The Effects of Nucleated Urban Growth Patterns on Transportation Energy Consumption - Aug 1980

Click HERE for graphic.

This document is disseminated under the sponsorship of the
Department of Transportation in the interest of information
exchange.  The United States Government assumes no liability for
its contents or use thereof.

4. Title and Subtitle :


5. Report Date: August, 1980

7. Author(s): William M. Weisel and Joseph L. Schofer

9. Performing Organization Name and Address:

   Department of Civil Engineering Northwestern University
   Evanston, Illinois 60201

11.   Contract or Grant No.: DOT-OS-50118

12.   Sponsoring Agency Name and Address:

      Office of University Research
      Research and Special Programs Administration
      U.S. Department of Transportation
      Washington, D.C. 20590

16.   Abstract:

This report continues the exploration of energy consumption un
urban passenger travel, land use, and transportation system
characteristics begun in the first volume issued under this
project, "The Impacts of Urban Transportation and Land Use Policies
on Transportation Energy Consumption," by Robert L. Peskin and
Joseph L. Schofer.  Findings are based on 88 policy experiments
using an integrated Lowry-type land use model incorporating mode
choice determination, network flow equilibrium and generalized cost
as the travel impedance measure.  The effects of decentralized
nucleated growth nodes, in the form of residential, retail, basic
employment and integrated land use clusters, were tested in detail. 
Results suggest that centralized growth is more energy-efficient
than decentralized development; further, off-center basic
employment nodes were least efficient, while similarly-located
population and retail clusters were more efficient than the former,
but not as attractive as centralized development.  Edge-clusters
are inclined to encourage energy-wasting urban sprawl.  Selected
TSM improvements and relaxed land use controls permitting natural,
centralized growth reduced energy consumption.  Guidelines for
encouraging more energy-efficient land use patterns are presented,
and likely obstacles associated with these guidelines are

                            EXECUTIVE SUMMARY
   This report describes a systematic exploration of the transpor-
tation energy requirements of urban travel associated with
different patterns of nucleated or clustered urban growth in a
medium size city, and with associated transportation system
improvements.  It was under-taken to determine which patterns of
clustered land development are most efficient from a transportation
energy perspective.
   The results are based on eighty-eight policy tests conducted
with an integrated land use and transportation simulation model,
which uses a Lowry-type land use model as its core.  The transit
share of the urban travel market is internally determined by this
model, as is the pattern of equilibrium network flows.  Generalized
cost, which is a linear combination of travel time and vehicle
operating costs, is used as the principle measure of travel
impedance.  The model incorporates activity density constraints and
location attractor variables to permit the simulation of a
realistic city in the size range 100,000 to 125,000 population.
   The model is based on an aggregate approach to location and
travel simulation; it functions with a hypothetical city form
having a welldefined outer edge which precludes direct observation
of urban sprawl beyond that boundary.  All locators and potential
development sites are treated as being homogeneous, and thus the
social and economic differences in location preferences and site
bidding power are necessarily ignored.
   The experimental program involved tests of the energy
consumption implications of four cluster types at various central
and decentralized


locations within the city's boundary.  The cluster types tested are
population, retail employment, basic or manufacturing employment,
and integrated (multiple land use) growth nodes.  For each of these
experiments, cluster type, size, and location with respect to the
center of the city are defined, and then the model-is run,
permitting its internal structure to simulate the location of
remaining, non-clustered activities, travel requirements, and
network equilibrium.  Tests are also conducted to explore the
efficacy, in terms of energy consumption, of a number of
transportation system management actions involving flow
facilitation and capacity improvement.  Measures of effectiveness
considered are total transportation energy consumption for
passenger travel, an aggregate measure of congestion on the
network, average trip length, the detailed location of all of the
activities simulated in the city at equilibrium, and aggregate
transit share of the travel market.
   Results indicate that centralized growth for all activity types
tested is more energy efficient than off center urban development. 
As clusters are moved further away from the center of the
hypothetical city, energy consumption tends to increase.  A
decrease in energy requirements is observed as clusters approach
the edge of the city'. but this is attributed to be an artifact of
the model, which does not permit growth to occur beyond the
predetermined boundary of the city.  It is expected that, in
reality, edge clusters would increase energy consumption because of
the resulting sprawl which they are likely to encourage.
   Of the decentralized cluster types tested, off-center basic em-
ployment nodes are found to be the least energy efficient,
apparently because they produce large-scale and wide-spread impacts
on location and travel patterns throughout the city.  Off-center
population and retail employment clusters are more energy efficient
than basic employment


clusters, but still less attractive from an energy perspective than
centralized growth.
   Certain kinds of transportation network improvements are found
to increase the energy efficiency of the hypothetical city.  These
can largely be interpreted as transportation system management
improvements involving congestion reduction and small scale
increases in-.capacity. Among the most attractive of such
improvements are the introduction of diagonal, radial links to the
rectangular grid network and capacity improvements which created
ring roads from existing arterial streets.  Such network
improvements reduce energy consumption in the hypothetical city
because they reduce network congestion and tend to encourage urban
land development patterns which take advantage of the service
offered by the improvements themselves.  Some small scale network
improvements intended to serve growth clusters are found to be
disadvantages from an energy perspective.
   Based on these interpretations of the results of the policy
experiments, the desirability of encouraging centralized growth in
urban areas is clear.  This may be accomplished by diverting new
growth from offcenter locations toward the central area, perhaps
through systematic zoning changes, the careful evaluation of
building permit request, the creation of tax incentives, and public
sector actions to maintain..and enhance the quality of central city
   The latter may include modest transportation improvements,
actions to control crime, and efforts to improve central city
schools.  The prospect for increased development in central areas
are relatively good considering the amount of vacant land, and the
number of obsolete structures, which are apparent in and near the
downtown areas of many American cities.


   When considering incentives for centralized growth, and
disincentives for off-centered growth, attention needs to be
devoted to the potentially inequitable impacts on low-income
residents of central cities, who might be priced out of the land
market if more affluent individuals and activities were attracted
toward the downtown.  This suggests the desirability of enhancing
central area housing opportunities for low-income residents.
   The potential implications of urban sprawl, in energy terms,
appear to be quite unattractive.  This suggests the desirability
for local governments to evaluate carefully proposals which support
a trend toward sprawl.
   One of the major obstacles toward achieving more centralized
urban development patterns is the fact that most urban regions are
made up of a multiplicity of governments, each financed through
taxes dependent on the level of local development.  The result
tends to be a strong competitive atmosphere in which central and
suburban governments work against each other to secure new
developments.  This suggests the potential desirability of
developing integrated forms of municipal financing to reduce or
eliminate the monetary incentives for encouraging urban sprawl as a
public policy.
   The spatial pattern of a city appears to be a major influence on
its consumption of energy for transportation.  Communities do have
some ability to control their land use pattern, and appropriate and
effective use of this ability appears to offer the possibility of
producing reductions in transportation and energy consumption in
the range of 5 to 10 percent within the next several decades,
independent of the nature of future improvements in transportation


                            TABLE OF CONTENTS

   LIST OF FIGURES                                                  viii
   LIST OF TABLES                                                     xi


  I   INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . .1

   A. The Problem . . . . . . . . . . . . . . . . . . . . . . . . . . .1
   B. Energy Conservation in Transportation . . . . . . . . . . . . . .2
   C. Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4
   D. Use of the Present Research . . . . . . . . . . . . . . . . . . .5
   E. Approach. . . . . . . . . . . . . . . . . . . . . . . . . . . . .7
   F. The Report. . . . . . . . . . . . . . . . . . . . . . . . . . . .8
   References to Chapter I. . . . . . . . . . . . . . . . . . . . . . .9

 II   BACKGROUND. . . . . . . . . . . . . . . . . . . . . . . . . . . 11

   A. Genesis: Edwards's Model. . . . . . . . . . . . . . . . . . . . 11
   B. MOD2: Focus on Policy Evaluation. . . . . . . . . . . . . . . . 18
   C. Policy Tests Applicable to Existing Cities:                     23
         The Work of Peskin . . . . . . . . . . . . . . . . . . . . . . 
   D. Other Contemporary Research . . . . . . . . . . . . . . . . . . 29
   References to Chapter II . . . . . . . . . . . . . . . . . . . . . 33

      EXPERIMENTAL PROGRAM. . . . . . . . . . . . . . . . . . . . . . 35
   A.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 35
   B.  Revisions to MOD3 Program. . . . . . . . . . . . . . . . . . . 35
      1. Conversion to CDC 6600 . . . . . . . . . . . . . . . . . . . 35
      2. Making the Experimental City More Realistic. . . . . . . . . 37


                      TABLE OF CONTENTS (continued)
Chapter                                                            Page 

      3. Determination of Appropriate Cluster Sizes . . . . . . . . . 46
   C. Caveats                                                         48
   D. Method of Analysis. . . . . . . . . . . . . . . . . . . . . . . 52
   E. Experimental Program. . . . . . . . . . . . . . . . . . . . . . 53
   References to Chapter III. . . . . . . . . . . . . . . . . . . . . 60

 IV   RESULTS AND ANALYSIS. . . . . . . . . . . . . . . . . . . . . . 62
   A. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . 62
   B. Base of Standard Incremental Runs in Square and
         Related Cities . . . . . . . . . . . . . . . . . . . . . . . 63
   C. Effects of Network Changes. . . . . . . . . . . . . . . . . . . 65
      1. Axial Freeways, Diagonal Freeways, and
            Rotated Cities. . . . . . . . . . . . . . . . . . . . . . 65
      2. Effects of Ring Roads. . . . . . . . . . . . . . . . . . . . 70
      3. Effects of Minor Capacity Improvements . . . . . . . . . . . 73
   D. Effects of Different Cluster Types. . . . . . . . . . . . . . . 77
   E. Effects of Cluster Location . . . . . . . . . . . . . . . . . . 90
   F. Effect of Density Constraints . . . . . . . . . . . . . . . . . 93
   G. Integrated Cluster Development. . . . . . . . . . . . . . . . .100
   H. Cluster Size. . . . . . . . . . . . . . . . . . . . . . . . . .102
   I. Comparative Roles-Transportation Improvement
         vs. Land Use Controls. . . . . . . . . . . . . . . . . . . .108
   References to Chapter 4. . . . . . . . . . . . . . . . . . . . . .115

  V   CONCLUSIONS AND GUIDELINES. . . . . . . . . . . . . . . . . . .116
   A. Limitations of the Research Approach. . . . . . . . . . . . . .116


                      TABLE OF CONTENTS (continued)

Chapter                                                            Page 

   B. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . .119
   C. Guidelines for Improving the Transportation
         Energy Efficiency of Cities. . . . . . . . . . . . . . . . .124
   Reference to Chapter 5 . . . . . . . . . . . . . . . . . . . . . .129


   A  RESULTS OF THE EXPERIMENTS. . . . . . . . . . . . . . . . . . .130
   B  PROGRAM LISTING . . . . . . . . . . . . . . . . . . . . . . . .134


                             LIST OF FIGURES
  Figure                                                           Page 

   2-1   Edwards's Cities . . . . . . . . . . . . . . . . . . . . . . 13
   2-2   Edwards's Model. . . . . . . . . . . . . . . . . . . . . . . 14
   2-3   Operation of Energy Simulation Model Revised by
         Bowman, et al. . . . . . . . . . . . . . . . . . . . . . . . 20
   2-4   The "Layering" Approach to Urban Growth. . . . . . . . . . . 25
   2-5   Experimental Process . . . . . . . . . . . . . . . . . . . . 25
   2-6   Peskin's Cities. . . . . . . . . . . . . . . . . . . . . . . 27
   3-1   Core of MOD3-the Transportation/Land Use Feedback. . . . . . 36
   3-2   New Concentric Ring "Diamond" City . . . . . . . . . . . . . 39
   3-3   Highway Network. . . . . . . . . . . . . . . . . . . . . . . 41
   3-4   Transit Network. . . . . . . . . . . . . . . . . . . . . . . 42
   3-5   Adjustment of Density Gradient . . . . . . . . . . . . . . . 45
   3-6   Distribution of Basic Employment . . . . . . . . . . . . . . 47
   3-7   Analysis Framework . . . . . . . . . . . . . . . . . . . . . 54
   3-8   Cluster Development Zones. . . . . . . . . . . . . . . . . . 56
   4-1   Network Changes to Standard Incremental Run. . . . . . . . . 66
   4-2   Ring Roads with Standard Incremental Run . . . . . . . . . . 71
   4-3   Results of Network Structure Experiments . . . . . . . . . . 74
   4-4   Total Energy of Experiments with Basic Employment
         Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . 75
   4-5   VMT of Experiments with Basic Employment Clusters. . . . . . 76
   4-6   Highway Congestion Index of Experiments with Basic
         Employment Clusters. . . . . . . . . . . . . . . . . . . . . 76


                       LIST OF FIGURES (continued)
 Figure                                                            Page 

   4-7   Total Energy of Experiments with Three Cluster Types . . . . 80
   4-8   VMT of Experiments with Three Cluster Types. . . . . . . . . 81
   4-9   Highway Congestion Index of Experiments with Three
         Cluster Types. . . . . . . . . . . . . . . . . . . . . . . . 81
   4-10  Distribution of Incremental Non-Clustered Basic
         Employment . . . . . . . . . . . . . . . . . . . . . . . . . 82
   4-11  Population and Service Employment Changes with Service
         Employment Cluster in Zone 36. . . . . . . . . . . . . . . . 86
   4-12  Population and Service Employment Changes with Population 
         Cluster in Zone 36 . . . . . . . . . . . . . . . . . . . . . 88
   4-13  Population and Service Employment Changes with Basic
         Employment Cluster in Zone 36. . . . . . . . . . . . . . . . 89
   4-14  Energy/Cluster Location Relationship with and without
         Control of Sprawl. . . . . . . . . . . . . . . . . . . . . . 94
   4-15  Population and Service Employment Changes with Basic
         Employment in Zone Z . . . . . . . . . . . . . . . . . . . . 96
   4-16  Density Gradients for Constrained and Unconstrained
         Control Area Development . . . . . . . . . . . . . . . . . . 98
   4-17  Energy Consumption for Clusters with and without
         Density Constraints. . . . . . . . . . . . . . . . . . . . .101
   4-18  Total Energy of Integrated and Cluster Type
         Experiments. . . . . . . . . . . . . . . . . . . . . . . . .104
   4-19  WIT of Integrated and Cluster Type Experiments . . . . . . .105


                       LIST OF FIGURES (continued)
 Figure                                                            Page 
   4-20  Highway Congestion Index of Integrated and Cluster
         Type Experiments . . . . . . . . . . . . . . . . . . . . . .105
   4-21  Residential Cluster Size and Energy Consumption. . . . . . .107
   4-22  Total Energy of Experiments with Ring Roads and
         Integrated Clusters. . . . . . . . . . . . . . . . . . . . .109
   4-23  VMT of Experiments with Ring Roads and Integrated
         Clusters . . . . . . . . . . . . . . . . . . . . . . . . . .110
   4-24  Highway Congestion Index of Experiments with Ring
         Roads and Integrated Clusters. . . . . . . . . . . . . . . .110


                             LIST OF TABLES
   Table                                                           Page 

   2-1   Urban Structural Variables . . . . . . . . . . . . . . . . . 15

   3-1   Link Classification. . . . . . . . . . . . . . . . . . . . . 43
   3-2   Experimental Program . . . . . . . . . . . . . . . . . . . . 58

   4-1   Comparison of S.I.R. and Base Runs . . . . . . . . . . . . . 64
   4-2   Axial and Diagonal Network Additions to Square and
         Rotated City (% Changes From S.I.R.) . . . . . . . . . . . . 67
   4-3   Comparison of Secondary Measures for S.I.R., Axial
         Freeways, Diagonal Freeways, and Diagonal Arterial
         Experiments. . . . . . . . . . . . . . . . . . . . . . . . . 69
   4-4   Average Highway Congestion in S.I.R. for Link
         Classification A to D. . . . . . . . . . . . . . . . . . . . 72
   4-5   Primary Results for Ring Road Improvements . . . . . . . . . 72
   4-6   Average Work Trip Length With and Without Selected
         Highway Improvements . . . . . . . . . . . . . . . . . . . . 78
   4-7   Population Clusters. . . . . . . . . . . . . . . . . . . . . 83
   4-8   Service Employment Clusters. . . . . . . . . . . . . . . . . 83
   4-9   Transit Ridership and Trip Length in Basic Employment,
         Population, and Service Employment Clusters. . . . . . . . . 91
   4-10  Comparison of Constrained and Unconstrained Standard
         Incremental Runs . . . . . . . . . . . . . . . . . . . . . . 99
   4-11  Integrated Clusters. . . . . . . . . . . . . . . . . . . . .103
   4-12  Comparison of Transit Ridership and Trip Lengths in
         Integrated Clusters with Other Cluster Types . . . . . . . .106


                                CHAPTER I

A. The Problem

   The availability and price of energy resources have, within the
short span of a decade, established themselves at the forefront of
the collective consciousness of this nation.  Current and almost
certain future shortages of oil threaten the internal fibre of
American society, as well as the security of the nation itself. 
While the debate rages as to whether the periodic oil shortfalls
are real or merely the artificial creation of the petroleum
industry, there is no question that the American people are
consuming the fixed petroleum resource at a high rate; consumption
continues to grow as manifested in continually growing demand for
gasoline.  In 1977, the transportation sector consumed more than 3
1/2% more energy (97% of that is 'petroleum-based) than in the
previous year, and 24.6% more than in 1970. (Shanka, 1979, p. 2-
10).  Contrasted against a world of finite resources and dwindling
oil reserves, the continued ability to satisfy this demand seems in
jeopardy.  Recent lines at the gas umps through the country clearly
signify the end of the era of cheap, plentiful fuel.  To lessen the
severity of the repercussions of this apparently long-term trauma,
the country must find ways of increasing petroleum supply and/or
reducing demand.  It is the second of these-the reduction in
demand-to which this research addresses itself.



B. Energy Conservation in Transportation

   More specifically, this work concentrates on exploring a
particular opportunity for reduction in the demand of
transportation energy.  Transportation, according to 1977 figures,
is the single largest user of petroleum, accounting for roughly 54%
of this nation's consumption Shanka, 1979, p. 2-10).  Logically
then, it may also be an area-with great potential for energy
savings through improved efficiency of use, technology, and
management of available resources.
   What are the potential areas for conservation of energy in
transportation?  First of all, the improvement of traffic
operations through small-scale Transportation System Management
actions (TSM; see, for instance, INTERPLAN, 1977 or Arnold, 1978)
can be quite effective in reducing energy consumption.  These
include a wide range of actions such as improved signalization or
channelization, selective parking regulations, or the introduction
of bus priority or high occupancy vehicle (HOV) lanes to encourage
transit ridership and carpooling.  Actions such as these tend to be
attractive to decision-makers since, in many cases, sizeable
returns are achievable in a short time frame with important, while
longer-term benefits may be passed over.  This "pennywise, dollar-
foolish" approach can, therefore, have grave repercussions for the
nation in the long run. (Schofer and Stopher, 1979).
   A second potential source for energy conservation in transporta-
tion is in the area of technological advancement, particularly with
regard to the internal combustion engine.  Important savings have


already been realized through mandated and market-forced increases
in the average fuel economy of the nation's fleet of automobiles. 
At the same time, considerable research is underway toward
developing new, more efficient automobile engines such as the
electric or hybrid vehicles, the gas turbine engine, or the
stirling cycle engine (see Shanka, 1979; Hittman, 1974).  Since,
however, the practical, widespread applicability of any of these
technologies is still far in the future, their potential for energy
savings can only be realized over a longer time horizon.  Hence,
both in the short-term and in the longterm, technological
improvements in the auto hold great potential for energy
   A third area of potential energy conservation in
transportation -- one which has perhaps received less attention
than the other two-involves the relationship of transportation to
the urban system in which it operates.  Past evidence has indicated
that urban form or spatial structure-the arrangement of different
land uses within the urban area--is a major factor in determining
energy consumption.  The implication here is that by controlling
and directing the shape of a city and the distribution of
activities within it, one can have an impact on such
characteristics of the transportation system as accessibility, trip
lengths, congestion, environmental impacts and-consequently energy
   Work in this area with respect to transportation-efficient new
towns has gone on in Europe for many years (Stockholm, 1972).  In
this country, interest in the subject has been more recent.  Work
such as that by Edwards & Schofer (1976), Peskin & Schofer (1977),
T. Owen Carroll, et al. (1978) and Clark (1976) have primarily
attempted to


identify the energy consumption characteristics of various
hypothetical or generalized urban forms.  While these efforts have
met with varied levels of success, they have at least indicated an
apparent potential for energy conservation with certain generic
urban forms, principally those that concentrate activities heavily
in a single central area or around several "polynucleated" centers. 
Although these results are helpful in defining idealized end-states
for urban areas, they provide only limited guidance-to regional
planners and municipal authorities in dealing with the type of
immediate problems and decisionmaking opportunities which they face
on a daily basis.

C. Purpose

   It is neither possible nor desirable (from an energy standpoint)
to abandon our most energy inefficient cities and to start again
with a clean slate.  Nonetheless, land use development continues to
occur and spatial attributes of cities change, albeit slowly.  The
challenge for planners is to learn to cope with existing cities and
to develop strategies for continually moving, through day-to-day
decisions as well as longer term policies, toward more desirable
endstates.  Once such desirable end-states have been defined, it
will be necessary to derive the set of most appropriate actions
which can be taken today to reach them.  As stated above, it is
this point which previous efforts have not addressed.  While they
suggest an optimal final state or form for cities, they do not
provide the planner with sufficient clues as to how to achieve it. 
Only Peskin (1977) begins to deal with policy questions in such
areas as highway and transit network improvements, pricing, and
land use control.


   It is the purpose of this research to begin to fill the gap in
the planning methodology between the desired land use/energy end-
state and the immediate issues facing local officials.  Toward
reaching this aim, this research, using simulation techniques,
explores the energy consumption implications of shifts in the locus
of growth in a hypothetical, yet realistic medium-sized city.  More
specifically, various growth nuclei (e.g., residential, service
employment, and basic employment activities) are simulated in both
central and noncentral locations and their energy use
characteristics are scrutinized.  Using the results of these
experiments, a series of guidelines are developed to serve as aids
to local planners and decision-makers faced with the task of
evaluating the constant flow of land use-change proposals and
opportunities.  In this way, it may be possible to provide present
land use decision-making, in both the public and private sector,
with a framework designed to produce long-term energy savings.

   D. Use of the Present Research

   With the legislative and regulatory powers which the public
sector, on different levels, possesses, government is
unquestionably a primary actor in the area of land use management. 
None will contest that decisions are often made in response to
valid and varying community and market pressures and do not
necessarily represent the optimal choice in terms of a single
public policy goal such as energy conservation; nevertheless, local
governments often play a major role, and in many cases have the
final permissive authority, over major land development and change
decisions within their jurisdiction.  Hence it makes sense for the
results of this research to be aimed at this body


of decision-makers.
   Planners and decision-makers concerned with growth management
should find the guidelines useful on several levels.  First, they
provide a framework within which to consider such common requests
as zoning variances, building permits, driveway cuts, public
financing, etc., which require immediate decisions.  Clearly, many
factors must be taken into account by local governments when making
such decisions, including the public welfare, effect on the tax
base, public infrastructure requirements, and environmental
impacts.  Guidelines of the sort developed in this research provide
the tool for the inclusion of another important factor--energy
consumption in urban passenger transportation in this list of
   On a second level, the ideas for energy-wise decision-making
developed in this research are helpful in managing growth through
public-sector-initiated actions.  A zoning change is one such
action that can be used in limiting (or encouraging) the type and
level of development in various locations.  Urban renewal or other
publicly subsidized activities are also an effective tool for
controlling development through direct public sector involvement. 
Another tool which can be effectively used to control sprawl is the
annexation of lands previously outside of the local jurisdiction. 
These lands can then be zoned for park land or other uses which
help to contain the spread of the city-or for commercial and
residential activities, to encourage another growth nucleus. 
Finally, infrastructure investment (e.g., transportation, water
supply, waste water treatment facilities) can be selected to
encourage or discourage particular development


   To the extent that research can show that certain patterns of
public sector actions produce more energy efficient land use
patterns, it is appropriate for local governments to consider using
these actions as opportunities for energy conservation.  Since
decisions which affect localized land use patterns in modest ways
are made quite frequently by local governments, ignoring such
opportunities amounts to throwing away an opportunity to conserve
energy in the long run. (See Dawson, 1977.)

E. Approach

   The approach adopted here is to test alternative urban forms
using digital computer simulation with an integrated, aggregate
model assembled from various separate models representing aspects
of the urban development and travel behavior processes.  No new
models are developed; rather, existing models are assembled into a
structure with a reasonable probability of providing realistic
estimates of general trends in energy conservation in urban
passenger travel.  As will be repeated several times throughout
this report, the numerical outputs cannot be precisely applied to
any particular city, but should indicate relative differences in
energy consumption due to different urban growth and transportation
   The results should prove helpful to local planners and deciSion-
makers by offering general guidance regarding the passenger
transportation energy implications of development options. 
Consideration of detailed, local consequences of specific actions
can only be accomplished through more sophisticated and locally-
specific methods.

The results presented here should assist local analysts in pursuing
such methods.

   F. The Report

   In the report that follows, Chapter 2 provides a brief overview.
of the three research efforts to which this present effort is
directly tied, as well as a brief discussion of other related work. 
Chapter 3 presents the significant changes and improvements from
the simulation methodology used by Peskin & Schofer, the antecedent
effort in this series (Peskin & Schofer, 1977).  This is followed,
in Chapter 3, by a discussion of the model's limitations and the
philosophy for interpretation of these results, as well as a brief
description of the experimental program.  Chapter 4 contains the
analysis of the results of the 88 simulation experiments conducted. 
Chapter 5 summarizes the conclusions reached in the analyses and
consolidates these conclusions into a set of guidelines designed
for use by planners and decisionmakers in making short-term land
use decisions with positive longterm energy impacts.


References to Chapter I

   1. D. B. Shanka, ed., Transportation Energy Conservation Data
      Book, Oak Ridge National Laboratory, Oak Ridge, Tenn.,
      February 1979.

   2. INTERPLAN Corporation, Transportation System Management:
      State of the Art, Urban Mass Transportation Administration
      Report No. RI-06-0008, U.S. Department of Transportation,
      February 1977.

   3. Eugene D. Arnold, Jr., Opportunities for Energy Conservation
      in Transportation Planning and Systems Management', Virginia
      Highway and Transportation Research Council Report VHTRC 79-
      R24, in cooperation with the U.S. Department of
      Transportation, Federal Highway Administration.

   4. Joseph L. Schofer and Peter R. Stopher, "Specifications for a
      New Long-Range Urban Transportation Planning Process,"
      Transportation, Vol. 8, 1979, pp. 199-218.

   5. Douglas G. Harvey and W. Robert Menchen, The Automobile,
      Energy, and the Environment, Hittman Associates, Inc.,
      Columbia, Md., March 1974.

   6. Stockholm Urban Environment, Stockholm City Building
      Department, Stockholm, Sweden, 1972.

   7. Jerry L. Edwards and Joseph L. Schofer, "Relationships
      Between Transportation Energy Consumption and Urban
      Structure: Results of Simulation Studies," Transportation
      Research Record 599, Transportation Research Board,
      Washington, D.C., 1976.


   8. Robert L. Peskin and Joseph L. Schofer, The Impacts of Urban
      Transportation and Land Use Policies on Transportation Energy
      Consumption, Report No. DOT-05-50118, U.S. Department of
      Transportation, Office of University Research, April 1977.

   9. T. Owen Carroll, Robert Nathans, and Phillip F. Palmedo, Land
      Use Energy Utilization, Final Report, Brookhaven National
      Laboratory/State University of New York Land Use-Energy
      Utilization Project for the Office of Conservation and
      Environment, Federal Energy Administration, BNL Report 50635,
      June 1977.

   10.   J. W. Clark, The Effect of Urban Form on Travel Demand and
         Transportation Energy Consumption: Theoretical
         Considerations, unpublished doctoral dissertation,
         Department of Civil Engineering, University of Washington,
         Seattle, Washington, 1976.

   11.   Grace Dawson, No Little Plans, The Urban Institute,
         Washington, D.C., 1977.

                               CHAPTER II


   The groundwork for the present research was established primar-
ily by three previous research efforts at Northwestern University. 
In the original research, Edwards (1975) developed the basic core
of the simulation model in order to investigate the relationships
between transportation energy use and urban form.  Building on this
foundation, Bowman, et al. (1976), contributed important
improvements to the simulation model resulting in increased realism
and increased sensitivity to potentially energy conserving urban
development policies.  Finally, Peskin and Schofer (1977) conducted
an extensive experimental program to explore a broad range of
policy changes concerning urban growth, transportation networks,
and transportation pricing." (Peskin, p. 1) The following sections
explore each of these studies in greater detail in terms of their
objectives, methodology, findings, and limitations.  This review
draws heavily on the summary presented by Peskin (see Chapter 2 of
Peskin, 1977).  The reader is referred there and to the original
works for further details.

   A. Genesis: Edwards' Model

   The work of J. L. Edwards was based on the perception, largely
in response to the 1973-74 Arab oil embargo, of a need to explore
not only the short range solutions to this energy shortage, but
also the



long-range impacts of-urban growth policies on energy consumption. 
This required the development of a better understanding of
fundamental relationships between urban spatial structure, urban
transportation networks, and energy consumption in urban passenger
travel.  To accomplish this task, Edwards sought to develop a model
with which he could simulate location and travel activity of urban
dwellers in various urban forms.  The urban forms that he developed
were several hypothetical but realistic urban structures, which he
referred to as the concentric ring, polynucleated, cruciform, and
linear urban forms (Fig. 2-1).
   The overall structure of Edwards' model is shown in Figure 2-2. 
At the core of the model lies a Lowry-type land use model which
"allocates commercial and residential land uses based on the
exogenously specified distribution of basic employment and the
free-flow travel times between all zone pairs." (Peskin, 1977, p.
13) A full discussion of the original Lowry model and its
successors can be found in Chapter 3 of Edwards' report (Edwards,
1975).  Of course, before the Lowry model can be applied, the test
city must be defined through a set of attributes, some of which are
fixed over all tests, while others vary.  These constants and
variables are presented in Table 2-1.
   Application of the Lowry model produces simultaneously a
distribution of land uses and a description of the daily person
trips needed to connect these different uses (work trips and
service trips).  Non-home based and social trips were subsequently
derived from the land use pattern through the use of common
gravity-type trip distribution models, supported by trip generation
estimates from previous studies.  All of these trips were then
allocated to transit or auto based on

Click HERE for graphic.

Click HERE for graphic.

                      Figure 2-2.  Edwards's Model


                                TABLE 2-1

                       URBAN STRUCTURAL VARIABLES

   Activity Variables                    Interaction Variables
   (Fixed Attributes of                  (Varied Over the Series
   Case Study City)                          of Urban Designs)

Total Population                         Urban Form

Employment, by Category                  Transportation Network, for
                                         Each Mode

Labor Force Participation Rate           Network Modes (Technologies)

Population Serving Ratios *              Network Levels of Service

Floorspace Requirements for              Mode Choice, by Trip Type
   Employee, by Type

Interzonal Friction Factors, by
   Trip Type**

Auto Occupancy, by Trip Type

Trip Rates per Capita, by
   Trip Type

 *  Defined to be the number of service workers by type per capita.
**  Defined in the context of the gravity trip distribution model
(Source: Edwards, 1975, 21]



an externally specified mode share fraction.  Once the trips were,
divided between auto and transit, each trip was then assigned to
its particular route within the network, using an all-or-nothing,
non-capacitated assignment algorithm.  Consumption coefficients
were then applied to determine the total auto and transit energy
consumption, the sum of which represented the estimated total daily
passenger transportation energy consumption for the particular
hypothetical city.
   It should be noted that additional measures, referred to as
accessibility measures,' were also derived by Edwards from the
results of the Lowry model.  Edwards pointed out that "while
recognizing the paramount need to identify the transportation
energy required of alternative urban structures, it is also
necessary to characterize the general nature of the proximity to
important activities which the residents of each structure
experience." (Edwards, 1975, p. 9) As might be expected, Edwards
finds that energy intensity increases with increased accessibility,
"a reflection of the fact that accessibility improves as the
dependence on the automobile and level of service on the highway
network increases." (Peskin, 1977, p. 13)
   Edwards conducted 37 experiments in which he explored the effect
(on energy consumption and accessibility) of varying such factors
as urban shape (concentric ring, linear, polynucleated, or
cruciform), the extent of central city concentration, and the
nature of the transportation network.  Generally, the compact urban
forms--linear, polynucleated, and cruciform-proved to have high
levels of energy efficiency but, at the same time, low levels of
regional accessibility to population.  For each particular urban
form, it was found that energy intensities


increased with:

   -  the expansiveness of land covered by development;

   -  the spread of population and employment, as measured by the
      second moments of their spatial distributions;

   -  the predominance of the automobile in the transportation
      system; and

   -  the level of congestion ("stop and delay") on the highway

   In many ways, the work done by Edwards set the pattern for all
of the research work which has followed.  Most importantly, it
established the Lowry-type land use model as the heart of the
simulation program.  While, like all aggregate models which attempt
to simulate an overwhelmingly complex urban system, it is in many
ways unrealistic, its strengths result from its "minimal data
requirements and (its] strong, intuitive causal structure
permitting the inclusion of many modifications." (Peskin, 1977, p.
14).  It is certainly still among the best methodologies available
for this type of land use modelling.  In addition, the decision by
Edwards to develop hypothetical cities rather than to model
existing American cities made it possible to avoid many of the
pitfalls inherent in gathering the data that would be necessary for
such modelling.  Realizing the wisdom of this choice, the practice
of working with hypothetical cities has been maintained throughout
all of the subsequent research efforts in this series.  Related to
this, the decision to confine the study to small-to-medium size
cities, a practice maintained here, has produced considerable
savings in terms of computational time (hence, cost) and
difficulty.  In short, it has made


the problem manageable.
   it is important to mention one additional aspect of Edwards'
work that has been critical in providing a framework within which
to assess and apply the products of his and all of the subsequent
research work.  His investigation of the important conflict between
two desirable ends, energy conservation and a high level of
accessibility, serves to caution those wishing to operationalize
his findings that any attempt to achieve energy conservation
through land use direction may be successful only At the expense of
other factors which the public generally perceives as important. 
Thus, any attempt at energy conservation will need to be tempered
by a concern for these other factors.  This subject will be
explored further in Chapter 4.
   While it is clear from this discussion that Edwards introduced
many important innovations toward developing a model with which to
better understand the relationship of urban form and energy
conservation, his work also had serious limitations.  Most of these
were related to assumptions and simplifications contained in the
model, especially pertaining to mode split and network assignment,
which resulted in a reduction of policy sensitivity and realism. 
It is to these limitations that.Bowman, et al., addressed their

   B. MOD2: Focus on Policy Evaluation

   With a goal of incorporating into Edwards' model increased
sensitivity co various energy conservation policies, Bowman, et
al., defined four distinct objectives:

   1) to identify relevant policies which should be incorporated in
      the design of any transportation model;


   2) to identify and incorporate the changes to a core model
      necessary to build the MOD2 (second generation) simulation

   3) to test the validity of the model; and

   4) to use the improved model package to test policies. (Bowman,
      et al., 1976, pp. 6-7)

   In order to accomplish the first objective, responses were
solicited from selected planning agencies.  A questionnaire was
developed which "requested information on policies being considered
to promote energy efficient land use patterns, to encourage more
efficient modes of travel, or to reduce congestion." (Bowman, et
al., 1976, p. 6) From the responses of 27 planning agencies,
several key policy areas and issues were identified for inclusion
in the MOD2 package.
   The major contribution made by Bowman, et al., came in the
fulfillment of the second objective.  First of all, in order to
increase the policy sensitivity, it was deemed necessary to improve
the model in general by improving the relationship between
transportation and land use. (A flow-chart of the revised model is
shown in Fig. 2-3.) Toward this aim, a considerable improvement was
made in the method for estimating modal split.  Rather than
defining the mode split externally, inclusion of a binary logit
mode split model developed and calibrated by Charles River
Associates enabled MOD2 to relate mode split to appropriate
transportation service variables.
   A second important improvement to the model involved the network
assignment procedure.  MOD2 replaced Edwards' non-capacitated, all-
or-nothing algorithm with an iterative equilibrium assignment

Click HERE for graphic.

            Figure 2-3.  Operation of Energy, Simulation Model
                         Revised by Bowman et al. (Source:  Peskin,
                         1977, p. 16]



algorithm (see LeBlanc, 1973).  The new algorithm directly
considered the critical factor of congestion by increasing the time
on any route as the volume on that route increases.  The iterative
aspect of the algorithm has a dual purpose.  First of all, it would
be far too cumbersome and time-consuming to continuously recompute
travel times for every route after the addition of each trip. 
Rather, trips are added in increments of decreasing size (6
increments in this research) and travel times are recomputed only
after each increment is assigned.  This use of increments also
provides the opportunity for recomputation of the modal split over
all of the remaining trips using the revised travel times for the
highway and transit networks.  Because these capacity restraint
iterations took place within the land use simulation process, the
equilibrium assignment algorithm greatly increases the responsive-
ness of land use decisions to the conditions on the surrounding
transportation network.
   A new transit simulation algorithm was also developed, "because
the original algorithm in the core model was incapable of
accommodating the complex characteristics of a real transit
network." (Bowman, et al., 1976, p. 46) Many features were
introduced through the new algorithm that allowed greater
flexibility in experimenting with potential energysaving transit
policies.  Further improvements to the land use attraction balance
factors and energy consumption estimation models added increased
realism to the overall simulation package.
   The improvements introduced in MOD2 act to give the model
greater flexibility, increased realism, and a stronger relationship
in the simulation process between the estimation of the land use


and the structure and performance of the transportation network. 
Because of this, it is, when compared with Edwards' model, "a much
more accurate and versatile tool for use in simulating the impacts
of policy changes in transportation energy consumption." (Bowman,
et al., 1976, p. 67)
   To complete the fulfillment of the second objective it was
necessary to include in the model those variables needed to
represent the important policies or cases that had been identified
in the first phase of the research.  Some of the variables needed
to test policies, such as density constraints or attractor
variables, had already been included in the original program as
inputs to the Lowry model.  Others had been added as a result of
the program modifications described above.  For instance, the use
of the equilibrium assignment algorithm allowed the user to explore
the effects of congestion because the model now accounted for
congested flow characteristics of links, thus becoming more real-
istically responsive to changes such as street widening or
signalization.  Still other variables, such as parking costs or
gasoline cost, had to be added individually into the program.
   The accomplishment of the third of the four objectives,
validation, involved primarily the -testing of the model's ability
to simulate an actual city.  It was clearly felt that such a
validation was necessary in order to build confidence in the model
as a policy evaluation tool.  The main component of this phase of
the research involved the use of data from a 1964 transportation
study in Amarillo,, Texas, a city similar in size (127,000) to the
hypothetical cities being considered.  The output from the model
was compared with actual data from the study using standard
regression techniques.  Partly due to the


difficulty in obtaining sufficient data to fulfill the MOD2 input
requirements, the results were less encouraging than had been
expected.  Nevertheless, although the numerical results obtained
from MOD2 differed from the actual figures by a large factor, MOD2
did evidence a capability to reproduce the direction of real world
trends.  This serves to strengthen the assertion that will be made
in this report as well, that principal attention should be given
not to the actual numerical outputs of the model, but to trends
which they indicate as a function of policies tested.
   The final phase of the research conducted by Bowman, et al., was
designed to identify promising energy-saving policies through the
use of the revised MOD2 package.  While only a limited experimental
program was carried out, among their conclusions were that travel
behavior variables seem to be more influential than land use
policies in determining transportation energy consumption and that
fringe residential growth can be relatively energy efficient if it
is tied to similar growth in service activities and employment.
   The main task left undone by Bowman, et al., was a more exten-
sive experimental program.  It was also clear that, while the
improvements to the model made in MOD2 were significant, there was
still room for further improvements.  Robert L. Peskin pursued
these two areas in the next phase of the research (Peskin, 1977).

   C. Policy Tests Applicable to Existing Cities: The Work of

   The focus of Peskin's research was the development of an
extensive-. experimental program in order to attempt to "answer
some key policy questions regarding the relationships between land
use, transportation


system characteristics, travel behavior, and transportation energy
consumption." (Peskin, 1977, p. 1) First, some significant
improvements were made to the MOD2 program.  The experimental
program followed, along with an extensive analysis with the
objective of formulating "practical guidelines for the development
of both national and local policies regarding transportation.and
land use planning-" (Peskin, 1977, P. 1)
   Peskin made two major improvements to the MOD2 program developed
by Bowman, et al.  First, he introduced the concept of generalized
cost of travel to replace time as the sole measure of impedance to
travel in the land use model.  This made it possible to assess the
impacts of changes in the price of travel, including fuel and
parking costs, on spatial patterns of activities.  Like many other
aspects of the evolving model, the generalized cost formulation was
based on a combination of logic and past (reported) research, and
could not be directly validated due to the absence of appropriate
   Second, in order to accommodate the notion that energy policies
would be implemented in the context of existing cities but that
their effects would principally influence future growth, the idea
of the "layering" approach to dynamic urban growth simulation was
introduced, as shown schematically in Fig. 2-4.  The concept
portrayed here is that a base city is first simulated using the
Lowry model representing some actual or hypothetical city.  Onto
this city, a growth layer of basic employment with its associated
increase in population and retail employment is introduced which
incorporates the particular policy under consideration (Fig. 2-5). 
Peskin's experimental program concerned itself

Click HERE for graphic.



primarily with this second or incremental mode of the program
operation.  Thus, he was able to evaluate the impacts of policies
implemented in existing cities, a capability which was absent in
previous work.
   Peskin conducted 112 experiments across three hypothetical
cities.  These test cities-the concentric ring, one-sided, and
polynucleated (Fig. 2-6)-were designed to be "representative of
typical urban forms in the U.S." (Peskin, 1977, p. 78).  The first
task of this program involved the comparison of the standard
incremental runs across all of the test cities; these experiments
were so named because they simulated a 25% growth increment in the
cities with no changes in transportation or land use policies. 
This was designed to identify the important differences between the
urban forms in order to guide later experiments.  Next, 12
experiments, one for each of 12 representative policy actions, were
conducted over each of the test cities.  These policy actions were
ranked in terms of their effectiveness in terms of energy
consumption, the highway congestion index, the work trip congestion
index, and total automobile VMT.  Finally, the remaining
experiments explored a broad range of policy excursions, pursuing
the most promising approaches to reducing transportation energy
consumption. (Peskin, 1977, p. 77)
   Many interesting results were generated regarding the relative.'
energy efficiency of various policy actions.  While it would not be
possible to include in this review a discussion of all of the
findings, the most important will be mentioned.  For a more
complete discussion, the reader is referred to Chapter 5 of the
work by Peskin & Schofer (Peskin & Schofer, 1977).  First, with
regard to the standard incremental

Click HERE for graphic.

                      Figure 2-6.  Peskin's Cities



run (no policy change) over each of the test cities, the
polynucleated city was found to be over twice as energy efficient
as the concentric ring city, with the one-sided city falling
between the two.  While MOD3 (as Peskin referred to his program)
was vastly different from the first model developed by Edwards,
this result agreed with the results that Edwards had obtained. 
Edwards discovered at the same time, however, that the energy
efficiency of the polynucleated city was achieved at the expense of
decreased regional accessibility.  Also notable was the marked
difference in the work trip congestion index between the different
cities.  The polynucleated city exhibited a congestion value much
lower than the value for the concentric ring or one-sided city.
   Congestion also appeared to be the most critical factor affect-
ing transportation energy consumption in the policy effectiveness
ranking experiments.  Other findings indicated that highway
improvements could be expensive solutions to reducing energy
consumption because of their less-obvious land use impacts, while
certain types of directed urban growth could be highly energy
   The remaining experiments covered a wide variety of policy
actions.  These included highway and transit network improvements,
transportation pricing changes, automobile occupancy variation, and
many land use changes.  Several additional experiments were made in
which two or more policy actions were combined.  Directed urban
growth, when coordinated with appropriate transportation services,
proved to be much more energy efficient than sprawled growth. 
Congestion was found to be the single most important determinant of
transportation energy consumption.  It was found that often simple
TSM-type highway


improvements could prove more effective than expensive highway con-
struction projects in reducing congestion, reducing energy
consumption without producing adverse land use impacts.  TSM
actions such as carpooling and increased commuter parking fees were
also found to be effective in reducing energy consumption. 
Provision of additional transit service did not seem to be
particularly effective although it appeared that maintenance of a
base level of peak-period service was essential in reducing highway
congestion and thus energy consumption.

   D. Other Contemporary Research

   The present work constitutes a further development of the work
begun by Edwards & Schofer, and subsequently expanded upon by
Bowman, et al., and Peskin & Schofer as described above.  Before
turning our attention to the particular improvements/changes
instituted within the present work, however, it is worthwhile to
consider briefly some of the other relevant contemporary research
   The joint effort of the Brookhaven National Laboratory and the
State University of New York at Stony Brook completed in 1977
(Carroll, 1977) produced a Lowry-based model designed to simulate
the energy requirements of alternative urban forms.  It is
comprehensive in that it considers not only energy used by
transportation, but also that used by the commercial, industrial,
and residential sectors.  In addition, it attempts to relate these
demand estimates to the supply side of the equation in terms of
both quantity and type (oil, electricity, etc.)
   As one would expect, the comprehensiveness of the model is
achieved at the expense of detail and realism throughout its


The transportation sector, for instance, is handled in a rather un-
sophisticated fashion.  First, transit is not considered as a
possible mode of travel.  This makes it impossible to investigate a
whole realm of energy-conserving possibilities that involve greater
use of transit.  Secondly, and most importantly, the assignment of
trips onto links is accomplished using an unconstrained minimum.
time path algorithm.  This means that congestion, which according
to Peskin & Schofer is a factor at least as important to energy
consumption as vehicle miles of travel (VMT), is never considered. 
This is crucial in that it greatly affects the land allocation
process itself.  Hence, failure to consider congestion can result
in a gross underestimation of transportation energy.
   It is worthwhile to note that a part of the project involved the
development of The Planner's Energy Workbook (Carroll, 1977), a
workbook "designed to present a straightforward set of
calculational procedures and worksheets which community planners
and designers can use to carry out their own evaluation of
alternative land use planning and design programs." (Carroll, 1977,
Final Report, p. 25)
   The development of the concept alone (more so than the actual
output) of a planner/decision-maker oriented tool of this type is
an important contribution to the field.  As noted in the
introduction to this report, it is of utmost importance that the
planners and the decision makers be able to utilize the results of
research efforts such as these.  The conclusions in Chapter 5
reflect this concern.
   Stephen H. Putman has done considerable work in the area of
integrated simulation transportation/land use models.  Of interest


is his comparison of the EMPIRIC and Lowry-derivative models
(Putman, October 1976).  He concludes that whereas both types of
models are able to yield close statistical fits to observed data in
the base year, the Lowry-derivative is far superior with respect to
its response to changes in input.  Hence, a Lowry model is
preferred when dealing with policy changes, a fact which lends
credibility to Edwards' original choice of the Lowry model as the
foundation of his work.
   Further, Putman's Progressive Land Use Model (PLUM, a Lowry
derivative) contained the important feature of an incremental
feedback .from transportation to land use that resulted in the
continuous alteration of the land use pattern caused by exogenously
specified changes in the transportation network (Putman, December
1976).  This feature was used by Peskin in the development of a
realistic policy simulation.  Another model developed by Putman,
DRAM, was calibrated with success against five different U.S.
metropolitan areas and is now included in the Integrated
Transportation and Land Use Package (ITLUP).
   An effort presently underway by Romanos and Hatmaker represents
a different approach to land use/transportation/energy
relationships.  In their model, energy cost is used as the
determinant in the allocation of new land activities.  Using a
linear programming approach, the objective function calls for the
minimization of transportation costs (i.e., energy) given a set of
system constraints.
   This model, too, suffers from the lack of feedback from the
transportation network to the land allocation scheme.  Furthermore,
as the authors admit themselves, the cost of energy is not, as yet,
the chief component of most locational decisions.  The use of a


generalized cost function as employed by Peskin would improve the
overall realism of the model since it takes into account the time
cost, an important locational determinant, as well as the energy
cost of travel.  Several other studies have explored different
aspects of the transportation/energy/urban form relationship.  A
comprehensive review of much of the related literature can be found
in doctoral dissertations by Clark (1976) and Rice (1975).  Again,
the reader is also referred to the review presented by Peskin &
Schofer in the second chapter of their work (1977).
   While the work of all of these researchers is highly relevant to
the transportation/energy/land use issue, and the experimental
program conducted by Peskin produced many interesting results and
constituted the most complete experimental program and analysis
thus far completed, several additional directions still exist for
further research consideration.  These will be addressed in Chapter
3 of this report.


References to Chapter II

   1. Jerry L. Edwards, Relationships Between Transportation Energy
      Consumption and Urban Spatial Structure, unpublished Ph.D.
      dissertation, Department of Civil Engineering, Northwestern
      University, June 1975.

   2. Larry A. Bowman, Daniel H. Goetsch, and Steven E. Polzin, A
      Model for Evaluating the Energy Conserving Potential of
      Transportation and Land Use Policies: Development and
      Preliminary Application, unpublished M.S. research report,
      Department of Civil Engineering, Northwestern University, May

   3. Robert L. Peskin and Joseph L. Schofer, The Impacts of Urban
      Transportation and Land Use Policies on Transportation Energy
      Consumption, Report No. DOT-OS-50118, U.S. Department of
      Transportation, Office of University Research, April 1977.

   4. Larry J. LeBlanc, Mathematical Programming Algorithms for
      Large Scale Network Equilibrium and Network Design Problems,
      Ph.D. dissertation, Transportation Center at Northwestern
      University, 1973.

   5. T. Owen Carroll, Robert Nathans, and Philip F. Palmedo, Land
      Use and Energy Utilization, Final Report of Brookhaven
      National Laboratory/State University of New York (Stony
      Brook) Land Use and Energy Utilization Project for the Office
      of Conservation and Environment of the Federal Energy
      Administration, June 1977.

   6. T.. Owen Carroll, Robert Nathans, Philip.F. Palmedo, and
      Robert Stein, The Planner's Energy Workbook, Brookhaven
      National Laboratory/State


      University of New York (Stony Brook) Land Uses and Energy
      Utilization Project for the Office of Conservation and
      Environment of the Federal Energy Administration, June 1977.

   7. Stephen H. Putman, Laboratory Testing of Predictive Land-Use
      Models:.  Some Comparisons, Report No. NSF-GI-38978 for U.S.
      Department of Transportation, Office of Transportation
      Systems Analysis and Information, October 1976.

   8. Stephen H. Put-An, The Integrated Forecasting of
      Transportation Land Uses, prepared for presentation at the
      seminar on Emerging Transportation Planning Methods, for U.S.
      Department of Transportation, December 1976.

   9. Michael C. Romanos and Michael L. Hatmaker, "Urban Activity
      Allocation Under Criteria of Transportation Energy
      Efficiency," International Journal of Energy Research 4
      (Jan/March 1980):1-10.

   10.   J. W. Clark, The Effect of Urban Form on Travel Demand and
         Transportation Energy Consumption: Theoretical
         Considerations, unpublished doctoral dissertation,
         Department of Civil Engineering, University of Washington,
         Seattle, Washington, 1976.

   11.   R. G. Rice, Performance Characteristics of Two-Mode
         Transportation Systems on Varying Urban Form, unpublished
         doctoral dissertation, Department of Civil Engineering,
         University of Toronto, Toronto, Ontario, Canada, 1975.

                                CHAPTER 3


                        THE EXPERIMENTAL PROGRAM

A. Introduction

   The simulation model used in this research is largely taken from
the work of Peskin and Schofer (1977), whose report provides ample
documentation of the model (MOD3).  A general flow chart of the
core of the overall system, based on the Lowry land use model, is
shown in Figure 3-1.  Several modifications to the model and its
parameters were made to increase the efficiency and realism of that
earlier version.  These are discussed here, after which a number of
qualifications or caveats to be applied when interpreting the
results are introduced.  Finally, the overall experimental program
is reviewed.

B. Revisions to MOD3 Program

1. Conversion to CDC 6600

   Because of the recently expanded capacity of the CDC 6600 system
at Northwestern, it was deemed feasible and desirable to convert
the model from Peskin's version for the IBM 370-195 system.  This
necessitated several minor changes.  The two systems have somewhat
different methods of satisfying mass storage requirements, thus
requiring revision of all statements dealing with mass storage
input/output.  The only other revision involved the treatment of
the double precision variables in the


Click HERE for graphic.

         Figure 3-1.  Core of MOD3-The Transportation/Land Use
                      Feedback (Source: Peskin, 1977, p. 41]



MATINV (matrix inversion) subroutine of MOD3.  Eventually, this
subroutine was replaced altogether by a packaged matrix inversion
routine (LINVSF of the IMSL package).  It is important to note that
no change in the output resulted from the substitution of this
system subroutine.  Unfortunately, the new subroutine did not
produce any noticeable gains in efficiency.
   Whereas all experiments had been submitted in straight batch
mode previously, in this work maximum use was made of online
input/output methods in order to increase efficiency.  All program
changes were made from a CRT terminal using the EDITOR6 routine. 
Later, all data sets were manipulated using the same routine, and
all experiments were submitted in batch mode from a CRT terminal. 
It is felt that this greatly increased the efficiency of the
experimental program.

2. Making the Experimental City More Realistic

   At the outset of the present research, a decision was made to
confine the experimental program to the concentric ring city which
had been found to give results representative of the one-sided city
and more realistic than those for the polynucleated city (Peskin &
Schofer, 1977).  The concentric ring city developed in Peskin's
work consisted of 52 zones arranged in a square 10 miles on each
side (see Figure 2-6).  Reviewers had suggested, however, that most
concentric ring cities do not develop in such a square fashion but
rather tend to develop along perpendicular axial highways in more
or less a diamond shape.  It had also been suggested that such a
diamond shape would tend to reduce the impacts of diagonal freeways
and arterials which were found to produce dramatic energy savings
in Peskin's experiments.

   Many options were available for accomplishing the rotation of
the city.  Basically, these fell into two categories: rearranging
the zones around the existing transportation network or,
conversely, rotating the network within the existing zonal
arrangement.  Although the flexibility exists in the program for
virtually any number of zones, it was also decided that, for
comparability, a number of zones similar to that used-by Peskin was
desirable.  It was also deemed desirable to maintain the basically
perpendicular. grid network and to have link intersections occur
only at zone centroids.  Under these considerations, the rotated
city presented in Figure 3-2 was developed.  Applying the same
scale used in the previous research, the new city would have an
area of 98 square miles.  For a city of 125,000, this would yield
an average gross density of 1275/square mile, or about 2
people/acre.  This value for density appears quite low when
compared with actual values for cities of roughly the same
population.  Rather, a representative value seems to be close to
2500/square mile (see Peterson, 1977).  Hence, a factor of .7 has
been applied to each dimension of the city, yielding a city with an
area of 47.04 square miles and an average gross density of
2657/square mile.  Shrinking the size of the rotated city has the
added advantage of creating a denser transportation network.  That
is, for example, instead of the links being separated by .5 miles
in the CBD, the distance is now .35 miles.  It is felt that a
change of this type makes the model more realistic.  Of course,
this change further limits the comparability between this effort
and that of Peskin.
   Like the old city, the new city has 52 zones.  With 180 one-way
interzonal links, the link-to-node ratio is slightly lower than

Click HERE for graphic.

         Figure 3-2:  New Concentric Ring "Diamond" City



at 3.462 (Peskin's: 3.538). All in all, there are 162.8 one-way
roadway miles which is a 30 percent reduction from the previous
concentric ring city.  The zones in the new, rotated city define
five concentric rings.
   In developing the transportation network for the new rotated
city, the highway links were divided into four groups.  Capacities
and maximum speeds were obtained from data contained in the CUTS
Manual (DeLeuw, Cather, 1977) and are shown-in Table 3-1.  The four
link classifications can be seen in Figure 3-3.  It should be noted
that all of the Group A streets were converted to one-way, which is
not atypical of arterials passing through a CBD.
   The transit network is similar to the one used by Peskin in his
square concentric ring city (Figure 3-4).  All zones are served by
at least one of the six routes (each route operates in two
directions), and each route begins and ends at the city periphery
and serves the CBD, as is common in many smaller United States
cities.  Route spacing and average link speeds are also
representative of urban bus routes in typical United States cities. 
Frequency, however, was decreased from six (value used by Peskin)
to three buses per hour, a value considered more representative of
the type of service observed in smaller cities.
   In creating the rotated base city of 100,000, it was observed
that the population density gradient the city used in the previous
work was extremely steep, with densities of greater than 30,000/sq
mi in the CBD, and around 400 in the outer zones.  In fact, over 50
percent of the population was found to be living in the first two
rings (zones 1-12) which represents less than 10 percent of the
total area.  This resulted

Click HERE for graphic.

         Figure 3-3.  Highway Network


Click HERE for graphic.

                      Figure 3-4.  Transit Network



                                TABLE 3-1
                           LINK CLASSIFICATION

                             Link Capacity      Maximum Speed
      Location               (veh/hour)            (mph)

   A. CBD                       1400                  20

   B. Fringe                    1600                  25

   C. Residential               1600                  30

   D. Outlying Residential      1100                  30

in extremely low simulated interzonal traffic volumes on links in
and near the CBD, indicating that many people were -both living and
working in the downtown area and, most likely, walking to work. 
This is not totally realistic, especially for a city of the type
and size being considered in this research; however, it indicated
that the model's land use allocation process, when unconstrained,
was behaving as would be-expected, i.e., people were locating as
closely as possible to jobs.  The model, however, fails to account
for the negative externalities associated with living in such a
densely populated area (e.g., lack of privacy, high crime rate)
that are evident in virtually every American city.  These must be
accounted for externally through the use of both the attractor
variables and density constraints.
   To determine a more appropriate density gradient, data on actual
values in real American cities were sought.  Because good data do


exist on density gradients per se, two related sources were
The first involved the use of the rather simplistic negative
exponential density function, as presented in Peterson (1977) and

                                DEN = ae

   DEN   =  gross population density
   k     =  distance from the urban center
   a,b   =  parameters
   e     =  the natural logarithm base.

Using estimates for the coefficients a and b based on information
presented by Peterson, a rough estimate of a density gradient was
sketched (Figure 3-5).
   A second perspective came from examining Census Block statistics
(U.S. Bureau of the Census, 1974) for several representative cities
(Topeka, Kansas; Cedar Rapids, Iowa; Montgomery, Alabama; and
Amarillo, Texas).  Radials were sketched out from the center, and
densities were computed for city blocks at one mile increments from
the center of the city.  This procedure produced a weak
relationship between density and distance from the CBD.  It also
pointed out one important fact which, while fully expected, is not
accounted for by an exponential density function: because
residential locations in the CBD are preempted by other land uses,
gross residential densities in the CBD are typically quite low. 
While these procedures are neither sophisticated nor productive of
conclusive results, they did provide enough information to ,give a
reasonable idea of a representative density gradient.

                     Figure 3-5.  Density Gradients



   Based on this information, density constraints and residential
attractor variables in the model were adjusted until: a
satisfactory density gradient was achieved.  The final base run and
standard incremental run gradients are shown along with Peskin's
standard incremental run density gradient in Figure 3-5.  The
flattening of the density gradient increased the energy consumption
of the city substantially.  It is felt that this density
modification is a major contribution toward increasing the realism
of the results of this research.
   Other changes made to the base data sat, introduced to reflect
more recent conditions, were of a relatively minor nature.  These
included raising the gasoline price to a level of 70/gallon,
raising the work trip car occupancy value from 1.2 to 1.3 and the
social/recreation trip car occupancy from 1.8 to 2.0 (COMSIS,
   The standard incremental run (S.I.R.) was developed by intro-
ducing 25,000 new residents into the existing base city of 100,000. 
To accomplish this, 3,750 basic employment jobs were distributed
around the city.  The model then located the population and retail
employment with respect to this new basic employment.  The
distribution of the basic employment increment used was similar to
the one used by Peskin and shown in Figure 3-6.

3. Determination of Appropriate Cluster Sizes

   Following the development of the base and standard incremental
cities, one task remained to be completed before the main portion
of the experimental program could be undertaken.  A decision had to
be made as to realistic sizes for the three types of cluster
development to be tested:    basic employment, service employment,
and residential.

Click HERE for graphic.

              Figure 3-6.  Distribution of Basic Employment



   For basic employment it was argued that the number of jobs in-
troduced into a community is a function of the work force available
which, in some respect, can be looked upon as a function of the
rate of unemployment.  If a maximum long-term rate of unemployment
is assumed to be 8 percent, and the basic labor force participation
rate is given as 22 percent of the population, an assumption can be
made that a realistic maximum basic employment cluster might be on
the order of:

                  100,000 x .22 x .08 = 1760 employees

From this simple assumption, a reasonable maximum size for a basic
employment cluster was assumed to be 1,500 employees.
   For retail employment and residential clusters, data were more
easily available.  For retail, the 1972 Census of Retail Trade
(U.S. Bureau of the Census, 1976) revealed that, although shopping
centers in large urban areas sometimes reach 4,000 employees and
higher, a reasonable size for a large shopping center in a small
urban area ranged from 1,200 to 1,500 employees.  For residential
clusters, several large residential developments were checked in
the U.S. Census Block Statistics (U.S. Bureau of the Census, 1974). 
These developments also seemed to range from 1,000 to 1,500

C. Caveats

   Before proceeding to the experimental program itself, it is
necessary to consider some of the key limitations of the simulation
model being employed in this research; that is, the MOD3 program
developed by Peskin (see p. 37), along with the changes outlined in
the first part of this chapter.


   To use the results of this research, other researchers,
planners, and decision makers must have sufficient confidence in
the validity of the outputs, the process through which they were
obtained, as well as the interpretations reported here.  The
validity of the model is its ability to produce a sufficiently
accurate representation of the behavior of the real world; however,
it is in the nature of every model to be limited in the ability to
do this.  Assessment of validity should focus on understanding
these limitations.
   To assure feasibility of development and application, a model
must simplify the real world it is to represent.  Hence,
simplifications are present in all practical models of urban and
transportation processes. -The model used here is further limited
by the state of the art of the component theories and modeling
technologies selected for use as well as the innovativeness of the
particular-modeler.  A careful consideration of the particular
limitations of any model is a necessary step in understanding the
results produced by that model and the conclusions based upon them. 
Therefore, a brief overview of the limitations of this model is
presented here, especially as they relate to the work done in this
research.  For a more detailed discussion, the reader is again
referred to the work by Peskin & Schofer (1977, pp. 61-75).
   A major limitation arises out of the aggregate nature of the
Lowry-type land use model.  All of the households are considered to
be homogenous in their travel behavior and locational preferences. 
Housing stock is considered uniform throughout the city, although
subjective treatment of the residential attraction variables allows
the possibility of some gross differentiation.  Combined with the
static equilibrium


approach of the activity allocation process, these limitations
restrict the realism in the simulation of the housing market.  In a
realistic situation, this might be reflected through the social and
amenity factors which have encouraged the flow of higher income
families out of the central area and into newer housing stock.
   A second major limitation involves the use of current travel
behavior characteristics to forecast the impacts of various
policies in the context of a population increase of 25%.  It
impossible that during the period of growth numerous forces could
act to produce dramatic .changes in travel behavior which might
alter locational preferences, trip distribution, mode split, and
energy consumption patterns.  Among the many externally determined
variables in this model (see Table 2-1) which could be subject to
change are the trip generation rates, trip time and cost
distributions used to generate trip time and cost preference
(friction factor curves), the value of travel and wait time, auto
availability (which can affect non-work trip generation rates),
auto occupancy, transit routes and levels of service, auto fuel
efficiency, the coefficients of the mode split model, and gasoline
prices.  It is understandably difficult to anticipate the magnitude
of the changes that might occur in these variables.
   The recent, unanticipated, precipitous rise of gasoline prices
beyond the $1.00/gallon level serves as a clear illustration of the
dangers of assuming the future stability of these variables. 
Furthermore, the large number of variables acts to restrict the
feasibility of either alternative futures or sensitivity analysis
approaches (Schofer & Stopher, 1979) to deal with uncertainty in
these variables.


   A third major limitation of the model is in the iterative
algorithms used to allocate land uses and travel on the highway and
transit networks.  As with most iterative procedures, the preferred
end-state is convergence at an equilibrium value; however, exact
convergence requires excessive amounts of computation time.  Thus,
some trade-off must be reached between computational efficiency and
level of convergence.  Research by Bowman, et al. (1976) determined
that for the transportation network assignment algorithm, execution
of more than four or five iterations for each basic employment
increment yielded very diminished returns in terms of level of
convergence.  Most extreme shifts in link volume were actually
eliminated in the first two to three passes, while the next two or
three iterations served to bring the link volumes closer to their
equilibrium values.  Beyond that, the minor shifts in volume from
one link to the next did not justify the cost of performing the
additional iterations.  Hence, it was determined that five
iterations would be performed on each basic employment increment. 
The present research adopted this number also.
   A further serious limitation is involved in the application of
the model in the present experimental program.  To keep
computational costs to a minimum, the number of zones and,
consequently, the number of links in the network, is kept small. 
This is particularly unrealistic in the CBD, where there are
usually many streets with high volumes of traffic.
   Finally, and perhaps most critical, is the artificial boundary
created by the edge of the city.  As will be postulated in the
analysis, this has an apparently important impact on estimates of
consumption and amount of travel since it artificially controls the
urban sprawl--new development at the edge of existing cities.  This
will tend to underestimate the magnitude of travel and energy
There are numerous, less-serious limitations.  Applying the 25%


population growth increment in one piece is not realistic and yet
it is not clear what size increments in what sequence would be more
realistic.  Also somewhat unrealistic is the assumed static
character of existing land uses.  In an actual city, buildings grow
old, are abandoned, or are demolished, and replaced by new
buildings.  Large urban renewal projects can, over an extended time
period, have an important effect on the redistribution of-
population and other land uses in the urban area.  Further,
external trips (those beginning and/or ending outside-the urban
area) are ignored by this model.  Likewise ignored are trips by
trucks and commercial vehicles, the negative environmental impact
of which can have considerable influence on activity location, and
which may consume significant quantities of energy.  Both result in
some underestimation of travel volumes, congestion, and energy
consumption in the urban area.  Congestion effects of transit and
non-work trips are also ignored by the models.  Finally, the
assumption of a one-hour peak period is subject to some question,
and has a direct influence on congestion and energy estimates.

D. Method of Analysis

   The spectrum of limitations presented here (and in more detail
in Peskin, 1977) should serve to warn the reader against assigning
great meaning to small numerical differences between experiments. 
This leads the experimental program toward investigating broad
urban development questions as opposed to minor changes in policy
   Thus, the analysis attempts to identify trends among a series of
experiments rather than concentrating on the precise numerical
variances between specific experiments.  Since there are no exact


methods for doing this, a more subjective methodology is used to
assist in the development of useful conclusions.  The heart of this
procedure is the use of intuition to guide the interpretation of
the results.  As Peskin & Schofer note, "A well designed,
structured intuitive assessment of the experimental results can be
both efficient and relevant, considering that it will be the
intuitive judgment on the part of planners and decision makers that
will determine whether and how the results are used."  (Peskin &
Schofer, 1977, p. 73) This is well supported by recent studies of
the use of analytic models in policy development (Greenberger, et
al., 1976).

   Peskin's analysis, and that adopted by this research, follows
the framework shown in Fig. 3-7.  First, an experiment is judged as
to whether the energy consumption agrees with intuition in
comparison with other experiments.  If so, confidence in the model
is increased.  If not, a logical analysis is attempted which, if
successful, alters intuition on future experiments and still
increases confidence in the model.  If a logical analysis is not
possible, further experiments, commonly of a parametric type, can
be performed, or the experiment is "flagged" as an anomaly.  It is
through this procedure, always with deference to the many
limitations that were outlined above, that the conclusions are
drawn regarding various urban development and transportation

E. Experimental Program

   The decisions leading to the development of the base (#513) and
standard incremental #514) cities have been discussed in detail in

Click HERE for graphic.

                     Figure 3-7.  Analysis Framework



this chapter.  For purposes of comparison, an important experiment
(#996) was carried out in which the density constraints were
removed from rings 1 and 2 in the standard incremental run.  These
two standard incremental experiments (#514 and #996) are compared
in detail in Chapter IV.
   The main body of the experimental program began with experiments
520 to 524 which investigated the effect of clustering 1,500 basic
employment jobs in each of five different zones.  The particular
zones chosen--2, 7, 18, 36, and 50-were selected in order to
provide information on the effect of moving clusters axially
outward from ring 1 to ring 5 (Fig. 3-8).  Experiments 530 to 534
added incentives for residential clustering around the basic
employment clusters by altering density and residential attractor
variables in zones adjacent to the employment nuclei.  Experiments
550 to 554 added to these an extra highway lane to all links to and
from the nucleus zone under consideration.  Meanwhile, experiments
541 through 543 investigated the impacts on the standard
incremental run of adding axial freeways, diagonal freeways, and
diagonal arterials, respectively.  These were pursued to provide a
comparison with the work of Peskin, who identified significant
reductions in energy consumption-with the introduction of diagonal
highway improvements.
   Since the residential clustering cannot be controlled explicitly
through an input variable in this model, as is the case with basic
employment, the residential attractors had to be altered
judgmentally until a cluster of acceptable size could be created. 
This resulted in 21 experiments in the 560 to 564 series with
several different residential

Click HERE for graphic.

                 Figure 3-8.  Cluster Development Zones



cluster sizes at each location.  Likewise, service employment
clusters were created by altering the retail attractor variable of
the selected zone until the desired size of service employment
cluster was achieved.  Hence, 19 experiments were run in the 570 to
574 series.  The impacts of cluster type, size, and location will
be investigated in the following chapter.
   With this portion of the research completed, two additional
network improvements were introduced: an inner ring road (#547) and
an outer ring road (#548).  These ring roads increased street
capacities in ring 3 and ring 4, respectively.  The effect of the
ring roads on the clustered developments was investigated by
introducing the appropriate ring road into each of three
experiments dealing with the three different cluster types in zone
18 (inner ring road with experiments 552, 562E, and 572A) and in
zone 36 (outer ring road with experiments 553, 563C, and 573A). 
These six experiments were numbered 552B, 582, 592, 553B, 583, and
593, respectively.
   The final task involved developing an integrated cluster
development with basic employment, residential, and retail
employment clusters of about 1,000 persons each in each of the five
cluster locations.  These also required several trials to reach
desired sizes, leading to 12 experiments in the 600-604 series. 
Experiments 610-614 consisted of the inner ring road added to each
of the integrated clusters, while experiments 620-624 consisted of
the outer ring road added to each of those same cluster
developments.  A total of 88 experiments were conducted, the
results of which will be analyzed in Chapter 4. (Table 3-2 sum-
marizes the full set of experiments conducted.)

                                TABLE 3-2

                          EXPERIMENTAL PROGRAM

   Expt# Description

   513   Base Run

   514   Standard Incremental Run'

   520   Basic Employment Cluster (1500) in Ring 1 (Zone     2)
   521   Basic Employment Cluster (1500) in Ring 2 (Zone     7)
   522   Basic Employment Cluster (1500) in Ring 3 (Zone     18)
   523   Basic Employment Cluster (1500) in Ring 4 (Zone     36)
   524   Basic Employment Cluster (1500) in Ring 5 (Zone     50)

   530   Basic Employment w/Population Cluster added around Zone 2
   531   Basic Employment w/Population Cluster added around Zone 7
   532   Basic Employment w/Population Cluster added around Zone 18
   533   Basic Employment w/Population Cluster added around Zone 36
   534   Basic Employment w/Population Cluster added around Zone 50

   541   Axial Freeways w/S.I.R.
   542   Diagonal Freeways w/S.I.R.
   543   Diagonal Arterials w/S.I.R.

   547   Inner Ring Road w/S.I.R.
   548   Outer Ring Road w/S.I.R.

   550   Basic/Population Cluster w/Single Lanes Added to and from
         Zone 2
   551   Basic/Population Cluster w/Single Lanes Added to and from
         Zone 7
   552   Basic/Population Cluster w/Single Lanes Added to and from
         Zone 18
   553   Basic/Population Cluster w/Single Lanes Added to and from
         Zone 36
   554   Basic/Population Cluster w/Single Lanes Added to and from
         Zone 50

   552B  Basic/Population Cluster in Zone 18 w/Inner Ring Road
   553B  Basic/Population Cluster in Zone 36 w/Outer Ring Road

   560   Population Cluster (1386) in Ring 1 (4 runs)
   561   Population Cluster (1266) in Ring 2 (6 runs)
   362   Population Cluster (1613) in Ring 3 (5 runs)


                            TABLE 3-2 (cont.)

   Exp#  Description

   563   Population Cluster     (1328) in Ring  4 (3 runs)
   564   Population Cluster     (1600) in Ring  5 (3 runs)

   570   Service Employment Cluster (1254) in Ring 1 (3 runs)
   571   Service Employment Cluster (1281) in Ring 2 (4 runs)
   572   Service Employment Cluster (1545) in Ring 3 (3 runs)
   573   Service Employment Cluster (1478) in Ring 4 (3 runs)
   574   Service Employment Cluster (1288) in Ring 5 (3 runs)

   582   Population Cluster     in Zone  18 w/Inner Ring Road
   583   Population Cluster     in Zone  36 w/Outer Ring Road

   592   Service Employment     Cluster  in Zone 18 w/Inner Ring Road
   593   Service Employment     Cluster  in Zone 36 w/Outer Ring Road

   600   Integrated  Cluster in  Ring 1  (1 run)
   601   Integrated  Cluster in  Ring 2  (2 runs)
   602   Integrated  Cluster in  Ring 3  (3 runs)
   603   Integrated  Cluster in  Ring 4  (3 runs)
   604   Integrated  Cluster in  Ring 5  (3 runs)

   610   Integrated  Cluster in  Zone 2  w/Inner Ring Road
   611   Integrated  Cluster in  Zone 7  w/Inner Ring Road
   612   Integrated  Cluster in  Zone 18 w/Inner Ring Road
   613   Integrated  Cluster in  Zone 36 w/Inner Ring Road
   614   Integrated  Cluster in  Zone 50  w/Inner Ring Road

   620   Integrated  Cluster in  Zone 2 w/Outer Ring Road
   621   Integrated  Cluster in  Zone 7 w/Outer Ring Road
   622   Integrated  Cluster in  Zone 18 w/Outer Ring Road
   623   Integrated  Cluster in  Zone 36 w/Outer Ring Road

   624   Integrated  Cluster in  Zone 50 w/Outer Ring Road
   996   Unconstrained Standard Incremental Run


References to Chapter III

1. Robert L. Peskin and Joseph L. Schofer, The Impacts of Urban
   Transportation and Land Use Policies on Transportation Energy
   Consumption, Report No. DOT-05-50118, U.S. Department of
   Transportation, Office of University Research, April 1977.

2. Jack Stein, International Mathematical and Statistical
   Libraries, Edition 6, Northwestern University, Vogelback
   Computing Center, Document No. 294 (Rev.  D), revised March

3. J. Norstad, Editor Version 6.0, Northwestern University,
   Vogelback Computing Center, Bulletin No. 70, May 10, 1978.

4. George E. Peterson, George Reigeluth, and Dale Keyes,
   Metropolitan Development Patterns (Washington, D.C.: Urban
   Institute, to be published).

5. U. S. Bureau of the Census, Block Statistics, Census of Housing,
   1970, U. S. Government Printing Office, Washington, D.C., 1974.

6. DeLeuw, Cather & Co. and Rock Creek Associates, Characteristics
   of Urban Transportation Systems: A Handbook for Urban
   Transportation Planners, Office of Planning Methods and Support,
   UMTA, July 1977.

7. COMSIS Corporation, Quick-Responses Urban Travel Estimation
   Manual Techniques and Transferable Parameters-A Users Guide,
   prepared for NCHRP Project 8-12A, Transportation Research Board,
   Washington, D.C., November 1977.

8. U. S. Bureau of the Census, Census of Retail Trade, 1972, U. S.
   Government Printing Office, Washington, D.C. 1976.

9. Joseph L. Schofer and Peter R. Stopher, "Specifications for a
   New Long-Range Urban Transportation Planning Process,"


   Vol. 8, 1979, pp. 199-218.

10.   Larry A. Bowman, Daniel H. Goetsch, and Steven E. Polzin, A
      Model for Evaluating the Energy Conserving Potential of
      Transportation and Land Use Policies: Development and
      Preliminary Application, unpublished M.S. research report,
      Department of Civil Engineering, Northwestern University, May

11.   M. Greenberger, M. A. Genson, and B. L. Crissey, Models in
      the Policy Process, Russell Sage Foundation, New York, 1976.

                                CHAPTER 4

                          RESULTS AND ANALYSIS

A. Introduction

   This chapter analyzes the results obtained from the experimental
   There are three primary measures used to perform these analyses:
total energy consumption (in gallons of fuel), total auto vehicle
miles of travel (VMT), and the average highway congestion index (a
weighted average of the ratio of estimated to pre-specified free-
flow travel times on each link).  Although, as would be expected,
there is strong correlation between these three measures, there are
many experiments in which they do not covary, in which case the
divergence is often illustrative.
   Secondary measures which are used in the analysis include auto
work trip VMT, auto work trip length (in miles), percent of the
work trips on transit, the second moment of population (a measure
of spatial dispersion of population), and ring population
densities.  Mappings of population and retail employment
distribution (presented as percent zonal changes from the S.I.R.)
are an important tool used to illustrate certain impacts.
   The analysis begins with consideration of the base and standard
incremental runs and then a set of network changes, especially as
they relate to similar experiments in the square city used by
Peskin & Schofer (1977).  This is followed by a discussion of some
additional network 62

changes.  Considerable attention is given to different types and
locations of clusters, logically leading to a discussion of
integrated cluster developments.  Following a brief discussion of
the effects of, cluster size, the chapter then concludes with the
analysis of combinations of large scale clusters and network

   B. Base and Standard Incremental Runs in Square and Related

   The base run and standard incremental run (S.I.R.) have already
been introduced in the discussion of density gradients in the
previous chapter.  While it is not possible, due to the many
changes made in this work, to compare the base and S.I.R.
experiments developed here to those developed previously by Peskin,
it is worthwhile to compare the relationship of each S.I.R. to its
own base run.  The comparison is presented in Table 4-1.
   The total energy impact of the standard incremental runs (25
percent population growth increment) is roughly the same for both
cities.  As would be expected, the smaller, more densely populated
rotated city is more sensitive to the 25 percent growth increment
in terms of congestion and less sensitive in terms of VMT.- The
combination of these two effects results in the similarity with
regard to total energy consumption.  Table 4 -1 illustrates that
change in VMT alone is not always a good indication of change in
energy consumption.  Here it can be seen that congestion and VMT
together provide a more comprehensive explanation of energy

                                TABLE 4-1

                   COMPARISON OF S.I.R. AND BASE RUNS

                Square (Peskin)          Rotated

                   Base      S.I.R.   %      Base     S.I.R.       %

Total Energy )     98,100    135,375  +38    59,901   81,865    + 39  
Total Auto VMT
  (102 miles)      3,535      5,652  +60     3,512     4,545   + 29

Average Highway
 Congestion Index   1,343      1,511  +49     1,305     1,633   +108

Work Energy         9,816     12,784  +30     5,381     7,693   + 43
 (106 BTU)

Auto Work Trip Length
   (miles)         2.704       2.837  + 5      2.648    2.771   + 5

Percent Work Trips
on Transit         6.58       13.29   +102    14.24     17.82   + 25



   C. Effects of Network Changes

   1. Axial Freeways, Diagonal Freeways and Diagonal Arterials in
      Square and Rotated Cities

   To investigate the suggestion that the diagonal arterials and
freeways would produce a less dramatic impact on a rotated city
than on the square city, a series of three experiments was
conducted in which three different network changes were introduced
to the S.I.R.: axial freeways (#541), diagonal freeways (#542) and
diagonal arterials (#543).  These are presented on the maps in
Figure 4-1.  The results of these experiments are compared with the
results of similar experiments conducted by Peskin in Table 4-2.
   Two salient conclusions can be drawn from examination of these
data.  First, the impacts of each of the three network changes are
attenuated in the rotated city.  This results from the combination
of two contributing factors: (1) the limits to population
redistribution imposed by the density constraints, particularly in
the inner ring, and (2) the relatively short length of the diagonal
routes, reflecting one of the key differences between the square
and rotated city forms.
   Second, the introduction of either diagonal arterials or
freeways is still more effective than the conversion of axial
arterials into freeways in reducing energy consumption.  This is
the case despite the low highway congestion index in the axial
freeway experiment (#541).  The primary difference between the
diagonal and axial facilities lies in the fact that the diagonals
constitute actual additions to the highway network while the axial
freeways are only capacity and speed improvements applied to
existing links.  The added availability of routes yields more
efficient route choices by users (i.e., shorter trip lengths) and

Click HERE for graphic.

                 Figure 4-1.  Network Changes to S.I.R.


                                     TABLE 4-2
                              (% CHANGES FROM S.I.R.)

                Square (Peskin)                 Rotated          

                Axial    Diagonal  Diagonal   Axial      Diagonal  Diagonal
                Freeways Freeways  Arterials  Freeways   Freeways  Arterials
Experiment      171      149       191          541      542       543

Total Energy    - 1.0    -17.9     -16.4      - 7.0      -11.1     - 9.7

VMT             +25.7    +11.0     + 0.8     +11.3       - 1.1     - 2.8

Average Highway
Index           +40.7    -45.2     -44.0     -23.1       -10.3     -15.5


consequently an increase in overall energy efficiency.  Hence, the
advantage of the diagonals stems in part from reduced auto work
trip length, which subsequently accounts for the low values of VMT
when compared with the axial freeways.  Furthermore, axial freeways
have the effect of pushing population into the fourth and fifth
rings, presumably because of reduced travel friction to the CBD,
while the diagonals, because they only extend into ring 3, do not.
   The measures in Table 4-3 all bear out this difference in terms
of population distribution between the two diagonal experiments and
the axial freeways.  The trip length, as indicated on the table, is
lower with diagonal roadways.  The second moment of population is
also lower, indicating a stronger tendency for population to locate
centrally.  This is substantiated by the figures on ring population
densities which show that the diagonal roadways produce higher
density in Ring 3, but lower densities in the outer rings (4 and
   All three of the network improvements yield large increases in
downtown service employment (up 15-30 percent from the S.I.R.),
caused by the increased accessibility to the central areas.  As
noted above, this shift in downtown service employment is unmatched
by a similar shift in downtown population because of externally
imposed density constraints which restrict population in central
   Thus, compared with Peskin's experiments with a square city,the
rotated city shows reduced sprawl and energy effects of diagonal
roadway improvements.  Still, the positive effects predicted here
are substantial.  In interpreting these outcomes, it is essential
to keep in mind that the hypothetical city tested here was

                                TABLE 4-3

                                Axial        Diagonal    Diagonal
                        S.I.R.  Freeways     Freeways    Arterials
Experiment                 514      541         542          543

Auto Work 
Trip Length              2,771     2,857        2,520        2,597

2nd Moment of Population
      (x 103)           1,077     1,085        1,079        1,076

Average Population/sq mi

   Ring 3                4,194     4,096        4,161        4,167

   Ring 4                2,048     2,015        1,979        1,977

   Ring 5                  686       700          678          668



constrained in its fringe growth by the city boundaries, which
absolutely precluded further sprawl.  Such improvements in a real
city are likely to increase the pressure for sprawl, and thus to
introduce a countertrend of increasing energy consumption in
travel.  To gain the full measure of benefits from such highway
improvements, rigorous control of sprawled growth would be

2. Effects of Ring Roads'

   The results of two additional network improvements, the inner
ring road and the outer ring road, in rings 3 and 4 respectively
(Figure 4-2), yield information which is useful in the
understanding of the dynamics of the rotated city., These
experiments were introduced in response to high levels of
congestion observed in rings 3 and 4 and clearly apparent in the
S.I.R. data presented in Table 4-4.
   The results of the ring road experiments in Table 4-5 show, as
expected, large decreases in total energy consumption and highway
congestion, especially for the outer ring road.
   It might be expected that a ring road, by reducing the
congestion of links in the outer rings, would serve, by making
these zones more accessible, to draw population into these outer
rings.  This would result in increased VMT as people would travel
farther to work and shop, thus offsetting the energy gains produced
by the reduced congestion.In the ring road experiments, however,
both population and service employment tend to cluster near the
ring roads themselves, causing a reduction in VMT and yielding
energy-efficient cluster development along efficient highway links. 
This combination of naturally occurring population and service
employment clusters (i.e., those which occur due to

Click HERE for graphic.

                   Figure 4-2.  Ring Roads with S.I.R.


                                TABLE 4-4

                       LINK CLASSIFICATION A TO D

   Class        Location              Rings        Congestion

   A            CBD                   1,2             1.001

   B            Fringe                2,3             1.698

   C            Residential           2,3,4           1.788

   D            Outlying Resi-
                dential               3,4,5           1.542

                All Links                             1.633

                                TABLE 4-5

                S.I.R.   Inner Ring Road        Outer Ring Road       
Experiment                         % Change           % Change
   No.           514       547     from SIR     548   from SIR

   Energy       81,865   75,275       -8        69,716   -15

   VMT          454,501  454,435      -         445,906  - 2

   Congestion*    1,633    1,320      -49         1,294  -54

*  % change computed using 1.0 as base.



the transportation change alone) with network improvements focused
on congestion reduction proves to be highly energy-efficient.
   In Figure 4-3, which summarizes results of the network structure
experiments, the efficacy of the outer ring road relative to the
other improvements is clear.  It is also evident that the
congestion index undergoes much greater excursions than either
total energy consumption or VMT, the least sensitive measure.  The
rank ordering of energy consumption and VMT are more similar than
the ranking of energy and congestion; but the magnitude of the
changes is more similar between energy and congestion.

   3. Effects of Minor Capacity Improvements

   Up to now, virtually all of the network structure experiments
have demonstrated energy savings.  This is not the case in tests
that were made with selected lane additions to links leading
immediately into clusters of basic employment and population.  The
basic employment clusters were distributed in various non-central
zones as discussed in the following section, but population density
constraints were reduced and attractor variables were increased in
each nucleus zone and all contiguous zones to encourage population
clustering around the basic employment cluster.  The results are
shown in Figures 4-4 through 4-6.  In each case, the highway
improvement tested was the addition of a single running lane in
each direction on each link entering the nucleus zone.
   Figure 4-4 shows a small but noticeable increase in energy con-
sumption when the highway improvements are added to the basic
employment/ population clusters in rings 3 and 4. Figure 4-5 shows
that this increase over the clusters alone (without the highway
improvement) is

          Figure 4-3.  Results of Network Structure Experiments

Click HERE for graphic.


Click HERE for graphic.

     Figure 4-4.  Total Energy of Experiments with Basic Employment


Click HERE for graphic.



reflected in the VMT.  One possible explanation of this phenomenon
is the possibility that the capacity additions reduced congestion,
improved level of service, and resulted in increased travel (VMT,
trip length) due to trip diversions.  This effect is more
noticeable in rings 3 and 4 where such diversions are more likely,
than in ring 5, where the benefits to travelers of such diversions
is limited because the improved links only lead to the nucleus
zones.  This interpretation is supported by the work trip length
data shown in Table 4-6.
   This seems to emphasize the point that highway improvements need
to be carefully selected not simply to eliminate the worst con-
gestion, but also to avoid encouraging VMT increases.  The ring
roads are more successful at achieving these objectives because
they provide An overall and symmetrical level of service
improvement, rather than a localized change.
   While the observations based on these experiments offer useful
insight into the processes likely to occur in reality, they do not
provide a strong basis for selecting specific transportation system
improvements.  They do suggest the merit in using simulation tools
to assess the energy impacts of any significant network
improvements prior to implementation.  This should be clear because
the energy consequences can sometimes be important and may
typically be counter-intuitive.  The latter is particularly true
because of the apparent, and sometimes countervailing, influence of
both congestion and VMT on overall transportation energy

D. Effects of Different Cluster Types

Figures 4-7 to 4-9 present the total energy, VMT, and highway

                                TABLE 4-6



                                   Work Trip Length                  
Ring of Nucleus    Basic, Population  Basic, Population      Percent
Location              Cluster         Cluster w/Impvts       Difference
   1                     2.77            2.78                + .4
   2                     2.78            2.73                1.8
   3                     2.71            2.77                + 2.2
   4                     2.78            2.85                + 2.5
   5                     2.71            2.69                0.7



congestion impacts of locating each of the three cluster types in
each of the five rings.  The size of the basic employment cluster
is always 1,500, leaving 2,250 additional basic jobs to be
distributed throughout the city as shown in Figure 4-10. (The
rationale behind the size of the basic employment cluster was set
forth previously in Chapter 3).  The number in parentheses in
Figure 4-10 represents the total basic employment of the base
together with incremental experiments.  As indicated in the
previous chapter, the size of the population and service employment
clusters cannot be controlled explicitly, but are implicit
functions of the attractor variables, among other things. 
Experiments were conducted with a range of attractor variables;
only one example from each cluster type (as explained in Chapter 3,
1,500 was the desired cluster size for residential and service
employment as well) at each location was selected for use in the
analysis.  These are presented in Tables 4-7 and 4-8.
   In Figure 4-7 it is apparent that the population and retail
clusters are more energy-efficient than either the S.I.R. or the
basic employment clusters.  In rings 3 to 5, the basic clusters are
less energy efficient than the S.I.R. One should note that the
S.I.R. is an example of a city-with a strongly centrally-oriented
basic employment cluster.  This suggests that centralizing basic
employment may be more attractive than decentralization, at least
in terms of transportation energy consumption.
   Also apparent from Figure 4-7 is that, at least in the cases of
basic employment and population clusters, rings 3 and 4 become
increasingly energy intensive locations while, for all three
cluster types,

Click HERE for graphic.

          Figure 4-7.  Total Energy of cluster Type Experiments


Click HERE for graphic.


Click HERE for graphic.

         Figure 4-10.    Distribution of Incremental Non-Clustered
                         Basic Employment


                                TABLE 4-7

                           POPULATION CLUSTERS

   Ring                  Experiment             Cluster Size

   1                         560C                  1,386

   2                         561F                  1,572

   3                         562E                  1,613

   4                         563C                  1,328

   5                         564C                  1,600

                                TABLE 4-8
                       SERVICE EMPLOYMENT CLUSTERS

   Ring                  Experiment             Cluster Size
                                             Total (Type N /Type S)

   1                         570C               1,254  (821/432)

   2                         571B               1,281  (878/403)

   3                         572A               1,545  (1013/532)

   4                         573A               1,478  (925/553)

   5                         574B               1,288  (810/478)

*  Type N service employment opportunities are jobs in service
   industries the markets for which are relatively insensitive to
   the location of those industries (e.g., appliance stores).  Type
   S opportunities are in industries the markets for which are
   sensitive to their location (e.g., fast food outlets, grocery


there is a subsequent decline in energy use in the fifth ring.  A
large decrease in energy consumption is observed when population is
clustered in ring 3.
   Figure 4-8 indicates that with VMT, as with total energy, ser-
vice employment and population clusters yield values less than
either the standard incremental run or the basic clusters.  Most
noticeably, both show a marked decrease in VMT for clusters in ring
3. Interestingly, VMT for a basic employment cluster in ring 3,
while greater than in ring 2, is less than the S.I.R., whereas its
total energy is greater.  This implies that the resulting spatial
pattern of activities which reduces VMT for the other clusters in
ring 3 also acts to some extent on the basic employment cluster. 
Otherwise, the trajectory of VMT for each cluster type is rather
similar to that of total energy, lower in the central city and at
the edges, and somewhat higher in rings 4 and 3.
   Figure 4-9 exhibits a similar pattern using the congestion index
as the measure of effectiveness.  Note that the ring 3 basic em-
ployment cluster shows higher congestion than for rings 1 and 2,
perhaps explaining the observed divergence between energy and VMT
for this experiment.  The service and basic employment curves are
similar in shape in that they both show a slight decrease from ring
1 to ring 2, large, steady increases from ring 2 to ring 4, and a
large decrease from ring 4 to ring 5. The congestion associated
with population clusters also increases from ring 3 to ring 4 and
decreases in ring 5, but also shows a major decrease in ring 3.
   The population and service employment distribution maps, Figures
4-11 to 4-13, help to give some insight into the increased energy


efficiency of clustering population and service employment relative
to clustering basic employment.  Figure 4-11 shows the percent
changes in population and service employment for each zone when
compared with the S.I.R. for clustered service employment in zone
36 (ring 4).  This service cluster represents almost 5 percent of
the total service employment, and thus leads to a decrease (more
correctly, smaller increase than the S.I.R.) in service employment
in every other zone in the city.  Because of the strong ties
between service employment and population in the model, this causes
a decrease in population in many zones as we In zones close to the
service employment cluster, the positive incentives for population
clustering around the large retail cluster are strong enough to
attract residents from the other zones.  This serves to induce a
"multi-use" nucleus which is strongly centered in zones near the
service employment cluster.  One other factor works toward spatial
confinement of this cluster.  Whereas for work purposes people are
willing to travel somewhat longer distances and times, there is a
preference for shorter trips when the purpose of the trip is to
shop.  This is reflected in the trip length (time and cost)
impedance factors in the model, which operationalize typical
traveler preferences in simulating the location of population with
respect to employment and services.  Hence, the area of impact is
less for a retail cluster than it might be for a basic employment
cluster because population and retail services tend to be located
nearer to each other.
   As noted above, clustering service employment in zone 36 causes
a retail "decline" in every other zone (where "decline" here means
less growth than the S.I.R.) of the city.  This potentially
negative economic

Click HERE for graphic.

         Figure 4-11:    Population and Service Employment
                         Clustering Around Service Employment
                         Cluster in Zone 36



impact may be more important to a community than the benefits due
to energy conservation.  Although only the transportation energy
impacts of various urban development strategies are being
considered here, it is important to remain aware that there may be
many other impacts due to the policies tested.  For example, there
may be other non-transportation related energy impacts, such as the
impact on residential, commercial, and industrial energy use. 
There may also be important transportation impacts unrelated to
energy consumption, such as microscale changes in accessibility,
effects on personal mobility, and goods movement impacts.  Finally,
there may be impacts unrelated to energy or transportation, such as
social, economic or environmental impacts (Carroll, et al., 1977). 
It is not within the scope of this project to determine just what
all of these impacts might be.  It is most important, therefore,
that the reader view the results with the proper perspective.
   The clustering of 1,328 residents in-zone 36 had relatively
minor impact on population distribution throughout the urban area,
as evidenced by Figure 4-12.  This is likely attributable to the
fact that the size of the population cluster is relatively small
(1.2 percent of total population).  This, in turn, limits the
service employment clustering to zones near the residential cluster
in zone 36.  Hence, the spatial spread of clustering effects is
even more confined than in the case of the service employment
cluster; further, for the residential cluster, the overall
magnitude of locational changes is less than for the service
   The population and service employment distribution in response
to a basic employment cluster in zone 36 is shown in Figure 4-13. 
In this

Click HERE for graphic.

      Figure-4-12:    Population and Service Employment Clustering
                      Around Population Cluster in Zone 36


Click HERE for graphic.

      Figure.4-13:    Population and Service Employment Clustering
                      Around Basic Employment Cluster in Zone 36



situation, the basic employment cluster causes a cluster of
population and, consequently, service employment in and near zone
36.  This clustering is self-proliferating and leads to further
clustering in adjacent zones more remote from zone 36.  Hence, the
clustering of population and service employment is spread over a
much larger area as a response to a basic employment cluster than
in the cases of population or service employment clusters.
   Although this provides a reasonable explanation as to why the
population and service employment clusters have similar impacts
which are different than the impacts from the basic employment
clusters, it does not, of itself, explain why those clusters are
better, i.e., more energy efficient'.  The reasons for this appear
to be twofold.  First, the transit ridership in the more spatially
confined population and service employment cluster experiments is
higher than in the basic employment cluster experiments (see Table
4-9).  This causes a reduction in the number of auto work trips. 
Second, there  is a slight, though relatively consistent, reduction
in average trip length in the population and retail employment
cluster experiments.  This would be expected since the clustering
is denser than in the basic employment runs due in part to the fact
that retail clusters demand shorter population retail trip lengths
than basic employment clusters due to the nature of the travel
impedance functions utilized.  The combination of these two factors
acts to reduce total VMT and overall network congestion and,
consequently, the total energy.

E. Effects of Cluster Location

   Figure 4-7 indicates that total transportation energy consumed

                                TABLE 4-9

                             a.  % of Total Work Trips on Transit

   Location of                     Cluster Type
   Cluster (Ring)     Basic        Population      Service

   1                  17.74        18.22           19.53
   2                  16.33        18.17           18.43
   3                  15.53        18.40           17.63
   4                  15.28        17.91           17.50
   5                  14.99        17.76           17.49

      b.  Auto Work Trip Length

   Location of                     Cluster Type
   Cluster (Ring)     Basic        Population      Service

   1                  2.77         2.75            2.70
   2                  2.72         2.73            2.69
   3                  2.74         2.69            2.67
   4                  2.78         2.74            2.74
   5                  2.71         2.75            2.64



in a city is lowest for any type of cluster when that cluster is
placed in the center (i.e., in rings 1 and 2).  The only exception
to this is a population cluster in zone 18 (ring 3), an analysis of
which will be presented later.  The principal cause for this energy
efficiency seems to lie in the low highway congestion indices (see
Figure 4-9).  Hence, despite sometimes higher levels of VMT for
centralized clusters, the lower highway congestion more than
compensates to yield low levels of total energy consumption. 
Again, a contributing factor to the low highway congestion is the
higher transit ridership observed with clusters in central zones
(see Table 4-9).  The resultant decrease in the number of auto work
trips tends to limit peak hour congestion, this being the critical
energy conservation role of (CBD-oriented) transit as identified by
Peskin (1977).  Furthermore, excepting the ring 3 retail and
population clusters, even the VMT trend is upward as clusters are
decentralized.  This is intuitively satisfying in that any off-
center cluster should tend to produce more circuitous travel due to
the asymmetry of the resulting activity distributions.  This VMT
impact of asymmetry should be less where the cluster itself has
sufficiently large impact on other land uses to create a new
symmetry of its own.  This may help to explain the lower VMT in the
off-center retail and population clusters compared to the basic
employment cluster.  The large VMT increases for ring 4 clusters
may reflect a maximum-asymmetry impact;. the decline in VMT in ring
5 (except residential) may reflect the tradeoff between an
asymmetrical travel pattern (e.g., higher VMT, energy) and the fact
that the farthest off-center clusters are limited in their impacts
by "edge-effects." That is, they simply draw less travel


because (in this model) they are linked to the rest of the city on
only one side.
   The decrease in total energy consumption as clusters are shifted
from ring 4 to ring 5 can be attributed to this edge-effect and to
the artificial constraint that is imposed on urban sprawl by the
model.  Because activities cannot locate beyond ring 5, the impacts
on VMT and congestion are sharply and unnaturally curtailed.  In
most real cities, activities attracted to edge-clusters could
locate beyond the boundaries of the currently-developed city,
leading to increases in VMT and transportation energy.  This may
lead to a continuation of the upward trend in energy consumption as
clusters shift from rings 1 and 2 through ring 5. It is possible
that energy consumption increases for ring 5 (and beyond) clusters
would be limited due to low levels of congestion in "deep suburban"
areas, but more likely, severe sprawl would increase trip lengths
and thus increase energy consumption.
   This suggests that if a city can control sprawl at some artifi-
cial boundary, then clustering in the outer ring is generally
preferable to middle ring clusters in terms of energy efficiency. 
If urban sprawl cannot be controlled and the city is allowed to
spread beyond its initial boundaries, it is likely that the implied
energy gains would not be achieved.  The overall trend in the
relationship between energy consumption and cluster location might
appear as shown in Figure 4-14.

   F. Effect of Density Constraints

   As noted, the population cluster located in the third ring
yields some apparently inconsistent results in team of the trend in
energy consumption and congestion with increasingly off-center

Click HERE for graphic.

         Figure 4-14.    Energy/Cluster Location Relationship With
                         and Without Control of Sprawl



These inconsistencies cannot be fully explained, and may be
attributable to an anomaly in the model or data since other ring 3
population experiments with only slightly larger clusters yield
more regular results (see Appendix A, experiments 562A, 562B). 
These are shown on Figures 4-7 and 4-9 as "*".  Nevertheless, at
least the third ring effect on VMT for each of the cluster types
can be explained logically with respect to the constraints applied
to the model.
   Figure 4-15 shows the population and service employment shift
from the S.I.R. to experiment #520, the basic employment cluster in
zone 2 (ring 1).  What is apparent is that the population
clustering and, consequently, the service employment clustering in
the adjacent ring 1 and ring 2 zones is quite weak.  Some
clustering is observed in rings 3 and 4, but this follows an
irregular pattern and is already at considerable distance from zone
   The cause for this pattern is that the density constraints are
prohibiting any further population from entering either rings 1 or
2. The density constraints of 12.5 persons/acre in ring 1 and 18.8
persons/ acre in ring 2 are binding in the base run, which
restricts any further growth in these rings in the incremental
runs.  This is not unrealistic if one accepts the notion that, in a
city of this type, the central area has little vacant land and will
not undergo sufficient redevelopment over a short time period to
change the density pattern dramatically.  An alternative philosophy
argues that, whereas density constraints should be utilized to
develop a representative base city, they could be removed in the
incremental run to enable the model to estimate a longterm
distribution of land uses in an unrestricted manner.  A conceptual

Click HERE for graphic.

         Figure.4-15:    Population and Service Employment
                         Clustering Around Basic Employment Cluster
                         in Zone 2


comparison of the constrained and unconstrained density gradients
is suggested in Figure 4-16.  It is the first of these notions,
that central area density should be constrained, that was adopted
from the start in this research.
   Although there is some trend in large cities today toward new
high density residential communities in the center of the city,
this is still more the exception than the rule, especially in the
smaller cities being considered here.  For the purposes of this
research, it was felt that a stable downtown residential and
employment picture, implying little capital redevelopment, would be
a more likely future scenario.  Obviously, this was a critical
decision which has an important impact on the results.  A
comparison of population density constrained and unconstrained
standard incremental runs (experiments 514 and 996) in Table 4-10
bears this out.  The unconstrained S.I.R. shows large decreases in
each of the three key measures.  VMT decreases by 11,000 miles as
more people live and work near the center of the city.  Highway
congestion decreases 21.0 percent as a result of these shorter
trips and higher transit ridership for work trips.  Consequently,
it is no surprise that energy consumption in such a city is
decreased considerably.  It is likely,.as well, that the growth
policies like those applied throughout these experiments would have
greatly attenuated effects, since it would be difficult to convince
many in this huge block of centrally located people and jobs to
move to the periphery.
   Returning to the consideration of ring 3 clusters, the following
analysis is logical.  Because of the density constraints,
clustering does not occur near activity nuclei in ring 1 or ring 2,
leading to artificially longer trip lengths, which cause VMT to
rise higher than


Click HERE for graphic.

         Figure 4-16.    Density Gradients for Constrained and
                         Unconstrained Central Area Development


                               TABLE 4-10
                        STANDARD INCREMENTAL RUNS

                             Constrained        Unconstrained
                                (#514)             (#996)          % 

   Total Energy                 81,865             74,953       -  8.4

   Work Auto Energy              7,693              6,852       - 10.9

   VMT                         454,501             443,770      -  2.4

   % Work Trips on Transit       17.82               18.39      -  3.2

   Highway Congestion           1,663              1,500        - 21.0

   Work Auto Trip Length        2,771              2,659        -  4.0


expected.  When the cluster is centered in ring 3, where the
density constraints are not binding, clustering takes place in the
zones surrounding the nucleus zones, thus (relatively) shortening
trip length and lowering VMT.  Thus it is expected t hat a series
of unconstrained experiments would result in the VMT in rings 1 and
2 being less than in ring 3. It might be expected that the denser
clustering toward the center of the city would increase the
congestion in that area.  While this may be the case, the data in
Table 4-4 suggest that the improvements in highway congestion in
rings 3 and 4 will more than compensate for the increases in the
inner rings to yield an overall decline in highway congestion.  The
likely effect on total energy is presented in Figure 4-17.  The
combination of decreases in VMT and congestion may serve to
decrease dramatically the total energy consumption in rings 1 and
2. Such results would serve to rationalize the clustering
experiments described in the preceding sections.

   G. Integrated Cluster Development

   Up to now, each experiment has concentrated on only one type of
cluster development, either basic employment, residential, or
service employment.  The unintentional combination of clusters,
however, as in the case in residential clustering around service
employment-and vice versa, has been shown to be quite energy
efficient.  Intuitively, the greatest efficiency should lie in the
combination of all three cluster types into one unified cluster
   It was decided that such a cluster should include about a
thousand of each type of development (i.e., 1,000 basic employment
jobs, 1,000 service employment jobs, 1,000 residents).  Once again,

Click HERE for graphic.

         Figure 4-17.    Energy Consumption for Clusters With and
                         Without Density Constraints



accomplish this, it was necessary to experiment with the attractor
variables and density constraints to achieve the desired cluster
size, as shown in Table 4-11.  The total energy, VMT, and highway
congestion results are presented in Figures 4-18 to 4-20,
superimposed over the same results from each individual cluster
type.  Surprisingly, the integrated clusters, while an improvement
(especially in rings 3-5) over the basic employment above, are not
as effective in reducing VMT, highway congestion, or energy
consumption as are the residential or service employment clusters
   This immediately suggests some possible upper bound to cluster
size, i.e., that 3,000 is too big.  This is not supported by other
experiments with various cluster sizes.  More likely, it implies
that the integrated cluster is being dominated by the basic
employment and is subject to the same inefficiencies common to that
cluster alone.  The basic employment clusters affect a wide area of
the city, causing increased trip lengths, reduced transit
ridership, etc.  This implication is generally supported by the
figures in Table 4-12.

H. Cluster Size

   The implication that there may be a maximum desirable or optimum
sized cluster development (in terms of energy efficiency) warrants
further consideration.  Several cluster sizes were tested in
designing each cluster type, and the results of these "sizing"
experiments were used to determine whether there might exist a
preferred size range in terms of energy consumption.  Figure 4-21
presents the graph of residential cluster size vs. total energy,
and represents, in the hatched area, the most distinguishable trend
in any of the graphs.  Here, the intensity

                                    TABLE 4-11
                                INTEGRATED CLUSTERS

                                         Cluster Sizes

Ring  Environment#  Basic Employment  Population  Service Employment  Total

1        600          1245               725             1142      3112

2        601          1000               746             746       2492

3        602          1070               917             820       2807

4        603          1000               913             870       2783

5        604          1000             1066              871       2937


Click HERE for graphic.

         Figure 4-18.    Total Energy of Integrated and Cluster Type


Click HERE for graphic.


                               TABLE 4-12

                      a.  % of Total Work Trips on Transit        

Location of                        Cluster Type
Cluster (Ring)  Basic    Population      Service      Integrated

   1            17.74        18.22       19.53        18.31

   2            16.33        18.17       18.43        16.48

   3            15.53        18.40       17.63        15.26

   4            15.28        17.91       17.50        15.11

   5            14.99        17.76       17.49        14.89

                             b.  Auto Work Trip Length             

                                   Cluster Type              
Location of
Cluster (Ring)  Basic    Population      Service      ntegrated

   1  2.77  2.75   2.70  2.74

   2  2.72  2.73   2.69  2.72

   3  2.74  2.69   2.67  2.71

   4  2.78  2.74   2.74  2.77

   5  2.71  2.75   2.64  2.64


Click HERE for graphic.

Figure 4-21.  Residential cluster Size and Energy Consumption



of energy use appears to decline with clusters ranging in size from
425 to 2,500.  With larger clusters, there are insufficient data to
draw any conclusions, but energy use seems to be increasing.
   The trend does not appear to be particularly strong for this
cluster type, and-it was even less evident for other cluster types. 
While intuitively, cluster size should influence transportation
energy consumption systematically, these experiments do not confirm
this.  On the one hand, this may be because clusters of any size
have about the same energy efficiency.  On the other hand, the fact
that the model ignores the congestion due to intrazonal travel may
account for this outcome.
   The-graph of integrated (basic employment, residential, and
retail) cluster sizes vs. total energy yielded no trend.
The fact that even excessively large clusters are not any more or
less energy intensive is difficult to explain but suggests that
perhaps the clusters of this type maintain a self-sufficiency that
does not alter travel characteristics.  Nonetheless, the results of
this analysis fail to provide a reasonable basis for a-specific
conclusion regarding the link between off-center cluster size and
energy consumption.  It should be noted that, as shown in Figure 4-
7, larger CBD clusters (the S.I.R. vs. clusters of all types in
ring 1) do produce important reductions in energy consumption.

   I. Comparative Roles--Transportation Improvement vs. Land Use

   As pointed out earlier in this chapter, the introduction of ring
roads alone produced large energy savings.  Figures 4-22 to 4-24

Click HERE for graphic.

         Figure 4-22.    Total Energy of Experiments with Ring Roads
                         and Integrated Clusters


Click HERE for graphic.



the effects of adding the ring roads to the integrated cluster
developments.  Both ring roads yield improvement in energy
efficiency (Figure 4-22) over the integrated clusters alone, but in
none of the experiments is the combination of ring road and cluster
as efficient as that ring road itself.  Again, clusters in rings 1,
2, and 5 are the most energy efficient, while clusters in rings 3
and 4 show higher values for total energy.  Each highway congestion
curve is similar to its energy curve.  Again, each is an
improvement over the clustered development without the ring roads,
but neither is-as uncongested as the ring road alone.
   The VMT impacts (see Figure 4-23) show that the integrated
cluster developments alone yield VMT values generally lower than
the S.I.R. while the clusters together with the inner ring road
yield values of VMT greater than the inner ring road alone (and the
S.I.R. as well).  Once again, clusters with the outer ring road
yield the lowest values' of VMT.
   The fact that the ring roads when combined with the integrated
clusters are more energy-efficient than the clusters alone but less
efficient than the free clustering associated with the introduction
of ring roads alone is important but not surprising.  Peskin &
Schofer (1977) found that the addition of diagonal freeways or
arterials produced large energy savings.  When combined with
directed corridor growth along these routes, the energy savings,
while still large, were not. as large as with the route
improvements alone.  Thus, the precedent exists for this kind of
response from the model.
   This may suggest, then, that urban development directed only by
the careful location of appropriate urban transportation facilities


may yield a city with more overall energy efficiency than any in
which development is carefully directed toward strong clusters. 
The structure of the simulation model used in this research permits
the various land uses (except basic employment) to arrange
themselves in order to reduce transportation costs.  The ring roads
produce energy efficient patterns because (1) they reduce
congestion where it is quite significant, and (2) they encourage a
general shift in activity patterns to allow land uses to take
advantage of the improved transportation service.  Any form of
clustering (as tested in this research) produces an asymmetrical
pattern and seems to restrict the most efficient arrangement of
activities with respect to transportation.  This theory may be
borne out by the fact that the outer ring road proved so much more
energy efficient than the inner ring road.  Developmental shifts in
response to the latter were somewhat restricted by the a priori
population density constraints.  The shifts due to the former could
provide increased benefits because of the already low level of
congestion at the edges of the experimental city.
   Thus, it may be profitable to direct additional efforts toward
identifying those network improvements which are most effective in
different types of cities.  For instance, from this research it
could be suggested that for a diamond-shaped concentric-ring city
similar to the one that is being considered, an outer ring road
seems to be most effective.  Diagonal arterials and freeways and an
inner ring road also appear promising.  Other network improvements,
however, can be identified as insufficient and counterproductive,
i.e., single lanes added to all links in and out of zones with
basic employment and population clusters, as in experiments 550-
554, covered earlier in this chapter.  As is clear there,


these "improvements" actually decreased energy efficiency.
   Generally, the beneficial effects of such highway improvements
may be achievable through modest, TSM-type changes in existing
roadways.  For example, the outer ring road tested here had a
capacity of 2,000 vehicles/hour, increased from 1100 vehicles/hour. 
These improvements may be achievable in many cities without major
investments.  The question that remains is whether the impacts of
such highway improvements agree with intuitive judgment.  The
proliferation of urban highways combined with ineffective or non-
existent controls on development appear to h ave led to the energy-
inefficient urban sprawl so prevalent in American cities today.
   There are several important-differences between the model and
the real-world situation which must be considered prior to adopting
the promising transport policies identified above. -First, the edge
of the city acts as an artificial boundary in the model,
restricting urban sprawl.  In a real city, certain major
transportation improvements may lead to a spreading of the city,
and thus to increased energy consumption.  Secondly, the model does
not consider social differentiation in the location of households. 
Since it assumes a homogeneous population, all households are
satisfied to locate near any other household.  In real cities,
social and economic segregation restricts many households to
specific locations, a fact that not only must reduce the energy
efficiency of those cities, but also is likely to reduce the
energysaving benefits of the transportation improvements tested
   In response to the first of these points, the original
implication might be reformulated to state that urban development
directed only by


the careful location of appropriate transportation improvements,
given effective controls against urban sprawl, may yield a city
with greater overall energy efficiency than any in which
development itself is directed into dense nuclei.
   If it is assumed that congestion and VMT are the two primary
inputs in determining total energy, it is apparent that congestion
effects exert the stronger influence in the case of these graphs
(Figures 4-22 to 4-24), as they have throughout the experiments. 
This suggests the important conclusion that reduction in congestion
is a critical factor in reducing energy consumption.. This agrees
with the conclusion reached by Peskin & Schofer (1977) as a result
of their research (p. 172).  Nevertheless, VMT is an important
factor which acts together with congestion in the determination of
total energy consumption.- The task for planners and decision
makers is to find key needs for congestion reduction which, if met,
will not encourage major increases in VMT and sprawled development.


References to Chapter IV

1. Robert L. Peskin and Joseph L. Schofer, The Impacts of Urban
   Transportation and Land Use Policies on Transportation Energy
   Consumption, Report No. DOT-05-50118, U.S. Department of
   Transportation,-Office of University Research, April 1977.

2. T. Owen Carroll, Robert Nathans, and Philip F. Palmedo, Land Use
   and Energy Utilization, final report prepared by the Brookhaven
   National Laboratory/State University of New York Land Use-Energy
   Utilization Project for the Office of Conservation and
   Environment, General Energy Administration, BNC Report 50635,
   June 1977.

                                CHAPTER 5

                       CONCLUSIONS AND GUIDELINES

   This chapter serves to (1) review the limitations of the ap-
proach taken in this research; (2) summarize the major findings;
and (3) offer guidelines to support planners and decision-makers in
urban development and infrastructure decisions in light of the
likely transportation energy consumption outcomes.

   A. Limitations of the Research Approach

   The simulation methodology used in this research is of an ag-
gregate nature.  As such, all locators (e.g., households,
employment opportunities) and potential locations (including both
housing-stock characteristics and locational amenities) are
considered to be homogeneous.  This means, for example, that in the
simulation process, any household will be satisfied with the social
and physical amenities of any location in the city, provided that
that location meets the travel requirements of that household. 
Furthermore, it is assumed that the travel requirements of all
households are identical.  This means that the simulation process
does not account for preferences of members of like socioeconomic
groups to live near each other, or to live far away from members of
unlike socioeconomic groups.  It implies that all workers bid with
equal effectiveness and resources in the housing market, which
doesn't reflect the fact that higher income groups can often meet



of their locational needs in the household-choice process than can
members of lower income groups.
   In reality, non-transportation-related social and economic
factors, as well as attributes of the physical environment, may be
of considerable importance in the location decision, in particular
the choice of a residential site.  The result of this limitation is
that the model system produces a city which is likely to be much
more efficient in a transportation-energy sense, since activity
locations are entirely dependent upon transportation-related
factors.  While the residential attractor variables and density
constraints do have an influence on the operation of the model,
they can reflect socioeconomic realities of a city only in a
limited sense.  Although the energy efficiency of the simulated
city is likely to be overestimated, it seems reasonable to suggest
that the relative change in-energy consumption as a function of
cluster location and transportation policies may well be indicative
of the true impacts of such alternatives.
   The simulation approach is also limited because historical ob-
servations of travel behavior have been used to develop model-
calibration parameters.  Thus, it is assumed that travel behavior
characteristics do not shift in response to changes in the price
and availability of energy or to new trends in urban spatial
structure.  While this is a general limitation which faces most
travel-demand estimation work, it may be an important consequence
in this research if residents of a community find increasingly
effective and creative ways to overcome the constraints of a
limited energy future.
The modeling system itself represents a considerable simplification

of a complex process.  The degree of change permitted in the land-
use pattern during the 25% growth increment was selected to insure
'some stability while permitting a reasonable amount of change. 
The process chosen to represent this mixture of behaviors is at
best arbitrary.  Similarly, the interactions between travel-
behavior decisions and location choices have been specified to be
logical, but are not based on well established theories.
   The simulated city has a limited number of zones, which re-
stricts the level of detail in the model.  Furthermore, it has a
sharply defined edge, precluding development beyond the pre-
established boundary.  This artificial constraint results in an
unrealistic treatment of the potential for urban sprawl.  Any
attempt to expand this boundary, given the limitation on the number
of zones, would result in additional loss of detail throughout the
city, especially in the CBD.
   The iterative procedures built into the modeling system are
subject to questionable closure rates and oscillation.  In order to
promote efficiency in the computer solution, it is possible that
some stability-and possibly some accuracy--is sacrificed by
limiting the number of iterations in advance of the solution
process.  There are, however, several forces which act to attenuate
the impacts of these oscillations.  The oscillations occur
principally in the network assignment algorithm where, since there
is conservation of overall travel, an increase in the volume of
travel on one link will lead to the decrease on some other link. 
Furthermore, the consequent increase or decrease in congestion on
the impacted links is likely to be similar, since most links with
extreme reactions-that is, large increases in congestion with only
a small volume increase--have been eliminated in the first couple


iterations.  Finally, confidence in the final experimental results
should be strengthened by the fact that most of the trends reported
occur repeatedly in different experiments throughout the
experimental program.
   Because of the nature of the modeling system, this research has
not attempted to assess the consequences of infrastructure
investments and spatial alternatives in the broad framework which
would be necessary in the face of such decisions in a more
realistic context.  The social, economic, environmental, and
distributional consequences of such actions are likely to be of
considerable importance, and their significance is likely to vary
from city to city.  This places some important limitations on the
generalizability of results of the-research presented here.  Thus,
planners and decision-makers who wish to implement some of these
findings should develop more detailed, locally-specific analytic
approaches to studying specific actions in their own communities. 
For example, it may be useful to pursue more detailed travel-
simulation efforts to evaluate development proposals; furthermore,
quantitative and qualitative analyses of consequences other than
those related to energy must also be conducted.

   B. Conclusions

   Results of the simulation studies suggest that some approaches
to urban growth nucleation offer the potential for reducing
transportation energy conservation relative to non-clustered,
trend-line growth patterns.  In virtually every experiment, the
lowest total energy consumption resulted when growth nuclei were in
the first two or three rings.  These clusters were characterized by
low congestion levels, higher transit ridership, and lower values
for vehicle-miles of travel;


all of these descriptors are relative to the standard incremental
rum Retail employment clusters seem particularly energy-efficient,
but the promise for centralized population clustering appears even
greater.  This is indicated by the results of the single run
attempted without the central-area population-density constraints,
which resulted in much lower energy consumption values than those
of the other experiments in this series.
   Retail employment and population clusters are more energy-effi-
cient, independent of location, than are basic employment clusters. 
This appears to be because basic employment clusters have an impact
area far greater than that of the other two cluster types.  This-is
partly because of the long trip lengths associated with basic
employment centers.  It results in lower transit ridership and
higher values of vehicle miles of travel.  Congestion levels also
seem to be relatively high near basic employment clusters. 
Furthermore, the basic employment cluster seems to become even less
energy efficient as it is moved away from the center of the city. 
This off-center movement seems to have an important effect on
reducing the symmetry of urban structure, and thus of the travel
patterns in the city.
   The integrated development cluster seems to behave similarly to
the basic cluster, probably because the spatial and travel impacts
are dominated by the off-center basic employment nucleus.
   While edge locations for all cluster types seem to be relatively
efficient in terms of transportation energy consumption, this is
interpreted to be an artifact of the simulation model.  The model
does not permit development to occur beyond the established
boundary of the city, and thus cannot truly reflect the urban-
sprawl implications of such edge

locations for growth nuclei.  This suggests that energy consumption
can be expected to increase as clusters are moved farther from the
central area, unless some means is found for controlling the sprawl
effects.  Without the control of urban sprawl, a development
cluster in ring 5 might be expected to generate growth a
considerable distance outward from the city, since the travel
impedances on links in that region would be quite low.  Such an
edge cluster, then, might be expected to generate a much more
spread-out city than produced in these simulations.  If basic
employment were not permitted to locate beyond ring 5, the
structure of the model suggests that some finite boundary for the
region would be reached, beyond which, due to the trip cost
preference factor and distributions, no one would choose to
locate.' Nonetheless, given a choice, it may be easier to encourage
centralized locations than to discourage second-order sprawl
   Notable reductions in transportation energy consumption can be
achieved through the careful selection of simple, TSM-type highway
improvements.  This is a useful finding, in that it suggests that
planners ought to consider the energy implications of small-scale
improvements, on the one hand, while also looking for ways to save
energy by selecting the most effective TSM options.
   The major factor which should be considered in the selection of
these improvements is congestion reduction.  These simulation
studies have indicated that both congestion and vehicle-miles of
travel are relatively independent determinants of transportation
energy consumption.  This supports a similar conclusion reached by
Peskin & Schofer (1977).  In this study, the congestion index seems
to be even more highly correlated


with transportation energy consumption than it did in Peskin's
work.  This suggests that using VMT alone as an indicator of the
potential for energy saving is insufficient.  A much more
productive approach seems to involve searching out areas where
congestion is the highest, and finding ways to remedy those
   It is also clear that not all highway improvements save trans-
portation energy.  This is especially evident in the experiments
which added a single lane of capacity to -existing links entering
growth nuclei, where congestion was relatively high.  Such
improvements appeared to promote a sufficient amount of additional
travel, as indicated by VMT, such that gains in energy efficiency
due to congestion reduction were more than offset by increases in
the total amount of travel.
   Some kinds of highway improvements seem to encourage urban
sprawl, and thus to produce a negative effect on energy
consumption.  This was particularly evident with the axial freeways
emanating from the central business district.  These facilities
encouraged the movement of population away from the downtown,
increasing vehicle-miles of travel to such an extent that
congestion improvements were overshadowed.
   On 'the other hand, paralleling the findings of Peskin, diagonal
facilities originating in the CBD did produce reductions in energy
consumption.  This was attributed to the fact that such facilities
added an important new element to the transportation network,
rather than simply reinforcing an existing element.  In addition,
the energy savings achieved by adding diagonal arterial roadways
was found to be greater than the effects of adding diagonal
freeways.  Here is a case where a lower-cost solution seems to
produce a more promising result.  This is probably due to the
smaller degree of encouragement of land-use


decentralization which results from the arterial streets compared
with the freeways.
   The ring-road experiments conducted in this investigation
suggest a similar conclusion.  The improvements tested were rather
modest, amounting to small-scale TSM-type improvements to existing
facilities.  Still, the energy savings which resulted were
impressive.  Indeed, each of the ring roads-but especially the
outer ring road-produced particularly energy-efficient results
relative to other experiments.  This was due in part to congestion
reduction, but in addition the vehicle-miles of travel were
effectively controlled by these transportation actions because of
natural clustering around the ring roads which occurred in the
simulation process.  In effect, each of these facilities created
its own travel market, diverting an important amount of travel from
an in-out radial flow pattern to a circumferential flow pattern
carried on more efficient facilities.
   The ring-road experiments also indicate that natural clustering,
resulting from the logical response of locators to changes in the
quality of service on the transportation network, can produce more
energy-efficient spatial forms than might be achieved by an
organized clustering process which occurs due to local urban
policy.  To achieve such natural clustering, of course, it would be
necessary to relax some constraints while tightening others which
influence the location of activities within an urban area.  While
this offers a desirable path toward controlling transportation
energy consumption through infrastructure investments, it requires
a highly integrated policy approach affecting both land-use and
transportation investments.


C. Guidelines for Improving the Transportation Energy Efficiency of

    Because of the limitations of the simulation system used in this
research, and because of the unique attributes of each community,
translating the analytical results of this work to specific
situations is not a promising course of action.  Still, the general
trends in the results suggest that certain actions may be
profitable in a transportation-energy sense for most communities.
   In particular, the logic of encouraging central-area development
seems to be particularly strong.  A variety of actions seem
promising in this regard.  For example, any actions which divert
new growth from far-off-center locations toward the central area
are likely to produce energy benefits in the-long term.  Such
actions may include governmentoriginated, policy-oriented zoning
changes and more careful evaluation of private sector requests for
zoning changes which would facilitate far-off-center growth.  In
addition, fostering redevelopment in the central area, either
through larger-scale densification or through the creation of
growth clusters near the downtown, seems to be an attractive
option.  Such actions may include formal redevelopment efforts, but
they may also involve providing general economic incentives,
information, and encouragement to developers who have not yet
chosen a location for their investments.  In addition, maintaining
the quality of the central area transportation system will be an
important factor not only for direct energy conservation, but also
for encouraging more centralized urban growth.  This includes
maintaining the viability of CBD-focused transit systems, as well
as TSM-type investments which serve to reduce central-area


   A variety of actions which may serve to enhance the economic and
social viability of the central area may also be warranted.  These
include aesthetic improvements, actions to control and reduce
crime, and efforts to improve central-city school systems.  It is
apparent in many urban areas that the major force behind the flight
of population to the suburbs is the inadequacy of inner city
schools.  While this is an issue far afield from transportation and
land-use planning, its long-term consequences for transportation
energy consumption seem quite negative, and thus the quality of
schools and other public services must be considered in attempts to
improve the long-term energy efficiency of urban areas.
   Along with all of these inducements for central area
development, it would also be necessary to consider the question of
equity; that is, what would happen to the present
characteristically lower-income inner-city residents?  While, in
the short-term, there would probably be sufficient unused or
underused land to meet the increased central area demand for space,
such developments would tend to increase the value of the land sur-
rounding them.  Higher rents, if they were allowed to go unchecked,
would tend to force many of the poor residents to seek new, less
costly, housing.  To combat this problem, municipal governments may
choose to funnel some of the "profits" resultant from the new
inner-city growth into rent-subsidy programs or other appropriate
means through which they could make sure that these poor residents
would still have access to affordable housing within the city.
   Non-centralized basic employment nuclei should be evaluated with
considerable care.  These seem to produce important increases in
transportation energy consumption; where options exist for
encouraging such developments to occur in the central part of the
city, they should be pursued vigorously. Indeed, efforts are
warranted to explore direct disincentives


to discourage far-off-center developments of any type.  It would
seem worthwhile for local governments to conduct a formal energy
analysis of such developments prior to granting required zoning and
building permits, so that, at the very least, an understanding of
their energy implications can be introduced into the decision-
making process.  Such a formal energy analysis would be a desirable
part of any routine process for making changes in zoning
   The potential implications of sprawl-type urban development-, at
or beyond the fringe of the existing developed area in a community,
seem quite important.  Land-use planners have developed and
attempted to apply arguments against urban sprawl for many decades. 
The current transportation energy situation introduces a new, and
potentially more important, argument against sprawl.  While it is
not obvious that there are easy ways to control this development
trend, the strong motivation for such control provided by the
energy situation should be considered in local decision-making. 
Disincentives such as zoning and building permit processes, as well
as incentives, such as the clearance and redevelopment of central-
city land, may be useful in combination to discourage urban sprawl. 
Still, the current structure of urban government makes the
effectiveness of such controls doubtful.
   In particular, the fact that most urban regions are made up of a
multiplicity of governments, each financed through real estate,
sales, employment and other taxes dependent upon the level of local
development, provides a strong incentive for each jurisdiction to
try to attract the maximum level of development which it can
support.  The result is a strong competitive atmosphere between
central-city and suburban governments.

When the degradation of public services and building stock in the
central city is considered within this atmosphere, the
attractiveness of suburban locations becomes particularly strong. 
Efforts to conserve transportation energy through more centralized
development in this context may be fruitless.
   This suggests a need to give serious consideration to some re-
structuring of urban governments so that regional development
decisions can be made which take cognizance of the national concern
for energy conservation.  For example, some ways to share tax
revenues among communities as a function of need rather than simply
as a function of local tax-revenue-generating capability may remove
or reduce the suburban incentive to attract development.  A more
radical solution may involve mandatory regional decision-making for
certain kinds of choices, particularly those which affect the
overall spatial pattern of an area.  At the highest level, this
problem may be approached by moving towards formalized regional
government structures, such as those existing in one form or
another in such cities as Miami, the Twin Cities area,
Indianapolis, or Portland.  The ability of such innovative
governmental structures to implement serious land-use controls
should be monitored to assess the efficacy of governmental reform
as a means of long-term transportation energy conservation.
   Finally, communities should pursue selected transportation im-
provements which reduce congestion but do not encourage more
decentralized development and increases in vehicle-mile of travel. 
The results of this research suggest that even modest improvements
in the quality of transportation services in communities should be
assessed for their

energy implications.  Such implications are often counter-intuitive
and revealing them requires careful analytic studies.
   Certain structural additions to the highway network seem to have
the ability to reduce congestion without producing counter-
effective land-use consequences.  These include central-area-
roadway improvements and the development of circumferential
facilities which do not encourage the spread of urban development.
   At the same time, efforts must be-advanced to maintain the qual-
ity of any strong mass-transit services which exist, and to
introduce centrally-oriented, peak-hour services in communities
which have none at this time.  Transit itself will not solve the
transportation energy problem, but the absence of transit, or a
weak transit system, is likely to exacerbate the problem, even in
the short term.
   The spatial pattern of a city seems to have an important
influence on its consumption of energy for transportation.  A
community does have some ability to control its land-use pattern;
and appropriate use of this control, as well as expansion of this
control in a positive sense, is likely to produce reductions of
transportation energy consumption on the order of 5 to 10 percent
in the coming years, independent of the nature of future
transportation technologies and their energy efficiencies.


Reference to Chapter V

1. Robert L. Peskin and Joseph L. Schofer, The Impacts of Urban
   Transportation and Land Use Policies on Transportation Energy
   Consumption.  Report No. DOT-05-4011B, U. S. Department of
   Transportation, Office of University Research, April 1977.

                                    APPENDIX A

                              RESULTS OF EXPERIMENTS

      Total Energy           Highway            Work Trip       Cluster Size
Run      (gals)      VMT   Congestion Index     Length (miles)  (When Applic.)

513   58,901       351,205      1.305              2.648

514   81,865       454,501      1.633              2.771

520   80,764       453,208      1.677              2.770
521   80,335       453,613      1.656              2.718
522   84,691       454,130      1.798              2.738
523   85,003       457,414      1.931              2.783
524   83,070       455,449      1.715              2.709

530   80,146       453,835      1.670              2.770
531   84,262       455,735      1.761              2.784
532   84,530       452,567      1.764              2.711
533   85,528       456,490      1.869              2.778
534   83,144       454,726      1.706              2.705

541   76,111       505,817      1.487              2.857
542   72,748       449,713      1.563              2.520
543   73,955       441,925      1.535              2.597

547   75,275       454,435      1.377              2.755
548   69,716       445,906      1.294              2.549

550   82,918       454,561      1.710              2.782


                              APPENDIX A (continued)

      Total Energy           Highway            Work Trip       Cluster Size
Run      (gals)      VMT   Congestion Index     Length (miles)  (When Applic.)

551   80,620       454,630   1.645              2.733
552   85,539       456,021   1.771              2.769
552B  80,385       457,267   1.478              2.781
553   87,024       459,496   1.839              2.851
553B  74,910       450,633   1.395              2.613
554   81,392       453,638   1.657              2.687

560A  78,289       451,157   1.553              2.719              425
560B  79,771       452,241   1.573              2.746              1060
560C  78,317       451,686   1.548              2.752              1386
5600  76,122       449,746   1.505              2.706              1970
561A  78,109       452,338   1.530              2.717              3361
561B  76,329       449,835   1.516              2.690              3289
561C  75,327       449,823   1.496              2.686              2512
5610  77,734       451,230   1.575              2.711              1841
561E  78,547       451,131   1.540              2.706              1266
561F  78,311       451,130   1.590              2.728              1572
562A  77,191       449,033   1.536              2.691              1688
562B  77,474       449,121   1.562              2.704              1687
562C  78,036       449,928   1.558              2.707               950
5620  77,553       446,441   1.615              2.690              4700
562E  76,080       447,584   1.498              2.690              1613
563A  77,423       449,685   1.543              2.705              2015
563B  79,836       452,450   1.596              2.746               970
563C  79,606       452,171   1.508              2.744              1328
564A  80,814       455,092   1.597              2.773              1892
564B  78,087       452,656   1.541              2.734              1125


                              APPENDIX A (continued)

      Total Energy           Highway            Work Trip       Cluster Size
Run      (gals)      VMT   Congestion Index     Length (miles)  (When Applic.)

564C  79,310       454,109   1.596              2.751              1600

570A  78,531       451,887   1.569              2.746               629
570B  75,683       448,827   1.504              2.707              1054
570C  75,511       448,293   1.493              2.700              1254
571A  74,882       446,633   1.485              2.656              1939
571B  76,684       449,301   1.520              2.688              1281
571C  77,154       448,695   1.538              2.678              1614
5710  80,230       452,118   1.605              2.743              980
572A  77,301       447,176   1.568              2.674              1545
572B  80,121       450,140   1.627              2.720               992
572C  77,314       444,693   1.555              2.658              1896
573A  80,141       452,329   1.662              2.737              1478
573B  80,314       453,259   1.614              2.755              930
573C  79,466       451,260   1.644              2.716              1743
574A  77,676       451,233   1.554              2.649              1509
574B  76,388       449,882   1.504              2.636              1288
574C  76,644       449,655   1.494              2.632              1257

582   75,291       452,829   1.370              2.744              1274
583   70,071       444,193   1.296              2.556              2935

592   74,600       449,685   1.380              2.701              1475
593   72,711       446,469   1.379              2.583              1509

600   80,126       450,671   1.629              2.736              3112
601   82,607       452,399   1.656              2.717              2492
601A  83,902       455,998   1.755              2.757              3506

                              APPENDIX A (continued)

      Total Energy           Highway            Work Trip       Cluster Size
Run      (gals)      VMT   Congestion Index     Length (miles)  (When Applic.)

602   84,184       451,485   1.752              2.709              2807
602A  76,337       439,826   1.579              2.596              8033
602B  78,813       440,260   1.659              2.658              6844
603   85,958       456,128   1.823              2.765              2783
603A  80,689       446,946   1.671              2.678              7696
603B  84,641       454,895   1.795              2.758              3408
604   79,131       453,194   1.597              2.643              2938
604A  79,619       453,205   1.607              2.632              4404
604B  80,352       453,477   1.635              2.665              2796

610   77,055       454,933   1.436              2.794
611   77,163       456,026   1.419              2.755
612   81,828       455,848   1.576              2.786
613   79,023       457,015   1.487              2.758
614   78,106       458,397   1.460              2.721

620   71,613       447,019   1.348              2.604
621   72,151       446,778   1.342              2.546
622   73,773       446,913   1.404              2.551
623   75,639       451,173   1.466              2.616
624   72,038       449,233   1.341              2.522

996   74,953       443,770   1.500              2.659


Jump To Top