Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Network-based simulation of air pollution emissions associated with truck operations
(USC Thesis Other)
Network-based simulation of air pollution emissions associated with truck operations
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
NETWORK-BASED SIMULATION OF AIR POLLUTION EMISSIONS
ASSOCIATED WITH TRUCK OPERATIONS
by
Joongkoo Cho
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ENGINEERING)
May 2013
Copyright 2013 Joongkoo Cho
ii
Dedication
I dedicate this dissertation to my wife Suyeon,
my daughter Yein, my son Joshua,
my parents and parents-in-law,
and my brother and sisters.
iii
Acknowledgements
This dissertation was supported by the University of Southern California’s METRANS
Transportation Center of the U.S. Department of Transportation.
I would like to thank Professor Peter Gordon, James E Moore II, and Harry W
Richardson for their support. I am grateful for the important contribution of PhD student
Weihong Hu, Professor Jiyoung Park, Qisheng Pan, and Mansour Rahimi to this research.
I am also thankful for the support of Simon Choi, Frank Wen, Hsi-Hwa Hu and Sungbin
Cho.
iv
Table of Contents
Dedication ii
Acknowledgements iii
List of Tables vi
List of Figures ix
Abstract xi
Chapter 1: Introduction 1
1.1 Motivation 1
1.2 Research objectives 5
Chapter 2: Literature and Existing Model Review 6
2.1 Truck O-D estimation review 6
2.1.1 Classification of truck O-D estimation methodologies 7
2.1.2 General O-D synthesis methodologies 10
2.1.3 Truck O-D estimation methodologies 18
2.2 Air pollution emissions review 29
2.2.1 Factors affecting air pollution emissions 29
2.2.2 Previous air pollution emissions research review 31
Chapter 3: Methods 35
3.1 Origin-Destination (OD) flows estimation 35
3.2 Transportation impact model 61
3.3 Air pollution emissions model 64
Chapter 4: Scenarios 65
Chapter 5: Model Results 71
5.1 Model results for the Los Angeles MSA 73
5.2 Model results for Los Angeles County 82
v
5.3 Sensitivity analysis 91
5.3.1 Sensitivity analysis for the Los Angeles MSA 93
5.3.2 Sensitivity analysis for Los Angeles County 97
Conclusions 101
Bibliography 105
Appendices 113
Appendix A: Data for OD estimation 113
Appendix B: EMFAC 2007 and MOVES 2010a 128
vi
List of Tables
Table 1: Estimated commodity demand and trade flows attracted to ZIP code areas
of California (2008) 39
Table 2: Estimated commodity demand and trade flows attracted to ZIP code areas
of California for truck mode (2008) 40
Table 3: Los Angeles MSA import components from FAF data 41
Table 4: Los Angeles MSA export components from FAF data 42
Table 5: Los Angeles MSA demand 46
Table 6: Los Angeles MSA demand for truck mode 47
Table 7: Los Angeles MSA demand for truck mode 48
Table 8: Port of Los Angeles and Port of Long Beach throughput demand forecast
(baseline) 66
Table 9: Summary of vehicle miles traveled (VMT) results, Los Angeles MSA 74
Table 10: Comparison of baseline total volumes in the Los Angeles MSA 76
Table 11: Air pollution emissions results for baseline and scenarios in the Los
Angeles MSA 77
Table 12: Proportions of trucks originated from the Los Angeles County 79
Table 13: Percent change of air pollution results by applying scenarios in the Los
Angeles MSA 80
Table 14: Comparison of baseline total volumes in the Los Angeles County 82
Table 15: Vehicle miles traveled (VMT) in the Los Angeles County 85
Table 16: Air pollution emissions results for baseline and scenarios in the Los
Angeles County 87
Table 17: Percent change of air pollution results by applying scenarios in the Los
Angeles County 89
vii
Table 18: Results of sensitivity analysis for the Los Angeles MSA 94
Table 19: Results of sensitivity analysis for the Los Angeles MSA (percent change)
95
Table 20: Results of sensitivity analysis for the Los Angeles County 98
Table 21: Results of sensitivity analysis for the Los Angeles County (percent
change) 99
Appendix Table 1: Metropolitan Areas with Component Counties 113
Appendix Table 2: SCTG Sector descriptions 115
Appendix Table 3: Bridge of vehicle class categories between VIUS and EMFAC 116
Appendix Table 4: Truck use percentages by SCTG Sector 117
Appendix Table 5: Los Angeles MSA import trade proportions 118
Appendix Table 6: Los Angeles MSA import trade proportions for truck mode 119
Appendix Table 7: 2008 California air cargo statistics 120
Appendix Table 8: California sea ports unit: U.S. ton 121
Appendix Table 9: Foreign Import Mode proportion 123
Appendix Table 10: Foreign Export Mode proportion 124
Appendix Table 11: Domestic modes definition 125
Appendix Table 12: Foreign modes definition 125
Appendix Table 13: Reference Case Greenhouse Gas Emissions 127
Appendix Table 14: Air pollutants estimated in EMFAC 2007 130
Appendix Table 15: Emission processes in EMFAC 2007 131
Appendix Table 16: Three modeling modes in EMFAC 2007 132
Appendix Table 17: Total Activity Basis by Process 133
Appendix Table 18: MOVES Source Bin Definitions (other than Model Year
viii
Group) 134
Appendix Table 19: MOVES Source Bin Definitions (Model Year Group) 135
Appendix Table 20: Operating Mode Bin Definitions 135
Appendix Table 21: Comparison of EMFAC2007 and MOVES2010a 136
ix
List of Figures
Figure 1: GHGs emissions by economic sectors, 1990-2008 4
Figure 2: GHGs emissions by transportation modes , 1990-2008 4
Figure 3: Factors affecting air pollution emissions rates from freight movements 31
Figure 4: Truck OD estimation 36
Figure 5: Process of estimating domestic trades for the Los Angeles MSA region 44
Figure 6: Process of foreign import and corresponding domestic shipment
estimation for the Los Angeles MSA region 49
Figure 7: Process of foreign export and corresponding domestic origin estimation
for the Los Angeles MSA region 52
Figure 8: Procedures to estimate VMT based on the estimated truck OD matrix 61
Figure 9: Procedures to estimate air pollution emissions based on the estimated
VMT by truck type 64
Figure 10: I-710 Corridor Project EIR/EIS (Scenario2) 68
Figure 11: Possible development site of inland port at Mira Loma (Scenario 3;
Source: SCAG)(Zipcode:91752) 69
Figure 12: Total truck flows originated from California or destined to California at
2030 (Baseline) 71
Figure 13: Total truck flows originated from California or destined to California at
2030 (to show flows outside CA) (Baseline) 72
Figure 14: Simulated versus Observed (Modified AADTT30) Volumes in the Los
Angeles MSA 76
Figure 15: Percentage of air pollution emissions reduction for scenarios in the Los
Angeles MSA 81
Figure 16: Simulated versus Observed (Modified AADTT30) Volumes in the Los
Angeles County 82
x
Figure 17: Total truck flows around port of LA/LB with port growth at 2030
(Baseline) 83
Figure 18: Total truck flows around port of LA/LB and Mira Loma (Scenario 3) 84
Figure 19: Percentage of air pollution emissions reduction for scenarios in Los
Angeles County 90
Figure 20: Results of sensitivity analysis for the Los Angeles MSA (percent
change) 96
Figure 21: Results of sensitivity analysis for the Los Angeles County (percent
change) 100
Appendix Figure 1: California’s 58 Counties 114
Appendix Figure 2: Sea ports in California 126
Appendix Figure 3: Number of trips per weekday for vehicle class 1 to 4 129
xi
Abstract
Estimating greenhouse gases (GHGs) and other emissions (especially diesel particulates)
is an increasingly important basis for regional policy analysis. According to the EPA
(2010b), the transportation sector contributed 27.2 percent of total GHG emissions in
2008, and 50 percent of these are from truck operations. This research focuses on
estimating GHGs and other emissions (e.g. PM) from freight movements on roads in
California (a prototypical example because of its leadership in air quality policy making)
as well as the concurrent effects of various regulation scenarios. This work addresses
questions of sustainability and environmental policy as well as efficiency in freight
transportation. The research builds on important data sources such as, ZIP code-level
IMPLAN input-output data and the Freight Analysis Framework (FAF) which provides
information on interregional freight movements throughout the U.S. for 2002-2035.
These data are used to estimate interregional trade flows between ZIP code areas. The
estimated interregional trade flows are translated into vehicle miles traveled (VMT) by
applying a user equilibrium model. The estimated VMT in turn are used as inputs to the
emissions model to estimate GHGs and other emissions. Interregional freight flow data
can be an important data source for emission models. The results are useful not only for
estimating GHGs and other emissions based on estimated freight flows, but also for
evaluating environmental impacts of policy alternatives. The analysis shows that
emissions impacts vary by study area as well as by policy. A policy alternative that brings
a significant impact in a specific area may show a trivial impact in a broader region or
vice versa. Also an emissions reduction in one area may be because of emissions
xii
increases in another area. Therefore it is important to simulate possible emissions impacts
by applying a spatially disaggregated model to help decision makers weigh alternatives.
This approach can also be applied to analyze environmental justice concerns when the
emission results are disaggregated into small areas.
1
Chapter 1: Introduction
1.1 Motivation
Evaluating a regional transportation plan (RTP) in terms of air quality impacts is now
essential for local, state and federal governments. This is why the U.S. Environmental
Protection Agency (EPA) has developed the Motor Vehicle Emission Simulator (MOVES)
which is an emissions model at the national and sub-regional levels. The California Air
Resources Board (CARB) has developed the EMFAC model which is an emissions model
for California in which various emissions for major vehicle types are estimated. The
Center for Environment Research and Technology at the University of California,
Riverside, has also developed a Comprehensive Modal Emission Model (CMEM) with
sponsorship from the National Cooperative Highway Research Program (NCHRP) and
the U.S. EPA.
It has been estimated that the transportation sector has contributed over 25 percent of U.S.
greenhouse gases (GHG) since 1990, as shown in Figure 1.
1
Emissions from truck
operations have been increasing steadily ever since 1990 and accounted for more than 50
percent of GHG emissions by 2008, as shown in Figure 2. Learning more about GHG and
criteria pollutants emissions for the trucking mode is a critical aspect of addressing
transportation policy in California as well as other states and regions.
There are many difficulties associated with developing an emissions model. Useable
1
A more detailed list for California is reproduced in Appendix Table 13.
2
data are scarce and reliable parameters are hard to judge. Basically, emissions levels are
estimated by production of emission factors and vehicle activities (CARB, 2007; EPA,
2010a). Therefore, researchers have worked on estimating reasonable emissions factors
parameters, vehicle activities, or interaction between emissions levels and vehicle
activities (Barth and Boriboonsomsin, 2009). The MOVES and EMFAC models have
incorporated such research results and have been widely used by government agencies
and researchers. Although the two models may calculate incorrect emission estimates for
a small region (Barth et al, 1996), the models are useful for identifying trends of
emissions levels for large areas.
MOVES2010a is the latest version developed by EPA. Several improvements have been
made in the latest version (Bai et al.,2008; EPA, 2009). First, MOVES differentiates
vehicle classes by Vehicle Specific Power (VSP) and speed. This is a significant
improvement because different emissions rates within each vehicle class can now be
estimated. Second, the model includes the most up-to-date emissions parameters. The
model also includes vehicle classes consistent with the Highway Performance Monitoring
System (HPMS) so that vehicle activity data can be easily adapted to the model.
EMFAC 2007 has been specifically developed for California. The model includes various
types of vehicle classes, populations of vehicles by classes as well as vehicle model years.
It also includes necessary information such as speed, temperature, and relative humidity
by time of day for each county. This study uses the EMFAC model because the study area
is California.
3
Although EMFAC2007 provides comprehensive data, the key factor, vehicle miles
traveled (VMT), are estimated by the product of vehicle population and vehicle accrual
data. Vehicle population and accrual data are obtained from DMV registration data.
Although DMV registration data provide real information about vehicles, there are
several disadvantages of the approach. First, vehicles registered in an area are not
guaranteed to be operated only in that area. This is an important point for trucks because
trucks usually travel long distances beyond an area. Second, most truck companies that
have their offices in several areas consolidate registration processes in one DMV office.
Third, the data do not provide origin-destination flows so that policy analysis is limited.
The shortcomings may be resolved by using freight flows information because freight
flows are estimated between specific origin-destination pairs by industry sectors.
Consistent sub-state VMT estimates determined via simulation of actual trade flows and
consequent use of the road networks would make emissions models much more useful for
policy analysis.
4
Figure 1: GHGs emissions by economic sectors, 1990-2008
Source: EPA, 2010b
Figure 2: GHGs emissions by transportation modes , 1990-2008
Source: EPA, 2010b
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
40.0%
1990 1995 2000 2005 2006 2007 2008
ElectricPowerIndustry
Transportation
Industry
Agriculture
Commercial
Residential
U.S.Territories
0%
10%
20%
30%
40%
50%
60%
1990 1995 2000 2005 2006 2007 2008
PassengerCars
Trucks
Others except PC
and trucks
5
1.2 Research objectives
The research objective is to simulate air pollution emissions on road networks associated
with truck operations. The study region is California.
There are three sub-goals of the study. First, truck freight flows are estimated between
ZIP code areas based on IMPLAN data at the geographic level of ZIP code areas.
Estimating spatially disaggregated freight flows is essential for this study. ZIP code areas
are the most disaggregated spatial units for estimating freight flows by industry sectors.
The estimation is done based on the IMPLAN input-output data and FAF origin-
destination commodity flow data. Second, a highway network model is developed to
estimate VMT on the network based on the estimated freight origin-destination (OD)
flows. VMT is estimated by truck types. Third, the results from the transportation model
are used as inputs to an air pollution emissions model to determine small-area results.
The model is applied to various policy scenarios.
6
Chapter 2: Literature and Existing Model Review
2.1 Truck O-D estimation review
An origin-destination (O-D) trip table is a two-dimensional matrix where each cell
represents the number of trips between the corresponding O-D zone pair in the road
network for a specific region (Sivanandan, 1991). A truck O-D matrix, accordingly,
represents the distribution of truck trips among a set of O-D pairs. Truck O-D matrices
are central to freight forecasting in metropolitan areas. As a matter of fact, the majority of
the literature assumes all urban freight movements to be conducted by trucking. With this
assumption, truck O-Ds would be identical to freight O-Ds except for the measures used.
Henceforth I will not distinguish between the two terminologies unless necessary.
O-D matrices provide essential information required for transportation planning, control
and management in both passenger and freight sectors. Unfortunately, these matrices are
seldom, if ever, known completely and thus need to be estimated. Initially, freight
modeling largely adapts passenger traffic modeling techniques epitomized by the
classical four-step framework, and truck O-D estimation is no exception. However, it has
been widely accepted (Holguin-Veras et al., 2001; Wisetjindawat et al., 2006; Hunt and
Stefan, 2007; Giuliano et al., 2010; Chow et al., 2010) that freight modeling differs from
its passenger counterpart in the following ways:
• Freight demand is highly disaggregated due to heterogeneity of commodities.
The disaggregation refers to not only geographical but also industry sectorial
and even firm levels.
7
• Agent behavior and spatial interactions play a significant role in freight supply
chain decisions. The selection of shippers for a customer, carriers for a shipper,
routes for a carrier, etc., all stems from and reflects the economic and/or
logistical behavior of and interactions between these agents.
• Commercial vehicles do not frequently take independent direct routes between
origin and destination; instead, there are trip chains/tours where the composite
trips correlate in freight networks. Therefore, truck O-Ds generally cannot be
estimated directly.
• Freight flows are unbalanced in the front haul and the backhaul, which leads
to empty trips that should be considered in high-quality O-D estimation.
The above characteristics, the so-called “multidimensionality” of freight transportation,
add complexity to truck O-D estimation. In fact, sometimes just one of these
characteristics can become an issue as will be seen later. Hence it is not surprising that
freight O-D estimation has received more attention in recent years.
2.1.1 Classification of truck O-D estimation methodologies
Truck O-D estimation methodologies can be classified via various criteria. A first
criterion can be the data involved, which classifies the existing research into two major
groups: (1) direct sampling, and (2) estimation from secondary data sources, i.e., O-D
synthesis.
Direct sampling employs survey data obtained from straightforward survey methods such
as home interviews, questionnaires, license plate surveys, roadside surveys, etc., to set the
8
parameters of classical sampling theory estimators. The main drawbacks of such
techniques are threefold: (1) the variances and covariances of the O-D values depend on
the sampling technique and the estimator adopted, and thus may be unstable; (2) bias is
often introduced in the parameters due to lack of calibration and systematic errors in
survey work; (3) large-scale traffic surveys tend to be time-consuming and labor-
intensive, which can be exacerbated by the dynamic nature of transportation demand. In
the case of freight modeling, there also exists the problem of data reliability because
firms may be reluctant to report various operational details.
Estimation from secondary data sources is an effort to derive the desired O-D matrix by
matching the cells with observed or available secondary data conforming to predefined
rules. Inputs like link volumes (traffic counts) contain the most critical information about
O-D distributions and can be updated readily when dynamics are taken into account (Rios
et al., 2003). This enables such estimation methods to bypass the need for large surveys
and, as a result, they appear attractive and have been intensively studied in the literature.
Without loss of generality, O-D estimation based on secondary data can be interpreted as
the inverse of the traffic assignment problem, where one aims at finding an O-D matrix
that can reproduce the observed traffic or commodity flows on critical links. In highly
dense road networks for detailed urban traffic study, available observations tend to be
limited and unlikely to cover all the links, which in turn poses too few constraints and
underspecifies many potential solutions. Consequently an important question in O-D
estimation is how to define and generate the best solution. In this regard O-D estimation
9
methodologies can be divided into three broad categories: traffic modeling approaches
(gravity models, entropy models, and equilibrium models), statistical inference
approaches, and mathematical programming approaches. Traffic modeling approaches
utilize traffic modeling concepts of information minimization or entropy maximization.
Statistical inference approaches implement the ideas of maximum likelihood, generalized
least squares, or Bayesian inference. Mathematical programming approaches formulate
the estimation problem as linear or nonlinear programming models, and solve them with
efficient algorithms in operations research.
As mentioned in the previous section, freight O-D estimation is similar, but not
equivalent to, passenger O-D estimation. While the above methods work well for the
latter, they are fundamental and inadequate for the former. A pragmatic philosophy is to
customize the above methods to meet the needs of freight O-D estimation. The resulting
research varies with respect to the modeling platform employed. There have been two
major categories from this viewpoint: commodity-based and vehicle-trip-based (Holguin-
Veras and Thorson, 2000). The commodity-based approach models commodity types and
then converts commodity flows to vehicle trips using spatial interaction models and/or
complementary empty trip models, whereas the vehicle-trip-based approach models
vehicle trips directly and explicitly. Both approaches have pros and cons. Freight
transportation naturally arises from human economic and social activities, therefore
commodity-based models can better capture the underlying economic drivers and
behavioral mechanisms. Also, it is convenient to model multimodal systems by tracking
the classified commodity flows. In terms of input and output of the models, however, the
10
commodity-based approach requires large volumes of commodity data for calibration,
and worse, empty trips may not be easy to assess. On the other hand, calibration data for
vehicle-trip-based models is easy to collect, but the approach itself does not reflect cargo
features directly and so multimodal attributes can be a problem. The appropriate approach
to use, of course, should be determined case by case. In fact, freight transportation is so
complex that adapting any existing approach has problems.
Freight O-D estimation approaches also vary regarding whether to account for traffic
evolution over time. Because of data limitation, however, dynamic O-D estimation is not
a part of this study and so will not be discussed.
2.1.2 General O-D synthesis methodologies
Traffic modeling approaches
The first class of traffic modeling approaches involves gravity models. These often
include the idea of estimating trip distributions via a proportional or all-or-nothing
assignment. Depending on how the total number of trips produced at and attracted to
various transportation analysis zones (TAZs) is constrained, this class can be further
divided into three types: the unconstrained gravity model, the singly (either origin or
destination) constrained gravity model, and the doubly constrained gravity model. In such
models, traffic counts are mapped to O-D elements with a function whose parameters can
be calibrated with regression techniques. Generic constraints and objective are flow
conservation and minimization of the differences between observed volumes and
estimated volumes, respectively. Both linear and nonlinear regression techniques have
11
been proposed. The former can be found in Low (1972), Holm et al. (1976), Gaudry and
Lamarre (1978), and Smith and McFarlane (1978); the latter are in Robillard (1975) and
Hogberg (1976).
The main criticism of gravity models is that they enforce gravity patterns on the trip
matrix and so to some extent waste the information contained in the observed traffic
counts. A solution to this problem is entropy models originally developed by Van Zuylen
and Willumsen (1980). This class generally introduces an a priori matrix called the target
O-D matrix to attain an a posteriori O-D matrix based on two ideas: the first is to add as
little external information as possible to the target O-D matrix, i.e., to minimize
information; and the second is to make as much use as possible of information contained
in real counts, i.e., to maximize entropy. Both ideas are equivalent in the sense that the
desired O-D matrix is the most likely one consistent with available real information. The
target O-D matrix can be designed with old data and a reasonable source is an O-D table
for the base year.
Entropy models differ by the assignment rules employed, proportional (Willumsen, 1978;
Van Zuylen and Willumsen, 1980) or equilibrium-based (Nguyen, 1977; Jornsten and
Nguyen, 1979; LeBlanc and Fahrangian, 1982). Here is how the proportional assignment
works: each link flow is divided proportionally among its incident O-D pairs, which
enables the calculation of the probability that this portion of flow comes from a specific
O-D pair; one can then search for the a posteriori trip matrix that maximizes the overall
probability of reproducing the observed traffic counts. One problem with this approach
12
lies in its limited application to congested networks, where the existence of bottleneck
queues may invalidate the reliability of fixed proportions. Although Fisk (1988) suggests
an extension to the case of congestion by imposing user-equilibrium constraints, the
resulting problem is hard to solve due to the nonconvex structure of the variational
inequalities, and hence this contribution is only of theoretical interest. Another problem
concerns the role of the target matrix, which is simply to provide an initial condition and
thus the traffic counts are given priority. But both the target matrix and the observed
traffic counts present some level of uncertainty in reality, therefore it is not always a
strong assumption that the observed traffic counts are more trustable and the above
approach would not induce larger errors than otherwise. As a matter of fact, Brenninger-
Gothe et al. (1989) show that a weighting is made, explicitly or implicitly, in almost
every estimation process, and there is no ideal way to specify the weights in proportional
assignments.
As an alternative, equilibrium methods seek the user-optimally assigned matrix based on
the so-called “Equilibrium Principle” which assumes each user to behave rationally and
non-cooperatively so that his/her transportation cost can be minimized (Wardrop, 1952).
Nguyen (1977) is the first to formulate the equilibrium based O-D estimation problem.
With this formulation, the solution will reproduce the observed traffic counts if these
observations are at equilibrium, but under-specification remains an issue. Following up
this work, Jornsten and Nguyen (1979) propose a formulation of entropy maximization
that does not require a target O-D matrix. In the same spirit of seeking for a unique O-D
matrix, LeBlanc and Fahrangian (1982) formulate a least squares model that obtains the
13
O-D matrix not only user-optimal but also deviates least from a known target matrix.
Contrary to the proportional approach, equilibrium models determine the route choice
proportions endogenously and hence can be and have been widely adopted for congested
networks. An elaborate review of this approach is presented in Yang et al. (1994).
Statistical inference approaches
These models jointly use traffic counts and the target matrix to estimate the desired O-D
matrix, and a common characteristic is to trade off the aforementioned sources. The main
advantage is that they consider the stochastic nature of the problem directly, and possess
the large sample properties of (asymptotic) unbiasedness, normality and efficiency.
The maximum likelihood approach assumes the target O-D matrix and the observed
traffic counts to be observations of two independent random vectors. The motivation is to
maximize the probability of realizing the target O-D matrix and the observed traffic
counts conditional on the O-D matrix to be estimated. A representative example is Spiess
(1987). In this paper, a full target matrix is obtained by sampling Poisson variables and
three optimization models are suggested for estimation. All the models have similar
objective functions, but differ in constraints: the first model assumes proportional
assignment and treats link flow consistency as constraints; the second model is doubly
constrained, i.e., it incorporates both trip assignment and trip distribution constraints; the
third model eliminates the consistency requirement. The first two models can reproduce
the observed traffic counts but are computationally demanding, whereas the simpler third
model may reduce to an approximate maximum likelihood model if the assumption of
14
mutually independent traffic counts is weakened. Efficient algorithms are designed for all
three models. Other works include Geva et al. (1983), Watling and Grey (1991), and
Watling and Maher (1992). All have the nice property of definite feasibility regardless of
the target matrix, which is not guaranteed for entropy maximization methods.
The generalized least squares approach views the target O-D matrix and the traffic counts
as stochastic response variables to the desired O-D matrix. The errors associated with the
target O-D matrix and the traffic counts reflect the respective dispersion, and are assumed
to be random and mutually independent. Given the dispersion matrices as parameters, an
estimator can then be formed by minimizing the weighted mean square errors depending
on the dispersion matrices. Cascetta (1984) derives expressions for the mean and variance
of the estimator when nonnegativity constraints on the estimated O-D matrix are not
binding. The resulting O-D estimation is demonstrated to be better than the maximum
entropy approach even if the dispersion matrices are not exact but heavily approximated.
The relative independence of the results from the dispersion matrices may be explained
by Cascetta (1984) and Bierlaire and Toint (1995), whose experiments show that the
models appear much more sensitive to variations and inaccuracies in the target O-D
matrix and the traffic counts than to values of the parameters. Bell (1991) addresses the
issue of active nonnegativity constraints by taking the corresponding Lagrange
multipliers to the objective function. Yang et al. (1992) extend the basic model to a
bilevel programming model combining generalized least squares and equilibrium
assignment. The observed traffic counts are not required to be at equilibrium in this paper,
and several special cases of the parameter values are identified to coincide with Nguyen
15
(1977) and Fisk (1988). The relations between the generalized least squares estimator and
other estimators have been discussed by Dolby (1972) and Bell (1984), and the main
findings are twofold: (1) the generalized least squares estimator is actually the maximum
likelihood estimator if the dispersion matrices are multivariate normally distributed; (2)
the generalized least squares approach can provide a good approximation of the minimum
information estimator proposed by Van Zuylen and Willumsen (1980).
When considering the equilibrium assignment, there always exist such questions as
whether the observed traffic counts are of the user-equilibrium pattern, what effects that
would make, and how to convert observations in disorder to be at equilibrium. Yang et al.
(1994) partly answer these questions by claiming that traffic counts for feasible
underspecified equation systems are at equilibrium, but overall no standard procedure to
adjust arbitrary traffic counts has been known.
The Bayesian inference approach treats the target O-D matrix as a prior distribution of
the estimated O-D matrix, the observed traffic counts as sample information for the
likelihood distribution, and the desired O-D matrix as the posterior distribution. Given the
target O-D matrix and the observed traffic counts, one can then use Bayes’ rule to
calculate the O-D estimations. Maher (1983) examines a logarithm expression of the
Bayesian equation and verifies the equivalence of this approach and the entropy method
when the posterior distribution is multivariate normal. Proportional assignment is
assumed in this work. According to Cascetta and Nguyen (1988), the Bayesian inference
approach resembles the maximum likelihood approach and the generalized least squares
16
approach as well, except that the roles of the target O-D matrix differ: for the Bayesian
approach, it is a random vector associated with the posterior distribution, whereas for the
other two approaches, it is the parameter set corresponding to the sampling likelihood
function.
Mathematical programming approaches
Equilibrium based models are credited for their applicability to congested networks, but
the majority of this class have a nonlinear and bilevel structure that determines the O-D
estimation and the equilibrium assignment on two interconnected levels. Such a
complicated structure brings about computational difficulty and, accordingly, necessitates
exclusively designed techniques. Mathematical programming methods have found their
place in this field.
One approach is to apply heuristics or gradient algorithms and iterate between the two
levels until a predefined convergent condition is satisfied. Some studies previously
discussed follow this way, including Jornsten and Nguyen (1979), LeBlanc and
Farhangian (1982), and Fisk (1988). Jornsten and Nguyen (1979) utilize Benders
decomposition and test three small numerical examples. LeBlanc and Farhangian (1982)
solve the lower level problem, i.e., the equilibrium based assignment problem, by the
Frank-Wolfe method. Fisk (1988) sketches out a solution procedure, but does not report
any applications. Relevant work can also be found in Spiess (1990), Drissi-Kaitouni and
Lundgren (1992), Florian and Chen (1993), Chen (1994), Yang (1995), Codina and
Barcelo (2004), and Lundgren and Peterson (2008). Specifically, Spiess (1990) designs
17
an approximation algorithm with proportional assignment which, though not convergent,
works satisfactorily and is later adopted in a commercial transportation planning system.
Drissi-Kaitouni and Lundgren (1992) suggest general descent algorithms as a proper
resort for large-scale networks. Florian and Chen (1993) show that a descent direction of
Gauss-Seidel type may produce closer solutions. Chen (1994) analyzes an augmented
Lagrangian method as well as a heuristic Gauss-Seidel type method, which are
demonstrated to suit small networks and large networks, respectively. Yang (1995)
studies two heuristic algorithms that converge fast, namely a heuristic iterative algorithm
and a sensitivity analysis based heuristic algorithm. Codina and Barcelo (2004) develop a
subgradient method for non-differentiable problems. Lundgren and Peterson (2008) adopt
a projected gradient method where the search direction is computed by approximating the
Jacobian matrix for the link flows. The order approximation of the Jacobian matrix is
done by solving a set of quadratic programs.
Another approach is to consider computationally tractable formulations, mainly referring
to linear programming. Colston and Blunden (1970) are among the first to study O-D
distributions with linear programming methods. Unfortunately, the attempt fails to
perform well as the applications to general transportation problems do in practice. No
successful trial has come up until the 1990s. Sherali et al. (1994a) formulate the problem
as a path-based linear model, where the objective coefficients are defined as the time
impedances or costs on the routes corresponding to each O-D pair, or a constant number
big enough for the rest routes; the constraints include equilibrium and nonnegativity. The
optimal solution to this problem, if it exists, is shown to be of user-equilibrium pattern
18
and thus can reproduce the observed flows. Sherali et al. make two successive
modifications of the preliminary model to accommodate inconsistent flow data and prior
trip tables, respectively. Given that there are exponentially possible path variables,
column generation techniques are employed to implicitly enumerate all feasible solutions.
The subproblems are essentially shortest path problems and so can be solved efficiently.
The model assumes a complete set of observed flows for the entire network, and hence
there naturally arises the question how to obtain the desired O-D matrix in case of
missing link flows. Improved versions in this respect appear in Sherali et al. (1994b) and
Sherali et al. (2003), where the objective coefficients are updated by solving both linear
and nonlinear subproblems iteratively, or approximating the nonlinear model with a
sequence of linear models. The efficiency of such approaches has been tested on real road
networks.
2.1.3 Truck O-D estimation methodologies
Early studies on truck O-D estimation generally resemble passenger O-D estimation and
follow the methodologies previously discussed. For instance, gravity models can be
found in Meyburg (1976), Ogden (1978), Swan Wooster (1979), Southworth (1982),
Ashtakala and Murthy (1988), and Tamin and Willumsen (1988); mathematical
programming models can be found in Gedeon et al. (1993) and List and Turnquist (1994);
and heuristic solution techniques can be found in Tavasszy et al. (1994) and Al-Battaineh
and Kaysi (2005).
The problem with early studies is that the unique features of freight transportation are
19
largely ignored. Taking gravity models for example, the core assumption of these models
is the monotonically decreasing pattern of trip length distribution, which conforms to the
rational behavior of passenger transportation but deviates from reality in the case of
freight transportation (Jack Faucet Associates, 1999). The complexity of freight modeling
has motivated the development of exclusive models and methods.
Data extraction methods
Secondary freight flow data generally have three problems: first, different data sources
reveal different aspects of freight flows, but hardly can any single source describe the
complete flows regarding an area; second, they are not equally available for various
modes; and third, most are at an aggregate level whereas the desired analysis requires
more disaggregate data.
Giuliano et al. (2010) attempt to address the first two issues for commodity-based models.
The underlying logic is to estimate regional commodity-specific O-D matrices by
integrating international, interregional and intraregional trip attractions and productions.
The suggested data sources include IMPLAN, CFS, WISER, WCUS, and ITMS. For any
area, its flow set can be divided into five parts, namely international import, international
export, domestic import, domestic export, and intraregional flows. Since IMPLAN
contains information about import/export totals by industrial sector, to get international
and interregional flows one can first derive each flow part proportionally from IMPLAN,
and then assign the domestic import/export flows to mode with CFS and ITMS, and
international import/export flows with WCUS and WISER. The subsequent question is
20
how to account for the intraregional flows. To do this the authors generate intraregional
productions and attractions utilizing a regional input-output transactions table as well as
employment data for small areas. The approach is demonstrated applicable to a
geographic level as fine as traffic analysis zones. Once the interregional flows and
intraregional productions and attractions are obtained, flows are converted from dollars to
tons. Intraregional trips can be distributed together with a further conversion to truck trips
with conventional gravity models. Since an implicit assumption is that intraregional flows
are conducted by truck, the total number of non-truck trips generated by some baseline
model may well be used as a control. The distribution of interregional trips is confined to
a limited number of zones in the region to reflect their import/export shares, which are
based on attracted trips at internal TAZs.
Traditionally, the third issue mentioned above is solved by rough spatial disaggregation,
i.e., by factoring both the rows and columns of a given aggregated O-D matrix
simultaneously and directly. The row and column split coefficients can be determined
with various sources, socioeconomic data, trip generation equations, disaggregated VMT,
and individual traffic counts, to name a few. Easy to implement as it is, this approach
ignores the possible effect of special disaggregate-level interactions that are hidden or
averaged out at the aggregated level. To overcome this problem, Horowitz (2009)
proposes a new disaggregation method with traffic counts as the secondary data source
and Fratar biproportional least-squares models as the estimation technique. Six models
are developed to satisfy different needs. In the case of perfectly aggregated O-D
information, the model seeks the solution that deviates from ground counts least and
21
matches the given O-D matrix exactly. In the case of approximately aggregated O-D
information, flow conservation constraints are moved to the objective function and thus a
relaxation problem is formed compared with the previous case. A variation for these basic
models considers the effects of trip utility or spatial separation, and logit gravity models
of destination choice are introduced to calculate the correction coefficients and thus
enhance the objective functions. The other two variations for the case of approximate
aggregated O-D information incorporate link-to-link flows and special zone-to-zone
flows, respectively. Further variations such as factor bounds and congestion can also be
handled by adding new constraints or combining equilibrium models. The resulting
models are all nonlinear optimization problems and an iterative bilevel algorithm is
designed for solution. The method can be applied to both commodity based and vehicle-
trip based approaches.
A more modeling-specific contribution but in the same spirit of data saving, Sivakumar
and Bhat (2002) introduce an intuitive fractional split distribution model which later
enlightens the development of a trip-chaining model in Wisetjindawat et al. (2006). The
main difference from earlier studies is that this framework does not require production
and consumption levels at each geographical analysis zone to be determined
simultaneously in the commodity generation step; instead, consumption data suffices, and
the allocation of production levels (in fractional form) at the associated origins is left to
the fractional split distribution model, which describes the relationship between the
desired fractions and zonal explanatory variables as normalized multinomial logit
functions. Each relationship function is composed of two parts: a composite size measure
22
which represents the number of elemental commodity production points within a specific
zone, and a composite impedance measure which represents the marginal deterrence
between a specific O-D pair. The parameters involve both scalars and vectors, and can be
obtained by maximizing a set of quasi-likelihood functions. The fractional split approach
saves production data but captures the essence of demand-driven freight movements and
hence appears more trustable than gravity models. Indeed, Sivakumar and Bhat (2002)
showed an empirical application that produces better results than the gravity models. A
drawback is the limited application to interregional (statewide) commodity flow analysis.
Trip-chaining and behavioral models
One major concern of conventional O-D estimation methods is that they confine analysis
to a zonal level, which challenges the incorporation of agent behavior and spatial
interactions. Trip-chaining, a result of the underlying logistical decisions, tends to be
ignored as well. A plausible improvement can be agent-based analysis where the smallest
analysis unit is an individual firm rather than a geographical zone.
McFadden et al. (1986) is perhaps the earliest to work on agent behavior for commodity
flows. The behavioral element of the proposed model is in essence a logistics model that
jointly determines mode choice and shipment size by minimizing inventory costs.
Abdelwahab and Sargious (1985) use a discrete choice model for the same purpose. A
variation of this model is designed by Holguin-Veras (2002), where the formulation is
discrete-continuous in that shipment size variables are treated as continuous. A common
limitation of these models is that they merely account for the interaction between two
23
freight agents. Boerkamps et al. (2000) illustrate the procedures to incorporate the
interactions among all agents but no formulations have emerged.
Wisetjindawat et al. (2006) shine a light on comprehensive behavior modeling.
Analogous to conventional studies, they first generate the production amount of a
commodity for each shipper and the consumption amount of a commodity for each
customer. These amounts constitute the input of the core model -- the distribution model,
which then calculates the commodity flow between each shipper-customer pair by
multiplying the total consumption of a commodity of a customer by the fraction of
him/her purchasing the commodity from a shipper. The fraction can be decomposed into
three parts, namely the distribution channel probability, the zone choice probability, and
the shipper choice probability. The distribution channel probability reflects the supply
chain structure of freight flows and can be determined from empirical data. The zone
choice probability reflects the spatial interaction in location choice and can be obtained
via a spatial mixed logit model. The shipper choice probability reflects the purchasing
relationship between shipper-customer pairs and can be estimated ideally from survey
data or approximately by weighting the production amount based on utility functions.
Due to the complexity of the model, parameter calibration is conducted with simulated
maximum likelihood techniques.
In a supplementary paper, Wisetjindawat and Sano (2003) further develop a framework
for conversion of the commodity flows to vehicle movements. Three steps are taken
sequentially: first, the delivery lot size and frequency are determined for each shipper-
24
customer pair and each commodity with an unconstrained total cost minimization model;
second, the carrier and vehicle types are selected for each shipper-customer pair and each
commodity with a utility-based nested logit model; and third, the delivery route is chosen
for each shipper with a vehicle routing model constrained by both capacities and time
windows. Tour selection can also be found in Donnelly (2007), where vehicles are first
allocated and filled according to average payload weight and traveling salesman
algorithms are then utilized for optimization.
More recently, the consideration of trip chains has led to a new family of truck O-D
estimation approaches as an alternative to the four-step framework – tour-based
microsimulation where a tour is the smallest analysis unit. Relevant work can be found in
Gliebe et al. (2007), Hunt and Stephan (2007), and Wang and Holguin-Veras (2008,
2009). Gliebe et al. (2007) creates an intra-urban commercial vehicle model that
incrementally builds tours and reproduces observed traveling patterns. Hunt and Stephan
(2007) design a multi-modal, multi-sector, agent-based framework that covers attributes
including tour generation, vehicle and tour purpose, tour start, next stop purpose, next
stop location, and stop duration. Wang and Holguin-Veras (2008, 2009) propose an
efficient discrete choice model to generate a candidate tour set, a heuristic algorithm to
select the desired tours, and an entropy maximization formulation to determine the flows
along each tour.
All the above approaches are disaggregated except for Wang and Holguin-Veras (2008,
2009). The most outstanding advantage is that they incorporate the complex relationships
25
involved in freight transportation at a micro level and thus is responsive to small-scale
changes, whereas some obvious disadvantages may be the intensive data and
computational efforts required for calibration, validation, and solution.
Empty trip models
Empty trip models are usually designed to overcome the inability of implicitly
incorporating empty trips in commodity-based approaches. The evolution of such models
has gone through three phases: the naïve proportionality model, models that assume a
direct correlation of empty trips in one direction to commodity flows in the reverse
direction (Noortman and van Es, 1978; Hautzinger, 1984), and models that take trip
chaining into consideration (Holguin-Veras and Thorson, 2003; Holguin-Veras and Patil,
2008).
In the naïve proportionality model, the average payload (tons per trip) ratio is assumed to
be constant for any trip (loaded or empty) produced per unit of commodity flow,
hence the total number of vehicle trips between an O-D pair can be expressed as the
commodity flow (in trips) divided by this constant ratio. Simple and broadly applied
though it is, this model is problematic since the number of empty trips is assumed to
merely rely on the commodity flow in the same direction, which implies empty trips
would remain unchanged when the reverse flow changes, but of course, contradicts real
observations.
A first improvement is achieved in Noortman and van Es (1978), where the number of
empty trips is obtained by multiplying the commodity flow in the reverse direction by a
26
constant. This leads to a more reasonable formulation that relates the total trips between
an O-D pair to the commodity flows in both directions. A by-product of this model is that
the total trips between the complementary O-D pair (the pair obtained by exchanging the
origin and the destination of a pair) may deviate significantly from that between the O-D
pair in question, whereas empirical evidence shows a consistency between the two even
in extreme cases. In light of this, Hautzinger (1984) makes a second improvement by
introducing bi-directional empty trip ratios to the model. The ratios are non-constant, but
can be calculated with positively related functions of commodity flows in the reverse
direction. In this way equality of the total trips is guaranteed. There exists a problem in
both improvements, though: trip chains have been ignored.
Holguin-Veras and Thorson (2003) introduce the concept of order of a trip chain model
which sets the basis for developing more complicated models and unifies the above
models as well. The concept refers to the number of transient stops before reaching the
final destination in a commodity flow. By this definition the above models are all zero-
order, whereas the one developed by Holguin-Veras and Thorson is first-order. For
simplicity, the summation of all higher-order empty trips is approximated by multiplying
the expected first-order empty trips by a constant for all O-D pairs, and the zero-order
empty trips are expressed as the same function in Noortman and van Es’s model. As a
result, the desired number of empty trips between an O-D pair is a linear function of the
given commodity flows with four types of parameters: the constant for higher-order
empty trips, the probability of a zero-order trip chain, the probability of the destination
chosen as the next stop in a tour, and the probability of not getting a load. Two ways to
27
calibrate the parameters are suggested: an unconstrained search that finds the parameters
fitting a given data set best, and an error minimization model constrained by replication
requirements on specific measures. An analysis of the relationships between the first-
order model and the previous models reveals that Holguin-Veras and Thorson’s model
mediates between Noortman and van Es’s model and Hautzinger’s model regarding the
difference of a commodity flow from the reverse flow.
Holguin-Veras and Thorson’s empty trip model is later integrated into doubly constrained
gravity models for freight O-D estimation (Holguin-Veras and Patil, 2007, 2008). Three
versions are developed: single-commodity, multi-commodity with parameters calibrated
by minimizing total squared truck traffic errors, and multi-commodity with parameters
calibrated by minimizing total squared errors in both loaded and empty link volumes.
Comparative experiments confirm the superiority of models incorporating empty trips
over otherwise and the superiority of the multi-commodity formulation over the single-
commodity formulation in their ability to reproduce the observed traffic counts.
This research use secondary data sources to estimate the truck O-D matrix. Traffic
modeling approach is applied, including a doubly-constrained gravity model and an
equilibrium model, in terms of the O-D estimation methodology. This approach is
commodity based. The estimated O-D matrix is adjusted by minimizing the differences
between estimated volumes from secondary data and observed volumes which are
AADTT obtained from FAF data. When the O-D matrix estimated from the secondary
data are adjusted with AADTT, the adjusted truck O-D matrix includes empty truck trips
28
because AADTT obtained from FAF data includes empty trips although empty trips are
not estimated separately.
29
2.2 Air pollution emissions review
2.2.1 Factors affecting air pollution emissions
Air pollution emissions caused by transport activities can be grouped into two types:
greenhouse gasses (GHGs) and other pollutants. GHGs include Carbon dioxide (CO
2
),
Methane (CH
4
), and Nitrous Oxide (N
2
O) from fuel combustion and F-gases (fluorinated
gases) from vehicle air conditioning (Kahn Ribeiro et al., 2007). Other pollutants are total
gaseous Hydrocarbons (HC), Carbon Monoxide (CO), Oxides of Nitrogen (NOx),
Particulate matter (PM
10
, PM
2.5
), and Oxides of sulfur (SOx) (CARB, 2007; EPA, 2010a).
Efforts have been made to estimate GHGs and other pollutants caused by transport
activities. Estimation processes reflect an understanding of which factors affect emissions
rates. As shown in Figure 3, air pollution emissions rates from freight movements in an
area are affected by three prominent factors:
• V olumes and types of production
• Ambient conditions
• Vehicle operating characteristics
The volumes and types of production determine the amounts and types of freight flows
within and among surrounding areas. For example, agricultural products and related
materials would be the types of freight transported in and out of a rural area that consists
mostly of farms. If there are many productions in an area, freight flows would likely
increase. Amounts and types of freight flows will affect the number of transport activities
30
and types of transport equipment used which, in turn, affect the amount of air pollution
emissions.
Ambient conditions such as grades of roads, temperature and relative humidity of an area
are important factors determining air pollution emissions rates (Lents et al., 2011). As
grades of roads change, vehicles accelerate and decelerate accordingly resulting in
changing emission rates. When vehicles go uphill, engines generate more power at low
speeds causing imperfect combustion which creates more exhaust emissions. When
vehicles go downhill, brakes would be used more frequently resulting in more emissions
of particulate matter (PM). Ambient temperature and relative humidity are important
factors related to evaporative emissions.
Vehicle operating characteristics such as vehicle age, types of air pollution control
devices equipped with the engine, driver's habits, and congestion levels are important
factors determining air pollution emissions rates.
31
Figure 3: Factors affecting air pollution emissions rates from freight movements
Modeling practices reflect the current understanding of the relationships between
emissions rates and the three factors mentioned above. Two models have been publicly
adopted for use in the U.S. One is EMFAC and the other is MOVES. These two models
have various similarities and dissimilarities. APPENDIX B includes reviews of the
EMFAC and MOVES models.
2.2.2 Previous air pollution emissions research review
In the 1990s, there were several tests to estimate vehicle emission parameters. Equipment
such as data-logger or global positioning system (GPS) was installed to collect data from
vehicle operations (Magbuhat, S. and J. Long, 1996; Benjamin, M. and J. Long, 1995).
Data were collected to determine distributions of vehicle miles traveled (VMT), trips,
temperature, and speed during weekdays and weekends. Grades and other loads effects
Air pollution
Emissions from
freight
movements in an
area
Volumes and types of
production
Ambient conditions
Vehicle operational
characteristics
32
on emissions were analyzed (Cicero-Fernandez, P. and J. R. Long, 1995, 1996). Benefits
on emission rates of on-board diagnostics and inspection/maintenance (I/M) were studied
(Patel, D and M. Carlock, 1995). Based on the research results mentioned above, the
California Air Resources Board (CARB) developed an air pollution emissions model
called EMission FACtors (EMFAC).
Similarly, in the early 2000s, U.S. EPA released several study results. These studies
showed how emission rates were estimated for second-by-second vehicle movements
(Nam, E. K. 2003; North Carolina State University, 2002). Based on the study results,
EPA developed MOVES. Both EMFAC and MOVES provide parameters and necessary
input data for passenger cars and trucks. Therefore researchers focused on estimating
VMT, which is a primary input data for the two models.
Efforts have been made to estimate VMT more accurately. Four methods have been
applied to estimate truck VMT in sub-state areas. First, a travel demand model has been
used to estimate VMT from passenger car travels (Hatzopoulou and Miller, 2010). The
travel demand model estimates origin-destination flows based on socio-economic data.
Then VMT is estimated by applying a trip assignment algorithm on road networks. Truck
VMT is calculated by multiplying truck percentage to the estimated total VMT. The
method is well developed for personal trips but may not be appropriate for freight trip
estimation because of data limitations. Second, diesel fuel sales data has been used to
estimate truck VMT (Harley et al., 2004). Since fuel sales data includes passenger vehicle
and truck, proportion of truck counts were multiplied with fuel sales data to get truck
33
VMT. The method can be useful for validating emission inventory in a specific area. But
the application would be limited to large urban area.
Third, a top down disaggregation approach has been applied. FHWA developed the
Freight Analysis Framework (FAF) database. FAF contains 114 domestic zones and 17
ports of entry for the U.S. Forty-three commodity flows transported by trucks are
provided. After the Freight Analysis Framework (FAF) data were released, efforts have
been made to disaggregate the state level flows into sub-state areas (Anderson et al., 2008,
2009; Rowinski et al, 2008; Opie et al., 2002; Viswanathan et al., 2008; Harris et al.,
2009). Then, assignment algorithms were applied to estimate VMT, based on
disaggregated flows.
Fourth, the traffic counts method has been widely used and may be the most common
approach to forecast VMT. Truck counts are collected at sample roads. Truck VMT is
calculated by multiplying average annual daily truck traffic (AADTT) to the length of
roads or multiplying total VMT to the average truck percentage. Sub-regional estimates
are obtained by applying extrapolation. Historical traffic count data are used to calculate
growth factor and the growth factor is applied to estimate future VMT. The method is
efficient and appropriate for statewide estimation but it has limited capacity at the sub-
state level.
The four methods have limited capability for sub-state truck VMT estimation. This is
because of lack of data. Recently however, the IMPLAN input-output data at ZIP code
have been released. IMPLAN provides commodity flows for ZIP codes. We can now
34
obtain truck flows among ZIP code areas by applying a gravity model (Alam et al., 2007).
Truck flows indirectly estimated from input-output data may not reflect real truck flows
on roads. The problem can be adjusted by comparing the estimated truck flows with
observed truck counts on sample areas. A new approach can be used to obtain VMT
based on commodity flows and traffic counts.
35
Chapter 3: Methods
This research combines input-output data, a highway network model and an air pollution
emissions model. For California, EMFAC 2007 provides vehicle population and VMT
data. However, the data do not provide origin-destination flows so that opportunities for
policy analysis based on transportation network performance are limited. Freight flows
information can be an alternative basis for estimating VMT in local areas (Alam et al.,
2007). Several steps are needed to estimate sub-state freight flows from IMPLAN ZIP
code area input-output data.
3.1 Origin-Destination (OD) flows estimation
Estimating truck OD flows at the sub-state level is the first step for estimating truck VMT.
IMPLAN 2008 ZIP code level data are the basis for estimating truck OD flows among
ZIP code areas in California and between California and other States. Figure 4 shows the
necessary steps.
36
Figure 4: Truck OD estimation
IMPLAN data provide commodity outputs and demands in an area. The California data,
for example, provide the following information:
• Total Commodity Output produced in California and Total Commodity Demand
attracted to California.
• Local Supply which shows commodities supplied by producers located in
California.
• Foreign Exports and Foreign Imports
• Domestic Exports and Imports.
Similarly IMPLAN ZIP code data provide the following information:
Trade data ($) at ZIP code areas from IMPLAN
Truck O-D matrix among ZIP code areas in California and between
California and other States
Trade proportions among
metropolitan areas from
Freight Analysis
Framework (FAF)
Doubly constrained
gravity model
Truck per ton
conversion factor from
Vehicle Inventory Use
Survey
Ton per dollar conversion
factor from FAF
37
• Total Commodity Output produced in the ZIP code and Total Commodity
Demand attracted to the ZIP code.
• Local Supply which shows commodities supplied by producers located in the ZIP
code.
• Foreign Exports and Foreign Imports.
• Domestic Exports and Imports.
To estimate trade flows between ZIP code areas, individual ZIP code data for California
are first combined to estimate total local supply and domestic commodity flows by ZIP
code areas. Then ZIP code data are combined into the four major MSA areas and one
remainder area made up of other state MSAs, according to the spatial definitions of the
Freight Analysis Framework (FAF). The ZIP code data are aggregated into FAF areas for
validation purposes. There are few data sources to validate trade flow estimation. The
Commodity Flows Survey (CFS) and FAF are two of the few sources. These two use the
same definitions of geographic areas. The following are the types of data that can be
obtained from IMPLAN the model at the California and MSA levels:
California
• Total Commodity Output produced in ZIP code areas and Total Commodity
Demand attracted to ZIP code areas in California.
• Foreign Exports and Foreign Imports by ZIP code areas in California.
• Local Supply which shows commodities that are produced and consumed at the
same ZIP code areas in California.
38
• Domestic Exports of ZIP code areas and Domestic Imports into ZIP code areas.
Domestic trades also include flows between ZIP code areas.
MSA and remainder of MSAs area
• Total Commodity Output produced in ZIP code areas in each MSA area and
remainder of MSA areas and Total Commodity Demand attracted to ZIP code
areas in each MSA area and remainder of MSAs area.
• Foreign Exports and Foreign Imports by ZIP code areas.
• Local Supply which shows commodities that are produced and consumed in the
same ZIP code areas in each MSA area as well as remainder of MSAs area.
• Domestic Exports of ZIP code areas and Domestic Imports into ZIP code areas.
Domestic trades also include flows between ZIP code areas.
Table 1 shows the aggregated total demand in California and Table 2 shows the
aggregated total demand in California for the truck mode. Truck mode proportions
obtained from FAF data are applied to generate estimated demand for the truck mode.
39
Table 1: Estimated commodity demand and trade flows attracted to ZIP code areas of
California (2008)
Units: $ Million
SCTG Total Commodity Demand
Foreign
Imports
Local Supply
Domestic Imports
1 2,070 34 664 1,372
2 4,164 69 66 4,029
3 15,232 1,619 3,076 10,536
4 22,496 575 2,683 19,238
5 12,424 1,314 1,138 9,972
6 10,678 201 1,734 8,743
7 58,155 2,193 6,177 49,785
8 12,698 1,608 156 10,933
9 6,853 76 68 6,709
10 39 1 0.34 38
11 703 17 10 676
12 1,185 23 10 1,151
13 1,055 593 3 459
14 1,163 121 48 993
15 2,663 179 1 2,482
16 116,499 58,546 2,075 55,879
17 47,107 1,712 6,580 38,814
18 18,730 681 2,616 15,433
19 18,889 727 2,516 15,645
20 29,532 3,683 2,700 23,149
21 52,516 4,053 10,330 38,133
22 1,164 362 62 740
23 18,133 886 2,087 15,160
24 46,259 4,541 4,443 37,275
25 1,130 1 454 675
26 13,326 2,056 1,579 9,690
27 11,877 1,399 1 10,477
28 4,863 399 57 4,407
29 19,782 1,759 1,213 16,810
30 32,655 16,064 1,245 15,346
31 17,424 1,433 625 15,366
32 27,957 6,910 1,087 19,960
33 29,305 3,149 1,742 24,414
34 53,994 10,086 7,242 36,666
35 224,568 32,317 56,591 135,660
36 60,988 17,921 4,818 38,249
37 19,696 861 3,458 15,376
38 29,064 3,145 5,696 20,223
39 16,328 3,056 2,342 10,930
40 37,361 14,126 9,116 14,119
41 3,006 82 877 2,047
Total 1,103,730 198,582 147,389 757,760
Data: 2008 IMPLAN model
Local Supply= ∑ 𝐷 𝐷𝐷 𝐷𝐷 𝐷 𝐷 𝐷 𝐶 𝐷𝐷 𝐷 𝐷 𝐶 𝐷 𝐷 𝐶 𝑂𝑂 𝐷 𝑂 𝑂 𝐷 𝑁 𝑖 = 1
𝑓 𝑓 𝐷𝐷 𝑍𝑍 𝑍 𝐷 𝐷𝐶 𝐷 𝐷 𝐷𝐶 𝐷 𝑚
Domestic Imports= Total Commodity Demand- Foreign Imports- Local Supply
40
Table 2: Estimated commodity demand and trade flows attracted to ZIP code areas of
California for truck mode (2008)
Units: $ Million
SCTG Total Commodity Demand Foreign Imports Local Supply
Domestic
Imports
1 2,058 25 664 1,369
2 3,584 41 66 3,477
3 14,899 1,440 3,076 10,382
4 21,376 396 2,683 18,298
5 12,131 1,217 1,138 9,776
6 10,444 191 1,734 8,519
7 56,062 1,999 6,177 47,885
8 10,990 1,408 156 9,426
9 6,753 61 68 6,625
10 38 1 0.34 37
11 676 13 10 653
12 1,009 20 10 978
13 929 496 3 430
14 1,035 91 48 896
15 1,156 144 1 1,011
16 60,860 32,619 2,075 26,166
17 31,840 1,023 6,580 24,236
18 11,637 362 2,616 8,658
19 12,147 650 2,516 8,981
20 25,474 2,673 2,700 20,102
21 43,471 3,264 10,330 29,877
22 1,042 277 62 702
23 16,723 649 2,087 13,987
24 42,821 4,097 4,443 34,281
25 1,120 1 454 665
26 12,530 1,818 1,579 9,133
27 10,304 1,285 1 9,017
28 4,604 343 57 4,204
29 16,978 1,504 1,213 14,261
30 28,584 14,114 1,245 13,226
31 16,306 1,244 625 14,437
32 24,425 5,807 1,087 17,531
33 26,877 2,745 1,742 22,389
34 50,102 7,591 7,242 35,269
35 182,646 22,493 56,591 103,562
36 55,644 16,726 4,818 34,100
37 13,007 686 3,458 8,862
38 22,757 2,211 5,696 14,850
39 15,701 2,760 2,342 10,598
40 32,026 11,542 9,116 11,368
41 2,973 60 877 2,036
Total 905,739 146,089 147,389 612,261
Data: 2008 IMPLAN model
Although IMPLAN provides foreign imports and exports as well as domestic imports and
exports, only aggregate flows are provided. Data for commodity flows between regions
41
are not provided by IMPLAN. Therefore freight flow proportions between MSA regions
are estimated from FAF data and applied to the IMPLAN data to estimate freight flows
between MSA regions.
FAF data provide commodity flows between MSA and remainder of MSA regions by
SCTG commodity sectors. Table 3 and Table 4 show domestic and foreign imports and
domestic/foreign exports obtained from FAF data for the Los Angeles MSA region.
Table 3: Los Angeles MSA import components from FAF data
Los Angeles MSA
Domestic import
Los Angeles MSA
Foreign import
Origin Destination Foreign Origin Domestic Origin
Domestic
Destination
Los Angeles
Los Angeles
MSA
Foreign
country
Los Angeles
MSA
Los Angeles
Sacramento Sacramento
San Diego San Diego
San Francisco San Francisco
Remainder Remainder
Other States Other States
42
Table 4: Los Angeles MSA export components from FAF data
Los Angeles MSA Domestic
export
Los Angeles MSA Foreign export
Origin Destination Domestic Origin
Domestic
Destination
Foreign
Destination
Los Angeles
MSA
Los Angeles
Los Angeles
MSA
Los Angeles
Foreign country
Sacramento Sacramento
San Diego San Diego
San Francisco San Francisco
Remainder Remainder
Other States Other States
Figure 5 shows the process of estimating domestic trades for the Los Angeles MSA
region by applying FAF trade proportions to IMPLAN data. Four steps were involved, as
follows:
Step 1:
1) IMPLAN data at ZIP code areas are aggregated to the Los Angeles five-county region.
Similar diagrams can be constructed for all other regions in California.
2) Trade flows are provided in dollar values for 440 IMPLAN sectors. IMPLAN Sectors
are converted to 43 SCTG Sectors.
3) IMPLAN domestic trades include consumptions at the Los Angeles MSA and
shipments to other regions.
4) IMPLAN domestic trades provide flows coming out of each ZIP code area but do not
provide the final destinations.
5) IMPLAN data are not available by shipping mode.
43
Step 2:
1) Proportions of shipments using truck mode for domestic trades are estimated for the 43
SCTG Sectors.
2) Dollar and ton values are provided for all origin-destination pairs.
3) FAF data provide flows among MSA regions.
4) Similar diagrams can be constructed for all the MSA regions.
5) Even though FAF data provides flows by modes, IMPLAN data are used for
estimation because IMPLAN data provides zip code level information.
Step 3:
1) Proportions of shipments for truck mode and commodity sectors from FAF data are
multiplied to IMPLAN domestic trades.
2) Flows among ZIP code areas are not yet estimated.
Step 4:
1) Flows among ZIP code areas are estimated by applying a gravity model based on
IMPLAN data.
44
1. Domestic trades from IMPLAN data (2008) 2. Domestic trades and corresponding domestic
origins/destinations from FAF data (2007)
3. Domestic trades from/to Los Angeles MSA region
and corresponding domestic destinations/origins
by multiplying FAF proportion to IMPLAN data.
4. Domestic trades in Los Angeles MSA region
and corresponding domestic
origins/destinations of ZIP code areas.
Figure 5: Process of estimating domestic trades for the Los Angeles MSA region
Proportions of commodity flows between MSA and the remainder region are estimated
based on MSA region data. Appendix Tables 5 and 6 show the estimated proportions for
the Los Angeles MSA region. Then the estimated proportions are multiplied by domestic
imports and exports of each region. Domestic imports from Table 5 are the estimated
commodity flows between MSA regions. FAF data also provide mode information for
45
domestic trades. The results for trade flows of the Los Angeles MSA region are shown in
Table 5, Table 6, and Table 7.
46
Table 5: Los Angeles MSA demand
47
Table 6: Los Angeles MSA demand for truck mode
48
Table 7: Los Angeles MSA demand for truck mode
49
1. Foreign imports from IMPLAN data (2008) 2. Foreign imports and corresponding domestic
shipments from FAF data (2007)
3. Foreign imports to Los Angeles FAF region and
corresponding domestic shipments by
multiplying FAF proportion to IMPLAN data.
4. Foreign imports to gateways in Los Angeles
FAF region and corresponding domestic
shipments to ZIP code areas.
Figure 6: Process of foreign import and corresponding domestic shipment estimation for
the Los Angeles MSA region
Figures 6 and 7 show the process of estimating foreign imports and exports and
corresponding domestic shipment estimation for the Los Angeles MSA region. Four
steps are involved for the estimation, as follows:
Step 1:
1) IMPLAN data for ZIP code areas are aggregated to the Los Angeles five-county region.
Similar diagrams can be constructed for all the other regions of California.
50
2) Imports are provided by dollar values by 440 IMPLAN Sectors. IMPLAN Sectors are
converted to 43 SCTG Sectors.
3) IMPLAN foreign imports include consumption at the Los Angeles FAF and shipments
to other regions.
4) IMPLAN foreign imports data provide flows coming into each ZIP code area but do
not provide the final destinations.
5) IMPLAN data are not available by modes.
Step 2:
1) The flows are provided by different modes (air->truck, water->truck, rail->truck,
truck->truck) and 43 SCTG Sectors.
2) Dollar and ton values are provided for all origin-destination pairs.
3) FAF data provide flows among FAF regions.
4) Similar diagrams can be constructed for all the FAF regions.
5) Even though FAF data provides flows by modes, IMPLAN data are used for
estimation.
6) FAF mode proportions are calculated and applied to IMPLAN data.
Step 3:
1) Proportions of shipments by modes and commodity sectors from FAF data are
multiplied by IMPLAN foreign imports
Step 4:
1) Estimated foreign imports by modes are assigned to the corresponding locations which
are designated as gateways (e.g. air mode to airports, water mode to seaports).
51
2) Distribution of flows within a mode are based on available statistics (airports:
California air cargo statistics (2008), seaports: 2008 California seaport cargo stastics)
3) Flows from gateways to ZIP code areas and FAF regions are estimated applying a
gravity model based on IMPLAN data.
52
1. Foreign exports from IMPLAN data (2008) 2. Foreign exports and corresponding domestic
origins from FAF data (2007)
3. Foreign exports from Los Angeles FAF region
and corresponding domestic origins by
multiplying FAF proportion to IMPLAN data.
4. Foreign exports through gateways in Los
Angeles FAF region and corresponding
domestic origins of ZIP code areas.
Figure 7: Process of foreign export and corresponding domestic origin estimation for the
Los Angeles MSA region
53
3.1.1 Foreign import and exports mode split
Foreign imports and exports can involve multiple transport modes such as water-truck,
air-truck, and truck-truck. Major regional seaports and airports that handle cargo are
selected and the locations of the selected ports are identified.
Modes of shipments for foreign imports/exports
Air Truck mode
Freight that is imported to California MSAs from foreign countries by air and shipped by
trucks to domestic destinations is included in Air (foreign mode) Truck (domestic
mode) mode in the FAF data. Similarly freight that is shipped to California MSAs from
domestic origins by trucks and exported to foreign countries by air is included in Truck
(domestic mode) Air (foreign mode) mode in FAF data. Appendix Table 7 shows 2008
California air cargo statistics and the airports selected for this analysis.
Water Truck mode
Freight that is imported to California MSA from foreign countries by water and shipped
by trucks to domestic destinations is included in Water (foreign mode) Truck
(domestic mode) mode in the FAF data. Similarly freight that is shipped to California
MSAs from domestic origins by trucks and exported to foreign countries by water is
included in Truck (domestic mode) Water (foreign mode) mode in the FAF data.
Appendix Table 8 shows 2008 California seaport cargo statistics and the selected seaports
used in this study.
54
Truck Truck mode
Freight that is imported to California MSAs from foreign countries by trucks and shipped
by trucks to domestic destinations is included in Truck (foreign mode) Truck
(domestic mode) mode in the FAF data. Similarly freight that is shipped to California
MSAs from domestic origins by trucks and exported to foreign countries by trucks is
included in Truck (domestic mode) Truck (foreign mode) mode in the FAF data.
Unlike other modes, identifying origin countries of foreign trade would be necessary for
the truck mode. These origin locations are either North (Canada) or South (Mexico,
Central and South America). Foreign trade proportions between the two foreign locations
and each California MSA region are calculated by applying the FAF data. Then the
calculated proportions are multiplied by MSA level IMPLAN data to estimate foreign
trade coming into each California MSA via the truck mode. Flows from the foreign
countries to ZIP code areas in each California MSA are estimated by applying a gravity
model. Locations of foreign countries are identified at the border regions.
Water Multi-modes
Freight that is imported into California MSAs through seaports and shipped by rail to
domestic destinations are included in flows by water (foreign mode) multi-modes
(domestic mode) in the FAF data
2
. Similarly freight that is exported through seaports in
California MSAs and arrives by rail from domestic origins are included in flows of multi-
2
When domestic mode is multi-modes, over 99% of them are imported/ exported through seaports in 2007
FAF data.
55
modes (domestic mode) water (foreign mode) in the FAF data. Most seaports have rail
facilities in port terminals so that freights can be shipped to domestic destinations directly
by train. Then the freight that arrives at the rail yards in the destinations is shipped to the
ultimate consumers by truck. That is why rail mode traffic is usually expressed via multi-
modes.
When imported freight is shipped by train from seaports, the distances from the ports to
destinations are usually greater than 500 miles (Port of Los Angeles, 2004: page 9, figure
2-1). So it is unlikely that freight is shipped by train when the destinations are inside
California. Similarly when freight is shipped by train to be exported through seaports in
California, the origin rail yards are likely located outside California. There may still be
freight shipped by train and truck inside California. However, this amount is not
significant according to the recent report from the Port of Los Angeles mentioned above.
Therefore ‘multi-modes’ flows are excluded from estimates of truck flows related to
foreign trade in California.
3.1.2 Gravity model
After estimating freight flows between MSA regions, a doubly-constrained gravity model
is applied to estimate freight flows between ZIP code areas in each MSA region and
between MSA regions. A gravity model consists of trip productions/attractions, and a
travel distance friction factor (Mao and Demetsky, 2002). Trip productions/attractions are
obtained from the IMPLAN input-output data. Travel distance friction factors are
calculated based on shortest path distances between centers of ZIP code areas. The FAF3
56
network is used to estimate these shortest paths. (Lindall et al, 2005).
There are two conditions to be satisfied for a doubly-constrained gravity model, as
follows:
Condition1: Sum of all trade flows from a region = that region’s total supply.
Condition2: Sum of all trade flows into a region = that region’s total demand.
Estimated values meeting these two conditions are typically achieved via iteration.
Equation (3) shows how trade flows between regions are estimated.
=
∑
∑
j
ij
ij
j
j
j
d
d
D
D
W
ij
(1)
∑
=
j
ij
ij
W
W
P
ij
(2)
ij i
ij P O T =
(3)
Where
j, to i region from flow for trade lues weight va is
ij
W
j, to i region from factor gravity is ,
ij
P
j, to i region from flows trade is ij T
i, region in g originatin commodity the of supply total is
i
O
and commodity, for the demand total js region is
j
D
j. and i region between distance is
ij
d
57
Condition1 (Sum of all trade flows from region i = Total supply of region i) is
automatically met because
i i
O P O T P P
j
ij
j
ij
j
ij i
= = = =
∑ ∑ ∑
, 1 (4)
For each region i.
To satisfy Condition2 (Sum of all trade flows to region i = Total demand of region i),
j
D region j’s total demand is divided by the estimated total inflows, resulting in the
following ratio:
j
j
j
T
D
B =
. (5)
Then each initial supply-constrained estimate of ij T is multiplied by
j
B to obtain the
demand-constrained estimate which is
ij i j
ij
j
D
ij
P O B T B T = =
. (6)
To satisfy Condition1 (Sum of all trade flows from region i = Total supply of region i)
again,
i
O , region i’s total supply for the commodity is divided by the estimated total outflows,
resulting in following ratio:
i
i
i
T
O
A =
. (7)
Each demand-constrained estimate
D
ij
T to origin i is then multiplied by
i
A to obtain the
second supply-constrained estimates which is
58
ij ij j i
D
ij i
S
ij
P T B A T A T = =
. (8)
Ieration is continued until the ratios
i
A and
j
B are approximately one. The results are
balanced trade flows.
FAF data provides dollar and ton values of trade flows between all MSA regions. Dollar
values are converted to ton values by applying the dollar-ton relationships from the FAF
data. Then trade flows by ton values between ZIP code areas are estimated by applying
the gravity model.
VIUS (Vehicle Inventory Use Survey) 2002 data are used to estimate the types of trucks
for shipments. Appendix Table 3 shows the percentage by truck types. By multiplying the
percentages with the trade flows between ZIP code areas, trade flows between ZIP code
areas by truck types are estimated.
Then trade flows data are converted to the number of trucks by applying average payload
factors. FHWA provides average payload by vehicle group of Vehicle Inventory Use
Survey. Appendix Table 3 show the average payload for California
Truck flows between ZIP code areas by truck types are estimated by dividing the gravity
model results with the average payload factors. The estimated truck flows are the initial
truck O-D matrix which is used as an input for estimating an adjusted O-D matrix.
59
3.1.3 O-D matrix adjustment
O-D matrix estimated by commodity flows may be different from real traffic flows.
Therefore an adjustment of the initial O-D matrix is needed to reflect real traffic flows.
Real traffic counts such as Highway Performance Monitoring System (HPMS) at state
and national levels or local survey data are often used to adjust an initial O-D matrix.
FAF data provides Average Annual Daily Truck Traffic (AADTT) which is driven from
HPMS. HPMS include traffic count data submitted by each State. So I used AADTT data
obtained from FAF.
AADTT is only available for 2007 or 2040. The scenario of the model is for year 2030.
Therefore AADTT for year 2030 (AADTT30) is calculated by interpolating the two
points. Similarly link capacity for year 2030 (CAP30) is calculated by interpolating 2007
and 2040 data
3
. AADTT30 includes truck flows for the nation, but the model only
includes truck flows originated from California or destined to California. California
AADTT30 is calculated using ton value of the projected freight flows at year 2030 from
FAF Origin-Destination data. Equation (9) shows the calculation process.
𝐶𝐶 _ 𝑍𝑓 𝐷 =
∑ 𝑇 𝑇𝑇
𝑖𝑖
1 2 3
𝑖 = 𝐶𝐶 , 𝑖 = 1
+ ∑ 𝑇 𝑇𝑇
𝑖𝑖
1 2 3
𝑖 = 1, 𝑖 = 𝐶𝐶
∑ ∑ 𝑇 𝑇𝑇
𝑖𝑖
1 2 3
𝑖 = 1
1 2 3
𝑖 = 1
(9)
Where CA_Pro is the proportions of freight flows originated from or destined to
California out of total freight flows. i=origin, j=destination.
3
AADTT30=AADTT07+((AADTT40-AADTT07)/33)*23. Similarly CAP30=CAP07+((CAP40-CAP07)/33)*23.
60
The calculated proportions are multiplied to the AADTT30 to estimate a modified
AADTT30. The modified AADTT30 is further multiplied by VMT proportions to
calculate AADTT30 by truck types
4
(tru_AADTT30). So, tru_AADTT30 shows average
annual daily truck traffic by truck types originating from or destined for California at
2030.
The TransCAD O-D matrix estimation procedure is applied to the tru_AADTT30 and
CAP 30 data to adjust the initial O-D matrix. Practice of Southern California Association
of Government (SCAG) has applied to convert the truck flows to passenger car
equivalent (PCE) which are 1.2 for Light truck, 1.5 for Medium truck, and 2 for Heavy
truck as PCE factors.
The estimated truck flows are the Origin-Destination matrix which was used as an input
for the transportation impact model to estimate Vehicle Miles Traveled (VMT) on each
link of the network.
4
EMFAC model requires VMT by truck types to estimate emissions. Therefore VMT has to be estimated by truck
types. To be consistent with EMFAC model, VMT by truck is obtained from EMFAC model and the proportions are
calculated. The calculated proportions are as follows: LDT1=0.15, LDT2=0.48, MDT=0.21, LHDT1=0.05,
LHDT2=0.01, MHDT=0.04, HHDT=0.06.
61
3.2 Transportation impact model
The user equilibrium (UE) model is applied to estimate a VMT baseline and to estimate
effects of various scenarios. Figure 8 shows the procedures used to estimate VMT based
on the estimated truck OD matrix.
Figure 8: Procedures to estimate VMT based on the estimated truck OD matrix
User Equilibrium (UE) assignment model
A UE assignment model is applied for assigning truck flows on road networks. Sheffi
(1985) introduced user equilibrium as follows:
Min
∑
∫
=
a
x
a
a
d t x z
0
) ( ) ( ω ω (10)
subject to
∑ ∑ ∑
=
o d k
od
k
od
k a a
f x
,
d a ∀ (11)
Truck O-D matrix between ZIP code areas in California
and between California and other States
VMT by truck type at county level
US road networks
Road capacity from
FAF data
Multi-Modal Multi-
Class Assignment
Modified AADTT
data
62
od
k
od
k
q f
∑
= d o, ∀ (12)
0 ≥
od
k
f d o k , , ∀ (13)
where
a
x is the total flow on link a,
) ( ω
a
t is the link cost-performance function,
od
k a,
d is the incidence relationship variable; equal to one if link a belongs to path
k connecting OD pair o and d,
od
k
f is flow on path k connecting origin o with destination d,
od
q is total trip between origin node o and destination node d,
The link performance function is shown as follows:
] ) ( 1 )[ 0 (
β
α
a
a
a a
C
x
t t + = (14)
where ) (x t
a
is the performance function to calculate average travel cost on link a,
and
) 0 (
a
t is the free-flow travel cost on link a,
a
x is the total flow on link a,
a
C is the capacity of link a,
Historically
α
and
β
have been set as 0.15 and 4, respectively. However, different
values may be applied according to simulation scenarios (Caliper, 2004).
The equilibrium model can be implemented in the following steps,
63
Step 0: Initialization. Perform all-or-nothing assignment based on ) 0 (
a a
t t = which
means there is no congestion. This step yields Link flows
1
a
x .
Step 1: Update. ) (
n
a a
n
a
x t t = , a ∀ .
Step 2: Find direction. Perform all-or-nothing assignment based on
n
a
t , which yields
a set of auxiliary flows {
a
y }.
Step 3: Line search. Find
n
α that solves
ω ω
α
α
d t
a
x y x
a
n
a
n
a
n
a
) (
) (
0
1 0
min
∑
∫
− +
≤ ≤
Step 4: Move. Set ) (
1 n
a
n
a n
n
a
n
a
x y x x − + =
+
α , a ∀
Step 5: Convergence test. If a convergence criterion is met, stop (current solution,
{ }
1 + n
a
x , is the set of equilibrium link flows); otherwise, set n:=n+1 and go to step 1.
The estimated VMTs are then used as inputs for the emissions model.
64
3.3 Air pollution emissions model
Air pollution emissions are estimated by applying an EMFAC model. Figure 9 shows the
procedures to estimate air pollution emissions based on the estimated VMT by truck type.
Figure 9: Procedures to estimate air pollution emissions based on the estimated VMT by
truck type
To estimate air pollution emissions, base emission rates are first adjusted by area specific
data such as Inspection and Maintenance (I/M) program, temperature, and relative
humidity. Then total emission inventories are estimated by multiplying the adjusted
emission rates with total vehicle activity. These adjustments and estimations are
accomplished by applying EMFAC model.
VMT by truck type at county level
Air pollution emissions by county
EMFAC2007 model
65
Chapter 4: Scenarios
The model developed for this research includes an origin-destination (OD) matrix for
domestic and foreign trade by commodity sector. To account for the effects of
interregional and international trade, the locations of a region’s international gateways for
trucking, such as airports, seaports, and border regions are identified. The model includes
road and highway networks that trucks utilize when traveling between OD pairs. The
model is appropriate for identifying and analyzing changes in commodity flow patterns or
changes of road network utilization and the corresponding consequences resulting in
various air pollution emissions. The key idea is to implement this for various emissions
control policy scenarios. In the discussion below, scenario results are compared to
projected baseline trends.
The model’s OD matrix, however, is not yet differentiated by time of day such as AM
peak, PM peak, and off peak. And the model does not include passenger flows. Therefore
congestion effects cannot be fully analyzed although the user equilibrium algorithm
includes a congestion function.
Baseline: Future growth of foreign trade in SPB
This is the reference case that was used to compare and evaluate the various scenario
results. The baseline shows network and emission responses for projected growth paths.
The results show the impacts on link volumes and air pollution emissions when trade via
local area seaports grows in the near future. Table 8 shows projected growth at San Pedro
Bay which includes the Port of Los Angeles and the Port of Long Beach. To compare
66
the results with other scenarios, a simple projected growth path to 2030 is developed.
Growth rates from 2008 to 2030 are multiplied by 2008 data for foreign trade via the
seaports of Los Angeles County. These results show how the expected growth of trade via
the ports affects commodity flows and air pollution emissions.
Table 8: Port of Los Angeles and Port of Long Beach throughput demand forecast (baseline)
Unit: 1,000
Actuals Forecast Increase
Actual/Forecast TEU 2008 2030 TEU %
Import Loads 7,328 19,801 12,473 170%
Export Loads 3,470 5,938 2,468 71%
Source: San Pedro Bay Container Forecast Update (July 2009), available at
http://www.portoflosangeles.org/pdf/SPB_Container_Forecast_Update_073109.pdf
Scenario One: Truck replacement scenario- Replacing older trucks with newer trucks
The Clean Truck Program (CTP) at the port of Los Angeles and the port of Long Beach
has been successful reducing truck related emissions around the ports
5
. CTP was applied
to drayage operations (short haul cargo container trips). Scenario One assume that a
similar program will be applied to all diesel truck in Los Angeles County so that the ages
of all diesel trucks would be less than 20 years in 2030 in the County. Truck populations
greater than 20 year of age are shifted to earlier ages based on current age distributions.
5
According to the port of Los Angeles (http://www.portofla.org/ctp/idx_ctp.asp), CTP reduced port truck
emissions by more than 80 percent in 2012.
67
Scenario Two: Network & truck improvement scenario- Developing zero emission truck
lanes on I-710
Route I-710 is a major freight corridor from the port of Los Angeles and the port of Long
Beach to various domestic destinations. Because communities around the freeway have
been impacted by air pollution emissions, there have been various studies and plans to
reduce emissions while expanding the capacity for truck flows of the freeway.
Developing zero emission truck lanes is one of the plans that is considered relatively
cost-effective and technically available. Based on the proposed plans
6
as shown in
Figure 10, assume that four lanes of eight lanes on I-710 from the ports to SR60 are
converted to zero-emission truck lanes by 2030. Assume that hybrid trucks that can be
operated by electricity and by diesel engine simultaneously are operated on the converted
lanes. So, 50 percent of the total traffic flows on I-710 from ports to SR60 are converted
to zero emission truck flows.
Scenario Three: Land use scenario- Inland port (intermodal facility) at Mira Loma
industrial area
Developing an inland port, connected by rail to the existing seaports, has been considered
as a long term project to reduce truck traffic and air pollution emissions around the ports
and highways. The Mira Loma industrial area is one of the candidates for such a
development (Rahimi et al., 2008). Assume that the inland port begins operations in 2030.
A possible development site from SCAG is in Figure 11. 50 percent of truck flows in the
port of Los Angeles and the port of Long Beach will be moved from the ports to the
6
http://www.metro.net/projects/i-710-corridor-project/i710-swg-meetings
68
inland port for this scenario.
Figure 10: I-710 Corridor Project EIR/EIS (Scenario 2; Source: Metro.net available from
http://www.metro.net/projects_studies/30-
10_highway/images/AHP_I_710_Corridor_EIS_EIR.pdf)
69
Figure 11: Possible development site of inland port at Mira Loma (Scenario 3; Source:
SCAG available from
http://www.scag.ca.gov/goodsmove/pdf/2007/gmtf053007fullagn.pdf) (Zipcode:91752)
70
Chapter 5: Model Results
This chapter explains model results for the Los Angeles MSA and the Los Angeles
County. Figure 12 shows a simulated total truck flow for the baseline estimates of the
model. Because the model only includes truck flows originated from California or
destined to California, a relatively high percentage of truck trips occur within California.
Figure 13 is similar to Figure 12, but shows flows from California to other states. Note
that the scale of the total flows in the legend is changed from 50,000, 25,000, 12,500 to
2,000, 1,000, 500 respectively. To estimate the OD matrix applied in the model, 2008
IMPLAN data for ZIP codes are used to estimate initial truck flows between ZIP code
areas as explained in Chapter 3. Then truck flows from and to the ports of Los
Angeles/Long Beach are modified to reflect 2030 port growth.
71
Legend
---- Road networks
---- FAF region boundary
Scenario 3 locations
Figure 12: Total truck flows originated from California or destined to California at 2030 (Baseline)
72
Figure 13: Total truck flows originated from California or destined to California at 2030 (to show flows outside CA)
(Baseline)
Legend
---- Road networks
---- FAF region boundary
Scenario 3 locations
73
5.1 Model results for the Los Angeles MSA
This section explains results for the Los Angeles MSA region. To obtain VMT for the
MSA region, VMT by vehicle classes for each scenario are aggregated into each county
within the MSA. Then, the aggregated VMTs are used as inputs for EMFAC model.
Table 9 summarizes model results giving VMT for the Los Angeles MSA region,
including Los Angeles County, Orange County, Riverside County, San Bernardino
County, and Ventura County. The Table shows separate results for combined counties
based on results for Scenario Three. Los Angeles, Orange, and Ventura counties are
combined because the three counties have a decrease in VMT in Scenario Three.
Riverside and San Bernardino counties are combined because two counties show an
increase in VMT in the scenario. Note that there is no change in VMT for Scenario One
because the VMT in Scenario One is assumed to be that of the baseline. In Scenario Two,
VMT for vehicle classes of MHDT and HHDT are reduced by 10,910 miles per day and
16,407 miles per day, respectively, due to the assumption of zero emission vehicle lanes
on I-710. Total VMT reductions are 27,317 miles per day, which is a 0.07 percent
reduction.
74
Table 9: Summary of vehicle miles traveled (VMT) results, Los Angeles MSA
Units: Miles per day
Region Vehicle class Baseline
Scenario VMT Change
1 2 3
Los Angeles MSA
(Los Angeles +
Orange + Ventura +
Riverside + San
Bernardino County)
LDT 23,971,075 0 0 65,143
MDT 7,990,359 0 0 20,925
LHDT 2,284,008 0 0 2,301
MHDT 1,527,658 0 -10,910 2,173
HHDT 2,308,083 0 -16,407 6,368
Total
Number 0 0 -27,317 96,910
% 0.00% 0.00% -0.07% 0.25%
Los Angeles + Orange
+ Ventura County
LDT
13,501,956 0 0 -258,623
MDT
4,500,652 0 0 -86,700
LHDT
1,285,775 0 0 -27,309
MHDT
859,145 0 -10,910 -17,353
HHDT
1,287,066 0 -16,407 -24,388
Total
Number 0 0
-27,317 -414,373
% 0.00% 0.00%
-0.13% -1.93%
Riverside + San
Bernardino County
LDT
10,469,120 0 0 323,766
MDT
3,489,707 0 0 107,625
LHDT
998,232 0 0 29,610
MHDT
668,513 0 0 19,526
HHDT
1,021,016 0 0 30,756
Total
Number 0 0
0 511,284
% 0.00% 0.00%
0.00% 3.07%
Note:
LDT: Light-Duty Trucks, MDT: Medium-Duty Trucks, LHDT: Light HD Trucks, MHDT: Medium HD
Trucks, HHDT: Heavy HD Trucks
Interestingly, in Scenario Three, model results show that VMT for vehicle classes are
increased when 50 percent of the truck flows are moved from the ports of Los
Angeles/Long Beach to Mira Loma area. The total VMT increase is 96,910 miles per day
which is a 0.25 percent increase. This result may be because there are no changes
75
assumed for network attributes around the Mira Loma area. If an inland port is developed
in the Mira Loma area, there likely would be new developments of highways and major
arterials to improve network accessibility of the area and the network model results may
be different than the results reported here. Even though road networks are not fully
updated to analyze the scenario, there is an important implication for policy applications
from the model results:
Shifting transport activities from one location to another may help to reduce
environmental problems for the specific area, but the benefits may be offset by increased
problems in other locations. Therefore analyzing the impacts of policy scenarios in
various regions is useful for local area policy makers. A comparison of Figure 17 and 18
make this implication obvious. VMT around the ports area has decreased. More
explanations are developed by comparing the Los Angeles MSA results with the Los
Angeles County results later in this chapter.
76
Figure 14: Simulated versus Observed (Modified AADTT30) V olumes in the Los Angeles
MSA
Figure 14 displays a scatterplot of simulated and modified AADTT30 for the Los Angeles
MSA. When the simulated and observed volumes agree 100 percent, the observations fall
on the 45-degree line. The correlation coefficient for model results shows about 84
percent agreement. Table 10 shows the comparison of total volumes in the Los Angeles
MSA. The difference in total volume of trucks between simulated values and AADTT30
values is about 900,000 trucks. In other words, total volumes of the modified AADTT30
and simulated agree over 98 percent in the aggregate.
Table 10: Comparison of baseline total volumes in the Los Angeles MSA
Total volume of truck
Difference (Simulated-AADTT30)
Modified AADTT30 Simulated (Base scenario)
Number %
V olumes
48,471,251 47,548,530 -922,721 -1.90%
R² = 0.8363
0
10000
20000
30000
40000
50000
60000
0 10000 20000 30000 40000 50000
Simulated Volumes
Modified AADTT30
77
Table 11: Air pollution emissions results for baseline and scenarios in the Los Angeles
MSA
Units: tons per day
Baseline
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG 5.59 23.14 12.09 5.71 0.96 1.35 2.98 51.82
CO 15.33 69.74 43.16 24.35 4.52 12.24 18.92 188.28
NOx 0.95 5.2 3.04 11.96 2.58 4.55 34.77 63.07
CO2 (1000) 3.76 12.08 7.32 1.81 0.36 2.41 6.12 33.83
PM 0.29 1.51 0.69 0.09 0.01 0.22 0.53 3.34
SOx 0.04 0.11 0.07 0.01 0 0.02 0.06 0.33
Difference from baseline
Scenario 1
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG - - - - - 0 -0.02 -0.04
CO - - - - - -0.09 -0.11 -0.21
NOx - - - - - -0.31 -0.23 -0.53
CO2 (1000) - - - - - 0 0 0
PM - - - - - -0.01 -0.03 -0.02
SOx - - - - - 0 0 0
Scenario 2
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG - - - - - 0 0 -0.01
CO - - - - - -0.02 -0.03 -0.05
NOx - - - - - -0.02 -0.05 -0.07
CO2 (1000) - - - - - -0.02 -0.03 -0.05
PM - - - - - 0 0 -0.01
SOx - - - - - 0 0 0
Scenario 3
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG 0 -0.01 0 0 0 0 0.01 -0.01
CO 0 0.05 0.01 0.01 0 0.01 0.02 0.07
NOx 0 0.01 0.01 0.01 0 0 0.01 0.01
CO2 (1000) 0.01 0.01 0.02 0.01 -0.01 0 0.01 0.06
PM -0.01 0.01 0 0 0 -0.01 -0.01 0.01
SOx 0 0 0 0 0 0 0 0
Table 11 displays the Los Angeles MSA results for air pollution emissions based on
applying the network model results for the baseline and the three scenarios. Note that
there are no changes for vehicle classes of LDT1, LDT2, MDV, LHDT1, and LHDT2 in
78
Scenarios One and Two because the two scenarios are only involved in MHDT and
HHDT. Scenario One shows the biggest reduction in all pollutants among all the
scenarios. In particular, NOx and PM are reduced by 0.54 and 0.04 tons per day
respectively. CO2 does not change because VMT remains at the same level with respect
to the baseline. Scenario Two displays relatively small changes compared to the other
scenarios. Change in PM is very small compared to the baseline. Scenario Three shows
increases in several of the air pollution emissions. PM for all vehicle classes except
LDT2 is reduced in Los Angeles MSA, although total VMT for the region is increased as
shown in Table 9. This is because PM reductions in Los Angeles, Orange, and Ventura
counties are bigger than PM increase in Riverside and San Bernardino counties.
In Scenario One, when old trucks in Los Angeles County are replaced with newer models,
it will affect air pollution emissions in Los Angeles County and other Counties as well.
The estimated origin-destination (OD) matrix is used to estimate the effects in each
county. Truck proportions originating from Los Angeles County are estimated by using
the estimated OD matrix. Table 12 shows the calculated proportions for the Los Angeles
MSA, including Los Angeles County, Orange County, Riverside County, San Bernardino
County, and Ventura County. Results for Los Angeles County, for example, show that 73
percent of the trucks operating in the County, including both medium heavy-duty trucks
(MHDT) and heavy heavy-duty trucks (HHDT), originate within the County. In Orange
County 30 percent of the trucks originated from Los Angeles County. Percentages for
other Counties can also be interpreted in the same way.
79
Table 12: Proportions of trucks originated from the Los Angeles County
County
MHDT HHDT
Los Angeles 0.73 0.73
Orange 0.30 0.30
Riverside 0.23 0.22
San Bernardino 0.22 0.21
Ventura 0.35 0.35
Source: estimated origin-destination matrix
80
Table 13 displays changes of the Los Angeles MSA air pollution emissions percentages
for three scenarios. Scenario One shows relatively larger impacts than other scenarios.
Total change in percentage is shown graphically in Figure 15.
Table 13: Percent change of air pollution results by applying scenarios in the Los Angeles
MSA
Scenario 1
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG - - - - - 0.00% -0.67% -0.08%
CO - - - - - -0.74% -0.58% -0.11%
NOx - - - - - -6.81% -0.66% -0.84%
CO2 - - - - - 0.00% 0.00% 0.00%
PM - - - - - -4.55% -5.66% -0.60%
SOx - - - - - 0.00% 0.00% 0.00%
Scenario 2
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG - - - - - 0.00% 0.00% -0.02%
CO - - - - - -0.16% -0.16% -0.03%
NOx - - - - - -0.44% -0.14% -0.11%
CO2 - - - - - -0.83% -0.49% -0.15%
PM - - - - - 0.00% 0.00% -0.30%
SOx - - - - - 0.00% 0.00% 0.00%
Scenario 3
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG 0.00% -0.04% 0.00% 0.00% 0.00% 0.00% 0.34% -0.02%
CO 0.00% 0.07% 0.02% 0.04% 0.00% 0.08% 0.11% 0.04%
NOx 0.00% 0.19% 0.33% 0.08% 0.00% 0.00% 0.03% 0.02%
CO2 0.27% 0.08% 0.27% 0.55% -2.78% 0.00% 0.16% 0.18%
PM -3.45% 0.66% 0.00% 0.00% 0.00% -4.55% -1.89% 0.30%
SOx 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
81
Figure 15: Percentage of air pollution emissions reduction for scenarios in the Los
Angeles MSA
-1.00% -0.50% 0.00% 0.50%
TOG
CO
NOx
CO2
PM
scenario 3
scenario 2
scenario 1
82
5.2 Model results for Los Angeles County
Figure 16: Simulated versus Observed (Modified AADTT30) V olumes in the Los Angeles
County
Figure 16 displays a scatterplot of simulated and modified AADTT30 for Los Angeles
County. The result is similar to the one for the Los Angeles MSA. The correlation
coefficient shows over 85 percent of agreement between simulated volumes and modified
AADTT30. Table 14 shows the comparison of total volumes in Los Angeles County.
Similar to the Los Angeles MSA result, total volumes of the modified AADTT30 and
simulated agree by more than 98 percent.
Table 14: Comparison of baseline total volumes in the Los Angeles County
Total volume of truck
Difference (Simulated-AADTT30)
Modified AADTT30 Simulated (Base scenario)
Number %
V olumes 27,753,969 27,803,178
49,208 0.18%
R² = 0.8518
0
10000
20000
30000
40000
50000
60000
0 10000 20000 30000 40000 50000 60000
Simulated Volumes
Modified AADTT30
83
Mira Loma
Port of LA Port of LB
Figure 17: Total truck flows around port of LA/LB with port growth at 2030 (Baseline)
84
Mira Loma
Port of LA Port of LB
Figure 18: Total truck flows around port of LA/LB and Mira Loma (Scenario 3)
85
Figure 17 shows total truck flows around the ports of Los Angeles/Long Beach and Mira
Loma for the Baseline. Figure 18 shows total truck flows around the Ports of Los
Angeles/Long Beach and Mira Loma for Scenario Three. 50% of truck flows are taken
from the two ports and are assigned to Mira Loma area. By comparing the two maps we
can see that truck flows around the two ports are decreased, and flows around the Mira
Loma area are increased as a consequence of the scenario.
Table 15: Vehicle miles traveled (VMT) in the Los Angeles County
Units: Miles per day
Baseline and Scenarios Baseline
VMT change from scenario
1 2 3
LDT 10,012,255 0 0 -215,567
MDT 3,337,419 0 0 -72,287
LHDT 953,527 0 0 -22,799
MHDT 637,983 0 -10,910 -14,401
HHDT 954,370 0 -16,407 -20,145
Total
Number 0 -27,317 -345,199
% 0.00% -0.17% -2.17%
Note:
LDT: Light-Duty Trucks, MDT: Medium-Duty Trucks, LHDT: Light HD Trucks, MHDT: Medium HD
Trucks, HHDT: Heavy HD Trucks
Table 15 shows VMT for the Baseline and VMT changes for the three scenarios. For
Scenario One, old trucks are replaced into newer ones but no change in VMT is assumed.
For Scenario Two, VMT of MHDT and HHDT are reduced because 50 percent of truck
flows for two truck classes are converted to zero emission vehicle trips on I-710. VMT
for other vehicle types remain at the same level.
Scenario Three shows a relatively larger decrease in VMT when 50 percent of truck flows
are moved from the Ports of Los Angeles/Long Beach to the Mira Loma area. This result
86
is different from the one for the Los Angeles MSA. Total Los Angeles MSA VMT is
increased in Scenario Three, as shown in Table 9. A part of the reason of this difference is
that the Mira Loma area is located in San Bernardino County. Because this table only
includes VMT within Los Angeles County, the results show decreased VMT.
87
Table 16: Air pollution emissions results for baseline and scenarios in the Los Angeles
County
Units: tons per day
Baseline
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG 2.56 11.46 5.67 2.69 0.45 0.7 1.04 24.58
CO 6.85 32.78 19.7 11.91 2.11 6.46 6.57 86.38
NOx 0.42 2.42 1.4 5.59 1.19 2.18 11.09 24.29
CO2 (1000) 1.62 5.32 3.18 0.79 0.16 1.02 2.37 14.45
PM 0.13 0.68 0.3 0.04 0.01 0.1 0.22 1.48
SOx 0.02 0.05 0.03 0.01 0 0.01 0.02 0.14
Difference from baseline
Scenario One
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG - - - - - 0 -0.02 -0.03
CO - - - - - -0.07 -0.09 -0.15
NOx - - - - - -0.21 -0.18 -0.38
CO2 (1000) - - - - - 0 0 0
PM - - - - - -0.01 -0.01 -0.02
SOx - - - - - 0 0 0
Scenario Two
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG - - - - - 0 0 -0.01
CO - - - - - -0.02 -0.03 -0.05
NOx - - - - - -0.02 -0.05 -0.07
CO2 (1000) - - - - - -0.02 -0.03 -0.05
PM - - - - - 0 0 -0.01
SOx - - - - - 0 0 0
Scenario Three
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG 0 -0.01 0 0 0 0 0 -0.02
CO -0.05 -0.22 -0.12 -0.01 0 -0.02 -0.04 -0.45
NOx -0.01 -0.02 -0.01 -0.01 0 -0.03 -0.06 -0.13
CO2 (1000) -0.03 -0.09 -0.05 -0.01 -0.01 -0.02 -0.04 -0.25
PM -0.01 -0.01 0 0 0 -0.01 -0.01 -0.03
SOx 0 0 0 0 0 0 0 0
Table 16 displays air pollution emissions results for the baseline and the three scenarios.
There are no changes for vehicle classes of LDT1, LDT2, MDV, LHDT1, and LHDT2 in
Scenario One and Two because these two Scenarios only involved MHDT and HHDT.
88
Scenario One shows the biggest reduction in NOx and TOG among all scenarios.
Scenario Three shows the biggest reduction in CO, CO2, and PM. Scenario Two shows
the least impact in terms of reducing emissions for the county. A part of the reason for
small impact of Scenario Two may be that emissions reductions in the specific area do
not have much impact for the county as a whole. Although the network model results are
produced for smaller local areas, the EMFAC model is not appropriate for emissions
estimations of smaller areas than counties.
89
Table 17: Percent change of air pollution results by applying scenarios in the Los Angeles
County
Scenario One
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG - - - - - 0.00% -1.92% -0.12%
CO - - - - - -1.08% -1.37% -0.17%
NOx - - - - - -9.63% -1.62% -1.56%
CO2 - - - - - 0.00% 0.00% 0.00%
PM - - - - - -10.00% -4.55% -1.35%
Sox - - - - - 0.00% 0.00% 0.00%
Scenario Two
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG - - - - - 0.00% 0.00% -0.04%
CO - - - - - -0.31% -0.46% -0.06%
NOx - - - - - -0.92% -0.45% -0.29%
CO2 - - - - - -1.96% -1.27% -0.35%
PM - - - - - 0.00% 0.00% -0.68%
Sox - - - - - 0.00% 0.00% 0.00%
Scenario Three
Vehicle class LDT1 LDT2 MDV LHDT1 LHDT2 MHDT HHDT Total
TOG 0.00% -0.09% 0.00% 0.00% 0.00% 0.00% 0.00% -0.08%
CO -0.73% -0.67% -0.61% -0.08% 0.00% -0.31% -0.61% -0.52%
NOx -2.38% -0.83% -0.71% -0.18% 0.00% -1.38% -0.54% -0.54%
CO2 -1.85% -1.69% -1.57% -1.27% -6.25% -1.96% -1.69% -1.73%
PM -7.69% -1.47% 0.00% 0.00% 0.00% -10.00% -4.55% -2.03%
Sox 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
Table 17 displays changes of air pollution emissions in percentage for the three scenarios.
Scenario One shows significant reduction of NOx and PM in medium-heavy duty trucks
(MHDT) and heavy-heavy duty trucks (HHDT). Total percentage change is also shown
graphically in Figure 19.
90
Figure 19: Percentage of air pollution emissions reduction for scenarios in Los Angeles
County
-2.50% -2.00% -1.50% -1.00% -0.50% 0.00%
TOG
CO
NOx
CO2
PM
scenario 3
scenario 2
scenario 1
91
5.3 Sensitivity analysis
This section explains the results from various sensitivity analyses. Three different levels
of implementation are applied to each scenario to see how sensitive the model results are.
Summary of the sensitivity test results
The sensitivity test results show that the model works almost linearly for Scenarios One
and Two. Emissions almost linearly decrease when more old trucks are replaced with new
trucks in Scenario One or when more lanes are converted to zero-emission truck lanes in
Scenario Two. Scenario Three shows varied results by pollutants and levels. These results
would change if a different inland port site other than the Mira Loma area is selected.
Overall, the model performs as expected.
The sensitivity test results show different implications for each scenario;
Scenario One: TOG, CO2, PM SOx are not changed by replacing old trucks because
truck populations, VMT, and fuel type are same regardless of the level of
implementations. CO and NOx, however, are changed although the amounts are small.
The reason for small changes may be because the EMFAC model has limited capability
to assess technology improvement. For example, natural gas trucks would not be included
in the EMFAC model unless natural gas trucks are first produced and tested to determine
emission parameters. If alternative fuel trucks such as natural gas trucks become
popular, the simulated and real impacts could be much larger.
92
Scenario Two: Emissions for all pollutants except SOx change because of VMT
decreases on I-710. The change is small because the VMT decrease on I-710 is less than
1 percent in the Los Angeles County total. Although truck traffic on I-710 is heavy, it is a
small portion of the amount for Los Angeles County.
Scenario Three: Emissions for all pollutants except SOx are changed because of VMT
decreases around the ports of Los Angeles/ Long Beach. But the change is small perhaps
because the VMT decrease around the ports is about 1 percent for all of Los Angeles
County.
The results in Tables 18 and 20 may suggest that the level of policy implementation does
not make much difference. However, Figures 20 and 21 show that it does have impacts.
Important implications of the results are that infrastructure projects at a specific location
would not make much impact for the whole County or MSA. Moreover, just replacing
old diesel truck to newer diesel trucks would not bring much reduction unless an
innovative technology is developed. Applying cleaner fuel such as natural gas would be
more promising.
93
5.3.1 Sensitivity analysis for the Los Angeles MSA
Table 18 shows air pollution emissions results for three scenarios for the Los Angeles
MSA. Each scenario includes three different levels which are -25 percent, 0 percent, and
25 percent. For Scenario One, -25 percent, 0 percent, and 25 percent mean 50 percent,
75percent, and 100 percent (original scenario) replacement of old trucks into new trucks
in Los Angeles County, respectively. For Scenario Two, -25 percent, 0 percent, and 25
percent mean 25 percent, 50 percent (original scenario), and 75 percent reduction of
medium-heavy duty truck (MHDT) and heavy-heavy duty truck (HHDT) on I-710
respectively. For Scenario Three, -25 percent, 0 percent, and 25 percent mean 25
percent, 50 percent (original scenario), and 75 percent reduction of truck flows at the port
of Los Angeles and Long Beach. It also means 25 percent, 50 percent, and 75 percent
increase of truck flows in the Mira Loma area.
In Table 18, TOG shows little change for various levels in each scenario. This is because
emissions of TOG mostly depend more on vehicle population than VMT. The numbers of
vehicles are assumed to be same for all scenarios. SOx shows no changes across
strategies. SOx emissions are calculated by multiplying a weight factor of sulfur in fuel
by gallons of fuels consumed. Even though gallons of fuels consumed are changed by
different levels of scenarios, the changes are not significant enough to make a difference
so that SOx levels remain at the same level. Other pollutants show more reductions
when more trucks are replaced in Scenario One or when more lanes are converted to
zero-emission truck lanes in Scenario Two. Scenario Three, however, shows mixed
94
results by pollutants and truck types. NOx, for example, remained at the same level then
decreased from 34.78 tons per day to 34.77 tons per day when more HHDT flows are
moved from the port of Los Angeles/Long Beach to the Mira Loma area. CO emissions,
on the contrary, increased first then decreased when more HHDT flows are relocated.
Table 18: Results of sensitivity analysis for the Los Angeles MSA
Units: tons per day
MHDT HHDT Total
-25% 0% 25% -25% 0% 25% -25% 0% 25%
TOG
Scenario1 1.35 1.35 1.35 2.97 2.96 2.96 51.79 51.78 51.78
Scenario2 1.35 1.35 1.35 2.98 2.98 2.97 51.82 51.81 51.81
Scenario3 1.35 1.35 1.35 2.99 2.99 2.98 51.82 51.81 51.81
CO
Scenario1 12.20 12.18 12.15 18.87 18.83 18.81 188.16 188.13 188.07
Scenario2 12.23 12.22 12.21 18.90 18.89 18.87 188.25 188.23 188.21
Scenario3 12.24 12.25 12.25 18.92 18.94 18.92 188.29 188.35 188.34
NOx
Scenario1 4.40 4.32 4.24 34.66 34.60 34.54 62.81 62.66 62.54
Scenario2 4.54 4.53 4.52 34.74 34.72 34.69 63.04 63.00 62.97
Scenario3 4.55 4.55 4.55 34.78 34.78 34.77 63.07 63.08 63.09
CO2
(thousand)
Scenario1 2.41 2.41 2.41 6.12 6.12 6.12 33.83 33.83 33.83
Scenario2 2.40 2.39 2.38 6.11 6.09 6.07 33.81 33.78 33.76
Scenario3 2.40 2.41 2.41 6.13 6.13 6.13 33.85 33.89 33.89
PM
Scenario1 0.21 0.21 0.21 0.51 0.50 0.50 3.32 3.32 3.32
Scenario2 0.22 0.22 0.21 0.53 0.53 0.52 3.33 3.33 3.33
Scenario3 0.22 0.21 0.21 0.53 0.52 0.52 3.34 3.35 3.35
Sox
Scenario1 0.02 0.02 0.02 0.06 0.06 0.06 0.33 0.33 0.33
Scenario2 0.02 0.02 0.02 0.06 0.06 0.06 0.33 0.33 0.33
Scenario3 0.02 0.02 0.02 0.06 0.06 0.06 0.33 0.33 0.33
Note: For scenario 1, -25%, 0, 25% mean 50 percent, 75 percent, 100 percent replacement of old trucks in
the Los Angeles county respectively.
For scenario 2, -25%, 0, 25% mean 25 percent, 50 percent, 75 percent reduction of MHDT and
HHDT on I-710 respectively.
For scenario 3, -25%, 0, 25% mean 25 percent, 50 percent, 75 percent reduction of truck flows at the
port of Los Angeles and Long Beach.
95
Table 19: Results of sensitivity analysis for the Los Angeles MSA (percent change)
MHDT HHDT Total
-25% 25% -25% 25% -25% 25%
TOG
Scenario1 0.00% 0.00% 0.34% 0.00% 0.02% 0.00%
Scenario2 0.00% 0.00% 0.00% -0.34% 0.02% 0.00%
Scenario3 0.00% 0.00% 0.00% -0.33% 0.02% 0.00%
CO
Scenario1 0.16% -0.25% 0.21% -0.11% 0.02% -0.03%
Scenario2 0.08% -0.08% 0.05% -0.11% 0.01% -0.01%
Scenario3 -0.08% 0.00% -0.11% -0.11% -0.03% -0.01%
NOx
Scenario1 1.85% -1.85% 0.17% -0.17% 0.24% -0.19%
Scenario2 0.22% -0.22% 0.06% -0.09% 0.06% -0.05%
Scenario3 0.00% 0.00% 0.00% -0.03% -0.02% 0.02%
CO2
Scenario1 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
Scenario2 0.42% -0.42% 0.33% -0.33% 0.09% -0.06%
Scenario3 -0.41% 0.00% 0.00% 0.00% -0.12% 0.00%
PM
Scenario1 0.00% 0.00% 2.00% 0.00% 0.00% 0.00%
Scenario2 0.00% -4.55% 0.00% -1.89% 0.00% 0.00%
Scenario3 4.76% 0.00% 1.92% 0.00% -0.30% 0.00%
Note: SOx is excluded from the table because emissions are same for all scenarios
Table 19 shows the difference of -25 percent and 25 percent levels from 0 percent level
for each scenario. Figure 20 is a graphic representation of Table 19. CO emissions of
MHDT in Scenario One, for example, -25 percent (50 percent old vehicles replacement)
show 0.16 percent change which means 0.16 percent more emissions compared to the 0
percent level (75 percent old vehicles replacement) (12.20-12.18)/12.18=0.16 percent ).
For Scenario One, emissions of CO and NOx are reduced more when more trucks are
replaced. CO2, however, does not change for the various levels.
For Scenario Two, emissions of all pollutants are reduced when more lanes are converted
to zero-emission truck lanes. Scenario Three shows mixed results by levels and pollutants.
96
CO (MHDT) CO (HHDT) CO (Total)
NOx (MHDT) NOx (HHDT) NOx (Total)
CO2 (MHDT) CO2 (HHDT) CO2 (Total)
PM (MHDT) PM (HHDT) PM (Total)
Figure 20: Results of sensitivity analysis for the Los Angeles MSA (percent change)
-0.40%
-0.20%
0.00%
0.20%
-25% 0% 25%
-0.20%
-0.10%
0.00%
0.10%
0.20%
0.30%
-25% 0% 25%
-0.04%
-0.02%
0.00%
0.02%
-25% 0% 25%
-4.00%
-2.00%
0.00%
2.00%
4.00%
-25% 0% 25%
-0.20%
-0.10%
0.00%
0.10%
0.20%
-25% 0% 25%
-0.40%
-0.20%
0.00%
0.20%
0.40%
-25% 0% 25%
-0.50%
0.00%
0.50%
-25% 0% 25%
-0.50%
0.00%
0.50%
-25% 0% 25%
-0.20%
-0.10%
0.00%
0.10%
-25% 0% 25%
-10.00%
-5.00%
0.00%
5.00%
10.00%
-25% 0% 25%
-4.00%
-2.00%
0.00%
2.00%
4.00%
-25% 0% 25%
-0.40%
-0.20%
0.00%
-25% 0% 25%
97
5.3.2 Sensitivity analysis for Los Angeles County
Table 20 shows results of sensitivity analysis for Los Angeles County. The results of
Scenario One and Two are similar to the ones for the Los Angeles MSA. Scenario Three,
however, is different from the results for the Los Angeles MSA. Emissions are decreased
as more truck flows are moved from the port of Los Angeles/ Long Beach to the Mira
Loma area. This is because the Mira Loma area is located in Riverside County and Table
20 include only Los Angeles County. Table 21 and Figure 21 show percent changes by
various levels within each scenario.
The model makes it possible to test scenario results of VMT changes at the sub-county
levels. The current state of the EMFAC model, however, does not permit us to go to this
next step. If and when EMFAC is suitably elaborated to treat smaller areas, the model
will be suitably useful. Many policies have effects at the sub-county level, and the
model will be useful in analyzing these.
98
Table 20: Results of sensitivity analysis for the Los Angeles County
Units: tons per day
MHDT HHDT Total
-25% 0% 25% -25% 0% 25% -25% 0% 25%
TOG
Scenario1 0.70 0.70 0.70 1.03 1.02 1.02 24.56 24.55 24.55
Scenario2 0.70 0.70 0.70 1.04 1.04 1.03 24.58 24.57 24.57
Scenario3 0.70 0.70 0.70 1.04 1.04 1.03 24.57 24.56 24.54
CO
Scenario1 6.42 6.41 6.39 6.53 6.50 6.48 86.30 86.27 86.23
Scenario2 6.45 6.44 6.43 6.55 6.54 6.52 86.35 86.33 86.31
Scenario3 6.44 6.44 6.43 6.54 6.53 6.51 86.12 85.93 85.70
NOx
Scenario1 2.07 2.02 1.97 11.00 10.96 10.91 24.10 24.00 23.91
Scenario2 2.17 2.16 2.15 11.06 11.04 11.01 24.26 24.22 24.19
Scenario3 2.16 2.15 2.14 11.05 11.03 10.99 24.22 24.16 24.10
CO2
Scenario1 1.02 1.02 1.02 2.37 2.37 2.37 14.45 14.45 14.45
Scenario2 1.01 1.00 0.99 2.36 2.34 2.32 14.43 14.40 14.38
Scenario3 1.01 1.00 0.99 2.35 2.33 2.31 14.31 14.20 14.07
PM
Scenario1 0.09 0.09 0.09 0.21 0.21 0.21 1.46 1.46 1.46
Scenario2 0.10 0.10 0.09 0.22 0.22 0.21 1.47 1.47 1.47
Scenario3 0.10 0.09 0.09 0.22 0.21 0.21 1.46 1.45 1.44
SOx
Scenario1 0.01 0.01 0.01 0.02 0.02 0.02 0.14 0.14 0.14
Scenario2 0.01 0.01 0.01 0.02 0.02 0.02 0.14 0.14 0.14
Scenario3 0.01 0.01 0.01 0.02 0.02 0.02 0.14 0.14 0.14
Note: For scenario 1, -25%, 0, 25% mean 50 percent, 75 percent, 100 percent replacement of old trucks in
the Los Angeles county respectively.
For scenario 2, -25%, 0, 25% mean 25 percent, 50 percent, 75 percent reduction of MHDT and
HHDT on I-710 respectively.
For scenario 3, -25%, 0, 25% mean 25 percent, 50 percent, 75 percent reduction of truck flows at the
port of Los Angeles and Long Beach.
99
Table 21: Results of sensitivity analysis for Los Angeles County (percent change)
MHDT HHDT Total
-25% 25% -25% 25% -25% 25%
TOG
Scenario1 0.00% 0.00% 0.98% 0.00% 0.04% 0.00%
Scenario2 0.00% 0.00% 0.00% -0.96% 0.04% 0.00%
Scenario3 0.00% 0.00% 0.00% -0.96% 0.04% -0.08%
CO
Scenario1 0.16% -0.31% 0.46% -0.31% 0.03% -0.05%
Scenario2 0.16% -0.16% 0.15% -0.31% 0.02% -0.02%
Scenario3 0.00% -0.16% 0.15% -0.31% 0.22% -0.27%
NOx
Scenario1 2.48% -2.48% 0.36% -0.46% 0.42% -0.37%
Scenario2 0.46% -0.46% 0.18% -0.27% 0.17% -0.12%
Scenario3 0.47% -0.47% 0.18% -0.36% 0.25% -0.25%
CO2
Scenario1 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
Scenario2 1.00% -1.00% 0.85% -0.85% 0.21% -0.14%
Scenario3 1.00% -1.00% 0.86% -0.86% 0.77% -0.92%
PM
Scenario1 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
Scenario2 0.00% -10.00% 0.00% -4.55% 0.00% 0.00%
Scenario3 11.11% 0.00% 4.76% 0.00% 0.69% -0.69%
100
CO (MHDT) CO (HHDT) CO (Total)
NOx (MHDT) NOx (HHDT) NOx (Total)
CO2 (MHDT) CO2 (HHDT) CO2 (Total)
PM (MHDT) PM (HHDT) PM (Total)
Figure 21: Results of sensitivity analysis for the Los Angeles County (percent change)
-0.40%
-0.20%
0.00%
0.20%
-25% 0% 25%
-0.50%
0.00%
0.50%
1.00%
-25% 0% 25%
-0.40%
-0.20%
0.00%
0.20%
0.40%
-25% 0% 25%
-4.00%
-2.00%
0.00%
2.00%
4.00%
-25% 0% 25%
-0.50%
0.00%
0.50%
-25% 0% 25%
-0.50%
0.00%
0.50%
-25% 0% 25%
-2.00%
-1.00%
0.00%
1.00%
2.00%
-25% 0% 25%
-1.00%
0.00%
1.00%
-25% 0% 25%
-1.00%
0.00%
1.00%
-25% 0% 25%
-20.00%
-10.00%
0.00%
10.00%
20.00%
-25% 0% 25%
-10.00%
-5.00%
0.00%
5.00%
10.00%
-25% 0% 25%
-1.00%
0.00%
1.00%
-25% 0% 25%
101
Conclusions
Estimating GHGs and other pollutants is an important basis for regional transportation
planning. Accounting for the trucking sector has been a challenge because of data
limitations. This research demonstrates how input-output data at the ZIP code level along
with Freight Analysis Framework (FAF) data can be applied to estimate truck flows
between sub-state areas, and how the estimated truck flows can be used to evaluate
various scenarios of reducing air pollution emissions.
The model developed here has been applied to evaluate three plausible policy alternatives:
1) How much air pollution emissions such as PM and NOx are reduced by replacing old
trucks with newer models in Los Angeles County and how great are the impacts
throughout the Los Angeles MSA due to the truck upgrade in Los Angeles County.
2) How much are air pollution emissions reduced by introducing zero emission lanes on
I-710 in Los Angeles County,
3) How much are air pollution emissions reduced by developing an inland port at Mira
Loma area in Los Angeles County as well as throughout the Los Angeles MSA area.
A truck replacement strategy can be effective for reducing air pollution emissions in both
Los Angeles County and the surrounding MSA. Introducing zero emission lanes on a
major truck highway may deliver small impacts in the County or surrounding MSA
region, although it may have a significant impact on reducing air pollution emissions in
102
specific local areas. Developing an inland port, however, can increase air pollution
emissions in the MSA, although it can reduce emissions around the port areas.
As explained in Chapter 5, the EMFAC model provides county level air pollution
emissions estimations. In Scenario Two, unlike the other two scenarios, emissions
reductions occur only on the link of I-710 which is in the scenario area. It is possible that
restricting estimation to the area surrounding the I-710, the impact of Scenario Two can
be significant. The argument becomes clearer by comparing the percent changes of
Scenario Two in Tables 13 and 17. In Los Angeles County, for example, CO reduction in
percentage was 0.03 percent but 0.06 percent in the Los Angeles MSA. Estimating small
areas below the county level will be a next step of this research
Analyzing and comparing the results of three scenarios provides various lessons. First,
when considering a policy alternative to reduce air pollution emissions, it is important to
make the objectives clear. A strategy that reduces air pollution emissions in a specific
area can also increase emissions in the county or the MSA. Similarly there can be a
strategy that reduces air pollution emissions in the county or MSA, although the
reduction in a specific area is unlikely or even negative. If the objective is to reduce
overall air pollution emissions in large areas, the vehicle replacement strategy seems to
be promising. If the objective is to reduce air pollution emissions in a specific area such
as near highway segments, developing zero emission truck lanes could be an option.
Second, moving transport activities from one site to another could have both positive and
negative impacts. The total air pollution emissions may not be changed, although
103
emissions in a local area can be reduced. There are also possibilities to increase overall
emissions if proper developments of infrastructure are not implemented. For example, if
an inland port is developed in an area, there should be new developments of highways
and major arterials to improve network accessibility of the area.
The model developed here has limitations. First, the model may not evaluate congestion
effects properly because only freight flows are included, and passenger car flows are not
yet added in the trip assignment. When both passenger car flows and truck flows are
added, the results could be different.
Second, new technologies can change the model results. For the truck replacement
scenario, old diesel trucks are assumed to be replaced with newer diesel trucks. Recently
however, significant new natural gas reserves have been developed in the U.S. It is
possible that natural gas trucks will be more popular in 2030 because natural gas is likely
to be cheaper than diesel fuel. Of course there must be investments in developing
efficient trucks and proper infrastructure must be established to make natural gas trucks
popular. Natural gas trucks could not be included in the truck replacement strategy
because the EMFAC model does not yet include this fuel category. If natural gas trucks
are included in the model, there could be more reductions in air pollution emissions.
Electric trucks can also change the model results. EMFAC does not het accommodate
electric truck types. When the EMFAC model includes electric truck types, the results of
this analysis may be different.
104
Third, changes in supply chains resulting from the Panama Canal expansion may change
the model results. The baseline origin-destination truck flows matrix does not take into
account the Panama Canal expansion. It is not yet known the extent to which the
expansion will be a game changer or if current trends will continue.
These limitations of the model suggest the next steps in this research. Because including
passenger vehicles is important to estimate congestion effects, both passenger trips and
freight trips should be combined in the model. To do this it may be necessary to change
the study area to that of local Metropolitan Planning Organizations. Also the model can
be updated when more fuel types such as natural gas are accommodated in the EMFAC
model.
Emissions results have been estimated at the MSA and County levels in this study
because Counties are the smallest geographic units in the current version of the EMFAC
model. However, when methods become available to disaggregate county level emissions
down to smaller areas, the model can also be useful for analyzing environmental justice
questions since air pollution emissions (in particular particulate matter) affect local area
public health (Correia et. al., 2013).
105
Bibliography
Abdelwahab, W. M., and M. A. Sargious (1985) “A Simultaneous Decision-Making Approach to
Model the Demand for Freight Transportation,” CANADIAN JOURNAL OF CIVIL
ENGINEERING, 18 (3): 515-520.
Al-Battaineh, O., and I. A. Kaysi (2005) “Commodity-Based Truck Matrix Estimation Using
Input-Output Data and Genetic Algorithms,” TRANSPORTATION RESEARCH RECORD,
1923: 37-45.
Alam, M., E. Fekpe, M. Majed, and Battelle (2007) “FAF2 Freight Traffic Analysis,” Report
submitted to Office of Freight Management and Operations (HOFM), Federal Highway
Administration, Washington, D.C. Available at
http://ops.fhwa.dot.gov/freight/freight_analysis/faf/faf2_reports/reports7/c1_intro.htm.
Anderson, M. D., G. A. Harris, and K. Harrison (2010) “Using Aggregated Federal Data to Model
Freight in a Medium-Size Community,” TRANSPORTATION RESEARCH RECORD, 2174:
39-43.
Anderson, M., G. A. Harris, S. Jeereddy, S. Gholston, J. Swain, and N. Schoening (2008) “Using
a Federal Database and New Factors for Disaggregation of Freight to a Local Level,”
PROCEEDINGS OF THE 10TH INTERNATIONAL CONFERENCE ON APPLICATION OF
ADVANCED TECHNOLOGIES IN TRANSPORTATION, Athens, Greece, May.
Ashtakala, B., and A. S. N. Murthy (1988) “Optimized Gravity Models for Commodity
Transportation,” TRANSPORTATION ENGINEERING JOURNAL, 114 (4): 393-408.
Bai, S., D. Eisinger and D. Niemeier (2009) “MOVES vs. EMFAC: A Comparison of Greenhouse
Gas Emissions Using Los Angeles County,” Presented at Transportation Research Board 88th
Annual Meeting, Washington D.C.
Barth, M., and K. Boriboonsomsin (2009) “Traffic Congestion and Greenhouse Gases,” TR
NEWS, 268: 26.
Barth, M., F. An, J. Norbeck, and M. Ross (1996) “Modal Emissions Modeling: A Physical
Approach,” TRANSPORTATION RESEARCH RECORD, 1520: 81-88.
Bell, M. (1984) “Log-linear Models for the Estimation of Origin-Destination Matrices from
Traffic Counts,” PROCEEDINGS OF THE NINTH INTERNATIONAL SYMPOSIUM ON
TRANSPORTATION AND TRAFFIC THEORY, Delft, The Netherlands.
Bell, M. (1991) “The Estimation of Origin-Destination Matrices by Constrained Generalized
Least Squares,” TRANSPORTATION RESEARCH B, 25 (1): 13-22.
Benjamin, M., and J. R. Long (1995) “Application of the Global Positioning System (GPS) to the
Collection of Vehicle Dynamics Data,” Presented at the 5th CRC On-road Vehicle Emission
Workshop, San Diego, California.
Berwick, M., and M. Farooq (2003) “Truck Costing Model for Transportation Managers,” Upper
106
Great Plains Transportation Institute, North Dakota State University, available at
http://www.mountain-plains.org/pubs/html/mpc-03-152/index.php.
Bierlaire, M., and Ph. L. Toint (1995) “MEUSE: an Origin-Destination Estimator that Exploits
Structure,” TRANSPORTATION RESEARCH B, 29 (1): 47-60.
Boerkamps, J. H. K., A. J. V. Binsbergen, and P. H. L. Bovy (2000) “Modeling Behavioral
Aspects of Urban Freight Movement in Supply Chains,” TRANSPORTATION RESEARCH
RECORD, 1725: 17-25.
Brenniger-Gothe, M., K. O. Jornsten, and J. T. Lundgren (1989) “Estimation of Origin-
Destination Matrix from Traffic Counts Using Multiobjective Programming Formulations,”
TRANSPORTATION RESEARCH B, 23 (4): 257-265.
California Environmental Protection Agency Air Resources Board (CARB) (2007) “EMFAC2007
Version 2.30 User’s Guide,” Available at
http://www.arb.ca.gov/msei/onroad/downloads/docs/user_guide_emfac2007.pdf.
Caliper Co. (2004) “Travel Demand Modeling with TransCAD,” Available at
http://www.caliper.com/PDFs/TravelDemandModelingBrochure.pdf.
Cascetta, E. (1984) “Estimation of Trip Matrices from Traffic Counts and Survey Data: a
Generalized Least Squares Estimator,” TRANSPORTATION RESEARCH B, 18 (4-5): 289-299.
Cascetta, E., and S. Nguyen (1988) “A Unified Framework for Estimating or Updating
Origin/Destination Matrices from Traffic Counts,” TRANSPORTATION RESEARCH B, 22 (6):
437-455.
Chen, Y. (1994) “Bilevel Programming Problems: Analysis, Algorithms and Applications,” PhD
thesis, report CRT-984, Centre de Recherche sur les Transports, Universite de Montreal,
Montreal, Canada.
Cicero-Pernandez, P., and J. R. Long (1995) “Grades and Other Loads Effects on On-road
Emissions: an On-board Analyzer Study,” Presented at the 5th CRC On-road Vehicle
Emission Workshop, San Diego, California.
Cicero-Pernandez, P., and J. R. Long (1996) “Assessment of Commuting under Grade Loads and
Ramp Metering: Preliminary On-road Emissions Findings,” Presented at the 3rd World Car
Conference, Riverside, California.
Codina, E., and J. Barcelo (2004) “Adjustment of O–D Trip Matrices from Observed V olumes: an
Algorithmic Approach Based on Conjugate Directions,” EUROPEAN JOURNAL OF
OPERATIONAL RESEARCH, 155: 535–557.
Colston, M., and W. R. Blunden (1970) “On the Duality of Desire Line and Land Use Models,”
PROCEEDINGS OF AUSTRALIAN ROAD RESEARCH BOARD, 4 (1): 170-183.
Correia, A. W., A. Pope III, D. W. Dockery, Y. Wang, M. Ezzati, and F. Dominici (2013) “Effect
of Air Pollution Control on Line Expectancy in the United States,” EMIDEMIOLOGY, 24 (1):
23-31.
Dolby, G.R. (1972) “Generalized Least Squares and Maximum Likelihood Estimation of
107
Nonlinear Functional Relationships,” JOURNAL OF THE ROYAL STATISTICAL SOCIETY.
SERIES B (METHODOLOGICAL), 34 (3): 393-400.
Donnelly, R. (2007) “A Hybrid Microsimulation Model of Freight Flows,” PROCEEDINGS OF
THE 4TH INTERNATIONAL CONFERENCE ON CITY LOGISTICS, ed. Taniguchi, E. and R.
G. Thompson, 235-246. Institute for City Logistics, Crete, Greece.
Drissi-Kaitouni, O., and J. Lundgren (1992) “Bilevel Origin-Destination Matrix Estimation Using
a Descent Approach,” Technical report LiTH-MAT-R-92-49, Department of Mathematics,
Linkoping, Institute of Technology, Linkoping, Sweden.
Fisk, C. S. (1988) “On Combining Maximum Entropy Trip Matrix Estimation with User Optimal
Assignment,” TRANSPORTATION RESEARCH B, 22 (1): 69-73.
Florian, M., and Y. Chen (1993) “A Coordinate Descent Method for the Bilevel OD Matrix
Adjustment Problem,” Presented at the IFORS Conference in Lisbon, Portugal.
Gaudry, M., and L. LaMarre (1978) “Estimating Origin-Destination Matrices from Traffic Counts:
A Simple Linear Intercity Model for Quebec,” Publication No. 105, Centre de Recherche sur
les Transports, Universite de Montreal, Montreal, Canada.
Gedeon, C., M. Florian, and T. Crainic (1993) “Determining Origin-Destination Matrices and
Optimal Multi-product Flows for Freight Transportation over Multimodal Networks,”
TRANSPORTATION RESEARCH B, 27 (5): 351-368.
Geva, E., E. Haur, and U. Landau (1983) “Maximum Likelihood and Bayesian Methods for the
Estimation of OD Flows,” TRANSPORTATION RESEARCH RECORD, 944: 101–105.
Giuliano, G., P. Gordon, Q. Pan, J. Y. Park, and L. L. Wang (2010) “Estimating Freight Flows for
Metropolitan Area Highway Networks Using Secondary Data Sources,” NETWORKS AND
SP ATIAL ECONOMICS, 10: 73-91.
Gliebe, J., O. Cohen, and J. D. Hunt (2007) “A Dynamic Choice Model of Urban Commercial
Vehicle and Person Activity Patterns,” Presented at the 86th Transportation Research Board
Annual Meeting, Washington D.C.
Gordon, P., J. K. Cho, J. E. Moore, J. Y. Park, H. W. Richardson, and S. S. Yoon (2009) “Adding
a Freight Network to a National Interstate Input-Output Model: Implications for California,”
Available at http://www.metrans.org/research/final/07-19%20Final.pdf.
Harley, R. A., S. N. Griddings, and L. C. Marr (2004) “Decadal Trends in Air Pollutant Emissions
from Motor Vehicles in Central California,” Prepared for San Joaquin Valleywide Air
Pollution Study Agency and California Air Resources Board, contract 00-14CCOS.
Harris, G. A., P. A. Farrington, M. D. Anderson, N. Schoening, J. Swain, and N. Sharma (2009)
“Developing Freight Analysis Zones at State Level: Cluster Analysis Approach,” Presented at
Transportation Research Board 88th Annual Meeting, Washington, D.C.
Hatzopoulou, M., and E. J. Miller (2010) “Linking an Activity-based Travel Demand Model with
Traffic Emission and Dispersion Model: Transport’s Contribution to Air Pollution in Toronto,”
TRANSPORTATION RESEARCH D, 15: 315-325.
Hautzinger, H. (1984) “The Prediction of Interregional Goods Vehicle Flows: Some New
Modeling Concepts,” Presented at the 9th International Symposium on Transportation and
108
Tra ffic Theory, July, The Netherlands.
Hogberg, P. (1986) “Estimation of Parameters in Models for Traffic Prediction: a Nonlinear
Regression Approach,” TRANSPORTATION RESEARCH, 10 (4): 263-265.
Holguin-Veras, J. (2002) “Revealed Preference Analysis of Commercial Vehicle Choice Process,”
JOURNAL OF TRANSPORTATION ENGINEERING, 128 (4): 336-346.
Holguin-Veras, J., and E. Thorson (2000) “Trip Length Distribution in Commodity Based and
Trip Based Freight Demand Modeling: Investigation of Relationships,” TRANSPORTATION
RESEARCH RECORD, 1707: 37-48.
Holguin-Veras, J., G. List, A. Meyburg, K. Ozbay, R. Paaswell, H. Teng, and S. Yahalom (2001)
“An Assessment of Methodological Alternatives for a Regional Freight Model in the
NYMTC Region,” New York City Metropolitan Transportation Council (NYMTC), May 30.
Holguin-Veras, J., and E. Thorson (2003) “Modeling Commercial Vehicle Empty Trips with a
First Order Trip Chain Model,” TRANSPORTATION RESEARCH B, 37 (2): 129-148.
Holguin-Veras, J., and G. R. Patil (2007) “An Integrated Commodity Based / Empty Trip Freight
Origin-Destination Synthesis Model,” TRANSPORTATION RESEARCH RECORD, 2008: 60-
66.
Holguin-Veras, J., and G. R. Patil (2008) “A Multicommodity Integrated Freight Origin-
Destination Synthesis Model,” NETWORKS AND SP ATIAL ECONOMICS, 8(2-3): 309-326.
Holm, J., T. Jensen, S. K. Nielsen, A. Christensen, B. Johnsen, and G. Ronby (1976) “Calibrating
Traffic Models on Traffic Census Results Only,” TRAFFIC ENGINEERING AND CONTROL,
17(4): 137-140.
Horowitz, A. J. (2009) “Origin-Destination Table Disaggregation Using Fratar Biproportional
Least Squares Estimation,” Working Paper 09-1, Center for Urban Transportation Studies,
University of Wisconsin-Milwaukee, Milwaukee, Wisconsin.
Hunt, J.D., and K. J. Stefan (2007) “Tour-based Microsimulation of Urban Commercial
Movements,” TRANSPORTATION RESEARCH B, 41(9): 981-1013.
Jack Faucett Associates, Inc. (1999) “Research and Development of Destination, Mode, and
Routing Choice Models for Freight,” Final Report Prepared for DOT SBIR Office, DTS-22,
May 20.
Jones & Stokes Associates, Inc. (2004) “Port of Los Angeles Portwide Rail Synopsis Review
Draft,” Prepared for Port of Los Angeles. Available at
http://www.portoflosangeles.org/DOC/REPORT_Draft_Rail_Synopsis.pdf.
Jornsten, K. O., and S. Nguyen (1979) “On the Estimation of a Trip Matrix from Network Data,”
Research Report LiTH-MAT-R-79-36, Linkoping University, Linkoping, Sweden.
Kahn Ribeiro, S., S. Kobayashi, M. Beuthe, J. Gasca, D. Greene, D. S. Lee, Y. Muromachi, P. J.
Newton, S. Plotkin, D. Sperling, R. Wit, and P. J. Zhou (2007) “Transport and its
Infrastructure,” CLIMATE CHANGE 2007: MITIGATION. Contribution of Working Group
III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, ed. B.
Metz, O. R. Davidson, P. R. Bosch, R. Dave, and L. A. Meyer, Cambridge University Press,
Cambridge, UK and New York, USA.
109
LeBlanc, L. J., and K. Fahrangian (1982) “Selection of a Trip Table which Reproduces Observed
Link Flows,” TRANSPORTATION RESEARCH B, 16: 83-88.
Lents, J., M. Walsh, K. He, N. Davis, M. Osses, S. Tolvett, and H. Liu (2011) “Handbook of Air
Quality Management,” Available at http://www.aqbook.org/.
Lindal S., D. Olson, and G. Alward (2006) “Deriving Multi-Regional Models using the IMPLAN
National Trade Flows Model,” JOURNAL OF REGIONAL ANALYSIS AND POLICY, 36 (1):
76-83.
List, G., and M. Turnquist (1994) “Estimating Truck Travel Patterns in Urban Areas,”
TRANSPORTATION RESEARCH RECORD, 1430: 1-9.
Litman, T. A. (2009) “Transportation Cost and Benefit Analysis: Techniques, Estimates and
Implications,” Victoria Transport Policy Institute. Available at http://www.vtpi.org/tca/.
Low, D. E. (1972) “A New Approach to Transportation Systems Modeling,” TRAFFIC
QUARTERLY, 26 (3): 391-404.
Lundgren, J. T., and A. Peterson (2008) “A Heuristic for the Bilevel Origin-Destination Matrix
Estimation Problem,” TRANSPORTATION RESEARCH B, 42 (4): 339-354.
Magbuhat, S., and J. R. Long (1996) “Improving Califonia’s Moter Vehicle Emissions Inventory
Activity Estimates through the Use of Datalogger-equipped Vehicles,” Presented at the 6th
CRC On-road Vehicle Emissions Workshop, San Diego, California.
Maher, M. J. (1983) “Inferences on Trip Matrices from Observations on Link Volumes: A
Bayesian Statistical Approach,” TRANSPORTATION RESEARCH B, 17 (6): 435-447.
Mao, S., and M. J. Demetsky (2002) “Calibration of the Gravity model for Truck Freight Flow
Distribution,” Research Project Report, Mid-Atlantic Universities Transportation Center
(MAUTC). Available at http://cts.virginia.edu/docs/UVACTS-5-14-14.pdf.
McFadden, D., C. Winston, and A. Boersch-Supan (1986) “Joint Estimation of Freight
Transportation Decisions under Non-random Sampling,” ANALYTICAL STUDIES IN
TRANSPORT ECONOMICS, ed. Daugherty, A. Cambridge University Press, Cambridge, UK.
Meyburg, A. H. (1976) “Modeling in the Context of Urban Goods Movement Problems,” 2ND
CONFERENCE ON GOODS TRANSPORTATION IN URBAN AREAS, ed. Fisher, G.P., 127-
168. Engineering Foundation, New York.
Nam, E. K. (2003) “Proof of Concept Investigation for the Physical Emission Rate Estimator
(PERE) to be Used in MOVES,” Assessment and Standards Division Office of Transportation
and Air Quality, U.S. EPA. Available at http://www.epa.gov/otaq/models/ngm/r03005.pdf.
Nguyen, S. (1977) “Estimating an O-D Matrix from Network Data: A Network Equilibrium
Approach,” Publication No. 87, Centre de Recherche sur les Transports, Universite de
Montreal, Montreal, Canada.
Noortman, H. J., and J. van Es (1978) “Tra ffic Model,” Manuscript for the Dutch Freight
Transport Model.
110
North Carolina State University (2002) “Methodology for Developing Modal Emission Rates for
EPA’s Multi-Scale Motor Vehicle and Equipment Emission System,” Assessment and
Standards Division Office of Transportation and Air Quality, U.S. EPA. Available at
http://www.epa.gov/otaq/models/ngm/r02027.pdf.
Odgen, K. W. (1978) “The Distribution of Truck Trips and Commodity Flow in Urban Areas: A
Gravity Model Analysis,” TRANSPORTATION RESEARCH, 12 (2): 131-137.
Opie, K., J. Rowinski, and L. N. Spasovic (2009) “Commodity-Specific Disaggregation of 2002
Freight Analysis Framework Data to County Level for New Jersey,” TRANSPORTATION
RESEARCH RECORD, 2121: 128-134.
Park, J. Y. (2008) “The Economic Impacts of Dirty Bomb Attacks on the Los Angeles and Long
Beach Ports: Applying the Supply-Driven NIEMO (National Interstate Economic Model),”
JOURNAL OF HOMELAND SECURITY AND EMERGENCY MANAGEMENT, 5 (1): 10.
Park, J. Y., P. Gordon, J. E. Moore II, and H. W. Richardson (2008) “The State-by-State
Economic Impacts of the 2002 Shutdown of the Los Angeles–Long Beach Ports,” GROWTH
AND CHANGE, 39 (4): 548-572.
Patel, D., and M. Carlock (1995) “A study of the Relative Benefits of On-board Diagnostic and
Inspection and Maintenance in California,” SAE Technical Paper 951944, California Air
Resource Board. Available at http://www.arb.ca.gov/msei/onroad/downloads/pubs/obdim.pdf.
Rios, A., L. K. Nozick, and M. A. Turnquist (2003) “Value of Different Categories of Information
in Estimating Freight Origin-Destination Tables,” TRANSPORTATION RESEARCH
RECORD, 1783: 42-48.
Robillard, P. (1975) “Estimating an O-D Matrix from Observed Link Volumes,”
TRANSPORTATION RESEARCH, 9 (2-3): 123-128.
Rowinski, J., K. Opie, and L. N. Spasovic (2008) “Development of Method to Disaggregate 2002
FAF2 Data down to County Level for New Jersey,” Presented at Transportation Research
Board 87th Annual Meeting, Washington, D.C.
Sheffi, Y. (1985) “Urban Transportation Networks: Equilibrium Analysis with Mathematical
Programming Methods,” Prentice-Hall, Inc., Englewood Cliffs, NJ. Available at
http://web.mit.edu/sheffi/www/selectedMedia/sheffi_urban_trans_networks.pdf.
Sherali, H. D., R. Sivanandan, and A. G. Hobeika (1994a) “A Linear Programming Approach for
Synthesizing Origin-Destination Trip Tables from Link Traffic Volumes,”
TRANSPORTATION RESEARCH B, 28 (3): 213-233.
Sherali, H. D., R. Sivanandan, A. G. Hobieka, and A. Narayanan (1994b) “Estimating Missing
Link Volumes in a Traffic Network – a Linear Programming Approach,” Presented at the
TRB Annual Meeting, Washington, D.C.
111
Sherali, H. D., A. Narayanan, and R. Sivanandan (2003) “Estimation of Origin-Destination Trip-
Tables Based on a Partial Set of Tra ffic Link Volumes,” TRANSPORTATION RESEARCH B,
37 (9): 815-836.
Sivakumar, A., and C. Bhat (2002) “Fractional Split-Distribution Model for Statewide
Commodity-Flow Analysis,” TRANSPORTATION RESEARCH RECORD, 1790: 80-
88.Sivanandan, R. (1991) “A Linear Programming Approach for Synthesizing Origin-
Destination (O-D) Trip Tables from Link Traffic Volumes,” PhD thesis, Virginia Polytechnic
Institute & State University, Blacksburg, Virginia.
Smith, R., and W. McFarlane (1978) “Examination of a Simplified Travel Demand Model,”
TRANSPORTATION ENGINEERING JOURNAL, 104 (1): 31-41.
Southworth, F. (1982) “The Spatial Accessibility of Truck Terminals in the Presence of Multi
Destination Truck Circuits,” MODELING AND SIMULATION, 13 (3): 1073-1080.
Spiess, H. (1987) “A Maximum Likelihood Model for Estimating Origin-Destination Matrices,”
TRANSPORTATION RESEARCH B, 21 (5): 395-412.
Spiess, H. (1990) “A Descent Based Approach for the OD Matrix Adjustment Problem,”
Publication No. 693, Centre de Recherchesur les Transports, Universite de Montreal,
Montreal, Canada.
Swan Wooster Engineering Co. Ltd. (1979) “Evaluation of Urban Trucking Rationalization in
Vancouver – phase 1 and 2,” Transport Canada, Montreal, Canada.
Tamin, O. Z. and L. G. Willumsen (1988) “Freight Demand Model Estimation from Traffic
Counts,” PROCEEDINGS OF THE 16TH PTRC SWNMER ANNUAL MEETING, University
of Bath, England.
Tavasszy, L. A., J. E. Stada, and R. Hamerslag (1994) “The Impact of Decreasing Border Barriers
in Europe on Freight Transport Flows by Road,” PROCEEDINGS OF THE 36TH ANNUAL
CONFERENCE OF THE TRANSPORTATION RESEARCH FORUM, Florida, USA.
U.S. Environmental Protection Agency (EPA) (2009) “Draft Motor Vehicle Emission Simulator
(MOVES) 2009 Software Design and Reference Manual,” Available at
http://www.epa.gov/oms/models/moves/420b09007.pdf.
U.S. Environmental Protection Agency (EPA) (2010a) “Motor Vehicle Emission Simulator
(MOVES) User Guide for MOVES2010a,” Available at
http://www.epa.gov/otaq/models/moves/MOVES2010a/420b10036.pdf.
U.S. Environmental Protection Agency (EPA) (2010b) “Inventory of U.S. Greenhouse Gas
Emissions and Sinks: 1990-2008,” Available at
http://www.epa.gov/climatechange/emissions/usinventoryreport.html.
Van Zuylen, H. and L. G. Willumsen (1980) “The Most Likely Trip Matrix Estimated from
Traffic Counts,” TRANSPORTATION RESEARCH B, 14 (3): 281-293.
Viswanathan, K., D. F. Beagan, V. Mysore, and N. N. Srinivasan (2008) “Disaggregating Freight
Analysis Framework Version 2 Data for Florida: Methodology and Results,”
TRANSPORTATION RESEARCH RECORD, 2049: 167–175.
Walting, D. P., and D. R. Grey (1991) “Analysis of Partial Registration Plate Data Using a Model
112
with Poisson Input and Output,” PROCEEDINGS OF THE INTERNATIONAL
CONFERENCE ON MATHEMATICS IN TRANSPORT PLANNING AND CONTROL,
Institute of Mathematics and Its Applications, University of Wales College of Cardiff, Oxford
University Press, Sep. 1989.
Walting, D. P., and M. J. Maher (1992) “A Statistical Procedure for Estimating a Mean Origin-
Destination Matrix from a Partial Registration Plate Survey,” TRANSPORTATION
RESEARCH B, 26 (3): 171-193.
Wang, Q., and J. Holguin-Veras (2008) “Investigation of Attributes Determining Trip Chaining
Behavior in Hybrid Microsimulation Urban Freight Models,” TRANSPORTATION
RESEARCH RECORD, 2066: 1-8.
Wang, Q., and J. Holguin-Veras (2009) “Tour-Based Entropy Maximization Formulations of
Urban Freight Demand,” Presented at Transportation Research Board 88th Annual Meeting,
Washington D.C.
Wardrop, J. G. (1952) “Some Theoretical Aspects of Road Traffic Research,” INSTITUTION OF
CIVIL ENGINEERS PROCEEDINGS (P ART II), 1: 325–378.
Willumsen, L. (I 978) “Estimation of an O-D Matrix from Traffic Counts: a Review,” Working
Paper 99, Institute for Transport Studies, University of Leeds, Leeds, UK.
Wisetjindawat, W., and K. Sano (2003) “A Behavioral Modeling in Micro-simulation for Urban
Freight Transportation,” JOURNAL OF THE EASTERN ASIA SOCIETY FOR
TRANSPORTATION STUDIES, 5: 2193-2208.
Wisetjindawat, W., K. Sano, and S. Matsumoto (2006) “Commodity Distribution Model
Incorporating Spatial Inter-actions for Urban Freight Movement,” TRANSPORTATION
RESEARCH RECORD, 1966: 41-50.
Yang, H., Y. Iida, and T. Sasaki (1994) “The Equilibrium-based Origin-Destination Matrix
Estimation Problem,” TRANSPORTATION RESEARCH B, 28 (1): 23-33.
Yang, H., T. Sasaki, Y. Iida, and Y. Asakura (1992) “Estimation of Origin-Destination Matrices
from Link Traffic Counts on Congested Networks,” TRANSPORTATION RESEARCH B, 26
(6): 417-434.
Yang, H. (1995) “Heuristic Algorithms for the Bilevel Origin-Destination Matrix Estimation
Problem,” TRANSPORTATION RESEARCH B, 29 (4): 231-242.
113
Appendices
Appendix A: Data for OD estimation
Appendix Table 1: Metropolitan Areas with Component Counties
Metropolitan Areas Component Counties MSA (FAF)
1 Bakersfield Kern County 69
2 Chico
Butte County 69
Glenn County 69
Colusa County 69
Plums County 69
Lassen County 69
Modoc County 69
Lake County 69
3 El Centro Imperial County 69
4 Fresno Fresno County 69
5 Hanford-Corcoran Kings County 69
6 Los Angeles-Long Beach-Santa Ana
Los Angeles County 61
Orange County 61
7 Madera
Madera County 69
Mariposa County 69
Mono County 69
Inyo County 69
Amador County 69
Alpine County 69
Calaveras County 69
Tuolumne County 69
8 Merced Merced County 69
9 Modesto Stanislaus County 69
10 Napa Napa County 64
11 Oxnard-Thousand Oaks-Ventura Ventura County 61
12 Redding
Shasta County 69
Trinity County 69
Siskiyou County 69
Tehama County 69
Mendocino County 69
Humboldt County 69
Del Norte County 69
13 Riverside-San Bernardino-Ontario
Riverside County 61
San Bernardino County 61
14 Sacramento--Arden-Arcade—Roseville
El Dorado County 62
Placer County 62
Sacramento County 62
Yolo County 62
Nevada County 62
15 Salinas Monterey County 69
16 San Diego-Carlsbad-San Marcos San Diego County 63
17 San Francisco-Oakland-Fremont
Alameda County 64
Contra Costa County 64
Marin County 64
San Francisco County 64
San Mateo County 64
18 San Jose-Sunnyvale-Santa Clara
San Benito County 64
Santa Clara County 64
19 San Luis Obispo-Paso Robles San Luis Obispo County 69
20 Santa Barbara-Santa Maria Santa Barbara County 69
21 Santa Cruz-Watsonville Santa Cruz County 64
22 Santa Rosa-Petaluma Sonoma County 64
23 Stockton San Joaquin County 69
24 Vallejo-Fairfield Solano County 64
25 Visalia-Porterville Tulare County 69
26 Yuba-Sutter
Sutter County 69
Yuba County 69
Sierra County 69
Note:
61: Los Angeles MSA (LA), 62: Sacramento MSA (SA), 63: San Diego MSA (SD), 64: San Francisco
MSA (SF), 69: Remainder of MSA (RE)
114
Appendix Figure 1: California’s 58 Counties
115
Appendix Table 2: SCTG Sector descriptions
SCTG Context
VIUS
01 Live animals/fish 1
02 Cereal grains 3
03 Other ag prods. 4
04 Animal feed 2
05 Meat/seafood 11
06 Milled grain prods. 10
07 Other foodstuffs 13
08 Alcoholic beverages 9
09 Tobacco prods. 12
10 Building stone 36
11 Natural sands 37
12 Gravel 34
13 Nonmetallic minerals 38
14 Metallic ores 35
15 Coal 32
16 Crude petroleum 33
17 Gasoline 40
18 Fuel oils 39
19 Coal-n.e.c. 42
20 Basic chemicals 5
21 Pharmaceuticals 7
22 Fertilizers 6
23 Chemical prods. 8
24 Plastics/rubber 41
25 Logs 14
26 Wood prods. 18
27 Newsprint/paper 17
28 Paper articles 15
29 Printed prods. 16
30 Textiles/leather 29
31 Nonmetal min. prods. 21
32 Base metals 20
33 Articles-base metal 19
34 Machinery 26
35 Electronics 24
36 Motorized vehicles 30
37 Transport equip. 31
38 Precision instruments 28
39 Furniture 25
40 Misc. mfg. prods. 27
41 Waste/scrap 44
Source: VIUS: US Census Bureau (http://www.census.gov/svsd/www/vius/products.html)
SCTG: U.S. Department of Transportation Bureau of Transportation Statistics (www.bts.gov)
116
Appendix Table 3: Bridge of vehicle class categories between VIUS and EMFAC
VIUS EMFAC
Adjusted Avg.
payload (lbs)
Vehicle
group
Gross
Vehicle
Weight
Avg. Payload(lbs)
for California
Vehicle
class
Description
Weight
Class(lbs)
Group 1
Less than
6,000 lbs.
-
LDT1
Light-Duty
Trucks
0-3750
2,116
LDT2
Light-Duty
Trucks
3751-
5750
Group 2
6,001 to
10,000 lbs.
2,116
MDT
Medium-Duty
Trucks
5751-
8500
LHDT
1
Light-Heavy-
Duty Trucks
8501-
10000
Group 3
10,001 to
14,000 lbs.
3,945
LHDT
2
Light-Heavy-
Duty Trucks
10001-
14000
3,945
Group 4
14,001 to
16,000 lbs.
4,560
MHDT
Medium-Heavy-
Duty Trucks
14001-
33000
11,797
Group 5
16,001 to
19,500 lbs.
5,097
Group 6
19,501 to
26,000 lbs.
8,518
Group 7
26,001 to
33,000 lbs.
29,012
Group 8
More than
33,000 lbs
31,550 HHDT
Heavy-Heavy-
Duty Trucks
33001-
60000
31,550
Data: Vehicle Inventory Use Survey 2002
(http://ops.fhwa.dot.gov/freight/freight_analysis/faf/faf2_reports/reports9/s501_2_3_tables.htm#_Toc16939
9555), EMFAC model
Note: Group 1 of VIUS has too little sample to calculate average payload
Same payload is applied for LDT1, LDT2, MDT, and LHDT1
117
Appendix Table 4: Truck use percentages by SCTG Sector
SCTG LDT1 LDT2 MDT LHDT1 LHDT2 MHDT HHDT TOTAL
1 3% 3% 3% 3% 8% 20% 58% 100%
2 0% 0% 0% 0% 0% 36% 62% 100%
3 2% 2% 2% 2% 4% 26% 62% 100%
4 3% 3% 3% 3% 4% 30% 55% 100%
5 1% 1% 1% 1% 1% 18% 78% 100%
6 3% 3% 3% 3% 10% 34% 44% 100%
7 1% 1% 1% 1% 1% 33% 61% 100%
8 0% 0% 0% 0% 1% 57% 41% 100%
9 1% 1% 1% 1% 1% 12% 82% 100%
10 1% 1% 1% 1% 3% 14% 80% 100%
11 1% 1% 1% 1% 1% 20% 76% 100%
12 0% 0% 0% 0% 1% 13% 84% 100%
13 0% 0% 0% 0% 1% 21% 76% 100%
14 0% 0% 0% 0% 4% 7% 89% 100%
15 0% 0% 0% 0% 1% 3% 94% 100%
16 0% 0% 0% 0% 1% 14% 85% 100%
17 0% 0% 0% 0%
9% 90% 100%
18 1% 1% 1% 1% 2% 47% 48% 100%
19 0% 0% 0% 0% 2% 54% 42% 100%
20 1% 1% 1% 1% 1% 20% 76% 100%
21 6% 6% 6% 6% 4% 28% 43% 100%
22 1% 1% 1% 1% 2% 32% 62% 100%
23 3% 3% 3% 3% 7% 19% 63% 100%
24 2% 2% 2% 2% 7% 22% 63% 100%
25 1% 1% 1% 1% 3% 12% 81% 100%
26 2% 2% 2% 2% 5% 31% 57% 100%
27 1% 1% 1% 1% 1% 13% 83% 100%
28 1% 1% 1% 1% 2% 22% 71% 100%
29 6% 6% 6% 6% 9% 27% 40% 100%
30 3% 3% 3% 3% 7% 29% 50% 100%
31 0% 0% 0% 0% 1% 7% 90% 100%
32 1% 1% 1% 1% 5% 23% 67% 100%
33 4% 4% 4% 4% 8% 28% 47% 100%
34 1% 1% 1% 1% 3% 15% 77% 100%
35 5% 5% 5% 5% 12% 22% 45% 100%
36 3% 3% 3% 3% 6% 35% 47% 100%
37 1% 1% 1% 1% 4% 20% 74% 100%
38 10% 10% 10% 10% 13% 17% 28% 100%
39 2% 2% 2% 2% 7% 21% 64% 100%
40 4% 4% 4% 4% 6% 26% 54% 100%
41 1% 1% 1% 1% 3% 22% 70% 100%
Data: VIUS 2002
118
Appendix Table 5: Los Angeles MSA import trade proportions
Percentage of domestic imports by origin
SCTG LA SA SD SF RE OS
1 0.5684 0.0000 0.0623 0.0000 0.0002 0.3691
2 0.5826 0.0313 0.1279 0.0000 0.0120 0.2463
3 0.7804 0.0037 0.0709 0.0139 0.0579 0.0732
4 0.6927 0.0000 0.0489 0.0115 0.0247 0.2221
5 0.6986 0.0003 0.0389 0.0145 0.0602 0.1875
6 0.7220 0.0161 0.0235 0.0066 0.0470 0.1849
7 0.6499 0.0191 0.0143 0.0598 0.0839 0.1730
8 0.8481 0.0000 0.0037 0.0323 0.0654 0.0505
9 0.8980 0.0000 0.0125 0.0000 0.0268 0.0626
10 0.7877 0.0000 0.0126 0.0872 0.0530 0.0595
11 0.9040 0.0000 0.0048 0.0158 0.0065 0.0690
12 0.9374 0.0028 0.0069 0.0000 0.0398 0.0131
13 0.7028 0.0015 0.0175 0.0117 0.0245 0.2421
14 0.0518 0.0000 0.0007 0.0000 0.0006 0.9470
15 0.1088 0.0000 0.0019 0.0000 0.0017 0.8876
16 0.1475 0.0887 0.1253 0.1440 0.4826 0.0119
17 0.9618 0.0002 0.0012 0.0238 0.0079 0.0051
18 0.8843 0.0000 0.0072 0.0377 0.0215 0.0493
19 0.5764 0.0000 0.0016 0.0084 0.0123 0.4014
20 0.4236 0.0003 0.0111 0.0143 0.0067 0.5440
21 0.7049 0.0058 0.0028 0.0285 0.0143 0.2437
22 0.9039 0.0011 0.0063 0.0012 0.0535 0.0340
23 0.6578 0.0009 0.0102 0.0169 0.0106 0.3036
24 0.7097 0.0039 0.0222 0.0184 0.0357 0.2100
25 0.8986 0.0000 0.0117 0.0000 0.0103 0.0794
26 0.7311 0.0238 0.0158 0.0099 0.0611 0.1583
27 0.6298 0.0000 0.0003 0.0171 0.0124 0.3404
28 0.7316 0.0007 0.0106 0.0155 0.0328 0.2088
29 0.6278 0.0095 0.0187 0.0176 0.0155 0.3110
30 0.6401 0.0009 0.0363 0.0294 0.0377 0.2556
31 0.8073 0.0017 0.0365 0.0136 0.0606 0.0802
32 0.6748 0.0001 0.0115 0.0383 0.0269 0.2485
33 0.7786 0.0024 0.0304 0.0190 0.0366 0.1329
34 0.8579 0.0028 0.0150 0.0043 0.0234 0.0966
35 0.5088 0.0091 0.0864 0.1408 0.0183 0.2366
36 0.6570 0.0013 0.0777 0.0187 0.0110 0.2343
37 0.5276 0.0125 0.0833 0.0026 0.0175 0.3566
38 0.4653 0.0054 0.0214 0.0941 0.0050 0.4090
39 0.7510 0.0029 0.0116 0.0128 0.0095 0.2122
40 0.6155 0.0034 0.0361 0.0066 0.0294 0.3091
41 0.9248 0.0001 0.0059 0.0009 0.0320 0.0363
Data: FAF database 2007
Note: LA: Los Angeles MSA, SA: Sacramento MSA, SD: San Diego MSA, SF: San Francisco MSA, RE: Remainder
of MSA, OS: Other States
119
Appendix Table 6: Los Angeles MSA import trade proportions for truck mode
Percentage of domestic imports by origin
SCTG LA SA SD SF RE OS Total
1 0.5684 0.0000 0.0623 0.0000 0.0002 0.3660 0.9969
2 0.5826 0.0313 0.1279 0.0000 0.0120 0.0881 0.8419
3 0.7671 0.0037 0.0707 0.0134 0.0566 0.0683 0.9799
4 0.6927 0.0000 0.0489 0.0115 0.0247 0.1622 0.9401
5 0.6934 0.0003 0.0389 0.0145 0.0602 0.1710 0.9782
6 0.7174 0.0151 0.0234 0.0066 0.0470 0.1671 0.9766
7 0.6446 0.0183 0.0142 0.0582 0.0814 0.1484 0.9651
8 0.5799 0.0000 0.0037 0.0320 0.0650 0.0280 0.7086
9 0.8957 0.0000 0.0125 0.0000 0.0268 0.0573 0.9924
10 0.7875 0.0000 0.0126 0.0870 0.0530 0.0574 0.9976
11 0.9023 0.0000 0.0048 0.0158 0.0065 0.0364 0.9657
12 0.9356 0.0028 0.0069 0.0000 0.0398 0.0120 0.9971
13 0.6894 0.0015 0.0175 0.0117 0.0245 0.1941 0.9386
14 0.0518 0.0000 0.0007 0.0000 0.0006 0.9319 0.9849
15 0.1088 0.0000 0.0019 0.0000 0.0017 0.0002 0.1126
16 0.0020 0.0133 0.1117 0.1260 0.4118 0.0002 0.6650
17 0.5496 0.0002 0.0012 0.0136 0.0074 0.0027 0.5747
18 0.3231 0.0000 0.0072 0.0116 0.0215 0.0006 0.3640
19 0.5162 0.0000 0.0016 0.0061 0.0120 0.0281 0.5640
20 0.4090 0.0003 0.0109 0.0135 0.0067 0.3920 0.8324
21 0.5029 0.0029 0.0025 0.0243 0.0118 0.1942 0.7388
22 0.9031 0.0011 0.0063 0.0012 0.0535 0.0252 0.9903
23 0.6199 0.0009 0.0086 0.0151 0.0105 0.2719 0.9268
24 0.6773 0.0032 0.0204 0.0173 0.0351 0.1700 0.9233
25 0.8986 0.0000 0.0117 0.0000 0.0103 0.0421 0.9627
26 0.7228 0.0238 0.0158 0.0099 0.0584 0.1312 0.9618
27 0.6178 0.0000 0.0003 0.0170 0.0124 0.2309 0.8784
28 0.7085 0.0007 0.0103 0.0147 0.0327 0.1828 0.9496
29 0.5738 0.0053 0.0173 0.0164 0.0146 0.2368 0.8641
30 0.5792 0.0005 0.0323 0.0225 0.0335 0.1994 0.8675
31 0.7862 0.0016 0.0349 0.0132 0.0606 0.0673 0.9637
32 0.6306 0.0001 0.0110 0.0338 0.0264 0.1932 0.8951
33 0.7090 0.0020 0.0265 0.0162 0.0341 0.1118 0.8995
34 0.8380 0.0021 0.0147 0.0038 0.0226 0.0800 0.9612
35 0.4260 0.0051 0.0728 0.0885 0.0152 0.1609 0.7686
36 0.6193 0.0012 0.0744 0.0176 0.0107 0.1732 0.8964
37 0.3912 0.0084 0.0693 0.0013 0.0165 0.2213 0.7079
38 0.3924 0.0027 0.0141 0.0697 0.0047 0.2574 0.7408
39 0.7382 0.0029 0.0114 0.0126 0.0094 0.1954 0.9698
40 0.5387 0.0022 0.0261 0.0056 0.0284 0.2185 0.8196
41 0.9247 0.0001 0.0059 0.0009 0.0310 0.0296 0.9923
Avg. 0.6150 0.0037 0.0260 0.0201 0.0364 0.1538 0.8550
Data: FAF database 2007
Note: LA: Los Angeles MSA, SA: Sacramento MSA, SD: San Diego MSA, SF: San Francisco MSA, RE: Remainder of
MSA, OS: Other States
120
Appendix Table 7: 2008 California air cargo statistics
Unit: U.S. ton
Airport name 2008 total
FAF region ZIP county
Arcata 664.9
69 95519 Humboldt
Bob Hope 42,908.90
61 91505 Los Angeles
Fresno-Yosemite Int'l 9,741.10
69 93727 Fresno
John Wayne 16,829.80
61 92707 Orange
LA Ontario Int'l 481,283.00
61 91761 Los Angeles
Long Beach 44,352.60
61 90808 Los Angeles
Los Angeles Int'l 1,797,780.00
61 90045 Los Angeles
March ARB (Air reserve base) 26,044.20
61 92518 Riverside
Merced Municipal 71.7
69 95341 Merced
Metro Oakland Int'l 679,117.50
61 94621 Alameda
Modesto 312.1
69 95354 Stanislaus
Monterey 618
69 93940 Monterey
Murray Field 6,331.90
69 95501 Humboldt
Palm Springs Int'l 26
61 92262 Riverside
Redding Muni 1,675.90
69 96002 Shasta
Sacramento Int'l 79,319.30
62 95837 Sacramento
Sacramento Mather 77,100.10
62 95655 Sacramento
San Diego Int'l 133,913.10
63 92101 San Diego
San Francisco Int'l 543,197.60
64 94128 San Mateo
San Jose Int'l 81,222.20
64 95110 Santa Clara
San Luis Obispo 1,332.90
69 93401 San Luis
Obispo
Santa Barbara Muni 2,797.00
69 93117 Santa
Barbara
Sonoma County 672.8
64 95403 Sonoma
Source: California Department of Transportation, 2008 California Air Cargo Statistics
(http://www.dot.ca.gov/hq/planning/aeronaut/documents/2008Cargo2009Apr.pdf)
Note: 2008 total includes imports and exports. I selected airports that have more than 100,000 ton at 2008.
Selected 5 airports handle more than 90 % of total air cargo.
121
Appendix Table 8: California sea ports unit: U.S. ton
Seaport name 2008 Import 2008 Export
FAF
region
ZIP county
Main Cargo
Types of imports
Main Cargo Types
of exports
Benicia 64 94510 Solano
Hueneme 1,216,595 62,424 61 93044 Ventura
Autos, Produce
Liquid Fertilizer
Nuts, Bulk Liquid
Autos
Produce
General Cargo
Humboldt Bay 69 95502 Humboldt
Logs
Petroleum
Logs, Wood chips
Lumber
Long Beach 45,186,084 22,084,935 61 90802 Los Angeles
Crude oil
Electronics
Plastics
Furniture
Clothing
Petroleum coke
Petroleum bulk
Chemicals
Waste Paper
Food
Los Angeles 32,732,756 20,180,533 61 90731 Los Angeles
Furniture
Automobile parts
Apparel
Electronic
Products
Footwear
Wastepaper
Scrap Metal
Animal Feeds
Cotton
Resins
Oakland 6,497,039 8,631,041 64 94607 Alameda
Furniture
Plastic ware, tiles
Computers
Machinery/parts
Machinery
Fruit, Nuts
Beverages
Meats
Machinery
Lumber
Redwood City 1,310,112 299,832 64 94063 San Mateo
Cement
Gypsum
Bauxite
Sand, Building
Aggregates
Scrap Metal
Rock
Non-ferrous
metals
Richmond 13,044,242 2,898,576 64 94804 Contra Costa
break-bulk,
bulk,
project cargo
chemicals,
pharmaceuticals,
forest products,
machinery, frozen
seafood,
produce, bottled
water from
Iceland,
recreational
campers, steel,
steel products,
stone, tobacco
leaf, aluminum,
project
cargo, vehicles,
recreational boats,
wire coils, wire
rods, pipe, bulk
grain, minerals,
and livestock
West Sacramento 476,983 347,710 62 95691 Sacramento
San Diego 1,463,243 16,343 63 92101 San Diego
San Francisco 803,968 55,539 64 94111
San
Francisco
Steel Products
Boats / Yachts
Wind Turbines
Project Cargo
Aggregate
Sand
Tallow
Vegetable Oil
Stockton 1,218,654 513,469 69 95203 San Joaquin
Cement
Molasses
Steel Products
Palm Oil
Sulphur
Bulk Rice
Bagged Rice
Machinery
122
Machinery
Boric Acid
Lumber
Fertilizer
Windmills
Anhydrous
Ammonia
Wheat
Steel Scrap
Petroleum Coke
Safflower Seed
Iron Ore / Cole
Source: American Association of Port Authorities
(http://aapa.files.cms-
plus.com/Statistics/2008%20U.S.%20PORT%20RANKINGS%20BY%20CARGO%20TONNAGE.pdf)
Main Cargo Types : California DOT
(http://www.dot.ca.gov/hq/tpp/offices/ogm/fact_sheets_index.html)
Main cargo types that are not available at California DOT are obtained from port website.
Port of LA: http://www.portoflosangeles.org/about/facts.asp
Port of Richmond: http://www.richmondgov.com/PortOfRichmond/index.aspx
Note: No data is available for Port of Benicia. I selected ports that have more than 10,000,000 tons of
trades at 2008. Selected four ports handle over 95% of total cargo.
123
Appendix Table 9: Foreign Import Mode proportion
SCTG
Air(including air-
truck)
Water-
Truck
Water-
multiple
Truck-
Truck
Multiple-
truck Others
1 55.78% 8.42% 0.04% 35.69% 0.01% 0.06%
2 0.48% 17.99% 3.27% 11.85% 0.00% 66.41%
3 4.93% 45.79% 9.31% 38.21% 0.03% 1.73%
4 19.48% 64.17% 13.17% 1.08% 0.03% 2.08%
5 0.31% 77.19% 10.47% 3.20% 0.00% 8.83%
6 0.26% 77.11% 8.36% 12.22% 0.00% 2.04%
7 1.20% 74.06% 12.26% 9.46% 0.01% 3.01%
8 0.55% 63.62% 12.10% 16.18% 0.03% 7.51%
9 11.16% 74.12% 11.34% 2.29% 0.12% 0.98%
10 0.00% 90.33% 8.33% 0.00% 0.00% 1.34%
11 0.00% 24.97% 74.53% 0.00% 0.00% 0.50%
12 0.00% 78.13% 10.58% 0.00% 0.00% 11.28%
13 0.78% 50.82% 36.31% 10.62% 0.02% 1.45%
14 16.53% 60.16% 18.68% 0.00% 0.00% 4.64%
15 0.00% 55.44% 41.07% 0.00% 0.00% 3.49%
16 0.00% 0.00% 0.00% 0.00% 0.00% 100.00
17 0.00% 47.68% 0.04% 0.00% 0.00% 52.28%
18 0.00% 31.34% 8.47% 0.00% 0.00% 60.19%
19 0.58% 89.90% 2.49% 0.00% 0.00% 7.03%
20 27.33% 42.94% 16.34% 0.07% 0.00% 13.32%
21 27.00% 22.78% 5.37% 43.26% 0.06% 1.53%
22 0.05% 42.96% 29.95% 0.05% 0.00% 27.00%
23 29.81% 39.27% 15.92% 6.88% 0.43% 7.68%
24 1.64% 64.00% 26.85% 6.19% 0.35% 0.97%
25 0.00% 73.07% 22.03% 0.00% 0.00% 4.90%
26 0.93% 68.86% 24.34% 4.86% 0.20% 0.82%
27 0.00% 80.37% 16.99% 0.54% 0.00% 2.09%
28 2.03% 42.73% 24.22% 29.90% 0.26% 0.87%
29 3.22% 60.62% 30.59% 3.25% 0.14% 2.19%
30 4.37% 65.17% 23.81% 3.74% 0.03% 2.89%
31 3.09% 64.91% 22.49% 6.71% 0.44% 2.35%
32 2.89% 65.78% 12.45% 6.37% 0.38% 12.13%
33 1.69% 54.01% 29.36% 9.11% 0.24% 5.59%
34 33.00% 38.37% 20.54% 5.57% 0.28% 2.25%
35 18.08% 33.94% 17.73% 22.94% 1.71% 5.61%
36 0.57% 76.04% 17.82% 3.33% 0.01% 2.24%
37 11.08% 51.08% 30.38% 6.86% 0.03% 0.57%
38 28.58% 32.09% 17.35% 20.96% 0.78% 0.24%
39 1.07% 64.84% 24.39% 8.08% 0.03% 1.59%
40 16.04% 49.73% 26.27% 3.05% 0.18% 4.73%
41 0.00% 75.47% 12.60% 0.00% 0.00% 11.94%
43 62.93% 0.47% 0.00% 35.72% 0.76% 0.12%
Source: FAF 2007 data
Note: Rail-truck mode proportion is zero.
124
Appendix Table 10: Foreign Export Mode proportion
SCTG
Air(including truck-
air)
Truck-
Water
multiple-
Water
Truck-
Truck
Truck-
Multiple Others
1 74.36% 0.82% 0.45% 23.77% 0.13% 0.48%
2 0.03% 8.43% 42.11% 0.94% 0.00% 48.48%
3 3.29% 60.38% 26.95% 6.77% 0.09% 2.52%
4 0.35% 50.80% 43.11% 2.04% 0.00% 3.70%
5 0.32% 70.43% 10.86% 11.74% 0.01% 6.64%
6 0.97% 49.57% 19.69% 23.58% 0.00% 6.20%
7 4.64% 49.43% 11.58% 29.23% 0.01% 5.11%
8 2.87% 62.26% 20.35% 10.39% 0.00% 4.13%
9 0.05% 76.79% 15.88% 0.03% 0.00% 7.25%
10 0.00% 78.08% 13.05% 0.00% 0.00% 8.87%
11 0.00% 16.73% 22.58% 0.00% 0.00% 60.69%
12 0.00% 60.05% 33.89% 0.00% 0.00% 6.06%
13 0.52% 74.19% 14.72% 5.99% 0.00% 4.59%
14 0.39% 14.64% 81.21% 0.10% 0.00% 3.67%
15 0.00% 68.78% 17.16% 0.00% 0.00% 14.06%
16 0.00% 62.97% 16.60% 0.00% 0.00% 20.42%
17 0.00% 47.31% 0.04% 0.00% 0.00% 52.65%
18 0.00% 21.48% 23.96% 0.00% 0.00% 54.56%
19 0.11% 56.54% 3.66% 8.62% 0.02% 31.04%
20 5.06% 49.12% 18.11% 1.48% 0.00% 26.23%
21 75.06% 9.16% 14.06% 1.10% 0.00% 0.62%
22 0.00% 23.56% 36.61% 32.97% 0.00% 6.86%
23 34.43% 28.73% 22.13% 12.04% 0.00% 2.68%
24 5.38% 25.01% 34.61% 29.12% 0.00% 5.88%
25 0.00% 48.54% 40.19% 0.00% 0.00% 11.27%
26 1.17% 16.86% 11.77% 63.41% 0.02% 6.78%
27 0.00% 45.32% 40.21% 11.23% 0.00% 3.24%
28 4.15% 6.44% 4.02% 82.07% 0.05% 3.26%
29 32.50% 28.02% 8.75% 29.75% 0.00% 0.97%
30 16.85% 20.54% 19.73% 41.71% 0.03% 1.14%
31 13.46% 37.06% 21.56% 25.77% 0.01% 2.14%
32 11.08% 25.69% 11.18% 48.27% 0.02% 3.75%
33 16.52% 21.78% 7.78% 52.16% 0.02% 1.73%
34 49.10% 23.11% 10.02% 16.29% 0.00% 1.48%
35 70.82% 5.20% 2.96% 20.23% 0.01% 0.79%
36 5.10% 38.05% 18.30% 29.39% 0.00% 9.16%
37 77.11% 9.18% 3.25% 0.83% 0.00% 9.63%
38 75.21% 9.75% 5.83% 8.70% 0.00% 0.50%
39 12.09% 32.97% 19.36% 33.24% 0.01% 2.33%
40 53.01% 27.20% 8.48% 10.26% 0.00% 1.05%
41 0.00% 45.57% 46.06% 0.00% 0.00% 8.37%
43 3.07% 0.00% 0.00% 21.34% 0.00% 75.59%
Source: FAF 2007 data
Note: Truck-rail mode proportion is zero.
125
Appendix Table 11: Domestic modes definition
Mode ID Mode Description Remarks
1 Truck
Includes private and for-hire truck. Private trucks are owned or
operated by shippers, and exclude personal use vehicles hauling
over-the-counter purchases from retail establishments.
2 Rail Any common carrier or private railroad.
3 Water Includes shallow draft, deep draft and Great Lakes shipments.
4 Air (include truck-air)
Includes shipments typically weighing more than 100 pounds that
move by air or a combination of truck and air in commercial or
private aircraft. Includes air freight and air express. Shipments
typically weighing 100 pounds or less are classified with Multiple
Modes and Mail.
5
Multiple modes &
mail
Includes shipments by multiple modes, parcel delivery services,
U.S. Postal Service, and couriers. This category is not limited to
containerized or trailer-on-flatcar shipments.
6 Pipeline Includes shipments by pipeline and from offshore wells to land.
7 Other and unknown
Any mode not included within the other mode definitions and
unknown modes of transport.
8 No domestic mode Applies to some intra zonal movements of imports
Source: FAF database 2007
Appendix Table 12: Foreign modes definition
Mode ID Mode Description Remarks
1 Truck
Includes U.S. trade with Canada or Mexico that crosses the border
on a private or for-hire truck.
2 Rail
Includes U.S. trade with Canada or Mexico that crosses the border
on any common carrier or private railroad.
3 Water
Includes U.S. imports and exports that enter or exit the United
States through a seaport.
4 Air
Includes U.S. imports and exports that enter or exit the United
States through an airport.
5
Multiple modes &
mail
Includes U.S. imports and exports that enter or exit the United
States by multiple modes of transport, parcel delivery services,
U.S. Postal Service, couriers, and U.S. imports and exports
transhipped thru Canada or Mexico by a land mode (e.g. truck,
rail, etc.) from/to a third country. This category is not limited to
containerized or trailer-on-flatcar shipments.
6 Pipeline
Includes U.S. imports and exports that cross the U.S.-Canada or
U.S.-Mexico border by pipeline.
7 Other and unknown
Any mode not included within the other mode definitions and
unknown modes of transport. Includes flyaway aircraft, vessels
and vehicles moving under their own power from the
manufacturer to a customer and not carrying any freight, and
imports into Foreign Trade Zones (FTZs).
Source: FAF database 2007
126
Appendix Figure 2: Sea ports in California
Source: California Department of Transportation http://www.dot.ca.gov/hq/tpp/offices/ogm/seaports.html)
127
Appendix Table 13: Reference Case Greenhouse Gas Emissions
GHG Emissions (Mt) 2006 2010 2015 2020
Avg. Annual
Growth Rate
2006-2020
Residential 27.3 27.0 27.9 29.7 0.6%
Commercial 14.0 12.4 12.1 12.1 -1.0%
Industrial 80.0 86.2 92.8 102.8 1.8%
Energy Intensive Industry 52.5 47.8 48.6 49.2 -0.5%
Other Industry 27.5 38.4 44.2 53.6 4.9%
Mining 13.2 13.0 13.0 12.2 -0.6%
Agriculture 27.4 29.1 29.8 31.0 0.9%
Transportation 213.3 211.5 222.7 227.8 0.5%
Passenger 167.6 162.0 168.5 168.8 0.1%
Freight 45.7 49.5 54.2 58.9 1.8%
Power Sector 102.0 89.1 93.1 100.0 -0.1%
Domestic Power Sector 43.2 40.0 37.7 39.1 -0.7%
Electricity Imports 58.8 49.1 55.3 60.8 0.2%
Waste and Other 9.8 10.9 11.5 12.4 1.7%
Total 486.9 479.3 502.8 527.9 0.6%
Source: California Air Protection Agency | Air Resources Board
Updated Economic Analysis of California’s Climate Change Scoping Plan | March 24, 2010
128
Appendix B: EMFAC 2007 and MOVES 2010a
EMFAC2007
The California Air Resources Board (ARB) has developed EMission FACtors (EMFAC) models.
The latest model is EMFAC2007. It includes all motor vehicle data from motorcycles to heavy
duty trucks. Emission rates are estimated for vehicles operated on highways, freeways, and local
roads in California. Emission rates are calculated via the following equation:
E
ij
c
= EF
ij
c
× CF
ij
× TA
ij
c
Where
E
ij
c
are emissions in tons per day by region i, calendar year j and vehicle class c
EF
ij
c
are emissions factors (in grams per mile, grams per trip, and grams per vehicle)
CF
ij
are correction factors
TA
ij
c
are vehicle activities
Correction factors reflect area-specific information affecting emission rates such as ambient
temperature, relative humidity, and speed.
Vehicle activity refers to vehicle population, vehicle miles traveled (VMT) on a weekday, and
vehicle trips for each vehicle class, fuel type and geographic area. Geographic areas can be one of
four types: statewide, 15 air basins, 35 air pollution control districts, or 58 counties. EMFAC
contains vehicle population data by vehicle classes, fuel types, regions, and vehicle age from 1 to
45 years. Vehicle populations are estimated by utilizing DMV vehicle registration data from base
years 2000 to 2005. Data for 1970 to 1999 and 2001 to 2040 are estimated by back-casting and
forecasting of the base year data. VMT is calculated by multiplying vehicle population to the
vehicle accrual or total miles a vehicle traveled a year. VMT varies by vehicle age, class, and time
of the day. Vehicle trips per day are the number of starts made per weekday. For vehicle classes 1
to 4, trips are estimated based on travel survey data and assumed to linearly decrease from 6.56
when vehicle age is 1 to 3.72 when vehicle age is 45 years. Appendix Figure 1 show the linearly
decreasing graph. Trips for other classes are obtained either from engineering judgment or
instrumented data. The number of trips per day is used to estimate starting exhaust.
129
Appendix Figure 3: Number of trips per weekday for vehicle class 1 to 4
Source: adapted from CARB, 2007
Air pollutants
The model estimates emission inventories for four pollutants, Hydrocarbons (HC), Carbon
monoxide (CO), Carbon dioxide (CO2), Nitrogen oxides (NOx), and Particulate matter (PM10,
PM2.5). Fuel consumption is calculated by applying a carbon balance equation showing the
relationship between fuel consumption and emission inventories such as CO, CO2 and HC.
Emission inventories of Oxides of sulfer (SOx) are calculated by multiplying fuel consumption
with the percentage of SOx in a gallon of fuel.
3.00
3.50
4.00
4.50
5.00
5.50
6.00
6.50
7.00
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45
Trips
Vehicle age
Number of trips
Number of trips
130
Appendix Table 14: Air pollutants estimated in EMFAC 2007
Pollutant Full name Description Unit
Direct estimation
HC Hydrocarbons
HC is equivalent to TOG(total
organic gases), ROG (reactive
organic gases), THC (total
hydrocarbon), or CH4 (methane)
grams Emission
rates are
estimated
directly
from
vehicle
activities
CO Carbon monoxide vehicle activities grams
CO2 Carbon dioxide grams
NOx Nitrogen oxides grams
PM Particulate matter
Particulate matter 10 microns or
less in diameter (PM10),
Particulate matter 2.5 microns or
less in diameter (PM2.5)
grams
Indirect estimation
Sox Oxides of sulfur grams Emission
rates are
estimated
indirectly
by applying
fuel
consumptio
ns
Pb Lead
Estimated until year 1991 and
Zero from year 1992.
grams
Source: Summarized from CARB, 2007
131
Emission processes
Nine emission processes are considered in the EMFAC model as shown in Appendix Table 16.
Appendix Table 15: Emission processes in EMFAC 2007
Process Emitted areas Activities of emission Applied vehicle Pollutants
Exhaust
Running exhaust Tailpipe While traveling on the
road
All vehicles CO, NOx,
CO2, SOx
Idle exhaust tailpipe While operating for
loading and unloading
goods
Heavy-duty
trucks
CO, NOx,
CO2, SOx
Starting exhaust tailpipe While starting a
vehicle
Only for
gasoline fueled
vehicles
CO, NOx,
CO2, SOx
Evaporative
Diurnal Fuel system, fuel
hoses, connectors,
carbon canister
From 35 minutes of
sitting after finishing
operation and ambient
temperature is
increasing.
All vehicles HC
Resting loss Fuel system, fuel
hoses, connectors,
carbon canister
From 35 minutes of
sitting after finishing
operation and ambient
temperature is not
increasing.
All vehicles HC
Hot soak Fuel injector, Fuel
hoses
Immediately after a
trip end until 35
minutes
All vehicles HC
Running losses Fuel system, carbon
canister
While operating All vehicles HC
Wear
Tire wear Tires While moving All vehicles PM
Break wear Brake While using brakes All vehicles PM
Source: Summarized from CARB, 2007
Vehicle class and technology group
Emission rates are estimated separately for 13 vehicle classes in the model. Vehicle classes are car
types such as passenger cars, trucks, motorcycles, buses, and motor homes. Truck class is broken
down into 7 sub classes by vehicle weights. Vehicle classes are broken down further into
technology groups. The basic assumption of a technology group is that vehicles of each
technology group have the same emission rates due to installed emission control devices in
vehicles. A technology group can include more than one vehicle class. There are two types of
technology groups: exhaust and evaporative technology groups. Exhaust technology groups are
related to emissions such as CO, NOx, CO2 and SOx that come out of the tailpipe while
operating. Evaporative technology groups are related to HC emissions that are evaporated from
fuel systems.
132
Three modeling modes in EMFAC 2007
EMFAC 2007 supports three modeling modes such as Burden, Emfac and Calimfac.
Appendix Table 16: Three modeling modes in EMFAC 2007
Burden Emfac Calimfac
Result Total emissions in tons
per weekday. Vehicle
population,
VMT(mi/day), and trips
(per day)
Emission factors in
grams per vehicle
activity (grams per mile,
grams per hour, grams
per start and depends on
emissions process)
Basic emission rates
(g/mi)
Common classification For each pollutant by 13
vehicle classes,
geographic area, season,
calendar year, emission
processes, vehicle model
year
For each pollutant by 13
vehicle class,
geographic area, season,
calendar year, emission
processes, vehicle
model year
For each pollutant by
13 vehicle classes,
geographic area,
season, calendar year,
emission processes,
vehicle model year
Specific By temperature, relative
humidities, speed,
By technology group
and vehicle age,
with/without I/M
program
Source: Summarized from CARB, 2007
MOVES 2010a
The U.S. Environmental Protection Agency (EPA) has developed a comprehensive air pollution
emissions estimation model, Motor Vehicle Emission Simulator (MOVES). The latest version is
MOVES2010a. The basic concept of emission estimation processes is similar to the one used in
EMFAC2007. The emission calculation process is similar to EMFAC. However, primary activity
data that is used to estimate emission inventory is significantly different. Initial data to estimate
running exhaust emissions is VMT which is the same as for EMFAC. The VMT, however, are
converted into Source Hours and Source Hours Operating (SHO).
MOVES consists of five major frameworks: activity generator, source bin distribution generator,
operating mode distribution generator, energy consumption calculator, and emission calculator.
VMT is converted to source hours and source hours operating (SHO). Each activity basis for
emission processes is explained in Table 4. Source bin refers to vehicle classes that are similar
to technology group in EMFAC. Table 5 explains source bin. An operating mode is a
combination of Vehicle Specific Power (VSP) and speed. Table 6 shows operating mode bins.
133
Whole processes of emission rate calculation can be simplified (Bai, 2009). Base emission rates
are first adjusted by area specific data such as Inspection and Maintenance (I/M) program,
temperature, and relative humidity. Then the adjusted emission rates are weighted by source bin
and operating mode bin fractions. Finally total emission inventories are estimated by multiplying
total activity with the weighted emission rates.
Appendix Table 17: Total Activity Basis by Process
Emission
Process
Total
Activity
Basis
Description
Running
Tire wear
Brake
wear
Source
Hours
Operating
(SHO)
Total hours, of all sources within a source type, spent operating on the
roadway network for the given time and location of the run spec. The same
as number of sources * per-source hours operating
Evaporati
ve Fuel
Permeatio
n,
Vapor
Venting
and
Leaking
Source
Hours
Total hours, of all sources within a source type for the given time and
location of the run spec. This is equivalent to the population of the source
type times the number of hours in the time period.
Start Number of
Starts
Total starts, of all sources within a source type, for the given time and
location of the run spec. The same as number of sources * per-source starts
Extended
Idle
Extended
Idle Hours
Total hours, of all sources within a source type, spent in extended idle
operation for the given time and location of the run spec.
Source: EPA, 2009: page 39
134
Appendix Table 18: MOVES Source Bin Definitions (other than Model Year Group)
Fuel Type
(All
Pollutants)
Engine Technology
(All Pollutants)
Loaded Weight
(Energy)
Engine Size
(Energy)
Regulatory Class (All
pollutants except energy
and evap permeation)
Gas
Diesel CNG
LPG
Ethanol
(E85)
Methanol
(E85) Gas
H2
Liquid H2
Electric
Conventional IC
(CIC)
Advanced IC (AIC)
Hybrid - CIC
Moderate
Hybrid - CIC Full
Hybrid - AIC
Moderate Hybrid -
AIC Full Fuel Cell
Hybrid - Fuel Cell
Electric
Null
< 500 (for
motorcycles) 500-
700 (for
motorcycles)
> 700 (for
motorcycles)
<= 2000 lbs 2001-
2500 2501-3000
3001-3500
3501-4000
4001-4500 4501-
5000
5001-6000
6001-7000
7001-8000
8001-9000
9001-10,000
10,001-14,000
14,001-16,000
16,001-19,500
19,501-26,000
26,001-33,000
33,001-40,000
40,001-50,000
50,001-60,000
60,001-80,000
80,001-100,000
100,001-130,000
>=130,001
Null
< 2.0 liters
2.1-2.5 liters
2.6-3.0 liters
3.1-3.5 liters
3.6-4.0 liters
4.1-5.0 liters
> 5.0 liters
Null
Motorcycle LDV
LDT
HD gasoline GVWR <=
14K lbs HD gasoline
GVWR > 14K llbs.
LHDD
MHDD
HHDD
Urban Bus
Source: EPA, 2009: page 34
135
Appendix Table 19: MOVES Source Bin Definitions (Model Year Group)
Model Year Group
Energy CH4, N2O HC - Evap HC, CO,
NOx, PM
start, running
HC, CO,
NOx, PM
extended idle
Sulfate PM
(ratios to
energy)
1980 and
earlier
1981-85
1986-90
1991-2000
2001-2010
2011-2020
2021 and later
1972 and
earlier
1973
1974
1975
.
.
.
1999
2000
2001-2010
2011-2020
2021 and later
1970 and
earlier
1971-1977
1978-1995
1996-2003
2004
2005
.
.
2019
2020
2021 and later
1980 and
earlier
1981-1982
1983-1984
1985
1986-1987
1988-1989
1990
1991-1993
1994
1995
.
.
2019
2020
2021 and later
1980 and
earlier
1981-85
1986-90
1991-2000
2001-2006
2007-2010
2011-2020
2021 and later
1980 and
earlier
1981 and later
Source: EPA, 2009: page 34
Appendix Table 20: Operating Mode Bin Definitions
Braking Bin 0
Idle Bin 1
VSP\Instantaneous Speed 0-25 mph 25-50 <50
<0 kW/ton Bin 11 Bin 21
0 to 3 Bin 12 Bin 22
3 to 6 Bin 13 Bin 23
6 to 9 Bin 14 Bin 24
9 to 12 Bin 15 Bin 25
12 and greater Bin 16 Bin 26 Bin 36
6 to 12 Bin 35
<6 Bin 33
12 to 18 Bin 27 Bin 37
18 to 24 Bin 28 Bin 38
24 to 30 Bin 29 Bin 39
30 and greater Bin 30 Bin 40
Source: EPA, 2009: page 40
136
Comparison of EMFAC and MOVES model
Table 7 shows a comparison of EMFAC2007 and MOVES2010a
Appendix Table 21: Comparison of EMFAC2007 and MOVES2010a
EMFAC2007 MOVES2010a
Geographic
area
California state,
15 air basins,
35 air pollution control districts, or
58 counties
U.S. as a nation,
53 States (District of Columbia, Puerto Rico,
U.S. Virgin Islands are considered to be
states),
3222 counties,
5 Links in each county
Pollutants Hydrocarbons (TOG, ROG, THC, or
CH
4
)
Carbon monoxide (CO)
Carbon dioxide (CO
2
)
Oxides of Nitrogen (NOx)
Particulate matter (PM10, PM2.5)
Oxides of sulfur (SOx)
Lead (Pb)
Fuel consumption
Hydrocarbons (TOG, VOC, THC, or CH
4
)
Carbon monoxide (CO)
Carbon Dioxide (CO
2
: depends on total
energy con.)
CO2 equivalent (CO
2
e)
Oxides of Nitrogen (NOx, NO, NO
2
)
Nitrous Oxide (N
2
O)
Particulate matter (PM10, PM2.5)
Sulfur Dioxide (SO
2
)
Total Energy Consumption (Petroleum and
Fossil Fuel)
Ammonia (NH
3
)
Naphthalene (C
10
H
8
-depends on PM10)
Below emissions depends on VOC
Benzene (C
6
H
6
)
Ethanol (C
2
H
6
O)
methyl tertiary butyl ether (MBTE)( C
5
H
12
O)
1,3-Butadiene(C
4
H
6
)
Formaldehyde(CH
2
O)
Acetaldehyde(C
2
H
4
O)
Acrolein(C
3
H
4
O)
Vehicle class PassengerCars
Light-DutyTrucks(0-3750)
Light-DutyTrucks(3751-5750)
Medium-DutyTrucks(5751-8500)
Light-Heavy-Duty(8501-10000)
Light-Heavy-Duty(10001-14000)
Medium-Heavy-Duty(14001-33000)
Heavy-Heavy-Duty(33001-60000)
Other Buses
Urban Buses
Motorcycles
School Buses
Motor Homes
Passenger Cars
Passenger Trucks
Light Commercial Trucks
Refuse Trucks
Single Unit Short-haul Trucks
Single Unit Long-haul Trucks
Combination Short-haul Trucks
Combination Long-haul Trucks
Intercity Buses
Transit Buses
Motorcycles
School Buses
Motor Homes
Fuel type Gasoline
Diesel
Gasoline
Diesel
137
Electricity Electricity
Compressed Natural Gas (CNG)
Liquid Propane Gas (LPG)
Ethanol (E85)
Methanol (M85)
Gaseous Hydrogen
Liquid Hydrogen
Emission
process
Running Exhaust
Starting Exhaust
Idle Exhaust
Diurnal
Hot soak
Resting loss
Running losses
Tire Wear
Brake Wear
Running Exhaust
Starting Exhaust
Extended Idle
Evaporative Fuel Permeation
Evaporative Fuel Vapor Venting
Evaporative Fuel Leaking
Refueling Spillage Loss
Refueling Displacement Vapor Loss
Tire Wear
Brake Wear
Time period Calendar years 1970-2040.
Output by hour of weekdays, month,
season (summer, winter), and year
Calendar years 1990 and 1999 through 2050.
Output by hour of the day, weekday,
weekends, month, and year
Vehicle model
year
1965 – 2040 1960-2050
Activity data
for running
exhaust
Vehicle Miles Traveled (VMT) Source Hours Operating (SHO): operating
time by combination of Vehicle Specific
Power (VSP) and speed
Road Type Not available Rural Restricted Access (i.e. freeways and
interstates)
Rural Unrestricted Access
Urban Restricted Access (i.e. freeways and
interstates)
Urban Unrestricted Access
Off of the highway network (for start, idle,
evap.)
Source: Summarized from CARB, 2007 and EPA 2009, 2010a
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Environmental effects from a large-scale adoption of electric vehicle technology in the City of Los Angeles
PDF
Evaluating city logistics using two-level location routing modeling and SCPM simulation
PDF
Developing an agent-based simulation model to evaluate competition in private health care markets with an assessment of accountable care organizations
PDF
Investigation of health system performance: effects of integrated triple element method of high reliability, patient safety, and care coordination
PDF
Chemical and toxicological characteristics of particulate matter in urban environments with a focus on its sources, associated health impacts and mitigation policies
PDF
Integration of truck scheduling and routing with parking availability
PDF
A framework for comprehensive assessment of resilience and other dimensions of asset management in metropolis-scale transport systems
PDF
Physico-chemical characteristics and sources of ambient PM mass and number concentrations and their associated toxicity, and development of novel techniques for high time-resolution measurement o...
Asset Metadata
Creator
Cho, Joongkoo
(author)
Core Title
Network-based simulation of air pollution emissions associated with truck operations
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Industrial and Systems Engineering
Publication Date
04/01/2013
Defense Date
01/24/2013
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
estimation of truck origin-destination matrix,OAI-PMH Harvest,scenario test for emission reduction policies,simulation of air pollution emissions
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Moore, James Elliott, II (
committee chair
), Gordon, Peter (
committee member
), Rahimi, Mansour (
committee member
)
Creator Email
joongkoo.cho@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-229637
Unique identifier
UC11293929
Identifier
usctheses-c3-229637 (legacy record id)
Legacy Identifier
etd-ChoJoongko-1499.pdf
Dmrecord
229637
Document Type
Dissertation
Rights
Cho, Joongkoo
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
estimation of truck origin-destination matrix
scenario test for emission reduction policies
simulation of air pollution emissions