United States
Environmental Protection
Agency
Air
Office of Air Quality
Planning and Standards
Research Triangle Park, NC 27711
EPA-452/R-93-003
January 1993
JV EPA ISSUES AND APPROACHES TO
IMPROVING TRANSPORTATION
MODELING FOR AIR QUALITY
ANALYSIS
-------
ISSUES AND APPROACHES TO
IMPROVING TRANSPORTATION
MODELING FOR
AIR QUALITY ANALYSIS
OFFICE OF AIR QUALTTTY PLANNING
AND STANDARDS
and
OFFICE OF MOBILE SOURCES
U.S. ENVIRONMENTAL AGENCY
JANUARY 1993 U.S. Environmental Proton Agency
Region 5, Library IrL-i^Jj
77 West Jackson Boulevard, 12th
Chicago, IL 60604-3590
-------
Notice
This material has been funded by the United States Environmental
Protection Agency (EPA) under contract 68000102, Assignments 2-16
and 3-08 to Systems Applications International (SAI) and the Urban
Analysis Group (UAG). Significant technical material was developed
by SAI personnel including Julie Fieber, Lori Duvall, and Doug
Eisinger. For UAG, Edward Granzow, and Jo Ann Coxey made
significant contributions. Sharon Reinders of the Office of Air
Quality Planning and Standards and Hark Wolcott of the Office of
Mobile Sources served as co-Work Assignment Managers of the
project. Ted Creekmore assisted with the final review and
publication. This document has been approved for publication as an
EPA document. Mention of trade names or commercial products does
not constitute endorsement or recommendation for use.
-------
Executive Summary
Driven in large part by passage of the Clean Air Act Amendments (CAAA) of 1990,
increased attention and resources have been directed towards improving the procedures of
transportation modeling in order to arrive at information better suited for air quality analyses.
Several studies sponsored by the EPA, national organizations, and state and local agencies
have been initiated in the last two years to try to improve transportation modeling. This
report documents the results of one of these efforts. The purpose of this work was to
produce a list of current shortcomings both in transportation model structure and in the ways
transportation models are used, written in large part from the perspective of air quality
modelers. The intention has been to provide a document which would be of use to both
transportation and air quality modelers. In addition, a list of improvements to either the
models or transportation modeling procedures, augmented by sample model runs
demonstrating implementation of some of these suggestions, is provided.
-------
Contents
Executive Summary i
1 INTRODUCTION 1
Background 1
Organization of the Report 2
2 TRANSPORTATION MODEL SHORTCOMINGS WITH RESPECT
TO EMISSIONS 3
Development of Air Quality Modeling Inputs from
Transportation Models 3
Mobile Source Emissions 4
Vehicle Classes 5
Roadway Types 5
Spatial Resolution of Mobile Source Emissions 9
Temporal Resolution of Mobile Source Emissions 9
Limitations of Transportation Models From An Air
Quality Perspective 10
Model Algorithms 10
Model Inputs 13
Simplifications Used in Model Execution 16
Summary of Concerns and Their Relationship to Trip, VMT,
and Speed Estimation 19
3 HOW TO DIAGNOSE MODELING DIFFICULTIES 22
Key Questions to Ask Your Transportation Modeling Agency 22
Trip Generation 23
Land Use and Model Calibration Data 24
Generation of Productions and Attractions to TAZs 24
Trip Types 25
Structure of Traffic Analysis Zones 25
Trip Distribution 26
Development of Friction Factors and K-Factors 26
Intrazonal Travel 27
Mode Choice 27
Treatment of Public Transit 28
Determination of Public Transit Travel 28
Urban Bus VMT 29
92004rl 01
ii
-------
Trip Assignment 30
Choice of Travel Route 30
Choice of Initial Speeds 31
Assignment Methodology 31
Use of Feedback 31
Temporal Variation in Traffic 32
Summary of Key Questions 33
4 AIR AGENCY STAFF GUIDANCE: SUGGESTED PROCEDURES FOR
IMPROVING THE LINK BETWEEN TRANSPORTATION AND AIR
QUALITY MODELS 39
Overview 39
Suggested Procedures for Improving Transportation Modeling 39
General Improvement to Travel Demand Forecasting 41
General Procedural Improvements 43
Factoring and Cross-Classification Processes 43
Intractable Problems and Statistical Limitations 45
Modifications to Models to Address Air Quality 45
Demonstration of Selected Procedures . 46
Description of Base Case 46
Option A-3: Use Transit Speeds that Reflect Congested
Roadway Conditions 51
Option A-5: Examine the Impact of Using the Loaded
Highway Network 52
Option C-l: Postprocessing of Link-Based Travel to
Estimate Emissions 54
Recommended Follow-Up 54
References 60
Appendix A: Characteristics of Prototypical Model
Appendix B: Comparison of Option A-5 Results for Highway and Transit Networks
92004rl.01
iii
-------
1 INTRODUCTION
BACKGROUND
This document has been prepared to explain to transportation modelers the detailed input
requirements of air quality analyses and to explain to air quality modelers ways in which
the current practices in transportation modeling may not always be well-suited for
providing these inputs, largely because of data limitations. Some suggestions, largely
relating to improved communication between transportation and air quality personnel, are
made. An example of one suggestion explored in this document is focusing more
attention on the assumptions used in modeling intrazonal travel. Transportation models
were not designed for explicitly treating this type of travel and yet from an air quality
perspective it is important. Many of the other suggestions offered in this document are
actually post-processing steps, intended to incorporate more detail into the vehicle
emission inventories prepared from transportation model outputs. As an example, we
have included an example of one procedure which illustrates the importance of properly
matching fleet mix to road type when determining what average fleet emission rates
should be used with the vehicle activity estimates produced by transportation models.
Suggestions are included that are appropriate for areas with preexisting transportation
models as well as for areas that are beginning development of a transportation model or
that are in the process of updating a preexisting transportation model.
Transportation demand analyses can be envisioned as comprising four steps: the
estimation of trip generation, trip distribution, mode choice, and trip assignment. These
four steps are often referred to as the "urban transportation planning, or modeling
system" (e.g., see Meyer and Miller, 1984). Computer models or manual procedures,
such as sketch planning methods, exist as tools for transportation demand analyses (SAI,
1989). There is great variety in the approaches available for transportation demand
analyses.
Both transportation and vehicle emission models have a variety of uncertainties and
difficulties associated with their use. Generally, both provide better results on a macro
rather than microscale basis. Thus, transportation models are likely to be more accurate
at predicting total VMT on a county-wide basis than at the level of individual links, and
vehicle emission factor models are better at developing regional emission estimates than
estimates for a particular intersection.
92004r2.04
-------
Transportation models are generally designed for the evaluation of urban transportation
needs and network designs. This type of use does not require the same type of detailed
transportation activity data as is required when developing emission inventories for air
quality analysis. For example, an air quality analysis might require seasonal, day-of-
week, and/or hourly estimates of travel activity (including average speeds) in order to
properly account for the effects of temperature on vehicle emission rates or to develop
vehicle emission estimates that reflect travel behavior on a specific day or portion of day.
Spatial allocation of vehicle miles traveled and even trip starts and trip ends is typically
required when conducting an air quality analysis. Typical transportation models are not
designed to provide such detailed information, yet information developed in the
transportation modeling process can help supply the data needs of air quality modelers.
For example, though it is generally unreasonable to attempt to survey traffic and run
transportation models on an hourly basis, it is reasonable to use the models to provide
predictions of peak and off peak travel activity, and even to differentiate between
morning and afternoon peak periods. Interpolation can then be used to estimate hourly
travel activity. Most transportation models are already capable of providing estimates of
travel activity by roadway link and estimates of trip starts, trip ends, and intrazonal travel
by zone, thus providing the detailed information on the spatial distribution of travel
required for air quality analysis.
ORGANIZATION OF THE REPORT
The focus of Chapter 2 of this report is to describe the specific requirements placed on
transportation models to provide necessary inputs for air quality analyses, and some of
the qualities of the current generation of transportation models which limit their use in air
quality analyses. Chapter 3 develops these ideas further through a list of questions which
should be discussed between transportation and air quality modelers regarding each of the
four steps of the transportation modeling process. The goal is to improve the air quality
modeler's understanding of the relative strengths of vehicle activity estimates produced by
transportation models. Finally, in Chapter 4 procedures are suggested for tailoring the
results of transportation model exercises more specifically for air quality analysis,
accompanied by results from sample model runs which demonstrate a few of these
suggestions.
It should be noted that some of the issues of concern with transportation modeling
discussed in this report will generally be of most importance if model predications are
used to prepare modeling inventories for urban-scale air quality models, such as the
Urban Airshed Model (UAM). Such applications are often required for areas that have
not attained ozone and carbon monoxide standards. Not every air quality analysis needs
to have as detailed information on the temporal and spatial distribution of vehicle activity
as is needed for photochemical grid models like the UAM; sometimes annual regionwide
estimates are sufficient. The suggestions included in this report should, in general, result
in the more accurate representation of vehicle activity for these applications as well as for
the production of modeling inventories.
92004r2.04
-------
2 TRANSPORTATION MODEL SHORTCOMINGS
WITH RESPECT TO EMISSIONS
This chapter sets out a description of what is required from transportation models in
order for air quality personnel to develop vehicle emission inventories. This is followed
by a discussion of areas where travel demand models do not perform as well as needed
for emission inventory development, and what the implications of these model limitations
are for emission estimates calculated with travel demand model output.
DEVELOPMENT OF AIR QUALITY MODELING
INPUTS FROM TRANSPORTATION MODELS
In the past, air quality control strategies were based on relatively simple emission inputs;
generally county-wide annual average seasonally adjusted emissions. However, the Clean
Air Act Amendments (CAAA) of 1990 have required that some regions use
photochemical grid modeling, which places more stringent demands on emission inputs.
The primary requirements of the emission inventory to be used in grid modeling can be
summarized in the following manner:
Estimates of precursor pollutants must be provided for each individual grid cell of
a modeling domain rather than at a county level; these may be as small as 2 km x
2 km.
Hourly emission rates must be provided instead of annual average emissions.
Note that estimates of peak and off-peak period travel activity provided by most
transportation models can be used to estimate hourly travel activities without
resorting to the collection of hourly data.
Total organic gas (TOG) emissions must be disaggregated into individual chemical
classes corresponding to the chemical mechanism of the photochemical model.
The emission factors used to estimate emissions from on-road motor vehicles vary non-
linearly with a variety of parameters, including vehicle type, vehicle speed and
acceleration, fuel volatility, vehicle fleet characteristics, ambient temperature, diurnal
temperature variations, and vehicle fleet inspection program characteristics. These
emission factors (which are usually reported in terms of grams pollutant/vehicle mile or
92004r2.02
-------
event) are then used with an activity level (e.g., VMT or number or event?;) to generate
on-road vehicle emissions estimates. Ideally, link-specific traffic parameters will be used
to generate the emission estimates in order to retain information on the spatial distribution
of emissions. Various inventory classification schemes may then be employed to
aggregate these emissions into a manageable number of categories; two examples include
vehicle class and road type.
To properly understand the requirements placed on transportation models for developing
relatively accurate inputs for air quality analyses, an understanding of motor vehicle
emission rates is useful. The following discussion briefly describes the parameters which
affect mobile source emission estimates. The elements discussed are emission
components, vehicle classes, roadway classifications, spatial distribution, and temporal
distribution. Areas where information either is currently or could be supplied by
transportation models are indicated in the discussion.
Mobile Source Emissions
Mobile source emissions should be disaggregated into their components in order to
properly distribute them spatially and temporally, and in order to properly speciate the
hydrocarbon emissions. The individual components of the on-road vehicle emissions are
defined below:
Exhaust emissions: vehicle tailpipe TOG, oxides of nitrogen (NOX), carbon
monoxide (CO), particulates, and oxides of sulfur (SOX) emissions which occur
during the operation of the vehicle. These emissions occur along transportation
links, and are highest during peak vehicle operating hours.
Evaporative emissions: TOG emissions which include diurnal emissions, resting
losses, and hot soak emissions. Diurnal emissions result from fuel vapor
expansion occurring during periods of rising ambient temperatures. Resting losses
occur due to fuel vapor permeating through fuel lines, loaded canisters, and liquid
leaks. Hot soak emissions consist of the evaporation of fuel by engine heat
immediately following the end of a trip. All are associated with stationary
vehicles, and their amount is a function of fuel volatility, and ambient
temperature.
Running loss emissions: evaporative TOG emissions that occur during the
operation of the vehicle. These emissions also occur along transportation links,
and are highest during peak vehicle operating hours.
Exhaust and evaporative emissions must be differentiated because of the different TOG
species profiles (which affect ozone production and toxicity) for these two categories, and
because the spatial distributions of these emission modes are different. Exhaust emissions
92004r2.02
-------
occur primarily along roadways; evaporative emissions, because they occur while a
vehicle is parked, occur primarily in residential or business districts.
As mentioned throughout this report, the vehicle emission rates for all of these emission
modes are strongly dependent upon temperature and, for exhaust and running loss
emissions, vehicle speed. For example, the temperature dependence of light-duty vehicle
emission rates is demonstrated for total organic gas (TOG) emissions in Figure 2-1, and
the speed dependence of CO emission rates in Figure 2-2.
Vehicle Classes
The registered vehicle fleet can be divided into sub-groups, or classes, such as autos,
light-duty trucks, and heavy-duty trucks. The emission factors associated with each
vehicle class vary widely because of differing emission certification standards and
pollution control equipment. For example, the MOBILE model (EPA, 1991)
distinguishes eight vehicle classes, listed in Table 2-1, based upon gross vehicle weight
(GVW) and fuel type (gasoline or diesel). Inventories will typically use some
combination of these eight vehicle classes to report emissions, such as LDGV, LDGT,
HDGV, and HDDT.
Currently, transportation models do not report VMT by vehicle type. In order to
estimate emissions, assumptions must often be made to disaggregate VMT and trip ends
by vehicle type, and a way found to distinguish between commercial traffic, transit buses,
and fleet operations. For some purposes, VMT associated with alternative fuels must
also be disaggregated.
Roadway Types
On-road mobile source emissions should also be distinguished by road type in the
inventory. Some of the road types for which the Federal Highway Administration
(FHWA) maintains statistics are listed in Table 2-2; these road types are commonly used
in mobile emission inventories. Emission factors will vary by road type because of
changes in speed and fleet distributions associated with different road types.
For example, a rural or urban interstate might be preferentially used as a transportation
route for trucking operations, or for the movement of farm products during harvest
periods. Hence, under such conditions, one might expect a higher concentration of
heavy-duty diesel vehicles on this roadway type, which would cause the fleet distribution
to be markedly different from that which would be found on roadway types which are
used more for commuter traffic, such as urban collector roadways.
92004r2.02
-------
30
40 50 60 70 80
Ambient Temperature (°F)
90
99
-* Exhaust TOG + Evaporative TOG ^ Running Losses & Resting Losses
Figure 2-1. 1992 light-duty gasoline vehicle TOG emission rates as a
function of temperature.
-------
100
10 19.6 30 40 50 60
Speed (mph)
65
70
75
Figure 2-2. 1992 light-duty gasoline vehicle exhaust CO
emission rates as a function of speed.
-------
TABLE 2-1. Vehicle class definitions used by the MOBILE model.
Vehicle Class (Abbreviation)
Light-duty gasoline vehicles (LDGV)
Light-duty gasoline trucks 1 (LDGT1)
Light-duty gasoline trucks2 (LDGT2)
Heavy-duty gasoline vehicles (HDGV)
Light-duty diesel vehicles (LDDV)
Light-duty diesel trucks (LDDT)
Heavy-duty diesel vehicles (HDDV)
Motorcycles (MC)
GVW*
Specification
Not applicable
Less than 6500 Ib
6500 to 8500 Ib
More than 8500 Ib
Not applicable
Less than 8500 Ib
More than 8500 Ib
Not applicable
Typical
Percent of
Total VMT
62.3
17.3
7.7
3.5
0.8
0.2
7.4
0.8
*Gross vehicle weight
TABLE 2-2. Commonly used road type designations in emission inventories.
Rural and Urban Interstate
Rural and Urban Other Principal Arterials
Other Freeways and Expressways
Rural and Urban Minor Arterials
Rural and Urban Major Collector
Rural and Urban Minor Collector
Rural and Urban Local
92004r2.03
-------
Most transportation models can report VMT by roadway type. However, often both
VMT and speed are reported by roadway type averaged over an entire county or
transportation modeling region. Improved emissions and air quality calculations can be
made if these data are reported on a link-by-link or some other sub-county level, because
this allows travel activity to be matched with emission factors which correspond to the
speed and temperature conditions under which it occurred. Again, this level of detail is
primarily of importance when using gridded photochemical models such as the UAM.
The location of emissions and the time of their release strongly affect the ozone
production and the development of CO "hot spots."
Spatial Resolution of Mobile Source Emissions
Exhaust and running loss vehicle emissions differ from most other area source categories
in that it is generally easier to determine where these emission occur in a region. Spatial
distribution of exhaust and running loss emissions can be accomplished using a link-based
rather than area-wide surrogate-based gridding procedure such as is used for most area
sources. Link-based spatial allocation results in distributing emissions only to those grid
cells that contain transportation pathways. In contrast, most evaporative vehicle
emissions occur at the location of trip ends, and depend on the number of hours a vehicle
is left parked. As a rough approximation, evaporative emissions can be spatially
distributed the same as exhaust and running loss emissions, using the gram per mile
evaporative rates provided by the EPA's MOBILE model. However, it is more accurate
to differentiate between the two emission types, and allocate exhaust and running loss
emissions to roadway links and evaporative emissions to the zones where trip ends occur.
Within each zone, evaporative emissions can either be assigned to the location of the zone
centroid, if the zone is small, or be distributed over the entire zone area using a spatial
surrogate such as population. Under most circumstances, the more detailed approach is
not necessary unless photochemical modeling inventories are being prepared.
Temporal Resolution of Mobile Source Emissions
Temporal adjustment of the mobile source inventory into monthly, daily, and hourly
totals is not a straight-forward process, since transportation models provide little
information about the temporal variation of traffic patterns. As noted earlier,
transportation models are generally calibrated for use in determining either average or
"peak" (e.g., morning or afternoon commute) traffic conditions. Due to the lack of
reported information, air quality modelers make assumptions about traffic patterns to
estimate the diurnal variation pattern for on-road motor vehicle emissions appropriate for
the modeling episode. As a special consideration for weekend emission inventories, note
that diurnal variation in weekend driving activity usually differs markedly from weekday
patterns. Also, air quality modelers usually assume that the mix of vehicle classes (i.e.,
92004r2.02
-------
problem which arises due to this assumption concerns early morning delivery schedules,
which would be expected to result in more activity from heavy duty vehicles in the early
morning and a preponderance of light duty vehicles during the am and pm-peak commute
hours. In most areas resources do not permit collection of data to explicitly address these
issues. However, transportation modelers may have available information on hourly or
daily variation in traffic activity that can be used to process the activity estimates from
transportation models to arrive at estimates relevant to the time period (e.g., weekday,
weekend, hour) of interest hi an air quality analysis. Alternatively, data that can be used
hi allocating daily or peak/off-peak period travel to each hour of the day or for
differentiating between weekday and weekend travel patterns may be available from
regions that have similar travel patterns to the area under study. Again, this effort need
not require development of a transportation model explicitly for weekends or for
modeling hourly travel, but instead makes use of data that may be available to
transportation modelers for interpreting the outputs from transportation models.
Having described the products from transportation models required by air quality
modelers, and the ways in which these products are employed in vehicle emission
estimation, the next section will discuss some problems with the information produced by
transportation modelers, from an air quality perspective, and implications of these
problems for emission estimates.
LIMITATIONS OF TRANSPORTATION MODELS
FROM AN AIR QUALITY PERSPECTIVE
Transportation models of varying levels of complexity have been available for use in
transportation and air quality studies since the 1950s. Although considerable differences
exist between models, they often share a similar framework of assumptions and
limitations. The weaknesses of the models from an air quality perspective arise from the
model algorithms, the quality of data fed into them, and the way in which a user chooses
to exercise them.
Model Algorithms
Model algorithms affect transportation model results throughout the process. Models
include different levels of feedback control, different levels of sophistication for the land
use and trip generation process, different speed assignment methodologies, etc. A
general limitation of travel demand models is that they have been designed for the
analysis of regional, corridor, or major faculty traffic patterns rather than for analysis of
project-level effects, e.g., development of suburban housing tracts or HOV lanes (jsmart,
1991). This limits the use of their outputs to generate accurate estimates of emissions
either on a regional basis, where much of the traffic occurs on minor facilities, e.g.,
urban streets, or on a project level.
92004r2.02
10
-------
All four-step transportation models require the availability of a data set which can be used
to characterize the trip generating power of each TAZ. For historical years, this data can
be based upon actual knowledge of land uses; relevant information is often collected by
local agencies. For future years, either land use models or individual judgement, or
more generally a combination of the two, must be relied upon to determine land use
patterns. One of the difficulties encountered in land use forecasting is to maintain
consistency between land use and the transportation system. This consideration must be
included in land use models.
First developed in the 1970s and currently enjoying a resurgence of use, land use models
have been used in several U.S. urban areas. Seattle, Los Angeles, San Diego and a
number of other urban areas are using such models to forecast land use. These models
do create an explicit linkage between accessibility, i.e., the transportation system, and
land use. The main components consider accessibility and historical inertia in developing
spatial allocations of development. Difficulties in the use of these models stem from the
extensive calibration process required and their insensitivities to certain variables in the
spatial allocation process. The model being used most extensively in this country is the
Disaggregated Residential and Employment Allocation Model (DRAM/EMPAL),
developed and supported by Dr. Steven Putman at the University of Pennsylvania,
although other models, such as Hammerslag's Spatial Allocation Model, have been
developed.
Nevertheless, the interrelationship between socioeconomic and demographic patterns and
transportation system changes has not been effectively treated by any model to date
(Deakin, 1991). Land use modeling in general is still lacking much of the sophistication
needed; no commonly used models treat multi-centered cities well, or capture economic
or gender-based differences in drivers, or predict the influence of crime upon land use
(Deakin, 1991). Because land use forecasts do not always accurately identify the
attractiveness of high growth areas, transportation models often fail to properly assign
future vehicle trips to these high growth areas. Inaccuracies in land use forecasts result
in inadequate representation of the transportation infrastructure that will be built to
service these growing areas, and inaccuracies in the amount of assigned trips. Emissions
estimates suffer accordingly. These deficiencies limit a model's ability to correctly
forecast the spatial distribution and volume of future travel.
Another problem with some transportation models is that there are conceptual problems
inherent in "one-pass" modeling that do not consider feedback effects. As an illustration,
changes in travel behavior (e.g., as might result from implementation of roadway tolls)
influence congestion and, consequently, travel times; such changes could lead to further
behavioral changes that would not be observed unless a model were run until equilibrium
conditions were reached. However, though more areas are beginning to use feedback in
their models, the sensitivity of models to feedback is still being assessed. A number of
different feedback cycles can be implemented in the modeling process. In general, the
shorter the cycle, the more common it is to find it used in the modeling process. The
92004r2.02
11
-------
most, common feedback cycle is the use of vehicle delay to recalculate optimum path. It
is the basic nature of the capacity restrained assignment process. Another reasonably
common feedback cycle is using travel times from congested traffic assignments to re-
estimate modal split. Models which re-estimate trip distribution and/or trip generation
based on such criteria are much less common.
Most commercially available software includes equilibrium assignment capabilities. The
other feedback cycles discussed simply require the ability to reinput the results of later
model stages in a form that can be used by the stage where feedback is desired. The
major commercial and public software packages also include this capability. The major
capability missing for implementing feedback cycles is a method of testing for
equilibrium (program exit criteria). EMME/2 has such a capability for assignment-modal
choice feedback, and TRANPLAN is currently implementing similar capabilities.
Speed estimates generated by transportation models are subject to substantial uncertainty
depending upon the assumptions built into the modeling exercise. In some cases, for
example, speed estimates are derived by assuming that travel occurs on relatively
uncongested roadway links. Because speed estimation is often coupled with capacity
restraint procedures, assigned speeds may be too high for congested periods, which can
significantly underestimate emissions (Cambridge Systematics, 1990). Typically, only
one speed estimate is provided for an entire day, or at most peak and off-peak speed
assignments are provided, which does not provide as complete a resolution of diurnal
variation in vehicle traffic patterns as is required for accurate emission estimation.
Capacity restraint procedures consist of a family of methods to calculate impacts of
congestion on vehicle operating speeds. Typically, these methods are implemented as
iterative procedures which first assign zone-to-zone traffic interchange volumes based on
a set of assumed speeds. These assumed speeds are used to calculate the shortest path
between each pair of TAZs in the network model. After all traffic volumes have been
assigned to the network, a revised speed and travel time is calculated for each network
link based on the relationship between link volume and link capacity. These revised
values are used to calculate a new set of paths and subsequently a new traffic assignment.
The process continues until program exit criteria are met.
Two basic types of exit criteria are normally used. Iterative and incremental capacity
restraint are heuristic procedures with the user defining program exit criteria based on
what is considered to be the optimum number of iterations. Equilibrium assignment is an
optimizing procedure that attempts to reach travel time equilibrium among alternative
routes. This means the models will attempt to iterate until all paths (considering
congestion) connecting a pair of zones have the same travel time, or in other words, until
there is no advantage to a user switching paths.
An additional problem with model speeds is that they are as much a model input as
output, since they are frequently chosen to allocate trips to balance the network. In some
92004r2.02
12
-------
circumstances this results in predicted volume to capacity ratios in excess of 1.0 (Applied
Management, 1990). A user may sometimes replace model speed assumptions with those
derived from Highway Capacity Manual volume to capacity ratios, but such relationships
are based upon national average data and do not represent regional speed variations.
Finally, models are not designed to predict mileage for non-network coded facilities.
Intrazonal travel is an example of this situation. Transportation modelers may typically
choose a single region-wide speed to represent such travel.
Model Inputs
Some of the inputs used by travel demand models are derived from the outputs of other
models, i.e., land use or trip generation models. These models have their own associated
errors which are often not considered when utilizing them for transportation modeling.
Likewise, while each model may have been validated individually against a set of
observed data, often the linked series of models are not validated together (Applied
Management, 1990). This allows for a compounding of model errors when output from a
model that may have appeared correct for the wrong reasons is used as input to the next
model.
Some of the difficulties encountered when using transportation model outputs to estimate
vehicle emissions arise not from problems with the models, but with the quality and
resolution of the data used in the models. Figure 2-3 illustrates the typical four-step
modeling process (SAI, 1989). Each step of this process requires inputs from the
previous step as well as exogenous model inputs, all of which introduce uncertainty into
the model outputs. Table 2-3 summarizes inputs and outputs from a typical four-step
travel demand model process.
Several of the inputs listed are obtained from regional origin and destination surveys,
which can be out of date. In some areas much of this data was collected during the
1960s. Demographic conditions shift (an example is the proliferation of dual income
households), VMT growth rates exceed population growth rates, and "suburbanization" of
the work force increases. It is important to insure that forecasted trip activity is based
upon realistic travel conditions. This is particularly true with respect to non-work trips.
With older origin and destination surveys, extensive financial resources were sometimes
combined with less sophisticated (in retrospect) survey procedures for a collection of
large amounts of data. In recent years the financial resources of most areas have not
allowed for such extensive surveys, so more sophisticated methods are used to
characterize travel behavior in an area, with more reliance based upon taking statistical
samples from subsets of a region's population. Uncertainty in this statistical sampling
affects travel demand model results, but is generally not reported (Cambridge
Systematics, Inc. 1990).
9200412.02
13
-------
SCHEMATIC
DESCRIPTION
DATA INPUTS
TRIP GENERATION
TRIP DISTRIBUTION
MODAL SPLIT
I TRIP ASSIGNMENT
SYSTEM
OUTPUTS
Inventories and forecasts (population,
travel demand, land use)
Predicts number of trips produced by, and
attracted to, each zonepredicts flows, but
not origin or destination (estimates trip
frequency)
Predicts production-attractions (P-A) of traffic
movementslinks trip ends produced by
trip generation component (estimates trip
length based on P-A)
Predicts percent of traffic movement using
each mode option (transit, car, walking, etc.)
for each P-A pair (determines mode choice)
Predicts placement of each P-A flow, by mode,
on specific travel routes (estimates travel route
use by time of day)
Predicts route volumes and- travel speeds fen-
each network link, by mode, by morning and
afternoon peak and off-peak periods
FIGURE 2-3. Urban four-step transportation forecasting modeling system and typical inputs/outputs.
iii/92004
-------
TABLE 2-3. Summary inputs and outputs for each of the four principal transportation demand
analysis steps. (Source: SAI, 1989).
Trip Generation
Inputs: Socioeconomic data, e.g., population, housing (from census tract data), income,
employment.
Outputs: Number of trips originating per time period analyzed (typically a day) in a traffic
analysis zone (TAZ) (by land use type); trips are broken down by trip type
(examples: home-based work, home-based shopping, home-based other, other-based
work, other-based odier).
Trip Distribution
Inputs: Trip generation data for TAZs and travel times.
Outputs: Distributes the trips generated in individual zones into "production and attraction"
pairs, producing trip tables.
Mode Choice
Inputs: Costs, travel times, traveler characteristics, and trip tables from previous step.
Outputs: Determines which travel mode a person chooses to make a trip; assigns motor
vehicle trips and forecasts transit ridership; choices vary depending upon model and
region (sample choices: driving alone, two-person carpool, three-person carpool,
transit/walk, transit/auto). Product is trip tables representing regional travel between
zones by trip type and travel mode.
Highway and Transit Trip Assignment
Inputs: Estimates of trips from one TAZ to another, allocated among the different travel
modes (i.e., trip tables); data (computer code) of available networks. Other inputs:
roadway characteristics (e.g., capacity, travel time from "node to node"), trips per
day (for given roadways), road locations, tolls.
Outputs: Trip types are aggregated into time periods (a.m. peak, p.m. peak, off-peak) based
upon user-input trip profiles by trip type, which allocate travel activity to peak/off-
peak periods. Models then assign each trip to a roadway link(s) and produce daily
traffic volumes and speeds, broken down by peak and off-peak periods (i.e.,
emission "activity" factors); note that traffic engineers are primarily concerned with
peak period travel-local model systems/data may not produce hourly or weekend
traffic forecasts.
92004r2.03
15
-------
Simplifications Used in Model Execution
User chosen options that affect the quality of transportation model estimates of travel
activity include:
Disablement of feedback loops to save time or money;
Application of adjustments to force model predictions to match observed
network flows;
Level of aggregation at which comparisons between synthesized and
observed base-year parameters are presented;
Model validation procedures (i.e., matching ground counts of vehicles to
model predictions);
Level of abstraction used to represent streets and transit routes for
computer modeling;
Choice of forecast years.
The effects of most of these actions are relatively simple to explain. Disablement of a
feedback loop fundamentally alters the ability of a model to correct itself. Self correction
is an abstract concept. While feedback loops do enable some form of equilibration, they
are open to criticism due to the tenuousness of the criteria and objective functions used to
equilibrate. In other words, "equilibrate" and "correct" cannot be used interchangeably.
Such models may or may not be good models of observed behavior.
Likewise, use of large calibration adjustments may destroy the conceptual basis of a
model. The validity of calibration adjustments used to improve the match between model
predictions and observed base year flows cannot be checked once a model is used for
forecasting purposes (Atkins, 1986).
Some of the major calibration adjustments available in travel demand models are
discussed next.
The form of trip generation model equations and coefficients is normally model specific
and developed on the basis of local surveys.
Trip distribution models and calibration technique depend upon the model form chosen.
The most common form is the gravity model. This has two functions which can be
calibrated. Friction or accessibility factors (or functions) are used to relate the measured
travel time to a destination and the travel time perceived by a driver. The factors are
specified as a trip type-specific curve based on travel impedance. Calibration is
92004r2.02
16
-------
performed by developing a function(s) that uses existing networks and associated travel
times to replicate the distribution of trip lengths observed via survey data. Unfortunately,
there may be other influences on trip lengths and spatial distribution of trips besides
simple accessibility. These are represented by K-factors which supplement friction
factors. K-factors are developed by comparing the survey trip matrices to those
generated by a "calibrated" model. Typically, trial and error is used to obtain a better fit
between spatial distributions using the analyst's knowledge of the area to determine where
and why K-factors must be used. The major problem with K-factors is their longitudinal
instability. Replication of existing trip patterns using K-factor adjustments may not
properly represent future conditions due to the possible short term nature of the trip
pattern influences and aberrations.
Mode choice models are basically similar to trip generation models in terms of
calibration. The form and the coefficients are developed based on statistical analysis of
local surveys. The most common form is the logit model, which uses the exponential of
a mode-specific utility function to predict the proportion of trips served by a particular
mode. Models evaluate different statistical fits, whether some input variables can be
predicted, and the relative importance of variables in order to provide a workable tool for
policy analysis.
Various factoring processes to determine both the percent of trips occurring in a given
time period and the directional balance of trips often are applied after mode choice.
These are typically based on parallel survey distributions. Also, sometimes correction
factors are applied to achieve a better correspondence between total survey trips by
purpose and model trips by purpose. This corrects for such problems as underprediction
of non-home based trips due to deficiencies in land use data and/or trip generation. One
of the most useful numbers for checking the reasonableness of the model available to the
transportation analyst is total number of trips by purpose. This provides a check on
magnitude of trip generation.
If equilibrium assignment is used, the assignment methodology is self calibrating. The
user can control the relative weights of distance and travel time in the composite
impedance. However, both link speed and capacity are powerful tools for calibration
purposes. Speeds can have a major impact on trip distribution and the typically limited
sample of network speeds means that assumptions based on experience, area
characteristics, and known facility characteristics must be made for many of the links.
Inconsistencies in the size and characteristics of roads along the length of a single link
means that, typically, professional judgement is used to estimate a representative capacity
for the link.
A separate issue is that the degree of aggregation at which model results are presented for
comparison with observed base-year data may mask differences between synthesized and
observed data which are important from an emissions perspective, if not necessarily from
the perspective of transportation modeling. An example of this occurs when VMT
92004r2.02
17
-------
estimates are compared at selected monitoring points, or on a county level. 1 ^c .-curacy
of model predictions of speed and volume for the majority of links in the net'* OIK must
be assumed on the basis of this limited comparison. While this type of mode' validation
is satisfactory when travel demand models are exercised to represent the flow of rjffic
on major facilities, it will not necessarily indicate that the model is performing adequately
for minor roadways, which is important from an air quality perspective since travel on
minor roadways contributes significantly to regional emissions.
Model validation procedures are generally not standardized. Air quality modelers should
be aware that there are substantial uncertainties associated with emissions, air quality, and
transportation models. Transportation modeling is referred to as an art rather than as an
attempt to mathematically represent various aspects of behavioral science. Air quality
officials need to understand the inherent limitations within the transportation data they are
using; therefore it is important that the model validation procedure and results be
accessible and understandable to the air quality community. Documentation of
transportation modeling procedures must be written to be understood by those outside the
transportation community.
Another issue arises because a degree of user choice is allowed in selecting the network
for transportation models. For example, it js left to the modeler to determine the size
and definition of TAZs. Standard practice, in part due to limitations in the size of the
calibration data base, is to have small TAZs in urban areas, which allows more detailed
treatment of productions and attractions, and large TAZs in outlying areas, where the
preponderance of travel of interest to transportation modelers is likely to be on major
rather than local roadways. Use of large zones results in much of the traffic being
lumped together as intrazonal travel, which the transportation models do not represent
well. The degree to which zone definition and sizing can affect model results is
significant, but may not be communicated to those interpreting transportation model
results.
Related to this is the amount of simplification used to represent a roadway system for a
computer model. For example, though important in TCM analysis, high occupancy
vehicle (HOV) lanes may not be treated as separate facilities within a model. Ideally,
sensitivity exercises would be followed by a modeler in determining both the variation in
model results attributable to network definition and to determine as robust, i.e., least
likely to have large effects upon model output, network definition as possible. The
question of whether network definition should change for forecast years is often not
explored either; trending of baseline land use patterns are generally used for forecast
years (Deakin, 1991).
A final issue concerns the endpoints used in transportation models. It is not uncommon
for air quality modelers to have to interpolate between sets of model predictions to
develop estimates for the years of concern from an air quality j.- .rspective. For example,
regional transportation agencies forecast roadway use for different time periods than the
92004r2.02
18
-------
planning horizons used by air quality planners. Often, air quality planners require data
over shorter horizons, such as for three to five year intervals; transportation modelers
may focus on ten to twenty year projections that do not overlap air quality planning
milestones. Interpolation schemes for calculating intermediate transportation planning
estimates may not accurately reflect those future transportation conditions.
SUMMARY OF CONCERNS AND THEIR RELATIONSHIP
TO TRIP, VMT, AND SPEED ESTIMATION
As detailed above, there are discrepancies between the needs of air quality models and
the ability of transportation models to predict travel activity. These are related to the
algorithms used in a typical four-step travel demand modeling procedure, the quality of
input data used in the models, and the user choices which affect how the models are
applied. Table 2-4 summarizes the problems discussed, their effects upon both
transportation and emission estimates, and a general estimate of the potential significance
of each issue. Note that estimates of significance depend greatly upon both the
characteristics of the region studied as well as the goals of an air quality analysis. The
reader should keep in mind that uncertainties are found in the processes used for
emissions and air quality modeling as well -as in transportation modeling.
92004r2.02
19
-------
TABLE 2-4. Potential issues of concern in using transportation models for air quality analysis.
Issue
Transportation
Parameters Affected
Emission Parms
Affected
Bias
Direction
Potential
Significance*
Accuracy of land use forecasts,
or forecasts developed in
response to policy decisions
Spatial allocation; congestion levels;
VMT, trips, speed
Spatial and speed
distribution, VMT, trips
M
Roadway capacity restraints and
average daily traffic volumes
used to estimate speed
Overestimates of speed under
congested conditions
Speeds too high at peak
travel periods, too low for
off-peak
Probably
underestimate
emissions
H
Lack of feedback among the
four steps of the modeling
process
Inconsistent volumes and speed;
travel behavior not allowed to
change with congestion
VMT, trips, and speed
M
Intrazonal travel modeled with
less detail than link-based travel
No detail on speed; volumes
underestimated; limited spatial
resolution
YMT, trips, and speed
Underestimates
emissions
M
No hourly traffic estimates
Only daily or peak-period traffic
estimates provided
Temporal disaggregation
of emissions accomplished
through post-processing
Probably
underestimates
emissions
Land use and travel forecasts
are generally for long time
increments
Year-to-year traffic variation not
treated
Emissions may be needed
for much shorter time
increments
M
Model validation procedures
and goals are not standardized
No uncertainty estimate for
transportation results
No uncertainty estimate
for emissions
Continued
92004r2.03
-------
TABLE 2-4. Concluded
Issue
Transportation
Parameters Affected
Emission Farms
Affected
Bias
Direction
Potential
Significance*
Range of variability in model
inputs (e.g., speeds, network
definition) is not quantified
No seasonal or
weekday/weekend
effects
No uncertainty estimate for
transportation results
Transportation results only available
as annual averages
No uncertainty estimate
for emissions
Emissions are often
required for specific
episodes
Outdated survey data used
Introduces extrapolation errors;
recent changes in travel behavior not
treated
More potential errors in
VMT, trips, and speed
H
Use of calibration adjustments
to allow models to match
historical observed travel
patterns
Modeled region does not
correspond to air quality
analysis region
May invalidate model forecasts
Activity estimates are missing or
have little detail in outlying areas
Could introduce artificial-
'ity into resulting
emissions
Vehicle activity levels
must be extrapolated or
adjusted for outlying areas
M
Underestimates
emissions on
edges of
modeling region
M
* H = high, M = medium, L = low
92004r2.03
-------
3 HOW TO DIAGNOSE MODELING DIFFICULTIES
In the previous chapter, potential weaknesses in transportation models from an emissions
perspective were discussed. Where possible, the resulting effects of these weaknesses on
estimates of trips, VMT, and speeds input to emission models were indicated. This
information was summarized in Table 2-4. This chapter will assist the reader in
developing a list of questions which should be discussed between air quality and
transportation analysts.
KEY QUESTIONS TO ASK YOUR TRANSPORTATION MODELING AGENCY
The development of vehicle emission inventories suitable for air quality modeling
purposes is a cooperative effort between air quality and transportation agencies. In order
for the effort to succeed, this cooperation must extend throughout the transportation and
air quality modeling process. In the following section we will provide a series of
questions which should be discussed between air quality and transportation modelers. It
is important that transportation modelers understand what air quality personnel require in
transportation modeling outputs and that air quality staff understand the limitations of
transportation models from an air quality analysis perspective.
Although ideally these discussions would take place as a transportation model is
developed for an area, the reality is that many areas already have transportation models
in place. Hence, these questions serve as tools air quality modelers can use to develop a
deeper understanding of an existing transportation model, or as issues to be discussed
during the update of a preexisting transportation model.
For example, an air quality modeler may intend to use the transportation modeling
outputs to develop a seasonal emission inventory. In many regions, there are significant
seasonal variations in traffic patterns. An agricultural region would experience increased
commercial traffic (i.e., heavy-duty diesels) associated with the transport of crops during
harvest times. Areas near recreational attractions, such as skiing, would have increased
automobile traffic during the winter months. There might be particular events specific to
the year an air quality modeler is interested in, such as freeway repair, which would have
significant short-term impacts on traffic patterns. At the same time, a transportation
modeler needs to communicate to air quality personnel the necessary assumptions
92004r2.05
22
-------
involved with transportation models, and features of the models which might be viewed
as limitations from an air quality perspective. For example, transportation models are not
designed for the detailed representation of intrazonai traffic, and specific assumptions are
required of a modeler in order to provide VMT and speed estimates for this type of
travel.
The questions discussed below are loosely grouped by the typical four steps of
transportation modeling which they concern. These are trip generation, trip distribution,
mode choice, and highway assignment. The characteristics of these four steps have been
discussed in general terms in Section 2. More detail will be given on each stage in the
following pages in order to show the context of these questions. Some of the suggested
questions affect more than one of the transportation modeling stages. In this case, we
discuss the question at the earliest stage of the transportation modeling process.
TRIP GENERATION
Trip generation is where socio-economic land use data and projections are used to
estimate the number of productions (trip origins) and attractions (trip destinations)
associated with each TAZ. As discussed in Section 2, some of the limitations of
transportation models, from an emissions perspective, arise from the land use data and
projections used at this stage of the modeling process. A transportation modeler must use
these data, which provide such information as the numbers of homes, shopping centers,
businesses, or distribution centers in a TAZ, along with estimates of the numbers of trips
produced and attracted by each of these land uses, to calculate total trip productions and
attractions for each TAZ.
There are several areas of discussion for transportation and air quality modelers at this
stage of the process. These concern.the following issues:
Land use and model calibration data;
Generation of productions and attractions in TAZs;
Trip types;
Structure of traffic analysis zones.
The significance of these issues varies depending upon the characteristics of the area
under study as well as the goal of the study. All have the potential to be very significant,
particularly in studies evaluating the comparative impacts of different emissions control
strategies, or in studies using photochemical modeling tools.
92004r2.05
23
-------
Land Use and Model Calibration Data
Air quality personnel should determine the age and source of the land use and model
calibration data used. Although the distribution of land uses responsible for most of the
home-based-work (HBW) trips that are the main interest of transportation modelers may
be relatively unchanged from when the data were gathered, shifts from urban to suburban
populations or in the income levels in zones can have significant effects on the number of
trips produced and attracted by each zone and may not be represented in older land use
data. As mentioned in Chapter 1, some regional models are based on outdated survey
data. Resources are frequently not sufficient to permit updating of these surveys with
their original level of detail. Air quality modelers need to be aware of the limitations of
the surveys that will be used to develop a transportation model, and the methods chosen
to update them. Such understanding can indicate where further refinement of the data is
required.
Land use forecasts to future years are a source of uncertainty in transportation modeling.
In the past the same spatial distribution of productions and attractions was sometimes
used for both a base and future year, and the number scaled based upon a set of
assumptions. In other cases policy decisions may have unreasonably influenced land use
projections. The development of land use forecasts should be an issue of concern for air
quality modelers, because the spatial distribution of emissions can significantly affect
production of ozone and carbon monoxide. Assumptions about future density limits and
zoning are important, especially in evaluating denser, transit-rich development
alternatives aimed at improving air quality. The methods used to predict the future land
use or development distributions need to be discussed between air quality and
transportation modelers to ensure that they are appropriate for emissions modeling.
Alternative development scenarios should be considered for possible regional plans to
reduce emissions.
Generation of Productions and Attractions in TAZs
Finally, the methodology used to estimate the number of productions and attractions in
each TAZ should be discussed among air quality and transportation modelers. This step
of the process results in a forecast, different between base and future years, of the
number of productions and attractions in each TAZ. Different estimation methods may
result in different strengths and weaknesses in the resulting modeling results. Some
methods that have been used for this include optimal adaption to census data, home
interviews, measurements of distances, traffic censuses, etc. (Hamerslag, 1980). The air
quality modeler should be aware of what parameters drive the trip generation models
(e.g., household or employment characteristics), and whether an attempt is made in the
models to relate travel demand to the transportation system capacity.
92004r2.05
24
-------
Trip Types
The socio-economic characteristics of a TAZ (e.g., number of households or employees)
will determine both the number and the types of trips which it produces and attracts.
Although a transportation modeler can choose to explicitly represent any set of trips with
a common purpose within the model, commonly used trip types are: home-based work,
home-based school, home-based shopping, home-based social/recreational, home-based
other, non-home-based trips, through trips (i.e., which neither begin nor end within the
modeling region), and trips which originate outside the modeling region but end within it
and vice-versa.
The types of trips a transportation modeler explicitly treats are of importance to the air
quality modeler. For example, there is generally little survey data available on
commercial travel, i.e., heavy duty diesels, yet this is the source of a significant amount
of emissions. Transportation modelers may arrive at VMT, trip, and speed estimates for
this component of regional travel through postprocessing steps, or by setting it up as an
explicit trip type in the model. Either method will require a series of assumptions as to
volume and distribution of commercial travel. These should be discussed, and the
method chosen to treat commercial traffic and any other trip type not directly modeled
understood by the air quality modeler.
Transportation models are best at treating home based work trips, in part because a major
source of information on these trips is available for all regions from data collected by the
U.S. Census. Frequently trips for shopping or recreational purposes are treated as a
general category, such as home-based other. The air quality modeler should determine
exactly what trip types are contained in these general categories, and work with the
transportation modeler to ensure that all trip types are included in the model. The level
of detail at which specific trip types should be treated depends to a large extent on its
anticipated contribution to regional emissions. To return to our example of commercial
travel, heavy duty trucks are a significant source of NOX emissions, particularly in future
years, yet no widely used models of goods movement processes are available. Given the
available data their treatment in transportation modeling should be as detailed as possible
and the limitations in the model predictions for this vehicle class understood by the air
quality modeler.
Structure of Traffic Analysis Zones
The air quality and transportation modeler should discuss the methodology followed in
defining TAZs. In particular, the region modeled with the transportation model should
match the region of interest for air quality analysis. Outlying areas can serve as
important sources of emissions which lead to exceedances of air quality standards. These
outlying areas may either be modeled with less detail or not modeled at all bv
92004r2.05
25
-------
transportation models even though they are significant from an air quality perspective.
Similarly, it is often not enough to include only a nonattainment area in the transportation
model, since the regions that are modeled in air quality models are usually larger than the
nonattainment area in order to include pollutant emissions that can be transported (by
wind) into the nonattainment area during the period of time modeled (typically one or
more days). Therefore, when first setting up or when updating a transportation model,
efforts should be made to ensure that it will cover a reasonable amount of the area
contributing to the degradation of air quality in a nonattainment area. In the case of
preexisting models that fall short of this goal, transportation modelers may be able to
suggest alternate sources of information on travel activity that can be used to develop
emission estimates for the outlying areas.
The level of detail used in modeling a region and the variation in this level of detail as
the network is defined for the entire region of interest should also be discussed. Since
the air quality perspective as to what constitutes important travel activity is not always the
same as the transportation modeling perspective, this communication will help to ensure
that travel activity that contributes disproportionately to air quality problems is adequately
modeled.
The next section will discuss questions which should be raised by an air quality modeler
concerning aspects of the trip distribution portion of the transportation modeling process.
TRIP DISTRIBUTION
Trip distribution takes the number of productions and attractions associated with each
TAZ, as determined in the trip generation stage of this process, and predicts how traffic
will be distributed between TAZs. Key parameters at this stage of the process are
friction factors, travel time, and K-factors, which indicate the rektive likelihood of travel
involving different levels of time, distance, or cost occurring between TAZs.
Two key areas of discussion for transportation and air quality modelers occur at this stage
of the process. The first is the development and use of friction factors and K-factors for
the region. The second concerns representation of intrazonal travel.
Development of Friction Factors and K-Factors
Friction factors (often functions) are generally developed from local survey data for each
trip type in the model. People's willingness to invest time or money in a trip depends
largely on the reason for the travel. For example, one would likely travel farther hi
order to reach one's pkce of employment than one would travel to buy groceries. K-
factors, on the other hand, perturb the trip distribution that would otherwise result. Take
the case of Chicago, for example. There a central business district is surrounded on the
v
92004r2.05
26
-------
west and south sides by low-income areas. Further west are affluent suburbs. K-factors
that diminish the effect of the friction factors are needed to bring the suburban traffic into
the central business district. At this point, the air quality modeler should discuss with the
transportation modeler the methodology and assumptions used to determine friction and
K-factors, paying particular attention to assumptions used to determine differences in
travel behavior by trip types. The error range of the friction factors and limitations
which may result from local surveys used to develop them, or from the representativeness
of the sample size and sampling methods, should be understood by the air quality
modeler. This can help to determine whether old surveys should be updated with spot
sampling techniques, or whether further survey information is required. K-factors, which
are developed to adjust travel behavior to more closely resemble current conditions, may
not be appropriate when forecasting travel activity. Hence, the effect of such factors on
travel activity estimates should be understood to determine if forecast travel activity is
unduly affected by the application of K-factors.
Intrazonal Travel
It is during the trip distribution stage of transportation modeling that the amount of
intrazonal travel is determined. At this stage, only the number of intrazonal trips in each
TAZ is estimated. The transportation modeler must develop a set of assumptions to
assign speed and VMT to these trips and to determine the fraction which is non-vehicle
trips (e.g., walk/bike trips). Since this component of regional travel is important from an
emissions rather than from a transportation modeling perspective, it is important that the
air quality modeler discuss with the transportation modeler how intrazonal travel can be
treated. Assumptions applied to this type of travel should be stated. Additional insight
into intrazonal travel might be obtained by examining the procedures used to estimate
travel on local facilities for the Federal Highway Administration's Highway Performance
Monitoring System (HPMS). As stated in the EPA's Interim Guidance for the
Preparation of Mobile Source Emission Inventories (Lorang, 1991), though local road
VMT estimates are generally not based on the statistical procedures used for other facility
types, HPMS reports do contain VMT estimates for local roads.
MODE CHOICE
Mode choice is an optional stage of the transportation modeling process; however, some
form of transit estimation is done in all areas. Mode choice models are used to
determine the proportion of trips to be modeled as public transit, which includes such
travel modes as buses or light rail systems. A separate computer simulation of the
roadway network is generally developed for public transit, and is in some ways simpler
than that used for other travel modes. For example, transportation models usually
assume uniform spacing between bus arrivals. (An exception is the VTS Interactive
Planning System (VIPS), developed by VTS Systems Corporation of Sweden, which uses
92004r2.05
27
-------
a probabilistic estimation of bus arrivals.) Few models treat multi-path public transit.
However, public transit models are generally based upon very explicit route information.
Mode choice is of particular importance from an air quality perspective if transportation
control measures or improved public transit systems are to be evaluated as tools for
improving air quality.
Areas of discussion at this stage of the transportation modeling are:
Treatment of public transit, i.e., whether it should be modeled separately from
other travel modes and what types of public transit scenarios should be evaluated;
Determination of public transit travel, i.e., how the proportion of vehicle trips
diverted to public transit is determined or how vehicle trips to central transit
facilities, such as park-and-ride lots, are treated;
Urban bus VMT.
Treatment of Public Transit
The decisions as to whether public transit should be modeled separately from other travel
and what types of transit scenarios should be evaluated depend upon the overall purpose
of the air quality modeling. If control measures that include public transit measures are
being evaluated, public transit should generally be modeled separately from other travel.
Specific scenarios relating to public transit issues that an air quality modeler intends to
model should be communicated to the transportation modeler, since scenario definition
will affect assumptions made at this stage of the modeling process. An example is an air
quality improvement measure meant to increase the cost of using a personal automobile
for commuting to work, in order to encourage drivers to switch to public transit.
Since cost is an important factor in an individual's decision to use public transit systems,
or to choose one public transit option over another, air quality modelers should ask how
cost is included in the public transit portion of the transportation modeling process. They
should also be aware of how income data was incorporated into the original land use
data, since income level correlates with public transit use, and how accessibility to public
transit is addressed.
Determination of Public Transit Travel
When mode choice models are used, air quality modelers need to be aware of what
assumptions were used to assign riders to the public transit network. The breakdown of
public transit travel into components such as buses or light-rail systems should be
understood. Survey data are often used to establish the amount of public transit travel.
92004r2.0S
28
-------
Spot surveys are difficult to develop for public transit use, since public transit represents
a small portion of overall travel. As a result, mode choice models are often the most
difficult to calibrate. Air quality modelers need to know the limitations and error ranges
associated with public transit surveys, in order to correctly interpret the results of
modeling exercises.
Air quality and transportation modelers should discuss how trips to public transit facilities
such as park and ride lots or subway stations are treated. Although these trips may not
be important from a transportation perspective, and are often intrazonal trips, they may
still be a significant source of vehicle emissions. This is because emissions from vehicles
that have not "warmed up" are higher than emissions from vehicles that have been in
operation for a few minutes, i.e., although the VMT associated with trips to public transit
centers may be small, the emission rates for this type of travel are often much higher
than vehicle emission rates during standard commutes. Another related topic of
discussion is whether any attempt is made to model increased vehicle usage from people
driving passengers to public transit centers. Historically, within transportation models the
transit network has not been integrated with the overall roadway network, which
increases the difficulty of modeling all travel effects arising from use of public transit.
One approach to improving transportation models is to maintain correspondence between
the two networks.
Urban Bus VMT
Air quality modelers involved in TCM analysis should know the proportion of VMT due
to urban bus travel. The emission rates of buses are significantly higher than those of
passenger cars, and this characteristic should be modeled in producing emission
inventories. This is important when evaluating the effects of TCMs which propose
increased transit use. Urban buses are also likely to be targeted for alternative fuel use,
since they are frequently fueled from a central fleet location. In such situations, an air
quality modeler needs to know the amount and location of travel associated with these
vehicles. Although transportation models may not readily lend themselves to tracking
urban bus travel, discussion of these needs with the transportation modeler and
understanding of the assumptions followed in modeling urban bus travel will certainly
improve representation of these emissions. Note that integration of the highway and
transit networks is one tool available in transportation modeling to improve estimates of
transit use. This technique is demonstrated in Section 4.
The next section discusses questions which should be posed by an air quality modeler
which concern the trip assignment stage of the transportation modeling process.
92004r2.05
29
-------
TRIP ASSIGNMENT
Trip assignment is where the actual routes taken for trips between zones are determined,
generally based upon the path which has the shortest travel time (alternatively distance or
cost can be used). There are various methods available for this step of the process. One
of the most common, currently, is for the model to try different routings of trips between
origin and destination pairs until a user-specified equilibrium is attained. This is an
iterative approach which recalculates trip assignments based upon the systemwide
congestion on links. Equilibrium in this case is defined as the point at which there is no
alternative way to assign a trip without producing a net increase in system-wide travel
time.
Link distance and estimated speeds are inputs to this process. Output speeds are then
calculated for each link, often based on volume to capacity relationships. Different levels
of feedback allow a model to achieve some stability in the final speed estimates. An
heuristic approach can also be followed in modeling network speeds, where the user
defines program exit criteria based on what is considered to be the optimum number of
iterations.
Questions which should be asked of the transportation modeler at this stage concern:
Choice of travel route;
Choice of initial speeds;
Assignment methodology;
Use of feedback;
Temporal variation in traffic.
Choice of Travel Route
Air quality modelers should ascertain the criteria for assigning the path taken between
zones. They should understand the conversion of cost factors, such as bridge tolls, to
distance or time increments on the affected link within the model. If certain types of
travel are restricted on specific links in the region, or if some links are more likely to
draw a specific type of travel even though they may not represent the shortest distance
between points, the way in which this is represented in the model, or if it is represented,
should be discussed. Examples of such situations are highways on which heavy-duty
truck travel is prohibited, or specific highways that draw a disproportionate amount of the
heavy-duty truck travel in a region. Diversion of certain types of travel, such as
commercial, to specific routes may affect the vehicle emission factors that should be used
to calculate emissions for these links.
92004r2.05
30
-------
Choice of Initial Speeds
Initial estimates of speeds for each link in the highway and transit networks may be taken
from actual surveys or be assigned on the basis of the road (i.e., facility) and area type of
each link. The air quality modeler should ask how initial speeds were determined, since
these can affect both how trips are distributed in a region and the final speed estimates
produced by the model. Possible approaches include starting the model with congested
speeds or speeds representative of the time of day modeled (e.g., off peak, morning
peak), or starting the model with freeflow speeds so that the model is allowed to develop
the congested speeds. Further research needs to be conducted to resolve this question.
Lack of local survey information on speeds may force reliance upon standard reference
manuals, such as the Highway Capacity Manual (Transportation Research Board, 1985)
to assign initial speeds based on area and roadway type. Error in model outputs due to
the choice of initial speed can be bracketed by the transportation modeler, as discussed in
Section 4, which would help in determining the importance of the initial speed estimates.
Assignment Methodology
The air quality modeler should be aware of what methodology was used in the trip
assignment model, and how a specific method may affect model output compared to a
different method. The weaknesses and strengths of the methodology chosen, in terms of
the spatial distribution of trips and VMT, and speed estimates, should be known. One
typical assignment method used is free-flow, whereby all selected trips are located on the
minimum paths (based on time, distance, cost, or user impedances) of the network.
Another method is restraint loading, which is similar except that the network parameter
time is adjusted link by link according to a capacity restraint formula. Still another
method is incremental loading, where for each iteration a user-specified percentage of
selected trips is loaded on the minimum paths determined during path building (the
network parameter, time, is adjusted link by link according to a capacity restraint
formula). The air quality modeler should discuss with the transportation modeler the
reasons for the methodological choice. Important assumptions are made at this stage of
the modeling process which will have significant effects upon the final VMT and speed
inputs to the emissions models, as well as the final spatial distribution of the vehicle
emissions. For. example, curves which are used to relate traffic delay and congestion can
either represent relative or absolute delays, a distinction which can affect how well model
treatment of traffic delay matches real world conditions.
Use of Feedback
Transportation models can use some type of feedback to refine their initial estimates of
VMT, trips, and speed in an attempt to reflect real-world conditions. The speeds
produced by the trip assignment model may significantly affect the travel times between
92004r2.05
31
-------
specific zones, making it more likely that travel will be routed to another zone than had
been predicted in the first pass through the trip distribution stage. Rerunning portions of
the model with these new predicted speeds can, in some cases, improve the model
predictions when coupled with knowledge of how congestion impacts destination choice.
Some models will even go back to the original trip generation stage, using model
predictions to modify the land use data. More areas are considering using feedback in
their transportation models, in part in response to a lawsuit filed by the Siena Club and
Citizens for a Better Environment against the San Francisco Bay Area's Metropolitan
Transportation Commission (MTC) and associated parties challenging some elements of
the Bay Area's 1982 State Implementation Plan. The MTC's MTCFCAST models have
feedback capabilities that, in the past, were rarely employed in practice by MTC staff.
This practice was one source of criticism of MTC in that lawsuit. An air quality modeler
should be aware of the amount of feedback, if any, employed in running the model, and
what criteria were used to determine when the final set of model outputs had been
reached.
Temporal Variation in Traffic
Air quality modelers often require hourly emission estimates. The usual practice is to
process the daily or peak and offpeak period estimates of travel activity available from
transportation models to arrive at hourly estimates. Transportation modelers typically
break daily average traffic estimates into estimates for peak and offpeak periods between
the mode choice and trip assignment steps in the typical four-step model. Working
together, transportation and air quality modelers can decide the level of breakdown
required given the anticipated use of the transportation model outputs as well as the data
available. For example, modeling travel activity separately for offpeak, morning,
midday, and afternoon peak periods is in most cases sufficient for developing defensible
estimates of hourly vehicle emission inventories. Alternatively, daily travel activity
estimates from a transportation model can be disaggregated into hourly activity estimates
as a postprocessing step to the transportation modeling. Through analysis of regional
survey data, transportation modelers are uniquely qualified to provide advice on
developing hourly activity profiles specific to the area of study.
Once the number of time periods to be modeled is decided, the temporal variation in
zonal characteristics must be determined. An air quality modeler should determine if
regional factors will be used to allocate traffic to specific time periods, or if subregional
or even zonal factors will be used. TAZs within a region will not be active at a uniform
level for each time period modeled. As an extreme example, zones with a high
concentration of offices might be almost deserted during off-peak hours, whereas zones
with recreational attractions could be at their most active during these times.
The same issues must be considered when transportation outputs are to be used to model
weekend episodes. One would expect less temporal variation in traffic patterns on
92004r2.05
32
-------
weekends. However, the number of productions and attractions associated with each
TAZ could be different than that calculated for weekdays. Business districts might be
relatively inactive, whereas recreational areas would be the primary source of attractions,
and activity might be more uniform throughout the day. Although it is uncommon to
model weekend travel with transportation models, since most areas lack sufficient data,
adjustments for estimating weekend travel activity from transportation model outputs
should be discussed when air quality episodes include weekends.
SUMMARY OF KEY QUESTIONS
The key issues of discussion between air quality and transportation modelers are
summarized in Table 3-1, grouped by the four transportation modeling steps. Current
practice as well as suggested improvements are discussed when possible.
At this time it is difficult to estimate the potential significance of all of these issues,
especially since the significance will vary depending upon the region studied and the
purpose of the study. Many air quality analyses compare emissions reduction strategies
which are very similar in terms of magnitude of emissions reduced, but which target
emission sources with different photochemical reactivity or which occur at different times
of day. The resulting impact of such strategies on air quality is nonlinear in nature.
Studies have found that simply shifting the time of occurrence of emissions by an hour
will have a significant effect upon the buildup of pollutants in an urban area (Ireson et
al., 1987). Rigorous quantification of the significance and the cost effectiveness of
addressing the issues listed here is outside the scope of this study, but may be addressed
in subsequent studies carried out by the EPA. In general terms, temporal variation, the
accuracy of land use data and projections, inclusion of intrazonal travel, methods used for
predicting speeds, and the use of feedback within the model are most significant among
the issues listed in Table 3-1.
92004r2.05
33
-------
TABLE 3-la. Summary of key questions: trip generation.
Area of Discussion
Question
Potential Effects and S
Temporal Variation
Land Use Data
Trip Types
Determination of
Productions and Attractions
What are the time
periods/days of
week being
modeled?
What is the vintage
of the land use data
being used by the .
transportation
modeler?
What types of trips
are or are not
normally treated by
the transportation
modeler?
How are estimates
of the non-treated
trips' activity
developed?
What methodology
is being used by the
transportation
modeler to generate
productions and
attractions in TAZs?
Current practice is generally to no*...-l
weekday travel for a peak and off-peak
period. Results are reported as daily totals.
Air quality (AQ) modelers need hourly
emissions estimates for both weekday and
weekend travel.
Division of modeling periods into AM peak,
PM peak, off-peak and between peak
periods to better characterize different travel
demand characteristics. Using transportation
models for weekend episodes should also be
discussed.
Changes in urban and suburban populations
can have significant effects on spatial
distribution of emissions.
Older surveys may not reflect population
distribution changes for the periods of
interest to air quality modelers.
Transportation models are most widely used
to characterize home-based work trips.
Non-work trips are often placed in a "home-
based other" category.
Lack of survey data for commercial traffic
has necessitated extensive assumption-
making when determining VMT, trip, and
speed estimates for this trip type. Different
modelers may have different methods of
handling this problem.
The contribution of commercial travel
(heavy duty trucks) to emissions is
significant. Air quality modelers must be
aware of the limitations of the transportation
models in handling this trip type.
Different trip generation methodologies may
result in varying strengths and weaknesses in
transportation modeling results.
920O4r2.06
34
-------
TABLE 3-1 b. Trip distribution.
Area of Discussion
Question
Potential Effects and Solutions
Development of Friction
and K-factors
What is the
methodology used to
determine friction
and K-factors?
Intrazonal Travel
What assumptions
does the
transportation
modeler make when
determining speed
and VMT for these"
trips?
AQ modelers need to discuss with
transportation modelers the assumptions used
to determine differences in travel behavior
by trip types. The error range of these
assumptions should be fully understood by
AQ modelers.
If AQ modeling is to be conducted for a
particular season or episode, conditions
which could affect the decision to travel
should be considered in friction and K-factor
development.
Only the number of intrazonal trips is
calculated by the transportation model. The
transportation modeler must make
assumptions to assign VMT and speeds, and
to determine die fraction of non-vehicle
trips.
VMT and speeds are much more important
from an emissions perspective; AQ modelers
should discuss how intrazonal travel will be
treated with die transportation modeler
before the process begins. Comparisons of
intrazonal travel estimates and assumptions
with procedures used for the HPMS.
92004r2.06
35
-------
TABLE 3-lc. Mode choice.
Area of Discussion
Question
Potential Effects and Solutions
Treatment of Transit
Determination of Transit
Travel
Urban Bus VMT
Seasonal or Episodic Effects
What scenarios does
the transportation
modeler intend to
model? How does
cost figure into these
scenarios?
What assumptions
are used to assign
riders to the transit
network?
How are trips to -
transit facilities (e.g.
park and ride lots)
treated by the
transportation
model?
What is the amount
and location of
travel associated
with these vehicles?
How can
transportation
models be used to
model a specific
season or episode?
Which transit scenarios are to be modeled
depends heavily upon the overall purpose of
the AQ modeling. EXAMPLE: If control
measures (e.g. transit) are to be modeled,
transit should be treated separately but not in
isolation from other modes.
Income level correlates with transit use; AQ
modelers should understand how this
information is incorporated into the
transportation model.
AQ modelers need to understand how
transit travel is broken down into
components (e.g. ridesharing, buses).
If survey data are being used, AQ personnel
should understand the limitations and error
ranges associated with these surveys.
Trips to transit centers, though not of great
importance from a transportation modeling
perspective, are very important from an
emissions perspective. AQ and
transportation modelers need to discuss
methods of handling this.
Emissions rates of these vehicles tend to be
much higher than those of light-duty
vehicles. AQ modelers need to understand
the assumptions made hi modeling this mode
to better represent its contribution to
emissions.
Transportation models are typically
developed for annual average applications;
however, more specific time periods could
be modeled.
Discussions between AQ and transportation
modelers concerning the characteristics of
these seasons/episodes are necessary if they
are to be modeled.
92004r2.06
36
-------
TABLE 3-Id. Trip assignment.
Area of Discussion
Question
Potential Effects and Solutions
Choice of Travel Route
What criteria are
used for determining
the minimum path
between zones?
How are cost factors
converted to distance
or time?
Choice of Initial Speeds
How are initial
speeds determined?
Assignment Methodologies
What methodology
was used in the trip
assignment model?
How can different
methodologies affect
model output?
Certain types of travel may be restricted
from using some routes within a network.
Also, specific links may be more likely to
draw certain types of travel, even though
use of a particular link may not represent
the shortest distance.
Diversion of certain types of travel (e.g.
heavy-duty truck traffic) may affect the
vehicle emission factors used to estimate
emissions on these links.
Speed estimates for links may be taken from
actual survey data or be assigned on the
basis of facility (e.g., arterial) and area type
(e.g., urban) of the link.
AQ modelers should ask how these estimates
are made since they can affect trip
distribution within the region and final
speed estimates produced by the model.
Lack of local data may result in the
transportation modeler relying upon
reference manuals for default values. AQ
modelers need to understand how this may
limit speed estimates and the error ranges of
these estimates.
Different methodologies used in trip
assignment have varying strengths and
weaknesses, in terms of trip distribution,
speeds, and VMT.
AQ modelers should discuss with
transportation modelers the reasons for
choosing one method over another.
AQ modelers should also discuss any
assumptions that may have been made by
the transportation modeler at this phase of
the process.
continued
92004r2.06
37
-------
TABLE 3-Id. Concluded.
Area of Discussion Question Potential Effects and Solutions
Use of Feedback What are the types AQ modelers need to understand the types
and amounts of of feedback being used in the transportation
feedback being used model, and the effects of this feedback on
by the transportation model outputs.
modeler?
AQ modelers should also understand the
How are VMT and criteria used to determine when the "final"
speeds calibrated set of model outputs have been reached
with the actual case, (equilibrium is an example of a common
i.e., how well does criteria).
the model predict?
38
92004r2.06
-------
AIR AGENCY STAFF GUIDANCE: SUGGESTED PROCEDURES
FOR IMPROVING THE LINK BETWEEN TRANSPORTATION
AND AIR QUALITY MODELS
OVERVIEW
This chapter discusses a range of procedures which could be used to improve the linkage
between transportation and air quality models. The ease with which these suggested
improvements can be applied using current transportation modeling systems is discussed
below. Sample model results for four of these options, developed using a prototypical
model, are provided and discussed.
Future EPA studies may be carried out to quantify the effectiveness of these suggestions
in terms of cost and impact. The complexity of transportation and air quality modeling
techniques make it difficult to estimate effectiveness without implementing these
suggestions with actual transportation models for a variety of urban areas and evaluating
their relative impacts for different scenarios. Some of these suggestions are simple and
inexpensive to implement and result in travel activity estimates which are better matched
to the needs of air quality modelers (this is the case for some of the postprocessing
techniques that are described), and therefore should more readily be considered when
using transportation models in air quality analyses. Obviously, some of the techniques,
such as those involving detailed treatment of transit, are useful only in a few situations,
such as in the evaluation of transportation control measures.
SUGGESTED PROCEDURES FOR IMPROVING
TRANSPORTATION MODELING
The suggested procedures for improving the link between transportation and air quality
models have been grouped into five general classes,(the procedures are summarized in
Table 4-1):
1. General Improvement to Travel Demand Forecasting. This includes
procedures that use available tools and information, such as reestimation of
origin and destination information using observed traffic volumes, but
which in general are more sophisticated and would require substantial
effort to implement.
92004r2.07
39
-------
TAHI li 4-1 Summary of options for improving transportation modeling.
General Improvements to Travel
Demand Forecasting
(A)
General Procedural
Improvements
(B)
Factoring and Cross-
Classification Processes
(C)
Intractable Problems &
Statistical Limitations
(D)
Modifications to Models to
Specifically Address Air
Quality
(E)
A-1 Intersection delay modeling
A 2 Weekend and episodic
modeling
A-.) Integration of transit and
highway networks
A 4 Analytical land use
forecasting processes
(DRAM/I-MPAL)
^ A 5 Consideration of congestion
0 in choosing destinations
A 6 Belter "peak spreading"
characterization
A 7 Improve HOV models
A X Rccslimale O&D
information using observed
traffic volumes
B-l Bracketing (all
parameters)
C-l Relating fleet mix to
network characteristics
D-l Need to bracket statistical E-l Estimation of emission
B-2 Conduct speed surveys to C-2 Match fleet to trip type
better validate models
C-3 Time-specific fleet
distribution
C-4 Diurnal speed profiles
C-5 "Correcting" speeds
based on V/C ratio
(TRFCONV)]
C-6 Deal with off-network
speeds/VMT, based on
socioeconomic data
C-7 Cross-classification of
parking duration by
zone/trip ends
limitations
D-2 Transportation control
measures (TCMs)
for transit using public
transit models (also
estimation of transit
access mode); need to
consider non-revenue
transit trips
E-2 Further disaggregalion
of zones
(rural/suburban areas)
Speed inputs based on
E-3 cross-classification of
network characteristics
A 9 Queuing models for toll
stations
A 10 Integration with GIS
databases
<»20(Mi I .OH
-------
2. Philosophical Perspective (General Procedural Improvements). These are
procedures that would generally be simple to implement, although they
might require additional survey data, as would be the case if models were
calibrated against observed speeds as well as volumes.
3. Factoring and Cross-Classification Processes (Post-Processing Data into
Look-Up Tables). These procedures are relatively simple to implement,
and the implementation process is relatively uniform for each suggested
improvement. These are generally post-processing steps which would
allow air quality modelers to have more detailed information available for
producing emission inventories. Many of these suggestions can only be
demonstrated in the prototype testing mode, since they rely upon survey
data which, while generally simple to gather, is available for few existing
models.
4. Intractable Problems and Statistical Limitations. Rather than procedures,
these are problems which, due to statistical limitations of data used in
developing these models, cannot be addressed with the current generation
of transportation models.
5. Reworking Transportation Models to Specifically Address Air Quality
Issues. These procedures would require receding of existing transportation
models so that they would directly provide more of the information which
is required by air quality modelers in developing emission estimates. An
example of such a procedure would be coding models to report link-
specific transit vehicle travel.
General Improvement to Travel Demand Forecasting
Ten optional procedures have been suggested which fall within the category of providing
general improvement to travel demand forecasting using available tools which, due to
their complexity, are often not utilized in transforation models. For reference, these
procedures have been labeled in Table 4-1 as A-l through A-10.
Option A-l, Intersection Delay Modeling. This type of modeling would require
incorporation of signalization models, such as FHWA's TRAF-NETSIM, into the overall
transportation model, in order to capture the effects of intersection queuing on overall
travel times and link impedances. Although a variety of signalization models are
available, it would be a complex effort to actually merge the two types of models. Some
research has currently been proposed in this direction by the California Air Resources
Board, which is initiating an effort to study the feasibility of incorporating signalization
models into transportation models.
92004r2.07
41
-------
Option A-2, Weekend and Episodic Modeling. This type of modeling requires no new
modeling tools but would require expansion of origin and destination surveys and ground
count and/or speed calibration data in order to gather data pertinent to weekends,
seasons, and/or particular episodes. What is envisioned here is the gathering of data
sufficient to demonstrate the statistical accuracy of such a model. The result of this effort
would be model outputs more representative of the typical travel activity a region might
experience during a particular season or day of week or time of day. This would be of
most use in areas which attract seasonal recreational activity, such as skiing or tourism.
Another practical result is gained if models specific to weekend and weekday travel
activity can be developed, since air quality episodes are not limited to weekdays and
photochemical (ozone) models require hourly inputs.
Option A-3, Integration of Transit and Highway Networks. This can be achieved
using existing models and is not resource intensive. It allows modeling the effects of
congestion on transit travel times, and by extension on mode choice. This is one of the
strategies demonstrated in this project using the TRANPLAN program INET (adapted
from UTPS software), and is discussed in more detail later in this chapter.
Option A-4, Analytical Land-Use Forecasting Processes. This is still a resource-
intensive process. Models such as DRAM/EMPAL have been developed for this
procedure, which refine the landuse forecasts used to develop regional transportation
models. At present, these models are still relatively resource intensive to use and
difficult to calibrate (past growth patterns are used for calibration).
Option A-5, Consideration of Congestion in Choosing Destinations. This is a
feedback step which allows modeling the impacts of longer travel times on destination
choice. It is relatively simple to implement. Trips for selected trip purposes are
redistributed after the highway assignment process based on congested travel times. This
is one of the strategies demonstrated in this project, and is discussed in more detail later
in this chapter.
Option A-6, Better Characterization of "Peak Spreading". This technique is used to
overcome the problem of predicted volume to capacity ratios exceeding one. Some
simple procedures using adjustments based upon peak period volume to capacity ratios by
link were developed to attempt to reduce this problem as part of research conducted by
the Arizona Department of Transportation (Loudon et al., 1988), but research still
continues to develop more accurate methods to represent the phenomenon.
Option A-7, Improved HOV Models. Improving HOV models is an ongoing topic of
research in the transportation modeling community. This becomes more important as
regions are forced to evaluate TCM packages as part of their response to new legislative
requirements, such as the CAAA. Improving models to include the effects of other
CAAA-mandated TCMs is also a matter of continuing research.
92004r2.07
42
-------
Option A-8, Reestimation of Origin and Destination (O-D) Matrices Using Observed
Traffic Volumes. A variety of different techniques using different combinations of
traffic and transport data are currently in use for creating O-D matrices. Although
caution must be exercised when using observed traffic volumes for this purpose, due to
the stochastic nature of the observations, this method has been successfully applied
(Hammerslag, 1980).
Option A-9, Using Queuing Models for Toll Stations. Use of queuing models is
similar in complexity to the use of intersection delay models.
Option A-10, Integration with GIS Data bases. This can be particularly useful for
incorporating information from various data bases, such as the U.S. Census TIGER files,
into land-use characterizations. For many agencies, this option is still unworkable due to
the resources required to effectively use GIS.
General Procedural Improvements
Two procedures have been suggested which fall within the category of being relatively
simple procedures which may, however, require more extensive survey data than is
usually available. For convenient reference, these procedures have been labeled in Table
4-1 as B-l and B-2.
Option B-l, Bracketing of All Parameters, is a simple exercise which can be conducted
by the transportation modeler. Inputs to the model, such as speed, can be set to the
extreme values of the anticipated variation which would be seen for these inputs. The
outputs produced by such an exercise express the range of variability which can be
attributed to the specific model input.
Option B-2, Conduct Speed Surveys to Better Validate Models. This is another
technique which would require more extensive survey data than is frequently collected for
a transportation model.
Factoring and Cross-Classification Processes
Seven procedures have been suggested in Table 4-1, labelled as C-l through C-8, which
are predominantly post-processing steps for the model outputs rather than actual changes
in modeling procedure. Many require more extensive socioeconomic and origin and
destination survey data than is commonly collected.
Option C-l, Relating Fleet Mix to Network Characteristics. This would rely upon
survey information to better characterize the vehicle fleejt composition expected on
specific facility types or subregions. This would allow a better match of emission rates
92004r2.07
43
-------
to vehicle activity for these links or zones. One near-term solution might be to use tube
counters that differentiate between light vehicles and heavy vehicles, and use local
registration distributions or national defaults to split those groups. In the future, one
might be able to rely on automatic vehicle identification equipment placed on selected
travel routes.
Option C-2, Relate Fleet Characteristics to Specific Trip Types. This is similar to
option C-l. It would develop trip tables for specific fleets, such as heavy-duty vehicles.
Such tables would vary by time of day, day of week, and facility class. This allows a
better match of emission rates to vehicle activity for this trip type. It does require the
transportation model to report travel activity by trip type and zone/link.
Option C-3, Time-Specific Fleet Characterization. This is similar to the previous
options in this category. It would use origin and destination survey data to alter average
fleet characteristics by time of day, in order to better match emission rates to travel
activity. The emission changes which can be seen from changing fleet characteristics are
illustrated with a sample model run discussed later in this chapter.
Option C-4, Diurnal Speed Profiles. This would change the speeds input to the
transportation model based upon time of day. For example, if model runs were
performed for peak and off-peak periods, one would use congested and freeflow speeds
for the two runs, respectively.
Option C-5, Correcting Speeds Based on V/C Ratios. This is a post-processing step
(although it could also be combined with feedback) which has been applied in air quality
modeling conducted for Phoenix (SAI, 1987). It serves to eliminate the problem of V/C
ratios exceeding one, but may be an overly simplistic remedy as it is presently applied
since there is no feedback to the assignment or mode choice process. A thorough
discussion of this technique has been provided in a recent paper by Cambridge
Systematics (CSI, 1991).
Option C-6, Use Socioeconomic Data for Developing Off-Network Speeds/VMT. This
is intended to provide better characterization of intrazonal travel, but relies upon the
availability of adequate survey data and interpretive skills in arriving at more accurate
estimates of this travel activity.
Option C-7, Cross-Classification of Parking Duration by Zone. This would allow
better characterization of vehicle starts into hot and cold starts, and better characterization
of the magnitude of resting losses and diurnal evaporative emissions from parked
vehicles. This type of information is not currently being provided by transportation
models, and still requires some developmental work.
92004r2.07
44
-------
Intractable Problems and Statistical Limitations
This category is not a list of suggested procedures but instead groups two problems facing
transportation modelers which currently cannot be handled effectively either because of
the complexity of the issue or because of statistical limitations within the model.
Referred to as options D-l and D-2, these are respectively, identifying the statistical
limitations and absolute uncertainty of model outputs, and properly modeling the effects
of transportation control measures (TCMs).
TCMs are particularly important at this time due to legislative requirements, including the
CAAA, which are forcing regions to develop TCM packages. Since many TCMs, such
as telecommuting, affect a very small percent of total travel activity, statistical limitations
inherent in the data used to develop transportation models do not allow the effects of such
measures to be accurately predicted by the models.
Modifications to Models to Address Air Quality
Three options are provided in this category. All require receding of existing
transportation models to implement, but the result would be that models would provide
information which is better suited to the needs of air quality modelers. These options are
referred to as E-l through E-3 in Table 4-1.
Option E-l, Estimation of Transit Emissions Using Public Transit Models. This
represents a refinement of the usual treatment of transit activity in transportation models
in that a model specifically developed for representation of transit activity would be
integrated into the transportation model. Ideally, such a model would provide link and
zone based estimates of transit activity (including non-revenue producing activity).
Option E-2, Further Disaggregation of Zones. This is intended to allow more specific
tailoring of the region covered by a transportation model to the region of interest in air
quality modeling. Ideally, the level of detail in outlying areas of the modeled region,
which can serve as important sources of emissions due to transport, would be as detailed
as could be practically managed, given the statistical accuracy of databases.
Simplification of the network in these outlying areas would not occur in order to conserve
resources, but only when data were not available to support a more detailed treatment.
Option E-3, Speed Inputs Based on Cross-Classification of Network Characteristics.
This is similar to option C-6, which applied to assumptions for intrazonal travel. This
suggested procedure would refine the methods used in assigning initial speeds to each link
in a network to take into account more information on the actual network. This
information could include the occurrence of episode specific events, such as roadway
construction, which would have significant effects on a subset of links in a network.
92004r2.07
45
-------
DEMONSTRATION OF SELECTED PROCEDURES
As mentioned earlier, three of the options described above have been demonstrated using
a prototypical transportation model developed for this project. Use of a prototypical
model allowed demonstration of a wider range of potential improvements to the
transportation model process within the resources allocated for this effort, and given the
absence of survey data required to actually implement many of these suggested
procedures. The model was developed by drawing from data bases established for
existing models and uses the TRANPLAN version 7.0 and NIS version 3.0 model
software (UAG, 1990). The exercises conducted with this prototypical model serve as
demonstrations of how these procedures can be implemented and confirm that the effects
can be significant. They do not quantify the actual magnitude of change to be expected
from their implementation in an urban transportation model. The prototypical network
used here is too small and conditions on it, particularly in terms of congestion, too
unrepresentative to allow one to assume that these results are indicative of results that
would be obtained with an actual urban transportation model. In future work, the EPA
may use actual urban transportation models to obtain more realistic estimates of the
effectiveness of some of these techniques.
Description of Base Case
The transportation network used in these modeling exercises represents a simplified
prototype network. The highway network geometry consists of 27 zones, 190 links, and
has a maximum node number of 317. The geometry of the transit portion of the network
contains the same number of zones, but has 174 links and a maximum node number of
314. Figures 4-1 and 4-2 are graphical representations of these networks. Appendix A
contains a summary of the land-use data used by the model.
The initial step in our analysis was to generate a base transportation network using the
TRANPLAN transportation modeling software. Figure 3-3 is a flow diagram of the basic
modeling procedure. Note that any modules not enclosed in the thick border are add-on
applications to the core modeling software. This modeling exercise provided baseline
values to be used for comparison in the test exercises detailed below. Note that in the
base case feedback is not used to reflect the effects of congestion on trip length and
speeds in both the highway and transit networks. Options A-3, A-5, and C-l (see Table
4-1) were chosen for modeling. The overall results of these modeling exercises are given
in Table 4-2.
92004r2.07
46
-------
FOX LANDING
HWY NETWOF
27 centroWs
FIGURE 4-1. Highway network employed in sample model runs.
92004
47
-------
SoSHI
~t310l
FIGURE 4-2. Transit network employed in sample model runs.
92004
48
-------
HIGHWAY SELECTED
SUMMATION
1
TRANSIT PARftM
& LINK DATA
TRANSIT LINKS /ROUTES
(UPDATED SPEEDS)
ZONAL LANDUSE
DATA
INTER:
HIGHWAY
IONAL
r SKIMS
TRIP
GENERATION
FRICTIOH FACTOR
CURVE DATA
!
ZONA.
AND
sUHHs
ll
BUILD INTRAZONAL
IMPEDANCES
ZONAL PRODUCTIONS
AND ATTRACTIONS
SUH-sH:::::
:SKS=K::S=
TOTAL P/A
TRIP TABLE
101
ON!
:::
|l
i
IS
s
BH
::
::=£::::::::::::
HIGI
::::::
IHA1
TAE
SH
r si
1LE
S:H
CIM
iSHHHU:
TRANSIT SELECTED
SUMMATION
(continued next page)
FIGURE 4-3. Structure of prototypical transportation model. The model
components are described in the TRANPLAN user's manual (Urban Analysis
Group, 1990).
92004
49
-------
MATRIX TRANSPOSE
HIGHWAY A/P
TRIP TABLE
MATRIX UPDATE
MATRIX UPDATE
HIGHWAY ORIGIN
TRIP TABLE
HIGHWAY DESTINATION
TRIP TABLE
MATRIX UPDATE
MATRIX MANIPULATE
HIGHWAY ORIGIN/DESTINATION
TRIP TABLE
LOAD HIGHWAY
NETWORK
Jl
LOADED HIGHWAY
NETWORK
NETCARD
PROGRAM
TRANSIT ORIGIN
TRIP TABLE
TRANSIT DESTINATION
TRIP TABLE
MATRIX MANIPULATE
J
TRANSIT ORIGIN/DESTINATION
TRIP TABLE
LOAD TRANSIT
NETWORK
r
LOADED TRANSIT
NETWORK
LOADED HIGHWAY
NETWORK (ASCII)
VEHICLE MIX BY
FACILITY TYPE
n
LOTOS
PROGRAM
LINK
SMI S3 IONS
FIGURE 4-3. Concluded.
92004
50
-------
TABLE 4-2. Summary of modeling exercise results.
Modeling
Exercise
A-3
A-5
Highway
Transit
C-l
Speed Conditions
Highway
Free
Congested
N/A
Transit
Congested
Free
N/A
Trips
-20%
+ 1%
-5%
N/A
Observed Change in
VMT
-27%
Higher
Lower
N/A
Time
-13%
Higher
Lower
N/A
Speed
Lower
Lower
Lower
N/A
Emissions
Lower
Mixed
Mixed
Lower
Option A-3: Use Transit Speeds that Reflect Congested Roadway Conditions
In initial base case free-flow speeds are applied to both the highway and transit networks. In
this exercise the loaded highway network file produced in the base case run was input into the
INET* module to generate transit speeds that would reflect congested roadway conditions.
The effects of this procedure on passenger trips, miles traveled, and travel time can be seen in
Table 4-3.
TABLE 4-3. Results of Option A-3 on transit trips, miles, and hours of travel.
Passenger
TRANPLAN Model Run
Base Case
Option A-3
% Difference
Trips
4172
3330
- 20 %
Miles
4401
3191
- 27 %
Hours
956
827
- 13 %
All three transit usage variables exhibit significant reductions when the loaded highway
network is used as input to INET. This occurs because modeling transit choice with congested
rather than free-flow speeds reduces its attractiveness.
* INET is a transit network program released by the U.S. Department of Transportation for use with the
Urban Transportation Planning System (UTPS) software. Its principal task is to calculate transit running
times. INET computes transit speeds as a linear function of highway speed.
92004r2.07
51
-------
Option A-5: Examine the Impact of Using the Loaded Highway Network
In this test case, the loaded highway network generated by the base run was input in to
the Highway Selected Summation stage of the modeling process, replacing the Build
Highway Network step used initially (see Figure 4-3). The predicted effect of this
procedure is that the average trip times would increase, reflecting the congested highway
speeds. This will affect the trip distribution, mode choice, and trip assignment processes.
In the base case, free flow speeds were used. Table 4-4 summarizes the results of this
exercise.
As illustrated by Table 4-4, the apparent effect of this test case on trip times is
negligible. This prompted further examination of modeling outputs, as it was expected
that there would be noticeable changes using the loaded highway network. These results
are described in greater detail below.
The next model run is an extension of the two procedures described above. In this
exercise, the loaded highway network from the base transportation model run was first
input into both INET and the Highway Selected Summation phases of the model, and then
the modeling process was run to its conclusion. The predicted effect of this method was
that the congested highway speeds would affect both the highway and transit networks at
the mode choice level. It was predicted that the increase in travel time on the transit
network would result in a decrease in transit usage, and subsequently an increase in
automobile trips. Table 4-5 illustrates the results of this mode choice comparison.
As Table 4-5 illustrates, there is a shift in the trips from transit to the highway network
when congested speeds are input into the transit network. These changes are fairly small,
however we believe that a larger, more complex transportation network would exhibit
more widespread and significant changes in mode choice.
Another result of this experiment was a subtle change in the origins/destinations and
volumes of trips at the zonal level. Appendix B contains the Report Matrix Comparison
output which contrasts the origin/destination tables of the base case with those from this
sample model run. In nearly every zone, changes in the number of trips is evident. The
degree of change is very zone-dependant. Particularly dramatic changes (greater than
10%) in trip origins/destinations are evident in zones 12, 13, 15, 17, 20, and 25.
Significant (greater than 10%) changes in volumes are also seen in the aforementioned
zones, as well as zones 9, 16, 18, 21, 22, 26, and 27. These changes occur because
congested speeds increase travel time disproportionately across the region, making some
zones less attractive to travel to. This results in a shift in the distribution of origins and
destinations. Trips are shifted onto shorter pathways, potentially resulting in increased
intrazonal travel.
92004r2.07
52
-------
TABLE 4-4. Effects of congested highway speeds on trip distribution by trip type.
Trip Type*
Type 1 (Base)
Type 1 (Option A-5)
% Difference
Type 2 (Base)
Type 2 (Option A-5)
% Difference
Type 3 (Base)
Type 3 (Option A-5)
% Difference
Trip-Hours
3255
3270
0.005 %
6522
6559
0.007 %
2702
2700
0
Average Trip
Length (min)
16.222
16.293
0.004 %
15.039
15.124
0.007 %
12.007
11.997
0
* Trip type 1 = home-based work
Trip type 2 = home-based other
Trip type 3 = non-home based
TABLE 4-5. Effects of congested speeds on mode share.
TRANPLAN Test Case
Highway Trips Summary
Base Case
Option A-5
% Difference
Transit Trips Summary
Base Case
Option A-5
% Difference
Origins/
Productions
24,354
24,695
1 %
8,128
7,722
-5 %
Destinations/
Attractions
24,354
24,695
1 %
8,128
7,722
-5 %
Total
48,708
49,390
1 %
16,256
15,444
-5 %
92004*2.08 53
-------
Option C-l: Postprocessing of Link-Based Travel to Estimate Emissions
In this example, MOBILE 4.1 NOX emission rates and vehicle flest mix information are
combined with data from the Loaded Highway Network file to estimate link-level NOX
emissions. This exercise illustrates the importance of matching fleet mix characteristics
to roadway class in estimating emissions for a region, and also illustrates the effects of
speed on vehicle emission rates.
Emission rates used correspond to a 1991 calendar year fleet in wintertime conditions
(minimum temperature of 20 F, maximum of 40 F). Two fleet mixes were modeled in
this exercise, one using a VMT by vehicle class distribution which is based upon national
averages, referred to as "Default" in subsequent tables, and the other reflecting the
arbitrary reduction in VMT attributed to heavy duty vehicles (both gasoline and diesel),
with a concurrent increase in the VMT attributed to light duty gas vehicles. This second
scenario is intended to roughly represent the fleet characteristics found on residential
streets, although the numbers given are provided purely for the purposes of
demonstration, and as such are not based upon actual survey data. This second fleet mix
is referred to as the "Reduced HDV" in subsequent table. Table 4-6 summarizes the fleet
mixes and emission rates used.
Three scenarios are considered. In the first, the default fleet mix is used to calculate the
fleet average emission rate, which is then applied to all travel on the network to calculate
total NOX emissions. The second uses the reduced HDV fleet mix to calculate the fleet
average emission rate, which is then applied to all travel on the network to calculate total
NOX emissions. The final scenario uses the default fleet mix to calculate the fleet
average emission rate applied to travel for facility types 0 and 1, and the reduced HDV
fleet mix to calculate the fleet average emission rate applied to travel for facility type 2.
Table 4-7 is an example of the spreadsheet developed for these link-based emission
calculations, and Table 4-8 summarizes the fleet average emission rates, VMT
accumulation, and emissions by speed range, facility type, and scenario.
As would be expected, the greatest difference is seen between emissions calculated by
assuming all travel in the model occurred with the default fleet mix as opposed to
assuming it occurs with the reduced HDV fleet. The reduced HDV fleet produces
approximately 40 percent less NOX over the network than does the default fleet.
However, since this is only a prototypical model, this exercise can only serve to indicate
that fleet mix can significantly affect emissions estimates, and that the emissions will
change significantly depending upon the speeds output by the model.
RECOMMENDED FOLLOW-UP
The resources of this project allowed only a small number of procedural improvements to
actually be demonstrated with sample model runs using the prototypical model developed
92004.r2.07
-------
Table 4-6. Summary of fleet mix and emission rates.
Vehicle NOX Rate
Class (g/mi), 20 mph
LDGV
LDGT1
LDGT2
HDGV
LDDV
LDDT
HDDV
MC
1.74
2.15
2.69
6.46
1.65
1.9
17.52
1.07
NOX Rate
(g/mi), 50 mph
1.65
2.11
2.75
8.16
1.74
2.00
18.44
1.60
NOX Rate
(g/mi), 65 mph
2.57
3.29
4.27
9.01
2.88
3.32
30.54
2.49
Default
Scenario VMT
Distribution
.62
.175
.077
.035
.008
.002
.075
.008
Reduced HDV
Scenario VMT
Distribution
.669
.224
.077
.004
.008
.002
.008
.008
92004rl.lO
55
-------
TABLE 4-7. Calculation of link-based emissions.
A Node
203
115
114
113
114
117
201
117
201
201
a! 205
203
207
107
209
203
112
111
110
111
111
108
B Node
205
300
120
114
116
309
203
123
301
207
208
303
214
108
211
212
113
121
201
112
1222
2141
Link
Facility
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
1
Information
Distance
0.6
0.4
0.4
0.4
0.6
0.6
08
0.4
0.5
0.6
0.3
0.5
0.3
0.3
0.3
0.5
0.4
0.5
0.8
0.2
0.3
0.3
Speed(A-B)
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
69.23
69.23
Volume
662
803
0
681
534
431
935
558
849
512
940
864
795
0
0
676
573
157
798
573
736
121
No
LDGV
443
537
0
456
357
288
626
373
568
343
629
578
532
0
0
452
383
10S
534
383
492
75
. of Vehicles
LDGT
148
180
0
153
120
97
209
125
190
115
211
194
178
0
0
151
128
35
179
128
165
21
HDGV
3
3
0
3
2
2
4
2
3
2
4 -
3
3
0
0
3
2
1
3
2
3
4
LDGV
265.73
214.88
0.00
182.24
214.35
173.00
500.41
149.32
283.99
205.52
188.66
289.01
159.56
0.00
0.00
226.12
153.33
52.52
427.09
76.67
147.72
22.51
VMT
LDGT
88.97
71.95
0.00
61.02
71.77
57.93
167.55
50.00
95.09
68.81
63.17
96.77
53.42
0.00
0.00
75.71
51.34
17.58
143.00
25.67
49.46
6.35
NOX Emissions
HDGV
53.38
28.78
0.00
24.41
43.06
34.76
134.04
20.00
47.54
41.29
18.95
48.38
16.03
0.00
0.00
37.86
20.54
8.79
114.40
5.13
14.84
1.91
LDGV
438.45
354.56
0.00
300.69
353.67
285.46
825.68
246.38
468.58
339.10
311.29
476.86
263.27
0.00
0.00
373.10
253.00
86.65
704.70
126.50
379.63
57.84
LDGT
187.73
151.81
0.00
128.75
151.43
122.22
353.53
105.49
200.64
145.20
133.28
204.18
112.72
0.00
0.00
159.75
108.33
37.10
301.73
54.16
162.72
20.90
HDGV
146.81
79.14
0.00
67.12
118.42
95.58
368.61
55.00
130.75
113.54
52.11
133.06
44.07
0.00
0.00
104.10
56.47
24.18
314.60
14.12
133.69
17.17
Continued
92004ri.l2
-------
TABLE 4-7. Concluded.
in
-4
A Node
IIS
116
208
108
110
117
209
208
110
BNode
1161
1171
2141
1101
1151
1191
2101
2091
1222
Link
Facility
1
I
1
1
1
1
1
1
2
Information
Distance
0.3
0.3
0.8
1
0.5
0.12
0.6
0.4
0.2
Speed(A-B)
69.23
69.23
69.57
69.77
69.77
69.9
70.59
70.59
70.59
Volume
1539
1552
1139
920
128
1453
1086
1086
928
No
LDGV
954
962
706
570
79
901
673
673
621
. of Vehicles
LOOT
269
272
199
161
22
254
190
190
208
HDGV
54
54
40
32
4
51
38
38
4
LDGV
286.25
288.67
564.94
570.40
39.68
108.10
403.99
269.33
124.17
VMT
LDGT
80.80
81.48
159.46
161.00
11.20
30.51
1 14.03
76.02
41.57
NOX Emissions
HDGV
24.24
24.44
127.57
161.00
5.60
3.66
68.42
30.41
8.31
LDGV
735.67
741.89
1451.91
1465.93
101.98
277.83
1038.26
692.17
319.11
LDGT
265.82
268.07
524.62
529.69
36.85
100.39
375.16
250.11
136.78
HDGV
218.40
220.24
1149.39
1450.61
50.46
32.99
616.45
273.98
74.92
TOTAL 13470.15
5329.187 6155.967
NOX EMISSIONS
(GRAMS)
NOX
EMISSIONS(LBS)
24955.3
54.91
92004rl.l2
-------
Table 4-8. Projected VMT and emission levels for comparison of fleet mixes.
Fleet-Average NOX Rate
Default Fleet (g/mi)
Reduced HDV Fleet
Total VMT
Facility Type 0 (Centroids)
Facility Type 1
Facility Type 2
Total NOX
With Default Fleet (grams)
With Reduced HDV Fleet (grams)*
With Default Fleet on Facilities
0 & 1, Reduced HDV Fleet on
Facility 2 (grams)*
For VMT
Accumulated at
20 mph
3.234
2.045
1273
0
0
4116.9
2603.3 (-37%)
41 16.9, (0%)
For VMT
Accumulated at
50 mph
3.308
1.999
0
0
11156
36903.1
22300.2 (-40%)
22300.2 (-40%)
For VMT
Accumulated at
65 rr.nh
5.161
3.112
12872
4629
1772
99468.9
59978.1 (-40%)
95838.3 (-4%)
Numbers in parenthesis reflect percent change from default fleet emission levels.
92004rl.ll
58
-------
for this effort. It would be useful to explore the potential benefits of s---ne cf the- -ther
procedures listed in Table 4-1 further in future modeling work using the protoryp;cal
model as well as actual travel demand networks. This would accomplish two goals: (a)
demonstration of the practical utility of a procedure and (b) demonstration of the
anticipated effects of a procedure. In short, further modeling efforts can answer fhe
questions of whether an idea can actually be implemented, and whether it would result in
enough improvement in the model predictions to make it worth whatever extra resources
are required for the implementation.
Future modeling work should not be limited to the ideas presented in this report. Given
the number of concurrent research efforts ongoing in the field of transportation modeling,
it would be worthwhile to select the best ideas from these other efforts also and assess
their practical merit. For example, some of the suggestions anticipated from the ongoing
work sponsored by the National Association of Regional Councils (Harvey and Deakin,
1991), as well as work in progress by Cambridge Systematics for Region DC of the EPA
(CSI, 1991), would be interesting to investigate through actual model runs. The
prototypical model that has been developed for this current work effort can be adapted for
the demonstration of these ideas with the goal of determining their practicality and the
anticipated benefits of their use.
92004r2.07
59
-------
References
Applied Management. 1990. "Traffic Congestion and Capacity Increases." Applied
Management and Planning Group, Los Angeles, California.
Atkins, S. T., 1986. Transportation planning models-what the papers say.
Traffic Engineering and Control. 27(9):460-467.
Cambridge Systematics, Inc. 1991a. "The Preparation of Highway Vehicle
Emissions Inventories." Prepared for the Transportation Research Board
70th Annual Meeting, January 13-17, 1991, Washington, DC.
Cambridge Systematics, Inc. 1991b. "Development of Information on Transportation
and Air Quality." Cambridge Systematics, Inc., Cambridge, Massachusetts.
CARB. 1990. "Derivation of the EMFAC7E Emission and Correction Factors for On-
Road Motor Vehicles. California Air Resources Board, Mobile Source Division,
Sacramento, California.
Deakin, E. 1991. "Seminar on Transportation Modeling for Emissions
Analysis." Prepared for presentation June 20, 21, 1991, Sacramento,
California, California Air Resources Board.
EPA. 1991. User's Guide to MOBILE4.1 Mobile Source Emission Factor ModeD.
U.S. Environmental Protection Agency (EPA-AA-TEB-91-01).
Hamerslag. R. 1980. "Spatial Development, Developments in Traffic and
Transportation, and Changes in the Transportation System." In J. B. Polak and
J. B. van der Kamp. Changes in the Field of Transport Studies. Martinus
Nijhoff, The Hague.
Harvey, G., and E. Deakin. 1991. "Toward Improved Regional Transportation
Modeling Practice." National Association of Regional Councils, Clean Air
Project Transportation Modeling Conference, Washington, D.C.
92004.09
60
-------
Ireson, R. G., J. L. Haney, L. A. Mahoney, M. C. Dudik, J. L. Fieber, and D. R.
Souten. 1987. "Carbon Monoxide Air Quality Modeling for the Phoenix
Metropolitan AreaEmission Reduction Requirements and Control Measure
Effectiveness." Systems Application, Inc., San Rafael, California (SYSAPP-
87/059).
Ismart, D. 1991. "Travel Demand Forecasting Limitations for Evaluating
TCM's." Air and Waste Management Association 84th Annual Meeting &
Exhibition, Vancouver, BC.
Lorang, P. 1991. "Interim Guidance for the Preparation of Mobile Source Emission
Inventories." Memorandum from Phil Lorang, Chief, Technical Support Staff,
U.S. Environmental Protection Agency (January 31, 1991).
Meyer, M. D., T. F. Humphrey, C. M. Walton, K. Hooper, R. G. Stanley,
C. K. Orski, and P. A. Peyser, Jr. 1989. "A Toolbox for Alleviating
Traffic Congestion." Institute of Transportation Engineers, Washington,
DC.
Ruiter, E. R. 1991. "Highway Vehicle Speed Estimation Procedures for Use in
Emission Inventories." Cambridge Systematics, Inc., Cambridge, Massachusetts.
SAL 1989. "Transportation Control Measures: State Implementation Plan
Guidance". Systems Applications, Inc. San Rafael, California (SYSAPP-
89/126).
Transportation Research Board. 1985. Highway Capacity Manual. National Research
Council, Washington, D.C. (Special Report 209).
Urban Analysis Group. 1990. "TRANPLAN Version 7.0 and NIS Version 3.0 User
Manuals." The Urban Analysis Group, Danville, California.
92004.09
61
-------
Appendix A
CHARACTERISTICS OF PROTOTYPICAL MODEL
92004rl.01
-------
1. FoxLandUse JL
Zone Zone Inhabitants
Type No.
CBD 01
CBD 02
CBD-Sum
Industr. 21
Industr. 23
Industr 2-8
Industr. 33"
Ind-Sum
Mixed -1-2
Mixed 4-3
Mixed 4-4
Mixed 25
Mix-Sum
Res.Flats 10
Res.Flats 1 1
Res.Flats 24-
Res.Flats 26
Res.Flats 29
Res.Flats 30
Res.Flats 32
Res.Flats 3rt
Res.Flats 35
Res.Flats 36
Flats-Sum
Res.OneF. 22
Res. One F. 27
Res.OneF 3ft
Res.OneF 37
OneF-Sum
TOTAL
Fiats
1000
2000
3000
: r 0
^ 0
n 0
ti o
0
5 3500
V 3000
" 1200
i 12000
9700
r 2500
' 2400
^ 1500
- 2000
'" 1000
', 2600
/2 3200
// 3000
/J 2800
><- 3000
24000
0
iff 0
/r o
0
0
36700
One-fan
0
0
0
0
0
0
0
0
500
0
300
0
800
0
0
0
300
0
400
800
0
0
200
1700
1800
1200
2000
3000
8000
10500
Incoming commuters from
External zone 104- 2',
102- ll
Outgoing commuters to
External zone -1-Ot -'-
1J02 ;-
Grand Total
Econ.
480
960
1440
JOBS -JOSS
Ind. Retail Office Misc. Tcial
1625
1230
605
820
4280
1025
985
615
935
410
1220
1615
1230
1150
1305
10490
685
455
760
1140
3040
19250
1000
550
20800
0 1000 2000
0 1000 1050
400 3400
600 2550
0 2000 3050 1000 6050
2000
1500
1500
1000
6000
600
850
400
1200
3050
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
140
100
0
100
340
0
20
0
220
0
50
30
20
0
200
540
30
20
20
50
0
0
0
0
0
600
700
200
500
2000
50
0
0
150
0
150
50
0
50
300
750
0
0
50
200
0
0
0
0
0
300
400
0
0
700
0
0
0
100
0
100
0
0
0
100
300
0
0
0
0
2000
1500
1500
1COO
6000
1640
2050
600
800
6090
50
20
0
470
0
300
80
20
50
600
1590
30
20
70
250
0 120 250 0 370
9050 3000 6050 2000 20100
300
400
20800
n.
-------
2. FoxLandllse 2
Re-edited in accordance with limitations in TRANPLAN.
INHABITANTS
INDUSTRY
RETAIL
OFFICE
Zone no.
MISC.
(01)
(02)
(12)
(13)
(10)
01)
(14)
(38)
(37)
(36)
(34)
(32)
(35)
(33)
(31)
(30)
(29)
(28)
(27)
(20)
(24)
(25)
(23)
(22)
(21)
Ext
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Ext 27
Total
CBD
1000
2000
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3000
Flat
0
0
3500
3000
2500
2400
1200
0
0
3000
3000
3200
2800
0
0
2600
1000
0
0
2000
1500
2000
0
0
0
0
0
33700
House
0
0
500
0
0
0
300
0
3000
200
800
0
0
2000
400
0
0
1200
300
0
0
0
1800
0
0
0
10500
Large Small CBD
0
0
0
0
0
0
0
0
0
0
0
0
0
1000
0
0
0
1500
0
0
0
0
1500
0
2000
0
0
6000
0
0
600
850
0
0
400
0
0
0
0
0
0
0
0
0
0
D
0
0
0
1200
0
0
0
0
0
3050
1000
1000
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2000
Other
0
0
140
100
0
20
0
0
50
200
20
30
0
0
20
50
0
0
20
220
0
100
0
30
0
0
0
1000
CBD
2000
1050
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3050
Other
0
0
600
700
50
0
200
0
200
300
0
50
50
0
50
150
0
0
0
150
0
500
0
0
0
0
0
3000
ii ^* ^^
400
600
300
400
0
0
0
0
0
100
0
0
0
0
0
100
0
0
0
100
0
0
0
0
0
0
0
2000
Col 11 Col 12 Col 13 Col 14
(38) 8 1000 000
(101)26 0 1000 300 3830
(102) 27 0 550 400 3145
Total 0 1550 700 6975
Figures in the four last columns do not refer to any explicit land use
category. For all other zones than the mentioned 8,26, and 27 the values
are 0 in these columns.
-------
The forecast example will be carried out as a traditional application
of the four-step gravitational model. Data and parameters are ourely
fictitious but reasonable. Some parts of the approach are a iitiie
more sophisticated than is usually used; the intention is to get
familiar with the possibilities of TRANPLAN and get ideas for the
combined system VIPSPLAN.
3.1 Travel purposes
The trip production and the zonal distribution are carried out in the
form of three one-directional trip matrices:
Home to Work
Home to Visit
Non-home to Visit
3.2 Notes to the land use table
The Land use table is first given in a form that is rather common in
this type of studies. It is then re-'edited according to the
conventions in TRANPLAN.
The rate of economically active population is for people living in
CBD 48 %
Flats 41 %
One-family houses 38 %
The external area consists of the two zones 101 and 102.
In-coming commuters from zone 101: 1000
zone 102: 550
Out-going commuters to zone 101: 300
zone 102: 400 ,. ,-
!
Total number of trips to / from zone 101: 10000 - " - - ~ ~
zone 102: 8000 - "- - J";£-
3.3 Definitions
Trip is here defined as a movement from Origin to Destination by
car or public transport.
A commuter is a person with a given home zone and a given work
zone (within or outside the city.) This zonal combination results in a
trip a particular day only if the person really goes to work that day
and does it by car or public transport.
-------
3.4 Assumed trip rates
Trip production per day and person normally varies with various
socio - economical factors etc. These factors are here represented by type and
location of the home.
Average trip rate per person living in CBD is 1.5 and for the rest 2.0. Residents
in houses have a trip rate that is 25 % higher than those in flats. This gives the
following rates and number of trips by the inhabitants of Fox Landings:
Residents in CBD
Residents in flats
Residents in houses
Total
Trip rate
1.5
1.92
2.40
Number of trips
4 500
64 700
25 200
94 400
3.5 Trips across the border line
The Home / Work relations can be written in the following matrix form:
WORK
HOME
FOX LANDING
EXTERNAL 2.
TOTAL |
FOX LANDING
18550
1550
20100
EXTERNAL ZONES
700
0
700
TOTAL
19250
1550
20800
In order to transform these figures into trips the following assumptions are
made:
10 % are absent from work on an average day.
60 % of the internal and all the external commuting is made by car or transit.
This corresponds to the following generation and attraction factors for the
home-work trip matrix:
Generation: 0.9 x 0.6 = 0.54 applied to the value of economically
active population in each internal zone and 0.9 applied to the value of
incoming commuters.
Attraction: 0.54 applied to the number of jobs in each zone and 0.9
applied to outgoing commuters.
-------
WORK
HOME
FOX LANDING
EXTERNAL 2.
TOTAL |
FOX LANDING
20034
2790
22824
EXTERNAL ZONES
1260
0
1260
TOTAL
21294
279C
24084
The remaining trips (not Home-Work) we call "visiting trips" or "non-work trips".
Such trips between Fox Landing and the external zones amount to
18000-1260-2790 = 13950.
50 % of these are supposed to be produced by the people in the city and 50 %
by external people. Thus the city inhabitants make 94400 - 21294 = 73106 non-
work trips. The double directed matrix can be written
TO
FROM
FOX LANDING
EXTERNAL 2.
TOTAL |
FOX LANDING
66125
6975
73100
EXTERNAL 2ONES
6975
0
6975
TOTAL
73100
6975
80075
3.6 Distribution on trip generators
The total number of trips is assumed to be generated as follows:
Dwellings
Work
Visit
65% = 47515
20%= 14620
15 %= 1Q965
SUM
100% = 73100
With the same relative trip rate for different types of dwellings as above, the trip
generation per inhabitant will be
CBD
Flats
Houses
0.755
0.966
1.208
-------
14620/20100 = 0.727.
Generation in chains of visits is put equal for all kinds of visits. Eacn £it;i. . a
visit consequently is assumed to generate
10965 / 73100 = 0.15 visiting trip.
Of the 73100 non-work trips 2000 are supposed to go to the recreation area in
zone 38. This estimate has to be based on special considerations as land use
in terms of dwellings or jobs does not exist or does not indicate the trip
attraction. Thus 71100 trips have to be distributed on the attraction side
according to the land use. The following assumptions are made:
Dwellings 20 % (evenly distributed over all inhabitants)
Industry 15 % (three times higher in mixed areas than
in the heavy industrial zones)
Retail 25 % (CBD -located shops generate twice as
many visits per employee as the rest)
Office 20 % (same relations as for retail)
Misc. 20 % (the same for all zones)
The resulting attraction factors become:
Dwellings 0.301 per inhabitant
Industry (heavy) 0.704 per employee
(light) 2.112 --"--
Retail (CBD) 7.110
(other) 3.555
Office (CBD) 3.125
(other) 1.563
Misc. 4.740
All factors so far refer to round trips with the two trip legs identical. This does not
present any difficulties for the construction of a trip matrix for one full day.
However, if it is desired to build a matrix for a peak hour or some other
dimensioning period of the day, it is better to create the matrices as one-
directional flows. In the following calculations, therefore, half the generation
and attraction factors given above will be used.
The number of trips between the forecast area and the surrounding world
cannot be calculated from any land use as no such is given. It has to be
estimated separately and then inserted as constants into the trip generation
-------
in TRANPLAN, four additional columns (11 - 14) were aadec .- : >'.-
FoxLandUse table only containing values for one or more of na zones
and 27.
Summary of trip generation factors
Variable
(column)
Home to Work
Gen. Attr.
Home to Visit
Gen Attr.
1
2
3
4
5
6
7
3
9
10
11
12
13
14
0.2655
0.2268
0.2102
0
0
0
0
0
0
0
0
0.9
0
0
0
0
0
0.54
0.54
0.54
0.54
0
0.54
0.54
0
0
0.9
0
0.3775
0.4830
0.6040
0
0
0
0
0
0
0
0
0
0
0.3250
0.0978
0.0978
0.0978
0.2288
0.6864
2.3108
1.1554
1.0156
0.5080
1.5405
0.6500
0
0
0.3250
Variable
(column)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Work/Visit to Visit
Gen. Attr.
0.0226
0.0226
0.0226
0.4163
0.5219
0.8968
0.6301
0.5979
0.4807
0.7190
0
0
0
0.1750
0.0527
0.0527
0.0527
0.1232
0.3696
1.2443
0.6221
0.5469
0.2735
0.8295
0.3500
0
0
0.1750
3.7 Zonal distribution
The zonal distribution for the three matrices is calculated by means of a
traditional gravity model with total travel time as impedance. The friction factors
are calculated as weighted means of the impedance values in the skim tables
from the highway and transit networks. As weights are used the approximate
-------
the rest. The formula reads.
f(d) = P x d(tr)a + (1.0 - P ) x d(car)a
The values of "a" are:
Home - work -1.5
Home - visit -2.0
Non-home - visit -2.5
Intrazonal trips and trips between the two external zones 26 an 27
are assumed to be « 0.
3.8 Modal split
3.9 Summing up to 24 hour trip matrices
3.10 Construction of max period matrices
-------
Appendix B
COMPARISON OF OPTION A-5 RESULTS FOR
HIGHWAY AND TRANSIT NETWORKS
J0310 92004
-------
Highway Network
J0310 92004
-------
XX
XX
XX XXXXXXXXX XXXXXXXXX
xxxxxx
XX
XX XXXXXXXXXX XXXXXXXXXX XXXXXXXX XXX
XX
XX
XX
XX
XX XX
XXXXXXXX
XXXXXXXXXX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XXXXXXXXXX
XXXXXXXX
XX
XX
XX
XX
XXXXXXXXXX
XXXXXXXXX
XX
XX
XX
XX
XX
XX
XX
XX
XX XX
XX XX
XXXXXXXXXX
XXXXXXXXXX
XX XX
XX XX
XXXXXXXXXX
XXXXXXXXX
XX
XX
XX
XX
XXXXXXXXXX
XXXXXXXXXX
XX
XX
XX
XX
XX
XX
XX
XX
XXXX XX
XX XX XX
XX XX XX
XX XX XX
XX XX XX
XX XXXX
XX XXX
XX XX
XX
XX
XX
XX
XX
XX
XX
XX
XX XX
XX
XXXXXXXXX
XXXXXXXXX
XX
XX XX
XXXXXXXXXX
XXXXXXXX
XX XX
XX XX
XXXX
XX
XX
XX
XX
XX
XX XX
XX
XXXXXXXXX
XXXXXXXXX
XX
XX XX
XXXXXXXXXX
XXXXXXXX
XXXXXXXXXX XXXXXXXXX
XXXXXXXXXX XXXXXXXXXX
XX
XX
XX
XX
XX
XX
XX XX
XX XX
XXXXXXXXXX
XXXXXXXXX
XX XX
XX XX
XXXXXXXX
XXXXXXXXXX
XX XX
XX XX
XX XX
XXXXXXXXXX
XXXXXXXXXX
XX XX
XX XX
XXX XX
XXXX XX
XX XX XX
XX XX XX
XX XX XX
XX XX XX
XX XXXX
XXXXXXXXX
XXXXXXXXXX
XX XX
XX XX
XXXXXXXXXX
XXXXXXXXX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX XX
XX XX
XX XX
XX XX
XXX XX
XX XX
XX
XX
XX
XX
XX
XX
XX
XX
XXXXXXXXXX
XXXXXXXXXX
XXXXXXXX
XXXXXXXXXX
XX XX
XX XX
XX XX
XXXXXXXXXX
XXXXXXXXXX
XX XX
XX XX
XX XX
XX XX
XXX XX
XXXX XX
XX XX XX
XX XX XX
XX XX XX
XX XX XX
XX XXXX
XX XXX
XX XX
«******»*»****«***»**********************«********»***»«
MATRIX COMPARSION REPORTS
MATRIX COMPARE OF HIGHWAY 0/0 TABLES
(BASE CASE VS TEST CASE 3)
»******»*««»****»»«*********»*************************»****
FILE CHARACTERISTICS
USER FILE IDENTIFICATION - HTRIPS.A3
FILE HEADER
MATRIX COMPARE OF HIGHWAY 0/0 TABLES
(BASE CASE VS TEST CASE 3)
GENERATING FUNCTION MATRIX MANIPULATE
TYPE OF FILE VOLUME
GENERATION FILE NAME IMAN5
GENERATION DATE 02JAN92
GENERATION TIME 15:48:15
FILE SIZE MAXIMUM ZONE
MAXIMUM TABLE NO. =
CURRENT DATE
CURRENT TIME
07JAN92
K:36:45
27
1
-------
UAG - URBAN/SYS
TRANPLAN SVSTEN
VERSION 7.0
MATRIX COMPARE Of HIGHWAY 0/0 TABLES
(BASE CASE VS TEST CASE 3)
VOLUME COHPARISON REPORT --
MAXIMUM CENTROID NUMBER = 27
- VOLUME DIFFERENCES AND RATIOS.
NUMBER OF PURPOSES
PAGE NO. 1
DATE 07JAN92
TIME 14:36:45
ZONE
TAPE 1
TAPE 2
DIFF.
RATIO
TAPE 1
TAPE 2
DIFF.
RATIO
TAPE 1
TAPE 2
DIFF.
RATIO
11
21
0
0
0
.00
55
35
20
.64
16
18
-2
1.13
377
379
-2
1.01
82
88
-6
1.07
37
39
-2
1.05
168
171
-3
1.02
49
56
-7
1.14
17
19
-2
1.12
99
104
-5
1.05
10
10
0
1.00
29
26
3
.90
136
135
1
.99
35
43
-8
1.23
56
57
-1
1.02
126
127
-1
1.01
45
46
-1
1.02
101
101
0
1.00
33
34
-1
1.03
13
17
-4
1.31
78
78
0
1.00
12
11
1
.92
14
16
-2
1.14
67
68
-1
1.01
23
22
1
.96
48
50
-2
1.04
36
39
-3
1.08
-------
IMG - IMBAN/SVS
TRAHPLAN SYSTEM
VERSION 7.0
MATRIX COMPARE OF HIGHWAY 0/0 TABLES
(BASE CASE VS TEST CASE 3)
SEPARATION COMPARISON REPORT FREQUENCY DISTRIBUTION (V1-V2).
NUMBER OF PURPOSES ' 1
PAGE NO. 2
DATE 07JAN92
TIME 14:36:45
INTERCHANGES UIIH
y
SEPARATION GRP
VI
0-
101-
201-
301-
401-
TOTAL
100
200
300
400
SOO
-SO
TO
-31
144
4
1
0
0
149
-30 -20
TO TO
-21 -11
0 0
0 0
0 0
0 0
0 0
0 0
ZERO
SEPARATION TAPE 1 =
PURPOSE 1
5
TAPE 2 = 7
NEGATIVE
-10
TO
-8
0
0
0
0
0
0
-7
TO
-6
0
0
0
0
0
0
-5
TO
-4
0
0
0
0
0
0
-3
TO
-3
0
0
0
0
0
0
-2
TO
-2
0
0
0
0
0
0
-1
TO
-1
0
0
0
0
0
0
-0
TO
+0
221
a
0
2
0
231
+1
TO
+1
0
0
0
0
0
0
+2
TO
+2
0
0
0
0
0
0
+3
TO
+3
0
0
0
0
0
0
+4
TO
+5
0
0
0
0
0
0
+6
TO
+7
0
0
0
0
0
0
POSITIVE
+8 +11 +21 +31
TO TO TO TO
+10 +20 +30 +50
0 0 0 314
0 0 0 22
0008
0003
0002
0 0 0 349
TOT
679
34
9
5
2
729
-------
IMG - URBAN/SYS
TRANPLAM SYSTEM
VERSION 7.0
MAXIMUM
MATRIX COMPARE OF HIGHWAY 0/0 TABLES
(BASE CASE VS TEST CASE 3)
VOLUME COMPARISON REPORT STATISTICAL CALCULATIONS.
CENTROID NUMBER = 27 NUMBER OF PURPOSES =
1
PURPOSE 1
VOLUME GRP
VI
0- 100
101- 200
201- 300
301- 400
401- 500
TOTAL
VOL.
TAPE1
14755
4735
2120
1795
949
24354
AVG.
VOL.
21.7
139.3
235.6
359.0
474.5
33.4
VOL.
TAPE2
1S042
4706
2156
1B03
953
24660
AVG.
VOL.
22.2
138.4
239.6
360.6
476.5
33.8
AVG.
OlFF.
-.42
.85
-4.00
-1.60
-2.00
-.42
STO.
OEV.
2.46
12.03
6.20
1.50
3.00
3.63
PRCNT
S.O.
.11
.09
.03
.00
.01
.11
PRCNT
TOTAL
60.59
19.44
8.70
7.37
3.90
100.00
UGHTD
AVG.
6.86
1.68
.23
.03
.02
10.85
ROOT MN
SO.
2.5
12.1
7.4
2.2
3.6
3.6
PRCNT
RMS
11.48
8.66
3.13
.61
.76
10.93
PAGE NO. 3
DATE 07JAN92
TIME 14:36:45
SUM OF
SO D1FF
4229.
4943.
490.
24.
26.
9712.
-------
UAG - URBAN/SYS
TRANPLAN SYSTEM
VERSION 7.0
MATRIX COMPARE OF HIGHWAY 0/D TABLES
(BASE CASE VS TEST CASE 3)
TRIP END COMPARISON REPORT -- PURPOSE 1
ZONE/OIST ORIG/PROD DEST/ATTR
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
DIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
10
1762
1789
-27
1.02
1711
172S
-14
1.01
1736
1753
-17
1.01
1569
1583
-14
1.01
701
699
2
1.00
656
661
-5
1.01
617
617
0
1.00
144
142
2
.99
1133
1149
-16
1.01
1280
1284
-4
1.00
3191
3258
-67
1.02
2827
2869
-42
1.01
1772
1793
-21
1.01
1896
1929
-33
1.02
400
389
11
.97
356
370
-14
1.04
624
621
3
1.00
345
340
5
.99
700
719
-19
1.03
996
1001
-5
1.01
4953
5047
-94
1.02
4538
4594
-56
1.01
3508
3546
-38
1.01
3465
3512
-47
1.01
1101
1088
13
.99
1012
1031
-19
1.02
1241
1238
3
1.00
489
482
7
.99
1833
1868
-35
1.02
2276
2285
-9
1.00
0
0
0
.00
399
403
-4
1.01
492
497
-5
1.01
457
456
1
1.00
31
31
0
1.00
34
33
1
.97
81
80
1
.99
0
0
0
.00
101
101
0
1.00
244
241
3
.99
TOTAL INTRATRIPS ZONE/OIST ORIG/PROO DEST/ATTR
11
12
13
14
15
16
17
18
PAGE NO. 4
DATE 07JAH92
TIME 14:36:45
TOTAL INTRATRIPS
20
990
997
-7
1.01
1332
1368
-36
1.03
940
983
-43
1.05
236
238
-2
1.01
749
780
-31
1.04
1045
1063
-18
1.02
336
349
-13
1.04
354
355
-1
1.00
452
454
-2
1.00
942
973
-31
1.03
516
511
5
.99
707
718
-11
1.02
501
531
-30
1.06
474
477
-3
1.01
414
421
-7
1.02
715
725
-10
1.01
169
175
-6
1.04
714
718
-4
1.01
243
245
-2
1.01
778
783
-5
1.01
1506
1508
-2
1.00
2039
2086
-47
1.02
1441
1514
-73
1.05
710
715
-5
1.01
1163
1201
-38
1.03
1760
1788
-28
1.02
505
524
-19
1.04
1068
1073
-5
1.00
695
699
-4
1.01
1720
1756
-36
1.02
54
60
-6
1.11
120
120
0
1.00
75
76
-1
1.01
12
12
0
1.00
68
66
2
.97
161
164
-3
1.02
10
10
0
1.00
24
22
2
.92
22
21
1
.95
180
180
0
1.00
-------
IMG URBAN/SYS
TRANPLAN SYSTEM
VERSION 7.0
MATRIX COMPARE Of HIGHWAY O/D TABLES
(BASE CASE VS TEST CASE 3)
TAPE 1
TAPE 2
DIFF
RATIO
TAPE 1
TAPE 2
OlfF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
DIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
DIFF
RATIO
27
TRIP END COMPARISON REPORT -- PURPOSE 1
ZONE/OIST ORIG/PROD DEST/ATTR
21
22
23
TOTAL
INTRATRIPS ZONE/DIST OR1G/PROO DEST/ATTR
470
473
-3
1.01
1149
1159
-10
1.01
349
354
-5
1.01
674
679
-5
1.01
472
480
a
1.02
1469
1468
1
1.00
1086
1085
1
1.00
244
250
-6
1.02
1433
1448
-15
1.01
705
722
-17
1.02
357
351
6
.98
958
976
-18
1.02
1262
1264
-2
1.00
1057
1056
1
1.00
714
723
-9
1.01
2582
2607
-25
1.01
1054
1076
-22
1.02
1031
1030
1
1.00
1430
1456
-26
1.02
2731
2732
-1
1.00
2143
2141
2
1.00
18
19
-1
1.06
362
362
0
1.00
21
21
0
1.00
36
36
0
1.00
27
27
0
1.00
99
98
1
.99
68
67
1
.99
TOTAL
PAGE NO. 5
DATE 07JAN92
TIME 14:36:45
1HTRATRIPS
TAPE 1
TOTALS
24354
24354
4B708
3196
-------
TAPE 2
OIFF
RATIO
24660
-306
1.01
24660
-306
1.01
A9320
-612
1.01
3203
-7
1.00
-------
JREPORT MATRIX COMPARISON
SFIIES
INPUT FILE = HATCOH 1, USER ID - SHTRIPS.A1S
INPUT FILE » HATCOH 2. USER ID = SHTRIPS.A3*
(HEADERS
MATRIX COMPARE OF HIGHWAY 0/D TABLES
(BASE CASE VS TEST CASE 3)
SOPTIONS
PRINT FREQUENCY DISTRIBUTION
PRINT ZONAL DIFFERENCES
PRINT TRIP END COMPARISON
PRINT STATISTICAL SUMMARY
(PARAMETERS
(END TP FUNCTION
-------
UAG - URBAN/SYS
TRANPLAN SYSTEM
VERSION 7.0
COMBINE HIGHWAY P/A AND A/P TABLES TO 0/0 FORMAT - ALT 1
PAGE NO. 1
DATE 07JAN92
TIME 14:36:45
INPUT FILE NAME
MATCOM1
FILE CHARACTERISTICS
USER FILE IDENTIFICATION - HTRIPS.A1
FILE HEADER COMBINE HIGHWAY P/A AND A/P TABLES TO 0/D FORMAT - ALT 1
I
I
GENERATING FUNCTION MATRIX MANIPULATE
ITPE OF FILE VOLUME
GENERATION FILE NAME TMAN3
GENERATION DATE 19DEC91
GENERATION TIME 15:09:52
FILE SIZE MAXIMUM ZONE = 27
MAXIMUM TABLE NO. = 1
INPUT FILE NAME MAT COM2
FILE CHARACTERISTICS
USER FILE IDENTIFICATION - HTRIPS.A3
FILE HEADER COMBINE HIGHWAY P/A AND A/P TABLES TO 0/D FORMAT - ALT 3
GENERATING FUNCTION MATRIX MANIPULATE
TYPE OF FILE VOLUME
GENERATION FILE NAME THAN}
GENERATION DATE 02JAN92
GENERATION TIME 15:48:15
FILE SIZE MAXIMUM ZONE = 27
MAXIMUM TABLE NO. = 1
CURRENT DATE
CURRENT TIME
07JAN92
14:36:45
CURRENT DATE
CURRENT TIME
07JAN92
14:36:45
-------
Transit Network
J0310 92004
-------
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
xxxxxxxxxx
xxxxxxxx
XXXXXXXXX
xxxxxxxxxx
XX XX
XX XX
xxxxxxxxxx
XXXXXXXXX
XX XX
XX XX
XX XX
XX XX
XXXXXXXXX
xxxxxxxxxx
XX XX
XX XX
xxxxxxxxxx
xxxxxxxxxx
XX XX
XX XX
xxxxxxxxxx
XXXXXXXXX
xxxxxx
XXKXXXXX
XX XX
XX XX
xxxxxxxxxx
xxxxxxxxxx
XX XX
XX XX
XX XX
XX XX
XX
XXX
xxxx
XX XX
XX XX
XX XX
XX XX XX
XX XXXX
XX XXX
XX XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
xxxxxxxx
xxxxxxxxxx
XX XX
XX
XXXXXXXXX
XXXXXXXXX
XX
XX XX
xxxxxxxxxx
xxxxxxxx
XX
XX
XX
XX
XX XX
XX XX
xxxx
XX
XX
XX
XX
XX
xxxxxxxx
xxxxxxxxxx
XX XX
XX
XXXXXXXXX
XXXXXXXXX
XX
XX XX
xxxxxxxxxx
xxxxxxxx
XXXXXXXXXX XXXXXXXXX
XXXXXXXX XX
XXXXXXXXXX XXXXXXXXXX XXXXXXXXXX XXX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX XX
XX XX
XXXXXXXXX
XX XX
XX XX
XX XX
XX XX
XX XXXX
XXXXXXXXXX XX
xxxxxxxxxx
xxxxxxxxxx
XX XX
XX XX
XX XX
XX XXXXXXXXX
XX XX
XX XX XX XX XX
XX XX XX
XX XX XX
XX XXXX
XX XXX
XX XX
XXXXXXXXX
XX
XX
XX
XX
XX
XX XXXXXXXXXX XX
XX XX
XX XX
XX XX XX XX XXXXXXXXXX XX
XX
XX
XX
xxxxxxxxxx
xxxxxxxxxx
XXXXXXXX XX
XXXXXXXXXX XXX
XX
XX
XX
XX XXXX
xxxxxxxxxx
xxxxxxxxxx
XX XX
XX XX
XX XX
XX
XX
XX
XX XX XX XX
XX XX XX XX
XX XX XX
XX XX XX
XX XXXX
XX XXX
XX XX
**************************************************************************************
*
HAIRIX COMPARSION REPORTS *
MATRIX COMPARE OF TRANSIT 0/0 TABLES
(BASE CASE VS TEST CASE 3)
**************************************************************************************
FILE CHARACTERISTICS
USER FILE IDENTIFICATIOM - TTRIPS.A3
FILE HEADER MATRIX COMPARE OF TRANSIT 0/D TABLES
(BASE CASE VS TEST CASE 3)
GENERATING FUNCTION MATRIX MANIPULATE
TYPE OF FILE VOLUME
GENERATION FILE NAME TMAN3
GENERATION DATE 02JAN92 CURRENT DATE 07JAN92
GENERATION TIME 16:36:56 CURRENT TIME 14:37:58
FILE SIZE MAXIMUM ZONE = 27
MAXIMUM TABLE NO. = 1
-------
UAG - URBAN/SYS
TRANPIAN SYSTEM
VERSION 7.0
MATRIX CGHPARE OF TRANSIT 0/0 TABLES
(BASE CASE VS TEST CASE 3)
VOLUME COMPARISON REPORT --
MAXIMUM CENTROIO NUMBER * 27
- VOLUME DIFFERENCES AND RATIOS.
NUMBER OF PURPOSES
PAGE NO. 1
DATE 07JAN92
TIME 14:37:58
ZONE
TAPE 1
TAPE 2
DlfF.
RATIO
TAPE 1
TAPE 2
OUF.
RATIO
TAPE 1
TAPE 2
OIFF.
RATIO
21
0
0
0
.00
12
13
-1
t.oa
7
7
0
1.00
245
247
-2
1.01
22
16
6
.73
12
10
2
.83
49
46
3
.94
16
8
8
.50
4
3
1
.75
35
33
2
.94
1
2
-1
2.00
11
10
1
.91
107
107
0
1.00
13
6
7
.46
8
8
0
1.00
98
102
-4
1.04
16
13
3
.81
13
15
-2
1.15
9
9
0
1.00
5
2
3
.40
12
9
3
.75
1
1
0
1.00
2
3
-1
1.50
28
25
3
.89
3
3
0
1.00
13
14
-1
1.08
12
7
5
.58
-------
UAG - URBAN/SYS
TRANPLAN SYSTEM
VERSION 7.0
MATRIX COMPARE Of TRANSIT 0/D TABLES
(BASE CASE VS TEST CASE 3)
SEPARATION COMPARISON REPORT FREQUENCY DISTRIBUTION (V1-V2).
PAGE NO. 2
DATE 07JAN92
TIME 14:37:58
INTERCHANGES UITH
SEPARATION GRP
VI
0-
101-
201-
301-
TOTAL
100
200
300
400
-50
TO
-31
133
1
1
0
135
-30
TO
-21
0
0
0
0
0
20
TO
-11
0
0
0
0
0
10
TO
-8
0
0
0
0
0
ZERO SEPARATION
TAPE 1 »
PURPOSE 1
100
TAPE 2 = 99
NEGATIVE
-7
TO
-6
0
0
0
0
0
5
TO
-4
0
0
0
0
0
-3
TO
-3
0
0
0
0
0
-2
TO
-2
0
0
0
0
0
-1
TO
-1
0
0
0
0
0
-0
TO
+0
254
3
0
0
257
+1
TO
+1
0
0
0
0
0
+2
TO
+2
0
0
0
0
0
+3
TO
+3
0
0
0
0
0
+4
TO
+5
0
0
0
0
0
+6
TO
+7
0
0
0
0
0
POSITIVE
+8 +11 +21 +31
TO TO TO TO
+10 +20 +30 +50
0 0 0 327
0003
0006
0001
0 0 0 337
TOT
714
7
7
1
729
-------
DAG - URBAN/SYS MATRIX COMPARE OF TRANSIT 0/D TABLES PAGE NO. 3
TRANPLAN SYSTEM (BASE CASE VS TEST CASE 3) DATE 07JAN92
VERSION 7.0 TIME 14:37:58
VOLUME COMPARISON REPORT STATISTICAL CALCULATIONS.
MAXIMUM CENTROIO NUMBER = 27 NUMBER Of PURPOSES - 1
PURPOSE 1
VOLUME GRP VOL. AVG. VOL. AVG. AVG. STD. PRCMT PRCNT UGHTD ROOT HN PRCNT SUM OF
VI TAPE1 VOL. TAPE2 VOL. DlfF. OEV. S.O. TOTAL AVG. SO. RMS SO OIFF
0- 100 5199 7.3 4818 6.7 .53 2.13 .29 63.96 18.70 2.2 30.14 3439.
101- 200 876 12S.1 876 125.1 .00 1.20 .01 10.78 .10 1.2 .96 10.
201- 300 1740 248.6 1753 250.4 -1.86 1.64 .01 21.41 .14 2.5 1.00 43.
30t- 400 313 313.0 315 315.0 -2.00 .00 .00 3.85 .00 2.0 .64 4.
TOTAL 8128 11.1 7762 10.6 .50 2.13 .19 100.00 19.12 2.2 19.64 3496.
-------
UAG - URBAN/SYS
TRANPUM SYSTEM
VERSION 7.0
MATRIX COMPARE OF TRANSIT 0/0 TABLES
(BASE CASE VS TEST CASE 3)
TRIP END COMPARISON REPORT -- PURPOSE 1
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
Olff
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
DIFF
RATIO
TAPE 1
TAPE 2
DIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
DIFF
RATIO
TAPE 1
TAPE 2
DlfF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
ZONE/D1ST ORIG/PROO OEST/ATTR
754
719
35
.95
869
846
23
.97
711
690
21
.97
627
616
11
.98
351
352
-1
1.00
352
347
5
.99
268
266
2
.99
19
19
0
1.00
334
316
18
.95
348
346
2
.99
1379
1294
85
.94
1253
1199
54
.96
673
642
31
.95
710
676
34
.95
175
181
-6
1.03
168
169
-1
1.01
216
204
12
.94
45
44
1
.98
199
187
12
.94
286
298
-12
1.04
2133
2013
120
.94
2122
2045
77
.96
1384
1332
52
.96
1337
1292
45
.97
526
533
-7
1.01
520
516
4
.99
484
470
14
.97
64
63
1
.98
533
503
30
.94
634
644
-10
1.02
0
0
0
.00
313
315
-2
1.01
290
291
-1
1.00
269
268
1
1.00
16
16
0
1.00
18
16
2
.89
49
46
3
.94
0
0
0
.00
56
56
0
1.00
140
138
2
.99
TOTAL INTRATRIPS ZONE/DIST ORIG/PROO DEST/ATTR
11
12
13
U
15
16
17
18
19
20
PAGE NO. 4
DATE 07JAN92
TIME 14:37:58
TOTAL INTRATRIPS
258
260
-2
1.01
385
343
42
.89
236
187
49
.79
58
56
2
.97
200
163 >
37
.81
382
360
22
.94
88
69
19
.78
79
79
0
1.00
117
113
4
.97
282
243
39
.86
144
152
-8
1.06
229
202
27
.88
129
113
16
.88
117
114
3
.97
114
90
24
.79
252
246
6
.98
46
36
10
.78
163
161
2
.99
65
62
3
.95
234
216
18
.92
402
412
-10
1.02
614
545
69
.89
365
300
65
.82
175
170
5
.97
314
253
61
.81
634
606
28
.96
134
105
29
.78
242
240
2
.99
182
175
7
.96
516
459
57
.89
26
28
-2
1.08
67
64
3
.96
40
39
1
.98
6
6
0
1.00
33
33
0
1.00
91
92
-1
1.01
6
5
1
.83
11
12
-1
1.09
10
10
0
1.00
104
103
1
.99
-------
UAG - URBAN/SYS
TRANPUN SYSTEM
VERSION 7.0
TAPE 1
TAPE 2
DIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
DIFF
RATIO
TAPE 1
TAPE 2
OIFF
RATIO
TAPE 1
TAPE 2
DIFF
RATIO
24
25
26
27
MATRIX COMPARE OF TRANSIT 0/D TABLES
(BASE CASE VS TEST CASE 3)
TRIP END COMPARISON REPORT -- PURPOSE 1
ZONE/DIST ORIG/PROO DEST/ATTR
21
22
TOTAL
INTRATRIPS ZONE/DIST ORIG/PROD DEST/ATTR
164
158
6
.96
450
438
12
.97
86
83
3
.97
171
167
4
.98
103
93
10
.90
243
243
0
1.00
193
190
3
.98
81
83
-2
1.02
555
541
14
.97
171
162
9
.95
98
95
3
.97
216
183
33
.85
210
218
-8
1.04
200
194
6
.97
245
241
4
.98
1005
979
26
.97
257
245
12
.95
269
262
7
.97
319
276
43
.87
453
461
-8
1.02
393
384
9
.98
11
11
0
1.00
209
214
-5
1.02
10
10
0
1.00
19
21
-2
1.11
14
14
0
1.00
55
56
-1
1.02
40
38
2
.95
TOTAL
PAGE NO. 5
DATE 07JAN92
TIME 14:37:58
INTRATRIPS
TAPE 1
TOIALS
8128
8128
16256
1903
-------
TAPE 2 7762 7762 15524 1902
D1FF 366 366 732 1
RATIO .95 .95 .95 1.00
-------
SREPORT MATRIX COMPARISON
$flLES
INPUT FILE = HATCOH 1. USER ID - $TTRIPS.A1$
INPUT FILE - HAICOM 2, USER ID = (TTRIPS.A3*
(HEADERS
MATRIX COMPARE Of TRANSIT 0/0 TABLES
(BASE CASE VS TEST CASE 3)
(OPTIONS
PRINT FREQUENCY DISTRIBUTION
PRINT ZONAL DIFFERENCES
PRINT TRIP END COMPARISON
PRINT STATISTICAL SUMMARY
(PARAMETERS
SEND TP FUNCTION
-------
UAG - URBAN/SYS COMBINE TRANSIT P/A AND A/P TABLES TO 0/0 FORMAT - AIT t PAGE HO. 1
TRAMPUN SYSTEM DATE 07JAN92
VERSION 7.0 TIME U:37:58
INPUT HIE NAME MATCOMt
FILE CHARACTERISTICS
USER FILE IDENTIFICATION - TTRIPS.A1
FILE HEADER COMBINE TRANSIT P/A AND A/P TABLES TO 0/0 FORMAT - ALT 1
GENERATING FUNCTION MATRIX MANIPULATE
TYPE OF FILE VOLUME
GENERATION FILE NAME TMAN3
GENERATION DATE 19DEC91 CURRENT DATE 07JAN92
GENERATION TIME 15:10:55 CURRENT TIME 14:37:58
FILE SIZE MAXIMUM ZONE « 27
MAXIMUM TABLE NO. = 1
INPUT FILE NAME MATCOM2
FILE CHARACTERISTICS
USER FILE IDENTIFICATION - TTR1PS.A1
FILE HEADER COMBINE TRANSIT P/A AND A/P TABLES TO 0/D FORMAT - ALT 3
GENERATING FUNCTION MATRIX MANIPULATE
TYPE OF FILE VOLUME
GENERATION FILE NAME TMAH3
GENERATION DATE 02JAM92 CURRENT DATE 07JAN92
GENERATION TIME 16:36:36 CURRENT TIME 14:37:58
FILE SIZE MAXIMUM ZONE = 27
MAXIMUM TABLE NO. = 1
-------
TECHNICAL REPORT DATA
{Please react Instructions on the reverse before completing)
1. REPORT NO.
EPA 452/R-93-003
2.
3. RECIPIENT'S ACCESSION NO
4, TITLE AND SUBTITLE
Issues and Approaches to improving Transportation
Modeling for Air Quality Analysis.
5. REPORT DATE
January, 1993
6. PERFORMING ORGANIZATION CODE
. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Office of Air Quality Planning & Standards
Environmental Protection Agency
Research Triangle Pk., NC 27711
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68000102
12. SPONSORING AGENCY NAME AND ADDRESS
13. TYPE OF REPORT AND PERIOD COVERED
Director, Office of Air Quality Planning & Standards
Office of Air & Radiation
US EPA
Research Trianqle PK., NC 277U
14. SPONSORING AGENCY CODE
68A
15. SUPPLEMENTARY NOTES
16. ABSTRACT
Several studies sponsored by the EPA, national organizations, and state and local
agencies have been initiated to try to improve transportaion modeling. This report
documents the results of one of these efforts. The purpose of this work was to
produce a list of current shortcomings both in transportation model structure and in
the ways transportation models are used, written in large part from the perspective
of air quality modelers. The intention has been to provide a document which would
be of use to both transportation and air quality modelers. In addition, a list of
improvements to either the models or transportation modeling procedures, augmented
by sample model runs demonstrating implementation of some of these suggestions, is
provided.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lOENTlFIERS/OPEN ENDED TERMS c. COSATI Ftetd/Gcoup
Ai r Qua1i ty Analyst s
Air Quality Modeling
Improving Transportation Modeling
Improving
Transportation
Modeling
18. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (Tins Report/
Unclassified
21. NO. Or PAGcS
20. SECURITY CLASS /Tins page)
Unclassified
22. PRICE
EPA Form 2230-1 (R«v. 4-77) PREVIOUS KOITION is OBSOLETE
------- |