Table of Contents
Coupled Human- Biologic Systems in Urban Areas:
Towards an Analytical Framework Using Dynamic Simulation
Philip C. Emmi
College of Architecture and Planning
University of Utah
375 S. 1530 East, Rm 235
Salt Lake City, UT 84112-0370
Ph: (801) 582-0719 Fx: (801) 581-8217
emmi@ geog.utah.edu
Abstract
This paper hypothesizes a newly understood, autonomous third force that shapes urban
expansion even without demographic or economic pressures. It is a self-reinforcing feedback
relationship driving urban spraw.. I is represented as a multi-causal feedback loop whereby
roads beget roads through a process of ever declining developmental densities. The loop’s
dynamic structure is central to a framework for analyzing human-biologic systems in urban.
areas. The hypothesis is articulated as a system dynamics model calibrated on the Salt Lake
City-Ogden metropolitan area, closely validated against recent observations, and used to
explore the urban system performance implications of various policy options through to the year
2030. Other relevant dynamics are then outlined peripheral to this core. Thus this paper is about
the dynamics of sprawl and how it serves as a dynamic organizing principle around which to
conceptualize further explorations of human-biologic systems in urban area.
Key words: urban sprawl, system dynamics, travel demand, policy simulation
1, Urban metabolics and coupled human-biologic systems
Urbanization is increasingly salient to inquiry conceming atmospheric futures and global climate
change. Historically humans have disproportionately favored a mural and agriculturally based
settlement pattem. By the middle of the 20° Century, thirty percent of our species lived in urban
areas. Mankind is about to cross a cultural threshold when in 2007 we become a predominantly
urban species. By 2030, sixty percent of the world’s population will be urban. Ninety-four
percent of a projected increase of 2.1 billion inhabitants will be urban (United Nations, 2002).
Urban areas presently occupy 3 percent of the earth’s habitable surface area. With expected
increases in urban populations, continued urban density declines and continued loss of habitable
landscapes, urban areas could occupy 8-to-9 percent of the earth’s habitable surface area by
2030. Declining urban densities are linked to increased energy consumption levels. Urban-to-
tural per capita energy consumption ratios are about four-to-one and increasing. These trends
suggest that urban land use and metabolic pattems merit closer scrutiny.
This is particularly the case when considering biospheric-atmospheric exchange, the human
dimensions of global climate change and the local determinants of human health.
Figure 1 identifies key relationships between urbanization and _biospheric-atmospheric
exchange. Demographic and economic drivers of urbanization precipitate a pattem of urban land
development to accommodate a variety of urban activities - working, residing, traveling - and
entail a variety of landscape ecologic changes - paving, planting, imigating. These are mediated
through an iterative process of urban developmental decision making and collaborative leaming.
Human activities and urban biota emit and absorb a variety of trace gases and suspended
particulates. Most significant are water vapor, volatile organic compounds (VOC), and the
oxides of carbon, nitrogen and sulphur (CO, CO2, NOx and SOx). Trace gases play important
roles in the global accumulation of greenhouse gases, the local formation of tropospheric ozone,
plant metabolics and the exacerbation of cardiopulmonary health conditions.
The dynamics of urbanization and biospheric- atmospheric exchange are not understood in an
integrated or systemic way. They are related to the metabolic rates of cities. Figure 2 shows one
aspect of this - the relationship between urban population density and per capita energy
consumption for selected urban agglomerations. Note the marked change observed in energy
consumption rates when comparing Vienna, Toronto and Phoenix. The suggestion for the
moment is that as densities decline, energy consumption rates increase thus driving upward rates
of trace gas emissions.
But the matter is not quite so simple. Other relevant processes include the dynamics of urban
demo- economics, the dynamics of urban sprawl, the prospects for CO2 sequestration via urban
forest expansion, and the modulation of ozone formation via altemative urban biotic regimes.
The purpose of this paper is to delineate an analytical framework for addressing these issues.
The effort begins with the recognition that all urban processes affecting bio-atmospheric
exchange are initially land-based. The strategy is to focus on urban densities, urban land use and
the processes that affect the appropriation of rural land into the urban sphere.
Demographic growth and economic change have long been recognized as key in determining
the pace of urban land development But a newly understood, autonomous third force causes
urban land development to expand even in the absence of demo-economic pressures. It is the
self-reinforcing feedback relationship underlying the well-known phenomenon of urban sprawl.
Urban sprawl is grounded in a multi-causal feedback loop whereby roads beget roads through a
process of ever declining developmental densities. The dynamic organization of this loop is
given a central role in the framework for analyzing coupled humar-biologic systems in urban
areas. Other relevant dynamics are then elaborated peripheral to this core. Thus the main thrust
of this paper is about the dynamics of sprawl.
In the section immediately below, we outline the basics with respect to America’s
consumption of urban land and urban road-building as preliminaries to the articulation of a self-
reinforcing feedback loop of congestion, road-building, sprawl, more congestion and even more
road-building. The third section focuses on the feedback loop itself and the evidence supporting
its veracity. The fourth section articulates the structure of this dynamic as set forth in a model
validated against data drawn from the Salt Lake City-Ogden, Utah, metropolitan area. The fifth
section outlines findings from simulation of policy altematives, and the sixth section suggests
how elaborations might be added to enhance the fullness of our analysis.
2. Urban land development, travel demand and road building in the U.S.A.
Between 1950 and 1990, urbanized lands in America quintupled while urban populations
doubled. This brought on a two and one-half-fold decline in urban land use densities.
Metropolitan rates of change in per capita urban land use consumption appear to be accelerating.
A recent map interpretation of urban land use change for the Chicago-Milwaukee region shows
urban growth between 1955 and 1975 covering a spatial extent equivalent to all pre-1955
growth, while urban growth between 1975 and 1995 covers a spatial extent equivalent to all pre-
1975 growth. In a region with moderate population growth, urban land appears to be doubling
every twenty years (Acevedo, 1999).
In both rapidly growing and slowly growing metropolitan regions, urban land development
consistently outpaces population growth. In the New York, Chicago and San Francisco, growth
in urban land between 1950 and 1990 outpaced population growth by factors of between 2.5 and
5. Yet during the same period, the Cleveland area expanded by 33 percent even as the population
declined by 11 percent (Diamond and Noonan, 1996). Fulton (2001) presents data from which
one can compute the change in urban land relative to the change in population by major U.S.
Census region for the period 1982 - 1997. There we leam that this ratio in the fast-growing West
is 1.6:1 while in the slow- growing Northeast it is 5.7:1.
Contrary to expectations, rates of urban land development appear to be poorly correlated with
rate of urban population or job growth.
Figure 3 shows an image of urban land development in the continental United States. It is
based on a 1993 to 2001 comparison of U.S. Air Force Defense Meteorological Satellite data on
night-lights (Mitchell, 2001). The gray and blue tones show development before 1993 while the
yellow and red tones show development from 1993 to 2001. The raw acreage developed during
the latter period closely matches the acreage developed during the earlier and much longer
period.
Figure 4 show how this process is evolving before and after 1993 in the rapidly growing
Southem Appalachian Piedmont region. The satellite image is placed along side data from
metropolitan areas in the region from the Fulton report to show results in both graphic and
numeric forms. The populations of the cities shown there grew during the 1982 - 1997 period
from between 9 and 61 percent while their urban land areas grew from between 29 and 139
percent. The median rate of growth in urban land was 3.3 times the median rate of population
growth.
Figure 5 show how this process is evolving in the declining region of Ohio, westem
Pennsylvania and northem West Virginia. The populations of the cities show there changed
during the 1982-1997 period from between 2 and -15 percent while their urban land areas grew
from between 25 and 53 percent. The median rate of growth in urban land was 7.6 times the
median rate of population decline.
Since 1980, suburban populations have grown ten-times faster than central-city populations.
Four-fifths of the country’s urban growth is located in dispersed suburban settings. In contrast,
European urban regions have three to four times the density. Notably, the comparison holds for
the newly developed suburban areas of European cities as well (Newman and Kenworthy, 1989).
From 1950-90 urban land grew at an average annual rate of 9.0 percent. For the nation’s
urban road network to keep up, it would have had to grow at about 4 percent per year
(approximately a square root relationship: 1.04 = (1.09)"”). A nearly comparable statistic on
substantially paved urban roads indicates a 3.8 percent growth rate over a similar time period
(Bureau of Transportation Statistics, 2000). Data from 1990-99 on urban lane miles shows a 4.1
percent annual average growth. Urban land and urban road lane miles appear to grow in an
appropriate relation to one another.
As urban road lane mile grew, national passenger vehicle miles increased by an average of
4.2 percent per year during the period 1950 - 1990 (Benfield, 1999). During the 1980’s growth
in vehicle miles traveled (VMT) was more than four times the growth in driving-age population.
By 1995 VMT per capita was over 2.3 times the rate for 1965 and roughly double that of
Europe’s more prosperous countries (Bureau of Transportation Statistics, 2000). In North
American metropolitan regions, growth in urban land use, urban road lane miles and VMTs have
consistently outpaced growth in population, eamings or jobs.
The disassociation between urban land development and population growth prompts one to
ask what other factors might be driving urban land consumption.
A few comparisons with European practice prompts one to consider whether rapid rates of
urban land development might not be related to land use and surface transportation policies.
Focusing briefly on the U.S in comparison to Great Britain and France, one notes first that a
higher percentage of surface transportation expenditures in the U.S. go to roadway construction
(86%) than in Britain (70%) or France (57%). Correspondingly a lower percentage goes to transit
facilities in the U.S. that typically require higherdensity, European-style neighborhoods to
provide service efficiently. Gasoline is 3 times more costly in Great Britain and 3.6 times more
costly in France than in the U.S. due predominantly to different fuel tax policies. Personal
vehicle miles traveled per capita are 47% lower in Great Britain and 45% lower in France than in
the U.S. Transit trips per capita are 3.7 times higher in London and 4.2 time higher in Paris than
in Philadelphia. The density of new urban land developments in the U.S. is typically 3to-4 times
lower than in Europe. There appears to be something about the relationship between urban land
development and U.S. policy and practice regarding surface transportation and urban land use.
3. The dynamic organizing principle governing humar-biologic systems in urban areas
Against the backdrop of these pattems, one needs also to note a changing urban
environmental context. Ours is no longer a manufacturing society. We can no longer point to
smoke stacks and industrial pollution as principal determinants of urban environmental
conditions. Urban land development, vehicle trip generation and roadway capacity expansion
now drive urban environmental dynamics. The transport sector uses two-thirds of U.S. oil
supplies. It produces a third of U.S. carbon emissions, 90 percent of urban CO emissions and 50
percent of tropospheric ozone precursors (Benfield, 1999). Urban land use and land development
pattems affect long-term trends in municipal and industrial demand for water, water supply
systems and stream quality. Newman and Kenworthy (1989) find that low-density suburbs
requires four to five times more water than do medium-density suburbs. Y et without remediation
measures, stream degradation can be a more serious issue in medium-density cities where
impervious materials predominate (Amold and Gibbons, 1996). Urban land development,
vehicle trip generation and roadway capacity expansion now mediate between urban demo-
economics forces and the environmental dimensions of urban system performance - between
people and jobs on the one hand and fuel use, trace gas emissions, stream quality, wildlife habitat
and solid waste on the other.
Expansion of an urban road network induces a reduction in developmental densities and
stimulates even further travel demand and further roadway capacity expansion. Anthony
Downs’s (2001) congressional testimony is based on his insights into the implication of this
process. He comes to very pessimistic conclusions.
Newman and Kenworthy (1999: 140) describe this nexus with particular cogency:
Most major cities that built extensive freeways then found that this process spread out
land use and generated more and more traffic, until very soon after completion the
freeways were already badly congested. The obvious response to the failure of freeways
to cope with traffic congestion is to suggest that still more roads are urgently needed.
The new roads are then justified again on technical grounds in terms of time, fuel, and
other perceived savings to the community from eliminating congestion. This sets in
motion a vicious circle of self-fulfilling prophecy of congestion, road building, sprawl,
congestion and more road building. Automobile dependence is inevitable in such traffic
engineering. Awareness of this phenomenon, called induced or generated traffic, is now
much more common in the literature. In fact, traffic is now being referred to not as a
liquid that flows where it is directed, but as a gas that expands to fill all available space.
Recognizing and modulating the strength of this feedback relationship is strategic to the long-
term management of human- biologic systems in urban areas.
Figure 6 shows the cover from the industry bulletin, Asphalt, dated April 1966. It shows
clearly the industry’s view of Newman and Kenworthy’s feedback loop. From the industry's
point of view, this loop guaranteed an increasing schedule of business activity. Do note the
special phrases, “roads stimulate travel,” “more road-tax revenue means extra paved roads,” and
“improved roads ... develop even more traffic.” Today, the industry would quickly disavow
these statements: the politics of road building were simpler and more supportive in 1966 than
now.
The idea captured by the special phrases in Figure 6 is now referred to as induced traffic or
induced travel demand. It is the idea that roadway investments induce longer, further and more
numerous tips (‘if you build it, they will come”), that induced traffic absorbs roadway capacity
intended to ease congestion (“you can’t pave your way out of traffic congestion”), and that
investments in transit and land use controls should have priority over new roadway construction
(“transit first”). As Robert Cervero (2003) notes, “Few issues in the urban transportation field
have sparked as much controversy ... as claims of ‘induced demand.’”
Seeking to quantitatively verify these claims has become a small “industry” in its own right
In 1994, the British Standing Advisory Committee on Trunk Roads and Generation of Traffic
(Goodwin, 1996) was the first to assess the effects of additional highway lane miles on induced
traffic volumes. They estimated a one percent increase in road lane miles brought on a one
percent increase in vehicular traffic. This finding caused the govemment to abandon its policy of
building more roads in response to projected growth in vehicular traffic.
In the U.S., the Transportation Research Board (1995) of the National Academy of Science
addressed the subject with reserve but did concede that, "Major highway capacity additions such
as a freeway bypass or a major interchange reconstruction are likely to attract further
development."
In a Califomia-based study, Hansen and Huang (1997) found that adding lane miles to the
state road network did induce substantial new traffic. At the metropolitan level, a one-percent
increase in lane miles induced a 0.9% increase in vehicle miles traveled within four years.
Missing from these studies was the description of a mechanism or a behavior that would
explain why added roadway capacity yields added traffic. Fishman (1989: 143) had previously
suggested it had to do with the relationship between transportation systems investments and
urban land speculation. Speaking of developer Henry Howard Houston and the Wissahickon
Heights section of Chestnut Hill, Philadelphia, built in the late 1890's, Fishman notes:
. a basic point which held true not only in the railroad era but also in the highway era
that followed it: The ultimate purpose of suburban transportation lines is not to move
people; it is to increase the value of the land through which it passes. The best suburban
land developer must also be a rail developer so he can direct his lines through land he
already owns and retain the bulk of its increased value for himself.
Transportation system investments do raise land values but not so much so as to discourage
developmental density declines. Lower density is, after all, the featured attraction of newly
developed lands. Stratham et al. (2000) measured this effect and found that a 1% increase in per
capita roadway capacity leads individuals to reside at a 0.25% lower density and work at a 0.22%
lower density. In addition to inducing added travel demand, new roadways also induce added
land development at consistently lower densities.
Closing the circle, the Urban Transportation Center (1999) showed that in metropolitan
Chicago the prior period’s population growth induced the present period’s roadway capacity
expansion. Roads attract populations and populations attract roads.
Cervero (2003) sought to sort out the effects of induced travel demand, induced building
growth and induced roadway investment by employing path analytic techniques to a model
composed of four simultaneous equations. With these he defined the direct and indirect effects of
lane mile growth on roadway speed, travel demand, development activity and back again to lane
mile growth.
Cervero’s model is useful in sorting out the mechanics of the processes. Added roadway
capacity improves average travel speeds and both spur further land development. The improved
speeds and the added development both contribute to increased traffic. The increased traffic then
spurs even further roadway capacity improvements. Elasticities are provided for each effect as
captured over a 2-to-4 year time horizon. But then he hedges his position by noting that these
effects “generally accumulate with time.”
4. A systems thinking - system dynamics approach
Cervero’s last remark gives a clue as to why one might want to expand upon statistical analyses
while using a more integrative approach. When effects generally accumulate through long period
of time, why would one want to assess roadway capacity expansion impacts with elasticities
estimated over a 2to-4 year time horizon? Cities do not equilibrate in 2-to-4 years. Cities do not
equilibrate, period. They are emergent systems. The characteristics of any complex system
emerge out of the rich interactions among its elemental parts (Gilbert, 1995). Einstein's remark
about problem solving comes to mind: “You can’t solve a problem with the same kind of
thinking that created the problem.” Experimentation with other approaches is required.
Consider instead the following two hypotheses: (1) that the vicious cycle described by
Newman and Kenworthy (1999) accounts for increases in urban land consumption, vehicle miles
traveled, roadway capacity, and urban environmental degradation at rates disproportionately high
relative to the underlying rates of demographic and economic change and (2) that recognizing
and regulating (dampening) the strength of this reciprocally reinforcing feedback loop is
essential for the successful management of not only urban densities, trip generation, and traffic
congestion but also urban economic and environmental indicators such as infrastructure costs,
water demand, local governmental fiscal capacity, energy use, and urban trace gas emissions.
To test these hypotheses, consider a system dynamics model that simulates over the 50 years
from 1980 to 2030 an approximation of Newman and Kenworthy’s dynamic nexus and identifies
long: term impacts on urban system performance characteristics.
At this early stage in model development, it is useful to have the model striped to its barest
essentials. The model recognizes only three stocks. While implications for other stocks will be
inferred, these are less central. Stocks at the core of the model include population, urban land,
and urban roadway capacity.
The structure of the model is simple, strategic and telling. Population growth is exogenous
and “drives” land development. Newly developed land generates additional trips. Additional
trips increase the region’s road-building goal: building roads is a goal-attainment process.
Pursuing the goal gets roads built, lowers developmental densities, and requires incrementally
greater amounts of land per building structure. The greater amounts of land required then
generate incrementally more trips per acre and thus increase the road-building goal. The
reinforcing feedback effect transforms the impacts of a nearly linear “driver” (population) into
exponentially growing land consumption and traffic generation functions. This dynamic structure
is formally a goal attainment process nested within a reinforcing feedback loop. Its reference
behavior is a pattem of incessantly more fervent activity (road building) in pursuit as an ever-
receding goal (traffic congestion relief) with ever diminishing success. It is the simplest way to
explain why growth in urban land, traffic and road building so consistently outpaces changes in
population, eamings and jobs and why road building so infrequently succeeds in meeting its
goals. The model is named SprawlSim I.
The model is initialized with data from the Salt Lake-Ogden metropolitan area for the year
1980. For validation purposes, simulations are run with policy variables in default positions, year
1999 simulation values are noted, and these values are compared with observations from 1999.
Validation results are reported below.
Modulating the feedback dynamic during the projected time horizon is a core function of
local land use and transportation policies. Elementary policy options include controlling land
developmental densities, altering the rates of road construction, diverting traffic to altemative
travel modes, and filling the road gap with intelligent transportation systems capacity
enhancements instead of building roads. Each policy individually is insufficient to effectively
modulate the dynamics of sprawl, but in reasonable combination the process can be substantially
managed, densities can be largely maintained, and traffic congestion made bearable.
Figure 7 shows a causal loop diagram for the SprawlSim model. The model and the causal
loop diagram were made using High Performance System’s STELLA® system dynamic software
package. Figure 7 shows the basic structure of the model’s dynamic organizing principle. Start
with the region’s stock of Urban Land, which expands because of population growth. As Urban
Land develops, it causes a specific flow of persontrips to be generated each year by people
seeking to get where they need to go. These are calculated on the basis of a monotonically
increasing function relating person-trips per acre to time. As TripGeneration goes up, the desired
roadway capacity increases. This flow is called the TargetLaneMiles. It represents the added
highway capacity transportation engineers would want to see built given tends in
TripGeneration. TargetLaneMiles are compared to the actual regional roadway system capacity
also measured in lane miles and any deficit is defined as the RoadGap. A positive RoadGap
drives the building of roads as a goal-attainment process. This increases the region’s roadway
capacity measured in Lane Miles and in LaneMile\l000Persons. An increase in
LaneMile\1000Persons drives new development densities down maryinally by an amount
suggested by a research provided by Stratham (2000). These now marginally lower densities
govem urban land requirements during the next round of developingAcres. These then feeds
back into the volume of Urban Land in the region, and the process starts all over again.
Figure 8 shows a system map for the SprawlSim model that corresponds to the causal loop
diagram of Figure 7. To navigate through the system map, start with Pop%GrowRate. This
defines the population growth called net Growing. It causes new building structures
(NetBuildingGrowth) to be built on newly developed urban land @evelopingAcres). The process
is govemed by a monotonically increasing function relating buildings per person to time as is
appropriate given changing demographic age structures in the region. This adds to the stock of
Urban Land. Person-trips are generated as a time-dependent function of Urban Land. These are
converted into desired urban road lane miles (TargetLaneMiles), after consideration for (1) the
diversion of persontrips to nonautomotive modes of travel (NonAuto%*NonAutoT rafficE quiv)
and (2) the diminution of the need for added highway capacity by altematively expanding the
existing roadway capacity through intelligent transportation system _ investments
(ITS%Capacityincr). The rest of the system map follows the logic defined above for the causal
loop diagram of Figure 7.
Several policy sliders facilitate experimentation with feedback modulation. These, too, can
be seen on the system map in Figure 8. Developmental densities can be increased with a slider
that presumes a user-specified percentage increase in new development densities as might be
achieved through changes in local land use and land development policies ctDensityIncrease).
For example a 21 percent increase in density could be achieved by following the
recommendations developed for Envision Utah by ECONorthwest (1999) as part of their Quality
Growth Strategy. These recommendations pertain to a modification of the structure of new
housing starts by building structure type so as to match the pattem of structure types needed
given the proportions of owner - and rental occupancy in the region.
The percentage of trips made by nonautomobile modes of transportation can be increased
with a slider that presumes increased investment in and use of buses, regional rapid transit,
bicycle lanes and walking (NonAuto%). Considerable investments in these altematives are
recently under way within the Salt Lake-Ogden area and are being responded to with notable
user enthusiasm.
Intelligent transportation systems (ITS) refer to a class of road capacity management
strategies designed to get more capacity out of existing roads. They include activities like
freeway onramp metering, traffic signal synchronization and well-enforced peak-hour parking
limitations on arterials. An ITS slider allows one to increase the traffic handling capacity of
existing roads by a user-specified percent (TS%Capacitylncr). It uses the presumption that when
these policies are implemented demand for newly added roadway capacity is proportionately
reduced.
The fraction of the road gap built per year (FrOfGapBt\Yr) reflects the intensity of the
regional road building effort relative to perceived road building need. Changing this slider from
its default value lets the user simulate more intense or less intense road building activity relative
to perceived needs.
All policy slider take effect in the year 2004. Any change in the FrOfGapBtl\yr takes effect
without delay. All other policies have an average implementation lag time of 3 years. This means
most of the policy’s effects take 7 years for near- complete implementation.
The model was initialized with data from 1980 for the Salt Lake-Ogden metropolitan area.
Initial values were defined for urban population (InitPop) using data on county population from
the Govemor's Office of Planning and Budget (GOPB, 2001). County population was converted
to urban population using a ratio obtained by inspection of Federal Highway Administration data
on urbanized areas within the state (FHWA, 1999). Data on Urban Land was obtained from the
Utah office of the USDA’s Natural Resources Conservation Service where current and historic
records from the Natural Resources Inventory are maintained (Grow, 2002). This provided the
input for InitUrbLand. Data on urban roadway lane miles is available for the study area reaching
back to 1992 (FHWA, Table HM-71, various dates). Data on roadway lane miles for all urban
areas in Utah is available reaching back at least to 1980 (FHWA, Table HM-60, various dates).
A variety of interpolation methods were used to estimate the initial value for the region’s urban
roadway lane miles (InitLaneMi).
A baseline simulation was used to validate the model’s performance against recent
observations. Initialized with 1980 data, the model was run until 1999 and stopped. Simulation
values were noted and compared to observations gleaned from the data sources given above. The
comparison is summarized in Table 1 as a set of percentage error terms. These range from a low
of 0.3 percent to a high of 2.1 percent for an average absolute error among the five observations
of 1.3 percent. Percentage errors of this low magnitude lend credibility to the use of the model as
a projection tool especially since it is stripped to the barest essential relationships and does not
include several enhancements that would improve its accuracy.
5. Experimental policy simulations: defeating urban sprawl and traffic congestion
The outcomes of baseline simulations to the year 2030 are shown in Figures 9 and 10. Figure 9
shows trajectories from 1980 to 2030 with respect to key indicators of urban system performance
- urban land per capita, urban roadway lane miles per 1000 persons and average daily person
trips per capita. Land per capita grows from 0.2 gross urban acres per person to 0.4 acres per
person or altematively from 2/10ths of an acre per person to 3/10ths of an acre per person -
roughly a 50 percent density decline. Urban roadway lane miles per 1000 grow from 3.5 to 7 for
about a 100 percent increase. Person trips per person grow from 1.5 per day to 3.7 for about a
125 percent increase. Note that smaller percentage changes in densities are associated with larger
percentage changes in trip-making and road-building. Also note that each of these three measures
of urban system performance show distinct signs of exponential growth.
Figure 10 tracks changes as measured by percentage increases from initial values. This helps
show relative rates of change among measures of urban system performance. There we track
change in urban population, urban land, urban roads and person trips. Urban population has the
lowest percentage rate of change. Urban land expands at roughly twice the speed. Urban
roadway capacity exceeds the rate of population growth by a factor of three, while trip- making
explodes at four times the rate of population growth. The percentage change in the road gap is
not shown until Figure 14. It is a rough proxy for roadway congestion. Under the baseline
simulation, it grows eight times faster than population.
Given the untoward nature of baseline results, one can safely assume that community leaders
will not permit historic pattems and practices to be expended unmodified into the future. But
what combination of system modification will be effective in managing the untoward results of
an undampened feedback loop?
To explore this question, two experiments were designed and min in comparison with the
baseline simulation (trajectory 1). These are tracked in Figures 11 - 14 with the comparative
graph trajectories labeled 1, 2 and 3. Remember the simple model show above in Figure 8 has
been modified so that all policy changes take effect only after 2004 and kick in with an average
three-year implementation delay (STELLA ® function [SMTH1, 3]).
In the first experiment, aggressive changes were made in all policy sliders. The fraction of
the road gap built per year was cut in half. A set of land use policies sufficient to increase
densities by 22 percent was assumed. The percent of non-automobile traffic was presumed to
change from a default value of three percent to a value of 22 percent. And ITS improvements
were presumed to substitute for 14 percent of needed roadway capacity. This tough combination
of policies is called “shock- and- awe.”
The second experiment was made identical to the first except that the intensity of road
building was increased by 50 percent instead of being cut by 50 percent. Because this scenario
seeks to accommodate society’ s road- building impulse, it is call “cake- and- eat: it-too.”
Figure 11 shows the outcomes of the baseline simulation and the two experimental mms on
population densities measured in people per gross urban acre. The baseline shows strong decline
in densities. Both experimental runs show some control over continued density declines.
Densities in 2004, when policy changes are initiated, stand at 4.2 persons per acre. Without
policy change, they would decline to 3.2 persons per acre by 2030. Both experimental runs limit
further declines to the 3.8 - 3.9 persons per acre range. Figure 11 shows that some effective
control over sprawl is possible and that both experimental policy sets have roughly the same
general level of impact.
Figure 12 shows the outcomes of the baseline simulation and the two experimental mms on
trip generation measured by percentage change in trips per capita. The baseline shows a 125
percent increase in trip generation by 2030. Both experimental runs show some control over
continued increases in trip generation. Both experimental runs limit further increases to the 80
percent range. Figure 12 shows that some effective control over trip generation is possible and
that both experimental policy sets have roughly the same general level of impact.
Figure 13 shows the outcomes of the baseline simulation and the two experimental runs on
roadway capacity as measured by the percentage change in roadway lane miles. The baseline
shows a 125 percent increase in per capita roadway capacity by 2030. The shock-and-awe
strategy shows a meager 10 percent increase in per capita roadway capacity, while the cake-and-
eat: it-too strategy shows only a 25 percent increase. Both experimental runs show a tempering of
road building to rates well in proportion to population growth. Figure 13 shows that need to
build roads can be brought under control and that the difference between the two experimental
policy strategies in modest.
Figure 14 shows the outcomes of the baseline simulation and the two experimental nuns on
roadway congestion as measured by percentage changes in the road gap - the difference between
desired and actual roadway capacity. The baseline shows a 1000 percent increase in the road gap
by 2030. This implies a seriously growing problem with traffic congestion under the baseline
scenario. The shock-and-awe strategy reduces this growth substantially. It reduces congestion
substantially and allows 16 years before congestion retums to 2004 levels. But then it continues
to increase at an all-too-rapid pace that parallels the baseline case. By 2030, it shows about a 700
percent increase in congestion. The cake-and-eat-it-too strategy shows a different trajectory. It,
too, cuts substantially into the road gap. But then it equilibrates at about a 250 percent increase -
a level of congestion first encountered in 1997. This outcome is interesting and merits some
further explanation.
It appears that seeking to meet desired roadway capacity standards without simultaneously
controlling developmental densities, diverting traffic to altemative modes, and improving the
capacity of the existing road network is totally ineffective. It also appears that both experimental
policy sets are substantially effective at containing sprawl. This is because they both dampen the
feedback effects defined by the causal loop diagram in Figure 7.
But the shock-and-awe strategy requires an unnecessary sacrifice. It unnecessarily imposes a
continued problem of roadway congestion and travel delay.
The cake-and-eat-it-too strategy provides nearly the same level of sprawl control but also
provides substantial traffic congestion relief. There is a simple reason why this happens. The
policies on land use densities, altemative travel and existing road capacity sufficiently dampen
reinforcing feedback effects and lower the perceived need to build roads so as to pemit an
appropriately lowered volume of road building to proceed without its counter- productive impacts
on induced development and induced traffic.
SprawlSim offers strong evidence for the conclusion that sprawl can be controlled and that
roads can be build to meet their intended purpose of congestion relief. But the latter result holds
only under very restrictive conditions, namely that an appropriately lower level of road-building
be accompanied by vigorous programs promoting developmental density controls, abundant
travel by altemative modes and effective enhancements to existing roadway capacity.
6. From SprawlSim to a coupled human-biologic urban system model
The argument advanced in Section 3 is that a sprawlinducing feedback mechanism is the
dynamic organizing principle shaping urban developmental trajectories in North American cities
today. This principle constitutes a third driving force, an autonomous element that can take its
place along side demographic and economic forces as determinants of urban system trajectories.
Having discemed a way to model this force and to modulate its impacts, the next step is to
outline how that model might be enhanced to represent a series of other important urban-based
human and biotic processes.
The first task is to improve the basic SprawlSim model. Track both net job formation and the
stock of job opportunities. Link jobs to demographic migration. Differentiate between
commercial and residential building activity. Simulate change in average tip lengths. Expand the
determinant to trip making. Include a calculation of average daily vehicle miles traveled (VMT),
and assess desired road capacity based on VMT. Use simulated values for traffic volume (VMT)
and capacity to estimate average system-wide travel speeds.
The second task is to extend the basic model to include a sector on state and local fiscal
matters. Show how land development requires added regional and local infrastructure, how the
amount required varies with developmental densities and how this influences the rate of state and
local expenditures. But also show how these same expenditures add to property value and urban
activities, how these additions increase property and sales tax revenues without increased tax
rates, and how these revenues can, if densities are not too low, provide sufficient revenues to
fund required infrastructure expansion. Show how, if land use and transportation policies are
used to constructive effects, state and local budgets can be readily balanced and how, if not,
easily driven to deficit. Link infrastructure expenditures back to urban developmental densities
as they, too, like road building, push population densities down.
The third task is to extend the basic model to include a sector on anthropogenic trace gas
emissions. With simulations on jobs and urban commercial and residential land use, calculate
point- and area-source trace gas emissions. With simulations on vehicle traffic and speed,
simulate linear- source emissions with particular attention to SOx, VOC and NOx.
The fourth task is to add a sector on urban flora. While the region was originally treeless, it is
now about two-fifths forested. The change was enabled by the introduction of extensive
imigation.
Now water is scarce. Citizens are urged to conserve. Xeriscaping is encouraged. Urban
foresters wonder whether they should replace dying trees of which there are many.
In the short run, conservation will reduce the need to expand regional water development
systems. It will spare the region a great expense and a significant environmental loss. In the long
Tun, water conservation will permit the economic development of the region to continue.
But there is a second altemative. It relates to water, trees and tropospheric ozone. Ozone
formation is very temperature-sensitive once temperatures exceed 90 degrees F. Trees cool the
city. They offset an increasingly intense urban heat island effect. They emit water vapor that
reacts with O+ thus reducing the formation of ozone (03). They absorb ozone precursors. But
the wrong kind of tree will emit VOC, another 03 precursor. Fortunately we have been planting
and could continue to plant tree like maples that are low VOC emitters. If so, an expansion of
the urban forest will not only sequester more CO2, it will slow the formation of ozone.
Use the sector on urban flora to explore the tradeoff in the use of water between sustained
urban growth and cleaner urban air.
The fifth task is to forge linkages between the sector on anthropogenic trace gas emissions
and the sector on urban flora. Show how anthropogenic trace gases affect the long-run health of
trees. Show how biogenic trace gases and other biogenic effects interact with anthropogenic trace
gases to alter ozone formation and airbome particulates.
Finally link all the above to a human health sector with linkages back to the local fiscal
sector and to the propensity to form jobs and to migrate into the region.
For each model sector, organize an interest group with domain experts, stakeholders and
policy decision makers. Involve them in the task of group-based model building. Rely on each
group to validate the logic of that sector's model structure. Integrate each sector into the model
one at a time and, calibrate with local data, and seek to re-validate the model's accuracy at each
stage. Explore and document the policy implications gleaned from each interest group's
interaction with the process of model building and policy simulation.
With appropriate collaboration underwritten by extensive and extended research support, the
prospects are good for elaborating a useful framework for the design and development of a
dynamic simulation model of human- biologic systems in urban areas.
7, Summary
In 1950, 30 percent of the world’s population was urban. Fifty years later, urban areas take up 3
percent of the earth’s habitable surface area. United Nations projects that the world’s population
will be 60% urban by 2030 by which time nearly all growth will be urban growth. With
continued density declines, urban areas could take up 8-to-9 percent of the earth’ s habitable
surface area. This is clearly a period during which the urban dynamics commands our attention.
Forging an integrated understanding of the dynamics of urbanization and its relationship to
biospheric- atmospheric exchange is one of the most pressing aspects of the urban challenge.
To press forward with the challenge, I hypothesize an autonomous third force that causes
urban expansion in the absence of demographic and economic pressures. This force strongly
shapes the trajectories of urban system performance. This force is characterized as a self-
reinforcing feedback loop whereby roads beget roads through a process of ever- declining
developmental densities. Urban sprawl and traffic congestion are emergent properties of an
urban system impelled undiminished by the logic of its structure.
Earlier efforts have offered partial characterizations of this force. These appear as research on
induced traffic, induced travel demand, induced land development and induced road building.
We have come to understand that urban roads attract urban populations and urban populations
attract urban roads.
To integrate these ideas and test the feedback hypothesis, I present a system dynamics model
designed to simulate urban sprawl and traffic congestion. Its logic is simple. Newly developed
land generates additional trips. A dditional trips increase the region’ s road- building goal.
Pursuing the goal gets roads built, lowers new development densities, and requires incrementally
more land per building structure. This then generates incrementally more trips per acre and thus
increases the road- building goal. The logic of sprawl- inducing feedback transforms the impacts
of a nearly linear “driver” (population) into exponentially growing land consumption and traffic
generation.
Formally, the dynamic structure of this process is a goal attainment process nested within a
reinforcing feedback loop. Its reference behavior is a pattem of incessantly more fervent activity
(road building) in pursuit as an ever- receding goal (traffic congestion relief) with ever
diminishing success. It is the simplest way to explain why growth in urban land so consistently
outpaces changes in population and why road building so seldom succeeds in meeting its goals.
The model is named SprawlSim I.
Mathematically the model is a complex set of simultaneous differential equations. It is
initialized with 1980 data from the Salt Lake City- Ogden metropolitan area. The model’s
nominal time horizon extends to the year 2030. It contains user- adjustable policy variables that
simulate policy initiatives launched in the year 2004 with 90% implementation by 2011. These
simulate the effects of four land use and transportation policy options that collectively have the
potential to dampen the model's sprawl-inducing feedback mechanism.
To validate the model, simulations were stopped during the year 1999, five simulation values
were noted and then compared to observations for the same year. Simulation errors range from
0.3% to 2.1% with an average error of 1.3%. Percentage errors of this low magnitude lend
credibility to the use of the model as a projection tool.
A baseline and two experimental policy simulations were mun to test the prospects for
defeating urban sprawl and traffic congestion. Given the untoward nature of baseline results, one
can safely assume that community leaders will not permit historic pattems and practices to be
expended unmodified into the future. The “shock- and- awe” experimental policy scenario
effectively dampens the reinforcing feedback loop but exacts an unnecessary sacrifice. It
imposes a level of traffic congestion that need not be endured. The “cake-and- eat-it-too” scenario
constrains developmental densities, meets travel demand with both demand and supply
management strategies and addresses remaining travel demand in an even shorter period than
now. The experiments show what combination of policies can be used to both control sprawl and
relieve traffic congestion.
Having modeled America’s sprawl-inducing feedback mechanism, the next step is to outline
how that model might be enhanced to represent a series of other important urban- based human
and biotic processes. Nine measures are recommended. (1) Enhance the existing SprawlSim
model to account for jobs, land use types, trip lengths, vehicle miles traveled and average vehicle
speeds. (2) Add a sector to the model on state and local fiscal practices with a density- dependent
infrastructure expenditure function showing the conditions under which expenditures add to land
value, draw upon fiscal resources, and induce further developmental density declines. (3) Adda
sector on anthropogenic trace gas emissions. (4) Add a sector on urban forests to explore the
tradeoff in water use between sustained urban growth and cleaner urban air. (5) Link trace gas
emissions to urban forest health. (6) Link all the above to human health, local fiscal effects and
demo- economic responses. (7) For each sector, involve domain experts, stakeholders and policy
decision makers in group based model building. (8) Document the impacts of these activities on
participants’ comprehension of policy complexity and (9) define the implications for elaborating
amore comprehensive understanding of human- biologic interactions in urban areas.
Acknowledgements
The author would like to acknowledge the support of the National Science Foundation,
Directorate for Engineering, Biocomplexity in the Environment Program on the Dynamics of
Coupled Natural- Human Systems. The author also acknowledges support from the Southwest
Center for Environmental Research and Policy (SCERP), a bi- national, multi- university
consortium serving the U.S.- Mexican border.
Figure 1. Key relationships between urbanization and biospheric- atmospheric exchange.
80 Houston
PI ;
oF on. United States
Denver
1 Los ‘Angeles
Chicago
New York
Melbourne
Adelaide, Australia and Canada
Oi
Sydney O Paris Europe
Bis Berlin Asia
aa)
LJ Singapore
London Vienna~ [Tj
60 80
Energy consumption per capita (1,000 millions of joules)
Population density (people per hectare)
Figure 2. The relationship between population density and energy consumption in cities.
Figure 3. Continental U.S. urban land development pre- and post 1993.
Range of Change in Population
and Urbanized Land, 1982-97
Spraw ina
Growing Region
(land)=3.3* (pop)
The Piedmont of the Southem A pnalachian Mountains
Figure 4. Sprawl in a growing region represented graphically and numerically.
Sprawl ina Ran . .
we . ge of Change in Population
Declining Region and Urbanized Land, 1982-1997
(land) =-7.6* (pop) Maximum $= 2.3% 53%
Minimum -15% 25%
Median “5% 38%
Ohio, Westem Pennsylvania and West Virginia (Columbus,
Cleveland, Y oungstown, Pittsburgh and Wheeling)
Figure 5. Sprawl in a declining region represented graphically and numerically.
GTH ASPHALT
AP STENT ORE PAVED ROA MEME,
OEE wines OADS fp ,NT
PRON FEWER TAX DOLLARS" SOR
iiny
2
=
a
o
a
s
M es
RE RoaD.Tax pevenve WONG
DOLLARS 10 MAINTAIN
LO EXTRA PAVED ROMD?
Figure 6. Roads beget roads: From the cover of Asphalt Bulletin, A pril 1966.
| Urban_Land
A ™
developingAcres
| Acres\Building R
LaneMiles\10...
%
Lane_Miles
x.
TripGeneration
TragetLaneMiles
,
RoadGap
building
Figure 7. A causal loop diagram of a sprawl- inducing feedback mechanism.
net Growing
Pop%GrowRate
FrOfGapBit\Yr
O
NetBuildingGrowth
People
building
Acres\Building
LaneMiles\1000Person
Lane Miles
RoadGap
(@, O
InitPop InitUrbLand InitLaneMi
developingAcres
TragetLaneMiles
Buildings\Person
PctDensityIncrease
Urban Land
VehTrips\Ac.
ao
TripGeneration
DesiredLaneMiles\Trip
NonAuto%
NonAutoTrafficEquiv
ITS%CapacityIner
Figure 8. A system map of the SprawlSim system dynamics model.
Table 1. Validation Statistics for SprawlSim from the Salt Lake City- Ogden Area.
ITEM 1980 1999 Simulation Error
Urban Population 827,000 | 1,184,000] 1,187,780 | +0.3%
Urban Land (acres) 166,000 272,000 267,694 | - 1.6%
People/A cre 4.98 4.35 444 +2.1%
Road Lane Miles 2,821 4,636 4,610 | +1.1%
Lane Miles/1000 people 3.41 3.92 3.87 - 1.3%
1: Land\Capita 2: LaneMiles\1000Person 3: Trips\Person
wuno
Sow
600
nr
ee}
|_ 23: @
2
30 |-<——3———
1980 1990 2000 2010 2020 2030
Years
Figure 9. Key ratios: density, road capacity and travel behavior, baseline simulation.
1: Pop%Chng 2: Land%Chng 3: Roads%Chng 4: Trips%Chng
500 y,
P04
rs r
im
250 mA 4 - i |
Pa
3
ea
(| no)
3 eee |"
tes Ly
a
0 eee 4 i _
1980 1990 2000 2010 2020 2030
Years
Figure 10. Percentage change in urban activities, baseline simulation.
People\Ac: 1- 2- 3-
5.0 7=
Tis
| on
ba
“2
a:
o | hi ae
: SN aes
Ce
3.0
1980 1990 2000 2010 2020 2030
Years
Figure 11. Population density as people per acre - three scenarios: (1) baseline, (2) shock- and-
awe, (3) cake-and-eat-it-too.
Trips\Capita%Chng: 1 - 2- 3 -
200 4
100 1
hee]
129-3
3 wet
=
yor
oe a ae
0 1.4=2=3
1980 1990 2000 2010 2020 2030
Years
Figure 12. Travel behavior as percentage change in trips per capita - three scenarios: (1)
baseline, (2) shock- and- awe, (3) cake- and- eat- it-too.
LaneMiles\Capita%Chng: 1- 2- 3-
150.07
val
J
50.0 5
“4 ite | ase
, , a 3
=2- SS
23 1
-50.0
1980 1990 2000 2010 2020 2030
Years
Figure 13. Roadway capacity as percentage change in lane miles per capita - three scenarios: (1)
baseline, (2) shock- and- awe, (3) cake-and-eat-it-too.
RdGap%Chng: 1 - 2- 3 -
1500 4
500 + 1 2
te | ile
12 3 oT
a as
-500
1980 1990 2000 2010 2020 2030
Years
Figure 14. Traffic congestion as percentage change in the road gap - three scenarios: (1)
baseline, (2) shock- and- awe, (3) cake- and- eat- it-too.
References Cited
Acevedo, William. 1999. Analyzing land use change in urban environments, U.S. Geological
Survey, USGS Fact Sheet 188-99, URL: http://edcdgs9.cr.usgs.gov/urban/factsht pdf.
Amold, Chester L. Jr. and C. James Gibbons. 1996. Impervious surface coverage: The
emergence of a key environmental indicator. Joumal of the American Planning Association,
62(2):243- 258.
Benfield, F. Kaid, Matthew D. Raimi, and Donald D. T. Chen. 1999. Once There Were
Greenfields. N.Y .C.: Natural Resources Defense Council.
Bureau of Transportation Statistics. 2000. National Transportation Statistics 2000. BTS01-01.
Washington, DC: U.S. Govemment Printing Office.
<http://www.bts.gov/publications/nts/index.html>.
Cervero, Robert. 2003. Road expansion, urban growth and induced travel: A path analysis.
Journal of the American Planning Association, 69(2): 145-161.
Diamond, Henry L. and Patrick F. Noonan. 1996. Land Use in America. Washington, D.C.:
Island Press.
Downs, Anthony. 2001. The Future of U.S. Ground Transportation from 2000 to 2020.
Testimony to the Subcommittee on Highways and Transit of the Committee on Transportation
and Infrastructure, House of Representatives, 107 Congress of the United States. Washington,
D.C., March 21.
ECONorthwest. 1999. Greater Wasatch Housing Analysis. Salt Lake City: Envision Utah.
Fishman, Robert. 1989. Bourgeois Utopias. New Y ork City: Basic Books.
Fulton, William. 2001. Who Sprawls Most? Washington D.C.: The Brookings Institute.
Gilbert, N. 1995. Emergence in social simulation. In N. Gilbert and R. Conte (eds), Artificial
Societies. London: UCL Press.
Goodwin, Philip B. 1996. Empirical evidence on induced traffic: A review and synthesis.
Transportation, 23: 35-54.
Govemor's Office of Planning and Budget. 2001. State and county data profiles. Salt Lake City:
GOPB, Demographic and Economic Analysis.
Grow, Ray. 2002. Personal communication from the Utah office of the USDA Natural Resources
Conservation Service. Salt Lake City: USDA.
Hansen M. and Y. Huang. 1997. Road supply and traffic in urban areas: A panel study.
Transportation Research A, 31(3): 205-218.
Mitchell, John G. 2001. Urban sprawl, National Geographic, July, 48-73.
Newman, P. and J.R. Kenworthy. 1989. Cities and Automobile Dependence: An Intemational
Sourcebook. Aldershot, UK: Gower.
Newman, P. and J. Kenworthy. 1999. Sustainability and Cities: Overcoming Automobile
Dependence. Washington, D.C.: Island Press.
Strathman, James G. et al. 2000. Analysis of induced travel in the 1995 NPTS. Center for Urban
Studies, Portland State University. Presented at the A ssociation of Collegiate Schools of
Planning Conference, Atlanta, November 5 - 10.
Transportation Research Board. 1995. Expanding Metropolitan Highways, Special Report 245.
Washington DC: National Academy of Science.
United Nations. 2002. World Urbanization Prospects: The 2001 Revision. Department of
Economic and Social A ffairs, Population Division.
U.S. Department of Transportation, Federal Highway Administration. 1999. 1999 Status of the
Nation’s Surface Transportation: Conditions and Performance Report. Washington, D. C.:
USDOT, URL: http://www .fhwa.dot.gov/policy/1999cpr/report.htm/.
U.S. Department of Transportation, Federal Highway A dministration, Office of Highway Policy
Information. 1999. Highway Statistics 1999. Washington, D. C.: USDOT, URL:
http://www .fhwa.dot.gov/ohim/hs99/index.htm/.
U.S. Department of Transportation, Federal Highway Administration, Office of Highway Policy
Information. 1995. Highway Statistics 1995. Washington, D. C.: USDOT, URL:
http://www .fhwa.dot.gov/ohim/1995/.
Urban Transportation Center. 1999. Highways and urban decentralization. Chicago: University
of Illinois at Chicago.
Back to the Top
