Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 117
TCRP H-12 Final Report
.
5.0 REVIEW OF CURRENT MODELS
5.l Introduction
This chapter reviews a selection of operational integrated urban models. It describes the structure
and capabilities of these models, and compares them to those described in Chapter 4 for the ideal
model.
The reader should be cautious in interpreting the review- strictly as an evaluation of today's
operational models. A more appropriate interpretation, and the one intended by the investigators,
is that the review constitutes the taking of an inventory of what is available currently. The objective
is to arrive at a description of the current general state of practice by examining in detail several
representative models rather than to identify a "best" or "recommended" model.
Six models are reviewed. Section 5.2 describes how- these six were selected from a larger list
of candidates. Section 5.3 outlines the framework that is used for the review and its derivation.
Section 5.4 provides a general description of the six models. Section :.5 reviews the six, and
Section 5.6 comments upon and summarizes the review.
5.2 Selection of Models for Review
It is fair to say that a significant number of 'integrated' or semi-integrated urban models exists
, ~ ~ ~ ~ ~ ~ ~ Wegener ~1 995]
identified twenty active urban modeling centers (i.e., models) around the world, of which
approximately twelve integrated urban models were sufficient operational to have been used for
around the world. certainly In varying degrees of completeness arid usability.
actual research anchor policy analysis in particular urban areas. South`~-orth ti995] identifies a
fimher thee models.
The TCRP Research Project Statement for this work specified that three integrated urban models
were to be considered: ITEUP (also often referred to as DRAM/EMPAL), MEPLAN arid TRANUS.
The three models have the following features in common:
· they are operational. commercially-ax affable packages;
· each has an established history of use; and
· each has been applied in the United States In at least one practical setting (i.e.. in an MPO).
To this group. wee elected to add another three models. These models are MUSSA, NYMTC-
LUM and UrbanSim. Thev were included for two main reasons:
9:
OCR for page 118
TCRP H-}2 Final Report
· each is currently operational, or sufficiently close to being operational, in a practical setting;
and
· each contains a significant market representation (i.e., there is an explicit treatment of prices
in land development).
It is noteworthy that UrbanS~m is a recent developments d therefore was not included in Wegener's
1995 list.
As indicated by Me Wegener and Sou~wor~ reviews, many other integrated urban models exist.
Noteworthy examples include examples of microsimulation models tMackett. 199Ob; Wegener
1993, Miller, et at., 1987; Miller and Salvini, 1998; Oskamp, 1997], optimization models [Brotchie,
et al., i980 Caindec and Prastacos, 1995; Dickey and Leiner, 1983; Kim 19893; land accounting
type models Landis, 1994, Yen and Fncker, 1997~- and other European, Japanese and Aus~aiian
models ~Anderstig and Mattson, ~ 991; Eliasson arid Mattsson. ~ 997; Gu. et al., ~ 992; Mackett, ~ 983,
1985a, 1990a;Nakamura,et al., 1983,Wegener, 1982a, 1982b, 1986;Wegener~etal.. 19913. While
each of these models is of potential interest in one or more ways, it was felt that the six models
chosen for detailed review are sufficient for the purposes of this study in that they are representative
of the state of operational practice.
5.3 Framework
There are several reviews of integrated urban models in the literature. This review is
distinguished from these earlier efforts in one key respect: the review criteria are derived from the
descnption of the "ideal" mode] described ire Chapter 4. In other words, the review has a somewhat
normative/prescnptive orientation ("what should bed. as opposed to a descriptive basis ("what isle.
Nonetheless, some recent reviews are instructive. Wegener t! 995] compared or perhaps more
appropnately, classified thirteen of the twenty models that he cited. He noted that inclusion in the
list Eras based upon the available information, as opposed to being a comment on the respective
quality. The list included DRAM/EMPAL (ITEUP), MEPLAN, TRANUS, MUSSA and
METRO SIM (a forerunner of NYMTC-LUM). The classification was made according to the
following criteria:
Comprehensiveness, in which a mode] was deemed to be comprehensive if it could mode!
at least two of eight identified subsystems, all thirteen met this criterion.
.
Overall structure. Two types were identified: "unified" -- i.e., a tightly integrated structure --
or "composite," a series of independently-structured, loosely organized submodels. The
structure alas seen as art important determinant of the mode! techniques used by each model,
and the dynamic behavior of each. MUSSA arid METRO SIM were classified as unified,
while ITEUP, MEPLAN arid TRANUS were classified as composite.
- 94
OCR for page 119
TCRP H-12 Final Report
Theoretical basis. Twelve of the models had a random utility basis for simulating the
behavior of the actors; beyond this one commonality, the models had a variety theoretical
bases for simulating the land/development markets, some simulating market equ~libnum, and
for simulating have} activity/transportation network behavior.
· Techniques included land-use equilibnum, network equilibrium or both; spatial-interaction
location models or the use ot accessibility indicators to simulate location choice, input-output
methods to represent the flows of goods, etc.
Dynamics, including the treatment of time lags, and the short- and Tong-term processes ot
Barbara development.
Data requirements.
Calibration versus validation (i.e., application).
· Operationality (i.e.. transferability in addition to actually having been applied).
· Applicability (i.e.- the range of problems to which these models could be applied).
In this review. we examine the six models from the perspective of the design issues listed in
Table 4. I. In particular, we discuss how each model deals with: physical system representation,
representation of decision makers, representation of processes (especially market processes). as well
as the generic and implementation issues listed in Table 4.~. The model evaluation cutena identified
in Table 4.2 are also used. The logic ur~derlying the selection of both the design issues arid the
evaluation criteria has been discussed in detail in Chapter 4. Throughout the review, a primary
concern is dete~mirung how the current state of practice compares with the "ideal" model described
in Chapter 4.
5.4 General Description of the Six Models
5.4.1 ITLUP
ITLUP (Integrated Transportation and Land Use Package) was developed by Dr. Stephen H.
Putman at the University of Pennsylvania, over the course of the last 25 years. ITLUP comprises
a number of sub-models, the best known of which are DRAM (Disaggregate Residential Allocation
Model) arid EMPAL (Employment Allocation Model). DRAM and EMPAL are Lowry-derivative
models. The model allocates household categories (usually four income categories, though further
categorizations are possible, e.g., by household structure), employment categories (usually by four
types, though more detailed SIC groupings are possible) and travel patterns (public and private
modes), using exogenous study-area forecasts of employment, population and Lips activity rates and
95
OCR for page 120
TCRP H-12 Final Report
,
household type. Detailed documentation of the mode] is provided in Putman tI983. 1991 1994,
199S, among other sources]. Useful summary- descriptions are provided in Wegener tI995],
Sou~wo~ tipsy], Webster et at. tI 988] an] Putrna~ tI996], from which this discussion is drawn.
Notable features of ITEUP are:
IThUP (more specifically, DRAM and EMPAL) are the most ~dely-used spatial allocation
models in the United States today. A recent count indicates over a dozen active US
applications tPutrnan. 19974. although over 40 calibrations have been performed across the
United States arid elsewhere.
IThUP contains a multinomial logit modal split submodel, as well as a trip assignment
submode] that Carl support various (auto) assignment aigonthms. Trip generation and
distr~buhon are developed within DRAM, simultaneously with household location.
However, DRAM and EMPAL often are used separately, and have been linked in actual
applications with other commercial Gavel demand forecasting models (including EMME/2-
~ITS T ~- ~T ~T T~_T ~P ~. ' ~ · ~. ~
~ An' ~
1-KAN PLAN and U ~ And. ~ Bus, conserve teeming or travel demand and Ravel costs can
be provided through exogenous links.
· The model has been applied to a wide range of situations. ranging from lon~-term
A- , ~ ~ ,,.
transportation network planning, to the evaluation of different urban forms and. notably,
environmental impacts and region-wide vehicle emissions. How-ever, although in theory
there appear to be no constraints to including transit travel 'costs' in the accessibility
definitions, most applications appear to have emphasized road network improvements.
· Compared with other models, DRAM/EMPAL is considered to have relatively parsimonious
data requirements. Southworth ~1995] notes that an important advantage of DRAM/EMPAL
is their basis in generally available data (i.e.. related to population, households and
employment). However, he also notes that this reflects a weakness of the approach; namely,
that the model does not account for land market clearing processes (or, it follows. other
market clearing processes).
A recent development is METROPlLUS. which is aimed at improving linkages with GIS
data bases and In revising He mode] structure towards greater system modularity. The mode]
operates within an ArcView shell, which supports a linkages with an ArcView GIS data base.
The shell also is W~ndows-compatible, which in turn provides linkages to off-the-shelf data
packages such as Excel [Putman, 1997].
5.4.2 MEPLAN
.
The MEPLAN package is proprietary software developed by Marcial Echenique and Piers Ltd
In the United Kingdom. The MEPLAN package draws on 25 years of practical integrated urban
- 96
OCR for page 121
TCRP H-17 Final Report
modeling experience. win work on He package itself beginning in 1 985. It has been applied to over
95 regions throughout the world, including Sacramento. Califorrna. Detailed documentation Carl be
found in Echenique tI985] Echenique, et al. ti969, 19904. Echenique and Williams reshow. Hunt
tI994], Hunt and Simmonds tI993] and Hurlt and Echenique ~19933.
, .,
MEPLAN is based on a generals highly flexible framework. This framework has an aggregate
perspective: space is divided into zones quantities of households and economic activities (called
'factors' or 'sectors') are allocated to these zones, and flows of interactions among these factors in
different zones give rise to flows of transport demand. The heart of the framework is a spatially
disaggre~ated input-output matnx. or social accounting matnx, extended to include vanable
technical coefficients. labor sectors arid space sectors. All economic activities, including
households are treated as producing and consuming activities. with consumption patterns expressed
using technical coefficients. Spatial disaggregation is accomplished by having the further production
arising to satisfy consumption allocated among the spatial zones according to discrete choice models
reacting to the prices for such production. The resulting interactions among zones gives rise to the
demand for travel.
Temporal change is simulated by having the model consider sequential points in time. Space
(both land and floorspace) is "non-transportable" and must be consumed in the zone where it is
produced. The supply of space in each zone is fixed at a given point in time. The technical
coefficients for the consumption of space are elastic with respect to price and prices for space that
ensure demand equates with supply in each zone are established endogenously as part of an
equilibrium solution established for each point in time considered. Prices for the outputs of other
sectors are established endogenously running back along the 'chains' of production-consumption.
Travel demands arising for a given point in time are allocated to a multimodal network using logit
functions representing mode and route choice, taking account of congestion. Transport disutilities
feed back into He next time period representing lags in response to transport conditions. Exogenous
demand (analogous to the 'Lowry' basic sector) provides the initial impetus for economic activity.
Changes ~ study-v~de exogenous demand and in the quantity of space in each zone from one time
period to the next fuel economic change. with these charges allocated among zones.
5.~.3 MUSSA
MUSSA ("Modelo de Uso de Suelo de SAntiago") is an operational model of urban land and
floor space markets. developed by Dr. F~cisco Martinez for Santiago, Chile. It is ''filly connected"
with a very good four-stage model (known as ESTRAUS); together, the combined models are
referred to as 5-L,UT~ and provide equilibrated forecasts of larld-use and travel demand for Santiago.
The model has been used to examine venous transportation and/or land-use policies. usually
involving transit as a central component of the policy. Documentation of the model is provided in
Martinez [1992a' 1992b' 1997a 1997b] and Martinez and Donoso [1995]. Notable features of
MUSSA include:
- 97
OCR for page 122
TCRP H-19 Final Report
It is consistently based throughout on an extremely rigorous and compelling application of
microeconomic theory.
It is an equilibnum mode} of building stock supply aIld demand. Demand for building stock
(whether by households or firms) is based on Weir w~naness to pay (WP). Buyers attempt
to max~nize their surplus (WP - price actually paint, while sellers attempt to maximize price
paid. Building stock is supplied by developers so as to maximize profits, given the apparent
demand. Building stock prices are endogenously determined within the equilibration
process.
· The model solves for a static equilibnum in the forecast year, by adjusting the amount of
building stock supplied. a supply response, and consumers. expectation levels (expected
utility to be obtained from their housing), a demand response, until demand and supply
balance. The model end state is path-independent and does not require solution for
intermediate year results. although such intermediate results can also be generated.
.
.
.
The model uses traffic analysis zones as its spatial unit of analysis (264 zones in the Santiago
application), thereby pros icing a very fine level of spatial disaggregation relative to many
other current models. In addition, extensions to more micro-levels of spatial analysis are
being investigated (Martinez, 1997b).
The model is highly disaaaregated relative to most other currently operational models. The
Santiago implementation has 65 household types and could be run using ~ large weighted
sample of observed households (and their associated detailed attributes) in essentially a
"static micros~mulation" format.
Extensions are being Investigated to incorporate zone-level environmental impact
(emissions) calculations into the modeling, system tO'Ryan, et at., ~ 9973.
· As wall the NYMTC-~UMfMETROPOLIS family of models, MUSSA provides operational
evidence of the usefulness of integrated urbar~ models in the analysis of both the impacts of
major transit projects on land-use and the impacts of land development on transit.
S.4.4 NYMTC-LUM
NYMTC-LUM is currently under development by Dr. Alex Anas on behalf of New York
Metropolitan Transit Co~runission (MTC). It is a simplified version of METROPOLIS. also
developed by Dr. Anas. It is the most recent of a series of land-use and housing market models
developed by Dr. Anas over the last two decades tAnas, ~ 982, 1992, ~ 994, ~ 995; Anas and Arnott,
1993, 1994, Anas and Brown, 1985, Anas et al.. 1987, Anas and Duann, 1986~. Notable features
of NYMTC-LUM include:
- 98
OCR for page 123
TCRP H-12 Final Report
· It is consistently based throughout on macroeconomic theory.
· It simultaneously models the interactions between residential housing commercial floor
space, labor, and non-work travel markets, with explicit representations of Remarry and
supply processes in each case.
Housing pnces. floor space rents and workers' wages are all endogenously determined within
the mode} and are used to mediate between demand and supply processes within their
relevant markets.
The mode! solves for a static equilibrium In Me forecast year, by finding the prices and wages
which cause Remarry and supply in the markets being modeled to balance. The mode} end
state is path-~ndependent arid does not require solution for intermediate year results.
The mode! uses traffic analysis zones as its spatial urut of analysis (up to 3,500 zones in the
blew York application), thereby providing a very fine level of spatial disaggregation relative
to many other current models.
In its current state of Implementation, the mode} does not contain much disa~gregation of its
main "behavioral units" (households, employment. buildings).
In its current implementation, the mode} is not integrated with a travel demand model.
Rather, it is "connected" to We existing MTC travel demand mode! in terms of receiving as
inputs mode} utilities from the MTC mode choice model. This is similar to the case for
IThUP and UrbanSim.
From the perspective of this work. perhaps He single most important point to note about this
mode! is that it is being developed for a transit property. explicitly for transit-related planning.
Features of the mode! which facilitate this type of application include: use of small traffic zones as
the spatial unit of analysis; access to detailed transit network representations and mode choice
models in the MTC travel demand model; and the macroeconomic structure of the model which
permits a range of economic evaluation measures to be computed (property values. consumers.
surplus, producers surplus, etc.~. Earlier models have similarly been applied to the es aluation of
the impacts of He proposed South Corridor rapid transit line In Chicago fusing the C ATLAS model,
see Anas and Dunn tI9863) and the assessing the impacts of a range of road and transit service
changes in the New- York (using NYSIM, see Anas [199541.
5.4.5 TRANUS
.
The TRANUS package is proprietary software developed by Modelistica in Venezuela, a private
film rural by Tomas de la Balsa arid his family. It draws on much the same modeling experience as
99
OCR for page 124
TCRP H-12 Final Report
MEPLAN win We elements of the package fast coming together in the early 1 980s. The focus with
TR~NUS would appear to be a somewhat more restricted set of functional forms and modeling
options within the framework allowing a more "set" approach to mode] development reiat~ve to
MEPLAN. It has been applied to a number of regions In Central and South America and In Europe.
TRANUS models of the Sacramento and Baltimore Regions and the State of Oregon have been
completed or are ureter development. For demled documentation. see de la Barra tI989, 1989], de
ia Barra et at. ~! 984] and Modelistica tI9983.
.
5.4.6 UrbanSim
UrbanS~m is an operational model of dreary bared and floor space markets. developed by Dr. Paul
Waddell for the States of Hawaii, Oregon, and Utah. A prototype has been completed in Eugene-
Springfield~ Oregon and a histoncal validation process is beginrung there. It is "fills` connected"
with a traditional four-stage mode] in Eugene-Spr~ngfield, Oregon. and is being integrated with a
new activity--based travel model in Honolulu, Hawaii. Although the Meal development of the model
was undertaken by Dr. Waddell through the consulting firm of Urban Analytics, further development
and support of the mode] is being done at the University of Washington. The model and software
has been placed in the public domain by the Oregon Department of Transportation and the
University of Washington will support its release and dissemination through the Internet as part of
an NCHRP project 8-32~3) "Integration of Land Use Planning and Multimodal Transportation
Planning." Documentation of the mode] can be fourth in Waddell ~1998a, 199Sb 1998c] and
Waddeli et al ti9983. Access to documentation and to the model will be available from
ww~-.urbansim.org on He Intemet as of August, ~ 998.
Notable features of UrbanSim include:
It is consistently based throughout on an extremely rigorous and compelling application of
microeconomic theory, following a similar framework to that used in MUSSA, but diffenug
in significant aspects such as the assumption of equilibnum.
It is a disequilibrium model of building stock supply and demand with annual time
increments. Demand for building stock (whether by households or firms) is based on their
willingness to pay (WP), or bid (observed prices paid rather than hypothetical WP that
cannot be observed). Buyers attempt to maximize their surplus (WP - price actually paid)
while sellers attempt to maximize price paid. Building stock is supplied by developers so
as to maximize profits, given the apparent demand. Building stock prices are determined
~ ithin the market clearing process, which occurs at the submarket level of the traffic analysis
zone and property- type.
· The mode] operates as a dynamic disequilibrium in each year, with the supply component
developing and redeveloping individual land parcels on the basis of expected profits
- 100
,
OCR for page 125
TCRP H-17 Final Report
(expected revenue less costs). Expected revenue is based on prices lagged b! one ~-ear' and
new coin choices are not assumed to be available for occupancy until the subsequent
year. Demand is based on lagged prices and current supply d prices are adjusted based
on the balance of demand and supply in each submarket in each year.
is path-dependent and requires solution for each intermediate year.
The model end state
· The demand side of Me mode] uses traffic analysis zones as its spatial unit of analysis (271
zones In the Eugene-Spnngfield application 76 ~ in Honolulu. and over 1.000 in Salt Lake
City), thereby providing a very fine level of spatial disaggre~ation relative to many other
current models. In We supply-side, the model uses the individual land parcel as the unit of
land development and redevelopment, making this the only mode} to date to use the parcel-
leve] as the fundamental unit of analysis.
· The mode] is highly disag~are~ated relative to most other currently operational models. The
Eugene-Springfield implementation has ~ ~ ~ household types and could be run using a large
weighted sample of observed households (and their associated detailed attributes) in
essentially a "static micros~mulation" format.
· The mode! is based on the analysis of policy scenarios that include comprehensive land-use
plans. growth management regulations such as urban growth boundaries. minimum and
maximum densities, mixed use development, redevelopment environmental restrictions on
development, and development pricing policies as well as the range of transportation
infrastructure and pricing policies handled by the linked travel demand models.
5.5 Review
The review is organized In terms of a series of tables. These are introduced in sequence below.
References for statements made throughout this section are generally not given. These sources,
however. have been cited in Section 5.4 under the general descriptions of each of the models.
Table 5. ~ presents some very general facts about the operational history and availability- of each
of Me six packages. As Indicated In the table. operational experience varies dramatically across the
packages, from being currently Implemented for the first time ~YMTC-LUM) to over 25 years of
experience in multiple urban areas (IThUP, MEPLAN/TRANUS). The packages run on a variety
of platforms although most are Ilow- typically PC-based. Most are supported by (at best) a small
consulting group, which typically depends heavily on the mode} developer. One exception is
MEPLAN, which is supported by a fair-sized consulting firm.
- Cot
OCR for page 126
OCR for page 127
TCRP H-12 Final Report
Table 5.2 summarizes how Me six models deal with the primary physical entities within urban
areas: time' Carol. and developed space (buildings). Five of the six models are static equilibrium
models, which either jump directly to the end-year equilibrium state or are moved from one
equilibnurn point to the next in Epically five-year steps. The exception is UrbanSirn, which operates
on a one-year time step arid which does not assume equilibnum.
All six models are zone-based, with the three older models (IThUP, MEPLAN TRANUS)
typically being implemented using a very coarse zone system. ME' SSA and N7YMTC-LUM operate
at the ~aff~c zone level. Urbar~S~m uses two levels of spahai detail: the traffic zone level for travel
demand calculations, and the individual parcel for land supply and demand calculations. The
developers of MUSSA are also experimenting web more m~cro-scale analysis capabilities, but these
are far from operational at time of writing. Five of the six models explicit represent developed
space In one way or another. The exception is IThUP. In which households consume land directly,
win no explicit representation of the built environment being maintained.
Table 5.3 discusses the representation of the transportation system within the six packages. All
six interact win a multimodal ~ar~sportation network model. MEPLAN~ and TRANTUS have buiTt-in
network modeling capabilities; MUSSA, NYMTC-LUM and UrbanSim are "connected" to stand-
alone four-stage modeling systems, and IThUP can function either way.36 In addition at time of
writing, UrbanSim is being integrated with art activit,~-based travel mode} in Honolulu.
In all cases, the information passed Tom the network mode! to the land-use component takes the
form of venous "accessibility measures" typically Serif ed from random utility theory. That is, the
accessibility teens typically are so-called "possum" or "inclusive value" terms taken Tom the posit
. . . ., ,. . ~. .
,~, ~
models used within the travel Lemma modeling system. In general. the ability of the integrated
mode] to analyze transportation policy impacts (on tras e! or land-use) depends on the quality and
capabilities of the four-stage travel Remarry mode} being used, rather than on He integrated modeling
system per se. All six models explicitly include trar~sit although He large-area zones typically used
in the older models (ITEUP, MEPL`AN~, TRANUS) clearly limit the sensitivity of these models to
transit system impacts; also, ITEUP's internal travel demand mode] does not include a transit
· ~ · t · ~ ~ ~ ~ ~ ~_% ~ _ ~ ~ . ~
assignment capabllltv. M~FLAN and 1 lU\N U b are me only systems which explicitly model goods
movements. bothers' modes (such as HOV/carpoolina and paratr~sit) typically are not modeled
although' again' this is prLmanly a Unction of the travel demand model being used' rather than the
integrated model per se.
|6 MUSSA and NYMTC-LUM me somewhat inte~ediate ~ this respect. MUSSA is explicitly designed to
connect with ESTRAUS, the Santiago travel demand model. The combined GLUT package can be
considered an integrated system. NYMTC-LUM is a somewhat simplified version of METROPOLIS,
which is intended to be a fully integrated modeling system.
- 10,
OCR for page 146
TCRP H-12 Final Report
All models reviewed are much less able to address regulatory policies, especially with respect
to transportation. TOM and ITS impacts are generally not well addressed. These limitations
primarily reflect well known weaknesses in the four-stage travel demand modeling system, upon
which these integrated models heavily rely, rather than any fundamental problem with land-use
models per se. Thus, as operational travel demand models continue to improve (through the efforts
of TMIP and other research and development initiatives), integrated models can be expected to
directly benefit as well.
Not surprisingly, all models reviewed can only respond to educationJmarketing policies aimed
at changing people's values/attitudes through exogenous changes in the mode] parameter values,
where these parameters are intended to capture decision-makers' tastes and preferences. All models
are calibrated against observed, historical data. Given this, at best, they capture the behavioral
preferences manifested in the historical data. To use these models in a forecasting mode, we must
assume that these behavioral preferences wall hold in the future as well. While the increasing
availability of time-series data (both pane! and repeated cross-sections) is beginning to provide us
with the opportunity to study the evolution of tastes and preferences over time,~7 such research is a
long way from providing us with operational models of parameter formation arid evolution.
Finally, Tables S.8a and S.8b summarize much of the previous discussion in terms of the
operational characteristics of the models. Some key points to note from this table include:
Historically, integrated models have been quite aggregate spatially. The current mend among
newer models is towards a finer spatial scale, typically the traffic zone level. UrbanSim is
the most disaggregated of the models reviewed, modeling land development at the parcel
level arid travel demand at the traffic zone level.
· Most models are based on strong equilibrium assumptions. UrbanSim is a noteworthy
exception to this rule.
.
.
Most models are still temporally very aggregate, using, at most five-year time steps. Again,
UrbanSim is the exception to this rule, in which a one-year time step is used.
Data requirements and implementation effort obviously vary, depending on mode}
complexity. Any integrated urban model however, requires significant investment in time
arid money to implement.
· Significant, on-going technical support is required to operate these models.
|7 For a good overview of the use of panel data in travel demand analysis see Golob, Kitamura and Long
[1997~.
- 1~)2
OCR for page 147
c~ -
~ - u)
·- u) .= ~
~ 3 - -
_ . Q _
=0 ~
V - V,
- V - -
- ~ _ ~s
- - _
v ·- 3 _ '
~Q) - ^ V
_ .- ·- - 0
tV _ ~ - :e
~ =~ _
'~) ~ U: U. V' -
X o,) ~ ~ ~ ~ V-,
_.
~.
Z - ~
~ _ _
e~ 5 et: c~ C)-
V: ~ - C ~
CO _
V _ V
v, c, 5 5 ~ ,- v, ^
tC - ~ C ~ ~ ~
tt,) ~ _ ~ _ _ =, _
~ e :10 C :t =- ' _
-
V-, ~
_ _ _
sL) _ -
~ V ~ :e
CO ~ _ C =~
V r ~ Cx ~V.
ec {.> ~V c#
_ ~ - _ ~
_ _ _ _ ~ _ _
3, _ _ 3 · = _ . ~
3 - _ ~ ·_ ^O _
_ ~) tD CC ~- -
* _ ~- ~ _ ~ _
S ._ ~ ~ ~ ~ V
. _ -=
OD >` ~ '~
C o -._
- - G) ~ '-
O O - i= . _
_ '- >` C) C -~; ~
~U_ ~-e--
_ ~ ~ C;
_ C~
O _ C^, _ ,.
_ _ _ = ~
, ~ - V ~;~ V. ~ - V V
_ ~ ~ -3 ~_ _ *
~ ~_
_ _
C C
._
_ _ _
_ - V
_ C: _ ~
_ _ C~, . _
_ ~ _
~ Uj ~ V)
_ ~ ~ _
- Ce> - -
_ ~
_ _
_ ~ 0 0
_ ~ ~) _
_
-, = 0 =
_ _ __
C~ ~
_ _
_ ~ _
_ ~ ~ _
_ ,_ . _ 5
~ ~ , _ . _
~V _
3 j _ ec
~ V
· _ [,,
- ~ eC -
_ Q) - 3
e~ . _ ~ _
V _ ;~ ~J
~-~ ~ ~
. _
e C~ ~> ~ ~ V ~ C~ e
'v ~V C~ ~ >, Cc - ;-
_ .
-3 V
C C.~:
_ ~
_ _
_ _
_ ._
_ _
._ ^,
_ ~
._
_ _
._ _
_
G
V
'~
._ _
c,) ,_ ^ V -1
V V U) ~ V V
c ~ _ 3 ~ ~ - ~
~ 3,, _
3 _ ~
_ _
_ ~_
_ _ _
, _ _
_ ~ _
.-
. _
- _
_ _ _
:_ ~ e~
_ _
_ = U,
C.] .- -
- - tV->
t) _ __
- - V
C) ~ _
,.;~ V _
C =_ tC
_
:
~V
_
_ .
CC
_
_
_
V,
Z
<
C~
CO
c
_ . _
-
1
OCR for page 148
~ -
=- - -
3 CV,
V:
C: 0,) V_
_ ~ '.:C
CC ~ _ .~
ee U ~ ~
.--_V
~ fi
. >, V
, ~
. . U.
.,.,.,
, Q)
~ V, ~
~. - , ....
V ~ ~ - C V ~ V
ec-~ j 3 ~ - 3 et
:~ 5 . ~ - -o t~ -
- °- Ct'- - V _
V C~ ~ ~ V ^ ~ -
L) - V U 4; ~ ~ ~ V
·_ 0,
:t 5
_' ~ _
_ r~} =d _
_ ~ >_ _
_ ~ 3 -
V ~ Z Q)
a ~ .- ._
_
V
_ _
~; ~
i~ c)
_ _
U
_ ~
,, _
- _
- V
- ~
_ _
- _
V ~
C) -
_ .
V _ _
t' ., ~ eo
V _ _
l
U~
_ ._
V~
t-_
-C~
·_ E_ _
_~ ~ _
_ _
-~ _ _ _
-~ .= .=
v
Z
~ . .' ! 11
.
~ ,~ ~ ~ . ~
_ ~ ~ 3 3 ~ _
5
_ _
-
_
C) - V _
- e~ - ~ :C -
- 3 - eo_ 5
_ _ >~ ~ U. _ _
- - ~ Go - =
:^ = ~ ~ ~r.
= _ +
_
~ ~ 5 _ . G)
r, - ;> ~ ~.~ _
cc V C4 :e c) u, Ce
- V ,= ,~ _ ~d V V :)
~ ~ u~ - ce .= _ ~
_ t%; ~ ~ ._ C~ C) V
_ - ,~ _ ~ _ ~ ~ _
~ - ~ ~ ~ V~ ~ :~:
-
._ _
_ ~
,, fi
3 - _
C~ ~ _
_ ~ V
_ __ C ~
= 3~, ~
'-) {L) ~ C~ _
r, _ _ _
_ ,~,) ~ _~
3 _ = _ _
. ~ ~) 5 V
_ ~ V ~ _
~_-
~ ~ V V V
^. - ?-
-
1
OCR for page 149
:
:~
·-l
·~S -
3
~ :~C)
== ~,
o ~ ~o
.=
C) c,}
~ o _
~E o
C e
U) _
. t:
L
_
t3
_
._
C)
.
-3 ~
'- ~
~ ~ ~'~r
_ C)
~ Ct ~O
_ =; ._
0D,~ _
_ - C
C ;~ 3
_ Z O
_ _
_ ~_ _
_ _ _
_ _ _
t~ 3 -
- V: C~
oC o,o ~
. . ,= _
- C ~
_ CD O
O _
r=,
-
C)=
~ C,) -
_~ _
V ~= .`r
_ _ V _
_ ~ ~ _
= 0 ~ ~ ~o
C) _ ~? U
_C; ~ _ _
_ C-) _ U ' ~
__
- _- ~
~C)
_ >~_
_ __
_ C,,_
~ C~
_ ~_
V ~_V.
~ ~·)
~ Ce)~
_ V~_
U ~_ ~
O -_- 3
;~ C._
_ .-C:
:S} - =G
- U. _ -
._ ~, ~ ~_
:,) - _
_ C.) _
- V: _ _
_ ;, ~_ =
_ _ - _
C) _ U. _
_ >: ~ ,-t,, _
.. :, - ,= '~
_ _ ; - _ _
~ ~ ~ - _
;~ - _ _ C~,
,~
_ _ ~
_ _ ._
~ ~ -
_ - ' I C _
') '- - _
~ U. _ ~ V,
- V: ~ ~ =
,:) ~ _- ~ _
C) ;C ~ >~
~ ~ C, ~ r. _
._~ E ~ '~;
e O ~- :,
e _ ~ ta ~ _ rr
~ ~ ~ 0 _ ~ _
Z
0 <
=~.
=
._
_ ur,
_ _
Z ,_
'~ ~ ~r.
_'_ ~
_ _ _
r ~
_ t_ ~
~ 00=
· · ._ _
_C_
_
_ =_ .-
~ ~ ~ i ~ PC ~ ~ t~ig~
;~
_
Z
_
OCR for page 150
-
c
-
·
cc
A
-
-
·
·
u,
en
1
-
c}
to
G
-
So
V
TIC
E
Cut
x
Cat
Cat
., , , - j ! f - I '
~ c ~ 3 =C _.= doe, ~ _ _
_ it,, 3
. - . - ~ _
e -ivy
:- x 3 ~ _
-to - o _
--Van C) .> o
.~ _ .~ OF
':v _ ~ =.
=: 1 c~
_ ~
~ j 4~$ ' ~ = ~)
1
1
OCR for page 151
TCRP H-19 Final Report
5.6 Summary
Table 5.9 attempts to summarize the findings of this review, by comparing the general state of
practice (as defined by this review) with the ideal mode] described in Chapter 4. In this table, the
attributes of the ideal model. first presented In Table 4.7, are reproduced in small type, followed by
a summary of current practice in normally sized type. In this way one can begin to assess how'
current operational practice compares with what we are positing as the ideal or ultimate model.
towards which we should presumably be striving.
In assessing the contents ofthis table, as well as its predecessors. one should recognize that wee
have tried as best as possible to characterize the attributes and performances of the six models as
they are "typically" implemented in practice. Any specific application may, in some instances, either
exceed or fall short of what has been depicted here. In addition, the reader is reminded that the
objective of this review has not been to pick "winners arid losers," but rather to achieve as objective
an assessment of current modeling practice as possible. Thus, this general state of practice is of
much more direct interest than the strengths/weaknesses of any single model.
Summanar~g even further from Table S.9. as well as from the discussion In the previous section,
several key points emerge. These include the following.
All models are sensitive to transit - land-use interactions to varying extents. The two
major limitations In this regard have been the large zones traditionally used in these models,
and the level of transit representation in the travel demand component of the modeling
system. Use of traffic zones in more recently developed models. plus general improvements
in travel demand modeling mesons. bode well for improved Posit sensitivity in integrated
urban models. Applications of models such as MUSSA and the NYMTC-LUM family of
models to the analysis of major transit alternatives provide good examples of the usefi~Iness
of integrated urban models for transit planning.
All currently, operational models fall short of the ideal mode! to varying degrees. Areas
of significant shortfall in most models Include:
0 excessive spatial aggregation;
o excessive reliance on static equilibnum assumptions (with associated assumptions of
large time steps and lack of path dependencies)
overly aggregate represerltations of households and firms, as well as a lack of
representation of individuals as decision-making units separable from their households:
o lack of endogenous demographic processes;
0 lack of endogenous auto ownership processes; and
reliance on four-stage travel demand modeling methods.
- 127
OCR for page 152
TCEP H-17 Final Report
Table 5.9 Summary of Current Modeling Capabilities Relative to the Ideal Mode}
PHYSICAL SYSTEM
Time: Dynamic evolution of the system state in one-year time steps. System sate generally not in equilibrium.
Interactions between long-run and short-run processes are "properly" accounted for.
Most models are static equilibrium models with large tone steps (5 years or more). One-
year tune steps certainly feasible, as demonstrated by UrbanS~m.
Land: The basic unit of land is the individual lot.
All models are zone-based, often with very large zones. Recently developed/emerging
models tend to use traffic zones. UrbanS~m uses individual parcels for modeling land-
use. MUSSA experimenting with micro-level analysis capabilities.
Building Stock: Building stock is explicitly represented. Each lot has a certain amount of floor space.
characterized by type, price, etc.
Building stock explicit in most models. Small number of building types usually used.
Transportation Networks: Full, multimodal representation of the transportation system used to move both
people and goods. Sufficient spatial and temporal detail to properly model flows. network performance, emissions, etc.
Ideally, a dynamic. twenty-four hour network model to be used.
All models involve use of a multimodal network. Spatial detail often could be improved.
Static, temporally aggregate representation of flows. Goods movement only represented
in MEPLAN and TRANUS.
Services: Sufficient representation of other services for the purpose of modeling land development decisions.
Non-transportation services generally not explicitly represented in the models.
- 128
OCR for page 153
TCR~ H-12 Final Report
.
Table S.9
St~n~mary of Current Modeling Capabilities Relative to the Ideal Model, cont'd
DECISION MAKERS
Persons & Households: Both persons and households are explicitly maintained (with appropriate "mappings"
between the two entities) in sufficient detail tO model the various processes of interest.
All models are household-based, with very little explicit representation of individuals are
separate decision makers. Households are grouped into relatively homogeneous
categories, often using a relatively small number of categories. MUSSA and UrbanSun
provide examples of a high level of household disaggregation. in the limit approaching a
totally disaggregated "microsimulation" approach.
Firms: Explicitly represented. Firms at least as important as households in the overall system: they occupy land /
floor space; they employ workers; and they buy / sell goods and services from-/ tO themselves and households. Firms
are modeled in sufficient detail to capture adequately their behavior within these various roles.
Oider models typically mode} employment directly, rather than fimns. More recently
developed/emerging models model firms explicitly, usually using a small number of firm
types.
Public Authorities: Represented within the model tO the extent they generate purely endo~enous effects
(employer of workers; demander/supplier of services; etc.~. Will remain largely represented by exogenous inputs to the
model.
Public authorities are almost entirely exogenous the models. The only exceptions are
MEPLAN and TRANUS, which include "Government" as an explicit sector in their
regional I/O models.
L I
- 129
OCR for page 154
TCRP 1:~-19 Final Report
Table 5.9 Summary of Current Modeled Capabilities Relative to the Ideal Model, cont'd
PROCESSES
Markets: Land development residential housing. commercial floor space and labor all occur within economic
markets which possess demand and supply components and price signals which mediate between demand and supply.
These economic markets must be explicitly modeled if their behavior over time is to be captured properly.
Most models have explicit supply-demand market interaction models for land development and
building stock allocation. Some mode! the job market explicitly; some use work trip distribution
models to link workers with jobs. Some link housing demand and job allocation processes so
closely that the two processes are essentially treated as one. Strong static equilibrium assumptions
are generally imposed to balance demand and supply in each time step (UrbanSim is the one non-
equilibriurn based model considered using a one-year time step).
Demographics: Demographic processes should be modeled endogenously so as to ensure that the distribution of
population attributes (personal and household) are representative at reach point of time being modeled and are
sufficiently detailed to support the behavioral decision models being used.
No mode! currently includes a significant demographic component. Households typically
categorized by the, often with only a few household types used. Totals by household type
typically supplied by exogenous inputs.
Regional Economics: Components of urban production/consumption processes which are endo~enous tO urban
areas should be modeled endogenously. The model should also be sensitive tO macro exogenous factors such as interest
rates, national migration policies. erc.
MEPLAN and TRANUS are built around a regional input/output (~/O) model. MUSSA takes
output totals from a regional I/O model. Otherwise, regional economic data are all exogenously
generated.
Activity/Travel: The travel demand component of the integrated model should be activity-based and sufficiently
disa~regated so as to properly capture trip-makers' responses to a full range of transportation policies. including ITS
and TDM.
All models use variations on conventional four-stage modeling systems, and so are susceptible to
all the usual criticisms of these models [Stopher, 1993; Deakin and Harvey, 1993, Miller and
Hassounah. 19933. UrbanSim is being integrated with an activity-based travel mode] in
Honolulu.
Automobile Holdings: Household auto holdings (number of vehicles, by type) should be endogenously
determined ~ ithin the model.
All models essentially ignore auto holdings. At best, the household categorization scheme may
- 1,0
-
OCR for page 155
TC8P H-12 Final Report
-
Of the ~nodels reviewed. UrbanS~m clearly comes closest to the ideal mode] specification
with respect to most of these points, including: spatial disa~regation (use of parcels to
mode] land development), temporal aggregation (use of one-year time steps) dynamics
(disequilibnum model), detailed disaggregations of households and firms (an a~bute shared
by MUSSA), =d use of activi~-based travel models (on-going In the Honolulu application).
· At the same time, current models indivi(lually and collectively display many strengths
and generally provide a solid basis for further evolutionary improvements. Strengths
include:
generally strong microeconomic formulations of land and housing/floorspace market
processes;
coherent frameworks for dealing with transportation - Arouse interactions;
multimodal transportation network analysis capabilities; and
experience with developing and using large-scale integrated models.
· Despite the scope for significant evolutionary development among existing models, a
"new generation" of integrated models drill need to be developed in order to fully
achieve the ideal model. While newer models such as MUSSA and UrbanSim point the
way to more disaggregate andlor more dynamic models. much research and development
must be undertaken in order to filly achieve the idea] model. This will include development
of and experimentation with mode! structures which are explicitly designed to operate in a
more disaggregate, dynamic non-equilibrium framework.
These observations provide the starting point for the development of an integrated short- and
lon~-run research and development program for integrated urban models. which will be discussed
in detail in the next chapter.
I,]
OCR for page 156
ACRE H-12 Final Report
- 1~2
Representative terms from entire chapter:
six models
