Jean-Paul Rodrigue (2017), New York:
Routledge, 440 pages.
Transportation / Land Use Modeling
Author: Dr. Jean-Paul Rodrigue
"Essentially, all models are wrong but some are useful." George
To gain a better understanding of the behavior of urban areas, several
operational transportation / land use models (TLUM) have been developed.
The reasons behind using TLUM are numerous, such as the ability to forecast
future urban patterns based on a set of economic assumptions or to evaluate
the potential impacts of legislations pertaining to environmental standards.
Other uses of TLUM relate to testing theories, policies and practices
about urban systems. With a simulation model, urban theories can be
evaluated and the impacts of policy measures, such as growth management
and congestion pricing can be measured. It is not surprising that since
TLUM are planning tools per se, their development and application has
mainly been done by various government agencies related to transportation,
regional planning and the environment.
Broadly taken, a model is an information construct used to represent
and process relationships between a set of concepts, ideas, and beliefs.
Models have a language, commonly mathematics (expressed as functions
in various computer programming languages), an intended use and a correspondence
to reality. There are four levels of complexity related to the modeling
transportation / land use relationships:
On average, models tend to be relevant for constrained and well
structured problems with a specified number of variables, well-defined
goals, and firmly established technical solutions. This in itself limits
significantly the applicability of TLUM as urban systems are complex
entities. Still, these models have pros and cons:
- Static modeling. Express the state of a system at a given
point in time through the classification and arithmetic manipulation
of representative variables. Measuring accessibility can be considered
as static modeling.
- System modeling. Express the behavior of a system with
a given set of relationships between variables. The gravity model
is an example of system modeling as it tries to evaluate the generation
and attraction of movements.
- Modeling interactions between systems. Tries to integrate
several models to form a meta-system (a large and complex system).
A transportation / land use model
offers such a perspective.
- Modeling in a decision-taking environment. This not only
implies the application of a transportation / land use model, but
the analysis and reporting of its results in order to find strategies
and recommendations. Geographic Information Systems are useful tools
for that purpose as they can include the modeling, its graphic display
as well as being the platform over which decision making can take
2. Four Stages Transportation / Land Use Modeling
The core foundation of TLUM involves two obvious components; land
use and transportation. The land use
component, which is based on the location of housing, industrial and
commercial activities, tends to be more stable than the transportation
component which is highly dynamic. Most of TLUM have been applied regionally,
mainly at the urban level, as a larger scale would be prohibitively
complex to model. The modeling of the transportation components is particularly
relevant and is divided in four sequential
stages for the estimation of travel demand, from where movements
originate, how they are allocated, which modes are used and finally what
segments of the transport network are being used:
- Advantages. Incites a conceptualization of urban economic
and spatial processes. Advances in urban and regional sciences are
often linked with advances in modeling, notably conceptual representations
such as land economics. Moving from an urban concept to an urban
model is simply a step forward, albeit an important one. The data
requirements of TLUM are often an incentive to perform surveys from
which useful information about urban mobility and spatial structure
can be gathered. This information has the additional advantage of
triggering studies that are not necessarily related to TLUM, but
that contribute to advances in the understanding of the dynamics
of urban systems.
- Drawbacks. Models may gear towards a mechanistic approach
of urban dynamics where processes are compartmentalized. This leads
to difficulties about "thinking outside the box" and by its nature
modeling often fails to grasp significant economic, technological
and social changes. They may also give the impression that a system
can effectively be controlled since all its major elements have
been summarized. Solving a problem is thus a matter of tweaking
parameters. TLUM are consequently privileged tools of government
agencies as they fit such an unimaginative mindset.
Applying TLUM requires an extensive range
of data, most of it related to spatial units, land use, spatial
interactions and the transportation network. The most important information
for TLUM is however origin-destination data. A variety of survey
methods are used to collect this data including roadside questionnaires,
telephone interviews, and detailed activity modeling. Data availability
and limitation is an important factor behind the applicability of such
models and there is a constant trade-off between the costs of fulfilling
the data requirements and the benefits supplementary data may offer.
Additionally, data needs to constantly be updated as demographic, economic
and technological changes are taking place. This is one of the major
reasons why the transportation / land use modeling process, although
theoretically and conceptually sound, has not been applied comprehensively.
Among the major types of variables, it is possible to identify:
- The first stage is called trip generation and deals with
trip rate estimates, usually at the zonal level. The most common
methods for trip generation are cross-classification (also referred
to as category analysis) and multiple regression analysis. Cross-classification
seeks to identify specific socioeconomic groups within the population
that have common trip generation characteristics. The trip generation
of a zone will thus be the outcome of its composition. Regression
analysis estimates the number of trips generated by a zone (dependent
variable) as a function of a series of independent variables.
- The second stage is referred to as trip distribution
and deals with spatial movement patterns; the links between trip
origins and destinations. The most common technique for estimating
trip distribution is the gravity model. There are various forms
of the gravity model and various calibration techniques as well.
Cross-classification and multiple regression can also be used to
estimate the number a trips a zone would attract.
- The third stage is modal split; the proportion of trips
made by automobile drivers and passengers, transit, cyclists, and
walking. Logit modeling is commonly used as it evaluates the preference
of each user in terms of probability of using a specific mode for
a specific origin / destination pair.
- Finally, once the spatial patterns of movements by various
modes are estimated, trips are assigned to the various transport
links. This is done mostly by using operations research methods
aiming at minimizing travel costs or time over a transport network.
There are a wide variety of TLUM, most of them developed during the
quantitative revolution that transformed geography in the 1960s and
1970s. Among the best known are:
- Land use data. Include socio-economic variables pertaining
to the area under investigation, such as population, employment,
income level, commercial activity, etc. Such data is used to estimate
or calibrate the amount of travel generated and attracted
by each zone.
- Travel generation factors. Considering the available
land use data, these factors estimate the number of trips, people
and/or freight, each level of economic activity generates. They
consider a multitude of issues such as income, modal preferences
and consumption levels. Most of this information can be gathered
using surveys or inferring from observations made elsewhere.
- Friction of distance factors. They represent the difficulty
of traveling between different locations of the area under investigation,
commonly measured in terms of time, distance of cost. There is a
significant variance according to mode and purpose of travel. Friction
of distance factors enable to assess trip distribution and
- Calibration factors. It is uncommon that the results
produced by an uncalibrated model are corresponding to the reality.
Calibration factors thus try to match the results produced by the
model with data based on observations, surveys or common sense.
Calibration can often be an obscure process, because it tries to
incorporate factors that are not explained by the model itself.
- Transportation networks. A representation of the structure
and geometry of transportation within the area under investigation,
mainly composed of nodes and links. Transportation networks are
commonly divided by modes. For road transportation, a node could
represent an intersection, a stop or a parking lot, while a segment
could be linked with attributes such as permitted speed, distance
and capacity. For public transit, a node could represent a bus stop
or a metro station, while a segment could have attributes such as
capacity and frequency of service. Transportation networks, along
with origin-destination matrices, are fundamental elements of the
traffic assignment procedure.
The core of most transportation / land use model is some kind of
regional economic forecast that predicts and assigns the location
of the basic employment sector. As such, they are dependent on the reliability
and accuracy of macro-economic and micro-economic forecasting. Traditionally,
such forecasting tends not to be very accurate as it fails to assess
the impacts of economic, social and technological changes. For instance,
globalization and the emergence of global commodity chains have significantly
altered the dynamics of regional economies.
Additionally, few TLUM are dealing with freight transportation. This
can be explained by the fact that passengers transportation in urban
areas tends to be highly regulated by governmental agencies (e.g. public
transit) while freight transportation is dominantly controlled by private
interests. Paradoxically, while freight related activities such as terminals
and distribution centers tend to occupy a large amount of space, they
do not generate a large amount of passenger traffic.
- Lowry Model. Considered to be the
first transportation / land use model (1964), it links two spatial
interaction components. The first calculates spatial interactions
between basic employment activities and zones of residence, while
the second calculates spatial interactions between service employment
activities and zones of residence. The Lowry model is discussed
in more details here.
- ITLUP. The Integrated Transportation
and Land Use Package is composed of a residential allocation model,
an employment allocation model, and a travel demand model.
- MEPLAN. This model is a derivative
of the Lowry model, since it is based on the economic base theory.
It considers the two components of the transportation / land use
system as markets, one market for land use and one market for transportation.
- TELUS. Transport Economic Land Use System. Designed to
evaluate the impacts of transport improvements over the regional
economy and land use.
- Cellular automata. A new range of models where space
is represented as a grid (raster) with a set rules enforced to govern
the state of a cell depending on the configuration of its adjacent