Full text of DOI:10.1080/10473289.2001.10464248

This article was downloaded by: [Van Pelt and Opie Library]
On: 17 October 2014, At: 04:35
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK
Journal of the Air & Waste Management Association
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/uawm20
Observations and Model Simulations of Carbon
Monoxide Emission Factors from a California Highway
Anthony E. Held a , Daniel P.Y. Chang a & Debbie A. Niemeier a
a Department of Civil and Environmental Engineering , University of California , Davis,
California , USA
Published online: 27 Dec 2011.
To cite this article: Anthony E. Held , Daniel P.Y. Chang & Debbie A. Niemeier (2001) Observations and Model Simulations
of Carbon Monoxide Emission Factors from a California Highway, Journal of the Air & Waste Management Association, 51:1,
121-132, DOI: 10.1080/10473289.2001.10464248
To link to this article: http://dx.doi.org/10.1080/10473289.2001.10464248
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions

TECHNICAL PAPER
Held, Chang, and Niemeier
ISSN 1047-3289 J. Air & Waste Manage. Assoc. 51:121-132
Copyright 2001 Air & Waste Management Association
Observations and Model Simulations of Carbon Monoxide
Emission Factors from a California Highway
Anthony E. Held, Daniel P.Y. Chang, and Debbie A. Niemeier
Department of Civil and Environmental Engineering, University of California, Davis
ABSTRACT
one to estimate fleet averaged emissions factor (EF), such
as the California Department of Transportation’s
(CALTRANS) CT-EMFAC (EMission FACtor) model, and
one to model dispersion of pollutants, such as the
CALINE4 model.
A series of twelve intensively monitored 1-hr CO disper-
sion studies were conducted near Davis, CA, in winter
1996. The experimental equipment included twelve CO
sampling ports at elevations up to 50 m, three sonic an-
emometers, a tethersonde station, aircraft measurements
of wind and temperature profile aloft, and a variety of
conventional meteorological equipment. The study was
designed to explore the role of vehicular exhaust buoy-
ancy during worst-case meteorological conditions, such
as low winds oriented in near-parallel alignment with the
road during a surface-based nocturnal inversion. From the
study, field estimates of the CO emission factor (EF) from
a California vehicle fleet were computed using two differ-
ent methods. The analysis suggests that the CT-EMFAC/
EMFAC (EMission FACtor) models currently used to con-
duct federal conformity modeling significantly overpredict
CO emissions for high-speed, free-flowing traffic on Cali-
fornia highways.
In October 1995, the University of California, Davis
(UCD), in conjunction with the UCD-CALTRANS Air Qual-
ity Project, began an investigation to determine if the
CALINE family of models adequately parameterized the
effect of vehicular exhaust buoyancy during “worst-case”
meteorological events. The intensive sampling effort, con-
ducted near California Interstate 80 (I-80), included a vari-
ety of sampling towers, meteorological balloons, and aircraft
sampling. In addition to specifying near-roadway meteo-
rological fields, the collected data proved ideal for calcu-
lating field-estimated CO EFs. This paper details how
site-specific CO EFs can be determined based on integrat-
ing near-roadway wind and pollutant profiles. A novel ap-
proach to “back-calculate” a CO EF based on the CALINE4
dispersion model is also developed. Lastly, a comparison
between the regulatory model CT-EMFAC and experimen-
tally determined CO EFs is presented.
To better understand why this analysis is insightful,
it is useful to understand the history and development of
the EMFAC and CT-EMFAC models. EMFAC is designed
for use with area inventory models such as BURDEN. These
area inventory models use input from EMFAC and stan-
dard travel demand models, which produce outputs of
vehicle miles traveled (VMT), to estimate mobile-source
pollutant inventories for urban airsheds. The EMFAC
emission rates are based on cycles that represent average
trips and use average speed as an identifier for trip-based
mobile-source emissions.2 Obviously, it is desirable to have
fleet-averaged, aggregate EFs based on speed that is trip-
averaged, rather than instantaneous, to estimate airshed
pollutant loads.
INTRODUCTION
Transportation projects using federal funding are required
to undergo a review of their environmental impacts dur-
ing the planning and design process as specified in con-
formity regulations. Regional and localized air quality
studies are often part of the impact assessment procedure.1
Microscale modeling of air pollutants near roadways in
California is typically accomplished using two models,
IMPLICATIONS
The California Department of Transportation (CALTRANS)
currently uses EMFAC to predict regional emission fac-
tors. For lack of a more appropriate model, transporta-
tion planners routinely use CT-EMFAC, a streamlined ver-
sion of EMFAC, to conduct microscale modeling for fed-
eral conformity requirements. This study indicates that
CT-EMFAC/EMFAC models may significantly overpredict
CO emission factors for free-flowing, high-speed inter-
state highways. These results further emphasize the need
for appropriate emission models to increase the reliability
of microscale analyses.
CT-EMFAC was developed as a “front-end” to
EMFAC3,4 and became the de facto model used for
microscale modeling in California. However, given that the
EMFAC model was not designed for microscale modeling,
Volume 51 January 2001
Journal of the Air & Waste Management Association 121

Held, Chang, and Niemeier
Table 2. Experiment dates and configurations.
its performance is suspect when using it for this purpose.
One of the most significant drawbacks of using EMFAC
for microscale modeling is that the microscale modeler is
trying to estimate a modal emission factor based on a trip-
averaged, or aggregate, EF model. For instance, in this pa-
per we determined the CO EF from a California fleet that
is traveling 65–70+ mph (i.e., a modal EF).
A recent study by the National Research Council noted
the need for a toolbox of modeling tools based on the analy-
sis conditions.5 Our analysis supports this notion and the
findings of previous microscale modeling studies.6-8 We
clearly demonstrate that using a trip-averaged EF model to
determine localized emissions is inappropriate.
Sampling Period
Stations Used
Nov 14, 1996, 6:00–8:00 a.m.
Nov 21, 1996, 6:00–8:00 a.m.
Nov 26, 1996, 6:00–8:00 a.m.
Dec 1, 1996, 6:00–8:00 p.m.
Dec 3, 1996, 6:00–8:00 a.m.
Jan 8, 1997, 6:00–8:00 a.m.
Jan 10, 1997, 6:00–8:00 a.m.
A,B,D,E,G,H
A,B,D,E,F,H,I
A,B,C,D,H,I
A,C,D,E,F,G,I
A,B,C,D,E,F,G,H,I
A,B,C,D,E,F,H,I
A,B,C,D,E,F,H,I
Five-minute grab samples were collected at multiple
elevations using diaphragm pumps, neoprene tubing, and
Tedlar bags at each of the CO sampling stations. Samples
were typically collected every 10 min from 6:00 to 8:00
a.m. except during unfavorable meteorological conditions.
The Tedlar sampling bags were transported to a UCD lab
and analyzed with two Dasibi non-dispersive infrared CO
analyzers within 36 hr of sampling in all cases and within
24 hr in most cases.
SAMPLING SITE DESCRIPTION
In a multi-month study performed near Davis, CA, eight
fixed stations and one aircraft were used to collect pollut-
ant and meteorological data along the I-80 corridor. The
fixed stations included two 18-m and one 6-m sampling
tower, one sampling balloon, one tethersonde, one low-
resolution video camera, two CO autosamplers, and three
sonic anemometers. The function and usage of the vari-
ous sampling stations are presented in Tables 1 and 2.
The station locations and their relation to the observed
highway are shown in Figure 1. Station labels in this fig-
ure correspond with those used in Tables 1 and 2. The
meteorological instruments and CO sampling ports on
the balloon and tower stations are shown in Figures 2, 3,
and 4. Most of the instrumentation was placed south of
the roadway because the predominant daily seasonal wind
direction is from the north. Measurements were collected
on 7 days when meteorological conditions were forecasted
to be favorable. Six of the seven experiments were con-
ducted between 6:00 and 8:00 a.m.; the seventh experi-
ment was conducted between 6:00 and 8:00 p.m.
Meteorological field variables were collected at each
of the 18-m tower locations by data loggers at 1 Hz. Each
18-m tower included four temperature sensors housed in
both aspirated and non-aspirated radiation shields, two
conventional cup anemometer/wind-vane systems, and
one UVW propeller system. Campbell CSAT-3 3-D sonic
anemometers, which sampled wind speed and tempera-
ture at 10 Hz, were mounted on the northern 18-m tower
and the southern 6-m tower at an elevation of ~6 m. In
addition, a 1-D sonic was attached ~3.6 m from the trav-
eled way to a roadway call box (see Figure 1, Station B) so
that the nearside roadway heat flux could be determined.
The tethersonde, balloon, and aircraft measure-
ments complemented ground-based instrumentation
and allowed a detailed description of the planetary
boundary layer to be ascertained. The tethersonde sys-
tem was used to continuously profile wind speed, di-
rection, pressure, temperature, and relative humidity
up to an elevation of ~45 m. The aircraft collected up-
per elevation temperature and wind profiles. Finally,
eastbound and westbound hourly traffic data were col-
lected using CALTRANS automatic loop detectors lo-
cated near the experimental site. A low-resolution video
camera also continuously recorded the eastbound traf-
fic, which was useful in determining vehicle type and
speed estimates, but was not of sufficient resolution to
determine vehicle license plate numbers.
Table 1. Sampling station functions.
Station
Function
A. North 18-m tower (32)a
Heat and momentum fluxes; wind, temperature,
and CO profiles
B. Call box 1-D sonic (3.6)
C. Freeway 6-m tower (14)
D. South 18-m tower (54)
E. Balloon station (117)
Highway heat flux
Heat and momentum fluxes; CO profiles
Wind, temperature, and CO profiles
High-elevation CO profiles
High-elevation wind, temperature, and RH
profiles
F. Tethersonde station (177)
G. Aircraft (variable)
Boundary-layer wind and temperature profiles
Traffic volume
H. Camera station
Sampling Results
I. CO auto samplers (variable)
Remote determination of ground-level CO
concentrations
A complete description of the meteorological data and trace
gas concentrations collected during the study will not be
presented since they are not essential for the subsequent
aApproximate distance in meters to the nearest traveled way of I-80.
122 Journal of the Air & Waste Management Association
Volume 51 January 2001

Held, Chang, and Niemeier
Figure 1. Sampling station locations.
analysis. Interested readers can review Held et al.9 for ad-
ditional results and findings. One-hour averaged traffic
counts, wind speed, wind direction, vertical heat flux, tem-
perature, and CO concentrations collected at the 18-m
towers for the continuous sampling periods are detailed
in Table 3. The “background” CO concentration was as-
sumed to be the 18-m CO concentration at the upwind
tower (the north tower was typically upwind). The wind
Figure 2. North and south 18-m sampling towers.
Volume 51 January 2001
Journal of the Air & Waste Management Association 123

Held, Chang, and Niemeier
fluxes is quite complex (readers are referred to Wilson and
Swaters10 for additional details).
Trace Gas Error Analysis
Given the high signal-to-noise ratio present in the 5-min
averaged CO concentrations, an error analysis is essential
to determine the validity of the EF analysis presented in
this paper. Furthermore, since many sampling efforts
hinge on the averaging of representative “grab-
samples”, it is instructive to demonstrate that our
method of error analysis can be extended to similarly
conducted trace gas sampling efforts. The 5-min aver-
aged CO concentrations measured in this study were
typically less than 2 parts per million by volume (ppmv),
and the Dasibi CO analyzer used for this project was
accurate to ±0.1 ppmv. To determine the signal-to-noise
ratio for the 1-hr calculations, the functional form of
the 1-hr average was assumed to be
Figure 3. 6-m sampling tower.
directions follow standard meteorological convention (i.e.,
wind angle is the direction from which the wind is blow-
ing). In addition, wind angles were aligned so that a wind
blowing perpendicular to the road from north to south
was 0° and a wind parallel to the road from east to west
was 90°. The error associated with the standard meteoro-
logical equipment for 1-hr averages is variable and instru-
ment-specific. The cup-vane systems used in this project
are accurate to ±3° and within ±0.2 m/sec. Sonic anemom-
eter wind speed and direction error is essentially negli-
gible for our purposes, but the interpretation of the heat
(1)
where CO1HR is 1-hr average CO concentration (ppmv);
COi is 5-min averaged CO concentration (ppmv); and N
is number of 5-min periods considered in an hourly aver-
age (typically 6). The error in the 1-hr average CO mea-
surement can be computed as
(2)
where σAVG is the standard deviation of the 1-hr averaged
CO concentration (ppmv), and σi is the standard devia-
tion of the 5-min averaged CO concentration (ppmv).
Since the σi are all assumed to be identical, eq 2 reduces to
(3)
Equations 1–3 require that a sequential measure of
CO over a time period is identical for several simultaneous
samples. This is not an arbitrary assumption made to fa-
cilitate the error analysis. Rather, it is a direct consequence
of assuming that the wind and pollutant covariance struc-
ture near a roadway are statistically stationary for certain
averaging periods (in this case, 1 hr). Although most natu-
ral processes are far from statistically stationary, the as-
sumption is always explicitly, or implicity, made in the
development of steady-state models such as a Gaussian
dispersion model. Thus, calculations made with eqs 1–3
are suitable for comparison with any steady-state model
or any other measurement that assumes statistical
stationarity.
From the equations above, it can be shown the standard
deviation of a 1-hr average CO concentration will be 0.041
ppmv based on a 5-min standard deviation of 0.1 ppmv.
Figure 4. CO sampling balloon.
124 Journal of the Air & Waste Management Association
Volume 51 January 2001

Held, Chang, and Niemeier
Volume 51 January 2001
Journal of the Air & Waste Management Association 125

Held, Chang, and Niemeier
In calculations where a 1-hr averaged CO concentration
was subtracted from a 1-hr averaged background concen-
tration, the standard deviation of the difference can be
shown to be 0.058 ppmv. Since the random error of each
5-min CO measurement will in part cancel out, the sig-
nal-to-noise ratio for a 1-hr average is far superior to that
of the 5-min averages and can be used with greater confi-
dence. Inspection of Table 3 indicates that four CO con-
centrations are reported as negative (–0.017, –0.025,
–0.033, and –0.063 ppmv). All four of these measurements
occurred at the downwind 18-m sampling port where the
CO concentration approached the background at times.
These negative concentrations are obviously nonphysi-
cal and result from sampling error, but the spread of these
values is consistent with the analytical determination of
the 1-hr averaged CO concentration’s random error.
north tower approached a near-parallel wind speed of 0.8
m/sec. If these low, parallel wind conditions had persisted,
it would have been possible to conduct a buoyancy analy-
sis of the type originally envisioned.
The wind was from the south during the December 3,
1996, and January 8, 1997, sampling periods. Since the
experiment was designed for a northerly wind, the ex-
periment was not conducted for the full 2 hours on these
days. In addition, CALTRANS traffic counts were not avail-
able for the January 8, 1997, sampling, making a com-
plete analysis difficult. The last study was conducted
January 10, 1997. Wind speeds during this period were
also well above 1 m/sec and were qualitatively similar to
the first two sampling periods.
EMISSION FACTOR ANALYSIS
Three separate methods were used to determine the CO
EF for the I-80 vehicle fleet. The first method was based
on the CT-EMFAC (release 2.01) model, the second method
was based on back-calculating the CO EF from experimen-
tal measurements, and the last method was based on com-
puting the best fit between the observed data and
CALINE4-predicted CO concentrations.
Meteorological Conditions during Each
Sampling Day
The UCD CO sampling began November 14, 1996. At ap-
proximately 5:00 a.m., the 0.6-m temperature was ~4 °C
colder than the 18-m temperature, indicating a tempera-
ture increase of ~1 °C/5 m. The 10-m wind speed was
~2.7 m/sec with a 6° standard deviation of the horizon-
tal wind direction. The heat flux determined by the up-
wind sonic anemometer varied from –38 to –21 W/m2.
Thus, the stability class during this time period was stable
to extremely stable (class E–G depending on the stabil-
ity criteria applied). The combined traffic flow rate dur-
ing the November study approached 10,000 vehicles per
hour (VPH). The meteorological conditions during the
November 21, 1996, study were similar to the first sam-
pling day, but the temperature inversion was significantly
less intense.
The wind speed during the November 26, 1996, sam-
pling period was exceedingly high. The ground-level wind
speed was typically greater than 5 m/sec, and the wind
speed at 18 m exceeded 10 m/sec at times. Given the strong
winds, the balloon and tethersonde stations were not used,
and only the fixed tower stations collected CO data. Al-
though the wind speeds were strong, the wind directions
were approximately constant, with a 1-hr averaged stan-
dard deviation of less than 3° at the upwind station.
The December 1, 1996, sampling was the only night-
time sampling for this study and was selected because an
evening rush hour was expected due to the Thanksgiving
holiday weekend. The CALTRANS hourly counts indicate
that the hourly traffic flow rate approached 8500 VPH
during the sampling period (6:00–8:00 p.m.) and was com-
parable in magnitude to the morning commute-hour stud-
ies. The ground-level wind speed during the sampling
period varied between 1 and 2 m/sec. It is worth noting
that for a brief period, the ground-level wind speed at the
CT-EMFAC CO Emission Factor
The California Air Resources Board (ARB) has developed,
and currently supports, a modeling tool known as EMFAC
to estimate vehicular emission factors for various pollut-
ants. The model is similar to the U.S. Environmental Pro-
tection Agency model MOBILE, but it takes into account
the vehicle fleet, fuel, and maintenance programs spe-
cific to California. The model CT-EMFAC was developed
by CALTRANS to simplify estimation of composite EFs
based on user-supplied estimates of vehicle fleet operat-
ing modes (i.e., cold-start percentage) and vehicle mix
distributions (i.e., percentage of heavy-duty trucks).
CT-EMFAC is essentially a front-end to the EMFAC model
that enables microscale modelers to run analyses without
mastering the entire EMFAC model. Thus, CT-EMFAC re-
sults can be generalized to the EMFAC model as well.
Transportation planners can estimate CT-EMFAC in-
put parameters based on the California Carbon Monox-
ide Protocol11 (CCMP) recommendations and experience.
In this study, a variety of input parameters are used to
demonstrate the sensitivity of the CT-EMFAC model to
user input. The CCMP recommendation for a California
vehicle fleet distribution is presented in Table 4. Table 5
lists the hot-stabilized CO EF for various vehicle classes
based on a 1996 distribution of vehicle age. Table 5 indi-
cates that the LDA, LDT, and MDT CO EFs are essentially
identical at high speeds (see Table 4 for acronym defini-
tions). Heavy-duty gas trucks have significantly higher CO
EFs than LDAs at all speeds, whereas HDD CO EFs are
126 Journal of the Air & Waste Management Association
Volume 51 January 2001

Held, Chang, and Niemeier
Table 4. Vehicle distribution assumed for the calculation of CO emission factors.
start of 50% (Figure 5, diamond marker) does one see a
significant impact in CO EFs.
Vehicle Type
Percent of Fleet
The approximate average vehicle speed during the
study was calculated from video counts to be at least 70
mph; however, CT-EMFAC is only capable of estimating
CO emission factors for an average speed of 65 mph or
less. Inspection of Figure 5 demonstrates that the CO EF
sharply increases when the average vehicle speed is in
excess of 55 mph, but it is unclear if this trend continues
at higher speeds. To estimate CO EF at speeds greater than
65 mph, a quadratic extrapolation was assumed for the
55–65 mph CT-EMFAC EF values. This reasoning is con-
sistent with the argument that the engine load is propor-
tional to the square of the vehicle speed in this speed
range. In reality, residence time on the catalyst bed may
be a more realistic explanation for increased emissions,
which would suggest that the EF should be linearly ex-
trapolated.
Light-duty automobiles (LDA)
Light-duty trucks (LDT)
69
19.4
6.4
1.2
3.6
0.4
Medium-duty trucks (MDT)
Heavy-duty gas trucks (HDG)
Heavy-duty diesel trucks (HDD)
Motorcycles (MC)
Table 5. Hot-stabilized CO EF (g/vehicle-mile) for various vehicle classes operating
at 40 °F.
Speed (mph)
LDA
LDT
MDT
HDG
HDD
MC
50
55
60
65
3.2
3.6
3.2
3.8
3.4
3.8
18.5
20.2
23.3
28.4
6.7
7.0
7.6
8.6
5.9
5.7
5.2
3.7
The 65-mph EF for a 5% cold-start fleet is ~15 grams
per vehicle mile traveled (GPVMT). One would obviously
expect that the 70-mph EF would be greater than the
65-mph EF; thus, 15 GPVMT represents the lowest EF that
a traffic engineer could reasonably select using known or
predicted values. If a traffic engineer assumed that the
CT-EMFAC curve increased linearly, the calculated EF for
70 mph would be ~22 GPVMT; parabolically, it would be
28 GPVMT.
5.9
6.1
6.3
13.4
14.9
15.6
comparable at high speeds. Thus, the CO EF is relatively
insensitive to vehicle distribution unless the heavy-duty
gas vehicle fraction exceeds 10%, which did not occur
during the sampling effort. Therefore, it is unlikely that a
slight misestimation of the I-80 vehicle fleet distribution
will have a significant impact on the estimated CO EF.
In addition to the vehicle distribution, CT-EMFAC
requires the user to estimate the percentage of the vehicle
fleet operating in cold, hot, and stabilized running modes.
The CCMP recommends that hot spot modelers select a
cold-start percentage ranging from 1 to 15 for the observed
roadway. The range of possible cold-start percentages re-
sults from differing assumptions about the freeway classi-
fication, with the lower percentages representing segments
with fewer expected modal changes. The lack of any sig-
nificant on-ramps within 3–5 mi of the experimental site
suggests a lower cold-start fraction, 0–5%, for the vehicle
fleet was appropriate.
CO Emission Factor Estimation from
Field Measurements
It is also possible to directly calculate a composite vehicle
fleet CO EF from the field data collected in this study.
The relationship between line source strength, vehicular
flow rate, and EF is
(4)
where SS is CO source strength of roadway (g CO per mph);
EF is vehicular emission factor (g CO per vehicle per mi);
and Q is vehicular flow rate (VPH). Given Q, one can com-
pute an EF if a method for determining SS is available.
One way of determining SS from the field data collected
in this study is to perform a numerical integration of the
CO flux based on curve-fitting of the CO and wind pro-
files as shown in eq 5.
Figure 5 presents the modeled CT-EMFAC CO EFs for
various vehicle fleet and operation mode distributions for
an ambient temperature of 40 oF (a typical ground-level
temperature recorded during the sampling effort). The
100% LDA fleet operating in stable mode (square mark-
ers) represents the lowest CO EF value one could reason-
ably expect for the observed roadway. The CCMP vehicle
fleet with 5% cold start is a vehicle distribution that a
transportation engineer would presumably select as a de-
fault value. Clearly the differences between the baseline
and CCMP distributions are relatively insignificant at high
speeds. Only if one considers an unreasonably high cold
(5)
where U is average perpendicular wind speed (m/sec);
⊥
CO is average CO concentration (g/m3); z is elevation
(m); κ is unit correction factor; L is upper boundary
considered in the flux calculations (m); and z0 is rough-
ness length (m).
Volume 51 January 2001
Journal of the Air & Waste Management Association 127

Held, Chang, and Niemeier
Figure 5. CT-EMFAC emission factors for various fleet compositions. Diamond and triangle series based on vehicle distribution presented in Table 4.
Given that the topography near the highway (due to
oleanders in the highway median) is complex and the
freeway traffic is a significant source of turbulence, it is
unclear whether the log-wind profile or a diabatic wind
profile is more representative downwind of the roadway.
In this study, it was assumed that the wind speed was
adequately described by a neutral logarithmic wind pro-
file that did not have significant wind direction rotation
with height for the first few tens of meters above the
ground (see eq 6). The friction velocity and surface rough-
ness were determined using a Marquardt least-squares
nonlinear fitting for each 1-hr time period. These assump-
tions should be satisfactory for a first-order approxima-
tion of SS.
function also fit the data set quite well and was used for
comparison purposes. Once the functional forms of U and
CO were determined, the effective source strength of the
road was calculated by eq 5 and the EF was back-calcu-
lated from eq 4.
The method derived above is based on the following
assumptions:
•
•
•
The CO and wind fields are steady over the aver-
aging time of the CO bag sample;
The functional forms selected to represent the
CO and wind profiles are valid;
The CO profile results from an upwind infinite
line source (i.e., there are no edge effects at the
end of the line source);
•
•
•
•
There is no pollutant flux above the top control
element considered;
There is negligible pollutant flux between 0 m
and z0;
(6)
where u*⊥ is friction velocity based on perpendicular wind
(m/sec), and k is the von Karman constant (0.4).
The emission source strength is linearly propor-
tional to vehicle count; and
The along-wind diffusive flux is insignificant in
comparison with the advective flux.
The exact functional form of the CO profile is also
unknown. For a first-order approximation, it was assumed
that the CO profile decreased exponentially with height.
This functional form does not have a rigorous theoretical
basis; however, it exhibited good agreement with the
1-hr CO samples with the background removed. A linear
Given the meteorological and traffic conditions ex-
perienced in the field, these assumptions appear to be
valid. The most questionable assumptions made in this
128 Journal of the Air & Waste Management Association
Volume 51 January 2001

Held, Chang, and Niemeier
Table 6. CO emission factor (GPVMT) back-calculated from downwind CO and
analysis were the functional forms selected for the CO
and wind profiles. However, we feel that the assumed
CO and wind profiles were sufficiently accurate for a first-
order estimate. It should also be noted that the CO con-
centration profile typically approached 0 ppmv prior to
the upper elevation CO port. Thus, it is unlikely that there
was a significant flux of CO overlooked when the defi-
nite integral upper bound for eq 5 was considered finite.
For the linearly-fit CO profile calculations that follow, L
was varied from 16 to 23 m to ensure that the CO flux at
upper elevations could not be negative (upper-elevation
negative fluxes are expected given a linear functional form
of the CO profile). For the exponential fit, L was consid-
ered to be 4 times the L from the linear fit (~60 m) to
ensure that at least 99% of the CO flux was considered in
the flux integral.
A program was written to numerically back-calculate
a composite EF based on 1-hr averaged wind and CO pro-
files for each period that CALTRANS hourly traffic counts,
wind speed, wind direction, and CO measurements were
available. Results of the analysis are shown in Table 6.
The average EF for the twelve 1-hr periods considered was
calculated to be 6.4 GPVMT with a standard deviation of
2.8 for the linear CO fit, and 6.8 GPVMT with a standard
deviation of 3.3 for the exponential CO fit. Insufficient
data were available to determine why the CO EF for the
two periods of southerly wind were significantly greater
than the EF for the ten periods of northerly wind. Figure 6
shows the wind, CO, and CO flux profiles for two periods
for both the linear and exponential fit CO profiles. The
shape of the flux profile is as expected and demonstrates
that an insignificant amount of mass is lost above eleva-
tion L or below zo.
wind profiles.
Time Period
Linear CO Profile
Exponential CO Profile
Nov 14, 6:00–7:00 a.m.
Nov 14, 7:00–8:00 a.m.
Nov 21, 6:00–7:00 a.m.
Nov 21, 7:00–8:00 a.m.
Nov 26, 6:00–7:00 a.m.
Nov 26, 7:00–8:00 a.m.
Dec 1, 6:00–7:00 p.m.
Dec 1, 7:00–8:00 p.m.
Dec 3, 6:00–7:00 a.m.
Jan 8, 6:00–7:00 a.m.
Jan 10, 6:00–7:00 a.m.
Jan 10, 7:00–8:00 a.m.
Average
8.9
4.7
5.7
2.8
7.1
4.7
5.9
8.0
10.7
11.1
3.6
3.2
6.4
2.8
13.0
5.4
7.3
3.2
6.5
3.5
6.6
8.3
11.1
10.1
3.8
3.3
6.8
3.3
Standard Deviation
In this analysis, the wind speed used was from the up-
wind 4-m wind sensor. Noting that CO concentrations
are inversely proportional to wind speed, a wind speed
measured at a higher elevation will result in a lower
downwind CO concentration since wind speed typically
increases with height. Futhermore, if CO concentrations
are held constant, an increase in wind speed would re-
quire an increase in EF.
Figures 7 and 8 present the measured and CALINE4-
modeled CO concentrations based on two different emis-
sion factors for the November 21, 1996, 6:00 to 7:00 a.m.,
and December 1, 1996, 6:00 to 7:00 p.m., sampling peri-
ods. The 7.0 GPVMT line represents the minimized sum
of squares residual fit EF, and 22 GPVMT is considered
the EF that a traffic engineer would select based on the
CT-EMFAC analysis. Clearly an EF of 7.0 GPVMT matches
the measured CO concentrations more accurately than
22 GPVMT does. For instance, in Figure 8 the measured
and predicted CO concentrations at the lower freeway
tower sampling port for an EF of 7.0 are nearly identical,
whereas the EF of 22 results in an overprediction of ~340%.
The remaining 1-hr time periods also demonstrate the
trends presented in Figures 7 and 8 for the balloon and
tower stations. Thus, CALINE4 appears to provide reason-
able estimates of dispersion 30–60 m from a roadway if
accurate emissions input data are supplied during non-
worst-case meteorological conditions.
CALINE4 Analysis
By minimizing the total squared residual between the
measured and CALINE4-predicted CO concentrations at
the downwind 18-m and 6-m towers for the eleven 1-hr
periods with complete data, a CO EF for the vehicle fleet
traveling on I-80 could be computed. The CO EF calcu-
lated from this method was 6.5 GPVMT. For human health
concerns, only the lower-elevation CO predictions are of
interest. When only CO concentrations at elevations less
than 6 m were considered, the EF with the minimized
sum of squares residual was found to be 7.0 GPVMT.
The CALINE4 simulations were based on the col-
lected field data (i.e., field-measured σ was used rather
θ
than default settings). Although CALINE4 is a relatively
simple model to run, the wind-speed input is worth clari-
fying. According to the user manual, the wind speed
should be “measure[d] at 5 to 10 m or assume worst-
case. For localized sources and nearby receptors, wind
speeds measured at lower elevations (5 m) [are] desirable….”1
Since 7.0 GVPM is the best-fit EF, there are several
instances where CALINE4 underpredicted CO concen-
trations near the observed highway. Thus, it would ap-
pear warranted to multiply the best-fit EF by a safety
factor (SF) so that CO concentration estimates would be
conservative. However, applying an SF greater than
Volume 51 January 2001
Journal of the Air & Waste Management Association 129

Held, Chang, and Niemeier
Figure 6. Wind speed, CO concentration, and CO flux profiles downwind of I-80. Dashed lines are exponentially fit and solid lines are linearly fit for
the CO and CO flux subplots. The top three plots are for the January 8, 6:00–7:00 a.m. sampling period when the wind was from the south. The
bottom three plots are for the January 10, 6:00–7:00 a.m. sampling period when the wind was from the north.
Figure 7. Measured and CALINE4-predicted CO concentrations for the November 21, 1996, 6:00–7:00 a.m. sampling period for two different CO
emission factors. Wind angle orientation based on standard meteorological convention. ST = south tower; B = balloon station.
130 Journal of the Air & Waste Management Association
Volume 51 January 2001

Held, Chang, and Niemeier
Figure 8. Measured and CALINE4-predicted CO concentrations for the December 1, 1996, 6:00–7:00 p.m. sampling period for two different CO
emission factors. Wind angle orientation based on standard meteorological convention. ST = south tower; B = balloon station; FT = freeway tower.
3 would result in a CO EF similar to the one predicted by
CT-EMFAC. This appears to obviate the need for the ad-
ditional EF analysis performed in this study, since we
are suggesting the use of an EF similar to the CT-EMFAC
model. In reality, this is not an “apples to apples” com-
parison, because the CT-EMFAC model is intended to
produce exact, not conservative, EFs. As it stands, the
transportation planner has no practical intuition whether
an EF from CT-EMFAC or CALINE4 already has a built-in
SF, or whether one needs to be used. Efforts should be
made to provide transportation/air-quality modelers with
guidelines indicating where and when SFs should be used
for screening purposes and to clearly state what implicit
and explicit SFs are present in both emission factor and
dispersion models.
The downwind concentrations observed near the
roadway are quite low, typically ranging from 0 to 1
ppmv, and are in no danger of violating the California
or federal CO standard. Observed CO concentrations in
this study were low due to the modest traffic loading
and near-zero background concentration. In areas where
the background CO concentration is greater, the road-
way contribution necessary to exceed the air quality stan-
dard would be less. Thus, the overestimation of the CO
EF would probably be most significant near freeways with
greater traffic loading in an area with an ambient back-
ground CO concentration approaching the regulatory
standard.
CONCLUSIONS
Three methods were used to determine the CO EF for a
vehicle fleet traveling on a freeway near Davis. CT-EMFAC,
which would presumably be used by a traffic engineer to
perform a screening analysis of a proposed or existing
roadway, resulted in a CO EF prediction of approximately
15–28 GPVMT. A CO EF back-calculated from the inte-
grated highway CO flux was found to be ~6.4 or 6.8
GPVMT. Lastly, CALINE4 was used to minimize the error
between measured and modeled CO concentrations and
resulted in a CO EF of ~7 GPVMT. Figures 7 and 8 are
representative of the twelve 1-hr sampling periods and
demonstrate that an EF of 7.0 GPVMT provides excellent
agreement between measured and CALINE4-predicted
concentrations. It appears that a CO EF of 7 GPVMT is
much more appropriate for the observed roadway than
the 15–28 GPVMT that would be predicted by the
CT-EMFAC model. Futhermore, since a predicted down-
wind CO concentration is linearly proportional to the CO
EF, a 100% error in EF estimation will lead to a 100% error
in predicted CO concentrations.
Volume 51 January 2001
Journal of the Air & Waste Management Association 131

Held, Chang, and Niemeier
5. National Research Council. Modeling Mobile-Source Emissions; Wash-
ington, DC, Committee To Review EPA’s Mobile Source Emissions
Factor Model; Board on Environmental Studies and Toxicology; Com-
mission on Geosciences, Environment, and Resources; Transporta-
tion Research Board; National Research Council: 2000.
6. Moseholm, L.; Silva, J.; Larson, T. Forecasting Carbon Monoxide Con-
centrations near a Sheltered Intersection Using Video Traffic Surveil-
lance and Neural Networks; Transportation Research Part D: Transport
Environment 1996, 1D.
7. Scully, R.D. Vehicle Emission Rate Analysis for Carbon Monoxide Hot
Spot Modeling; J. Air Pollut. Control Assoc. 1989, 39, 1334-1343.
8. Zamurs, J.; Conway, R. Comparison of Intersection Air Quality Model’s
Ability to Simulate Carbon Monoxide Concentrations in an Urban
Area; Transport. Res. Record 1991, 23-32.
9. Held, A.E.; Chang, D.P.Y.; Carroll, J.J. Observations and Model Simula-
tions of Carbon Monoxide Dispersion; California Department of Trans-
portation: 1998.
10. Wilson, J.D.; Swaters, G.E. The Source Area Influencing a Measure-
ment in the Planetary Boundary Layer—The Footprint and the Dis-
tribution of Contact Distance; Boundary-Layer Meteorol. 1991, 55,
25-46.
There are some limitations associated with the tech-
niques employed in this study. Specifically, the downwind
CO and wind profiles were limited to three points for each
1-hr average. Although the three-point curves appear to
adequately fit the CO profile, the log wind profile and
the measured wind profile did not show strong agreement
for every 1-hr time period. While the functional form of
the CO and wind profiles may not be exact, the method-
ology for calculating the CO flux is still suitably robust to
demonstrate the significant difference between measured
and modeled CO EF.
The EMFAC model was originally designed for regional
air quality modeling of mobile emissions. The CT-EMFAC
model uses selected features of the EMFAC model to esti-
mate an aggregate EF for a vehicle fleet traveling at a trip-
averaged speed. However, for “hot-spot” analysis, a modal
EF representing vehicle fleet emissions by mode of opera-
tion (i.e., traveling at 25 mph or idling) is required. Thus,
improperly using regional emission factors to determine
modal emissions introduces significant error into
microscale analysis and highlights the need for new modal
emission factor models designed expressly to perform fa-
cility-specific microscale air quality studies.
11. Garza, V.J.; Franey, P.; Sperling, D. Transportation Project-Level Carbon
Monoxide Protocol; Institute of Transportation Studies, University of
California, Davis: Davis, CA, 1996.
ACKNOWLEDGMENTS
Support for this study was provided by CALTRANS under
the guidance and direction of Steve Borroum. Additional
assistance from CALTRANS was provided by Keith Jones,
Doug Eisinger, Marianne Larsen, and Tom Kear. The au-
thors gratefully acknowledge Dr. Roger Shaw, Dr. Kyaw
Tha Paw U, Dr. John Carroll, and Ed Tai for their support
and insight. Lastly, the authors would like to thank the
crew members, whose hard work made these experiments
possible: George Croll, Kellie Dougherty, Dale Uyeminami,
Sid Sok, Vicente Garza, and Pingkuan Di.
About the Authors
Anthony Held is currently a doctoral student at the Univer-
sity of California, Davis. The work presented in this paper
was the foundation of his Master of Science research. Please
address all correspondence to Anthony Held, Department
of Civil and Environmental Engineering, One Shields Ave.,
Davis, CA 95817; e-mail: aeheld@ucdavis.edu. Daniel P.Y.
Chang is a professor of Civil and Environmental Engineer-
ing and currently serves as chair of the department. His
research interests are in the areas of physico-chemical and
biological air pollution control, by-products of combustion
processes, dispersion modeling, and health effects. Debbie
Niemeier is an associate professor in the Department of Civil
and Environmental Engineering. Her research interests are
in the areas of transportation forecasting and modeling of
mobile emissions.
REFERENCES
1. Benson, P.E. CALINE4, a Dispersion Model for Predicting Air Pollutant
Concentrations Near Roadways; California Department of Transporta-
tion: 1984.
2. Nanzetta, K.; Niemeier, D.; Utts, J.M. Changing Speed-VMT Distribu-
tions: The Effects on Emissions Inventories and Conformity; J. Air &
Waste Manage. Assoc. 2000, 50, 459-467.
3. California Air Resources Board. Methodology for Estimating Emissions
from On-Road Motor Vehicles; Volume I–III, 7f; Prepared by Technical
Support Division Mobile Source Emission Inventory Branch, Califor-
nia Air Resources Board: September 1993.
4. California Air Resources Board. Methodology for Estimating Emissions
from On-Road Motor Vehicles; Volume I–VII, 7g; Prepared by Technical
Support Division Mobile Source Emission Inventory Branch, Califor-
nia Air Resources Board: December 1995.
132 Journal of the Air & Waste Management Association
Volume 51 January 2001