EPA-600/5-74-001
February 1974
Socioeconomic Environmental Studies Series
Simulation City Approach
for Preparation of
Urban Area Data Bases
UJ
O
Office of Research and Development
U.S. Environmental Protection Agency
Washington, D.C. 20460
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, Environmental
Protection Agency, have been grouped into five series. These five broad
categories were established to facilitate further development and appli-
cation of environmental technology. Elimination of traditional grouping
was consciously planned to foster technology transfer and a maximum inter-
face in related fields. The five series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. TJocioeconomic Environmental Studies
This report has been assigned to the SOCIOECONOMIC ENVIRONMENTAL STUDIES
series. This series includes research on environmental management, compre-
hensive planning and forecasting and analysis methodologies. Included are
tools for determining varying impacts of alternative policies, analyses of
environmental planning techniques at the regional, state and local levels,
and approaches to measuring environmental quality perceptions. Such topics
as urban form, industrial mix, growth policies, control and organizational
structure are discussed in terms of optimal environmental performance.
These interdisciplinary studies and systems analyses are presented in forms
varying from quantitative relational analyses to management and policy-
oriented reports.
EPA REVIEW NOTICE
This report has been reviewed by the Office of Research and Development,
EPA, and approved for publication. Approval does not signify that the
contents necessarily reflect the views and policies of the Environmental
Protection Agency, nor does mention of trade names or commercial products
constitute endorsement or recommendation for use.
-------
EPA-600/5-74-001
February 1974
SIMULATION CITY APPROACH FOR
PREPARATION OF URBAN AREA DATA BASES
by
Andrew C. Lemer
Contract No. 68-01-1805
Program Element 1HA096
Project Officer
Philip Patterson
Environmental Studies Division
WASHINGTON ENVIRONMENTAL RESEARCH CENTER
WASHINGTON, D. C. 20460
Prepared for
OFFICE OF RESEARCH AND DEVELOPMENT
U. S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D. C. 20460
For ale by UM SaperlnUndnt of DoeamaaU, U.S. Ooreniment Printtnf Offlee, WMhinfton, D.C. 30KB
-------
ABSTRACT
This report presents a preliminary exploration of feasibility of a Simula-
tion City Approach to the preparation of urban area data bases for
environmental policy analyses. The Simulation City Approach is pre-
sented as a means of making complex comprehsnsive environmental
planning models more accessible for a broad range of policy level ques-
tions by effectively reducing the difficulty and cost of applying these
models.
The basic hypotheses of the Simulation City Approach is that it is possible
to approximate the detailed input data base required for complex planning
models, given only a relatively gross description of a specific metropolitan
area and general knowledge of patterns of urban composition. This
approximation is accomplished at a substantial reduction in the costs
associated with data preparation and thus planning models, at relatively
little expense in terms of accuracy. Accuracy is here judged in terms of
the final decision to be made on the basis of planning model analyses. To
the extent that the hypothesis is valid in a particular application, a decision-
maker is given the opportunity to ask a range of questions at substantially
reduced cost and time expense.
The concept is first described in general terms, and then supported by a
review of theoretical and empirical studies which would be valuable in its
realization. The trade-off between cost and accuracy in modeling is then
more explicitly considered.
-------
The investigation of the Simulation City Concept was undertaken within a
context of application to the General Environmental Model (GEM), a
comprehensive environmental planning tool being developed by EPA. It
is concluded that the approach is a promising one for policy-level analysis,
and offers a means of bringing the cost and difficulty of application of
large planning models into line with the problems they are best suited to
address.
-------
CONTENTS
Page
Abstract ii
List of Figures v
Acknowledgements vi
Sections
I. Conclusions 1
II. Recommendations 4
III. The Simulation City Concept 5
What is the Concept of Simulation City 6
Components of a Simulation .8
Undertaking a Simulation 10
IV. Bases for Simulating a City 12
Types of Urban Areas 12
Infrastructure of the Urban Economy 16
Simulation of the Socioeconomic System 17
Historical Patterns and Location Theory 18
Transport Orientation 21
Distribution of Socioeconomic Factors 22
Simulation of the Spatial Pattern 23
Overview: Bases for Simulation 24
V. The Trade-Off of Accuracy and Cost 26
A General View of the Trade-Off 26
Losses in Accuracy Through the Simulation City Approach • • 31
Accuracy and Cost in the GEM Context 33
VI. References 35
VII. Appendices
Appendix A: Simulation City Model for GEM A-l
Appendix B: A Simulation City Model for GEM:
Computation Oriented Description • • B-l
Appendix C: The GEM Load Deck C-l
IV
-------
FIGURES
Page
Major Components in a Simulation City
2. Manufacturing Employment as Related to Population and
City Type 14
3. Retail Employment as Related to Population and City Type . 14
4. Parameters of the Exponential Distribution of Activity
Density in Cities 20
5. Hypothetical Relationship of Cost and Accuracy in Planning
Models 28
6. Influence of Uncertain Knowledge of Cost-Accuracy
Tradeoff Relationship 28
A-l. Macro-Scale Design of a GEM-Linked Simulation City
Model A-3
A-2. Proposed Input to Simulation City Model A-6
A-3. Variation of Activity with Distance From City Center . . . A-8
A-4. Historical Patterns of Urban Form Based Upon
Transportation Growth A-8
A-5. The First Steps in Simulation A-ll
A-6. Assignment of Locations A-15
A-7. Interactive Stage of a GEM-Linked Simulation City Model . A-20
B-l. Macro Flowchart for Simulation City B-2
-------
ACKNOWLEDGEMENTS
The work reported here was accomplished by a team of professionals in
the Environmental and Urban Planning Department at AMV in McLean,
Virginia, with Dr. S. J. Bellomo as the Principal Investigator, and Dr.
A. C. Lemer as Project Manager. Ms. C. Schlappi and Ms. J. McMeckan
provided considerable input to the work.
Dr. Philip Paterson and Mr. Albert Pines of the EPA Office of Research
and Monitoring administered this project. Their assistance is
acknowledged with sincere thanks.
Thi s report presents the results of a study into an innovative approach to
the preparation of input data for large scale planning models. This approach
is not without precedent; drawing upon prior work by the consultants,
utilizing a range of published research in urban systems structures — ,
21
and taking a measure of encouragement from work in other fields — ,
the Simulation City concept has a broad range of potential applications.
However, this study has been conducted with the intention of a direct applica-
tion to the General Environmental Model, under development by the
Research and Monitoring Division of EPA. No attempt has been made to
explain fully the details of GEM, and some knowledge of the model's charac-
teristics is assumed. The body of the report minimizes the level of this
assumed knowledge, however, and the bulk of the argument related
specifically to GEM is presented in appendices.
1. See the Voorhees reports in the bibliography.
2. See, for example, Rand Corporation work.
vi
-------
SECTION I
CONCLUSIONS
The purpose of this report has been to describe the Simulation City
approach to data preparation, to present arguments regarding the possible
feasibility of implementing the concept, and to describe in a preliminary
fashion (see appendices) the design of a Simulation City model intended to
provide input data for GEM. As has been explained, the Simulation City
concept is potentially applicable to a range of large-scale planning
models, although a principal motivation for the current study has been to
explore the concept for the particular case of application to GEM. Thus,
it is appropriate to state separately certain conclusions which are directed
to the GEM system. Further, recommendations will be made as to what
might be done to proceed with the work, to implement a GEM-linked
Simulation City model.
SOME GENERAL CONCLUSIONS
The Simulation City concept is proposed as a means of developing a data base
for a large scale planning model, at a significantly reduced cost and relatively
small loss in accuracy. Chapter II reviewed the background information
which supports the technical feasibility of the concept, and Chapter HI
addressed more directly the considerations of cost and accuracy. Based
upon these discussions, the following conclusions are suggested:
• 1. The Simulation City concept is a feasible and potentially
valuable approach to data preparation for large scale planning
models.
-------
2. A Simulation City model, linked with an appropriate planning
model, will be most useful in the analysis of planning questions
at the policy level. This conclusion is based primarily on con-
siderations of the overall loss of predictive accuracy anticipated
to result from dependence on data simulation.
3. A Simulation City model will not only be most useful at the
policy level, but will actually render the planning model
more accessible to the decision-making process. Frequently
there are questions asked at the policy level for which there
is too little time or too uncertain an understanding of the
magnitude of the problem to justify expenditure of the
resources required to complete a standard data collection and
analysis effort. The Simulation City model would make it
practical to ask a fuller range of questions.
CONG LUSIONS FOR GEM
Beyond the general conclusions stated above, there are statements which
may be made about the Simulation City concept as applied to GEM:
1. Because of the extremely comprehensive nature of GEM,
development of a Simulation City model for this application
will represent a test of the extent to which the concept may
be applied.
2. However, the unique features of GEM input requirements
makes the evaluation of a Simulation City model more difficult
in any terms other than with respect to final planning decisions.
The question of accuracy will require careful consideration.
3. The potential savings in data expenses and the consequently
increased availability of the GEM system for policy explora-
tion makes the implementation of a GEM-linked Simulation
City model an apparently worthwhile undertaking.
SOME IMPLICATIONS
The conclusions stated above may be used as a basis for considering the value
of the simulation concept and thereby for identifying implications for its use:
-------
1. Because the model would be based on a consistent set of
rules, application of a Simulation City Approach across
a variety of metropolitan areas would result in a consis-
tent set of data bases. These data bases would be
consistent in the sense that they would have similar
levels of detail, source of error, and statement of
characteristics. Such data bases would be of value in
undertaking consistent studies of EPA policy across a
variety of urban areas.
2. Linked with appropriate environmental analysis model,
a Simulation City Model could be used to develop sketch
planning tools for local area planners. Such tools would
be used in preliminary planning analysis, to identify
major types of plan alternatives which would be feasible
and beneficial on environmental grounds. A simple
sketch planning tool would make the concepts and prac-
tice of environmental planning more accessible to small
communities and the general public.
-------
SECTION n
RECOMMENDATIONS
Based upon the investigation and conclusions of this study, it is recommended
that an effort be made to implement a Simulation City model for GEM. It is
further recommended that the implementation be undertaken in several steps:
• A detailed design and coding of the simulation should first
be completed. These efforts should be coordinated closely
with ongoing developments in the GEM system.
• The Simulation City model should then be given preliminary
testing/ through generation of several data bases. These
data bases would be compared with bases created "by hand."
• A test of the entire GEM system should be made, using the
Simulation City model, to investigate the impact of the model
on GEM's role in decision-making.
This last step is especially important, for it is upon its potential value to
policy explorations that the justification for the Simulation City concept
rests.
A Simulation City model should be most accurate in those areas where its
related planning model is most sensitive. The simulation may be quite
acceptable with relatively large errors in other areas. No conclusive
judgment can then be made by looking at raw data alone.
In summary, this exploration of the Simulation City approach should con-
tinue. But it should continue in a context of its ultimate use, with concern
for its role in a planning and decision-making process.
-------
SECTION III
THE SIMULATION CITY CONCEPT
In addressing the myriad problems of planning and decision-making
for the environment, the development and application of computer based
models has emerged as a major area of activity. While there may be
considerable question regarding the validity of individual efforts (See,
i 21
for example, the growing literature surrounding Forrester's work it— ),
there can be no question that modeling ferver is widespread and represents
potentially a very promising approach-to problem solving.
Modeling for environmental analysis has extensive antecedent--in philosophy,
content, and common personnel—in urban and transportation planning.
Models for land use and travel prediction and analysis have been the subject
3/
of substantial study and debate (Lee's discussion — is instructive) for
many years, and are now providing the basis for many environmental
modeling efforts.
One outstanding feature which runs throughout all of these efforts is the
level of resources typically devoted to the collection of data and preparation
of this data for use in analysis. It has been the experience of this consultant,
for example, that a minimum of 20 percent of the cost of applying land use
models will be directly associated with data collection, while in the case of
standard travel analyses, 45 percent to 50 percent of the cost is likely
to be incurred before any real analytical efforts can be made. The nature
of these data requirements severely limits the range of situations to
which models may be applied. Only when a major study is undertaken,
or when there is clearly an ongoing need for analytical capability will the
level of resources available be adequate to support an application of models,
requiring large input data bases.
-------
In contrast to the situation in which application of models may be financially
feasible is the typical situation in which models might be most useful: public
decision-makers often have questions about the possible results of one or
another course of action, questions which arise in the ongoing formulation
and application of policy. An immediate solution is desired, and it therefore
is assumed that accuracy or certainty in prediction may be sacrificed in order
to achieve a quick decision. Models, as currently designed, provide little
assistance in this incremental decision making context.
Even in the slower, more detailed analyses in which models are most often
employed, the planner is seeking a level of accuracy and detail well beyond
that which the model, as an abstraction of a partially understood physical
and socio-economic system, can provide. Clearly, there is a need to shift
3 4/
the focus of such analyses.—*—
There is a need to perceive more clearly the capabilities and limitations
of models, and to balance the allocation of effort in utilizing models.
It is in response to the need for balance that the Simulation City Approach
is proposed. By simplifying and substantially reducing the effort required
to prepare a data base for environmental analysis models, these models
will be rendered considerably more accessible and useful to the planning
and decision-making process.
WHAT IS THE CONCEPT OF SIMULATION CITY?
The underlying hypothesis of the Simulation City Approach to environmental
data preparation is that the level of accuracy at which environmental models
operate does not justify the detail required in their input and, further,
that quite acceptable and useful results may be obtained by using less accurate
input. That is, it should be possible to utilize more approximate input data,
-------
with an associated reduction in acquisition and preparation costs, with
relatively little overall deterioration in the value of the modeling exercise.
The approach is based upon a concept that there is a sufficient body of
knowledge about the structure and operating character of the system of
activities comprising a metropolitan area to permit the simulation of a
considerable portion of the detailed description of an area. Based upon
a knowledge of the principal industry and median income in an area,
for example, one may infer a great deal about the socio-economic
character of the area. To do so, one needs some knowledge of the current
character of the country as a whole, and of the body of research available
in urban economics. Similarly, a knowledge of total population will allow
one to make a reasonable assumption about the amount of land area
occupied by urban development.
Simulation of a city, in terms of a detailed descriptions of that city's
characteristics, cannot replicate the actual, "real world" situation. How-
ever, a successful simulation will possess a sufficient similtude to the
real world that the environmental model which utilizes the data will induce
the same decisions as would be made with "perfect" knowledge of the
environment. To make an analogy, the dimensions of a building plan are
not specified in fractions of an inch. Decisions regarding the allocation
of space may be made on a coarser basis.
A Simulation City Model would thus be a means of approximating the fine
description of a metropolitan area, as required for input to one or more
planning models. To be useful, such a model must be able to produce input
having sufficient resemblance to the real city that the final analytical
results indicate the same conclusions as would have been reached with
standard methods. It is then apparent that any realization of a Simulation
-------
City Model will be influenced significantly by the analytical model or
family of models to which it will be providing input. Accuracy in simula-
tion is more important for those items of data to which the analytical
planning model is most sensitive, and the form of simulation output must
be accepted by the planning model.
These requirements complicate the development of a Simulation City
model: It is necessary to accommodate the input requirements of the
planning model while producing an acceptable representation of reality.
On the other hand, the planning model may limit the scope of the simula-
tion via its limited abilities to accept a variety of input.
COMPONENTS OF A SIMULATION
Figure 1 indicates, in a very general manner, how a Simulation City
model might be constituted. As shown, there would be two types of infor-
mation required for simulation: aggregate parameters which are
representative of the overall socio-economic system of the city, and
descriptors of detail, which give a sketchy picture of the city's spatial
distribution of activity. The simulation is intended to infer a good deal
more detail about the socio-economic system, given some knowledge of
what cities in general have in common, and to postulate how the socio-
economic activity is distributed over the skeleton of pattern provided by
the sketch descriptions. A picture of the level and distribution of
activities which comprise the basic structure of the city is thus developed.
Observations of how cities in general operate and grow permit some
aspects of the city's character to be inferred from this description of
basic structure. For example, population density and median income
5 /
are found to be negatively correlated in residential areas.— Provision
of adequate road access must generally precede development of an area,
8
-------
FIGURE 1 MAJOR COMPONENTS IN A SIMULATION CITY
GROSS AGGREGATED
PARAMETERS
Such as:
-Total population
-City type
-Urbanized area
-Median income
SKETCH DESCRIPTORS
OF DETAIL
Such as locations of:
-Interstate highways
-Harbors
-Major park lands
-Rapid transit routes
o
ec.
\-
co
o
i
to
SOCIO-ECONOMIC SYSTEM
Theories of urbanization,
agglomeration, industrial
development, etc., are
used to predict:
-Employment distribution
-Housing characteristics
-Income distribution
-Industry mix
SPATIAL PATTERN
Theories of location
and spatial interaction
are used to distribute:
-Industry
-Residences
-Commercial space
cc.
=>
o
Ul
BL
O
I
STRUCTURAL SERVICES
Historical observations
suggest that activities
will have more or less
adequate:
-Transportati on-access
-Uti lity service
-Sewer service .
1
RENTS, INCOMES, SALARIES
Historical trends indicate
the association between
patterns of activity and
measures of the economic
value or prestige
NON-STRUCTURAL SERVICES
There are apparent
associations between
such factors as incomes
and structural services,
and such services as:
-Schools
-Government services
-Health facilities
I
COMPLETED SIMULATION
9
-------
so that a prediction that an area has been developed implies an assump-
tion that such access exists.
Finally, schools and other municipal services will be provided, but their
quality is quite variable and dependent upon the individual community.
Simulation thus requires some judgment or measure of the community's
character.
UNDERTAKING A SIMULATION
Asa general rule it may be suggested that the ability to simulate accurately
the characteristics shown in Figure 1 decreases as one moves down the
path from minimal input to final output. Predictive ability is increased by
the requirement of more extensive input, up to the point that an entire data
base is simply developed by hand. But the further one proceeds with the
simulation, and the more one minimizes the required input, the greater
is the saving in time, labor, and monetary expense in preparing the data
base. It is this trade off between accuracy and cost which represents the
central point in judging whether the Simulation City Approach is a viable
one.
In the following pages, the tradeoff will be explored. Chapter II presents a
summary of some of the theoretical and observational considerations which
suggest that, as a general concept, the Simulation City Approach is valid.
Chapter III addresses directly the question of accuracy, and the tradeoff
•
between accuracy and cost, in general and relative to the GEM system.
Chapter IV summarizes the considerations presented in preceding
chapters and suggest conclusions regarding the value of proceeding with
10
-------
development of the Simulation City Approach to data preparation.
Appendices present the exploration of a GEM-linked Simulation City
Model which provided background for much of the argument in the text.
11
-------
SECTION IV
BASES FOR SIMULATING A CITY
In considering the body of theory and observation which is available as a
basis for development of the Simulation City approach, it is important to
recognize that the Simulation City model is not intended to be predictive,
in the sense of most planning models. There is no consideration given to
future conditions, and hence no debate regarding whether the model is in
any sense normative. A Simulation City model is intended to be purely
descriptive -- descriptive of the present conditions in a given city. It is
then not unrealistic to expect that wherever theory is inadequate to
explain the patterns of urban development, statistics may fill the void.
With these points in mind, a search may be made to find what theories
and facts are available, and whether what'is available is adequate to
make development of the Simulation City approach a practical undertaking.
In this chapter, the findings of such a search are reviewed.
TYPES OF URBAN AREAS
Attempts to classify urban areas — cities -- into typical categories has
been a relatively popular area of research among geographers and
6/
planners. Work by Harris— was among the earliest to attract attention,
rjl
while that of Berry— reflects the recent tread toward applications of
advanced statistical methods. While it has been felt by many observers &
that there has been inadequate thought given in the conduct of such research
to why one should classify cities and what is thereby accomplished, such
work represents a resource for the simulation city approach: it is
12
-------
apparent that the statistics which identify groups and distinguish them
from one another are better predictors of the characteristics of an indi-
vidual city with the group than are the statistics for all groups combined.
Recent work by Pidot and Sommer, — for example (Golub, et. al:—
had similar findings), tested variables of population, education, social
character, housing quality, labor base, and industrial character for
standard metropolitan statistical areas in the U. S. (excluding such
unique cases as New York), and found what they considered to be an
optimal grouping into nine categories. One group was characterized,
for instance, as heavily oriented to recreational activity, while another
had high components of elderly population.
These categories suggest a way to predict more accurately certain
characteristics' of a metropolitan area. As Figure 2 illustrates, for
example, there is a quite noticeable difference in the fraction of
work force in manufacturing employment among different classes of
cities. In contrast, Figure 3 indicates that the effect of city type is much
less pronounced in the case of retail employment.
Classifications of city type, such as those mentioned here, are typically
most dependent upon population and industry characteristics for these
comprise the basis of the city's existence. It is hardly surprising then
that there should be empirical relations such as those shown in Figures 2 and 3.
But there is reason to suppose that there are other categories which could be
applied at a lower level of detail to give other types of useful relationships.
For example, there has been substantial literature produced in study of
the theory and process of agglomeration among businesses. According
to the theory, first postulated by Weber,— industries tend to cluster,
spatiaUy, to take advantage of economics of scale in minimizing trans-
portation costs for intermediate goods. Explorations of the theory have
led to the concept of industrial clusters, groups of industries which are
13
-------
(Hand fitted trend lines)
300 -
1
200 -
I
100 -
500
Population (thousands)
1000
Source: Statistical
Abstract of the
United States,
City Types,
P?dot and Sommer
FIGURE 2 MANUFACTURING EMPLOYMENT AS RELATED TO POPULATION AND CITY TYPE
200'
4*
£
City Type A
City Type C
City Type C
500
Population (Thousands)
1000
Source: Statistical
Abstract of the
United States,
City Types,
Pi dot and Sommer-?
FIGURE 3 RETAIL EMPLOYMENT AS RELATED TO POPULATION AND CITY TYPE
14
-------
found in many urban areas. Bergsman, et. al.,— found four widespread
clusters in their preliminary analysis, which they termed market center,
low wage apparel, labor intensive, and southern textile. Such work
suggests that indication of the major industry or cluster type in an area
might prove sufficient to simulate the industrial .composition of the area,
in terms of economic importance (such as sales, value added
13/
labor proportion, or other measures). Morrisett's— findings on
minimum employment levels as a function of city size supports this
concept.
On the other hand, a number of unique economic systems were noted
in this work. New York, Los Angeles, and Chicago are not surprising
in this category, but cities such as Philadelphia and Akron are perhaps
less expected as singularities. Looking at incomes in relation to
14 /
industry mix and employment, Bryce— suggested that roughly
30 percent of metropolitan areas have demographic profiles which
could best be termed atypical. Such findings indicate that it may not
be reasonable to carry the stratification of cities into type categories
on too detailed a level.
Somewhat related to the concept of distinct types, based on industry
and demography, is that of the economic base. Drawing from inter-
national trade theory, regional economists postulate (see Blumenfeld's
15/
discussion, for instance — ) that industries which export the bulk of
their production from the region, thus drawing cash into the region,
comprise the base of the region's economy. The employees associated
with these businesses have a need for services -- retail outlets,
building trades, etc. — which stimulates the flow of goods and services
in an urban economy.
15
-------
There are generally 15 percent to 30 percent of all workers employed in
these service industries*, with the percentage rising as city size grows.
This latter observation stems from the larger city's importance as a
center of a larger region of activity. It would seem then that one might
predict the number of workers in a city of given population, and the likely
split of these workers between basic and service businesses, without
reference to more detailed categories.
INFRASTRUCTURE OF THE URBAN ECONOMY
While attempts at stratification of urban areas focus upon differences
among types, there has been considerable theoretical and historical
support for the idea that there are certain consistencies of structure
underlying the economies of all cities. Location theory suggests that the
initial formulation of a settlement in an area (see Isard's introductory
16 /
discussion— ), is highly dependent upon its physical site. The site
therefore may reveal a good bit about the city. For example, location
on a river suggests a transport orientation, and shipping in or out of
heavy raw materials. Location at a source of natural resources indicates
information about the probable mix of labor.
More interesting, from the standpoint of a simulation city model, are
observations about supplies of services in urban areas. Another aspect
of the agglomeration process mentioned above is that urbanization supports
economies of scale, relative to provision of such infrastructure services
17/
as sewer, water, electricity, and schools. Borcherding, — for example,
was able to develop regression equations to predict with fair accuracy
the dollar expenditures on such services. Other investigations (see
Stone— and Sussna, — for example) have reported sufficient consistency
among urban areas to lend support to the idea that there may be an optimum
city size, relative to minimization of per capita service costs.
* Some definitions of service employment include as much as 50 percent
of the total employment
16
-------
Similar consistencies are found in the provision of transportation ser-
20 /
vices. Highway lane-miles— or percentage of land in highway use are
approximated by a function of population and urbanized land area.
Although public transit service is not so well defined, which may be due
21/
largely to the lack of consistent data, Wells— was able to develop
predictive regression equations for route-miles of supply.
SIMULATION OF THE SOCIO-ECONOMIC SYSTEM
The above discussion has been concerned with general characteristics of
urban demography and economy. Referring back to Figure 1, such
information may provide the basis for simulating the basic structure of
the socio-economic system of an urban area. At the extreme, simulation
might proceed on the basis of only an estimate of total population and a
city type designator or basic/service ratio.
From these two bits of data, the size of the workforce might be estimated,
and a split of this workforce into industry categories made. The split
might be simply basic versus service, or might deal with major activities
such as manufacturing, retail, and wholesale. Some inferences about
the distribution of income might be drawn.
Given a value for the median income, an income distribution is derived
and utilized to infer a distribution of land values and rental costs for the
area. Total supply levels of services and transportation facilities are
then estimated.
In this fashion, a general, aggregated picture of the city is developed.
Simulation of these aspects of the city might be considered analogous to
the reconstruction of a set of accounts for a company in which the records
have been lost. This part of the simulation has not directly addressed the
17
-------
spatial distribution of activities within a region, which will be discussed
in the following sections.
HISTORICAL PATTERNS AND LOCATION THEORY
The analysis of why certain activities are located in certain places, and
of how individual locations combine into the pattern of a metropolitan
region, has focused upon a trade-off between attractive features of a piece
of land and the problems of gaining physical access to that land. In these
terms, the location of a city at the site of a good harbor, rail terminus,
or concentration of mineral wealth is analogous to the location of higher
income residences within the metropolitan areas on sites with high ground,
good vegetation, and highway service.
Emphasis on attractions of particular areas within a region is reflected
22 /
in work such as that of Hoyt, — who was among the earlier analysts to
devise a coherent explanation of the patterns of activity within an area.
His sector theory, which addresses residential location and urban growth,
postulated that similar activities would tend to grow outward from the
center city in wedge-shaped land sections. The initial location of a sector
was determined by factors attractive to the activity -- for example, higher
ground for residential development: Hoyt pointed out that higher income
residences tended to cluster along ridges.
More recent work, such as Timms, 23. has extended Hoyt's work into
social analysis. Here, social rank as a whole, rather than simply
income, is said to be distributed along sectoral lines, while status
within a social class generally increases toward the center — until a
reversal occurs as one enters the central core area.
The extent to which the sector concept represents a contribution to urban
theory rather than a restatement of historical development, is uncertain.
18
-------
The radial patterns of highways and the street railways in the past several
decades certainly have encouraged such segregation of activities, and have
24 /
given characteristic forms to many cities.— The fact that the sector
concept was useful in guiding decision making in real estate investment
was adequate to justify its widespread acceptance.
An alternative focus is provided by the attempts to explore variations in
the level of activity within an urban area. Density of development is the
measure generally used, stated as people or jobs per unit area. Viewing
the urban region as a whole, reasonably good success has been achieved
with statistical fitting of an exponential function to patterns of density.
25/
Mills— presents one of the more recent and extensive efforts in this
area, using a model of the form
D(x) = DQe"yu
where D(x) is the density of activity at a distance x from the center of
the city, D is a central density measure, and Y is a gradient measure.
As Figure 4 shows, there seems to be some consistency of the estimates
of the parameters of this model within the city type categories discussed
previously. Some geographers have attempted to use this type of approach
26/
to define the limits of urbanized areas (see Grytzell, — for example), to
replace the artificial legal boundary as a limit of analysis. However, the
model seems most useful in replicating activity in the most stable sections
of the urban region rather than in the often decaying central core or newly
growing fringes. On a regional level, the fitted function has been shown to
provide a reasonable measure of the degree to which a city is centralized
or dispersed.
On a micro-scale attempts have been made to explain density on the basis
of particular factors of a site or subarea in a region. Chapin,—' for
example, listed five factors which influenced density of development:
19
-------
6=
0.7
0.6
0.5
y
0.3
0.2
0.1
o
o.o
10
20
30
FIGURE 4 PARAMETERS OF THE EXPONENTIAL DISTRIBUTION OF ACTIVITY
DENSITY IN CITIES
Source: Millsli/,
City Types from
PI dot and Sommer
20
-------
1. Poor drainage discourages development
2. Major transportation encourages development
3. Large employment potential intensifies development
4. Availability of services intensifies development
5. Proximity of blight and non-white areas discourages
development.
Knowledge about the pattern of location of any of these factors could be
useful in predicting the overall urban pattern.
TRANSPORT ORIENTATION
Something of a contrast to the above empirical studies is provided by the
analyses undertaken with a strong transport point of view. Typically
there has been an appealing model or theoretical approach to be applied,
28 /
to which data has then been fitted. Hansen's— work is an example of
one of the early efforts to formalize the role of distances among activities
within an urban region as a locational factor, drawing upon earlier work
in transportation where trip patterns were being predicted on this basis.
29/
(See Harris— also.)
The Lowry model of land use is perhaps the foremost representative of
30/
this approach (see Goldner's discussion— ), in which the level of
activity in a subarea is dependent upon the travel distance from that site
317
to all other activities in the region. Wilson— strengthened the case for
this type of model by demonstrating that it provides a maximum likelihood
estimator of the area's pattern of development, given limited information
on personal travel characteristics and the attractive features of a site, and
there is a quite extensive literature in this area. Linked with models to
predict (1) attractive-features of subareas— such as those discussed
above — and (2) mean trip length characteristics (see Voorhees—), the
Lowry type model provides a useful predictor of density of activity.
21
-------
A significant problem in these transport-oriented models (one present
also, to a lesser degree, in the other work reviewed) is that they start
from a knowledge of the locations of the principal --or basic -- industries
in an urban area. The factors determining the historical location of
industries and thus their current patterns of spatial distribution, within
a metropolitan region, have received little attention, although it might
be expected for example that industrial concentrations would occur close
to harbors and rail heads. Relative to current knowledge in this area,
location of basic industries might be judged a random process. This
situation obviously presents a potential problem for the Simulation City
concept, for it cannot be said with any certainty that it would be possible
to develop empirical predictors of industrial location as a function pri-
marily of natural features. Greater user input would then be required.
DISTRIBUTION OF SOCIO-ECONOMIC FACTORS
Beyond the simple description of. what activities occur in which areas of
a metropolitan region, and with what intensity, there are a range of
factors describing socio-economic qualities of these activities which
must also be distributed spatially throughout the region. Rents, income,
taxes, auto ownership, racial character, and many other such factors
32/
are of interest in various planning models. Bellomo, et.al.,— for
example, found that median income within subareas of a city was highly
correlated with residential density. Rent levels and auto ownership
would then be expected to be correlated with density, via their relation-
ship to income.
33/ 34/
Land values have been found (see Wieand— and Downing— for example)
to be inversely correlated with distance from the central business dis-
trict. As might be expected, commercial land value is higher along
35/
streets having higher traffic levels. Darling— was able to identify a
22
-------
similar increased value in vacant land and residential areas, attributable
to parkland having significant bodies of water.
The likely level of commercial sales at shopping centers was found
36/
(Lakshmanan— ) to be related to accessibility as well as size. The
accessibility of the center was computed as a function of the income of
population surrounding the center, weighted by the distance of the
population of given income from the center.
SIMULATION OF THE SPATIAL PATTERN
Referring again to Figure 1, the above paragraphs have been
concerned primarily with simulation of the spatial pattern of the city, and
to a lesser degree, with what is termed descriptive structure. Starting
only with a few major features of the urban area -- major highways,
rivers, lake or ocean frontage, location of city center --it would appear
to be feasible to simulate the locations of basic structural features:
land use types and levels of activity.
First, estimates are made of the overall pattern of activity intensity,
based upon city size and type. In general, intensity decreases
exponentially as distance from the center increases. More specific
information on intensity levels could then be inferred from the location
of major features. Activity intensity concentrates along highways, for
example.
The greater the amount of information which, is input to the simulation,
the greater will be the possible level of detail or accuracy. For
example, if it is possible to indicate, prior to simulation, which sub-
areas within an urban region are heavily developed, and which of these
23
-------
subareas are older, it will be more reasonable to attempt to simulate the
locations of services of all types -- sewer, utility, school, hospital.
These activities are typically located as a function of political factors
and land availability, and less often through consideration of access to
users of most desirable building sites. It is then rather difficult to
estimate location through information other than historical observation.
This difficulty is also a reflection of the problems inherent in any
simulation effort, as the level of detail to be simulated increases.
OVERVIEW: BASES FOR SIMULATION
The preceding discussion has attempted, through a brief review of a
selection of theoretical and empirical studies in urban economics and
»
geography, to illustrate that the concept of simulating a city --or rather
a detailed description of the city --is perhaps a reasonable undertaking.
Information such as that reviewed here provides a basis for devising a
Simulation City model.
As has been pointed out previously, the appeal of the Simulation City
approach stems from the possible trade-off between cost and accuracy in
preparing a data base. It has also been pointed out that accuracy is, in
this case, judged by how effectively the simulated data base reproduces
the results which would be obtained by using a very detailed, hand built
data base for input to a given planning model. It is this strong — and quite
necessary -- tie between a Simulation City model and the planning model
for which it will be used which makes it impossible to judge with certainty
how accurate, in general, a Simulation City might be. The quantity of
work reported in the literature suggests that the concept is at least
feasible, although simulation should not be carried too far.
24
-------
The next chapter will examine the trade-off between analysis cost and
accuracy which is the principal motivation for this study. If cost can be
reduced sufficiently to make planning tools more accessible to decision-
makers, without reducing accuracy so much as to render the model
results useless, then the Simulation City approach will be a successful
experiment.
25
-------
SECTION V
THE TRADE-OFF OF ACCURACY AND COST
As stated above, the principal motivation for exploring the Simulation
City concept lies in the possibility that an adequate data base might be
produced for a given planning model, at a significantly reduced expendi-
ture of time and other resources. The data base is judged to be adequate
if it produces results similar to those which would be produced by a
data base prepared by standard methods -- similar in the sense that the
planner-decision maker using the model would reach the same conclusion
with either base.
A GENERAL VIEW OF THE TRADE-OFF
The judgement of adequacy of a data base must ultimately be predicated
upon the final results of analysis. Figure 5 illustrates the general
concept of the postulated trade-off at this final level. It is an anticipated
increase in accuracy of prediction which motivates efforts to increase
the detail of analysis, which in turn is obtained at an increase in cost.
An increase in detail and cost may be associated with the analytical
model or with the data it utilizes as input.
Increasing cost will yield some increased accuracy, at a diminishing
rate, up to some range where the present limit of knowledge is
approached. In this range, current understanding of the functional and
empirical relationships involved in a process are inadequate to deliver
any increased accuracy, except perhaps on a purely chance basis.
Further efforts to increase the level of detail in analysis may actually
26
-------
reduce overall accuracy by introducing more potential sources of error
and uncertainty. —
The question to be addressed in judging the value of a Simulation City
approach is whether, given a particular analytical model for which the
data is to be provided, is the rate of trade-off of accuracy versus cost
low enough that the potential savings in the cost are not outweighed by
the losses in accuracy. Figure 6 illustrates this trade-off.
The potential differences in cost between a standard and a Simulation
City approach may be substantial. A typical large scale urban transpor-
tation study, utilizing a standard battery of models, may easily have a
cost in the range of $200,000 to $400, 000. A large percentage of this
cost -- up to 45 percent is not out of the ordinary -- may be associated
with basic data collection and preparation. Only 15-20 percent of the
4/
cost is typically associated with the investigation of decision alternatives— .
Experience with analyses using a typical land use projection model
provides another example. Here it is found that approximately $15, 000
to $20, 000 may be required in data preparation. This expense may
represent 30 to 50 percent of the total cost of the analysis.
•
Experience with analyses using a typical land use projection model has
indicated that approximately $15, 000 to $20, 000 may be required in data
preparation. This expense may represent 30 to 50 percent of the total cost
of the analysis.
It is estimated that preparation of a data base for the GEM model requires
approximately $20, 000 to $30, 000 per city. In this case, a significant por-
tion of the preparation work is judgmental, which tends to reduce costs.
27
-------
si
s
a.
Cost or Level of Detail
FIGURE 5 HYPOTHETICAL RELATIONSHIP OF
COST AND ACCURACY IN PLANNING MODELS
I Anticipated cost
/Simulation City Approach
Cost or Level of Detail
FIGURE 6 INFLUENCE OF UNCERTAIN KNOWLEDGE OF
COST-ACCURACY TRADEOFF RELATIONSHIP
28
-------
Clearly, there are great opportunities potentially available for the
reduction of data-related costs. If, as preliminary estimates indicate,
a Simulation City model could be implemented at a first cost of roughly
$60, 000, with a subsequent $5, 000 or less required per data base, then
the approach is quite promising. What makes the problem difficult is the
lack of a straightforward approach to assessment of accuracy.
Errors or inaccuracies in models come from two sources: There are
errors in measurement of data, which are carried through to the final
results; and there are errors in the specification of model relationships,
springing from lack of understanding or approximation of behavioral
relationships.
That is, even if there were perfect knowledge and measurement of the
conditions described by input data, there would be some degree of error
in output.
A model will be of potential value in decision-making if the level of total
error from these two sources is small relative to the critical ranges of
decision variables. For example, an error of two to three million
people in an estimate of the total population of the United States is of
little consequence in discussion of national growth policies, while a
similar absolute error would be clearly unacceptable for dealing with
sewer investments within a single metropolitan area. The amount of
error in a prediction must be viewed in proportion to the prediction
itself and how it will be used. In the present context, two conclusions
are implied: First, any judgment of a Simulation City approach must be
made with a basic assumption that the combined errors inherent in the
standard data preparation and utilization procedures are at an acceptable
level. Second, because it is generally the case that less detailed infor-
mation is required for policy decisions than for operating-design deci-
sions, increased error introduced by a Simulation City approach will
29
-------
tend to reduce the usefulness of the planning model for this latter type
of decision, although not necessarily for the former.
A major problem with current planning models, and the studies which
utilize them, is the lack of knowledge regarding levels and sources of
error. In a typical transportation study, the various models are able to
reproduce the base time period (observed) data on travel to within an
overall range of error of approximately 10 to 20 percent. However,
this level of accuracy is achieved through the manipulation of such a
large number of factors that it is unclear what the sources of error
37/
might be. One useful study in this area — suggests that the variation is
due largely to data, and in fact may represent an over-correction of
variance. Further, this level of-variation might lead, at the extreme,
to a shift in decisions about operational matters (such as numbers of
highway lanes, and in contrast to policy matters) in fewer than one-third
of all cases, a finding which casts light on the influence of error.
It might then be concluded that the base-line level of accuracy against
which a Simulation City approach would be compared might be measured
by a 20 percent average variation of simulated data from observation.
At this level, Simulation City data would be as effective as data prepared
by standard methods, and probably would lead to erroneous decisions in
no more than 30 percent of all cases. This guideline is useful only in
the absence of better information, for it is based upon observations of
models and decisions which are not necessarily comparable. In the
particular context of this study, there is little basis for believing that
the GEM model will behave like standard transportation models. Further,
only design questions have been treated, and nothing has been said about
policy-related questions.
30
-------
LOSSES IN ACCURACY THROUGH THE SIMULATION CITY APPROACH
The above discussion was directed at establishment of baseline informa-
tion on data accuracy, relative to its impact in dec is ion-ma king. The
next step is to investigate what losses in accuracy might result from
utilization of a Simulation City approach.
38/
Alonso's discussion of data— is instructive. He bases his discussion
on a formula which relates error in output to errors in input, when
applied to some predictive model Z = f(x,, x0, . . .,x ):
12 n
2 _ " 2 2 n n
z ~ ^ x e x + I Z ^ ^ e e r..
. - i i . i x-i xi x,- XJ
where:
e = error in the predicted variable (z)
Z
f = partial derivative with respect to input variable x.
xi l
r.. = correlation between variables x. and x.
13 l 3
This equation is exactly correct only in the event that the function f(x.) is
39 /
linear, but is a good approximation for many cases.—
Using this approximation, it may be seen that an increase in error in
one input variable may have a relatively significant influence upon
overall prediction error, if the input variable is correlated with others.
Large scale planning models are more likely than not to have such
correlation. If the errors in all input variables increase by the same
proportionate amount, then the error in prediction will increase by
this same proportion. This condition suggests that the upper limit of
31
-------
loss of accuracy due to increased error in any one or more input
variables is equal to the maximum increase in error in any one of the
variables.
If a Simulation City model can reconstruct a city data base to within a
20 percent range of error relative to standard data, for example, then
one might expect at most a 20 percent increase in error or uncertainty
in the predictions of the planning model for which the data were
prepared. That is, it may be expected that the final results of the
modeling process would exhibit no greater loss of accuracy than that of
the most poorly simulated variable in the Simulation City model.
The degree to which the overall loss in accuracy would be less than this
maximum would depend upon the planning model with .which the Simulation
City model is linked. If the model is relatively sensitive to variations in
particular variables, then it is desirable to minimize levels of error in
these variables.
As will be seen in discussion of Simulation City design in the appendices,
the conceptual descriptions of computational sequences commence with
those data items which are better understood and which may thus be
simulated with greater accuracy. For example, models to predict trip
length will typically account for greater than 90 percent of the variation
in observed data.
Regression studies such as those cited in Chapter II have developed models
for socio-economic factors which generally account for 60 to 90 percent
of observed variations in variables. These variables are included at the
second stage of Simulation City procedures. Items treated in the last
stages of the simulation are typically those for which there is little back-
ground information or for which results have been poor.
32
-------
ACCURACY AND COST IN THE GEM CONTEXT
As a test case, implementation of a Simulation City model for input to the
GEM model presents a number of difficulties. One of the principal ones
among these is that GEM is a new model, with relatively little history of
application. It is not clear to what variables GEM may be most sensitive.
Further, GEM utilizes an abstracted, representational data base; it is
thus unclear what meaning inaccuracies introduced by a Simulation City
approach might have.
Finally, the GEM model is planned to be among the most comprehensive
planning-analysis models in existence. It requires a wide range of data,
much of which may be difficult to obtain by standard collection procedures.
A Simulation City model, if implemented to supply a large portion of this
data base, will represent a severe test of the Simulation City concept.
Because of the lack of decision-making experience with GEM, there is a
tendency to assess the accuracy and cost of a GEM-linked Simulation City
model with respect to either the real metropolitan area, directly, or the
abstracted data base normally prepared for GEM. In the former case,
as has been pointed out throughout prior discussion, one is really asking
more of the simulation than it purports to deliver. In the latter case, there
are many aspects of the data which contain a measure of the judgment of
personnel preparing the data base. Hence, it may be unclear what are
the sources and resolution of discrepancies between standard and simulated
data.
In contrast to these difficulties, it is clear that GEM is intended to be used
at broader policy levels, and not for operational analysis of urban environ-
mental systems. A much higher level of inaccuracy might then be tolerated
in a Simulation City model, and there is much greater incentive to reduce
data costs.
33
-------
As was stated previously, it has been estimated that data preparation for
a moderate to large sized metropolitan area (0. 5 to 1. 5 million people)
would require approximately $30, 000. It is currently estimated that approxi-
mately $60, 000 would be the cost of implementation and preliminary testing
of a Simulation City model for GEM. The cost of application of the simulation
to prepare a data base might then be approximately $5, 000. There is thus
a fairly significant saving to be realized if a Simulation City model is
implemented and widely applied.
It would seem quite important, however, that any judgment on the Simulation
City concept be -made within a context of the total GEM system. Within
the context of a particular application, a Simulation City model could be
quite useful. However, maintenance of the context of application is critical.
34
-------
SECTION VI
REFERENCES
1. Forrester, J.W., World Dynamics, MIT Press, Cambridge, 1971.
2. Commoner, Barry, "Alternative Approaches to the Environmental
Crisis," Journal of AIP, May 1973.
3. Lee, Douglass B., "Requiem for Large-Scale Models, " Jpurnal of AIP,
May, 1973.
4. Hillegass, T.J., "Urban Transportation Planning-A Question of
Emphasis, " Traffic Engineering, June 1969.
5. Voorhees, Alan M. t and Associates, Factors and Trends in Trip Length,
NCHRP 48, Highway Research Board, Washington, D.C., 1968.
6. Harris, C. D., "A Functional Classification of Cities in the United
States, " Geographical Review, January 1943.
7. Berry, B. J. L., City Classification Handbook; Methods and Applications,
Wiley, New York, 1972.
8. Berry, B. J. L., and F. E. Horton, Geographic Perspectives on
Urban Systems, Prentice-Hall, Englewood Cliffs, 1970.
9. Pidot, G. B., and J. W. Sommer, Modal Cities (Statistical Phase),
Tech. Report 1, Project SUPERB, Dartmouth College, Hanover,
1973. (Draft)
10. Golub, T.F., E. T. Canty, R. L. Gustafson, Classification of
Metropolitan Areas for, the Study of New Systems of Arterial
Transportation, GMR-1225, GM Research Labs, Warren, 1972.
11. Friedrich, C.J., Alfred Weber's Theory of Location of Industries,
University of Chicago Press, Chicago, 1929.
12. Bergsman, J., P. Greenston, and R. Healy, "The Agglomeration
Process in Urban Growth, " Urban Studies, October 1972.
13. Morrisett, I., "The Economic Structure of American Cities",
Papers and Proc of the Regional Science Assoc., Vol. IV, 1958.
14. Bryce, H.J., "Identifying Socio-Economic Differences Between High
and Low Income Metropolitan Areas, " Socio-Economic Planning
Sciences, April 1973.
15. Blumenfeld, H., "The Economic Base of the Metropolis, " Jour, of
AIP, 1959.
35
-------
16. Isard, Walter, Location and Space Economy, M. I.T. Press,
Cambridge, 1956.
17. Borcherding, T.E., and R. T. Deacon, "The Demand for Services
of Non-Federal Governments, " American Economic Review,
Dec. 1972.
18. Stone, P. A., "The Economics of the Form and Organization of
Cities, " Urban Studies, Oct. 1972.
19. Sussna, S., "Residential Densities or a Fool's Paradise, " Land
Economics, Feb. 1973.
20. Bellomo, S. J., C. G. Turner, and D. K. Johnston, A System
Sensitive Approach for Forecasting Urbanized Area Travel Demands,
A. M. Voorhees & Assoc., for US DOT, FHWA, McLean, 1971.
21. Wells, J. D., et. cal., Economic Characteristics of the Urban
Public Transportation Industry, Inst. for Defense Analyses, for
US DOT, Arlington, 1972.
22. Weimer, A.M. and H. Hoyt, Real Estate, Ronald Press, New York,
1966 (5th Edition).
23. Timms, D. W. G., The Urban Mosaic: Towards a Theory of
Residential Differentiation, Cambridge Univ. Press, Cambridge
(U.K.), 1971.
24. Barton-Aschman Assoc., Inc., Guidelines for New Systems of
Urban Transportation, Vol. I: Urban Needs and Potentials, for
US HUD, Washington, 1968.
25. Mills, E.S., Studies in the Structure of the Urban Economy,
Resources for the Future, (Johns Hopkins), Baltimore, 1972.
26. Grytzell, K. G., The Demarcation of Comparable City Areas by
Means of Population Density, Gleeruf, Lund, 1963.
27. Chapin, F. S., "Patterns of Urban Development, " in Urban Growth
Dynamics, Chapin and Weiss, ed., Wiley, New York, 1962.
28. Hansen, W. G., "How Accessibility Shapes Land Use, " Jour, of AIPf
1957.
29 Harris, B., "A Note on the Probability of Interaction at a Distance, "
Journal of Regional Science, Vol. 5, #2, 1964.
36
-------
30. Goldner, W., "The Lowry Model Heritage, " Jour, of AIP, March,
1971.
31. Wilson, A. G., "The Use of Entropy Maximizing Models, " Journal
of Transport Economics and Policy, Jan. 1969.
32. Bellomo, S. J., R. B. Dial, and A. M. Voorhees, Factors, Trends
and Guidelines Related to Trip Lengths, NCHRP 89, HRB,
Washington, 1970.
33m Wieand, K. and R. F. Muth, "A Note on the Variation of Land
Values with Distance from the CBD in St. Louis, " Jour, of
Regional Science, Vol. 12, No. 3, 1972.
34e Downing, P. B. , "Factors Affecting Land Values: An Empirical
Study of Milwaukee, Wisconsin, " Land Economics, Feb. 1973.
35, Darling, A.H., "Measuring the Benefits Generated by Urban
Water Parks, " Land Economics, Feb. 1973.
36. Lakshmanan, T.R., A Theoretical and Empirical Analysis of
Intraurban Retail Location, Ph. D. Thesis, Dept. of Geography,
Ohio State Univ., 1965.
37. Creighton, Hamburg Planning Consultants. Data Requirements in
Metropolitan Transportation Planning, NCHRP 120, Highway
Research Board, Washington, 1971.
38. Alonso, W., "The Quality of Data and the Choice and Design an
Predictive Models," in Urban Development Models. Special Report
97, Highway Research Board, Washington, 1968.
39. Wilson, E. B., An Introduction to Scientific Research, McGraw-
Hill, New York, 1952.
37
-------
SECTION VII
APPENDICES
APPENDIX A
SIMULATION CITY APPLICATION FOR GEM:
AN ILLUSTRATIVE DESIGN*
As was explained in the body of the report, one of the motivating factors
for this study of the Simulation City concept has been the possibility that
the concept might be applied within the context of the Environmental
Protection Agency's on-going work on GEM, the General Environmental
Model. A preliminary design for such a Simulation City model serves as
an illustration of the entire concept, its various problems and possibilities,
One difficulty with the use of a GEM-linked model as an example is that
GEM itself differs in a number of ways from typical planning models:
First, the representation of geographic features in GEM currently is
based upon a square grid. All roads must follow the straight edges of
square parcels measuring one mile along the edge. These one square
mile parcels are the minimum land unit size, and may have only one
major type of land use per parcel. All economic components of the
metropolitan system are represented in discrete units -- population,
employment, production, consumption -- albeit they are small relative to
total city activities.
On the other hand, by making such approximations, it has been possible to
expand the purview of GEM, to make it one of the more comprehensive
models of urban systems activity currently available. The design and
implementation of a Simulation City model to "feed" GEM thus represents
an extremely broad and relatively severe test of the concept, and will
bring into focus the limits to which the simulation may be pursued.
*Throughout this discussion, a preliminary knowledge of the GEM
characteristics and intent is assumed.
38
-------
GENERAL APPROACH
As stated above, GEM representation of a metropolitan area is in terms of
discrete valued variables, i. e., measured in discrete, whole units. In
the case of such variables as population and employment, the units are
small relative to the entire system, so that no major problem is created.
Spatial distribution of activity, however, is expressed in terms of the
standard land parcel, upon which there may be only one type of land use.
Densely settled areas, particularly central cities, are thus not well repre-
sented, although it is unclear, on other than intuitive grounds, how seriously
the error influences model performance.
Because the GEM representations are unlike those used in many other planning
models, there is a major question to be answered early in the design of the
Simulation City model: Should simulation be carried on in the format
required for GEM, or should a more general format be employed and data then
adjusted to fit the required format? The decision was made to try to follow
the latter course, to provide more versatility in the simulation. On-going
study and revision of the GEM system makes such versatility especially desir-
able. But it must be recognized that this versatility may be achieved at some
expense in efficiency. Ideally, the Simulation City model would be closely
matched to the planning model to which it provides input.
Figure A-l illustrates the conceptual design for a GEM-linked Simulation City
model. * As shown, the design is developed in three stages: The first stage
deals with those items of data about which a relatively large amount is
known, and which are generally viewed as comprising the basic socio-economic
#
The conceptual description corresponds in only a general way with
anticipated computational sequences. See Appendix B for more detailed
discussion of the computational approach.
39
-------
Pattern
Descriptors
Business Goods
And Services
Locations
Personal Goods
Trip
Length
Characteristics
Locations of Municipal
And Government Services
Locations of Residences
Associated with Services
FIRST STAGE PARCEL ASSIGNMENTS
II.1
11.2
M.3
Levels for Schools,
Utilities, etc.
*—
Location Assignment
Of Soci o-Economi c 1 *•
Characteristics |
Demand Indices
11.4
Location Assignment
Of Structural Services
SECOND STAGE PARCEL ASSIGNMENTS
111.1
111.2
III.3
Effluent and
Pollution
Treatment
Taxes, Assessed
Values, and Other
System Measures
(
II 1.4
Value Ratios and
Other Operating
Measures
Educational
And Other
Social Measures
OUTPUT LOAD DECK
FIGURE A-l MACRO-SCALE DESIGN OF A GEM-LINKED SIMULATION CITY MODEL
40
-------
structure of a metropolitan area. In this portion of the model, all simulation
would be handled within the model itself. That is, the combination of
required input and information included in the model would be sufficient to
permit inferences about quantities and spatial distributions of population
and employment.
The second stage of the Simulation City model deals with aspects of the
infrastructure of a metropolitan area. While the activities treated at this
stage are always present in an area, their levels and relative locations vary
widely, in response to a variety of historical, economic, and political
conditions. In this stage of the model, the users judgment will be utilized
to resolve questions for which current theory and empirical findings are
inadequate. In this case, information will be displayed to the user, and he
will be requested to make certain decisions based upon this display. For
example, the locations of bus routes would be specified by the user, based
upon a display of the locations of residential and employment parcels and
an estimate of the number of miles of bus route that the city would have.
The third stage of the model treats a variety of data items required as input
for GEM, about which there.is too little knowledge to provide a reasonable
basis for simulation. These items are of two types: First there are those
which have not been subjected to sufficient study, and which in all likelihood
could not be predicted. The pollution control technologies in use at specific
sites in a city are an example of this type of item. It seems intuitively
impossible to predict this information in any general way. The second type
of data item is defined within the GEM system. The value ratio, for example,
while quite reasonable in concept, has a limited basis for practical measure-
ment and study. Such variables must be determined by the GEM framework
itself, or specified completely by the user.
41
-------
The progression of Figure A-l may be viewed as a series of operations arrayed
along a scale of trade-offs between objectivity—reflecting theoretical and
empirical regularities of urban areas—and subjectivity—reflecting concepts
highly abstracted from reality to accommodate the planning model. In the
following pages, which discuss Figure A-l in some detail, this trade-off will
be apparent.
THE CONCEPTUAL MODEL
The following paragraphs explain, item by item, the steps of the Simulation
City model presented in Figure A-l. As previously explained, this is pri-
marily a conceptual presentation, not necessarily corresponding to the compu-
tational sequence which would be followed in an implementation of the model.
Appendix B contains a more detailed step-by-step description, which is more
closely related to computational considerations.
Reference is also made to Chapter II, which reviewed a substantial portion
of the background literature for this discussion. An effort has been made to
simplify the discussion by avoiding excessive referencing of sources and
justifications of information upon which the design is based.
Input
Input for the Simulation City model is intended to represent the minimum amount
necessary to yield a useful simulation. In general, an effort has been made to
restrict input to those items contained in the Statistical Abstract of the United
*
States, and USGS topographical quadrangle maps (71/2 minute series).
The input is of two types, as shown in Figures A-l and A-2. Gross
descriptors (1.1*) refer to the overall socio-economic system which is the
* These numbers refer to the item designations in Figure 1.
42
-------
FIGURE A-2
PROPOSED INPUT TO SIMULATION CITY MODEL
Gross
Descriptors
Pattern
Descriptors
Social
School enrollment
Student-teacher
ratio*
Public assistance
payments
Count of schools.
per parcel*
Parks
Political
Local government
employees
Bus company
type*
Juris dictional
boundaries
Economic
Median income
Distribution of family
incomes
Manufacturing
employment
Wholesale employment
No. of housing units.
total and central city
% of housing in single
unit structure and
central city
Median value, single
family housing
City budgets
Industrial cluster
City type
Tax types*
Unemployment rate
Mean rent
Annual payrolls
Locations of industrial
concentrations *
Dense activity parcels
Physical
Total population
Population density.
total and central
city
Fraction of workers
using public transit
Mean trip length*
Highways
Rail terminals*
Rivers
Surface water
(lakes, ocean) .
Major facilities
(airport, harbor)*
Subway-rail system
Non-developable
land
Water intake, out-
flow points
•Will not be required, but very desirable.
43
-------
without reference to the specific geography of the metropolitan area. Pattern
descriptors (I. 2), in direct contrast, are intended to convey the historical and
geographic characteristics which have been instrumental in shaping the present
form of the area. The analogy might be made that pattern descriptors are
related to the skeleton, while gross descriptors are related to the levels,
distribution, and conditions of sinews encompassing the skeleton.
Gross Descriptors (I. 1) - Several aspects of the city's economic and social
system are included in the category:
• Spatial variation in activity intensity - It has been observed
that the density of population, industrial activity, and
commercial activity in a metropolitan area—and perhaps
such factors as land value, as well—may be estimated as a
function of distance from the center of the city. The form
for such estimation will be the exponential function,
a(d) = A e'Yd
°a(d)
where a(d) is the density of activity at distance d, A is a
measure of the density at the center, Y is the rate of
decrease of density. Alternatively, it has been observed
that certain activities, primarily population, exhibit
a "hole" in the city center, a sharp decrease in density
(see Figure A-3). In this case, a gamma functio_n might be
used, and estimation of two parameters would be adequate to
to model the likely variation in activity density as a
function of distance from city center. In both cases, circular
symmetry of the city is assumed.
• Economic activity composition - The overall nature of a
city's economy may be characterized by the proportion of
its employment in manufacturing, the ratio of basic to total
employment, description of the city as one of several stan-
dard types, median, income, etc. Knowledge of this overall
nature will provide a basis for allocating employment, pre-
dicting price levels, and determining infrastructure needs.
• Character of service - the term "character of service" is
here intended to cover a range of factors such as infra-
structure, school and governmental quality, education of
44
-------
in
0)
Gamma
Exponential
Distance
FIGURE A-3 VARIATION OF ACTIVITY WITH DISTANCE FROM CITY CENTER
Radial Arterial-Oriented
Early Metropolis
Grid Arterial-Oriented
Mature Metropolis
FIGURE A-4 HISTORICAL PATTERNS OF URBAN FORM BASED UPON TRANSPORTATION GROWTH
45
-------
the populace, income ranges, and amenities provided to
the populace. Simple regression models will be employed
to predict likely levels of sewer service and cost, parks
and recreational facilities, educational expenditures,
transportation levels of service, etc.
Through considerations such as these, the scales of various components of
the metropolitan system are estimated. Within the Simulation City model,
these estimates are expanded to describe more completely the city data base.
Pattern Descriptors (I. 2) - This set of descriptors is concerned with the
factors of geography and history which cause each city to develop an individual
form. These factors fall into several categories:
• Topographic features - rivers, major bodies of water,
marsh land, harbors, flood plain
• Political-development features - City center, major
government institutional development centers, major
highways, rapid transit lines, sanitary landfills.
• Historical development features - Transportation devel-
opment patterns (see Figure A-4), settlement orientation
(the recreation versus industrial orientation, leading to
dispersion versus concentration of activity).
Such locational factors are indicated within the grid which will become a mapped
presentation of the city data base. The Simulation City model will proceed
from these starting points, through inference, to distribute the development
estimated to be present in the metropolitan area.
Estimation of Levels and Patterns
The process of expansion of basic input information into a preliminary picture
of the metropolitan area has been partially described above, and is illustrated
m Figure A-4. This expansion consists of two parts: the estimation of the
levels of various economic activities in the city (I. 3), and the estimation of
measures indicating the relative likelihood that these activities will be
located in specific sub-areas in the metropolitan region (I. 4).
46
-------
Levels of Activity (I. 3) - The simulation of levels of activity in a GEM-
linked Simulation City model will be dependent upon the data rules of GEM
itself. This operation will be one essentially of using the Master Tables
directly to translate the input data. Beginning, for example, with the total
population and the manufacturing employment, the number of levels of
industry may be estimated. This estimate will be modified slightly, and
distributed among specific industries, by income distribution. Required
levels of business goods and services may be computed directly, and an allo-
cation of population units to employment in these two categories, estimated
from the census input. Similarly, required levels and allocation of workers
may be estimated for personal goods and services, municipal services,
schools, etc.
It will be necessary to go through the cycle more than once, in order to
balance population and employment, i. e., to attain the greatest number of
industrial levels compatible with the total population of the city, its distri-
bution of income, and its type and industrial character. It is anticipated
that the computational sequence may change with city size, for it is only
in large cities that goods and services, as treated in GEM, begin to have
significant influence on the distribution.
The numbers of parcels required to accommodate the city will be computed
directly from the allocation of levels, subject to constraints of central
city and metropolitan area population densities, jurisdictional controls,
and a desire to minimize the amounts of undeveloped land within the city's
built-up areas. This computation again is a relatively straight-forward
application of the Master Tables.
Proceeding in this fashion, the number of people, level of industry,
levels of employment, etc., as required for GEM, are estimated. As shown in
Figure A-5, these estimates might be displayed in tabular form, providing a
picture of the socio-economic system of the metropolitan area.
47
-------
PROPOSED INPUT TO SIMULATION CITY
Gross
Descriptors
Pattern
Descriptors
•
Social
School enrollment
Student-teacher
ratio*
Public assistance
payments
Count of schools,
per parcel*
Parks
Political
Local government
employees
Bus company
type*
Jurisdictions!
boundaries
*WII1 not be required, but very desirable.
W«
Economl c
Median Income
Distribution of faml 1y
Incomes
Manufacturing
employment
Wholesale employment
No. of housing units,
total and central city
1 of housing In single
unit structure and
central
city
Median value, single
faml ly housing
City budgets
Industrial
City type
Tax types*
cluster
Unemployment rate
Mean rent
Annual payrolls
Locations of Industrial
concentrations*
Dense activity parcels
I
• Er— '
1
fffgifj
Jjm
•^H rt^foKGR.
fft JE§
m&~~
••
Physical
Total population
Population density,
total and central1
city
Fraction of workers
using public transit
Mean trip length*
Highways
Rail terminals*
Rivers
Surface water
(lakes, ocean)
Major facilities
(airport, harbor)*
Subway-rail system
Non-developable
land
Water Intake, out-
flow points
LC| parks
e dense activity parcels
© rail terminals
$& major facl 11 ties
"" non-developable land
•^•••i river
h 1 ghways
l.i
Gross
Descriptors
1
1 1.3
Levels of
Activities
1
.2
Pattern
Descriptors
1
1.4
Attraction
Indices
m i
Activity
MF
NL
S'N
BG
BS
PG
PS
MS
SC
i
i
RA
r
Constructed
Levels
3
1
2
Associated Population
PH
13*
14*
PM
12
i
PL
12
I
Parcels
1
1
•
*Note shortage
of high Income
labor force. . .
simulated from
labor statistics
and Income.
^i^^^j
FIGURE A-5THE FIRST STEP
MATRICES
[.IA]
I- parcel number
A- activity type, e.g.
-heavy Industry
-light Industry and
commercial
-residential, high and
medium
-residential, low and
medium
IN SIMULATION
48
-------
Attraction Indices (I. 4) - The attraction index may best be understood—
and will be computed—as a probability that a particular activity will be
found on a particular parcel, given the information known about that parcel
and its surroundings. The attraction index of a parcel (i) for an activity
(k) would be computed as:
where:
A = an overall city measure of activity density
p., = a measure of features specific to the individual parcel
S, = scale factor, such that S, = I a • p..
The first component of the index, an overall density measure, would be
derived from gross descriptors input. As explained previously, estimates
of the parameters of an exponential distribution could be computed to give a
predictor of the density of heavy industry activity, in a city of a particular
type, as a function of distance from the center of the city. . Then:
-ykt di
aik « Dkt 6
where:
^kt = exponential parameters for heavy industry (k)
in a city type (t)
d. = center-to-center distance of parcel (i) from the
1 central parcel.
This component of the index will indicate a tendency of activities to locate
centrally, with some types of cities typically having greater concentration
or compactness.
49
-------
The second component, a parcel specificity indicator, would be derived
from pattern descriptors input. A range of considerations would be
reflected in this component, as suggested in Figure A-2. The component
could be formulated as:
I.. , + M., _ + M.
ikl ik2
where:
V
ik
ik
» fik2 = feasibility measures indicating whether it is
allowable that parcel (i) attract activity (k)
= a measure of the availability of space in parcel (i)
Mikl* Mik2' Mik3 = measures of amenities of parcel (i) relative to
activity (k).
For this specific example,
ikl =
fik2
-,
ik =
-ui
ikl
M
ik2
M
Ik3
1 if the parcel is developable
0 if it is not
1 if there is no activity already on the parcel which
conflicts with activity (k)
0 if there is such a conflict
a ratio of the percentage of undeveloped land in parcel (i)
to the percentage of a parcel required for one level of
activity (k)
access measure, the product of the number of sides
of the parcel having highway access multiplied by the
average level of the highways.
Quality measure, a defined scale giving value to the
presence of complementary features on the parcel
environs measure, a defined scale giving value to the
presence of complementary or conflicting activities
or features within the parcels immediately adjacent
to parcel (i).
50
-------
As a minimum, there would be two attraction index matrices, one for industry
and commerce, and one for residences. It is likely that in larger cities there
would be one each for heavy and light industries, low income residential,
and goods and services.
Attraction indices would be modified as each activity is allocated. In the
above example, parcels would have changes in f and M , dependent
upon the sequence of assignment. Figure A-6 illustrates this recomputation.
It is anticipated that some research would be required to implement this
component of the Simulation City model. Standard regression analysis
would be utilized to develop attraction indices equations of the types
described above.
Location of Activities
Actual simulation of the form of the metropolitan area begins with the location
of activities. Generally speaking, the Simulation City model will distribute
the previously estimated levels of socio-economic activity over the grid of
the metropolitan area, in accordance with the previously computed attraction
indices and various rules derived from theories of urban growth. Figure A-6
illustrates aspects of the procedure, which will start with the location of
industry.
Industry Locations (1.5. 1.6) - As mentioned in Chapter II, basic industry is
generally viewed as the central force in the metropolitan system, from the
standpoints of both economics and locational patterns. However, research
into the causal factors influencing location of basic industry has not been
noticeably successful. Hence, designation of parcels containing basic industry
is not an especially promising candidate for simulation, particularly in view
of the GEM constraint of one land use per parcel. It would be preferable for
51
-------
M,F
N'S
BG
BS
0.0 0.0
0.0
1.5 1.0
- s.o
- f.O S.O S.O 0.0
t.l ».l I.] S.O (.0 0.0 0.0
J.J t.l I.S J.I t.t 0.0 l.i
- (.0 3.J 0.0 I.I 1.1
HF Manufacture
NS National Services
BG Business Goods
BS Business Services
I '.3
Industry and
business goods
and services
1.6
Business Good
and Services
Locations
Levels of
Activities
1.5
Industry
Locations
Attraction
Indices '
r
i^i^
residential, high
and medium
o.o o.o - i.o ».j 7.1 s.t
o.o - - s.i J.I t.l s.o
...... o.o
- 0.0 0.0 0.0 0.0
- 0.0 0.0 0.0 0.0
• 0.0 0.0
BS Business Services
BG Business Goods„
I Industry
1.7
Trip
Length
Characteristics
1
Local
Res Id
Assoc
with
Actlv
|n
ons of
ences
la ted
Basic
Itles
.
1 1.9
Locations of
Personal Goods
and Services
1.10
Locations of Municipal
and Government Services
PH Population, High Income
PL Population, Low Income
PC Personal Goods
PS Personal Services
MS Municipal Services
SC School
1.11
Locations of Residences
Associated with Services
FIGURE A6 ASSIGNMENT OF LOCATIONS
52
-------
the user to designate these parcels, based upon a knowledge of the metro-
politan area.
In the absence of this knowledge - as a default measure - the attraction
index matrix for business goods and services might be modified to include
basic industry, which would be assigned in the same manner as other
activities.
The manner in which activity will be located is by selection of those parcels
having the greatest attraction indices, up to the required number of parcels
estimated previously. Figure A-6 illustrates this procedure, for six industrial
and commercial parcels. Levels of activity on individual parcels will be
allocated in proportion to the values of the attraction index in that parcel,
subject to the constraint that each parcel have at least one level of its
assigned activity.
Residential Locations (I. 7, 1.8, 1.11) - Residential locations wi 11 be
designated in basically the same fashion as industrial and commercial
(business goods and services), via the selection of parcels having the
greatest estimated attraction indices. However, residential attraction indices
will be computed differently from those for industry.
First, as illustrated in Figure A-6, the GEM constraint on land use
requires that residential attraction indices be identically zero for parcels
previously allocated to another land use. However, it would be expected
that only in dense, inner city areas would this constraint produce conflict.
A more significant modification would be made to account for travel distance.
As mentioned in Chapter II, the location of residences would seem theoretically
and empirically to be influenced by a desire to minimize the journey
53
-------
to work. The Lowry model and its successors, which Wilson has termed
"entropy maximizing" to indicate the character of the models as optimum
estimators of patterns of location, have been based largely on the journey-
to-work as the basis of residential location.
Based upon the Lowry model type of formulation, residential levels would
be allocated using the following formulation:
where
T>
iR •
P._, =. residential population potential of parcel (i)
ix\
T
ij = number of people (potentially) living in i and working in j.
A solution is found for the expression:
2d..
T.. = W. (E.T + E._x A
ij J Jl JB) AiR e 5
where
A. = residential attraction index of parcel (i)
iR
E. + E. = number of employees in industry and business goods
J * and services on parcel (j)
d.. = effective distance between parcels (i) and (j)
5 = mean trip length for work trips in the area
2d..
1/W = ?A.« r^— . a normalizing factor.;
j i iR e o
subject to the constraint that:
£ T = E r + E._,
i ij jl JB
54
-------
i. e., the total number of potential trips to a parcel (j) is equal to the number
of employees in that parcel. This solution then permits computation of P._,
IrC
for each parcel. Parcels having the greatest values of P.,, are
iJK
selected, up to the previously estimated required number; and levels are
allocated in proportion to this population potential.
In large cities, it may be desirable to perform this allocation on a stratified basis,
by income group. In this case, the above model would be applied consecu-
tively to each group with changing estimates of attraction indices and mean
trip length.
Obviously, a major component of these models is trip length. The travel
characteristics of a population are represented by a relative frequency dis-
tribution of trip lengths; and it is found that this distribution varies with
trip purpose, socio-economic characteristics of the trip-maker, and
characteristics of the transportation system.
As mentioned in Chapter II, it has been found that the distribution of trip
length may be approximated by the gamma distribution, specified by the
mean trip length. Hence, gross descriptors will be employed to estimate
mean trip length (s). City size, city type, and income distribution, would
be three parameters of major importance in this estimation.
Thus, using the Lowry model approach as a basis, residential locations
will be designated, and levels assigned, relative to previously located industry
and commerce. As Figure A-6 illustrates, this location could be done in two
stages, with personal goods and services assigned in relation to industry-
associated residences. Service residences would then be allocated with
respect to distance from this portion of commercial activity. Municipal and
other governmental services will be treated with perscnal goods, and services.
55
-------
Associated Services
The above paragraphs describe the progress of the GEM-linked Simulation
City model, through the first stage parcel assignments indicated in Figure A-l,
This first stage of assignments represents the aspects of a metropolitan
system about which a relatively significant amount is known. As has been
explained, in the second stage of assignments, an interactive approach
would be employed, an approach intended to make use of the user's judgment
in areas where rational predictive rules cannot be formulated. Figure A-7
illustrates this approach, for the case of bus routes.
As was mentioned in Chapter II, it may be possible to estimate the total
levels of various services in the city. These totals would then be displayed
to the user, to be assigned by him to particular sub-areas. The vehicle for
this assignment would be a status map, a display of particular aspects of
the simulation up to that point. Information displayed on the status map
would, in effect, define a spatial pattern of needs for services, allowing the
user to judge how needs are to be filled. In making this judgment, the user
can bring to bear all of his knowledge of actual conditions in the area, to
provide the GE M representation with as great a similarity to reality as may
be possible.
The approach to estimation of levels of services (II. 1) will be the same as
that for major activities, discussed previously. These levels represent
maximum amounts of these services which might be available in the metro-
politan region.
The demand indices (H. 3) would reflect an estimation of where these services
are most likely to be available. For example, rent and income levels in
residential parcels will be estimated (II.2), and together with job location,
would be of use in locating bus routes. The status map would then display
56
-------
| FIRST STAGE PARCEL ASSIGNMENTS |
H. 1 ||.2 H.3
Levels for Bus, Road
Utilities, etc.
—
Location Assignment
Of Socl o-Economl c I *
Characteristics |
Demand Indices
parcels where bus
service Mould
be likely
STATUS MAP
parcel requires
access
Location Assignment
Of Structural Services
bus routes
designated by
user
occupied
parcels
access designated
by user
FIGURE A-7INTERACTIVE STAGE OF A GEM-LINKED SIMULATION CITY MODEL
57
-------
the parcels containing high likelihood-of-service activity, along with
the available roads linking these parcels. The user would then designate
route location, up to the limit of mileage estimated by the Simulation City
model, as illustrated in Figure A-7.
Figure A-7 illustrates another aspect of this interactive stage of the simulation.
Because only the major highways are required to be input by the user,
location of activity in the first stage of the simulation will not be absolutely
constrained by the presence of access to a parcel. Thus it is possible that
in the second stage of simulation there would be parcels with activity and no
access to other parcels. Such parcels would be displayed on a status map,
for the user to designate a preferred access route.
Proceeding in thi s fashion, this Simulation City model will provide a basis
for the user to complete the description of a city's physical structure, in
terms of a GEM data base. The remaining data required for GEM, dealing
primarily with socio-economic conditions and pollution services, will be
treated in a third stage of the model, dependent heavily upon user knowledge
and judgment.
Operating Measures
The GEM model requires as input a number of data elements which are
especi ally diffi cult to treat in simulation, or which have a definite meaning
only within the GEM context. These items - such as value ratios, tax
rates, pollution control technologies, costs of goods and services outside of
the metropolitan regi on - refer primarily to operating characteristics of
the city, and have not been submitted to sufficient analysis to exhibit any
promise that their prediction would be feasible. The Simulation City model
will thus depend upon user input for these items, until such time as research
can be undertaken to develop the necessary data.
58
-------
As a default measure, the GEM-linked Simulation City model will include
standard values, or procedures for setting such values, for all remaining
items in the GEM load deck. It must be recognized, however, that this is
not a realistic simulation, relative to the previous stages of the model.
It is simply a stop-gap measure, intended to maintain the accessibility of
the GEM system for use on a broad range of policy-related questions.
SUMMARY OF THE GEM-LINKED MODEL
This chapter has provided a broad overview of a proposed application of the
Simulation City concept - a model to provide input to the General Environ-
mental Model. The GEM-linked Simulation City model is formulated in
three stages, representing decreasing knowledge of underlying relationships
describing a metropolitan area, and increasing dependence upon the rules
inherent in GEM itself.
The purpose of this overview has been two-fold: first, it has addressed
directly the manner in which a GEM-linked model might be constructed
(see Appendix B for more details), and has thus indicated the extent to which
such a model may be feasible to implement. Second, it has provided an
example of the Simulation City concept as a whole, and thus provides insight
into feasibility on a more general level.
The followi ng chapter examines aspects of the cost and accuracy of Simulation
City data, and thus addresses more directly the question of feasibility of
the approach. This question is one of more than simply technical capability,
but of cost and effectiveness as well.
59
-------
APPENDIX B
A SIMULATION CITY MODEL FOR GEM:
COMPUTATION-ORIENTED DESCRIPTION
Appendix A gave a general description of a GEM-linked Simulation City
model. This Appendix is intended to describe the design in a step-by-step
manner, and thus to provide a basis for complete design and implemen-
tation of the model. Figure B-l illustrates how the computational steps
presented here fit into the previous discussion. Appendix C accompanies
this description by presenting the GEM load deck as it will be treated by
the Simulation City model.
60
-------
^
Locations of Residences
Associated with Services
FIRST STAGE PARCEL ASSIGNMENTS
3.7.10.11
II.I
19,22
11.2
14-16
11.3
Levels for Schools,
Utilities, etc.
Location Assignment
SocIo-Economlc[
Characteristics
Demand Indices
23.24
11.4
Location Assignment
Of Structural Services
SECOND STAGE PARCEL ASSIGNMENTS
i
lll.l
25-30
II 1.2
32
111.3
Effluent and
Pollution
Treatment
Taxes, Assessed
Values, and Other
System Measures
i
II 1.4
Value Ratios and
Other Operating
Measures
Educational
And Other
Social Measures
R
i
i
OUTPUT LOAD DECK
FIGURE B-l MACRO FLOWCHART FOR SIMULATION CITY
61
-------
1. Determination of Population Levels
Total Population*, compared with a Master Table value, determines
the total number of Pi's in the city.
Median Family Income and the Distribution of Family Incomes are
used to split this total of Pi's into PL, PM, and PH.
2. Determination of Levels of Industries
Manufacturing Employment is used to estimate the total number of
levels of industry activity (including NS), leaving workers to be allocated
to goods and services.
Selection of one of several Standard Industry Clusters or designation
of a Percentage Distribution by the user determines the number of con-
structed levels of specific activities.
3. Determination of School and Teacher Levels
School Enrollments and an indicated-Student-Teacher Ratio (a
default will be available) are used to estimate the number of levels of
teachers. Percentage of Local Government Expenditure on Education
and the distribution of population by income (1)* are used to split these
levels between PM and PH.
The computed numbers of students from the Master Table and the
distribution of population by income (1), and the number of levels of
teachers are used to determine the number of constructed levels of schools.
4. Determination of Constructed Levels of Housing
The Numbers of Housing Units, Total and Central City, and the Per-
centages of One Unit Structures, Total and Outside Central City, are used
to estimate the percentage of housing levels in RA. The Central City
* Underlining indicates a direct input item. Numbers in parentheses
refer to other steps in the simulation.
62
-------
Population Density is used to estimate the split between RB and RC.
The distribution of population levels by income (1) provides an estimate
of space units demanded, thus permitting estimation of the constructed
levels of housing.
5. Determination of Municipal Services
The numbers of levels of industry (2), housing (4), and personal goods
and service's (8); preliminary estimates of business goods and services
(derived from industry levels); and the Number of Local Government Employ-
ees are used to estimate the number of levels of municipal services. PL and
PM employment levels are assigned to these services.
6. Estimation of Housing Conditions
The levejs of constructed housing (4) are allocated to population levels
by income (1), using standard allocations and a comparison of Median
Family Income and Median Value of Single Family Housing.
7. Determination of Utilities Levels
Housing conditions (6) and the levels of industry (2) are used to estimate
requirements for power, water treatment, and solid waste. Constructed.
levels of these activities are then estimated to provide for these requirements.
8. Determination of Levels for Personal Goods and Services
Housing conditions (6) and population levels .(1) are used to estimate
constructed levels of PG and PS. PH, PM, and PL worker units are assigned
directly to these goods and services.
9. Determination of Business Goods and Services
Levels of industry (2), personal goods and services (8), residences (4),
schools (3), municipal services (5), public transport (13, 14), and
highways (11) are used to estimate the constructed levels of BG and BS.
63
-------
Levels of PH, PM, and PL workers are assigned directly to these goods
and services.
10. Determination of Levels of Terminal
Industry (2) and business goods (9) are used to estimate the construc-
ted levels of terminal. Central City Population Density and area are used
to estimate the number of terminal locations.
11. E stimation of Minimum Highway Levels
Levels of population (1), bus service (13), business goods (9), and
industry (2) are used to estimate a minimum number of levels of highway
service required in the region.
12. Determination of Industrial Employment
Population distribution by income (1) is compared with labor alloca-
tions made previously (3, 5, 8, 9, 13) to determine the number of units
remaining. These remaining levels are allocated to industry in proportion
to demand (2).
13. Determination of Bus Services
Central City Size and Density, and the Percentage of Workers Using
Public Transport are used to estimate the miles of public transit route.
Units of PM labor are assigned directly to meet the required levels. Rail
Rapid Transit is subtracted from the estimate, leaving an estimate of miles
of bus route.
14. Estimation of Overall Density Measures
Total Populations and Population Densities in central city and total
SMSA are used to estimate the parameters of an exponential distribution. This
distribution gives a first approximation of the density of industrial and
population development in the region.
64
-------
15. Determination of Numbers of Parcels
The parameters of the exponential distributions (14), the levels of
housing (4), and goods and services (8, 9, 5) will be used to estimate the
number of parcels (equivalent) to be allocated. These estimates give
guidance for limiting the size of subsequent computational operations.
16, Estimation of Expected Activity Rates
The Input City Pattern and the exponential distribution (14) are used
to estimate the expected intensity of industrial-commercial and residential
land use activity within parcels. Adjustments will be made to permit
location of access to parcels where none is provided in the input.
17. Determination of Industry Locations
Parcels containing basic industries will be Indicated_by_the_Us_er or
selected as those having the highest expected activity intensities (16), up
to the limit of number of parcels previously estimated (15). Levels will
be assigned in proportion to expected activity intensities (16)%.
18. Determination of Business Goods and Services Locations
Parcels adjacent to industry locations (17) and having high expected
activity rates (16) are selected as locations for business goods and services.
Levels are .assigned to approximate overall density estimates (14).
19. Determination of Low and High Income Residential Locations
Expected activity rates (16) for residential parcels are modified to
account for industry-commercial locations (17, 18), and for Mean Trip Length.
PL population will be distributed by housing type (6). PH parcels will, be
designated, and levels distributed by housing type using a different set of
rates. Population densities on parcels will be assigned up to the proportion of
the appropriate income class within the total population (1),.
65
-------
20. Determination of Locations for Personal Goods and Services
Location of the estimated number of parcels (15) will be made
adjacent to residential locations - PL and PH (19) - according to industrial
activity rates (16).
21. Determination of Locations of Schools, Municipal Services, and Utilities
Locational assignment will be made to va cant space in previously
assigned parcels (17, 18, 19, 20), or to immediately adjacent developable
parcels. Water Intake and Outflow are assigned by user.
22. Determination of Location of Medium Income Residential Parcels
PM levels are assigned, by housing type (6), to vacant land on
previously designated residential parcels (19), up to expected activity rates
(16, 19). Additional parcels are selected as necessary, up to the total
parcels estimated for the city (15).
23. Provision of Additional Required Access
The distribution of activities is displayed to the user, with an indication
of parcels requiring access. If input mileage is below estimated minimum
required (11) the deficiency is noted for user action.
24. Location of Bus Routes
Distributions of residences and jobs will be displayed, for user assign-
ment of bus routes, up to the previously estimated mileage (13).
25. Determination of Public Transit Fares and Salaries
City Budgets. Bus Company Type, and Median Income will be used to
set fares and salaries. Relations between bus and rail fares, when both
are applicable, will be set relative to national experience.
66
-------
26. Determination of Salaries
Payrolls, by activity, and Median Income will be used to set the
salary scale for industry and commerce.
27. Determination of Rents
Mean Rent and the Distribution of Income will determine rent levels.
Assignment of rents will be based upon distance from the center and density
of population (19, 22).
28. Determination of Prices
A standard price structure will be applied, adjusted to Median Income.
29. Determira tion of Assessed Values
Assessed values of parcels will be computed by capitalizing prices
and rents (27, 28).
30. Determination of Government Transfers
Taxes and unemployment payments will be set by application of
standard structures, adjusted to City Type, Tax Type Indicators, Unem-
ployment Rate, and Public Assistance Payments.
31. Determination of Zoning
Zoning will be set to be consistant with assigned activities (17, 18,
19, 20, 22).
32. Determination of Educational Levels
Educational levels will be determined by previously assigned rents
and income characteristics (19, 22, 27).
67
-------
APPENDIX C
THE GEM LOAD DECK
Following is a list of items in the GEM load deck, which are to be
treated in the Simulation City model. Appended to various items to be
treated are brief comments to indicate sources of information or
computational procedure. Acronyms and variable names correspond
to GEM usage, and the reader is referred to forthcoming EPA
reports for their discussion.
68
-------
Land Parcel Cards
12-14
Constructed Level of
Economic Activity
15-17
18-21
22-25
30-33
34-36
37-39
40-43
44-46
47-49
54-56
60-67
68-75
Zoning
Value Ratio
Maintenance Level
No PH Residing on
Parcel
No PM Residing on
Parcel
No PL Residing on
Parcel
Salary PH/$100 if
non resident. Rent
per space unit if
resident
Salary PM/$100 if
non resident. Rent
per space unit if
resident
Salary PL/$100 if
non resident. Rent
per space unit if
resident.
Level of Utilities
installed
Price/CU in $100
Assessed Value of
land/$100,000for
all parcel
a) Manufacturing levels from
Statistical Abstracts and Master
Sheets for Economic Sector
b) Residential levels from Statistical
Abstracts which give RA level. RB
and RC will be calculated from area
density
c) Pa, Ps levels from population
levels and housing levels
d) Ba, Bs levels from levels of
industry, Pa, Ps, residences, schools
municipal services, utilities, consistent
with activity.
Person levels are split by income type
and assigned to parcels using attrac-
tion indices for (PH, PM) and (PM, PL) /
Attraction indices are calculated from
city density, highway, rivers, undeveloped
land, industry location, mean trip
length, etc.
Salary for PM from median income
in Statistical Abstract. PH taken to
be 1 1/2 x median income, PL taken to
be 1/3 median income. Similarly,
rents are computed from statistical
Abstract median rent
Use utility units from Rent Master
sheets
Mean income, location population,
industry, Ba, Bs, Pa, PS.
Prices, rents, location
69
-------
77-78
80
1-5
7-8
10-11
If industry on parcel
Effluent treatment type
Treatment plant level
Parcel Good Map
Team letter of owner
Economic Activity
Code
e. g., MF, Ba, RA.
The activity is assigned to parcel
using attraction indices
Industry assigned first Ba, Bs.
RA, RB, RC by income type
Pa, Ps, schools, municipal services,
utilities.
Water Intake Plant and Undeveloped Land
8
1-2
3
11-15
16-17
18-20
School Cards
9 1-2
3
11-15
16-17
18-20
21-23
24-26
27-29
30-32
WA
Jurisdiction no
Parcel location
Level of intake
treatment plant (0 if
under land)
Percent of parcel
(developed and un-
developed) owned by
Water Authority
SC
Jurisdiction
Parcel Location
Level of School
From map
Constructed levels industry etc,
required level
Schools located centrally
Statistical Abstract students
School quality teachers
with school master sheets
school level
% owned by dept.
Value ratio
Maintenance level
PM teachers (in Pi's) See above
PH teachers (in Pi's)
70
-------
Terminal Cards
10
1-2
3
11-15
16-17
18-20
TM
Jurisdiction
Intersection coord.
Level
Use map to decide location
From levels of industry, etc and
using Master Sheets
% land used by ter- Determined by user
minal on each of four
parcels surrounding
intersection
Park and Public Institutional Land
11 1-2 PZ or PI
3 Jurisdiction
11-15 Parcel containing
parkland
16-20 Percent of land in
park or public insti-
tutional use
Municipal Services
12 1-2 MS
3 Jurisdiction
11-15 Parcel location
16-17 Level of plant
18-20
21-23
24-26
27-29
30-32
Road Cards
13 1-2
3
11-15
% land owned by
department
V-alue ratio
Maintenance level
PL workers
PM workers
RD
Jurisdiction
Intersection where
road segment begins
From map
From map
Located centrally
Levels of industry, etc,
give level of MS based on Master
Sheets
From Master Sheets
From Master Sheets
From map
71
-------
16- Direction
19 Road Type
23-25 Value ratio of road
segment
Undeveloped Highway Land
14 1-2 HR
3 Jurisdiction
11-15 Location
16-20 % parcel owned but
undeveloped
Municipal Services Salaries
16 1-5 PL salary in $100
6-10 PM salary in $100
School Salaries
17 1-5 PM salary
From map
Any land available after all
assignments?
6-10
11-15
Fraction of Median Income
Median Income from Statistical
Abstract
Median Income from Statistical
Abstracts
Factor x Median Income may modify
Government expenditure on education
16-20
Bus Salaries
18 1-5
6-10
PH salary
Middle class part-
time units for adult
education and employ-
ment
High class part-time
units for adult educa-
tion and employment
Salary offered by Bus Bus use* company type, city budget
company/$100 and mean income
Salary offered by
Rail company/$100
Bus fares, bus salary give rail salary
72
-------
Bonds
19 1-2
3
4-5
6-7
8
9-14
Taxes
20 1
Rents
23 1-5
6-10
Bus Routes
24 1-3
5
6-10
Department name
Jurisdiction
Interest rate in 1/10%
Remaining term
"0" if current bond (2
<2 yr.)
M 1 II
1" if capital bond
(25 yr.-)
Amount/$10, 000
Jurisdiction
Land
Building
Resident-Income
Employment Income
Resident Auto
Employment Auto
Personal Goods
% bid price for land
bid
% above value for
outside construction
Location
Rent/space unit
Level of service
1 for bus
Route No
1
Mean Income and Tax designator
provide general tax structure
(See Land Parcel Cards - 6)
City size, density, area miles of
bus route
Activity map
assigned bus route
73
-------
Detailed Route Invormation
1-5 Starting point
6-10 Location of turn of
end point
Rail Stations
25
1-5
6-10
RLSTA
Number stations on
board
Subway map
11-15 Location
Unused Rail Land
26 1-5 RLLND
6-10 Twice the no. of
parcels on this card
11-15 Parcel location
16-20 % of land owned by
RR department but
unused
Track Segments
27
1-5
6-10
11-15
16-20
Rail Routes
28 1-3
5
6-10
11-15
RLTRK
Intersection pair. Subway map or USAS map
which defines segment
Location of parcel Subway map or USAS map
over which any part of
segment is above
ground
Level of Service
"0" for rail
Route No.
No. of turns and/or
stops
Additional Cards for each Rotf e
1-5 Intersection of stop
or turn
74
-------
10 0 if stop or stop and
turn
1 if turn only
Ba/Bs Contracts
29 2-7 Cards per department
Bus/Rail Fares
31 5 1 for rail 2-bus Set by Median income, bus
company type, and city budget
6-10 Base fare in cents
16-20 Increment/mile
21-25 Value ratio of equip-
ment
26-30 Maintenance Level
Highway Maintenance Levels
32
Education Level
33 1-5 Parcel location
6-8 Education Level - Standard table for city types
high class
9-11 Education Level -
middle class
12-14 Education Level -
low class
Welfare Payment
34 1 Jurisdiction
2-6 Welfare Payment per Statistical Abstracts
unemployed worker in
$100
Prices for Outside Purchase
35 1-10 Price/CU for outside
Pa, Ps
11-20 Price/CU for outside
21-30 Price/MA for outside
water
31-40 Price/MA for outside
water for residents and
private utilities
75
-------
41-50 Price /ton solid waste
dump
51-60 Price EPU for EPU
shortage
61-65 Multiplier for cost to
SWC to provide level
2 service
Topographical Restrictions and Pre-Empt Land
36 1-2 Row No. (12-60)
6-8 Three column per From map
board
up to Square for given row
78-80 containing % square
undevelopable
Government Employment Locations
37 1-10 Rail
similarly BUS MS 1, SC 1, as with industry
Federal State Employers
38 1-5
6-10
11-15
16-20
21-25
26-30
31-35
Surface Water
39 1-5
6-10
11-13
14-15
16-20
Location
PL job openings
PM job openings
PH job openings
PL salary
PM salary
PH salary
Location
Volume in MaD
% land area
Rate of flow (par-
cels/day)
Next parcel water
flows into
Government salaries in
Statistical Abstracts
Map
Map
Map
76
-------
Lake Parcels
40 1-5 Location
6-10 Water Quality
Municipal Treatment Plant
44 1-5 Location
7-8 Outflow type
9 Level
Intake and Outflow Points
45 1-5 Location (point)
6-10 Location of UT plant
15 0 - intake 1 - outflow
Sampling Stations
46 1-5 Location
Map
From Map
From level of industry, etc, and
Mast er Sheet
From Map
Map
From Map
6 P = business point source
A = ambient
M = municipal point source
E = all three kinds
Water Prices
47 1-2 Jurisdiction
4-5 Activity type
6
7-10 Price/Ma.
Typical Rents and Salaries
48 1-5 PL rent
6-10 PM rent
11-15 PH rent
16-20 PL salary
21-2 5PM salary
26-SOPH salary
Standard price structures
PM taken as median rent from
Statistical Abstract - city type income
and U. S. trends used to give PL, PH
payrolls and employment by business,
gov't used to give salaries
77
-------
Solid Waste Co - Developed and Undeveloped Land - 51
1-5 Location Map
8-10 % land owned by SWC
12-14 Landfill %
15 Type
16-20 Tons solid waste to fill
site
23-25 Level of incinerator
28-30 Level of incinerator
air treatment
33 Type A air treatment
34 Type B air treatment
35 Type C air treatment
Power Co - 52
1-5 Location From Map
8-10 % land owned by PC
12 Type
13-15 Level of power plant Levels of industry and Master Sheets
give level of power plant
18-20 Level of cooling tower
23-25 Level of air treatment
28 Type A air treatment
29 Type B air treatment
30 Type C air treatment
33-35 % fuel low grade coal
38-40 % fuel high grade coal
43-45 % fuel low grade oil
48-50 % fuel high grade oil
53-55 % fuel natural gas
78
-------
Solid Waste Collection Levels - 53
1 Level of Collection
6-10 Location of parcel served
Power Co. Prices - 54
1 Jurisdiction Standard price structure
2-5 Normal price
6-10 Heavy user price
15-20 Max annual Env.
(normal price)
Solid Waste Co. Prices - 55
1 Jurisdiction Standard price structure
3-4 Activity type
5-10 Price
Solid Waste Co. Contracts and Basic Land - 56
2 Jurisdiction of SWC Standard price structure
6-10 Location of industry
11-20 Price/year
MS and SC Fuel Mix - 57
2 Jurisdiction
/
3-5 % MS fuel high grade
coal
6-8 % MS fuel high grade
oil
9-11 % MS fuel natural gas
12-14 %SC fuel high grade coal
15-17 % SC fuel high grade oil
18-20 % SC fuel natural gas
Basic Industry Fuel Mix and Air Treatment - 58
1-5 Location
etc.
79
-------
SELECTED WATER
RESOURCES ABSTRACTS
INPUT TRANSACTION FORM
1. Report No.
w
4. Title
Simulation City Approach for Preparation of Urban
Area Data Bases
Author(^)
Andrew C. Lemer
Or> n
Alan M. Voorhees & Associates, Inc.
Westgate Research Park
McLean, Virginia 22101
5. Report Date
6.
8. F /formi: j Organization
Report No.
12. Sr
J5.
/c > Org^Vatftw^Environmental Protection Agency
ry .'• ie<.
Environmental Protection Agency
Report Number EPA-600/5-74-001,
February 1974.
68-01-1805
;. Type •-1 Repo: t and
Period Covered
Final Report
16. Abstract
The basic hypotheses of the Simulation City Approach is that it is possible to
approximate the detailed input data base required for complex planning models,
given only a relatively gross description of a specific metropolitan area and
general knowledge of patterns of urban composition. This approximation is
accomplished at a substantial reduction in the costs associated with data prepara-
tion and thus planning models, at relatively little expense in terms of accuracy.
Accuracy is here judged in terms of the final decision to be made on the basis
of planning model analyses. To the extent that the hypothesis is valid in a
particular application, a decision-maker is given the opportunity to ask a range
of questions at substantially reduced cost and time expense.
The concept is first described in general terms, and then supported by a review
of theoretical and empirical studies which would be valuable in its realization.
The trade-off between cost and accuracy in modeling is then more explicitly
considered.
17a. Descriptors
Simulation, Metropolitan Data, Planning Models
17b. Identifiers
•«. COH-RRFi'.ld & G-:
19. Security Class.
CRepoi )
'0. S.? -itityd ;s.
(Page)
21. No. of
Pages
-3. Pr •:
Send To:
WATER RESOURCES SCIENTIFIC INFORMATION CENTER
U.S. DEPARTMENT OF THE INTERIOR
WASHINGTON. D. C. 2O24O
Andrew C. Lemer
Alan M. Voorhees & Associates. Inc.
* U. S. GOVERNMENT PRINTING OFFICE : 1874 731-934/341
------- |