See discussions, stats, and author profiles for this publication at: [Link]

net/publication/5198247

Models of Transportation and Land Use Change: A Guide to the Territory

Article in Journal of Planning Literature · April 2007

DOI: 10.1177/0885412207314010 · Source: RePEc

CITATIONS READS

248 5,138

3 authors:

Michael Iacono David Levinson

University of Minnesota Twin Cities The University of Sydney
26 PUBLICATIONS 947 CITATIONS 517 PUBLICATIONS 12,969 CITATIONS

SEE PROFILE SEE PROFILE

Ahmed M. El-Geneidy
McGill University
248 PUBLICATIONS 11,108 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ahmed M. El-Geneidy on 26 May 2014.

The user has requested enhancement of the downloaded file.

 Journal of Planning Literature OnlineFirst, published on February 13, 2008 as doi:10.1177/0885412207314010

Models of Transportation
and Land Use Change:
A Guide to the Territory

Michael Iacono This article reviews some of the important theoreti-

David Levinson cal frameworks adopted by researchers to represent the
Ahmed El-Geneidy complex relationship between transportation and land

Modern urban regions are highly complex entities. Despite the MICHAEL IACONO is a research fellow in the Department of Civil
difficulty of modeling every relevant aspect of an urban region, Engineering at the University of Minnesota—Twin Cities. His
researchers have produced a rich variety of models dealing research interests include transportation planning, policy and eco-
with interrelated processes of urban change. The most popular nomics, and urban policy and economics. He is currently conduct-
types of models have been those dealing with the relationship ing research on multimodal measures of accessibility and models of
between transportation network growth and changes in land land use change for the Twin Cities Metropolitan Region.
use and the location of economic activity, embodied in the con-
cept of accessibility. This article reviews some of the more DAVID LEVINSON is an associate professor in the Department of
Civil Engineering at the University of Minnesota and director of
common frameworks for modeling transportation and land use
the Networks, Economics and Urban System (NEXUS) research
change, illustrating each with some examples of operational group. In academic year 2006–2007 he was a visiting academic at
models that have been applied to real-world settings. It then Imperial College in London examining the coevolution of transport
identifies new directions for future research in urban modeling and land use. He currently holds the Richard P. Braun/CTS Chair
and notes the important contributions of the field to date. in Transportation Engineering.

AHMED EL-GENEIDY is an assistant professor at the School of

Keywords: transportation planning; land use; mathe- Urban Planning, McGill University. His research interests include
matical models; urban growth; gravity model; land use and transportation planning, transit planning and opera-
microsimulation tions, and accessibility/mobility measures for urban areas. He earned
BS and MS degrees from the Department of Architectural
Engineering at the University of Alexandria, Egypt, and continued
INTRODUCTION his academic work at Portland State University, where he received a
Graduate GIS Certificate and earned a PhD in Urban Studies.
Models are the basic tool of analysis for planners
working in the fields of transportation and land use The authors would like to acknowledge the comments of two
forecasting. Current practice in these fields generally anonymous reviewers who helped to correct and clarify a number of
accepts the notion of some type of reciprocal relation- key ideas developed in this article. This research was undertaken as
ship between transportation and land use. For more part of the project Access to Destinations: Monitoring Land Use
than four decades, urban researchers have sought to Activity Changes in the Twin Cities Metropolitan Region, adminis-
tered and supported by the Minnesota Department of Transportation.
formalize this relationship using mathematical, statis-
tical, and logical methods, and to produce models Journal of Planning Literature, Vol. XX, No. X (Month XXXX)
capable of predicting changes to transportation and DOI: 10.1177/0885412207314010
land use systems as the result of policy measures. Copyright © 2008 by Sage Publications

Copyright 2008 by SAGE Publications.

2 Journal of Planning Literature

use. Each framework has guided the development of a the demand for travel to and from a particular loca-
number of different operational models, that is, models tion. This relationship is sometimes referred to as the
that have been applied using data from real-world transportation–land use “link” or “cycle,” emphasizing
metropolitan regions. Several of these models are a feedback relationship (Kelly 1994). The mediating
described in some detail to illustrate how each modeling factor in determining changes in the location of activi-
framework is used to represent the processes of urban ties and the demand for travel is accessibility, which
change.1 Before turning to the models, however, some measures the situation of a location relative to other
background is provided on the transportation-land use activities or opportunities (work, shopping, etc.) dis-
relationship and the chronological development of tributed in space. Changes in relative accessibility are
transportation and land use modeling. measured indirectly when researchers attempt to iden-
The first two modeling frameworks to be discussed tify the influence of new infrastructure, such as a high-
are those based on aggregate models of spatial interac- way link or transit station, on local land markets. In
tion and econometric models. These two modeling these cases, accessibility is usually approximated by
frameworks provide the vast majority of current oper- some measure of access to the transportation network,
ational models that are used in planning practice. We such as travel time or distance (Ryan 1999). Generally,
might refer to these first two frameworks as “top- the degree of land market impact is related to the
down” modeling frameworks, since they specify the impact of the new transportation link on regional
interaction between transportation networks and loca- accessibility, and so is roughly proportional to the
tion as a set of aggregate relationships based on increase in speeds (and reduction in travel time) per-
the behavior of a representative individual, usually the mitted by the new link (Cervero 1984).
mean calculated from a representative sample of the To operationalize the transportation–land use rela-
population. The third class of models to be introduced tionship within models of transportation and land use,
falls under the general category of microsimulation measures of accessibility are incorporated in determin-
models. These models cover a number of different ing the location of activities. It is typically assumed
approaches to representing the dynamics of land use that households wish to locate in areas with higher
change and travel behavior, but generally share the accessibilities to opportunities such as employment or
common focus of attempting to disaggregate the pop- shopping, while firms are assumed to locate in areas
ulation and to simulate changes from the “bottom up,” with higher accessibility to labor markets, perhaps
redefining the nature of actors in the model. Models of stratified by occupational type. In models where land
activity-based travel are discussed here, along with and floor space markets are considered explicitly,
multi-agent models and cell-based models, a special these accessibility factors can be important determi-
type of multi-agent model that offers an alternative nants of price. Since most models of transportation
mechanism for representing the dynamics of land use and land use contain a land use component that is
change. Some examples of prototype urban models integrated with, or at least loosely coupled with, a
that are being developed entirely within a microsimu- travel demand model containing a network assign-
lation framework are described. ment component, congested network travel times can
The later sections of the article review some of the be fed into the calculation of accessibility, thus provid-
common criticisms directed toward land use and ing a measure of the impact of congestion on regional
transportation models and note how these criticisms accessibility and activity location.
have (or have not) been addressed in the most recent To simulate these changes within models of metro-
generation of models. Some outstanding issues are politan regions, the region is typically broken down
discussed and suggestions offered as to important into a set of small geographic zones, similar (or in
future research directions. A concluding section fol- many cases identical) to the set of zones used for
lows with some general remarks on the state of trans- regional travel forecasting. Accessibility is typically
portation and land use modeling and its relationship calculated from each zone to all other zones in the
to planning as a discipline. region via the regional transportation network.
Changes to the travel network that alter zone-to-zone
travel times thus impact the relative accessibility of a
THE TRANSPORTATION–LAND USE RELATIONSHIP
location.
Transportation networks and the spatial patterns of
land use they serve are assumed to mutually influence
CHRONOLOGY OF MODEL DEVELOPMENT
each other over time. Changes to transportation net-
works, such as the construction of a new link or expan- The history of simulation models of transportation
sion of an existing one, eventually influence the and land use dates to the late 1950s (Batty 1979). While
location of investment in land, which in turn influences models of regional travel demand had been established
 Models of Transportation and Land Use Change 3

SPATIAL INTERACTION MODELS

Time
1960 Land Use and Transportation Models The earliest class of land use and transportation
simulation models are a set of highly aggregate
models based on principles of spatial interaction that
Spatial were popular in the regional science and quantitative
Interaction/
Gravity- geography fields in the 1950s and 1960s. There were
Based many different formulations of this type of model,
1980 Models
though most revolved around variations of the gravity
model, an adaptation from Newtonian physics. The
derivation of the gravity model from principles of
Econometric
Models
entropy maximization (Wilson 1967, 1970) was a major
Cell- accomplishment and formed the basis for many of the
Micro- Agent- Based
2000 simulation Based Models allocation mechanisms within spatial interaction
n Models models. A general form of the gravity model can be
Models
present expressed as:

FIGURE 1. Chronological Development of Land Tij = AiBjOiDj exp(−βcij)

Use and Transportation Models
where Tij represents trips (or other measures of
interaction) between two zones, Oi represents origins at
as far back as the early 1950s and some early experi- zone i, Dj represents destinations to zone j, and Ai and
ments with transportation and land use models were Bj are balancing factors to ensure that total origins
carried out in the following years, it was not until the equal total destinations. The exponential term in the
early 1960s that the first operational land use simula- model is used to capture the effect of decreasing inter-
tion model was built. The “Model of Metropolis” action as a function of travel cost, including travel time.
developed by Lowry (1964) is widely considered to be As mentioned previously, the first operational land
the first operational simulation model of urban land use simulation model was the model developed by
use. Lowry’s model was the first of a generation of Lowry (1964) for the Pittsburgh region. This model has
models based on theories of spatial interaction, includ- great importance, since many of the other land use and
ing the gravity model that was popular in quantitative transportation models that follow a spatial interaction
geography at the time. Models based on a spatial inter- framework have similar structures. A detailed review
action framework continued to be developed through of this model and its variations are provided in
the early to mid-1980s, when they became largely Horowitz (2004).
replaced by models grounded in random utility theory
The Lowry Model and Derivatives
and econometric methods.
Figure 1 describes this process and gives an approx- The land use model developed by Lowry was a spa-
imate timeline for the adoption of various modeling tial interaction model designed to simulate patterns of
frameworks within transportation and land use residential and service location in the Pittsburgh,
research. Several of the models that follow an econo- Pennsylvania, region. The impetus for building the
metric framework continue to be used today, although model was to be able to simulate the effects of urban
some, like the UrbanSim simulation system (Waddell renewal and slum clearance programs on the distribu-
2002a; Waddell et al. 2003) are being redeveloped tion of activities within the region. The model borrowed
within a microsimulation design. The broad class of from economic base theory, which divides a region’s
transportation and land use models that could fall employment into basic and nonbasic services. Basic
under the title of “microsimulation” began to be devel- industries are assumed to export much of their product
oped in the early 1990s, in parallel with major outside the region, generating additional income that
improvements in computational power that allowed can then support additional nonbasic services.
for their operation. These included prototype models Nonbasic industries then serve households (e.g., retail
of activity-based travel, cell-based models land use activities) and other industries within the region.
change, and the introduction of multi-agent models Lowry’s model assumed that the location of basic
for urban simulation. More recently, some researchers industries was fixed. This required an initial allocation
have begun to devote effort to developing comprehen- of basic employment to zones within the region.
sive urban microsimulation models that fully reflect Households were then allocated to zones from the
the dynamics of changes in the population and the initial basic employment locations, using a function
urban environment within which they make choices. similar to the deterrence function used in the trip
4 Journal of Planning Literature

distribution step of most trip-based travel forecasting Several models extended the basic Lowry frame-
models (Horowitz 2004): work in new directions. Table 1 lists some of these
models along with their distinguishing features. For
f(tij) = exp(−βtij) example, the Time Oriented Metropolitan Model
(TOMM) described by Crecine (1964) disaggregated
where f(tij) is a deterrence function value represent- the population into socioeconomic groups to improve
ing the inverse of the likelihood of workers working in the model’s representation. It also differed from the
zone i and living in zone j, and tij is a measure of the Lowry model in that only some of the nonbasic activi-
disutility of travel between zones, typically defined as ties in a region would be reallocated between model
travel time, and -? represents the marginal disutility iterations, reflecting a certain degree of inertia in loca-
per unit of time. This functional form implicitly tion. Garin (1966) recast the original Lowry model by
assumes that workers choose to locate near their work- proposing a matrix representation for the model’s
place and that only one household member is components and substituting a production-constrained,
employed outside the home. Lowry chose to define gravity-type interaction model as the basis for alloca-
this measure of disutility as the airline distance tion. Garin’s version also allocated all activities at each
between zones. He did this partially because of the dif- iteration, an improvement over Lowry’s formulation
ficulty of generating matrices of trips between zones since it improved the coupling between allocation and
using the travel models that existed at the time, but generation (Timmermans 2003). Another land use
also because he noted a high degree of correlation model designed by Goldner (1971) allocated activities
between observed airline and network distances in his according to an intervening opportunity model, a spe-
study region (Lowry 1964). Using the deterrence func- cial case of the gravity model (Wilson 1971). The
tion described above, f(tij), the number of workers design of the model also sought to improve realism by
working in zone i and living in zone j (defined here as using different dispersion parameters for each of the
Tij) could be calculated by using a modified expression nine counties of the San Francisco Bay area, where it
that included a value of attractiveness for each resi- was calibrated and tested.
dential zone (wj):
ITLUP/METROPILUS
eiwj f(tij) Building on the Lowry–Garin framework, Putman
Tij =
∑ wj f(tij) (1974, 1983) developed the Integrated Transportation
j
and Land Use Package (ITLUP), widely considered to
where ei is the employment in zone i. The residen- be the first fully operational transportation–land use
tial attractiveness measure as used in this formulation modeling software package. ITLUP has been applied
simply relates to the amount of land available for resi- in over a dozen locations within the United States, and
dential development in a particular zone. Deleting the has been calibrated over forty times (Hunt, Miller, and
variable for zonal employment in the above expression Kriger 2005). Designed in the mold of the Lowry
yields an expression for the probability of residing in a model, ITLUP initially contained a land use model that
zone given a fixed workplace location that is very sim- was similar to Goldner’s PLUM model. ITLUP offered
ilar to the probability expression in the multinomial a network representation that allowed for the incorpo-
logit model. This relationship is important, since it is ration of congested travel times in the distribution of
used extensively in transportation and land use activities. At the core of ITLUP were two allocation
models that derive from random utility theory, as will submodels: a household allocation submodel called
be discussed in the next section. DRAM, and an employment allocation submodel,
The process of worker/household allocation is fol- EMPAL. Trip generation and distribution functions for
lowed by a similar process in which the locations of the travel forecasting model are developed within
nonbasic industries serving households and other DRAM, simultaneously with household location,
(basic) industries are allocated assuming fixed loca- while mode choice and trip assignment are handled
tions for these quantities. Once these activities have with separate submodels. Travel times from runs of
been allocated, it is possible to couple the land use the travel model are fed forward to calculate new
model with a conventional, trip-based travel forecast- activity distributions.
ing model to produce a set of network flows. These More recently, the ITLUP model framework has
new flows and travel times can be used to modify the been updated to incorporate modifications to some of
deterrence function and produce a new allocation of its submodels and new data and visualization tools
households and nonbasic employment. (Putman 2001). The new package, called METROPILUS,
 Models of Transportation and Land Use Change 5

TABLE 1. Summary of Spatial Interaction / Gravity Models

Model Reference Distinguishing Features

Model of Metropolis Lowry (1964); Garin (1966) First recognized operational land use model; Garin provided
matrix representation
TOMM Crecine (1964) Disaggregation of population; incorporation of inertia effects in
activity allocation
PLUM Goldner (1971) Replaced standard gravity model with intervening opportunity
model; use of county-specific dispersion parameters
ITLUP Putman (1983) First complete software package for integrated modeling;
improved calibration techniques; improved network model
with multiple modes; incorporation of congestion effects in
activity allocation
LILT Mackett (1983) Use of accessibility function; car ownership submodel; land use
model capable of handling demolition, changing occupancy
and vacancy rates
IRPUD Wegener (1982) Contains seven separate submodels; microsimulation of land use;
use of differing spatial scales for submodels; separates
discretionary and non-discretionary travel

is housed within a geographic information system (GIS) rehabilitation and demolition, change of job, change of
environment that permits improved visualization of residence, and car ownership/travel demand. IRPUD
output. Other important features of METROPILUS is somewhat unusual in that it contains a microsimula-
include multivariate, multiparametric attractiveness tion model of land use, in which land uses are allowed
functions that include lag terms to better capture loca- to change through aging. Another desirable feature of
tion dynamics. The addition of zonal constraints can IRPUD’s design is that it allows different submodels to
limit allocation of activities to zones where land is not take place at different spatial scales (intra-regional loca-
available. Land supply in the model is managed by a tion takes places at a meso-scopic scale, while land
land supply function that translates the location development takes places at a micro/tract level). These
demands from employers and households from features are emulated in some of the newer, emerging
DRAM and EMPAL into land uses and intensities. urban microsimulation models. As a practical matter,
the IRPUD model can be classified along with other
LILT AND IRPUD
spatial interaction models, since it uses gravity models
Two other spatial interaction-based models merit to allocate the distribution of land use.
attention, since they have been extensively applied The first generation of land use and integrated
and tested. The first is the Leeds Integrated Land Use transportation and land use models based on spatial
(LILT) model, developed by Mackett (1983, 1991). LILT interaction formulations produced a multitude of
combines a Lowry-type land use model with a con- models that were tested and applied in numerous set-
ventional, four-step travel model. Forecasts of change tings. Some models, such as the METROPILUS plan-
in population are allocated to zones according to ning support system package, continue the legacy of
accessibility functions derived from work trips and these models to the present. However, very few exam-
zonal attractiveness functions. Other salient features of ples of this type of model framework remain. The
LILT include the ability to handle demolition, chang- shortcomings of these models were numerous: most
ing occupancy rates and vacancies, and a car owner- were static equilibrium models incapable of capturing
ship submodel, which estimates vehicle ownership as the dynamics of urban systems; none of the models
a function of network travel times and costs actually represented land markets with explicit prices;
(Timmermans 2003). zones were highly aggregate and lacked spatial detail,
The IRPUD model (Wegener 1982) was developed and the models were inadequately supported by
by Wegener and colleagues at the University of theory. Inadequate theory may have also been a reason
Dortmund in Germany. IRPUD is quite complex and that many of the models forecasted so poorly. There
contains seven interlinked submodels of aging, firm were many high-profile failures in terms of using the
relocation, residential and nonresidential construction, models for policy analysis purposes (Batty 1979).
6 Journal of Planning Literature

Some of these were seized on by Lee (1973) in a cri- input–output model, replacing the trip generation and
tique that highlighted some of mistakes of the first distribution steps in trip-based travel forecasting
generation of models. Lee characterized them as being models. The trade flows are converted to demand for
too complicated, overly aggregate, data hungry, commercial and passenger traffic through the applica-
wrongheaded, extraordinarily complicated, too tion of scaling constants. The generated traffic is then
mechanical, and expensive. Many of these criticisms fed into models of mode and route choice. Congestion
informed the next generation of models, which took and travel times from the transportation model are
their cue from developments in econometric modeling then fed back into the land use and economic model,
based on random utility theory. yielding time-lagged measures of accessibility, which
affect location choice. The structure of MEPLAN,
including its spatial economic model, makes it appro-
ECONOMETRIC APPROACHES
priate for modeling not only at an intraurban scale but
As noted previously, one of the major shortcomings also at an interurban scale. It has been used in a vari-
of the aggregate spatial interaction models was the ety of major applications, including modeling the
absence or use of inappropriate theory to describe regional impacts of the Channel Tunnel between
the behavior captured in the model. Developments in the England and France (Rohr and Williams 1994).
use of random utility theory to describe choices among The TRANUS model (de la Barra 1989) is similar to
discrete alternatives, such as the choice of travel mode, MEPLAN in that it incorporates a spatial input–output
provided the impetus for a new generation of models model as the basis of its generation and allocation of
based on the study of disaggregate behavior. When it activities. The regional economy is disaggregated into
was shown that discrete choice models could be sectors, with the demand for each zone and sector gen-
applied to problems such as residential location erated and then allocated to production zones and sec-
(Lerman 1976; McFadden 1978), researchers began to tors via a multinomial logit model. A land supply model
look for ways to model the interrelated choices indi- is also available to simulate the behavior of developers,
viduals made in terms of location and travel behavior. who choose where to build (new land versus existing
Land use and transportation models that follow sites), what type of space to build, and at what density.
econometric frameworks can be thought of as compris- This choice process is governed by explicit prices or
ing two types of models: regional economic models rents for new or replacement stock, demolition, and
and land market models. In these two types of simula- building costs. Another unique feature of TRANUS is its
tion models the economic model and the land market relatively advanced trip-based travel forecasting model.
model each form the core of a simulation system that Similar to MEPLAN, flows of traffic between zones are
includes the prediction of transportation flows. Both generated from input–output matrices. Personal travel is
types tend to have improved representation of land estimated by time of day by mode as a function of cost.
markets that include endogenously determined (deter- Trips are assigned to the network according to distinct
mined within the model) prices and market clearing mode–path combinations. Accessibility is calculated as a
mechanisms. A summary of these models and their logsum composite utility measure from the mode choice
characteristics are provided in Table 2. model and input directly to the land use model to gen-
erate a new set of spatial flows.
Regional Economic Models
A third model system that takes as its centerpiece a
Two of the most important and widely used trans- regional economic model is the PECAS system
portation and land use models grounded in economet- (Production, Exchange, and Consumption Allocation
ric modeling approaches, MEPLAN and TRANUS, are System), developed by Hunt and Abraham (2005).
largely built around a core of a regional economic PECAS is a generalization of the spatial input–output
model. MEPLAN (Echenique et al. 1990; Echenique modeling approach used in MEPLAN and TRANUS.
2004) is a model that began as a simpler model of The model system is based on a quasi-dynamic equi-
urban stock and activity (Echenique, Crowther, and librium structure with flows of exchanges, including
Lindsay 1969) and expanded into a more comprehen- goods, services, and labor, from production to con-
sive urban simulation model. Similar to other types of sumption based on technical coefficients. Flows of
models, MEPLAN has a zone-based structure. In con- exchanges from production to zones of exchange and
trast to spatial interaction models though, the activi- from exchange zones to consumption are based on
ties in zones are determined by a spatial input–output nested logit models that take into account exchange
model that predicts trade flows by sector between prices and transport disutilties. Similar to other spa-
zones of a region, driving the demand for space. tial input–output models, trade flows are converted
Production and consumption are linked in the spatial to transport demands and loaded onto networks to
 Models of Transportation and Land Use Change 7

TABLE 2. Summary of Econometric Models

Model Reference Distinguishing Features

CATLAS Anas (1982) Improved representation of economic agents and decision making; explicit
treatment of housing markets; economic analysis capabilities
MEPLAN Echenique et al. (1969); Incorporation of spatial input-output model with economic evaluation
Echenique et al. (1990) component; able to forecast commercial trip generation; travel treated as
a derived demand
TRANUS de la Barra (1989) Development supply model simulates choices of developers; sophisticated
travel model with combined mode-route choice
MUSSA Martinez (1992) Incorporation of bid-rent framework for land, floor space markets; detailed
representation of transit network in travel model; high level of household
type disaggregation
METROSIM Anas and Arnott (1994) Model extended to commercial real estate markets; addition of dynamic
CHPMM housing market model
NYMTC-LUM Anas (1998) Endogenous determination of housing prices, floor space rents, and wages;
high level of spatial disaggregation suitable for transit and land use policy
evaluation
DELTA Simmonds (1999) Microsimulation of demographic changes; treatment of quality in the market
for space
PECAS Hunt and Abraham (2005) Regional econometric model with microsimulation of land development
at the parcel level; ability to couple with an activity-based travel model
and to apply at supra-regional level

calculate congested travel times (disutilities). Exchange Anas and colleagues developed a series models
prices for space drive changes in available space, simu- (Anas 1982, 1998; Anas and Arnott 1994) designed to
lating developer actions. The model system is run in simulate the effects of transportation improvements on
one-year time steps, with travel disutilities and changes land markets and overall social welfare. The first such
in space in a given year influencing the flows of model, CATLAS, emphasizes a discrete choice frame-
exchanges in the next year (Hunt and Abraham 2005). work to describe both the supply and demand sides of
PECAS now features activity-based travel modules, the housing market. The supply side of the model con-
as well as microsimulations of land development, with tains vacancy–occupancy, construction, and demolition
land parcels as the unit of analysis. While PECAS is submodels that respond to factors such as construc-
run at the scale of a metropolitan region, it can, like tion costs, land prices, taxes and operating costs, and
other input–output models, be adapted to larger-scale expected future resale values. Developers are assumed
applications. Recent versions of the model system to be profit maximizers, and so select the location
have been applied in statewide models of land use and and type of construction to maximize profit. The demand
transportation for Ohio and Oregon, as well as metro- side of the model takes a nested logit choice model form,
politan-level applications in Sacramento, CA, and assuming that households have a fixed workplace loca-
Calgary and Edmonton, Alberta, Canada. tion and choose a residential location and travel mode to
maximize their utility. Only two workplace locations are
Land Market Models
considered in the model (CBD and non-CBD), though
Improved land market representation is a distin- commuters have a variety of modes available (auto, bus,
guishing characteristic of many of the econometric heavy rail, and commuter rail), depending on their resi-
approaches to transportation and land use models. In dential location. The model is calibrated with census
fact, several them have at their core markets for resi- data and can predict changes in mode splits, house
dential and commercial real estate, with transportation prices and rents, demolitions, and new construction
models linked into the overall model structure. Some activity (Anas 1987). The economic evaluation compo-
of these models, such as those developed by Anas nent of the model estimates changes in economic wel-
(1982, 1984), seized on theoretical advances in linking fare because of changes in modal utility arising from
the related strands of gravity-based models with those investment in different modes. The changes in utility are
based on the multinomial logit specification (Williams captured in an inclusive value (logsum) accessibility
1977; Anas 1983). measure and are capitalized into housing prices or rents.
8 Journal of Planning Literature

The original CATLAS framework was modified in simultaneously with both sides of an auction in a bi-level
an enhanced model called METROSIM (Anas and framework. The MUSSA system provides an equilib-
Arnott 1994), designed for the New York City metro- rium model of building stock supply and demand,
politan region. METROSIM incorporated a dynamic where buyers maximize their surplus, sellers maxi-
model of metropolitan housing markets (Anas and mize price, and builders maximize profits. Building
Arnott 1994), along with a model of commercial floor stock prices are then endogenously determined in
space markets. The full modeling system combined the model.
models of employment, residential and commercial The MUSSA system also includes a rather sophisti-
real estate, vacant land, households, work and non- cated four-step travel forecasting model that is linked
work travel and traffic assignment, which was absent to the land use component. The travel model features
in the CATLAS system. A recent extension of this sys- a detailed transit network representation and the abil-
tem is the NYMTC-LUM model (Anas 1998), a simpli- ity to forecast demand for eleven separate alternatives,
fication of METROSIM designed to facilitate the including road, transit, and mixed modes. The com-
evaluation of changes in transit policies for the New bined transportation and land use models are referred
York City transit system. The model is slightly refined, to as 5-LUT (indicating a five-step forecasting proce-
adding a local labor market submodel and using very dure), and are able to provide equilibrated forecasts of
small zones to better model transit and auto network land use and travel demand. A notable feature of
flows. The combined model determines housing prices MUSSA is that the model uses smaller-than-average
and floor space rents endogenously (within the zones as units of analysis to achieve a higher level of
model), and uses modal utilities from the mode choice spatial disaggregation. Likewise, there is an effort to
model as accessibility inputs to the land use model. disaggregate the treatment of households within the
A similar framework was adopted by Simmonds model, with the Santiago application containing sixty-
(1999) in developing DELTA, a land use model five different household types. This is an important
designed to form the basis of a dynamic model system step in the development of transportation and land
of land use and transportation interaction. The model use models, and one that is being replicated in the cur-
system is divided into processes that represent spaces rent generation of transportation and land use models
and those that represent activities. Processes dealing based on microsimulation techniques, as will be dis-
with activities include household formation and disso- cussed in the following section.
lution, employment growth or decline, location and Another transportation and land use simulation
property markets, and the employment status of indi- model that adopts this highly disaggregate structure is
viduals. Processes representing the change in spaces the UrbanSim model developed by Waddell and col-
predict the quantity and quality of floor space avail- leagues (Waddell 2000, 2002a). Like MUSSA,
able. The model system is designed to be run over a UrbanSim is primarily a model of land markets,
series of short steps of no more than one or two years, though extensions have been considered to add an
and was originally coupled with START, a transporta- activity-based travel forecasting model (Waddell
tion model developed for the city of Leeds, United 2002b), as well as an environmental analysis module
Kingdom. A distinguishing feature of DELTA is (Waddell and Borning 2004). Like MUSSA, UrbanSim
attempts to add a quality variable to the prediction of initially contained a highly disaggregated household
location choices. In the case of residential location, the treatment, with 111 distinct household types identified
quality variable relates to local income and vacancy in an early calibration of the model (Waddell 2000).
rates. Hence, the quality of development can change Demographic transition in population and household
over time. The DELTA model has seen several applica- formation are microsimulated within a separate sub-
tions in the United Kingdom and parts of Western model. Residential mobility of households is charac-
Europe and is currently being developed as a terized by a two-stage process in which households
microsimulation model system. decide whether to search and then whether to move.
An alternative framework for modeling land mar- Location choice of households and firms are repre-
kets in transportation and land use models was pro- sented by a multinomial logit model considering all
vided by Martinez (1992, 1996), who built an zones in the region within the choice set. While
integrated model called MUSSA for the city of UrbanSim makes extensive use of econometric models
Santiago, Chile. MUSSA adopted a modified version in its structure, predictions are based on Monte Carlo
of the “bid-rent” framework for land markets, first simulation methods, indicating that it also has the
articulated by Ellickson (1981). The “bid-choice” characteristics of a microsimulation model system.
framework used by Martinez combines bid-rent and UrbanSim’s structure is also unusual in that it
discrete choice approaches to land markets by dealing operates in disequilibrium from year to year, with no
 Models of Transportation and Land Use Change 9

general equilibrium in land markets assumed at the in several of the models, as well as improved accessi-
end of a time step, though market clearing does occur bility measures based on utility functions. Also, the
at the transportation analysis zone (TAZ) level. This introduction of model systems built around a regional
feature sets it apart from the preceding models that economic model allowed for the inclusion of commer-
incorporate land markets, which are typically static cial travel in forecasts and the general treatment of
within each time step of a simulation. Researchers in travel as a derived demand. Despite these advance-
the field of urban modeling have previously com- ments, many of the econometric models retained a
mented on the importance of modeling different ele- number of problems left over from the previous gen-
ments of urban systems at the time scales in which eration of models. For example, most of the models
they operate (Wegener 1994; Miller 2003). Since urban remained highly aggregate, despite the use of disag-
areas do not really ever reach a general equilibrium in gregate calibration methods. This became one source
land and travel markets, this disequilibrium structure of bias in the model forecasts. Also, with the exception
will likely be adopted in many future attempts to of UrbanSim, all of the models were essentially static
model land markets. in nature. Their structure forced them to reach a gen-
UrbanSim’s model of land markets also estimates eral equilibrium between each time step in the model;
supply at the parcel level, using parcel databases this was especially true of the models focusing on land
within a GIS. Demand for housing and floor space are markets. Furthermore, little advancement was made
calculated at the TAZ level in the original version of in the transportation component of the model. Most
the model, though subsequent versions are attempting models continued to use trip-based, four-step forecast-
to reconcile the spatial scale of the supply-demand ing procedures, where all submodels except mode
relationship. Land markets are simulated using the choice were run at an aggregate level. Much of the
bid-choice framework, similar to the MUSSA model current research into microsimulation methods is
(Waddell 2000). Land prices are estimated from hedo- attempting to address this issue, along with other
nic regressions containing building unit and neighbor- pressing research questions in the design of compre-
hood characteristics, and regional accessibility to work hensive simulation models of transportation and
and shopping. The neighborhood characteristics are land use.
determined by partitioning the region into 150-by-150
meter grid cells, each containing information about
DISAGGREGATE AND MICROSIMULATION MODELS
neighborhood composition and nearby land uses.
Further work on UrbanSim is focusing on convert- Since the late 1980s, advances in computing power
ing it to a comprehensive microsimulation modeling and efficiency of data storage have allowed researchers
system (Waddell et al. 2003). Many of the elements of to begin to build models that address many of the short-
the original model loaned themselves to this treat- comings associated with previous large-scale modeling
ment, including the high level of household type dis- efforts and represent important change processes in
aggregation and demographic transition submodel. cities with the detail they require. Examples of these
The land market simulation is already highly disag- include activity-based models of travel behavior, multi-
gregated and requires only further refinement of agent models of urban land use and transportation, and
developer behavior. The structure of the model system cell-based models of urban land use. The common con-
suggests that modified transportation submodels, ceptual underpinning of each of these models is that
such as an activity-based travel model, could be cou- they attempt to represent processes of change from
pled with the other elements in the model system. the bottom up, that is, they account for the behavior of
Long-term goals of the project include developing the individual agents in space and/or time, along with
software architecture to support an agent-based simu- interactions between agents. The use of the term
lation version of the modeling system and the explo- microsimulation can be applied to each of these types of
ration of new model structures. models, though it requires some definition. As defined
The experience with the generation of transportation by Miller (2003), microsimulation relates to “a method
and land use models based on econometric frameworks or approach (rather than a model per se) for exercising
has been valuable and addressed one of the most a disaggregate model over time.” All of the types of
pointed criticisms of the previous generation of spatial models identified above are what would be considered
interaction models, that of lack of theory. The use of disaggregate models and all have a significant temporal
random utility theory and advancements in discrete element. Microsimulation methods are particularly
choice modeling of individual behavior have allowed effective for modeling systems that are dynamic and
for the inclusion of economic evaluation components complex, which urban systems invariably are.
10 Journal of Planning Literature

Activity-Based Travel Models2 with an application to evaluate the impacts of control

measures in the Washington, D.C., metropolitan
Research into the foundations of travel behavior
region. Activity-based forecasting models incorporat-
dating back to the 1970s has identified many short-
ing GIS applications have also been developed by
comings in the use of sequential, trip-based travel-
McNally (1998). Bowman and Ben-Akiva (2001) struc-
demand forecasting models (Hagerstrand 1970;
tured a model of activity participation within a nested
Chapin 1974). However, there was little incentive until
logit framework to predict travel tours (clusters of
this time to attempt to recast travel forecasting proce-
chained trips). Their model was calibrated using travel
dures. Oil crises during the 1970s precipitated research
survey data from the Boston region. A model system
into various energy use reduction strategies, including
developed by Arentze and Timmermans (2004)
demand management measures and transportation
attempted to simulate learning behavior by agents
system management techniques. It was then that the
within the context of activity scheduling and travel
inability of existing forecasting models, which were
behavior. Perhaps the most ambitious effort to date in
mostly static and aggregate, to predict behavioral
the United States has been the research program asso-
responses to such policy measures became apparent
ciated with the TRANSIMS modeling system, which is
(McNally 2000).
designed to combine an activity-based forecasting
A combination of factors brought about resurgence
model with a region-wide traffic microsimulation sys-
in interest in reconceptualizing travel behavior for
tem (Barrett 1995).
modeling purposes during the 1990s. The completion
Activity-based models are necessarily disaggre-
of the interstate system and the difficulty of expanding
gate and attempt to simulate travel behavior within
existing urban road networks led many regional plan-
the limits of time and space. Because of spatial and
ning organizations to emphasize preservation and
temporal interdependencies, this process cannot be
management of transportation systems through such
modeled within a framework that treats trips as
policies as flexible working hours, travel information
independent and generates trips at an aggregate
provision, traffic flow improvements, and diversion of
level. An alternative, agent-based approach is typi-
some travel to alternate modes. The potential changes
cally adopted in formal travel forecasting applica-
in travel behavior implied by these policies cannot be
tions. This focus on the behavior of individual agents
forecast using existing methods, since trip-based
and addition of temporal elements makes activity-
models separate travel decisions from their broader
based travel models a natural complement to
context of activity participation and temporal con-
microsimulation models of transportation and land
straints. At the same time, improvements in comput-
use that focus on the activity of agents at an individ-
ing power and the use of geographic information
ual or household level.
systems have allowed for the formalization and testing
of models that previously existed only at conceptual or
Agent-Based Microsimulation Models
limited empirical levels. Support from the Federal
Highway Administration in the form of the Travel The state-of-the art in transportation and land use
Model Improvement Program (TMIP), which attempted modeling is defined by current research efforts aimed
to improve the state of practice in transportation mod- at building comprehensive microsimulation systems
eling and facilitate development of a new generation of urban areas, with representation at the level of indi-
of travel demand models, has also had a significant vidual agents (persons, households, firms, etc.) and
impact. simulations of the behavior of the entire population of
The first demonstration of an operational model of interest. The advantages of adopting such a modeling
activity-based travel preceded the TMIP, and was con- approach for urban systems are many (Miller 2003):
ducted by Recker, McNally, and Root (1986a, 1986b).
The STARCHILD model was developed to investigate • Urban systems are dynamic, with a significant time ele-
dynamic ridesharing, but was designed for research ment and components changing at different speeds.
• The behaviors of these systems are complex, with inter-
purposes only and required collection of data that is
acting agents, complex decision-making processes, and
still not commonly available (McNally 2000). Models significant probabilistic elements.
of activity chains and travel behavior were coupled • Closed-form mathematical and statistical representa-
with a mesoscopic traffic simulation in work by tions of urban systems often introduce large amounts
Axhausen (1990). Pendyala et al. (1997) developed an of bias and lead to poor forecasts.
activity-based simulation model capable of predicting
activity scheduling changes in response to transporta- The seeds of comprehensive microsimulation
tion control measures. They demonstrated their model models had been sown in a number of earlier models,
 Models of Transportation and Land Use Change 11

TABLE 3. Summary of Microsimulation Models of Transportation and Land Use

Model Reference Distinguishing Features

ILUTE Salvani and Miller (2005) Comprehensive urban system microsimulation model; structured to accurately
capture temporal elements urban change; activity-travel model includes
household member interactions; disequilibrium modeling framework
ILUMASS Moeckel et al. (2003); Descendent of IRPUD model; incorporates microscopic dynamic simulation
Strauch et al. (2003) model of traffic flows and goods movement model; designed with environmental
evaluation submodel;
Ramblas Veldhuisen et al. (2000) Entirely rule-based model framework; designed to simulate very large populations
UrbanSim Waddell et al. (2003) Land use model incorporating microsimulations of demographic processes land
use development; parcel-level land use representation; high level of household
type disaggregation; open-source software developed for general use

where one or more elements of the system were gov- participation and travel, as well as goods movement.
erned by a microsimluation process. For example, The activity-travel module uses data collected via a
Wegener’s IRPUD model contained microsimulations hand-held survey instrument. This innovation in data
of population and building stock. Mackett’s (1990) collection allows for near-real-time information on
MASTER model simulated location choices and travel activity and travel behavior, obviating the need for
decisions, and MUSSA and UrbanSim disaggregated respondents to recall their activities later on. The GIS
households at a level sufficient to operate them in a sta- component of ILUMASS combines raster-based and
tic microsimulation format, where a representative vector-based representations, allowing for the advan-
sample is used within a microanalytic framework for tages of spatial disaggregation in land use representa-
short-run applications. However, for long-term fore- tion and efficient network algorithms for the
casts, which most transportation and land use models transportation network model.
are designed for, the population must be synthesized or The ILUTE model (Salvani and Miller 2005), being
updated to represent the dynamics of individuals and developed by researchers at a number of Canadian
the environments within which they make choices. universities, chiefly the University of Toronto, repre-
An overview of some of the comprehensive sents the most complete microsimulation model to
microsimulation systems currently under develop- date. The product of a long-term effort to design an
ment are presented in Table 3. The UrbanSim system “ideal” simulation model of transportation and land
was the only simulation model to transition from a sta- use, ILUTE centers around a behavioral core consist-
tic simulation format to a dynamic microsimuation ing of four interrelated components: land use, location
model. As noted previously, the original version of choice, auto ownership, and activity/travel patterns.
UrbanSim contained a number of microsimulation The model system is highly integrated with feedback
submodels within its structure, thus eliminating the mechanisms whereby higher-level (longer-term) deci-
need for as radical a redesign as would be needed for sions, such as residential mobility, affect lower-level
many of the static, equilibrium models. (shorter-term) decisions, such as activity participation
The ILUMASS simulation system (Moeckel et al. and travel. ILUTE is not based on a single modeling
2003; Strauch et al. 2003), being developed by a technique (e.g., random utility), but rather uses a vari-
research team at the University of Dortmund, builds on ety of modeling approaches to represent the behavior
the experience of Wegener and others with the IRPUD of agents in the model, such as state transition models,
model in the 1980s. The design of ILUMASS embeds a random utility models, computational rule-based
microscopic dynamic simulation model of urban traffic models, learning models, and hybrids of previous
flows within a comprehensive model system incorpo- approaches.
rating changes in land use and building stock. ILUTE’s treatment of land markets explicitly
The microsimulation modules of ILUMASS include assumes a constant disequilibrium framework, indi-
models of demographic change, household formation, cating that a particular house could be on the market
firm lifecycles, residential and nonresidential con- for several months without selling, since no market
struction, labor mobility in a regional labor market, clearing is assumed. The time steps in the model are
and residential mobility in a regional housing market. brought down to the level of months, rather than years,
These modules are linked with models of daily activity to provide greater temporal detail. The disequilibrium
12 Journal of Planning Literature

framework and absence of market clearing also means Cellular Models

that projects with extended construction periods (e.g.,
greater than one year) can be accommodated. The The representation of land use in integrated models
housing market submodel within ILUTE assumes a of transportation and land use change has been one of
three-step process to describe residential mobility, the less satisfactory elements of these models (Chang
involving a mobility decision, a search process, and 2006). Until recently, land use had generally been rep-
bidding and search termination. resented by zones that served as convenient areal units
The transportation component of ILUTE is quite for the location of activities, and coincided with zonal
sophisticated and includes submodels for automobile designations for transportation models. Models that
transactions and activity scheduling. The activity provide greater simplicity and a clearer representation
scheduling submodels characterizes activities as occur- of the dynamics of land use change using cell-based
ring in time and space, with various scheduling depen- representations of regions have emerged within the
dencies to represent temporal constraints (Roorda, past two decades as an increasingly attractive land use
Doherty, and Miller 2005). Future plans include adding modeling alternative.
a network model, which is needed to provide travel Cell-based models, and particularly those based on
times and costs by mode, along with a formal model of cellular automata (CA) theory, arise from the applica-
activity participation. Like most comprehensive tion of complexity theory to cities (Batty 1997, 2005).
microsimulation models, ILUTE is still in the process of Complexity theory conceptualizes systems, such as
calibrating some of the submodels in the system, and urban systems, as being too complex to synthesize
has yet to be used in a full forecasting application, using closed-form, predetermined mathematical rep-
though the travel demand component has been applied resentation. Rather, these systems arise from the col-
in a policy simulation (Roorda and Miller 2006). lective interaction and self-organization of large
Another agent-based simulation model that merits numbers of individual agents that generate the
attention is the Ramblas model (Veldhuisen, observed macro-level states (Benenson 1998). Cell-
Timmermans, and Kapoen 2000, 2005). While it is not based models of land use can range from simple state
as comprehensive as the other models described here, transition models in which cells change states (land
Ramblas is designed to simulate the effects of land use uses) according to some observed probability, to the
and transportation planning policies, with an empha- more general form of CA, in which cell states are also
sis on the prediction of activity participation and traf- a function of states in neighboring cells. CA models
fic flows. An unusual aspect of Ramblas is that it is can be seen as extension of agent-based microsimula-
designed to simulate the effects of policies on the tion models, in which individual cells are the agents,
entire Dutch population (estimation at more than six- rather than persons or households.
teen million). The model also distinguishes itself by CA models generally require four basic elements: a
being entirely rule-based, rather than adopting a for- lattice of regular spaces or cells, a set of allowed states,
mal theoretical framework to guide the behavior of neighborhoods that are defined by the lattice, and a set
agents. These aspects of the model derive from its of transition rules governing the evolution of individ-
stated purpose of being a practical planning tool to ual cells in the system. Many CA models also add a
assess the impact of various transportation and land fifth, temporal element. CA models are basically deter-
use scenarios. ministic, rule-based models, using “if-then-else” logi-
Ramblas is run by selecting households, stratified cal statements to build their transition rules, though
according to size and structure. Individuals are classi- stochastic elements can be added to transition rules
fied according to one of twenty-four population seg- using probabilistic expressions and random number
ments, defined on the basis of age, gender, education, generation. Other types of modifications to CA models
and employment status. An activity agenda and trans- intended to introduce complexity include changes to
portation mode are drawn at random, with seven the structure and dimension of the lattice of cells,
activity types available. Destinations are randomly expansion of allowable cell states, expanded neighbor-
drawn from a choice set, sometimes delimited by hood definitions to include action at a distance, and
a given action space or distance constraint. Origin- changes to temporal elements, such as Markov chains
destination pairs are generated from the activity and (Torrens and O’Sullivan 2001).
mode allocations and traffic flows are then microsim- A precursor to many of the contemporary cellular
ulated, calculating travel times via a speed-flow models being used to describe the dynamics of urban
method. Output from the microsimulation of traffic is systems is the model of self-forming neighborhoods
used to forecast changes in land use, dwelling stock, presented by Schelling (1978). As part of a larger
and road construction. exposition of self-organizing principles, Schelling
 Models of Transportation and Land Use Change 13

demonstrated how “individually motivated” forms of to be an important determinant of residential develop-

segregation could arise through the interaction of ment and density.
many agents (households) pursuing their own objec- CA models appear to be growing more complex.
tives. Preferences for individuals of a different race, Their many applications reflect the relative ease and
income, or any other form of social stratification were flexibility with which they can be modified to describe
shown to lead to highly segregated outcomes under a processes of change. CA models are not without their
variety of initial conditions and preference structures. weaknesses, though. Their simplicity, which is one of
The compatibility of CA models with GIS, remote their most desirable attributes, is also a significant lim-
sensing data, and associated visualization capabilities itation. In most cases, they are inappropriate for mod-
make them particularly suitable for land use modeling eling systems with complex interactions. For example,
applications (Torrens and O’Sullivan 2001). It is here processes such as land development represent the
that they have received the most attention. One interaction between human and physical systems, but
example is the model of urban land use developed by CA models cannot capture both. Also, CA models are
Clarke, Hoppen, and Gaydos (1997) to estimate the not designed to be forecasting tools. Since they are cal-
regional impact of urbanization on the San Francisco ibrated on historical data and lack a strong behavioral
Bay Area’s climate. This model is an example of a self- interpretation, most forecasts have little meaning.
modifying CA, in which the CA can adapt to the cir- Rather, CA are better suited to idealized principles of
cumstances it generates. Clarke and Gaydos (1998) cities and urban design applications than large-scale
applied the same model to the Baltimore–Washington simulations or strategic planning (Batty 1997).
region to generate long-term urban growth predic-
tions. Jantz, Goetz, and Shelley (2004) also studied
growth in the Baltimore–Washington region using CA, RESOLVED AND ONGOING MODELING ISSUES
with the objective of simulating the effects of different
Old Issues
patterns of land use on the Chesapeake Bay water-
shed. Levinson and Chen (2005) describe the develop- The models currently being developed to describe
ment of a Markov Chain model of land use change for change in transportation and land use systems look
the Minneapolis–St. Paul region. Their model adopts very different than those that existed a few decades
the discrete-time version of a Markov Chain and pre- ago. One might question then to what extent these
dicts the evolution of transportation networks and newer models have overcome the deficiencies of ear-
land use patterns over the period from 1958 to 1990. A lier generations of models, such as the criticisms
next step for this model would be to add neighbor lodged against the first generation of spatial interac-
effects, which would move it to a CA–Markov Chain tion-based simulation models.
framework. Reflecting on the earlier experience, some modelers
Other applications of CA include simulating land claimed in the early 1990s that advances in computer
use density conditions, as in the model developed by processing power and data storage would obviate
Yeh and Li (2002). Their model incorporates a density many of the problems identified by Lee (1973) in his
gradient in the simulation of urban development for critique of the early modeling experience (Harris
different urban forms. The transition rules of their 1994). While these advances have undoubtedly
model specify a density, obtained from a distance- reduced some of the costs of building, operating, and
decay function, to be applied to cells as they are con- maintaining transportation and land use models, con-
verted to developed cells. Kii and Doi (2005) provide a comitant expansions in the scope of these models, as
similar application to demonstrate the effects of com- exemplified by the current generation of urban micr-
pact city form and mixed land use on total trip length, sosimulation models, ensures they will continue to be
energy consumption, and social welfare in Takamatsu, a resource-intensive effort. These models also remain
Japan. The model they present, MALUT, is a multi- highly complex, with many interacting submodels.
agent model of transportation and land use, where a Calibration is still a daunting task, even for models
CA model of land use is coupled with a microsimula- that are available as commercial packages. Data
tion model of travel. Accessibility can be incorporated requirements are still large, especially for dynamic
into a CA model of land use change, as demonstrated models that require synthesis of a population or con-
by Ottensmann’s (2005) LUCI2 model. LUCI2 was tinual updating of a sample.
designed to predict employment and land conversion It must also be recognized though, that a number
change over a forty-four-county region of Central of problems identified with earlier models have been,
Indiana, consisting of eight separate metropolitan sta- at least partially, resolved. Most microsimulation
tistical areas. The model found access to employment models are no longer static, and can simulate changes
14 Journal of Planning Literature

in transportation network performance and land use microsimulation models will reduce some of the bias
through time. Nearly all models now are able to model present in earlier, more aggregate models. However,
land markets with explicit prices and the ability to long-term forecasting models of many types necessar-
simulate the behavior of various agents in the land ily retain significant amounts of irreducible uncer-
development process. The level of aggregation of tainty, and the lack of available forecasting results
agents is being reduced, especially in comprehensive from applications of newer models leaves some room
microsimulation models. The size of zones in most for concern.
models is now much smaller, and should continue to
Treatment of Supply Side
decrease as computing power permits, though spatial
detail in many models could certainly improve. In most forecasting applications the supply side of
Perhaps most importantly, the theoretical basis of transportation, as represented by the extent and capac-
models has improved, especially in ongoing efforts to ity of networks, is held fixed or treated as a policy vari-
reconceptualize the relationship between individual able. The limited available evidence on the evolution
activity patterns and travel choices for travel demand of networks over time (Yamins, Rasmussen, and Fogel
forecasting. 2003; Yerra and Levinson 2005; Levinson and Yerra
2006; Zhang and Levinson 2007) suggests that scenar-
New Directions
ios such as alternative ownership regimes and their
The development of advanced models of trans- impact on transportation–land use systems are a topic
portation and land use change brings about opportu- worthy of exploration with more comprehensive
nities for exploring some important topics related to models.
the models themselves and their representation of
Agglomeration Effects
real-world urban regions. The following are some
issues worthy of more attention. Previous reviews of operational models of trans-
portation and land use (Berechman and Small 1988)
The Use of Theory
identified the absence of agglomerative effects as a
Some researchers question the continuing use of major weakness of the land use component of these
broad theoretical frameworks to guide agent behavior models. Recent work using multi-agent systems
in model systems. Timmermans (2003) points to the (Arentze and Timmermans 2003) suggests that model-
use of random utility theory to describe a wide range ing this effect is possible, and it is deserving of further
of spatial choices in many models. Noting that utility exploration.
is a concept that must be built up over several repeti-
Person-Based Accessibility
tive choice situations, he questions the applicability of
this concept to rare decisions such as mobility and res- Since accessibility is still seen as an important com-
idential location. Also, it is questionable whether dis- ponent of location choice in transportation and land use
crete choice methods and random utility theory are models, especially for residential location, it makes
applicable to entities such as firms, which comprise sense to pursue measures of accessibility that recognize
collective choice situations, as opposed to individual the importance of treating travel behavior as a process
agents. Models like Ramblas and ILUTE, which use constrained in time and space, as is reflected in activity-
rule-based or hybrid modeling approaches, suggest based travel models. Examples have been provided in
that tailoring the right tool to each model component work by Kwan and Weber (2003) and Miller (2005).
can overcome this issue. Timmermans also noted that Future work in these key areas holds some promise
most models are consumers of theory rather than pro- to improve the validity of land use and transportation
ducers, indicating that model development ought to models. Many of the suggested actions can be, and in
coincide with the process of theory development. some cases have been, incorporated into existing
models. Recent versions of UrbanSim have attempted
Forecast Accuracy
to simulate agglomeration effects by including a vari-
Recent studies that have sought to explore the prop- able in the utility expression for firm location choice
agation of uncertainty through transportation and reflecting the existence of employment in the same
land use models (Pradhan and Kockelman 2002; industry. As a proxy for agglomeration effects, this fea-
Krishnamurthy and Kockelman 2003; Clay and ture should improve location choice models by pro-
Johnston 2006) have identified a continuing trend of viding a complement to traditional accessibility
large variation in output from these models. measures as determinants of employment location.
Presumably, the addition of better model dynamics Incorporation of person-based accessibility mea-
and disaggregation of population groups within sures also seems feasible, and has been demonstrated
 Models of Transportation and Land Use Change 15

by Dong et al. (2006) in an application of activity-based and the older operational models still raise important
accessibility using an activity-based travel modeling questions about the utility of such complex tools.
system developed by Bowman and Ben-Akiva (2001). One must be more circumspect, though, in evaluat-
While this approach represents a definite improve- ing the transportation and land use modeling experi-
ment to the modeling of travel demand, it is unclear ence more generally. In reflecting on the experience
whether the use of activity-based measures of accessi- with the first generation of models nearly three decades
bility will greatly affect longer-term location decisions ago, Batty (1979) noted that models should be evaluated
as currently structured within land use models. in terms of their contribution to both science and design
Explicit treatment of the supply side in transporta- (i.e., policy). Many of the earliest models were failures
tion models would also improve their capacity to on both accounts, though there has arguably been some
reflect the dynamics of land use and network growth. success on the science side since then.
In a previous application, Levinson and Karamalaputi Models continue to represent an important means
(2003) demonstrated how factors such as existing traf- of testing theories and developing knowledge about
fic demand, demographic characteristics, and present the behavior of urban systems. For example, land use
network conditions could be used to predict network and transportation models have emphasized the role
expansion. Incorporation of a separate submodel of accessibility in location choices from their earliest
within current operational models seems a viable and origins. They have provided a method for formally
useful alternative to treating the supply side of trans- and quantitatively understanding this link and its
portation systems as fixed. effects on urban structure. Another contribution that
Perhaps the most important line of inquiry for modeling efforts have made has been to treat cities as
future work in land use and transportation modeling living, dynamic systems. While virtually all opera-
relates to understanding the accuracy and level of tional models include some type of feedback effect, the
uncertainty inherent in existing and proposed opera- increasing inclusion of dynamic effects in the form of
tional models. As indicated, some of this work has lagged responses to transportation network or land
already begun. Future work may continue to use sim- use changes has increased the realism and applicabil-
ulation methods, such as Monte Carlo or Latin ity of many models. Furthermore, recent modeling
Hypercube sampling (Hess, Train, and Polak 2006) to efforts are beginning to incorporate principles of self-
relate changes in inputs and model parameters to var- organization in representing urban growth and
ious model outputs. The outcome of this line of work change. Agent-based and microsimulation models
will provide important feedback to model users about demonstrate the importance of individual action, as
the acceptability of forecasts based on current opera- opposed to top-down decision-making, in represent-
tional model systems. ing dynamics of urban growth and change. Recent
microsimulation models such as UrbanSim have been
able to capture unique spatial effects, such as neigh-
CONCLUSION
borhood effects, in residential location choices.
Models of transportation and land use change have Agglomeration effects in firm location choice have also
evolved significantly since their early applications been added to reflect concentrations of an industry
more than four decades ago. In the search to design type in specific locations.
models that capture the recursive relationship Another observation by Batty related to the status
between transportation and land use, there has been a of planning as a science. He argued that planning was
general trend toward the disaggregation of the repre- (at the time) an “immature” science, marked by poor
sentation of people and space. Newer models repre- theoretical development, continuing controversy
sent in greater detail the dynamics of the about methods and results, and the tendency to follow
transportation–land use change process. Experiments “fashions.” In many respects this is true of the field
with bottom-up approaches to modeling urban sys- today, as in the continuing controversy over the influ-
tems, especially those that recognize the interactions ence of land use patterns on travel behavior. Batty sug-
between agents, provide an alternative means for gested that this status may be inherent to planning,
understanding their complexity. Yet, the ability to fore- which is considered a “policy” science, and hence
cast these processes for policy applications remains an subject to the dictates of short-term policy needs, albeit
important goal. Most of the newer generations of at the expense of long-term theory development. This
microsimulation models are designed with the objec- trend continues to the present and will likely do so in
tive of making them more policy sensitive. the future, as the needs for policy-oriented analysis
Unfortunately, few of them have yet reached a point (design) work continue to dominate planning practice.
where they can be fairly evaluated on this criterion, Batty noted though, that periodic reflection and critical
16 Journal of Planning Literature

review by those engaged in research can be seen as a Barrett, C. 1995. An operational description of TRANSIMS (No. LA-UR-
95-2393). Los Alamos, NM: Los Alamos National Laboratory.
sign of maturity. There has been much of this in the
Batty, Michael. 1979. Progress, success and failure in urban model-
field of transportation and land use modeling, as in the ing. Environment & Planning A 11(8): 863-878.
related field of travel demand analysis (see, for ———. 1997. Cellular automata and urban form: A primer. Journal of
example Pas 1990). Continued reflection and a com- the American Planning Association 63(3): 264-274.
mitment to developing models that reflect the relevant ———. 2005. Cities and complexity: Understanding cities with cellular
theoretical constructs of the behavior or system being automata, agent-based models and fractals. Cambridge, MA: MIT Press.
studied are seen then as the most promising paths Benenson, Itzhak. 1998. Multi-agent simulations of residential
dynamics in the city. Computers, Environment and Urban Systems
toward developing transportation and land use mod- 22(1): 25-42.
eling toward a more “mature” state and building more Berechman, Joseph, and Kenneth A. Small. 1988. Modeling land use
practically useful tools. and transportation: An interpretive review for growth areas.
Environment & Planning A 20(10): 1285-1309.
NOTES Bowman, John L., and Moshe E. Ben-Akiva. 2001. Activity-based dis-
aggregate travel demand model system with activity schedules.
1. Reviews that cover a larger number of models, including some Transportation Research, Part A: Policy and Practice 35A(1): 1-28.
that have seen less commercial application, are provided in recent Cervero, Robert. 1984. Light rail transit and urban development.
articles by Timmermans (2003) and Wegener (1994, 2004). Chang Journal of the American Planning Association 50(2): 133-147.
(2006) also provides a review of models based on mathematical pro- Chang, Justin. 2006. Models of the relationship between transport
gramming formulations, which are not discussed here. and land-use: A review. Transport Reviews 26(3): 325-50.
2. The literature on activity-based approaches to travel analysis is Chapin, F. Stuart. 1974. Human activity patterns in the city. New York:
quite extensive and dates to the 1970s. Thus, a comprehensive review Wiley.
of this literature is not possible here. Instead, the focus will be on cov- Clarke, K. C., and L. Gaydos. 1998. Loose-coupling a cellular
ering a few of the models that have been tested using real-world data automaton model and GIS: Long-term urban growth prediction
at least once. The interested reader is directed to articles by Axhausen for San Francisco and Washington/Baltimore. International
and Garling (1992), Ettema and Timmermans (1997), McNally (2000), Journal of Geographic Information Science 12:699-714.
Vovsha, Bradley, and Bowman (2005), and the collection of articles in Clarke, K. C., S. Hoppen, and L. Gaydos. 1997. A self-modifying cel-
the August 1996 issue of the journal Transportation, which describes lular automaton model of historical urbanization in the San
the early results of research work funded through TMIP. Francisco Bay area. Environment & Planning B: Planning and
Design 24(2): 247-61.
Clay, Michael J., and Robert A. Johnston. 2006. Multivariate uncer-
REFERENCES
tainty analysis of an integrated land use and transportation
Anas, Alex. 1982. Residential location markets and urban transportation: model: MEPLAN. Transportation Research, Part D: Transport and
Economic theory, econometrics and policy analysis with discrete choice Environment 11D(3): 191-203.
models. New York: Academic Press. Crecine, J. P. 1964. TOMM. Pittsburgh: Department of City and
———. 1983. Discrete choice theory, information theory and the Regional Planning.
multinomial logit and gravity models. Transportation Research, de la Barra, Tomas. 1989. Integrated transport and land use modeling:
Part B 17B(1): 13-23. Decision chains and hierarchies. Cambridge, UK: Cambridge
———. 1984. Discrete choice theory and the general equilibrium of University Press.
employment, housing and travel networks in a Lowry-type Dong, Xiaojing, Moshe E. Ben-Akiva, John L. Bowman, and Joan L.
model of the urban economy. Environment & Planning A 16(11): Walker. 2006. Moving from trip-based to activity-based measures
1489-1502. of accessibility. Transportation Research, Part A: Policy and Practice
———. 1987. Modeling in urban and regional economics. Chur, 40(2): 163-80.
Switzerland: Harwood Academic Publishers. Echenique, Marical H. 2004. Econometric models of land use and
———. 1998. NYMTC transportation models and data initiative. The transportation. In Handbook of transport geography and spatial sys-
NYMTC Land Use Model. Williamsville, NY: Alex Anas & Associates. tems, ed. David A. Hensher, Kenneth J. Button, Kingsley E.
Haynes, and Peter R. Stopher, 185-202. Amsterdam: Pergamon.
Anas, Alex, and Richard J. Arnott. 1994. The Chicago prototype
housing market model, with tenure choice and its policy impli- Echenique, Marcial H., D. Crowther, and W. Lindsay. 1969. A spatial
cations. Journal of Housing Research 5:73-129. model of urban stock and activity. Regional Studies 3:281-312.
Arentze, Theo A., and Harry J. P. Timmermans. 2003. Modeling Echenique, Marcial H., A. D. Flowerdew, J. D. Hunt, T. R. Mayo,
agglomeration forces in urban dynamics: A multi-agent system I. J. Skidmore, and D. C. Simmonds. 1990. The Meplan models of
approach. Paper presented at the 8th International Conference Bilbao, Leeds and Dortmund. Transport Reviews 10(4): 309-22.
on Computers in Urban Planning and Urban Management, May Ellickson, B. 1981. An alternative test of the hedonic theory of hous-
27-29, Sendai, Japan. ing markets. Journal of Urban Economics 9(1): 56-79.
———. 2004. A learning-based transportation-oriented simulation sys- Ettema, Dick, and Harry J. P. Timmermans. 1997. Activity-based
tem. Transportation Research, Part B: Methodological 38B(7): 613-33. approaches: An introduction. In Activity-based approaches to travel
Axhausen, Kay W. 1990. A simultaneous simulation of activity analysis, ed. Dick Ettema and Harry J.P. Timmermans, 1-36.
chains and traffic flow. In Developments in dynamic and activity- Oxford, UK: Pergamon.
based approaches to travel analysis, ed. Peter M. Jones, 206-25. Garin, Robert A. 1966. A matrix formulation of the Lowry model for
Brookfield, VT: Avebury. intra-metropolitan activity. Journal of the American Institute of
Axhausen, Kay W., and Tommy Garling. 1992. Activity-based Planners 32:361-64.
approaches to travel analysis: Conceptual frameworks, models Goldner, W. 1971. The Lowry model heritage. Journal of the American
and research problems. Transport Reviews 12(4): 323-341. Institute of Planners 37:100-10.
 Models of Transportation and Land Use Change 17

Hagerstrand, Torsten. 1970. What about people in regional science? Martinez, Francisco J. 1992. The bid-choice land use model: an inte-
Papers of the Regional Science Association 24:7-21. grated economic framework. Environment & Planning A 24: 871-85.
Harris, Britton. 1994. The real issues behind Lee’s “Requiem.” ———. 1996. MUSSA: A land-use model for Santiago city.
Journal of the American Planning Association 60(1): 31-34. Transportation Research Record 1552:126-34.
Hess, Stephane, Kenneth E. Train, and John W. Polak. 2006. On the McFadden, Daniel L. 1978. Modelling the choice of residential loca-
use of a Modified Latin Hypercube Sampling (MLHS) method in tion. In Spatial interaction theory and planning models, ed. Anders
the estimation of a Mixed Logit Model for vehicle choice. Karlqvist, 75-96. Amsterdam: North-Holland.
Transportation Research, Part B: Methodological 40B(2): 147-63. McNally, Michael G. 1998. Activity-based forecasting models inte-
Horowitz, Alan J. 2004. Lowry-type land use models. In Handbook of grating GIS. Geographical Systems 5(2): 163-184.
transport geography and spatial systems, ed. David A. Hensher, ———. 2000. The activity-based approach. In Handbook of transport
Kenneth J. Button, Kingsley E. Haynes, and Peter R. Stopher, 167- modeling, ed. David A. Hensher and Kenneth J. Button, 53-69.
83. Amsterdam: Pergamon. Amsterdam: Pergamon.
Hunt, John D., and John E. Abraham. 2005. Design and implemen- Miller, Eric J. 2003. Microsimulation. In Transportation systems
tation of PECAS: A generalised system for allocating economic planning: Methods and applications, ed. Konstadinos G. Goulias,
production, exchange and consumption quantities. In Integrated 12.1-12.22. Boca Raton, FL: CRC Press.
land-use and transportation models: Behavioural foundations, ed.
Martin Lee-Gosselin and Sean Doherty, 253-73. Amsterdam: Miller, Harvey J. 2005. Place-based versus people-based accessibility.
Elsevier. In Access to destinations ed. David M. Levinson and Kevin J.
Krizek, 63-89. Amsterdam: Elsevier.
Hunt, John D., Eric J. Miller, and David S. Kriger. 2005. Current oper-
ational urban land-use-transport modelling frameworks: A Moeckel, R., K. Spiekermann, C. Schurmann, and M. Wegener. 2003.
review. Transport Reviews 25(3): 329-76. Microsimulation of land use, transport and environment. Paper pre-
sented at the 8th International Conference on Computers in Urban
Jantz, Claire A., Scott J. Goetz, and Mary K. Shelley. 2004. Using the Planning and Urban Management, May 27-29, Sendai, Japan.
SLEUTH urban growth model to simulate the impacts of future
policy scenarios on urban land in the Baltimore/Washington Ottensmann, John R. 2005. Accessibility in the Luci2 Urban
metropolitan area. Environment & Planning B: Planning and Simulation model and the importance of accessibility for urban
Design 31(2): 251-71. development. In Access to destinations, ed. David M. Levinson
and Kevin J. Krizek, 297-324. Amsterdam: Elsevier.
Kelly, Eric D. 1994. The transportation land-use link. Journal of
Planning Literature 9(2): 128-45. Pas, Eric I. 1990. Is travel analysis and modelling in the doldrums?
In Developments in dynamic and activity-based approaches to travel
Kii, Masanobu, and Kenji Doi. 2005. Multiagent land-use and trans- analysis, ed. Peter M. Jones, 3-33. Brookfield, VT: Avebury.
port model for the policy evaluation of a compact city.
Environment & Planning B: Planning and Design 32(4): 485-504. Pendyala, Ram M., Ryuichi Kitamura, Cynthia Chen, and Eric I. Pas.
1997. An activity-based microsimulation analysis of transporta-
Krishnamurthy, Sriram, and Kara M. Kockelman. 2003. Propagation tion control measures. Transport Policy 4(3): 183-92.
of uncertainty in transportation land use models: Investigation
of DRAM/EMPAL and UTPP predictions in Austin, Texas. Pradhan, Anant, and Kara M. Kockelman. 2002. Uncertainty propa-
Transportation Research Record 1831:219-29. gation in an integrated land use-transportation modeling frame-
work: Output variation via UrbanSim. Transportation Research
Kwan, Mei-Po, and Joe Weber. 2003. Individual accessibility revis- Record 1805:128-35.
ited: Implications for geographical analysis in the twenty-first
century. Geographical Analysis 35:341-53. Putman, Stephen H. 1974. Preliminary results from an integrated trans-
portation and land use models package. Transportation 3(3): 193-224.
Lee, Douglass B. 1973. Requiem for large scale urban models. Journal
of the American Institute of Planners 39(2): 163-78. ———. 1983. Integrated urban models: Policy analysis of transportation
and land use. London: Pion.
Lerman, Steven R. 1976. Location, housing, automobile ownership
and mode to work: A joint choice model. Transportation Research ———. 2001. The METROPILUS planning support system: Urban
Record 610:6-11. models and GIS. In Planning support systems: Integrating geo-
graphic information systems, models and visualization tools, ed.
Levinson, David M., and Wei Chen. 2005. Paving new ground: A Richard K. Brail and Richard E. Klosterman, 99-128. Redlands,
Markov Chain model of the change in transportation networks CA: ESRI Press.
and land use. In Access to destinations, ed. David M. Levinson and
Recker, Wilfred W., Michael G. McNally, and Gregory S. Root. 1986a.
Kevin J. Krizek, 243-66. Amsterdam: Elsevier Science Publishers.
A model of complex travel behavior (I): Theoretical develop-
Levinson, David M., and Ramachandra Karamalaputi. 2003. ment. Transportation Research, Part A 20A(4): 307-18.
Induced supply: A model of highway network expansion at the
———. 1986b. A model of complex travel behavior (II): An opera-
microscopic level. Journal of Transport Economics and Policy 37(3):
tional model. Transportation Research, Part A 20A(4): 319-30.
297-318.
Rohr, C., and I. N. Williams. 1994. Modelling the regional economic
Levinson, David M., and Bhanu Yerra. 2006. Self-organization
impacts of the Channel Tunnel. Environment & Planning B:
of surface transportation networks. Transportation Science 40(2):
Planning and Design 21: 555-568.
179-88.
Roorda, Matthew J., Sean T. Doherty, and Eric J. Miller. 2005.
Lowry, Ira S. 1964. A model of metropolis. Santa Monica, CA: Rand
Operationalising household activity scheduling models:
Corporation.
Addressing assumptions and the use new sources of behavioural
Mackett, Roger L. 1983. The Leeds integrated land-use transport model data. In Integrated land-use and transportation models: Behavioural
(LILT). Crowthorne, UK: Transport and Road Research foundations, ed. Martin Lee-Gosselin and Sean T. Doherty, 61-85.
Laboratory. Amsterdam: Elsevier.
———. 1990. Exploratory analysis of long-term travel demand and Roorda, Matthew J., and Eric J. Miller. 2006. Assessing transporta-
policy impacts using micro-analytical simulation. In tion policy using an activity-based microsimulation model of
Developments in dynamic and activity-based approaches to travel travel demand. ITE Journal 76(11): 16-21.
analysis, ed. Peter M. Jones, 384-405. Brookfield, VT: Avebury. Ryan, Sherry. 1999. Property values and transportation facilities:
———. 1991. LILT and MEPLAN: A comparative analysis of land-use Finding the transportation-land use connection. Journal of
and transport policies for Leeds. Transport Reviews 11(2): 131-54. Planning Literature 13(4): 412-27.
 18 Journal of Planning Literature

Salvani, Paul, and Eric J. Miller. 2005. ILUTE: An operational proto- ———. 2002b. Design of an integrated land use and activity-based
type of a comprehensive microsimulation model of urban sys- travel model system for the Puget Sound region. Transportation
tems. Networks and Spatial Economics 5(2): 217-34. Research Record 1805:108-18.
Schelling, Thomas C. 1978. Micromotives and macrobehavior. New Waddell, Paul A., and Alan Borning. 2004. A case study in digital
York: Norton. government: Developing and applying UrbanSim, a system for
Simmonds, David. 1999. The design of the DELTA land-use model- simulating urban land use, transportation, and environmental
ling package. Environment & Planning B: Planning and Design impacts. Social Science Computer Review 22(1): 37-51.
26:665-84. Waddell, Paul A., A. Borning, M. Noth, N. Freier, M. Becke, and G.
Strauch, D., R. Moeckel, M. Wegener, J. Grafe, H. Muhlhans, G. Ulfarsson. 2003. Microsimulation of urban development and
Rindsfuser, and K.-J. Beckmann. 2003. Linking transport and location choices: Design and implementation of UrbanSim.
land use planning: The microscopic dynamic simulation model. Networks and Spatial Economics 3(1): 43-67.
Paper presented at the 7th International Conference on Wegener, Michael. 1982. Modeling urban decline: A multi-level eco-
Geocomputation, Sept. 8-10, Southampton, UK. nomic-demographic model of the Dortmund region. International
Timmermans, Harry J. P. 2003. The saga of integrated land use- Regional Science Review 7(1): 21-41.
transport modeling: How many dreams before we wake up? ———. 1994. Operational urban models: State of the art. Journal of
Paper presented at the 10th International Conference on Travel the American Planning Association 60(1): 17-30.
Behaviour Research, Aug. 10-15, Lucerne, Switzerland. ———. 2004. Overview of land use transport models. In Handbook of
Torrens, Paul M., and David O’Sullivan. 2001. Cellular automata transport geography and spatial systems, ed. David A. Hensher,
and urban simulation: Where do we go from here? Environment Kenneth J. Button, Kingsley E. Haynes, and Peter R. Stopher, 127-
& Planning B: Planning and Design 28(2): 163-68. 46. Amsterdam: Pergamon.
Veldhuisen, K. Jan, Harry J. P. Timmermans, and Loek L. Kapoen. Williams, Huw C. W. L. 1977. On the formation of travel demand
2000. Ramblas: A regional planning model based on the micro- models and economic evaluation measures of user benefit.
simulation of daily travel patterns. Environment & Planning A Environment & Planning A 9(3): 285-344.
32(3): 427-443. Wilson, Alan G. 1967. A statistical theory of spatial distribution
———. 2005. Simulating the effects of urban development on models. Transportation Research 1: 253-269.
activity-travel patterns: An application of Ramblas to the ———. 1970. Entropy in urban and regional modelling. London: Pion.
Randstad North Wing. Environment & Planning B: Planning and
Design 32(4): 567-80. ———. 1971. A family of spatial interaction models, and associated
developments. Environment & Planning 3(1): 1-32.
Vovsha, Peter, Mark A. Bradley, and John L. Bowman. 2005. Activity-
based travel forecasting models in the United States: Progress Yamins, Daniel, Steen Rasmussen, and David Fogel. 2003. Growing
since 1995 and prospects for the future. In Progress in activity- urban roads. Networks and Spatial Economics 3(1): 69-85.
based analysis, ed, Harry J. P. Timmermans, 389-414. Amsterdam: Yeh, Anthony G., and X. Li. 2002. A cellular automata model to sim-
Elsevier. ulate development density for urban planning. Environment &
Waddell, Paul A. 2000. A behavioral simulation model for metropol- Planning B: Planning and Design 29:431-50.
itan policy analysis and planning: Residential location and hous- Yerra, Bhanu, and David M. Levinson. 2005. The emergence of hier-
ing market components of UrbanSim. Environment & Planning B: archy in transportation networks. Annals of Regional Science 39(3):
Planning and Design 27(2): 247-63. 541-53.
———. 2002a. UrbanSim, modeling urban development for land Zhang, Lei, and David M. Levinson. 2007. The economics of trans-
use, transportation and environmental planning. Journal of the portation network growth. In Essays on transport economics, ed.
American Planning Association 68:297-314. P. Coto and V. Inglada, 317-40. Heidelberg: Physica-Verlag.

View publication stats