

c
o
s
t
s
0
.
0
6
0
0
.
3
3
4
0
.
3
3
4
0
.
3
3
2
0
.
0
6
1
0
.
3
3
4
0
.
3
3
4
0
.
3
3
2
R
o
a
d

a
c
c
i
d
e
n
t

c
o
s
t
s
0
.
0
3
3
0
.
3
3
4
0
.
3
3
4
0
.
3
3
2
0
.
0
3
3
0
.
3
3
4
0
.
3
3
4
0
.
3
3
2
T
o
t
a
l

t
r
i
p

t
i
m
e
0
.
0
3
4
0
.
3
3
7
0
.
3
3
5
0
.
3
2
8
0
.
0
3
3
0
.
3
3
5
0
.
3
3
1
0
.
3
3
4
E
n
v
i
r
o
n
m
e
n
t
a
l

c
o
s
t
s
0
.
0
4
9
0
.
3
3
4
0
.
3
3
4
0
.
3
3
2
0
.
0
4
8
0
.
3
3
4
0
.
3
3
4
0
.
3
3
2
C
a
p
a
c
i
t
y
0
.
0
9
0
0
.
3
1
1
0
.
3
0
9
0
.
3
8
0
0
.
0
7
9
0
.
3
0
8
0
.
3
1
0
0
.
3
8
2
A
c
c
e
s
s
i
b
i
l
i
t
y
0
.
0
5
4
0
.
2
5
8
0
.
1
0
5
0
.
6
3
7
0
.
0
5
2
0
.
2
5
8
0
.
1
0
5
0
.
6
3
7
C
o
n
n
e
c
t
i
o
n

t
o

t
h
e

b
u
s

t
e
r
m
i
n
a
l
0
.
0
1
0
0
.
1
2
4
0
.
3
6
5
0
.
5
1
0
0
.
0
0
9
0
.
3
8
7
0
.
2
4
7
0
.
3
6
6
A
i
r
p
o
r
t

c
o
n
n
e
c
t
i
o
n
0
.
0
0
3
0
.
3
3
3
0
.
3
3
3
0
.
3
3
3
0
.
0
0
3
0
.
3
3
3
0
.
3
3
3
0
.
3
3
3
C
o
n
n
e
c
t
i
o
n

t
o

t
h
e

b
o
s
p
h
o
r
u
s

c
r
o
s
s
i
n
g
0
.
0
3
9
0
.
3
3
2
0
.
2
5
4
0
.
4
2
4
0
.
0
3
9
0
.
3
5
6
0
.
2
3
4
0
.
4
1
1
C
o
n
f
o
r
m
i
t
y

t
o

t
h
e

M
P
0
.
0
3
0
0
.
3
0
2
0
.
2
9
9
0
.
4
0
0
0
.
0
4
2
0
.
3
0
2
0
.
2
9
9
0
.
4
0
0
S
t
a
g
i
n
g

f
l
e
x
i
b
i
l
i
t
y
0
.
0
6
3
0
.
3
5
5
0
.
3
5
7
0
.
2
8
8
0
.
0
6
4
0
.
3
5
5
0
.
3
5
7
0
.
2
8
8
E
x
p
r
o
p
r
i
a
t
i
o
n

d
i
f
f
i
c
u
l
t
y
0
.
0
5
8
0
.
2
7
2
0
.
3
9
6
0
.
3
3
2
0
.
0
5
8
0
.
2
7
2
0
.
3
9
6
0
.
3
3
2
C
o
n
s
t
r
u
c
t
i
o
n

p
e
r
i
o
d
0
.
0
7
2
0
.
4
7
7
0
.
3
1
3
0
.
2
1
0
0
.
0
7
3
0
.
4
7
7
0
.
3
1
3
0
.
2
1
0
E
f
f
e
c
t
i
v
e

L
R
T

u
s
e

0
.
0
1
6
0
.
2
6
5
0
.
3
7
0
0
.
3
6
6
0
.
0
1
5
0
.
2
6
5
0
.
3
7
0
0
.
3
6
6
O
v
e
r
a
l
l

p
r
i
o
r
i
t
y
0
.
3
3
5
0
.
3
2
1
0
.
3
4
4
0
.
3
4
5
0
.
3
2
0
0
.
3
3
5
from 23 percent to 33 percent, dominance of Network 6 over Network 1
becomes even more significant.
Network 6 is the most expensive alternative. Therefore it ranks last according
to the financial criterion. On the other hand, Network 1 is the best alterna-
tive under the financial criterion. When we performed the sensitivity analysis
219
Figure 4. Priorities of the Three Rail Transit Networks under both scenarios.
by varying the importance of the different criteria, Network 6 became as
preferable as Network 1 only if the importance rating of financial criterion were
raised to 75 percent from 30 percent under the Trend Scenario; under the
Master plan Scenario, it had to go up to 95 percent. Since it is unlikely that
the importance rating of the financial criterion would be dominant by that
much, Network 1s dominance was not sensitive to changes in the impor-
tance of financial criterion. We can even claim that whatever the importance
220
Figure 5. Sensitivity analysis under trend scenario.
rating of the financial criterion is, Network 1 is the best choice under the finan-
cial criterion.
6. Conclusions
Final decision of GIM
The Greater Istanbul Municipality (GIM), commissioned a consortium to carry
out a strategic rail transit network planning study on ways to improve the public
transportation system on the European side of the city. There were eight
alternative rail transit networks based on two land use scenarios that were
proposed.
The eight rail transit network proposals were first divided into three groups
with one alternative from each group being selected for the final competi-
tion. In this article, the final three alternatives were evaluated by using AHP.
They were prioritized according to the criteria that were defined for the
commercial, substitution and social functions of the public transport system.
Results indicated that overall the Alternative Rail Network 6 that would serve
densely populated areas was preferred. This network was also the best per-
former under the Trend scenario. The Alternative Rail Network 1, which would
serve the peripheries of the city, was the second choice. Network 1 supports
the master plan strategy mainly because it would help decentralize the popu-
lation by encouraging the development of sub-centers. It was the best one under
the Master plan scenario.
The public and the GIM both like the idea of preserving the historical and
natural structure of the city. This idea of preservation was one of the Master
Plan strategies. The Alternative Rail Transit 5 which radiates from the his-
torical peninsula creates a negative effect on the historical sites. As a result,
this network ranks last under both scenarios.
The sensitivity analysis indicated that if the likelihood that the city will
develop according to the Master Plan increases to 45 percent, overall
Alternative Rail Network 6 and Alternative Rail Network 1 become equally
preferable. This means that if local (including central planning) authorities
strengthen the control mechanism over the city according to the Master Plan
strategies, Alternative Rail Network 1 would receive the highest ranking
because it would be best for creating better living conditions in the city. This
situation directly depends on the control mechanism, which is under the power
of governmental authorities.
Since the priorities obtained, as a result of AHP application, for Network
6 and 1 were not significantly different, authorities came up with a new plan
altogether to construct a combination of Alternative Rail Networks 6 and 1,
221
a new Rail Transit Network 61 as a final decision (Figure 6). The Decision-
making team did not see any necessity to evaluate this new combination since
Network 6 and 1 performed very close to each other and much better than
Network 5. Authors caution the readers that this combination may not neces-
sarily be better than networks 6 and 1.
Contribution and benefits of AHP
The AHP has had appeal to decision-makers at all levels of decision-making.
It enables one to include both the strength of feelings needed to express
judgment and, the logic and understanding relating to the issues involved in
the decision. It elicits and combines the multiplicity of judgments in a sys-
tematic way to obtain the best outcome or mix of actions to be taken. Finally,
and more significantly, these outcomes are derived in an agreeable way that
is in harmony with our intuition and understanding and not forced on us by
technical manipulations. To sum up, the Analytic Hierarchy Process contributes
to solving complex problems by structuring a hierarchy of criteria, stakeholders,
and outcomes, by eliciting judgments to develop priorities, and synthesizing
these judgments throughout the structure to give the best choice.
222
1218: Peak hour travel volume in 1,000 passengers per hour per direction. Demand range covers
both scenarios.
Figure 6. Rail Transit Network 61.
Most researchers would use scoring methodology to solve multiple criteria
problems. However recent research has shown that, contrary to our intuition,
directly estimating importance rating for each criterion does not end up rep-
resenting the decision-makers real preference for one criterion over another.
In addition scoring methodology urges us to quantify all the criteria, while
the AHP can deal with intangibles, that is with concepts for which no measures
have yet been devised. The AHP overcomes most of the disadvantages of
scoring methods.
In general, because decision-makers fail to rank correctly the relative
accuracy of elicitation methods, there seems to be a need to direct decision-
makers toward better choice of techniques. In this study, the expert team
presented the relative accuracy and perceived ease of use of each of the
comparison methods provided by Expert Choice to the decision-making team
and let them choose which they preferred to use. Among the non-dominated
elicitation methods, the decision-makers did not select graphical multiple
bars because they understood that it would have low accuracy and they elim-
inated verbal pairwise comparisons because of the difficulty of translating their
judgments to words. The decision-making team thus selected the graphical
pairwise comparison mode as their preferred way to transmit their judgments.
The sensitivity analysis method of the AHP, by displaying different
outcomes depending on the priority given to high-level criteria, proved its
importance in the decision process. Sensitivity analysis is very important for
these kinds of studies, not only because there are enormous uncertainties
with regard to the future planning parameters, but also because the cost esti-
mates of the alternative rail networks are subject to change. Sensitivity analysis
was used to show how the preferred outcome would change as the circum-
stances changed, revealing under what assumptions the priorities of the
alternatives would be different and when they would be the same. Both were
very helpful to our decision-makers. They were much more comfortable with
their final decision.
Future research
In any decision all interested groups should have an opportunity to get involved
in the decision-making process. The technical role of the expert team is to
do an analysis that is valid, practical, relevant and constructive in clarifying
the issues that should be considered by all parties who will have an interest
in or some influence over the decision. In this decision, the main interested
groups were public transport users and operators, and nongovernmental orga-
nizations that represent the views of individuals and groups who will potentially
be affected, positively or negatively, by any of the actions being considered.
Among the stakeholders we have incorporated the views of the Mayor, the City
223
Council Members, and bureaucrats. Incorporating the rest of the stakeholders
such as representatives of the individuals and groups who will potentially be
affected by any of the decisions is an area for future research.
Another area for future research is incorporating dependence and feedback
among the criteria and the alternative decision actions. The Analytic Network
Process (ANP) feedback approach (Saaty 2001), is a generalization of the AHP
which replaces hierarchies with networks (www.creativedecisions.net/anp/
2001).
Metropolitan planning agencies face a daunting task when planning major
infrastructure projects. Multiple criteria must be considered in order to make
decisions that best satisfy concerns of cost, desirable economic and geographic
growth, and quality of life. This study has shown that AHP can be used as
an important tool for decision-makers in such multifaceted planning processes.
Using AHP and its sensitivity analysis helped decision-makers in Istanbul select
the best alternative for a new rail transit network. This rail system, Rail Transit
Network 61, is currently under construction in the Istanbul Metropolitan Area.
Note
1. Acquisition of ownership or property for public use.
Appendix 1. AHP methodology
The details of the methodology can be given as follows: We prefer explaining the method on
a hypothetical, simplified version of the rail transit network selection problem rather than on a
completely theoretical example. In order to set up an AHP model of the problem, first, the overall
goal should be identified, which is choosing the Best Rail Transit System. Then determine the
criteria that are important to achieve this goal. They are: Financial, Economic, System Planning
and Policy criteria. Assume we have three alternative rail transit networks (rail transit networks
1, 2 and 3) to be prioritized. Each alternative can be prioritized according to each criterion.
Figure A1-1 shows the AHP model of the problem.
The importance of each criterion with respect to others was given as follows:
Table A1-1. Comparing the criteria for importance with respect to the Best Rail Transit System.
Financial Economic System Policy Priorities
planning
Financial 1 3 2 3 0.461
Economic 1/3 1 2 3 0.262
System planing 1/2 1/2 1 2 0.153
Policy 1/3 1 1/2 1 0.124
Inconsistency ratio (IR) 7%.
224
The judgments in Table A1-1 are entered using the fundamental scale of the AHP; a criterion
compared with itself is assigned the value 1 so the main diagonal entries are always 1. The
numbers 3, 5, 7, and 9 correspond to the verbal judgments given in the paper, Table 1. Criteria
on the left are always compared with criteria at the top. The Financial Criterion is moderately
(3) more important than the Economic Criterion. If the Financial Criterion is moderately (3)
more important than the Economic Criterion while choosing the best alternative, then the
Economic Criterion is 1/3 of the importance of the Financial Criterion. Therefore 1/3 is entered
in the Economic Criterion Financial Criterion intersection of the matrix.
After forming an importance matrix, it is normalized by summing each column and by dividing
each entry by the sum. Entries of this new, normalized matrix are in comparable scales and
the average values of each row give us an approximation of the priority ranking of each
criterion. (Saaty (2001) explains this as raising the matrix to a large power to capture all the
interactions.) Since Expert Choice

is easily available, it is suggested that practitioners use the


software for calculations.
Expert Choice

calculates an inconsistency ratio (IR). Any inconsistency ratio above 10%


suggests reevaluation of pairwise comparisons. The software helps the decision-maker find the
most (second most, third most, etc.) inconsistent evaluation and offers the best fit value. However,
one should not change the evaluations unless one finds some mistake in them.
Comparative evaluations of each alternative with respect to others under each criterion are
given in Table A1-2 using the same nine-point scale given in Table 1 of the paper. The prior-
ities are derived in a similar way as it is explained while deriving priority ranking of each
criterion.
In Table A1-3, the preference ratings of each alternative have been multiplied by the impor-
tance of the criteria and summed to obtain the final ranking. For example the overall preference
rating of RTN1 is 0.479 (0.461 0.540 + 0.262 0.493 + 0.153 0.188 + 0.124 0.582 =
0.479). Constructing RTN1 is far more preferable than RTN2 or RTN3.
225
Figure A1-1. The best rail transit Network hierarchy.
Table A1-2. Pairwise comparisons and preference ratings of each alternative with respect to
different criteria.
A1-2a. Preference rating of each alternative A1-2b. Preference rating of each alternative
with Respect to Financial criterion. with respect to Economic criterion.
RTN1 RTN2 RTN3 Priorities RTN1 RTN2 RTN3 Priorities
RTN1 1 2 3 0.540 RTN1 1 2 2 0.493
RTN2 1/2 1 2 0.297 RTN2 1/2 1 1/2 0.196
RTN3 1/3 1/2 1 0.163 RTN3 1/2 2 1 0.311
IR = 0.01. IR = 0.01.
A1-2c. Preference rating of each alternative A1-2d. Preference rating of each alternative
with respect to System Planning criterion. with respect to Policy criterion.
RTN1 RTN2 RTN3 Priorities RTN1 RTN2 RTN3 Priorities
RTN1 1 3 1/5 0.188 RTN1 1 2 5 0.582
RTN2 1/3 1 1/7 0.081 RTN2 1/2 1 3 0.309
RTN3 5 7 1 0.731 RTN3 1/5 1/3 1 0.109
IR = 0.16. IR = 0.
Table A1-3. Synthesis of the priorities.
Criteria Financial Economic System Policy Priority
planning synthesis
Rail transit (0.461) (0.262 (0.153) (0.124)
network alternatives
RTN1 0.540 0.493 0.188 0.582 0.479
RTN2 0.297 0.196 0.081 0.309 0.239
RTN3 0.163 0.311 0.731 0.109 0.282
Inconsistency Ratio (IR) = 0.05.
226
References
Forman E & Saaty TL (1995) ECPro for Windows: Decision Support Software. Pittsburgh,
PA: Expert Choice, Inc.
Gakenheimer R (1999) Urban mobility in the developing world. Transportation Research, Part
A, 33: 671689.
Gerek H (2002) Mobility, Car Use and Traffic Congestion: A Comparative Analysis with Special
Reference to Istanbul. Paper presented to the erc/METU International Conference in
Economics VI, Ankara, Turkey.
Harper PT & Vargas LG (1990) Reply to remarks on the analytic hierarchy process.
Management Science 36: 269273.
Istanbul Metropolitan Area Sub-Region Master Plan (1995). The Greater Municipality of Istanbul.
ITU (1999) Istanbul Master Plan Transport Study. Technical University of Istanbul.
May AD (1991) Integrated transport strategies: a new approach to urban transport policy
formulation in the UK. Transport Reviews 11(3).
227
Appendix 2. Data for the three-rail transit networks for both scenarios
TREND MASTER PLAN
RTN1 RTN5 RTN6 RTN1 RTN5 RTN6
Operating and
maintenance costs
(1000 passenger-km
per hour) 0,005,043 0,005,362 0,005,336 0,004,669 0,004,897 0,004,944
Network construction
cost (million $ US) 0,00,0924 0,00,0937 0,001,286 0,00,0924 0,00,0937 0,001,286
Rail vehicle purchasing
costs (million $ US) 0,00,0673 0,00,0698 0,00,0602 0,00,0592 0,00,0698 0,00,0705
Road vehicle
operating costs
(1,000 pcu km/hour)* 0,007,943 0,007,951 0,008,035 0,007,623 0,007,613 0,007,654
Road vehicle
purchasing costs
(1,000 pcu km/hour) 0,007,943 0,007,951 0,008,035 0,007,623 0,007,613 0,007,654
Road accident costs
(1000 pcu km/hour) 0,007,943 0,007,951 0,008,035 0,007,623 0,007,613 0,007,654
Total trip time
(passenger hour
Peak time) 2,413,312 2,421,759 2,474,681 2,290,879 2,317,769 2,300,793
Environmental costs
(1,000 pcu km/hour) 0,007,943 0,007,951 0,008,035 0,007,623 0,007,613 0,007,654
* pcu: Passenger car unit.
Millet I (1997) The effectiveness of alternative preference elicitation methods in the analytic
hierarchy process. Journal of Multi-Criteria Decision Analysis 6: 4151.
Ocakci M & Southworth M (1995) Manufacturing and Changing Land Use Patterns of the Istanbul
Historical Peninsula. Working Paper 649, IURD, U.C. Berkeley.
Pucher J (1999) Transportation trends, problems, and policies: an international perspective.
Transportation Research, Part A 33: 493503.
Saaty TL (2001) Decision Making with Dependence and Feedback: The Analytic Network Process,
second edition. Pittsburgh, PA: RWS Publications, ,
Saaty TL (1995) Transport planning with multiple criteria: the analytic hierarchy process
applications and progress review. Journal of Advanced Transportation 29(1): 81126.
Saaty TL (1994) How to make a decision: the analytic hierarchy process. Interfaces 24: 1943.
Saaty TL (1990a) How to make a decision: the analytic hierarchy process. European Journal
of Operational Research 48: 926.
Saaty TL (1990b) The Analytic Hierarchy Process. New York: McGraw-Hill, 1980. Pittsburgh,
PA: RWS Publications.
Tekfen & Delcan (1996) Strategic Network Planning for the Rail Transit Systems in the European
Side of Istanbul. Final Report.
Vargas LG (1990) An overview of the analytic hierarchy process and its applications. European
Journal of Operational Research 48: 28.
www.creativedecisions.net/anp/ (2/5/2001).
www.expertchoice.com
About the authors
Haluk Gerek is a Professor of Civil Engineering at Istanbul Technical University.
His recent works are mainly in the field of urban transportation planning, mobility
and sustainable transportation.
Birsen Karpak is Professor of Management at Youngstown State University. Her
recent publications are mainly in applications of multiple criteria decision making
methodologies which have appeared in referred journals such as Journal of Multicriteria
Decision Analysis, the Journal of Applied Business Research, and European Journal
of Purchasing and Supply Management.
Tlay Klnaslan is a Professor of Transportation at the Department of Urban and
Regional Planning, Istanbul Technical University. Her recent studies are in travel
demand and transportation in Istanbul metropolitan area.
228

Anda mungkin juga menyukai