Professional Documents
Culture Documents
373-400, 1994
Copyright © 1994ElsevierScienceLtd
Pergamon Printed in Great Britain. All rightsreserved
0965-8564/94$6.00 + .00
0965-8564(94)E0001-P
KHALED A. ABBAS
The Egyptian National Institute of Transport, P.O. Box 34 Abbassia, Cairo, Egypt
and
MICHAEL G. H. BELL
Transport Operations and Research Group, University of Newcastle upon Tyne, Newcastle upon
Tyne, NE1 7RU, United Kingdom
(Received 12 October 1991; in revised form 21 February 1993)
Abstract -The main focus of this article is to review and evaluate the strengths and weaknesses of
system dynamics (SD) with respect to its suitability and appropriateness in transportation model-
ing. This evaluation helps in appreciating how the SD modelingstyle can contribute to understand-
ing better the relationships between elements of the transport system and between transport and
its environment. It establishes the ease with which system dynamics can be applied to construct
useful tools for testing alternative transport-related policies.
INTRODUCTION
Transportation systems are multi-dimensional in that they are multi-modal, multi-
sectoral, multi-faceted, multi-problematic, multi-purpose, multi-operational, multi-or-
ganisational, multi-effect, multi-ownership, multi-network, multi-technological, and
multi-disciplinary. In complex, large-scale systems, like transport, problems are rooted in
the basic structure of the system. Actions taken to deal with one problem may create
difficulties elsewhere.
Investment in transport affects, and is affected by, investments in other sectors.
Simplistic, step-by-step approaches do not cater to the dynamic interactions that exist
between the elements. What is needed is a system-based approach.
A system is a number of components integrated into a complex entity, and system
analysis simply means the consideration of the entity rather than the separate consider-
ation of individual components. The systems approach can be defined as an organised,
efficient procedure for representing, analysing and planning complex systems. It is a
comprehensive, problem-solving methodology that involves two main steps:
1. The rational and creative structuring of both quantitative and qualitative knowledge,
mainly in the form of models, to represent problems; and
2. The development of analytical techniques through which the problem can be analysed
and solved.
373
374 K . A . ABBAS and M. G. H. BELL
SYSTEM DYNAMICS
SD, originally called industrial dynamics, was developed at the Massachusetts Insti-
tute of Technology (see Forrester, 1961). It is a powerful methodology that derives its
roots from system theory, cybernetics, information science, organisational theory, feed-
back control theory, military games and tactical decision-making. The main function
of SD is to construct models of complex problems and to experiment with them on
computers.
SD pioneered the use of system concepts and computer simulation for the analysis of
complex problems in business and management. It is a methodology of wide applicability
that has become an appealing modeling style used by many different disciplines. Applica-
tions of SD cover a wide spectrum of different fields, disciplines and subjects. These
include applications in defence, urban and regional development, business, banking,
industry, economics, finance, manufacturing, biology, education, health, medicine, den-
tistry, engineering, forestry, fishery, energy, environment, transportation, psychology
and various others.
From its name it can be inferred that SD is a methodology designed to help in
understanding the dynamics of different real-world systems. SD is based on control
theory, sometimes referred to as the concept of servomechanisms. This presents a proce-
dure for investigating and understanding a system that is in the form of causal feedback
relations.
There are two types of feedback relations-namely, negative and positive feedback
loops. The negative feedback control procedure attempts to adjust the actual state of a
system to achieve a desired state. These changes are brought about by decisions to imple-
ment control actions, which are attempting to narrow and close discrepancies (whether
positive or negative) that exist between desired and actual conditions of the system. As
stated by Forrester (1975), "the basic structure of a feedback system is a loop, within
which the system condition provides the input to a decision process that generates action,
which modifies the system c o n d i t i o n . . , this does not imply that the system must reach
the equilibrium that is sought. The interacting efforts toward goals may keep the system
continuously out off balance" (see Fig. 1). On the other hand, the positive feedback loop
Goal
, Difference
(Gap) [
Action
Current 0
State
Delays
Actual
Performance
= Physical or information F l o w
q I
0 F e e d b a c k Loop
8oume: (Adapled from Formatmr, 1876)
AUXILIARY
l DELAY ~l'
LEVEL ~ RATE
CAUSAL REPRESENTATION
t
| PHYSICAL/IN FOF~4ATION~
~tCf
OTHER IDENTIFICATIONOF A PROBLEM
INFLUENCES ¢
DECISION
MAI,,~NG I DECISION'SUPPORT/ SYSTEM VERBALISATION
PROCESS SYSTEM UNDERSTANDI]NG [MENTAL MODELLING]
T
SYSTEM ANALYSIS SYSTEM CONCEPTULISATION
POLICY/SCENARiO/SENSITIVITY [DIAGRAMMATICMODELLING]
TESTS 1-CAUSAL DIAGRAM
t
MODEL SIMULATION
2-FLOW DIAGRAM
MODEL FORMULATION
l [MATHEMATICALiODELLING]
MODEL PROGRAMMING~
tCOM~UTE~ MODEmNCi
Identification o f a problem
Whenever practical difficulties a n d / o r limitations are encountered in dealing with
any real-world systems, politicians, scientists a n d / o r researchers could identify this as a
problem area. Consequentially, the potential problem requires careful thought and in-
spection prior to research efforts.
Definition of a problem
A model builder must establish a clear definition o f a problem in general terms. He
or she must identify the specific questions that the model is attempting to answer, and
consequentially the objectives and purposes o f the model. Another important aspect is to
identify the users who are expected to apply the model. These are crucial steps that can
often dictate the general characteristics, the context and the dimension demarcation of
the model to be developed. They involve establishing the purpose, the scope, the bound-
ary (the breadth) and the level of detail (depth) o f the model. In this context, a system
dynamicist tends to develop what is known as a reference mode. A reference mode
attempts to define the problem at hand in terms o f graphical or verbal representations.
Three types of X-Y graphical representations can be drawn. The first type displays the
behaviour of the major variables that are thought to describe the problem over time. In
doing so, available time series data can be used. The second type of graphical representa-
tions presents the expected futuristic behaviour of variables. The third type of graphical
representations depicts metaphysical variables such as happiness, anger, sadness, and
frustration. The dynamic behaviour of variables such as these can be specified by appeal-
ing to a modeler's "mental model" without the need for measured time series data.
According to Randers (1980), "the identification and description of a reference mode
greatly facilitates the selection of the basic causal structure of the m o d e l . . , it simplifies
decisions on what to leave o u t . . , keeps the attention of the modeler squarely fixed on
what m a t t e r s . . , and ensures that the initial model embraces the essential dynamics."
Knowledge acquisition. There are two levels o f knowledge elicitation. The first is the
individual level, which includes the following methods:
378 K. A. AaaAS and M. G. H. BELL
1. Clinical interviews with individuals to get an insight into their perceptions, beliefl
thoughts, views, values and impressions. Interviews can be structured or open ende
(i.e., unstructured interviews).
2. Nominal and judgmental questionnaires. These include choice, ranking and ratin
questions used to determine individual choices, preferences and attitudes. Examples o
these are the revealed preference, the stated preference and the attitudinal surve
questionnaires.
The second level of extracting knowledge and information is the group level. This involve
eliciting knowledge and soliciting information from experts. This can take the form o
expert opinions, mental concepts, consensus, practices and experiences. The followin
are some of these techniques:
These techniques for extracting and capturing knowledge are well established il
psychology and social sciences literature, as well as in behavioural and system structur
studies. A relatively recent development in this field is the introduction and developmen
o f computer software like S T E L L A and computer-based expert systems that are designe~
to help in the process of acquiring knowledge and information about a particular problen
or system. For a more detailed discussion and review of knowledge elicitation technique
that can be used in constructing SD models, see Richardson et al. (1989) and Fey ant
Trimble (1992).
The SD modeling procedure acts as a filter through which the aforementionet
sources of information could be synthesised. Synthesisation entails the explicit expres
sion, organisation, structural mapping and transfer of available information to a set o
quantitative relationships.
System verbalisation
Verbalisation mainly involves the clear explanation and description in writing of th,
mental concepts that form the foundation for developing a model of a particular system
Verbalisation entails stating explicitly all the relevant sources of information on which th,
logic of the structural relations of the model is based. The variables and parameters tha
describe the system are identified and selected. This includes choosing variable names
stating the directions of influence between variables, stating the types of influence
(whether positive or negative) and explaining the rationale and hypothesis involved il
formulating the model. The main advantage of the verbalisation step is that it acts as
communication medium. This allows people to scrutinise and understand the model ant
to gain better insight into and appreciation of the variables that are used in structurinl
the model relationships.
System conceptualisation
This involves developing a sound structural representation of a system in the form o
causal and flow diagrams. These diagrams are intended to permit a more precise qualita
tive description and appreciation of the structural components of the system. It is appro
priate at this point to assert the phrase, "One diagram could be worth a thousand words.
Diagrams illuminate the mental concepts and the verbal explanations of a system.
Causal diagrams. Causal diagrams, sometimes referred to as system maps or influ
ence diagrams, are simply the representation of the verbal description of a model in
graphical form. These graphical portrayals use the concept of signed digraphs, showinl
the full names of variables linked together by arrows that demonstrate the directions o
influences (dependencies) between the variables. The nature of the causal link could bq
System dynamics in transportation modeling 379
described in terms o f being either positive or negative. "Positive causality between A and
B could be defined as: A has a positive influence on B if A adds to B, or if a change in A
results in a change in B in the same direction. Negative causality between A and B could
be defined as: A has a negative effect on B if A subtracts from B, or if a change in A
results in a change in B in the opposite direction" (Richardson, 1986). Individual linkages
in a causal loop diagram are made ceterisparibus (i.e. with everything else held constant)
and they can be thought of as partial derivatives. Delays (namely, the existence of time
lags between a cause and an effect) should be identified and marked.
The causal interrelations form feedback loops, thus exhibiting the feedback mecha-
nisms that characterise the structure of a system. A feedback loop is identified when the
linkage between variables starts at one point and goes on in a cyclic path ending at the
same point and passing through at least one level variable. The polarity of a loop is
determined by multiplying the signs around the loop. If the outcome is negative the loop
has a negative net effect, whereas if the outcome is positive the loop has a positive net
effect. However, the preferred way of determining whether a loop is positive or negative
is to trace through the effects of an external shock to a variable (an increase or decrease)
as it propagates around the loop. If the shock is reinforced, the loop is positive. If it is
counteracted, the loop is negative. Conceptual errors can be caught and corrected in this
manner (i.e. people often place a minus sign where a plus sign should go, or vice versa).
However, it is fair to note that many SD modelers use causal loop diagrams just to
document a n d / o r present portions of an existing model to an audience. They use generic
structures, rather than causal loop diagrams, to assist them in the modeling process. A
generic structure could be defined as a causal building block that can be used repetitively
in mapping multiple models describing similar phenomena. For a more detailed discus-
sion of generic structures, see Paich 0985).
Flow diagrams. Using a set of specially devised SD diagram conventions, the causal
diagrams can be transferred from the signed digraph form into flow diagrams. These SD
conventions distinctly identify rates, levels, auxiliaries, constants, delays and exogenous
variables (see Fig. 2). The type of connections between variables, whether an information
or a physical flow, are distinguished. As a result, the system variables and the feedback
loops are classified in a manner that allows easy and quick computer programming and
simulation. To demonstrate the steps of system conceptualisation, two examples are
shown in Figs. 4 and 5. The examples are mainly concerned with the construction and
maintenance of roads.
It is obvious that SD modeling relies heavily on the diagrammatic representation of
systems. Diagrams are considered to be an essential step of the SD modeling procedure.
't
ROad Delay Road 1 I
Kllornetrel: a6 Conltruotton
Level Rate
ROad /.b
CAUSAL DIAGRAM ConltruGtlon
Fundi
TPRK
IRCRI i
FLOW DIAGRAM
Fig. 4. Main stages in developing a qualitative system dynamics model: an example of road construction.
380 K . A . ABBAS and M. G. H. BELL
't
M i n g l e d Road DIlay
l" 1
Road n
KIIometrel: ab Milntonlnce
Level RIm
RoM ,~"
CAUSAL DIAGRAM Malntlnlnol
Fundl
KRRM
-.0 nl
Fig. 5. Main stages in developinga qualitative system dynamics model: an example of road maintenance.
If diagrams are well designed and structured, formulating and developing SD models
mathematically become relatively quick and easy. In fact, there are commercial programs
that do both of these operations at once (e.g. the S T E L L A and I T H I N K software
packages for the Macintosh). These allow the modeler to connect level and rate icons
with a mouse, and the software automatically generates the corresponding equation in a
different window.
In summary, the main functions o f diagrams are to
Model formulation
Building on the aforementioned steps, the mathematical formulations that represent
the structural relations o f an SD model are written. These provide the means to examine
quantitatively the structure of a system. SD is used to model, understand and enhance the
management of physical and social systems. Although in social systems many relation-
ships are highly qualitative in nature and difficult to represent quantitatively, most of the
relationships describing physical systems can be quantitatively modeled. Sensitivity tests
could play an important role in the formulation of SD models. The mathematical formu-
lations o f an SD model define the modeled system in terms of a set of linear and nonlinear
equations.
System dynamics in transportation modeling 381
Computer programming
Having carefully formulated the structure of an SD model, D Y N A M O or any other
computer language can be used to program the algebraic relations of the model and code
them into an executable computer program. The software is then debugged and the logic
of the p r o g r a m m verified. I f D Y N A M O or any similar computer language is chosen, the
SD flow diagrams could be converted directly into a set of algebraic equations. This
substantially reduces the time and effort involved in the model formulation stage.
Simulation
Simulation is intended to show the behaviour of the system in quantitative terms.
The output of the simulation reflects the consequences of the structure used for formulat-
ing the model. C o m p u t e r simulation runs show the most likely impacts of alternative
policies on the state and performance of a system under study. Simulation serves this
purpose in a relatively cheap, short, accurate and safe manner. SD simulation entails the
numerical integration of the system equations over time.
System analysis
If the output of a simulation run shows inadequacies (namely, something odd or
wrong) the modeler asks why, looks for the answer by examining and assessing the
structure of the model, modifies and develops what he or she thinks is necessary, tests the
model by a new simulation run and so on. This cyclic procedure of model inspection,
interpretation, revision and simulation is one of the great strengths of the SD methodol-
ogy. If the modeler follows this procedure, the model improves greatly, and the modeler
gains a thorough understanding of the system being modeled. Additionally, the output of
models supports decision-makers in improving their decisions regarding the choice of
which policy to implement.
The mathematical relationships of an SD model require quantification. The values
of parameters and initial values of levels should be specified and supplied to the model.
Initial values of the variables of the model are set to represent the conditions at the
initial time of the simulation. From that point onward, the model computes variables
endogenously based on the specified mathematical relationships.
The model calculates the value of each of the variables at each point of the simulated
time. The computer keeps track of the ways in which these variables relate to one another
over time. These calculations repeat themselves until the specified period of simulation
has expired.
It is the time-dependent aspect of the SD methodology which is the most useful for
planning purposes. With it, the impact of specific changes can be traced over time and
evaluated. Without it, the dynamic aspect of change is lost. Therefore, an SD model
simulates performance patterns not as a result of extrapolation of trends, but through the
continuous application of rules and relationships that modify simulated conditions and
on which subsequent understanding and decision analysis are based. The system structure
is usually more important than the assumed numerical values in determining the modes
of behaviour that a system may exhibit.
Most o f the SD model variables are determined endogenously from the interplay
with other variables. The remaining variables, particularly the ones associated with state
long-term growth, are exogenously specified by the model user. These are permitted to
vary modestly as the remainder of the model is simulated in time. The cause-and-effect
forces, which produce the behaviour of these exogenous variables, are outside the scope
of the model. Some of these external conditions have an important influence on the
model behaviour. These can be easily varied, singly or in various combinations, from one
simulation to another to analyse the impacts of alternative scenarios and policy options.
Stated in general terms, SD is a useful tool for
2. Designing, formulating and testing different scenarios and policies by posing and
answering "what if" questions;
3. Providing useful information, both to policy- and decision-makers, thus giving support
to the decision-making process in the field of strategic planning; and
4. Improving the management and control of complex systems. SD is a methodology
through which powerful management tools are developed to enhance the abilities to
control complex systems.
Transportation models are developed to serve two main purposes. The first is to help
in reaching a better understanding of and gaining more insight into transport systems.
The second is to employ models, mainly in prediction a n d / o r policy analysis. However,
to date, and for a variety of reasons, forecasting models have had high margins of error.
The unreliability of forecasts can be attributed to the following reasons:
The level of uncertainty in planning activities increases due to the length of the planning
horizon and the difficultly of reversing decisions once implementation starts. Therefore,
it is of little use when forecasting for the long term to offer more time and money for the
acquisition of input data.
In the following subsections, the suitability and appropriateness of SD to transporta-
tion modeling are explored. However, it is fair to state at this point that SD models are to
be used for gaining understanding and for policy analysis, rather than for prediction. An
insightful demonstration of the impossibility of point prediction in social systems is
contained in Forrester (1961).
hence requires the integration of both deterministic and stochastic modeling efforts.
5. Policy-oriented planning techniques that can generate and test a variety of alternatives,
without the need to collect and analyse huge volumes of data, should be fully pursued.
Models which require data that are easy to collect, verify and use efficiently should be
developed. Simple, transparent, creative and user-oriented techniques for modeling
transportation systems should be explored and used.
6. Comprehensive, analytical models are required to simulate the nature and dynamics of
the decision-making process. If the resources available are to be employed to maximum
advantage, it is necessary to weigh the merits of various combinations of input. Re-
search is needed to look at the decision-making process and the allocation of resources
and to develop transport management tools intended to ease the dilemma faced by
transportation decision-makers when attempting to reach rational, informed decisions.
7. Evaluation of transportation policies should be based both on long- and short-term
impacts.
in fact, an SD transport model can act as a means for assessing the appropriate data
needed for future enhancement of the model.
5. The methodological advantages o f SD are such that the results o f an SD transport
model are mainly reached through the dynamic, causal feedback interactions of the
structural components of the model, as well as through empirically based formula-
tions. In contrast, in conventional statistical regression models results are reached
purely through empirically based formulations. The structural elements of an SD
transport model interact to produce a dynamic behaviour of the transport system.
Nonlinearities and time delays can be a c c o m m o d a t e d explicitly.
6. SD methodology can be used as a basis for constructing and testing a priori hypotheti-
cal models of different transportation problems. It can incorporate alternative policy
or structural assumptions, input data or empirical parameters. SD strives to incorpo-
rate relationships of subjective (behavioural) p h e n o m e n a rather than ignore and ex-
clude these relationships.
7. SD provides a rich and coherent conceptual working environment, which brings order
and makes the m a x i m u m use of the modeler's mental and modeling agility. This can
be achieved by formulating transport models, as much as possible, in clear, transpar-
ent and diagrammatic forms as well as in an algebraic form. Understanding gained
through building an SD transport model and manipulating that model can be as
useful as that gained f r o m long association with the transport system in the real
world. SD conceptualisation procedures can also provide a c o m m o n platform for
communication and understanding between the various parties that have interest in a
transport system.
8. SD can assist in developing experimental transport tools. These tools are meant to
provide flexibility in design and analysis and to enable tests to be made over a wide
variety of transport policy options and scenarios: The models can be used as training
tools to help familiarise transport personnel with the different transportation prob-
lems.
9. Through SD simulation, the short-term and long-term behaviour of a transport sys-
tem is traced. This can provide insight into the dynamic nature of the transport
problem and allow for timely adjustments to be made, if required.
10. Low cost, transparency, transferability and ease of updating are characteristics that
support the application of SD to model transportation problems.
11. Building transport models using the SD approach can help in identifying controls in
the transport system being modeled. These controls are meant to guide the efforts of
policy-makers regarding which parameters a n d / o r structures of a transport system
could be managed and controlled to produce better performance of the system.
12. According to Willumsen (1990), "a transport model is not simply a set of mathemati-
cal instructions used to simulate and forecast travel demand, it implicitly structures
the way in which transport problems are viewed and understood and the type of
solutions that are considered feasible and testable." SD methodology can provide a
foundation for structuring thoughts and building a better understanding of complex
transportation systems and their underlying problems.
1. SD, as a simulation technique, works mainly through the time dimension. Spatial
aspects and distribution effects are not easily accounted for.
2. Most SD models are aggregate models intended to show policy impacts in terms of
approximate magnitudes and directions of change. However, SD models could be
System dynamics in transportation modeling 385
further refined to describe the system in a more detailed form, thus providing numeri-
cally accurate output.
3. Generally, SD models are deterministic in nature, yet randomness and stochasticity can
be easily incorporated, thus accounting for future uncertainties and variabilities.
4. The main aim of an SD model is to aid policy-makers in reaching an optimum design
policy. In complex systems, heuristic optimisation (i.e. manual trial and error with the
system) can become a tedious and difficult task. A number o f computerised optimisa-
tion algorithms have been specially developed for the SD type of models, and these
should be more usable whenever available and required.
5. The validity of SD models has been a subject of debates throughout the years. In
general, it is difficult to establish the validity of structurally based simulation models
such as the SD type o f models. In SD, validity is interpreted as model usefulness rather
than numerical exactness. The SD literature includes a wide range o f both qualitative
and quantitative validity tests. These tests are meant to increase confidence in SD
models and to withstand any criticisms. Some of these tests are available in software
form. Further efforts to automate and facilitate the use of these tests should be pur-
sued.
CONCLUSION
way to represent the dynamic performance of such systems is for their behavioural dyna-
mism to be induced as a result of the feedback causal structures that describe such
systems. SD caters explicitly to the dynamic behaviour of systems. In general, dynamic
modeling has been widely applied in other disciplines. In transport it is steadily gaining
momentum in the midst of the conventional static planning approaches. This can be seen
in the growth of interest in dynamic assignment models.
According to Coyle (1978a), "the SD modeling approach seems to fulfill a need
which is not met by the standard planning and programming approaches, namely that of
providing for the concept of controllability." SD is a policy-oriented modeling technique
that provides a framework for the design of policies and the management of systems to
achieve improved system behaviour. The process of actually constructing an SD model
induces a substantial and extensive amount of intuition, understanding and learning
about a modeled system as well as about the available policy options and their anticipated
impacts on the behaviour of the system. SD is a versatile and flexible modeling approach.
In general terms, the main characteristics of modeling using the SD methodology are
as follows:
In concluding this article, it would seem appropriate to state the following quotations
from several different studies. The first is from an OECD (1974) report in which a
thorough review of possibilities for simplifying urban transport models was carried out.
In this report, the transport models were mainly divided into three categories: (a) four-
step models; (b) demand models; and (c) urban dynamic development models. In review-
ing the urban dynamic development models, the report stated that "a third group consists
of models structured by methods of systems analysis. This is the most promising group,
though they are at the moment only global models. The basis of these models is the
Forrester model, mentioned in Urban Dynamics." Along similar lines of thought, the
ECMT (1975) report, in discussing the impact of infrastructural investment on industrial
development within an integrated land-use and transport context, stated that "the main
progress in this field may come from formulation in terms of systems on the lines pro-
posed by J. W. Forrester in works that are now famous (Industrial Dynamics, World
Dynamics, etc.)."
The third quotation is from Hansen and Kahne (1975). At the end of their appraisal
of modeling large-scale systems using dynamic systems modeling techniques invented by
Forrester, they stated that "the benefit derived from the modeling efforts of large scale
systems suggests that the techniques may be beneficially applied to other large-scale
systems. Transportation provides many problems of large scale where new and more
effective approaches are needed."
The fourth quotation is from Hazel (1989), who recommended that SD be used for
transportation studies and who stated that "indeed, SD offers a potential way forward
for transportation planning in general, which seems to be becalmed between disbelief in
the traditional four-stage model and a lack of a credible alternative."
The utility of the SD methodology regarding its appropriateness and suitability for
application to modeling transportation systems has been critically reviewed in this article.
Although the SD modeling concept has its limitations, the motivations for pursuing it for
solving transport problems are overwhelming. Current research activities in the develop-
ment of modeling methodologies and planning approaches for handling transport-related
issues are extensively involved with the refinement, improvement and adaptation of al-
ready existing techniques. Although this is to be encouraged, research efforts should also
be directed toward developing and creating novel approaches. Exploring, researching
System dynamics in transportation modeling 387
and integrating other innovative approaches from other scientific disciplines should be
supported and encouraged. These research activities could make far-reaching contribu-
tions to the solution of rapidly changing transportation problems.
An important function of the various modeling approaches is to use them to develop
models (tools) that can act as credible supports to the policy- and decision-making pro-
cesses. Models that are carefully and diligently programmed act as intelligent amplifiers
that stimulate creativity and filter the mental models of users. These models are not in
any sense meant to replace decision-makers or even to inhibit their role in making deci-
sions; rather they are developed to help and support them in achieving better decisions.
The SD methodology can offer a lot in terms of better planning and solving transport-
related problems. SD should not be thought of as a methodology to replace or substitute
for the traditional transport modeling approaches. Rather it should complement and be
integrated with the existing approaches, to contribute, in a collective manner, to solving
transport problems. In particular, SD should be applied in strategic studies that are
concerned with policy analysis and decision-making in the field of transport.
REFERENCES
Abbas K. A. (1989). The use of System Dynamics methodology in modelling the transport development of new
cities in Egypt. Proceedings of the 21st Annual Conference of the Universities Transport Studies Group
(UTSG), Edinburgh, U.K.
Abbas K. A. (1990a). A road provision model using System Dynamics. In D. F. Andersen, G. P. Richardson
and J. D. Sterman (Eds.), System Dynamics '90, Proceedings of the 1990 International System Dynamics
Conference, Boston, MA. vol. 1, pp. 1-15. The System Dynamics Society, Cambridge, MA.
Abbas K. A. (1990b). A System Dynamics road provision model. In V. D'Amato and C. Maccheroni (Eds.),
Dynamic Analysis of Complex Systems, Proceedings of the 1990 European Conference of System Dynamics,
Milan, Italy, pp. 93-100. Franco Angeli, Milano, Italy.
Abbas K. A. (1991). The development of a road management system with particular reference to new cities in
Egypt: An application of System Dynamics methodology. Ph.D. thesis, Transport Engineering and Opera-
tions Division, Civil Engineering Department, University of Newcastle upon Tyne, U.K.
Abbas K. A., Bell M. G. H. and Crouch F. O. (1990). Computer-based support for the management of
investments in road infrastructure. Proceedings of the 18th Planning and Transport Research and Computa-
tion (PTRC) Summer Annual Meeting, Sussex, U.K., Seminar J/K, Highway Appraisal Design and Manage-
ment/Highway Construction and Maintenance, pp. 41-45 and 85-98.
Adler T. J. and Ison J. W. (1981). Analysis of long-term transportation energy use. Transpn. Res. Rec., 801,
33-44.
Adler T. J., Ison J. W. and Geinzer J. C. (1979). ENTRANS: A model of the interactions between energy
supply and transportation energy use. Proceedings of the 1979 Summer Computer Simulation Conference,
Toronto, Ontario, Canada.
Adler T. J., Ison J. W. and Geinzer J. C. (1980a). Interactions between energy supply and transportation-related
energy use. Report of U.S. DOT/RSPA/DPB-50/80/7. Thayer School of Engineering, Dartmouth College,
Hanover, NH.
Adler T. J., Ison J. W. and Geinzer J. C. (1980b). Interactions between energy supply and transportation-related
energy use. Report of U.S. DOT/RSPA/DPB-50/80/8. Thayer School of Engineering, Dartmouth College,
Hanover, NH.
Adler T. J., Stearns S. R. and Stephanedes Y. J. (1980c). Techniques for analyzing the performance of rural
transit systems. Report of U.S. DOT/RSPA/DPB-50/80/23. Thayer School of Engineering, Dartmouth
College, Hanover, NH.
AI-Dawood A. S. (1993). Corridor impact on the Roanoke area. Proceedings of the XI1 International Road
Federation (IRF) World Congress, Madrid, Spain. Working Session 8: Cost Benefit Analysis of Roads,
Impact on Economic Development, vol. IV, pp. 27-36.
Ali S., Bencosme A. and Dajani J. (1980). A dynamic simulation approach to transportation planning. Proceed-
ings of the IEEE International Conference on Cybernetics and Society, pp. 437-445.
Bencosme A. J. (1979). Infrastructure planning for regional development: A dynamic modeling approach.
Ph.D. thesis, Stanford University, Stanford, CA.
Bencosme A. J. and Dajani J. S. (1980). Dynamic appraisal of infrastructure investments. Proceedings of the
1980 IEEE International Conference on Cybernetics and Society, pp. 850-859.
Boyce D. E. and Goldstone S. E. (1966). A regional economic simulation model for urban transportation
planning. Highway Res. Rec., 149, 29-41.
Bradley M. A. (1985). A dynamic approach to investigate household car ownership and usage. Proceedings of
the 1985 International Conference of the System Dynamics Society, Keystone, CO, vol. 1, pp. 48-62.
Budhu G. (1981). A System Dynamics approach to rural transportation planning in less developed countries.
Ph.D. thesis, Virginia Polytechnic Institute and State University, Blacksburg, VA.
Budhu G. (1984). Coordination of urban/rural transportation investments in less developed countries . . . using
simulation and System Dynamics. Proceedings of the lOth Intern,:tional Road Federation World Meeting,
Rio De Janeiro, Brazil, pp. 485-495.
Budhu G. (1986). Comprehensive rural investment planning at the national level in an agri-based less developed
c o u n t r y - a system concept. Proceedings of the Sixth International Road Federation African Highway
Conference, Cairo, Egypt, pp. 35-46.
TR(A) 28:5-B
388 K. A. ABBASand M. G. H. BELL
Budhu G. and Grissom D. (1985). Development of a simulation model to study the impacts of rapid urban growth
on the transportation sector - the case of Charlotte, North Carolina. Transpn. Res. Rec., 1046, 45-49.
Budhu G. and Tran T. K. (1982). An integrated approach to studying urbanised regions. Proceedings of the
Fourth Annual Urban Affairs Conference, Chapel Hill, NC.
Charlesworth J. A. (1985). The dynamics of transportation system. Proceedings of the 1985 International
Conference of the System Dynamics Society, Keystone, CO, vol. 1, pp. 140-149.
Charlesworth J. A. (1987). A dynamic simulation model for the estimation of delays arising from medium-term
disturbances to a traffic network. Proceedings of the Third Mini Euro Conference on Operations Research
Methods in Transport Planning and Traffic Control, Herceg Novi, Yugoslavia, pp. 1-21.
Charlesworth J. A. and Gunawan E. (1987). The behavioural dynamics of route choice. Traffic Engineer.
Control, 28(2), 80-83.
Chen W. (1975). Transportation dynamics: A methodology for transportation systems analysis. Ph.D. thesis,
Asian Institute of Technology, Bangkok, Thailand.
Coyle R. G. (1978a). Management System Dynamics. Wiley-lnterscience, New York.
Coyle, R. G. (1978b). Tanker chartering: A System Dynamics case study. Eur. J. Oper. Res., 2, 86-96.
Drew D. R. et al. (1975). Bicol river basin development program simulation model. USAID Philippines Project
No. 492-5 l- 190-126. Manila, Philippines.
Drew D. R. (1978). The growth shapers: Infrastructure induced development. In N. Sharif and P. Adulbhan
(Eds.), System Models for Decision Making, pp. 119-190. Asian Institute of Technology, Bangkok, Thai-
land.
Drew D. R. (1989). System dynamics modeling of development induced by transportation investment. Transpn.
Res. Rec., 1274, 63-83.
Drew D. R. (1991). Transportation impact methodology for measuring user and non-user benefits. Proceedings
of the 1991 System Dynamic Conference, Bangkok, Thailand, pp. 163-172. The System Dynamics Society,
Cambridge, MA.
Duffy M. A., Eiden G. L. and Hamilton C. W. (1977). General aviation dynamics. Proceedings of the Eighth
Annual Pittsburgh Conference on Modeling and Simulation, pp. 267-279.
European Conference of Ministers of Transport (ECMT) (1975). Impact of infrastructural investment on
industrial development. Report of the Twenty-Fifth Round Table on Transport Economics, Paris, France.
Fey W. and Trimble J. (1992). The evaluation and development of knowledge acquisition in System Dynamics
studies. Proceedings of the 1992 System Dynamics Conference, Utrecht, The Netherlands, pp. 173-182. The
System Dynamics Society, Cambridge, MA.
Forrester J. W. (1961). Industrial Dynamics. MIT Press, Cambridge, MA.
Forrester J. W. (1975). Collected Papers of Jay IV. Forrester. MIT Press, Cambridge, MA.
Geinzer J. C. (1979). Travel in an energy-constrained future. M.Sc. thesis, Thayer School of Engineering,
Dartmouth College, Hanover, NH.
Gottschalk P. (1982). A System Dynamics model for long range planning in a railroad. Eur. J. Oper. Res., 14,
156-162.
Gunawan E. (1984). Management of traffic in an incident road. M.B.A. dissertation, University of Bradford,
Bradford, U.K.
Hansen R. C. and Kahne S. (1975). A transportation ecology model: Conceptual developments. IEEE Trans.
Systems, Man, Cybernet., 5(2), 157-166.
Hazel G. M. (1989). The estimation and effect of new transferred and pass-by private car trips to retail centres.
Proceedings of the 17th Planning and Transport Research and Computation (PTRC) Summer Annual
Meeting, Sussex, U.K., Seminar J, pp. 71-88.
Hirsh M. S. (1977). Toward a methodology for evaluating urban bus transit. Ph.D. thesis, Polytechnic Institute
of New York.
Hobeika A. J., Budhu G., and Tran T. K. (1981). System Dynamics approach to transportation planning in
developing regions. Transpn. Res. Rec., 820, 11-17.
Hobeika A. J., Budhu G., and Tran T. K. (1982). Indirect and spatial impacts of transportation investment in a
less-developed country. Transpn. Res. Rec., 887, 9-14.
Ison J. W. (1980). The effect of government policy on future automobile fuel efficiency. Report 203, Resource
Policy Centre, Thayer School of Engineering, Dartmouth College, Hanover, NH.
Khanna 1. K. (1986). Transportation planning and policy design for Delhi urban region: A System Dynamics
approach. Ph.D. thesis, Indian Institute of Technology, Delhi, India.
Khanna 1. K., Singh N. and Vrat P. (1985). System Dynamics in urban transportation planning and policy
analysis. Proceedings of the 1985 International Conference of the System Dynamics Society, Keystone, CO,
vol. 1, pp. 453-462.
Khanna I. K., Singh N. and Vrat P. (1986a). Some studies in policy design and analysis of Delhi's urban
passenger transportation system. Transpn. Sys. Anal. Policy Stud., 602-608.
Khanna 1. K., Singh N. and Vrat P. (1986b). A transportation policy planning and analysis model for air
pollution control. Indian J. Air Poll. Control, 7(3), 121-127.
Khanna I. K., Vrat P. and Singh N. (1989). A System Dynamics transportation policy planning model: Under-
standing syslem behaviour and policy alternatives. System Dynamics: Int. J. Pol. Model., 2(1), 13-21.
Krallman H. and Nestaas L. (1979). Optimale investitionsstrategien in tankermarket. OR Spektrum, 1, 35-49.
Kuroda J, and Tsaur T. H. M. (1990). Urban growth modeling under the limitation of transportation facilities-
case of Bangkok. In D. F. Andersen, G. P. Richardson and J. D. Sterman (Eds.), System Dynamics "90,
Proceedings t2/'the 1990 International System Dynamics Conference, Boston, MA, vol. 2, pp. 620-634. The
System Dynamics Society, Cambridge, MA.
Linstone H. A. and Turoff M. (1975). The Delphi Method: Techniques and Applications. Addison-Wesley,
Reading, N1A.
Organisation for Economic Co-operation and Development (OECD) (1974) Urban traffic models: Possibilities
for simplification. A report prepared by an OECD Road Research group.
Ortuzar J. D. and Willumsen k. G. (1990). Modeling Transport. John Wiley & Sons, New York.
System dynamics in transportation modeling 389
the 1989 International Conference of the System Dynamics Society. Stuttgart, Germany, pp. 460-485,
Springer-Verlag, Berlin, Germany.
Young S. H., Ma T. and Chao C. M. (1985). A simulation model for managing the parking systems of
Kaohsiung: A System Dynamics approach. Proceedings of the 1985 International Conference of the System
Dynamics Society, Keystone, CO, vol. 1, pp. 1081-1089.
Young S. H. and Santoso I. B. (1988). A System Dynamics approach to the car ownership trend in Taiwan
urban areas. Proceedings of the 1988 International Conference of the System Dynamics Society, La Jolla,
CA, pp. 522-534.
Zahn E. O. K. (1988). System Dynamics-instrument fur das strategische management. In Der Oberbau-Basis
der neuen Bahn, Publikation zur Fachtagung der Deutschen Bundeshahn und des, pp. 15-17.