TECHNICAL PAPER

Evaluating measurements of carbon dioxide emissions using a precision

source—A natural gas burner
Rodney Bryant,1,⁄ Matthew Bundy,1 and Ruowen Zong2
1
National Institute of Standards and Technology, Gaithersburg, MD, USA
2
University of Science and Technology of China, Hefei, Anhui, People’s Republic of China
⁄
Please address correspondence to: Rodney Bryant, NIST, 100 Bureau Drive, MS 8662, MD 20899, USA; e-mail: rodney.bryant@nist.gov

A natural gas burner has been used as a precise and accurate source for generating large quantities of carbon dioxide (CO2)
to evaluate emissions measurements at near-industrial scale. Two methods for determining carbon dioxide emissions from
stationary sources are considered here: predicting emissions based on fuel consumption measurements—predicted emissions
measurements, and direct measurement of emissions quantities in the flue gas—direct emissions measurements. Uncertainty for
the predicted emissions measurement was estimated at less than 1%. Uncertainty estimates for the direct emissions measurement
of carbon dioxide were on the order of ±4%. The relative difference between the direct emissions measurements and the predicted
emissions measurements was within the range of the measurement uncertainty, therefore demonstrating good agreement. The
study demonstrates how independent methods are used to validate source emissions measurements, while also demonstrating how
a fire research facility can be used as a precision test-bed to evaluate and improve carbon dioxide emissions measurements from
stationary sources.

Implications: Fossil-fuel-consuming stationary sources such as electric power plants and industrial facilities account for more
than half of the CO2 emissions in the United States. Therefore, accurate emissions measurements from these sources are critical
for evaluating efforts to reduce greenhouse gas emissions. This study demonstrates how a surrogate for a stationary source, a fire
research facility, can be used to evaluate the accuracy of measurements of CO2 emissions.

Introduction monitoring system (CEMS). The measurements for this

Carbon dioxide (CO2) accounts for approximately 84% of
total greenhouse emissions in the United States, with fossil fuel
combustion being the largest source. Since 1990, fossil fuel
combustion has been responsible for more than 90% of U.S.
CO2 emissions. In 2011, electric power generation had the
largest contribution of any sector. Fossil-fuel-consuming sta-
tionary sources such as electric power plants and industrial
facilities accounted for more than half of all CO2 emissions
in the United States (U.S. Environmental Protection Agency
[EPA], 2013). These large emission sources are therefore a
logical choice to examine the accuracy of reported CO2
emissions.
There are two primary methods for determining CO2 emis-
sions from stationary sources. The first method uses combus-
tion stoichiometry to compute theoretical amounts of CO 2
emissions from fuel consumption data. The fuel consumption
data come from precombustion measurements, and conver-
sion efficiency factors are applied to predict CO2 emission
amounts; hence, the method is hereafter referred to as pre-
dicted emissions measurements. The second method requires
measurements of the CO2 concentration and the volume flow
rate of the flue gas in the stack using a continuous emissions
method are postcombustion and are therefore direct measure-
ments of the CO2. This method is hereafter referred to as
direct emissions measurements.
Recent studies have examined the accuracy of reported
CO2 emissions derived from the two primary methods by
comparing data for individual electric power plants.
Ackerman and Sundquist compared CO2 emissions data for
2004 from two national databases, the EPA Emissions &
Generation Resource Integrated Database (eGRID) and the
U.S. Department of Energy Energy Information
Administration (EIA) database (Ackerman and Sundquist,
2008). Emissions data for stack measurements using CEMS
were extracted from the eGRID database, whereas emissions
predictions computed from fuel consumption data were
extracted from the EIA database. A comparison of matching
pairs of data records for individual power plants showed the
average of the absolute relative difference to be about 17%
for this subset of the two databases. In the most recent
investigation, the databases from 2009 were compared for
210 coal-fired power plants (Quick, 2014). The distribution
of the relative differences was approximately ±11% (95%
confidence level). Both studies were consistent in their report-
ing of distributions greater than ±10%. This is an indicator of

863
Journal of the Air & Waste Management Association, 65(7):863–870, 2015. This article not subject to U.S. copyright law. ISSN: 1096-2247 print
DOI: 10.1080/10962247.2015.1031294 Submitted November 17, 2014; final version submitted February 3, 2015; accepted March 10, 2015.
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/uawm.
864 Bryant et al. / Journal of the Air & Waste Management Association 65 (2015) 863–870

the level of uncertainty in emissions measurements at indivi-

dual power plants as well as other stationary sources. Cross-
checking emissions values with independent methods for
determining CO2 emissions is a necessary step to evaluate
the current level of practice of measurement science as
applied to emissions measurements.
The previously mentioned investigations identify uncertainty
in the CEMS measurements as a potential cause for the large
discrepancies in CO2 emissions when comparing the databases.
However, the investigations also recognize that predicting CO2
emissions from fuel consumption data and emissions factors is not
without uncertainty. The present study seeks to demonstrate how
a detailed uncertainty analysis of both direct emissions measure-
ments and predicted emissions measurements is useful to under-
standing the causes for discrepancy when the two methods are
compared. Figure 1. Photograph of the natural gas burner operating at a heat release of 8
Experiments were conducted at the National Institute of MW. The 1.2-m-wide nonpremixed tube burner can deliver controlled fires
Standards and Technology (NIST) Large Fire Research from 0.1 to 8 MW.
Laboratory (LFRL) to compare CO2 emissions as part of an
effort to evaluate the uncertainty of emissions measurements in
large combustion sources. A natural gas burner was used to Predicted emissions measurements
generate precise and accurate quantities of CO2 at near-indus- The natural gas burner (Figure 1) can operate at heat release
trial scale. Carbon dioxide emission rates for the flue gas were rates up to 8 MW. Measurements of volume flow rate, pressure,
derived from direct measurements of the CO2 concentration temperature, and gas composition are made in the natural gas
and the volume flow rate in the facility’s exhaust duct. At the delivery system just upstream of the burner (Figure 2). The
same time, measurements of the natural gas composition and system consists of a positive displacement flow meter, thermis-
flow to the burner provided fuel consumption data to predict tor temperature probes, pressure transducers, and a gas chro-
CO2 emission rates. This study simulates a periodic cross- matograph to analyze the gas composition. From these
check of CO2 emission values for a stationary source, similar measurements, it is possible to compute the amount of heat
to a relative accuracy test audit. It also demonstrates how the released from the burner.
conservation of mass is an excellent guiding principle for such The gas composition measurements also make it possible to
verification exercises. determine the molecular weight, compressibility, and carbon
content of the fuel. Carbon content measurements updated
every 180 sec and fuel consumption measurements updated
Experimental Methods every 1 sec allow for real-time CO2 emission predictions from
the natural gas burner. This treatment of the measurements
Facility makes the natural gas burner a precision source of carbon
dioxide. The CO2 emissions can be derived from the following
The NIST LFRL is used for the study of full-scale fires in expression, where the parameters are defined in Table 1.
buildings. During the routine fire experiments conducted in
the facility, the flow and concentration of effluents in the V_ ng Png Xc;ng ηb MCO2
exhaust duct are measured, much like CEMS measurements m_ CO2 ;p ¼ (1)
RTng Zng
at the smoke stack of a stationary source. The flue gas
measurements are used to derive the primary measurement
Assuming that all of the carbon mass in the fuel is converted to
parameter of the facility, the rate of heat released by a fire. A
CO2 and that all of the combustion products are captured by
1.2 m × 1.5 m tube-bed burner (Figure 1), which issues a
the canopy exhaust hood, eq 1 represents the mass flow rate of
turbulent diffusion flame of natural gas, is used as a refer-
CO2 generated by the fire and injected into the exhaust duct.
ence fire source for the heat release rate measurements
(Bryant et al., 2003). The LFRL is currently in the process
of a major construction remodel and expansion. It will
reopen as the National Fire Research Laboratory (NFRL),
equipped with an additional hood and floor space to accom-
modate fires with heat release rates as large as 20 MW.
Therefore, in addition to the contribution to fire research,
the added heat release rate capacity extends the range of use
of the facility as a near-industrial-scale surrogate to study the Figure 2. Photograph of the natural gas fuel delivery system used to control
issues related to emissions measurements from stationary and measure the mass flow rate and composition of the fuel that supplies the
sources. burner.
 Bryant et al. / Journal of the Air & Waste Management Association 65 (2015) 863–870 865

Table 1. Example of an uncertainty budget for predicted emissions measurements of CO2

Nondimensional
Relative Standard Sensitivity Percent
Measurement Component, xi Value Uncertainty, u(xi)/xi Coefficient, si Contribution, %

Gas volume flow rate, V_ ng (m3/sec) 0.02983 0.0019 1.0 22.9

Gas pressure, Png (Pa) 197719 0.0016 1.0 16.3
Gas temperature, Tng (K) 290.65 0.0017 −1.0 19.0
Gas compressibility, Zng (—) 0.9958 0.0005 −1.0 1.6
Gas carbon fraction, Xc,ng (mol/mol) 1.042 0.0020 1.0 26.2
CO2 molecular weight, MCO2 (g/mol) 44.0095 0.0000 1.0 0
Ideal gas constant, R (J/mol/K) 8.3144 0.0002 −1.0 0
Burner conversion efficiency, ηb (—) 1.0000 0.0015 1.0 14.0
Predicted CO2 emissions, m_ CO2 ;p (g/sec) 112.4 0.0040 (0.0080) Standard (Expanded) Uncertainty
Note: The CO2 was generated with a 2-MW fire from the natural gas burner.

Therefore, the equation represents the CO2 emissions com-
puted from fuel consumption and composition—the predicted
emissions measurements. Table 1 demonstrates nominal values
for the input measurements of eq 1. The burner was operated
with fires of 2 MW or less for this investigation due to the need
to conduct the velocity traversing experiments, which required
the burner to run for extended periods. Using lower heat
release rates limited the radiant heat exposure to the surround-
ing environment. When operating at full capacity, the natural
gas fire generates approximately 0.5 kg/sec of CO2. The burner
can be used to simulate steady-state and transient combustion
processes from a moderate size stationary source such as an
industrial plant.
An uncertainty analysis was performed to estimate the com-
bined uncertainty of the predicted emissions measurements of
CO2. Assuming that the input measurements for eq 1 were
mutually independent, the following equation was applied to
estimate the combined relative uncertainty:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u N  
uc ðyÞ u X uðxi Þ 2
¼t s2i (2)
y i¼1
xi
The standard uncertainty, u(xi), for each input measurement, xi,
used to compute the predicted CO2 emissions (y ¼ m_ CO2 ;p ) is
listed in Table 1. The nondimensional sensitivity coefficient,
given as,
@y xi
si ¼
@xi y
is also listed in the table to reflect the weighting applied to the
standard uncertainty of each component. Estimates of the rela-
tive expanded uncertainty (twice the relative standard uncer-
tainty for a 95% confidence interval) were nominally better
than ±0.017 for the predicted emissions measurements.
Improvements in the flow meter calibration and temperature
and pressure measurements reduced the relative expanded
uncertainty to nominally ±0.010 or less. The largest compo-
nents of uncertainty were the fuel carbon content and the
volume flow rate measurement. Exhaust stream measurements
of CO2, CO (carbon monoxide), and O2 (oxygen) were per-
formed to verify the burner conversion efficiency and plume
capture assumptions. Measureable amounts of CO were not
detected in the flue gas; therefore, complete carbon conversion
was assumed (ηb = 1) and the detection limit of the measure-
ment was used to estimate the uncertainty. A similar methodol-
ogy was used in a previous study of compartment fires to
estimate combustion efficiency (Bundy et al., 2007). A detailed
discussion of the uncertainty analysis for the predicted emis-
sions measurements can be found in a previous publication
(Borthwick and Bundy, 2011). Only data for experiments
with complete capture of the fire plume by the canopy exhaust
hood were included in this study.
Direct emissions measurements
Large canopy exhaust hoods were used to capture the
combustion products from the burner. The canopy hoods direct
the flow into the exhaust ducts that run along the roof of the
facility and were instrumented for measuring gas temperature,
velocity, and volume fraction of selected combustion products.
The maximum exhaust flow capacity is approximately
50 kg/sec of air and the operating pressure in the duct was
(3)
slightly below atmospheric.
Mean flow velocity in the exhaust duct was determined
from a collection of point velocity measurements conducted
by traversing two S probes, equipped with thermocouples,
across a section of the exhaust duct.1 The exhaust duct,
shown in Figure 3, runs horizontally along the roof of the

1
Note: Certain commercial entities, equipment, or materials are identified in this document in order to describe an experimental procedure or concept adequately.
Such identification does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the entities,
materials or equipment are necessarily the best available for the purpose.
866 Bryant et al. / Journal of the Air & Waste Management Association 65 (2015) 863–870
Figure 3. Direct emissions measurements were made in the exhaust duct that
ran along the roof of the Large Fire Research Laboratory.
facility with a series of turns. The cross-section for the velo-
city traverses had an inner diameter, D, of 1.504 ± 0.024 m
and was located 9.2 diameters downstream of a 180° bend.
Velocity profiles were measured on two perpendicular chords,
passing near the centerline of the duct (Figure 4). The point
velocity measurements were conducted according to the pro-
cedures defined by EPA method 2G (EPA, 2007), which
accounts for the angle of the flow in the plane perpendicular
to the traverse line—the yaw angle, and therefore determines
the near-axial velocity. Details of the mean flow velocity
measurements were discussed in a previous paper (Bryant
et al., 2014).
Figure 4. Schematic of the measurement section of the 1.5-m exhaust duct. Bulk flow was computed from a series of point velocity measurements made by
traversing S probes across the duct. Gas samples flowed continuously from the sampling tee to the gas analyzers.
Composition measurements of the flue gas were made by
continuously sampling the exhaust flow. The sample flowed
from a gas sampling tee, mounted inside the exhaust duct
(Figure 4), to a set of gas analyzers located in the facility
control room. The volume fraction of water vapor was mea-
sured with a thin film capacitive detector prior to drying the
sample. A portion of the dried sample was directed to a non-
dispersive infrared detector to measure the volume fraction of
CO2 and CO. Another portion of the dried sample was directed
to a paramagnetic analyzer to measure O2 volume fraction. The
water vapor measurement was used to convert the volume
fraction measurements to a wet basis.
The direct emissions measurement of carbon dioxide from
the exhaust duct or stack of a stationary source is mainly the
product of two measurements: the bulk flow of the flue gas and
the CO2 concentration. Direct emissions measurements of CO2
can be derived from the following expression, where the para-
meters are defined in Table 2.
πd 2 MCO2
m_ CO2 ;d ¼ Vexh ρexh XCO2 ;net;dry ð1  XH2 O;exh Þ (4)
4 Mexh
The net CO2 volume fraction (dry basis), XCO2 ;net;dry , is the
difference between the volume fraction measurements of the
exhaust gas during the fire and the ambient air; therefore, it is
the volume fraction of CO2 that is added to the air stream by
the natural gas fire. The estimates for relative expanded
 Bryant et al. / Journal of the Air & Waste Management Association 65 (2015) 863–870 867

Table 2. Example of an uncertainty budget for direct emissions measurements of CO2 in the exhaust duct of the LFRL

Nondimensional Percent
Relative standard sensitivity contribution,
Measurement component, xi Value uncertainty, u(xi)/xi coefficient, si %

Exhaust gas mean flow velocity, Vexh (m/sec) 20.91 0.0056 1.0 9.9
Exhaust duct diameter, d (m) 1.504 0.0079 2.0 77.6
Exhaust gas mean density, ρexh (kg/m3) 1.047 0.0034 1.0 3.6
CO2 net volume fraction—dry basis, XCO2 ;net;dry (m3/m3) 0.001819 0.0053 1.0 8.9
Exhaust gas H2O volume fraction, XH2 O;exh (m3/m3) 0.007947 0.0031 0.05 0
Exhaust gas molecular weight, Mexh (kg/kmol) 28.7734 0.0001 −1.0 0
CO2 molecular weight, MCO2 (kg/kmol) 44.0095 0.0000 1.0 0
Direct CO2 emissions, m_ CO2 ;d (g/sec) 107.3 0.0179 (0.0358) Standard (Expanded) Uncertainty
Note: The CO2 was generated using a 2-MW natural gas fire.

uncertainty of the direct emissions measurements of CO2 were

±0.042 or less. The largest contribution of uncertainty comes
from the duct diameter measurement. The effective diameter of
the duct was determined from the average of multiple length
measurements along the two chords. The lengths of the two
chords were in close agreement; therefore, a circle was chosen
to model the shape of the duct. In many cases, the cross-
sections of ducts and stacks are not perfectly circular but
elliptical. In these cases, other methods of measurement that
can accurately characterize the elliptical shapes should be
applied. The second largest contribution of uncertainty comes
from the mean flow velocity measurement. Methods to
improve the velocity traverse measurements have been demon-
strated and resulted in lowering the relative expanded uncer-
tainty estimates for CO2 emissions to ±0.036. A detailed
discussion of the velocity traverse measurements and their
uncertainty analysis is provided in a previous paper (Bryant
et al., 2014).

Results and Discussion

Emissions measurements comparison—CO2 mass
balance
Conservation of mass is the principle behind predicting CO2
emissions from fuel consumption data, where a known fraction
of the carbon atoms in the raw fuel are oxidized during com-
bustion to create CO2 molecules (EPA, 2008). Applying the
principle downstream of the combustion process, between the
inlet and outlet of a facility’s emissions control system, is also
a useful method of validating emissions measurements. For the
present study, the CO2 generated by the fire, the predicted
emissions, was captured by the exhaust hood and flowed
through the inlet of the exhaust duct (Figure 5). Assuming all
of the CO2 was captured by the hood, this predicted amount of
CO2 flowed through the exhaust duct unaltered. The exhaust
duct was under slight negative pressure, and the fire was the
only source of CO2. After sufficient mixing with the added
ambient air, measurements were conducted to determine the
Figure 5. Illustration of the CO2 mass balance as applied for the present study.
The direct emissions measurements in the exhaust duct are validated by the
predicted emissions measurements conducted prior to the exhaust duct.
amount of CO2 that would exit the exhaust duct, the direct CO2
emissions. Measurements of predicted and direct CO2 emis-
sions should agree, therefore confirming a CO2 mass balance
for the exhaust system.
To conduct an emissions comparison experiment, the natural
gas burner was ignited and the gas flow was adjusted to
achieve the desired set point value for heat output and CO2
injection rate. Continuous gas sampling measurements from the
868 Bryant et al. / Journal of the Air & Waste Management Association 65 (2015) 863–870
Figure 6. The predicted emissions measurements, CO2 output of the burner,
were repeatable to within ±1.0% (shown by dashed lines).
exhaust duct were performed for the duration of the experi-
ment. The burner and the exhaust system were allowed to reach
a pseudo-steady state before starting the velocity traverse mea-
surements to determine bulk flow. Figure 6 displays the data
for the predicted emissions measurements of CO2 with respect
to the burner set point. For two test series, the data were
repeatable to within ±1.0%, hence indicating a level of preci-
sion consistent with the combined uncertainty estimates for the
natural gas burner and fuel delivery system.
Direct measurements of CO2 emissions in the exhaust
duct compared well with emissions predicted from the bur-
ner. With the exception of a few data points, Figure 7
demonstrates that the relative difference between the paired
measurements is small enough for overlap of the uncertainty
estimates. The average relative difference is −0.024 and
demonstrates that the direct emissions measurements mostly
Figure 7. Relative difference between direct and predicted emissions measure-
ments of CO2 for natural gas fires. Error bars represent the expanded uncer-
tainty estimates for the direct emissions measurements, approximately ±4%.
The average relative difference is shown as the dashed gray line.
underestimate the predicted emissions measurements. Twice
the standard deviation of the relative difference is reported,
±0.067. This distribution is larger than, but of similar order,
as the uncertainty estimates for the direct measurements.
However, we expect that this distribution will decrease
with improvements in the precision of the direct emissions
measurements.
The agreement between predicted and direct emissions mea-
surements for a well-characterized fuel such as natural gas
provides greater confidence for the direct emissions measure-
ments of CO2 from other fuels. Predicted emissions measure-
ments for solid fuels such as coal or municipal solid waste are
difficult because determining the accurate carbon content of a
heterogeneous fuel requires great care and the carbon is not
always fully oxidized to CO2. In the future, the LFRL’s direct
emissions measurements along with added measurement cap-
abilities, such as solid fuel compositional analysis and volume
fraction measurements of flue gas particulates, can be used to
examine other fuel packages and confirm or improve their CO2
emission factors.
When a CO2 CEMS measurement is not available, it is also
acceptable to use an oxygen CEMS measurement to derive
CO2 concentration in the flue gas (U.S. National Archives
and Records Administration, 2014). Therefore, the oxygen
volume fraction measurements, XO2 ;amb;dry and XO2 ;exh;dry , can
be used to derive the net CO2 volume fraction applied in eq 4
and hence the CO2 emissions.
Fc XO2 ;amb;dry  XO2 ;exh;dry
XCO2 ;net;dry ¼ (5)
F XO2 ;amb;dry
The emissions factor, F, represents the theoretical dry volume
of combustion products generated per unit of heat content of
the fuel consumed. The emissions factor, Fc, represents the
theoretical volume of CO2 generated per unit of heat content
of the fuel consumed. Default values for both factors are
available for different hydrocarbon fuels. The default values
for natural gas are F = (2340 ± 30) × 10−4 m3/MJ (8710 ± 111
ft3/106 BTU) and Fc = (279 ± 6) × 10−4 m3/MJ (1040 ± 23 ft3/
106 BTU) (Shigehara et al., 1978; EPA, 2000; U.S. National
Archives and Records Administration, 2014). The results of the
direct emissions measurements of CO2 derived from oxygen
measurements also agree well with predicted emissions mea-
surements (Figure 8). The oxygen-derived emissions (dashed-
dotted line) underestimates the predicted emissions by similar
amounts when compared with the direct measurements for CO2
(dashed line). Estimates of relative expanded uncertainty for
the oxygen-derived CO2 emissions ranged from ±0.066 to
±0.070. The greater uncertainty is due to the contribution of
uncertainty from the default emissions factors.
Quality check of gas composition measurements
The results of Figure 7 and Figure 8 demonstrate that the direct
emissions measurements of CO2 in the exhaust duct agree with
stoichiometric predictions. This agreement can be further verified
by examining eq 5 in more detail. The equation can be rearranged
to represent the Fuel Factor, Fo, which is the ratio of the mole
 Bryant et al. / Journal of the Air & Waste Management Association 65 (2015) 863–870 869

from flue gas measurements (direct), Fo,d, agrees with the

default Fuel Factor for natural gas.
XO2 ;amb;dry  XO2 ;exh;dry
Fo;d ¼ (7)
XCO2 ;net;dry

This Fuel Factor was a very important parameter used to

quantitatively identify outliers for this study. If it was signifi-
cantly different from the default value, it indicated that the CO2
or O2 gas analyzers were malfunctioning or that the exhaust
flow was too low and combustion products accumulated under
the exhaust hood, to the point of spilling out into the labora-
tory. Either condition represented an outlier experiment that
was removed from the analysis.
In addition, the gas composition measurements from the nat-
ural gas supply can be used to predict a Fuel Factor for the natural
Figure 8. Relative difference between direct and predicted emissions measure- gas supplied to the burner, Fo,p. The natural gas supply was
ments of CO2 for the case of direct emissions derived from measurements of O2 composed mostly of alkanes (methane, ethane, propane, etc.)
volume fraction. Error bars represent the expanded uncertainty estimates for the and a small fraction of CO2. Assuming the CO2 acts only as an
direct emissions measurements, approximately ±7%. The average relative dif- inert, the general reaction for an alkane can be used to predict the
ference is shown as the gray dashed-dotted line, whereas the average from the stoichiometric mole fractions of O2 and CO2.
CO2 measurements, black dashed line, is shown for reference.
 0:7905
α þ β þ 3αþ1
Fo;p ¼  2 
0:2095
(8)
fraction (dry basis) of O2 consumed and the mole fraction (dry ðα þ βÞ 1 þ 0:7905
0:2095
basis) of CO2 produced for stoichiometric amounts of fuel and air.
where α and β are the number of moles of carbon atoms and
F XO ;amb;dry CO2 molecules, respectively, satisfying the general reaction for
Fo ¼ XO2 ;amb;dry ¼ 2 (6) an alkane:
Fc XCO2 ;net;dry
  
3α þ 1 0:7905
Default values for Fo can be derived from the emissions factors Cα H2αþ2 þ βCO2 þ O2 þ N2 !
F and Fc. The default value derived for natural gas is 1.755 ± 2 0:2095
  (9)
0.090. Published values range from 1.60 to 1.84 (Shigehara 3α þ 1 0:7905
et al., 1978; EPA, 2011). EPA recommends using the Fuel ðα þ βÞCO2 þ ðα þ 1ÞH2 O þ N2
2 0:2095
Factor as a data quality check for gas sampling measurements
of the flue gas (eq 7). Results for this data quality check are Results for the predicted Fuel Factor are shown in Figure 9
shown in Figure 9 and demonstrate that the Fuel Factor derived and also demonstrate good agreement with the default value
for natural gas. The predicted and direct measurements of Fo
also demonstrate good agreement, with the direct measure-
ments underestimating the predictions with an average rela-
tive difference of −0.010. The results of Figure 9 confirm the
quality of the gas composition measurements upstream of the
combustion process—at the fuel supply, and downstream of
the process—in the exhaust.

Figure 9. Data quality check of exhaust duct gas sampling measurements
(direct) and fuel supply gas composition measurements (predicted). Fuel
Factors computed from measurements are compared with the default value
for natural gas.
Conclusions
Fossil-fuel-burning stationary sources have accounted for
over half of all CO2 emissions in the United States and there-
fore play a significant role in the accuracy of greenhouse gas
reporting. Using a fire research facility as a near-industrial-
scale surrogate for a stationary source, two primary methods
for determining CO2 emissions, predicted emissions from fuel
consumption measurements and direct stack measurements,
have been compared.
A natural gas fire, issuing from a well-characterized burner
and gas supply system, served as a precision source of CO2.
870 Bryant et al. / Journal of the Air & Waste Management Association 65 (2015) 863–870
Predicted CO2 emissions, computed from the fuel consumption
measurements, were demonstrated with an expanded uncer-
tainty of ±1% or less. Direct measurements of CO2 emissions
from the exhaust duct were demonstrated with an expanded
uncertainty of ±4% or less. The relative difference between
pairs of predicted and direct emissions measurements was
generally less than the estimated measurement uncertainty,
therefore demonstrating good agreement. Fuel Factor values
computed from the gas composition measurements at the fuel
supply and in the exhaust duct compared well with the default
value for natural gas, hence confirming the quality of the
measurements.
This study demonstrates how the principle of the conserva-
tion of mass and independent measurement methods are used
to provide a cross-validation of CO2 emissions at a stationary
source. The study also introduces and demonstrates the concept
of a precision test-bed for the purpose of evaluating and
improving greenhouse gas emissions measurements from sta-
tionary sources.
Acknowledgment
The authors gratefully acknowledge the technical and engi-
neering support provided by Marco Fernandez, Laurean
DeLauter, Doris Rinehart, and Anthony Chakalis, data acquisi-
tion support provided by Artur Chernovsky, and data analysis
support provided by R. Paul Borthwick. We are also grateful
for the technical guidance provided by Anthony Hamins and
Jiann Yang. Research support by the NIST Office of Special
Programs—Greenhouse Gas and Climate Science Measure
ments, James Whetstone Program Manager—is gratefully
acknowledged.
References
Ackerman, K.V., and E.T. Sundquist. 2008. Comparison of two US power-plant
carbon dioxide emissions data sets. Environ. Sci. Technol. 42:5688–5693.
doi:10.1021/es800221q
Borthwick, R., and M. Bundy. Quantification of a precision point source for
generating carbon dioxide emissions. Paper presented at EPRI CEM User
Group Conference and Exhibit, Chicago, IL, June 8, 2011.
Bryant, R., O. Sanni, E. Moore, M. Bundy, and A. Johnson. 2014. An
uncertainty analysis of mean flow velocity measurements used to quantify
emissions from stationary sources. J. Air Waste Manage. Assoc. 64:679–
689. doi:10.1080/10962247.2014.881437
Bryant, R.A., T.J. Ohlemiller, E.L. Johnsson, A. Hamins, B.S. Grove, W.F.
Guthrie, A. Maranghides, and G.W. Mulholland. 2003. The NIST 3
Megawatt Quantitative Heat Release Rate Facility. NIST Special Publication
1007. Gaithersburg, MD: National Institute of Standards and Technology.
Bundy, M., A. Hamins, E.L. Johnsson, K.C. Sung, K. Gwon, and D.B. Lenhert.
2007. Measurements of Heat and Combustion Products in Reduced-Scale
Ventilation-Limited Compartment Fires. NIST Technical Note 1483.
Gaithersburg, MD: National Institute of Standards and Technology.
Quick, J.C. 2014. Carbon dioxide emission tallies for 210 U.S. coal-fired power
plants: A comparison of two accounting methods. J. Air Waste Manage.
Assoc. 64:73–79. doi:10.1080/10962247.2013.833146
Shigehara, R.T., R.M. Neulicht, W.S. Smith, and J.W. Peeler. 1978. Summary
of F Factor Methods for Determining Emissions from Combustion Sources.
EPA-450/2-78-042a. Washington, DC: U.S. Environmental Protection
Agency.
U.S. Environmental Protection Agency (EPA). 2000. Determination of Sulfur
Dioxide Removal Efficiency and Particulate Matter, Sulfur Dioxide, and
Nitrogen Oxide Emission Rates. EPA Method 19. Washington, DC: U.S.
Environmental Protection Agency.
U.S. Environmental Protection Agency (EPA). 2007. Determination of Stack
Gas Velocity and Volumetric Flow Rate with Two-Dimensional Probes. EPA
Method 2G. Washington, DC: U.S. Environmental Protection Agency.
U.S. Environmental Protection Agency (EPA). 2008. Direct Emissions from
Stationary Combustion Sources. EPA 430-K-08-003. Washington, DC: U.S.
Environmental Protection Agency.
U.S. Environmental Protection Agency (EPA). 2011. Gas Analysis for the
Determination of Emission Rate Correction Factor or Excess Air. EPA
Method 3B. Washington, DC: U.S. Environmental Protection Agency.
U.S. Environmental Protection Agency (EPA). 2013. Inventory of U.S.
Greenhouse Gas Emissions and Sinks: 1990–2011. EPA 430-R-13-001.
Washington, DC: U.S. Environmental Protection Agency.
U.S. National Archives and Records Administration. 2014. Code of Federal
Regulations: Title 40 Part 75, Continuous Emission Monitoring, Appendix F.
About the Authors
Rodney Bryant and Matthew Bundy are mechanical engineers at the National
Institute of Standards and Technology, Fire Research Division, in Gaithersburg,
MD.
Ruowen Zong is an associate professor at the State Key Laboratory of Fire
Science, University of Science and Technology of China, and was a visiting
guest researcher at the National Institute of Standards and Technology, during
the execution of this study.