Estimating and comparing greenhouse gas emissions with their uncertainties using different methods: A case study for an energy supply utility

Abstract

Energy supply utilities release significant amounts of greenhouse gases (GHGs) into the atmosphere. It is essential to accurately estimate GHG emissions with their uncertainties, for reducing GHG emissions and mitigating climate change. GHG emissions can be calculated by an activity-based method (i.e., fuel consumption) and continuous emission measurement (CEM). In this study, GHG emissions such as CO₂, CH₄, and N₂O are estimated for a heat generation utility, which uses bituminous coal as fuel, by applying both the activity-based method and CEM. CO₂ emissions by the activity-based method are 12–19% less than that by the CEM, while N₂O and CH₄ emissions by the activity-based method are two orders of magnitude and 60% less than those by the CEM, respectively. Comparing GHG emissions (as CO₂ equivalent) from both methods, total GHG emissions by the activity-based methods are 12–27% lower than that by the CEM, as CO₂ and N₂O emissions are lower than those by the CEM. Results from uncertainty estimation show that uncertainties in the GHG emissions by the activity-based methods range from 3.4% to about 20%, from 67% to 900%, and from about 70% to about 200% for CO₂, N₂O, and CH₄, respectively, while uncertainties in the GHG emissions by the CEM range from 4% to 4.5%. For the activity-based methods, an uncertainty in the Intergovernmental Panel on Climate Change (IPCC) default net calorific value (NCV) is the major uncertainty contributor to CO₂ emissions, while an uncertainty in the IPCC default emission factor is the major uncertainty contributor to CH₄ and N₂O emissions. For the CEM, an uncertainty in volumetric flow measurement, especially for the distribution of the volumetric flow rate in a stack, is the major uncertainty contributor to all GHG emissions, while uncertainties in concentration measurements contribute a little to uncertainties in the GHG emissions.

Implications:

Energy supply utilities contribute a significant portion of the global greenhouse gas (GHG) emissions. It is important to accurately estimate GHG emissions with their uncertainties for reducing GHG emissions and mitigating climate change. GHG emissions can be estimated by an activity-based method and by continuous emission measurement (CEM), yet little study has been done to calculate GHG emissions with uncertainty analysis. This study estimates GHG emissions and their uncertainties, and also identifies major uncertainty contributors for each method.

Introduction

Human activities, such as fossil fuel combustion and land usage changes, increase greenhouse gas (GHG) concentrations in the atmosphere. The increased GHG concentrations result in increasing average global temperature by absorbing the surface radiation (Le Treut et al., 2007). In 1997, under the United Nations Framework Convention on Climate Change (UNFCCC), the Kyoto Protocol was established to set legally binding obligations for developed countries (Annex I countries) to reduce GHG emissions to 5% below the 1990 levels (Barker et al., 2007). In 2010, the Cancun Agreement, mainly based on the Copenhagen Accord, was reached to limit the increase in average global temperature not more than 2°C relative to the preindustrial temperature by reducing GHG emissions significantly (UNFCCC, 2011). Under the UNFCCC, the Annex I countries should submit annual national inventory reports for GHG emissions. The inventories are useful not only for policymakers to develop effective strategies to reduce GHG emissions, but also for scientists to understand the sources and trends in GHG emissions. To support the UNFCCC, the Intergovernmental Panel on Climate Change (IPCC) has developed methodologies and guidelines for preparing national inventories of GHG emissions (IPCC, 1997, 2006). The 2006 guidelines consists of five volumes that describe the general guidance for developing emission inventories and the more specific guidance for estimating GHG emissions in different sectors such as energy, industrial processes and product use, agriculture/forestry/other land use, and waste from economic activities. The energy supply (e.g., stationary combustion) in the energy sector is an important source of GHG emissions as it contributes 25.9% of the global GHG emissions (Barker et al., 2007). Therefore, it is essential to properly estimate GHG emissions from the energy supply for mitigating climate change.

GHG emissions can be estimated by an activity-based (fuel consumption-based) method and by continuous emission measurement (CEM). The activity-based method estimates GHG emissions by using fuel consumption and emission factors, while the CEM directly measures GHG emissions through monitoring GHG concentrations and volumetric flow rate at a stack. In addition to the emission estimates, proper uncertainty estimation for the GHG emissions is also important for policymakers to make decisions for mitigating GHG emissions and for scientists to understand climate change. The IPCC has developed good practical guidelines for estimating and reporting uncertainties in the GHG emissions (IPCC, 2000, 2006). There are previous studies that estimate CO₂ emissions and their uncertainties for directly comparing those from both methods for coal-fired energy utilities (Ackerman and Sundquist, 2008; Evans et al., 2009; Quick, 2014).

In this study, GHG emissions (CO₂, CH₄, N₂O) from an energy supply utility fueled by coal were estimated by both activity-based and CEM methods. In addition, associated uncertainties in the GHG emissions were calculated by propagating the uncertainties of input variables for both methods. Based on results, important uncertainty sources were suggested to reduce uncertainties in the GHG emissions from energy supply utilities for improving the quality of GHG emission inventories.

Experiments and Methods

Fuel consumption estimation

A heat production plant that used bituminous coal as fuel was selected for this study. Monthly fuel consumption for August 2011 was estimated by using fuel mass that was carried into a plant storage tank and fuel mass that was left in the storage tank. The monthly fuel consumption was calculated as

1

where FM_i is the fuel consumption at i month (kg), INFM_i is the amount of fuel that is carried into a plant storage tank (kg) at i month, and LFM_i is the amount of fuel that is left in a plant storage tank (kg) at i month.

Flow measurement

A flow velocity inside a stack was measured by a flow rate measurement system (FTV-05, Tech Korea Inc., Korea), every 3 sec from August 1 to August 31, 2011. A volumetric flow rate was estimated as

2

where Q is the dry gas volumetric flow rate (m³/s), V is the flow velocity (m/s), D is the stack diameter (m), P is the static pressure (mmHg), T is the emission gas temperature (K), and X_w is the water content of emission gas.

The flow rate measurement system was connected to the national tele-metering system (TMS) for monitoring stack emissions, reporting 5-minute accumulated volumetric flow to the national TMS. The 5-minute accumulated volumetric flow was estimated as

3

where Q_5min is a 5-min accumulated volumetric flow (m³) and is a 5-min average flow velocity (m/sec). The 5-min accumulated volumetric flow was used to calculate the greenhouse gas emissions for the CEM method. Another flow measurement system (VOT-03 PS, Dooil Tech, Korea) was used to estimate distribution of flow rate across a stack, since manipulating the flow measurement system, which was connected to the national TMS, was not allowed during the study period.

Gas concentration measurement

CO₂, N₂O, and CH₄ were continuously measured by ULTRAMAT6 gas analyzers (Siemens, Munich, Germany) every 10 sec from August 1 to August 31, 2011. A sampling probe was installed at 0.5 m away from the center of an emission stack, and a sampling line was heated to prevent from condensing water vapor. Sampled emission gas was cooled at 4°C to remove water vapor prior to the gas analyzer. Prior to the continuous measurement, the distribution of CO₂ concentrations inside the stack (2.5 m as diameter) was determined by measuring the concentrations at seven points across the stack (one point at the center, two points at 0.5 m away from the center, two points at 0.9 m away from the center, two points 1.1 m away from the center). The maximum deviation from the mean concentrations was 0.3%. The results indicated that the CO₂ concentrations were not stratified across the stack, since the maximum deviation was less than 5% (U.S. Environmental Protection Agency [EPA], 1997). The sampling probe was installed at 0.5 m away from the center of the stack where the deviation from the mean CO₂ concentrations was 0.09%. The gas analyzer was calibrated against the primary standard gas mixtures (PSMs), which are traceable to Korea Research Institute of Standards and Science (KRISS), before and during the measurement. The 5-min averaged concentrations were calculated and then used to estimate GHG emissions for the CEM.

Emission estimation based on activity data

GHG emissions during the study period (from August 1 to August 31, 2011) were estimated based on activity data (e.g., fuel consumption, net calorific value), emission factor, and carbon content of fuel according to the 2006 IPCC guidelines (IPCC, 2006). Emission estimates, by using emission factor (EF) and net calorific value (NCV), can be expressed as

4

where E is the estimated emission (kg), FC is the fuel consumption (TJ), FM is the fuel mass (kg), NCV is the net calorific value (TJ/kg), and EF is an emission factor (kg/TJ).

By using carbon content (CC) and NCV, emission estimates can be expressed as

5

where E is estimated emission (kg), FM is fuel mass (kg), CC is the fuel carbon content (kg/kg), M_CO2 is the molecular mass of carbon dioxide, and MC is the atomic mass of carbon. Note that the CO₂ emission was slightly overestimated and its uncertainty was underestimated as the combustion efficiency (i.e., oxidation factor) was assumed as 1 in this study.

GHG emissions were estimated for four different cases: IPCC default EF (Tier 1) and NCV were used in case 1, IPCC default EF (Tier 1) and plant-specific NCV were used in case 2, plant-specific CC and NCV were used in case 3, and plant-specific CC for CO₂, technology-specific IPCC default EF for N₂O and CH₄, and plant specific NCV were used in case 4. For case 4, the technology-specific EFs (default Tier 3 EFs) for N₂O and CH₄ were used since they may be more realistic values compared to the default Tier 1 EF based on only fuel.

Emission estimation based on continuous emission measurement (CEM)

Unlike the activity-based method, the concentrations and flow rate of gas emissions at a stack were measured continuously by CEM during the study period. GHG emissions were estimated as

6

where E is the estimated emission (kg), E_5min,i is the 5-min accumulated emission of the ith measurement (kg), x_5min,i is the 5-min averaged concentration of the ith measurement (% or ppm), Q_5min,i is the 5-min accumulated volumetric flow of the ith measurement (m³), M_gas is the molar mass of an emission gas, MV is the molar volume of an ideal gas (22.414 L at 273.15 K and 760 mm Hg), and N is the total number of 5-min estimated emissions (N = 7693).

Uncertainty Estimation

Uncertainty components of GHG emission estimation for each method were identified, calculated, and then combined according to the Guide to the Expression of Uncertainty in Measurement (BIPM et al., 1995). In a companion article, more statistical methods (e.g., bootstrap, Monte Carlo method, etc.) were discussed for combining uncertainties with different probability distributions. Therefore, the statistical methods were not discussed in this study.

Uncertainty for fuel consumption

The model equation for the fuel consumption estimation can be expressed as

7

The combined standard uncertainty for the fuel consumption can be estimated as

8

The uncertainty of an amount of fuel carried into a plant storage, u(INFM_i), was estimated based on the calibration results of scales that were used to weigh a truck. The relative maximum deviation was 0.5%. Assuming a rectangular distribution, the relative standard uncertainty for the scales was 0.14%. The uncertainty of an amount of fuel left in the plant storage, u(LFM_i), was estimated based on the quality control and assurance reports of the plant. The relative maximum deviation was ±5%, which is the same value reported by DeSollar (2011) and Quick (2014). Assuming a rectangular distribution, the relative standard uncertainty for the amount of fuel left was 2.89%. The combined relative standard uncertainty for the studied month was calculated as 0.89%. The uncertainty of the plant-specific NCV (2%) and CC (2.9%) was estimated based on coal analysis reports for bituminous coal used during the study period. Quick (2014) has reported that the precision of heating value measurements varies from 0.8% to 1.7%.

Uncertainty for flow measurement

The model equation for the continuous flow measurement can be expressed as

9

where Q is the dry gas flow rate (m³), V is the flow velocity (m/s), C_p is the S-type Pitot tube coefficient, ΔP is the dynamic pressure (Pa), ρ is the density of emission gas (kg/m³), D is the stack diameter (mm), P is the static pressure (mm Hg), T is the emission gas temperature (K), and X_w is the water content of the emission gas. A typical S-type Pitot tube coefficient (C_p) is known as 0.85. Both a specific coefficient and its associated uncertainty, u(C_p), for each Pitot tube need to be determined by national metrology institute or accredited calibration laboratories when a Pitot tube is installed on a stack. For this study, the coefficient and its associated uncertainty was taken from the annual quality control and assurance report of the Pitot tube installed on the stack. The uncertainty for a dynamic pressure, u(ΔP), was determined by repeated pressure measurements with every 3 sec for 5 min and the uncertainty from annual variation of linearity results in the performance test of a Pitot tube and manometer. A density of emission gas (ρ) was determined by estimating a weighted average density based on the concentrations of major gas components (N₂, CO₂, O₂, Ar, H₂O) in the emission gas from the stack. The associated uncertainty of the estimated density was calculated by combining uncertainties in gas concentration, temperature, and pressure measurement. The uncertainty for density (ρ) of emission gas was determined by combining uncertainty in the measurement repeatability and the difference between measuring values and used values in the energy supply utility. The accuracy for the stack diameter was assumed as ±10 mm from manufactured technical specification and then its relative standard uncertainty became 0.23%, assuming a rectangular distribution. The uncertainty in the accuracy of a static pressure sensor was assumed as ±1 mm Hg (0.13% as relative standard uncertainty). The uncertainty for the repeatability of the static pressure sensor was determined by repeated pressure measurement at every 3 sec for 5 min. The uncertainty in the accuracy of a thermocouple was assumed as ±1 K. The uncertainty for the repeatability of the thermocouple was determined by repeated temperature measurement at every 3 sec for 5 min. The combined uncertainty for the static pressure and temperature measurement was equal to the positive square root of sum of square uncertainty of uncertainty components. A water content of emission gas was determined by gravimetrically measuring water mass out of a cold condenser prior to the gas analyzer. A portable flow velocity meter with S-type Pitot tube (VOT-03 PS, Dooil Tech, Korea) was used to determine the distribution of the flow velocity in the stack. There was a space limit to measure the flow velocity across the entire stack diameter. The distribution of the flow velocity (V_D) was measured from the center to only one end wall of the stack, instead of both ends. The uncertainty, u(V_D), due to velocity distribution in the cross section of the stack was determined by standard deviation of measured velocity distribution profile because only average velocity value in a certain position was used to calculate the flow velocity. Results showed that the uncertainty due to the distribution of the flow velocity was significant. The uncertainty budget for the flow measurement is summarized in Table 1. The combined standard uncertainty for flow measurement can be expressed as

Table 1. Uncertainty budget for flow rate measurement

Component(y)	Source	Distribution	Standard uncertainty	Combined standard uncertainty, u(y)	Sensitivity coefficient,	Contribution to u(Q),
Cp	Calibration	Rectangular	0.0045	0.0045	Q/Cp	5.5 × 10^−3Q
ΔP, Pa	Calibration	Rectangular	1.487	1.844	Q/2ΔP	6.8 × 10^−3Q
	Repeatability	Normal	1.097
ρ, kg/m^3	Deviation from a measured value	Rectangular	7.2 × 10^−5	0.0177	−Q/2 ρ	5.3 × 10^−3Q
	Concentration, temperature, pressure	Normal	0.0177
ΔVD, m/s	Variability	Normal	0.2233	0.2233	Q/ΔVD	1.5 × 10^−2Q
D, mm	Accuracy	Rectangular	5.774	5.774	2Q/D	4.6 × 10^−3Q
P, mmHg	Accuracy	Rectangular	1	1.0001	Q/P	1.3 × 10^−3Q
	Repeatability	Normal	0.0144
T, K	Accuracy	Rectangular	1	1.0002	−Q/T	2.4 × 10^−3Q
	Repeatability	Normal	0.0196
1-Xw, %	Accuracy	Normal	0.0015	0.2745	−Q/(1−Xw)	3.0 × 10^−3Q
	Deviation from a measured value	Rectangular	0.2745

Notes: The uncertainty distribution of each component was assumed as rectangular when its uncertainty was estimated based on expert judgment (B type uncertainty). The uncertainty distribution of each component was as normal when its uncertainty was estimated based on experimental measurements.

10

The relative expanded uncertainty for the flow measurement was estimated as 3.9% for a coverage factor (k = 2) at approximately 95% confidence level. The estimated uncertainty is similar to estimates (less than 5%) in previous studies (Evan et al., 2009; Boze, 2010).

Uncertainty for gas concentration measurement

The model equation for the continuous measurement of gaseous air pollutants can be expressed as

11

where x is the measured concentration, x_m is the uncorrected measured concentration, and x₀ is the zero offset correction. The model equation can be expanded as

12

where x_s is the span gas concentration (i.e., PSMs), R_s is the response of a measuring instrument when the span gas is introduced into the measuring instrument (repeatability), R is the response corresponding to an arbitrary concentration, f_{cal drift} is the fractional calibration drift that is the ratio of the response (R_s) of the measuring instrument corresponding to the span gas concentration (x_s) within a calibration period (calibration drift), f_linearity is the fractional linearity of the measuring instrument (linearity), and Δx₀ is the offset of the measuring instrument corresponding to zero gas within a calibration period (zero offset correction).

The estimate of x_s was a certified concentration (e.g., concentration reported in certificates of PSMs) for the span gas. The standard uncertainty, u(x_s), of the span gas was determined by dividing the reported expanded uncertainty of the span gas by a coverage factor, k:

13

where U(x_s) is the expanded uncertainty of the span gas, and k = 2.0 for approximately 95% confidence level.

For repeatability, a series of responses of the measuring instrument was observed by applying the span gas into the instrument. The arithmetic mean and experimental standard deviation was calculated for the n observations. The arithmetic mean was taken as the input estimate of R_s. The experimental standard deviation divided by a square root of the number (n = 20) of the observations, which was the experimental standard deviation of the mean of the observations, was the standard uncertainty for R_s:

14

where n is the number of R_s measurements.

The response (R) to an arbitrary concentration (x) varies as the measurand concentration changes. The standard uncertainty for R was assumed the same as the standard uncertainty of R_s:

15

The fractional calibration drift (f_{cal drift}) and zero offset (Δx₀) were determined through applying span and zero gas into the instrument, respectively. The standard uncertainty for the fractional calibration drift was equal to the experimental standard deviation of multiple estimates of the fraction calibration drift when the span gas was applied into the instrument:

16

where f_{cal drift} = R_s/x_s and n is a number of calibration tests (n = 4).

The standard uncertainty of the zero offset correction was estimated as the experimental standard deviation of multiple observations of the zero offset when the zero gas was applied into the instrument:

17

where n is the number of the zero offset tests (n = 4).

The fractional linearity (f_linearity) of the measuring instrument was determined by applying linearity gas (e.g., about a half concentration of the span gas) into the instrument. The standard uncertainty of the fractional linearity was estimated as the experimental standard deviation of multiple estimates of the fractional linearity:

18

where n is the number of the linearity tests (n = 3).

The combined standard uncertainty was estimated by combining the standard uncertainties of all uncertainty components in the measurement. The combined standard uncertainty can be estimated as

19

The uncertainty budget for each gas measurement was summarized in Table 2 and Figure S1 (Supplemental Materials). The standard uncertainty for each gas can be expressed as

Table 2. Uncertainty budget for GHG concentration measurements

				CO2		N2O		CH4
Component(y)	Source	Distribution	Sensitivity coefficient,	Standard uncertainty, u(y)	Contribution to u(x), |ci| × u(y), %	Standard uncertainty, u(y)	Contribution to u(x), |ci| × u(y), ppm	Standard uncertainty, u(y)	Contribution to u(x), |ci| × u(y),ppm
xs	Accuracy	Normal	x/xs	4.52 × 10^−3, %	2.50 × 10^−4x	4.01 × 10^−2, ppm	5.00 × 10^−4x	1.99 × 10^−2, ppm	5.00 × 10^−4x
Rs	Repeatability	Normal	−x/Rs	1.27 × 10^−3, %	6.99 × 10^−5x	5.48 × 10^−3, ppm	6.83 × 10^−5x	5.84 × 10^−3, ppm	1.47 × 10^−5x
R	Repeatability	Normal	x/R	1.27 × 10^−3, %	6.99 × 10^−5x	5.48 × 10^−3, ppm	6.83 × 10^−5x	5.84 × 10^−3, ppm	1.47 × 10^−5x
fcal drift	Drift	Normal	x/fcal drifit	3.54 × 10^−3	3.54 × 10^−3x	4.19 × 10^−3	4.19 × 10^−3x	6.56 × 10^−3	5.66 × 10^−3x
flin	Drift	Normal	x/flin	3.81 × 10^−3, %	3.81 × 10^−3x	2.51 × 10^−3	2.51 × 10^−3x	2.51 × 10^−3	2.51 × 10^−3x
Δx0	Drift	Normal	1	1.83 × 10^−2, %	1.83 × 10^−2	4.99 × 10^−2, ppm	4.99 × 10^−2	7.66 × 10^−2, ppm	7.66 × 10^−2

Notes: The uncertainty distribution of each component was assumed as normal as its uncertainty was estimated based on experimental measurements. The distribution of the uncertainty of the span gas was based on its certificate in which the distribution was specified as normal distribution.

20

21

22

The uncertainty equations indicate that the uncertainties in GHG concentration measurements vary as the measured GHG concentrations change. The contribution of each uncertainty component also changes as the measured GHG concentrations changes, shown in Table 2. The uncertainty contributions of the zero offsets are expressed as absolute values, while those of other uncertainty components are expressed as values relative to the measured GHG concentrations. As shown in Figure S1, the uncertainties in the zero offset corrections become the major uncertainty contributors for a lower measurement range of the gas analyzers (for less than 30% of a full range), whereas the uncertainties in the calibration drift and linearity become the major uncertainty contributors for a higher measurement range (for higher than 30% of a full measurement range). The relative uncertainties in GHG concentrations become larger, especially for the lower measurement range as the measured concentrations decrease. Therefore, for reducing the measurement uncertainty, it is a good practice to set the measurement range of gas analyzers properly so that measured GHG concentrations are not less than 30% of a full measurement range. As shown in Figure S1, typical concentration measurement uncertainties are from about 1% to 1.5%, while previous studies have reported less than 1% (Evans et al., 2009; Boze, 2010) and 7–12% (Quick, 2014).

Uncertainty for incomplete time coverage

Incomplete coverage of time period results in an additional uncertainty to the time average when there is a missing data for the time period in which the time average of a measurand (X) is calculated (ISO, 2002). The uncertainty due to the incomplete time coverage (or missing data) can be determined as

23

where is the uncertainty due to the incomplete time coverage, N is the number of the measurand (X), N_max is the maximum number (30) of the measurand (X) in a time period, and s(X_i) is the standard deviation of the measuerand values. The uncertainty due to the incomplete time coverage was combined with the uncertainty from concentration measurements when N is less than N_max.

Uncertainty for emission estimates

The standard uncertainties of emission estimates based on the activity-based method were calculated as

24

where u(FM) is the standard uncertainty of the fuel mass, u(NCV) is the standard uncertainty of the net calorific value, and u(EF) is the standard uncertainty of the emission factor.

The standard uncertainty of CO₂ emission estimates based on fuel carbon content was calculated as

25

where u(CC) is the standard uncertainty of a fuel carbon content, u(M_CO2) is the standard uncertainty of the molar mass of CO₂, and u(M_C) is the standard uncertainty of the atomic mass of carbon. The uncertainties for carbon dioxide and carbon are negligible compared to those in fuel mass and carbon content.

The standard uncertainties of emission estimates (for 5 min) based on CEM were calculated as

26

where u(x_5min) is the standard uncertainty of 5-min averaged concentration, u(Q_5min) is the standard uncertainty of 5-min accumulated volumetric flow, M_gas is the standard uncertainty of the molar mass of a gas, and u(MV) is the standard uncertainty of the molar volume of an ideal gas.

The standard uncertainties of emission estimates (for a month) based on the CEM were calculated as

27

where u(E_{5min, i}) is a standard uncertainty of emission estimate for 5 min and N is the number of data points (N = 7693). All emission estimates for every 5 min were correlated as the same measuring instruments were used. The uncertainties of monthly-accumulated emission estimates became the sum of the standard uncertainties of the 5-min emission estimates instead of the root square sum.

Results and Discussion

Uncertainties of input parameters for GHG emission estimates are summarized in Table 3. The emission estimates with their uncertainties for all cases are summarized in Table 4. For case 1, all values with their uncertainties are default ones from the 2006 IPCC guidelines (IPCC, 2006) except the fuel mass consumed. The major uncertainty contributor for CO₂ emission is the default NCV. For case 2, the same IPCC default EFs used in the case 1 are used for emission estimation while a plant-specific NCV, which is used to convert mass-based fuel consumption to energy-based fuel consumption, is determined from fuel analysis. The plant-specific NCV, 23.7 TJ/Gg, is within the uncertainty range of the IPCC default NCV (19.9–30.5 TJ/Gg). However, its uncertainty (±2.0%) is much lower than that of the IPCC default NCV, which results in a considerably reduced uncertainty (±6.0%) for the CO₂ emission estimate compared to case 1. The major uncertainty contributor for the CO₂ emission estimate is the IPCC default EF (±5.4%) due to the significantly reduced uncertainty in the plant specific NCV. For case 3 and case 4, CO₂ emission is determined using a plant-specific CC (0.62 kg/kg) determined from fuel analysis. For a comparison, the mass-based CC (0.62 kg/kg) can be converted to an energy-based CC (26.3 kg/GJ) by using the plant-specific NCV. The energy-based plant specific CC agrees with the IPCC default CC (24.4–27.2 kg/GJ), and its uncertainty (±3.9%) is lower than that of the IPCC default (±5.4%). However, the uncertainty of the energy-based CC is larger than that of the mass-based CC (±2.9%), which is used to calculate CO₂ emissions in this study, as the uncertainty in the plant specific NCV is combined with that of the mass-based CC for the energy-based CC. The smaller uncertainties of the plant specific NCV (case 2) and mass-based CC (case 3 and 4) results in significantly reduced uncertainties (±6.0% and ±3.4%, respectively) in CO₂ emission estimates compared to the case 1 (larger than ±19%). For N₂O and CH₄ emissions, uncertainties in the EFs (larger than ± 67% and up to 900%) are the largest uncertainty contributors for all cases, as the uncertainties are much larger than those of fuel mass and NCVs. For the CEM, the volumetric flow with its uncertainty of ±3.9% is the major uncertainty contributor for all GHG emission estimates with uncertainties of ±4.0–4.5%.

Table 3. Uncertainties of input parameters for GHG emission estimation

	Component	Values	Relative 95% lower limit, %	Relative 95% upper limit, %
Case 1	FM (kg)	—	−1.8	+1.8
	NCV (Tier 1 default, TJ/Gg)	25.8	−22.9	+18.2
	CO2 EF (Tier 1 default, kg/TJ)	94,600	−5.4	+5.4
	N2O EF (Tier 1 default, kg/TJ)	1.5	−67	+233
	CH4 EF (Tier 1 default, kg/TJ)	1	−70	+200
^aCase 2	NCV (plant specific, TJ/Gg)	23.7	−2.0	+2.0
^bCase 3	CC (plant specific, kg/kg)	0.62	−2.9	+2.9
^cCase 4	N2O EF (Tier 3 default, kg/TJ)	61	−90	+900
	CH4 EF (Tier 3 default, kg/TJ)	1	−100	+100
^dCEM	5 min accumulated flow volume (m^3)	—	−3.9	+3.9
	CO2 concentration (%)	—	−1.1	+1.1
	N2O concentration (ppm)	—	−1.0	+1.0
	CH4 concentration (ppm)	—	−2.1	+2.1

Notes: ^aThe EF values for the three gases are the same as those used in case 1, but the NCV value is a plant specific one for this study.

^bThe NCV value is the same as used in case 2. The CO₂ emission is based on a plant specific CC for this study. The EF values for N₂O and CH₄ are the same as those used in case 1.

^cTHE NCV and CC values are the same as those used in cases 2 and 3. The EF values for N₂O and CH₄ are the IPCC Tier 3 default utility source specific values.

^dUncertainties for concentration measurements are typical values for this study.

Table 4. Estimated emissions (as CO₂ equivalent) with uncertainties

	Gas species	Estimated emissions	Relative 95% lower limit, %	Relative 95% upper limit, %
Case 1	CO2	0.79	−24	+19
	N2O	3.75 × 10^−3	−71	+234
	CH4	2.10 × 10^−4	−74	+201
	Total	0.80	−23	+19
Case 2	CO2	0.73	−6.0	+6.0
	N2O	3.44 × 10^−3	−67	+233
	CH4	1.93 × 10^−4	−70	+200
	Total	0.73	−6.0	+6.1
Case 3	CO2	0.74	−3.4	+3.4
	N2O	3.44 × 10^−3	−67	+233
	CH4	1.93 × 10^−4	−70	+200
	Total	0.75	−3.4	+3.6
Case 4	CO2	0.74	−3.4	+3.4
	N2O	0.14	−90	+900
	CH4	1.93 × 10^−4	−100	+100
	Total	0.88	−15	+143
CEM	CO2	0.90	−4.0	+4.0
	N2O	0.10	−4.0	+4.0
	CH4	5.26 × 10^−4	−4.5	+4.5
	Total	1.00	−3.7	+3.7

Note: Estimated emissions (CO₂ equivalent) were normalized to the total CEM emission.

To compare total GHG emission estimates for all cases, the estimated N₂O and CH₄ emissions are converted to CO₂ equivalent emissions by multiplying with a gas-specific global warming potential (100-year time horizon), 298 and 25, respectively. The emission estimates by the CEM are larger than those by the activity-based method for all cases except N₂O emission for the case 4 (Figure 1). CO₂ emission estimates by the activity-based method are 12–19% less than those by the CEM. The differences in the emission estimates by two different methods are similar with those reported in previous studies (Ackerman and Sundquist, 2008; Evans et al, 2009). N₂O emission estimates by the activity-based method (cases 1–3) are two orders of magnitude less than that by the CEM. However, the N₂O emission from case 4 (based on default technology-specific EF, Tier 3 default EF) is 40% higher than that by the CEM, though the estimate agrees with that by the CEM within the associated uncertainty. The Tier 3 default EF, which is a factor of about 41 higher EF than the Tier 1 default EF, results in the higher emission, but the associated uncertainty is the largest among all emission estimates (from –90% to +900%), indicating that more plant-specific EFs should be developed to reduce the uncertainty. Total GHG emissions based on the activity-based method are about 12–27% lower than that by the CEM, which is mainly due to the lower CO₂ and N₂O emissions by the activity-based method. The total GHG emissions by the activity-based method agree with each other within their associated uncertainty ranges, while that by the CEM agrees with only those from case 1 and case 4.

Figure 1. Comparisons of GHG emissions estimates by different methods.

Summary

In this study, GHG emissions have been estimated for a heat generation utility, which uses bituminous coal as fuel, by applying both the activity-based and CEM method. In addition, associated uncertainties in the GHG emissions have been calculated by propagating the uncertainties of input variables for both methods. Based on results, important uncertainty sources have been identified to reduce uncertainties in the GHG emissions from energy supply utilities for improving the quality of GHG emission inventories. Results show that CO₂ emissions by the activity-based method are 12–19% less than that by the CEM, while N₂O and CH₄ emissions by the activity-based method are two orders of magnitude and 60% less than those by the CEM, respectively. Comparing GHG emissions (as CO₂ equivalent) from both methods, total GHG emissions by the activity-based methods are 12–27% lower than that by the CEM, as CO₂ and N₂O emissions are lower than those by the CEM. Results from the uncertainty estimation show that the major uncertainty contributors to the emission estimates are net calorific values (NCVs) and emission factors (EFs) for the activity-based method when the IPCC default values are used for the estimation. For the CEM, the major uncertainty contributor to the emission estimates is the volumetric flow, especially the distribution of flow rate in a stack. Although the uncertainties in GHG concentration measurements are minor uncertainty contributors for the GHG emission estimates in this study, it is important to properly calibrate gas analyzers and set an appropriate measurement range for the analyzers. To get more certain GHG emission estimates, it is a good practice to identify uncertainty contributors and quantify their uncertainties, whichever method is chosen for GHG emission estimation. Based on results from the uncertainty estimation, one can prioritize necessary work to improve the quality of input parameters in order to meet target (or required) uncertainties for GHG emission estimation.

Supplemental Material

Supplemental data for this article can be accessed at http://dx.doi.org/10.1080/10962247.2014.930078.

Funding

This work was supported by grants from the Korea Research Institute of Standards and Science (KRISS) under the Basic R&D Project of Establishment of National Gas Analysis Measurement Standards and Improvements of Calibration and Measurement Capabilities (Grant 13011020), and by grants from the Korea Research Council of Fundamental Science and Technology (KRCF) under the National Agenda Project of Development of Measurement Technology for Solving Climate Change (NAP-08-2).

Additional information

Notes on contributors

Sangil Lee

Sangil Lee and Jinchun Woo are principal scientists, and Jinsang Jung is a senior scientist with the Center for Gas Analysis in Korea Research Institute of Standards and Science.

Yongmoon Choi

Yongmoon Choi is a principal scientist and Woong Kang is a senior scientist with the Center for Fluid Flow and Acoustics in Korea Research Institute of Standards and Science, Daejeon, Republic of Korea.

Jinchun Woo

Sangil Lee and Jinchun Woo are principal scientists, and Jinsang Jung is a senior scientist with the Center for Gas Analysis in Korea Research Institute of Standards and Science.

Woong Kang

Yongmoon Choi is a principal scientist and Woong Kang is a senior scientist with the Center for Fluid Flow and Acoustics in Korea Research Institute of Standards and Science, Daejeon, Republic of Korea.

Jinsang Jung

Sangil Lee and Jinchun Woo are principal scientists, and Jinsang Jung is a senior scientist with the Center for Gas Analysis in Korea Research Institute of Standards and Science.