Title: Integrated Urban Models: Design Issues for Integrated Urban Models
1Integrated Urban ModelsDesign Issues for
Integrated Urban Models
2Definitions (Miller et al)
- Models are defined as mathematical algorithms
which simulate human activities, such as travel
behavior. Once they are calibrated against
existing (known) conditions, the models can be
used to forecast these activities. For present
purposes, however, it is appropriate also to
include theory and data, in addition to
algorithms, in the definition of a model.
3Definitions (Miller et al)
- Travel demand forecasting models are used to
predict travel on a transportation network.
Travel demand -- i.e., traveler behavior -- is a
function of both human activities (generally
represented in the models by land-use see below)
and network characteristics (including the
services provided and their costs where costs
generally are expressed in terms of time and/or
money expended by the user and/or the
transporter, and possibly intangibles such as
comfort, etc.). Therefore, travel forecasts will
vary according to changes in human activities or
network characteristics.
4Definitions (Miller et al)
- Land-use models generally are mathematical tools
that forecast demographic and economic measures
of land-based activities. These measures can be
described as meaningful attributes of the
population (e.g., age-group cohorts, income,
jobs, etc.), but also can be expressed in terms
of development (e.g., residential dwelling units,
commercial floor space, etc.). They take into
account trends and policies regarding demographic
and economic development, as well as the supply
of developable land. The shape of the
transportation network, and its possible
influence on the location of human activities are
taken into account, but usually in a generalized
manner. Among other uses, the results serve as
input to a travel demand forecasting model.
However, generally there is little or no feedback
from travel demand forecasting models into
land-use models.
5Definitions (Miller et al)
- Integrated transportation - land-use models
simulate the interaction of land-use and
transportation. They are intended to provide a
feedback mechanism between land-use models and
travel demand forecasting models. For example, a
new or improved transportation link can influence
the timing, type and extent of development in a
particular area, which in turn can influence the
need for further network improvements. Similarly,
the evolution of land-based human activities
influences the demand for improved transportation
services, which in turn further influences
development. Therefore, integrated models take
land-use models a step further, by also
simulating the location of human activities as a
function of transportation accessibility.
Similarly, integrated models do not simply
provide land-use input to travel demand
forecasting models.
6Definitions (Miller et al)
- Transit accessibility describes how well a given
point in an urban area is connected by transit to
other points within the area. "Accessibility" can
be defined in many ways - the absolute existence of a physical means of
traveling between two points, where no other link
exists (in this case, by transit) - the relative level of service (speed, cost,
comfort, etc.) of a particular transit link,
where alternate transit links exist and which may
offer different levels of service and - the relative level of service of a particular
transit link, compared with alternate modes
(e.g., auto) which provide the same link.
7Model Development Life Cycle
- Requirements Analysis
- Conceptual Design
- Model Specification
- Data Development
- Model Estimation
- Application Software
- Model System Validation
- Model Application
- Model and Software Documentation
8Model Development Life Cycle
- Requirements Analysis
- What questions should the model address?
- What level of detail is needed?
- Who will use it, and for what applications?
- What data are available to use in developing it?
- What resource and time constraints apply?
- How should trade-offs be made on design choices
- Accuracy
- Policy-sensitivity
- Complexity
- Resource requirements
9Model Development Life Cycle
- Conceptual Design
- What modeling framework best addresses the
requirements, under the time and resource
constraints? - Given a particular modeling framework, what are
the major model components that need to be
designed? - How should model components interact within a
model system? - Given a high-level conceptual design, are there
requirements that cannot be met? Should the
design or the requirements be modified to be
consistent? - How should data and resource requirements be
modified, given this conceptual design?
10Model Development Life Cycle
- Model Specification
- For each model component, how should the
equations be specified? - What is the dependent variable? (continuous,
discrete?) - What are the independent variables? What does
theory suggest should be included? - Are any effects likely to be nonlinear? What
transformations would be appropriate? (log,
quadratic) - Are there likely to be interactions between
variables? (if so, add interaction variables)
11Model Development Life Cycle
- Data Development
- What data are needed to estimate the model
parameters? - What processing needs to be done?
- What quality control checks should be done?
- How should missing data be handled?
- How should outliers be handled?
12Model Development Life Cycle
- Model Estimation
- For each specified model equation, need to
estimate the parameters associated with each
independent variable. - Depending on the nature of the model
specification, will need specialized econometric
software - Multiple regression
- Multinomial logit
- Nested Logit
13Model Development Life Cycle
- Application Software
- What kind of software is needed to implement the
model for use in forecasting and policy analysis,
using the estimated model? - What language should it be implemented in?
- How can it be modularized to make it easy to
maintain and evolve? - What kind of user-interface does it require?
- What other performance or functional requirements
are there?
14Model Development Life Cycle
- Model System Validation
- Once the estimated model is implemented in
application software, it is operational. - How should we test how well the model is working?
- Are the model predictions generally reasonable?
- Are the results sensitive to changes in model
inputs and assumptions in ways that seem
plausible? - Are there observed data that we can use to
compare the model results to, in order to
validate the model? - Are adjustments needed to equation constants to
improve the overall fit of the predictions to the
observed data?
15Model Development Life Cycle
- Model Application
- What initial real-world projects could be used
to apply the model to? - Helpful not to start with extremely large, or
politically sensitive topics - Results may still reveal problems in the model
that need further adjustment or refinement - Once model has been tested successfully on a few
non-critical projects, can be applied to more
important ones
16Model Development Life Cycle
- Model and Software Documentation
- Last, but not least, the model needs to be
documented - Good documentation of operational models often is
provided in three levels - Executive Summary
- Reference Guide
- User Guide