Uncertainty in Life Cycle Greenhouse Gas Emissions from United States Natural. Gas and its Effects on Policy SUPPORTING INFORMATION

Size: px
Start display at page:

Download "Uncertainty in Life Cycle Greenhouse Gas Emissions from United States Natural. Gas and its Effects on Policy SUPPORTING INFORMATION"

Transcription

1 Uncertainty in Life Cycle Greenhouse Gas Emissions from United States Natural Gas and its Effects on Policy SUPPORTING INFORMATION Aranya Venkatesh, Paulina Jaramillo, W. Michael Griffin, H. Scott Matthews Civil and Environmental Engineering Department, Department of Engineering and Public Policy, and Tepper School of Business, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, Pennsylvania Number of pages: 21 Number of tables: 6 Number of equations: 2 Number of figures: 6 S1

2 The Supporting Information contains detailed explanations of data, methods and modeling parameters used in the study. A higher heating value of 38 MJ/m 3 (1030 BTU/cf) for natural gas was assumed, as given by Jaramillo et al. (1). The life cycle system boundary and methods used are presented in Figure S 1. Figure S 1. Summary of domestic natural gas and LNG life cycle boundary and methods Domestic Natural Gas 10'2"&+',% 10'&#**.,5% 304,*(.**.',% A',/#%F40$'% *.("$4+',% -.*/0.)"+',%!"#$%&'()"*+',% Imported LNG 10'2"&+',% =.>"#?4&+',% 10')4).$./D% (.E/"0#% 54*%&',*"(#2%.,%/;#%C:% :;.<<.,5% 8#54*.9&4+',% A',/#%F40$'% *.("$4+',% 304,*(.**.',6% -.*/0.)"+',%7%!"#$%&'()"*+',% DOMESTIC NATURAL GAS Production The U.S. Energy Information Administration (EIA) reports the quantity of natural gas used as lease fuel by gas wells, fields and lease operations during the production of natural gas by state (2). They also report the dry production of natural gas by state (3), which was used to normalize the lease fuel consumed to obtain the volume of lease fuel required per unit volume of natural gas produced. A range of requirements was thus obtained along with a number of outliers, especially corresponding to the 3 states that contributed to less than 2% of total gas withdrawals. These were removed and a triangular S2

3 distribution was fitted to the remaining data using minimum, maximum and modal values from the data range. The U.S. EIA also reports the quantity of natural gas flared and vented by state (3) which was assumed as flared due lack of disaggregated data, as modeled by Jaramillo et al. (1). This quantity was normalized by the gas produced by state to obtain a range of values representing the volume of natural gas vented and flared per unit volume of gas produced. After removing outliers, an exponential distribution was found to best fit the remaining data. A range (minimum and maximum) of fugitive CO 2 emissions factors (due to accidental leaks from pressurized equipment) from natural gas production in developed countries is suggested in the 2006 IPCC Guidelines report (N 2 O emissions were reported as zero) (4). These were used as the parameters of a uniform distribution representing the associated uncertainty. The total CH 4 fugitive emissions from natural gas production in 2008 are reported for six U.S. regions consisting groups of U.S. states in EPA s 2010 Inventory of U.S. Greenhouse Gas Emissions and Sinks (5). Some of the U.S. states are covered by more than one region. Therefore, some of the states were aggregated to obtain emissions from four regions in all (West-coast, Rocky mountains, North-east and the aggregated region consisting of the Gulf-coast, Mid-central and South-west). In addition, some of the activities corresponding to the production emissions that occurred once or a fixed number of times during the lifetime of a well were also presented in the Inventory (5). These activities include methane emissions due to well drilling, completion flaring, and well workovers (remedial operations carried out to increase production (6)). These were removed and treated separately while the fugitive remaining CH 4 emissions were normalized by the gas production in 2008 in these regions to obtain emissions factors per unit volume of gas produced. These factors were used as parameters of a discrete distribution with the corresponding probabilities being the fractions of natural gas produced from that region. Emissions from well drilling and completion usually occur once in the lifetime of the well and the quantity of CH 4 vented per well is reported for each region in the Inventory (5). The quantity of CH 4 vented per well workover is also reported in the Inventory (5). The EPA reports updated national S3

4 average emissions factors for conventional well completions and workovers in Technical Support Documents for the Proposed Rule for Greenhouse Gas Mandatory Reporting Rule specific to the oil and natural gas systems, that was published recently (6) (0.71 metric tons CH 4 /well and 0.05 metric tons CH 4 /workover respectively). Using numbers in the Inventory (5) and EPA technical support document (6), emissions from well drilling and completion were estimated to be 0.72 metric tons CH 4 /well on average. Using an estimated average U.S. production of 0.15 MMscf/day of natural gas per well over a conservative lifetime estimate of 5 years, the normalized emissions were approximately 0.06 g CO 2 e/mj of natural gas. To estimate the emissions from well workovers, the number of workovers was assumed to be one per year as a conservative estimate. The emissions per year were normalized by an average estimate of 0.15 MMscf/day of natural gas production to obtain emissions of about 0.02 g CO 2 e/mj of natural gas. More reasonable estimates of well lifetimes are likely to be higher, and hence these emissions were considered negligible per unit volume of natural gas produced and were ignored. Similar emissions from unconventional wells are usually higher, but when normalized by the natural gas production over the lifetime of the well, these emissions are also likely to also be small on a per MJ basis. All distributions for domestic natural gas and their parameters are summarized in Table S 2. Processing The U.S. EIA reports the quantity of natural gas used as plant fuel at natural gas processing units by state (2). The quantity of natural gas processed by state also reported by EIA (7), was used to obtain the volume of lease fuel required per unit volume of natural gas processed, by normalization. After removing outliers, a lognormal distribution was found to best fit the remaining data. Uniform distributions using the range of emissions factors given in the 2006 IPCC Guidelines report (4) were used to represent the uncertainty in fugitive and flared CO 2 and N 2 O emissions. An average emissions factor of 1250 metric tons CH 4 /processing plant suggested by EPA s State Inventory Tool (8) was used along with the total number of processing plants in each state reported by the EIA (9), to obtain a range of emissions from processing plants. This was normalized by the quantity S4

5 of natural gas processed in each state. A lognormal distribution was fit to the data representing the CH 4 fugitive emissions per unit volume of gas processed. Pre-processed natural gas contains varying amounts of CO 2 that is separated and vented to the atmosphere before the gas reaches the transmissions pipelines, where it must contain less than 1% of CO 2 by volume. The amount of CO 2 separated from produced natural gas and vented, in order to meet transmission quality standards, was estimated using Equation S 1. Equation S 1. Vented CO 2 emissions from natural gas processing (modified from (5)) Raw CO 2 emissions from NG processing = (% CO 2 in produced gas 1% CO 2 in transmission system)*ng production The volume of CO 2 vented per unit volume of natural gas produced equals the difference between the % CO 2 content in produced gas and 1% CO 2 in the transmission system. The % CO 2 content in the produced gas has been reported for six U.S. region in EPA s 2010 Inventory of U.S. Greenhouse Gas Emissions and Sinks (10), two of which have less than 1% CO 2 by volume (thus requiring no CO 2 separation). The total quantity of CO 2 vented in the U.S. has been modeled as a discrete distribution, using the emissions for the six regions and the corresponding probability of occurrence as the fraction of total natural gas produced in that region. Since the resulting range is based on the quantity of natural gas produced rather than processed, it is not multiplied by the factor of Transmission and distribution As explained in Jaramillo et al. (1), some of the natural gas in the transmission system is supplied to large-scale consumers such as power plants, while the balance is supplied through local distribution systems to residential and commercial consumers. The U.S. EIA reports the quantity of pipeline fuel used by state (2), along with the annual flow rates of natural gas within the various pipeline transmission systems in 2008 (11). The total flow rate of natural gas within a state was estimated as the total flow of natural gas into a given state for the entire year. Therefore, a range of normalized quantities of pipeline fuel required per unit volume of natural gas transmitted was estimated. A generalized extreme value distribution was found to best fit the data range. Generalized extreme value S5

6 distribution refers to a family of continuous distributions that contain the Gumbel, Fréchet and Weibull distributions (12). Parameters include a location (mu), a scale (sigma) and a shape parameter (k). When the value of k is less than zero, the distribution is equivalent to the Weibull distribution, and when the value of k is greater than zero, it is equivalent to the Fréchet distribution. As k approaches zero, the distribution is equivalent to the Gumbel distribution. Based on available data for the pipeline fuel use parameter, the value of k was estimated to be , greater than zero, and hence this distribution was equivalent to the Fréchet distribution. Uniform distributions using the range of emissions factors given in the 2006 IPCC Guidelines report (4) were used to represent the uncertainty in fugitive and vented CO 2 emissions in transmission and storage systems. The fugitive CH 4 emissions from transmission systems are primarily from leaks from pipelines and gas compressor stations. The EPA s State Inventory Tool (8) reports an average emission rate of 0.62 metric tons CH 4 /mile of pipeline and 980 metric tons CH 4 /gas compressor station. Data on all the transmission pipeline systems in the U.S has been reported in Interstate Natural Gas Infrastructure Map Book (13), including the length of pipeline, average annual flow rate and number of gas compressor stations. Using this data along with the emissions factors, the CH 4 emissions per unit volume of natural gas in each of the transmission pipeline systems were estimated. An exponential distribution was found to best fit this data range. The fugitive CH 4 emissions from distribution systems are primarily from leaks in distribution pipelines and associated distribution services (pipeline connections to end-users). The length of pipelines in each state and services are reported in the Gas Facts 2007 Data report (14). The EPA s State Inventory Tool reports (8) an average of 0.54 metric tons CH 4 /mile of pipeline and 0.02 metric tons CH 4 /service. The CH 4 emissions from distribution systems estimated for each state were normalized by the flow rates of natural gas through the transmission systems. The latter was used, since only part of the natural gas in transmission systems is sent through distribution pipelines to smaller endusers such as residences and commercial property. S6

7 Combustion Emissions from combustion were estimated as explained in the main text of the paper. Updates to the emissions factors based on EPA s Technical Support Document The updated EPA technical support document (6) suggests that the emissions factors used in 4 areas of the Inventory need to be revised. These are listed and the corresponding update in this study is outlined. 1. Centrifugal compressor wet seal degassing venting The revised EPA technical support document (6) states that centrifugal compressors use high-pressure oil based seals that are depressurized in order to vent any absorbed CO 2 or natural gas and that these emissions are not recorded previously in the Inventory of U.S. Greenhouse Gas Emissions and Sinks (5). They estimate that 233 metric tons of CH 4 is vented per centrifugal compressor in the natural gas system per year. The Inventory (5) indicates that centrifugal compressors are used at natural gas processing plants and in the transmission and storage systems and records the total number of these compressors used in The wet seal degassing emissions are estimated at the processing plant based on the emissions factor reported in the revised EPA technical support document (6) and the number of compressors indicated in the Inventory (5). Based on the results, a 20% increase in methane emissions was estimated at processing plants. This percentage increase was used to update the default CH 4 emissions factor per processing plant reported by the State Inventory Tool (8) that was used in this study. Similarly, the emissions due to wet seal degassing were estimated for the transmission and storage systems, which were found to increase the total emissions for these systems in the Inventory (5) by 8%. This percentage increase was applied to the default CH 4 emissions factor per compressor station reported by the State Inventory Tool (8) that was used in this study. 2. Venting emissions from wells during liquid unloading These emissions are primarily estimated for liquid unloading from conventional wells required due to blowdowns. The revised EPA technical support document (6) reports an average of 11 metric tons CH 4 per well per year vented due to this activity. This factor was used along with the number of wells in each of the U.S. regions to obtain the total volume of emissions vented. The volume of methane thus S7

8 obtained was normalized by the total natural gas produced in 2008 to obtain an average emissions rate of 3 g CO 2 e/cf, assuming all of the gas was vented. Based on the estimates in the report, this may vary between 1 and 5 g CO 2 e/cf. This was added to the methane emissions previously estimated for the production of natural gas. 3. Well completion Updated emissions factors were used in estimating emissions in the production stage and are reported in that section. These were found to be negligible per unit volume of natural gas produced. 4. Well workovers Updated emissions factors were used in estimating emissions in the production stage and are reported in that section. These were found to be negligible per unit volume of natural gas produced. LIQUEFIED NATURAL GAS Production The mean value of GHG emissions for natural gas produced in the U.S. was estimated to range between 6 to 14 g CO 2 e/cf from our analysis. The ETH-ESU database in the life cycle analysis software Simapro (15) reports GHG emissions for a few European and African countries, ranging between 1.5 and 2.6 g CO 2 e/cf. In a similar deterministic analysis of life cycle GHG emissions from LNG in Europe, Arteconi et al. (16) consider production emissions based on previous studies on LNG imported to the U.S., by ARI Inc. and ICF International (17) and on LNG imported to Japan, by Okamura et al. (18). The ARI&ICF report (17) uses production emissions of 1 g CO 2 e/cf, while Okamura et al. (18) use a value of 0.9 g CO 2 e/cf. As suggested by Arteconi et al. (16), the emissions observed in other countries worldwide are possibly higher than values reported in these analyses where superior technologies are assumed. These data points, including those developed in this study, were used to estimate the parameters of a triangular distribution to represent emissions from the production of gas produced in countries where the U.S derives its imports. S8

9 Liquefaction Jaramillo et al. (1) consider emissions from liquefaction reported in Tamura et al. (19), ranging from 5.1 to 14 g CO 2 e/cf. Arteconi et al. (16) suggest using data from ARI&ICF (17) and Okamura et al. (18) of 7.1 and 8.8 g CO 2 e/cf respectively. Emissions for natural gas liquefied in Algeria, based on reported values in the Ecoinvent database in Simapro (15), were estimated to be 10 g CO 2 e/cf. A study on life cycle emissions of natural gas imported from Australia to the U.S. conducted by Climate Mitigation Services for the Environmental Defense Center (20) reports liquefaction emissions of 9.3 gco 2 e/cf, based on an existing Conoco-Phillips plant in Australia. These data points were used to estimate the parameters of a triangular distribution assumed to quantify the uncertainty in emissions from liquefaction of natural gas imported to the U.S. Shipping Port-to-port distances between LNG exporting countries and the U.S. were obtained from online distance calculators (21, 22). The shipping distance for a unit volume of LNG transported was modeled as a discrete distribution, where the corresponding probability that LNG is shipped a particular distance was estimated as the fraction of total imports that was shipped that distance. The EIA reports the quantities of LNG imported to the U.S. for the year 2008 (23) which were used to estimate fractions (and hence probabilities) for particular shipping distances. The data used to develop the parameters of the discrete distribution is summarized in Table S 1. Table S 1. LNG import volumes and shipping distances Quantity imported Distance Port-to-port distance Imports to port (mmcf) (nm) assumption Northeast Gateway - Imports from Trinidad and Tobago Trinidad to Boston Lake Charles, LA - Imports from Nigeria Bonny to Lake Charles Cove Point, MD - Imports from Egypt Port Said to Baltimore Sabine Pass, LA - Imports from Qatar Ras Laffan to Sabine S9

10 Elba Island, GA - Imports from Nigeria Bonny to Savannah Freeport, TX - Imports from Trinidad and Tobago Trinidad to Freeport Sabine Pass, LA - Imports from Nigeria Bonny to Sabine Lake Charles, LA - Imports from Egypt Port Said to Lake Charles Cove Point, MD - Imports from Trinidad and Tobago Trinidad to Baltimore Cove Point, MD - Imports from Norway Hammerfest to Baltimore Elba Island, GA - Imports from Egypt Port Said to Savannah Elba Island, GA - Imports from Trinidad and Tobago Trinidad to Savannah Everett, MA - Imports from Trinidad and Tobago Trinidad to Everett The emissions intensity of shipping by ocean tanker, represented as the GHG emissions per unit volume of LNG transported per unit distance was estimated from Jaramillo et al., in the Ecoinvent database in Simapro (with respect to the transport of Algerian crude) and Okamura et al. These data points were used as the parameters of a triangular distribution, which was further multiplied with the discrete distribution representing shipping distances, to obtain an estimate of emissions from transporting a unit volume of LNG to the U.S. by ocean tanker. Regasification Jaramillo et al. (1) describe and use regasification emissions as reported in Tamura et al. (19) and Ruether et al. (24). Therefore, these emissions were used as the parameters of a uniform distribution representing uncertainty in regasification emissions, due to lack of better data. A summary of modeling parameters and the distributions used to represent the uncertainty and variability in upstream emissions from LNG is presented in Table S3. S10

11 Table S2. Summary of types and parameters of distributions representing domestic natural gas life cycle activities Modeling parameters Distribution Distribution parameters Units Lease fuel use Triangular 0, 0.04, 0.13 cf/cf Gas vented/flared, production Exponential cf/cf Fugitive CO 2 emissions, production Uniform 0, g/cf Fugitive CH 4 emissions, production Discrete [0.10, 0.02; 0.13, 0.69; 0.23, 0.23; 0.24, 0.06] g/cf Fugitive CH 4 emissions, production (emissions from liquid unloading) Uniform 0.04, 0.2 g/cf Plant fuel use Lognormal -3.8, 1.0 cf/cf Flaring CO 2 emissions, processing Uniform 0.05, 0.14 g/cf Flaring N 2 O emissions, processing Uniform 9.5x10-07, 1.2x10-05 g/cf Fugitive CO 2 emissions, processing Uniform 0, g/cf Fugitive CH 4 emissions, processing Lognormal -2.5, 0.62 g/cf Vented CO 2 emissions, processing Discrete [0, 0.02; 0, 0.14; 0.61, 0.49; 1.1, 0.06; 1.4, 0.07; 3.4, 0.23] g/cf Pipeline fuel use Generalized extreme value 0.25, , cf/cf Fugitive CO 2 emissions, transmission Uniform 0, 5.6x10-05 g/cf Vented CO 2 emissions, transmission Uniform 2.5x10-05, 1.7x10-05 g/cf Fugitive CO 2 emissions, storage Uniform 2.8x10-06, 2.1x10-05 g/cf S11

12 Fugitive CH 4 emissions, transmission Exponential 0.04 g/cf Fugitive CH 4 emissions, distribution Lognormal -3.7, 0.86 g/cf CO 2 emissions, combustion Triangular 50, 52, 54 g/cf CH 4 emissions, combustion Triangular 2.8x10-04, 9.2x10-04, 2.8x10-03 g/cf N 2 O emissions, combustion Triangular 2.8x10-05, 9.2x10-05, 2.8x10-04 g/cf Table S3. Summary of types and parameters of distributions representing LNG life cycle activities Modeling parameters Distribution Distribution parameters Units Production emissions Triangular 0.9, 2.0, 14 g/cf Liquefaction emissions Triangular 5.1, 10, 14 g/cf Shipping emissions Triangular 3.3x10-04, 4.8x10-04, 6.2x10-04 g/cf-nm Regasification emissions Uniform 0.4, 1.8 g/cf Figure S 2 consists of plots representing the best-fitting continuous distributions for some of the most sensitive parameters in the domestic natural gas life cycle model (as presented in later in Figure S 4). For instance, a lognormal distribution was found to best-fit the data available for the emissions from plant fuel combustion, as shown in the top right panel of Figure S 2. The data for emissions from lease fuel combustion (shown in the top left panel of Figure S 2) did not pass standard goodness-of-fit tests when fitted to common probability distributions. Therefore, a triangular distribution was used to represent this model input, with the minimum, maximum and mode of the data representing the corresponding parameters of the triangular distribution. S12

13 Figure S 2. Continuous distributions fitted to available data representing sensitive inputs to the life cycle model Raw data Triangular fit Raw data Lognormal fit Density Density CO 2 lease fuel (cf/cf) CO 2 plant fuel use (cf/cf) 15 Raw data Lognormal fit Raw data Exponential fit Density 10 5 Density CH 4 fuguitives processing (g/cf) CH 4 fugitives transmission (g/cf) ` PROBABILITY MIXTURE MODEL The emissions obtained per MJ of domestic natural gas and LNG consumed were combined in a probability mixture model presented in Equation S 2. In this equation, the probability distribution representing emissions per MJ of natural gas consumed in the U.S. is represented by!! (!). Equation S 2. Probability Mixture Model f X (x) = 2! i=1 a i f Yi (x) where! (!) is the probability distribution representing life cycle GHG emissions from consuming 1 MJ of natural gas consumed in the U.S. S13

14 !!! (!) is the probability distribution representing life cycle GHG emissions from consuming 1 MJ of domestic natural gas!!! (!) is the probability distribution representing life cycle GHG emissions from consuming 1 MJ of LNG!! is the fraction of domestic natural gas consumed in the U.S. (98.5%)!! is the fraction of LNG consumed in the U.S. (1.5%) RESULTS The mean and range of GHG emissions from the life cycle stages of domestic natural gas are presented in Table S 4. Table S 4. Summary statistics for the GHG emissions from the life cycle stages of domestic natural gas, where COV is the coefficient of variation Life cycle stage Mean (g CO 2 e/mj) Standard Deviation COV 90% CIlower 90% CIupper Production Processing Transmission and storage Distribution Combustion Total Domestic Natural Gas Life cycle The mean and range of GHG emissions from the life cycle stages of domestic natural gas are presented in Table S 5. S14

15 Table S 5. Summary statistics for the GHG emissions from the life cycle stages of LNG, where COV is the coefficient of variation Life cycle stage Mean (g CO 2 e/mj) Standard Deviation COV 90% CIlower 90% CIupper Production Liquefaction Shipping Regasification Transmission and Storage Distribution Combustion Total LNG Life cycle The histograms representing the life cycle GHG emissions from domestic natural gas and LNG are presented in Figure S 3. Figure S 3. Histograms representing life cycle GHG emissions from domestic natural gas and LNG Domestic natural gas LNG 0.1 Density Life cycle GHG emissions (gco 2 e/mj) S15

16 An uncertainty importance analysis was carried out, using the Spearman s rank correlation coefficients between input parameters and output life cycle emissions to represent how strongly the input parameters influence the output. The squares of the coefficients were normalized by the sum of the squares of all the coefficients, and these fractions were assumed to represent the contribution of that input parameter to the total variance in output. CO 2 emissions from combusting lease fuel during production and from plant fuel during processing represent the largest uncertainty in life cycle emissions from domestic natural gas. For LNG imports, the emissions from production contribute significantly to the uncertainty. Note that LNG production was considered an aggregated activity with a single emissions factor (and a range), while the domestic natural gas production stage was modeled in more detail, with a number of contributing activities. The results from the uncertainty importance analysis are presented in Figure S 4 where the bars represent how much each of the inputs parameters contribute to the total variance in life cycle GHG emissions. Figure S 4. Uncertainty importance analysis contribution to variance of critical parameters Domestic Natural Gas CO lease fuel 2 CO 2 plant fuel Vented CO 2 production CH fugitives production 4 CH 4 fugitives processing CH 4 fugitives transmission CO 2 combustion CH 4 fugitives distribution Percent contribution to total variance Imported Liquefied Natural Gas Production Liquefaction Shipping distance CH 4 fugitives transmission CO 2 combustion CH 4 fugtives distribution Regasification Shipping emissions factor Percent contribution to total variance S16

17 The GHG emissions from the upstream activities of natural gas obtained in this study were found to encompass the results reported in a recent NETL study (24) which is yet to be peer-reviewed. The upstream emissions from domestic average natural gas are reported to be 11 g CO 2 e/mj (ranging from 7-20 g CO 2 e/mj) in the NETL study, while the upstream emissions obtained in this study averages 16 g CO 2 e/mj (ranging from 11 g CO 2 e/mj to 22 g CO 2 e/mj). A recent paper by Howarth et al. (25) indicates a methane leakage rate of % from the production and processing stages of both conventional gas and shale gas. In comparison, this study estimates a methane leakage rate of % from the production and processing stages. The Howarth study, however, is not directly comparable to this one since it is directed specifically at shale gas, uses non-standard 20-year GWP potentials, and transmission leakage rates more applicable to the Russian gas system than that found in the U.S. This study focuses on the domestic natural gas mix, which includes both conventional and non-conventional sources of natural gas, of which shale is only a fraction. Methane leakage transmission rates were estimated in this study using U.S. specific data and were found to be lower than the values reported by Howarth. Additionally, a study by Jiang et al. (26) shows far less emissions from Marcellus shale pre-production activities on a per MJ basis, compared to the Howarth study. As a bounding analysis for the comparison of replacing coal with natural gas in the power generation sector, the extremes of a range of power plant efficiencies obtained from EPA s Emissions & Generated Resource Integrated Database (egrid) (25) for 2007 were used. The cumulative distribution functions (cdfs) of the efficiencies of natural gas and coal plants as given in 2007 egrid database are plotted in Figure S 5. Natural gas plants greater than 500 MW were considered separately as these plants would be most likely to replace coal plants in being able to supply base-load power. The 90% confidence interval of capacity factor for these plants ranged between 1% and 57%, with the median value at 17% indicating that these do have the additional potential capacity to generate baseload power. The cdf representing all natural gas plants is shown for reference purposes. S17

18 Figure S 5. Cumulative distribution function plots of power plant efficiencies >500 MW natural gas plants All natural gas plants Coal plants Cumulative probability Power plant efficiency The 90% confidence intervals of the plant efficiencies are reported in Table S 6. Table S 6. Range of power plant efficiencies based on egrid data (25) Plant type 5 th percentile 95 th percentile Coal (all) Natural gas (all) Natural gas (>500 MW) For the bounding analysis, a low efficiency coal plant (5th percentile, 24% efficiency) was compared to a high efficiency natural gas plant (95th percentile, 49% efficiency). The curve representing this scenario is added to those shown (previously) in Figure 3 in the main text and is presented in Figure S 6. S18

19 Figure S 6. Trade-off between emissions reductions per MJ of natural gas consumed and the probabilities of achieving these reductions, including bounding analysis on power plant efficiencies Probability that expected emissions reductions are achieved by natural gas use Expected emissions reductions per MJ of natural gas used (g CO 2 e/mj natural gas) CNG Civic compared to gasoline Civic CNG Civic compared to diesel Jetta CNG bus (improved) compared to diesel bus Low efficiency coal plant (5 th percentile, 24% efficiency) compared to high efficiency natural gas plant (95 th percentile, 49% efficiency) NGCC power plant (50% efficiency) compared to coal power plant (34% efficiency) The range of emissions reductions that can be achieved by substituting a coal plant with a natural gas plant is large, and depends very strongly on the plant efficiencies considered. The dashed curve furthest to the right of the figure indicates a scenario that could be considered feasible replacing a low efficiency coal plant with a high efficiency natural gas plant and is capable of achieving significant emissions reductions. As a lower bound, a low efficiency natural gas plant (5 th percentile, 25% efficiency) could be compared to a high efficiency coal plant (95 th percentile, 37% efficiency). However, this represents a scenario that is most likely infeasible and capable of achieving little or no emissions reductions, similar to the scenarios where CNG is used as a transportation fuel. S19

20 REFERENCES (1) Jaramillo, P.; Griffin, W. M.; Matthews, H. S. Comparative Life-Cycle Air Emissions of Coal, Domestic Natural Gas, LNG, and SNG for Electricity Generation. Environmental Science & Technology 2007, 41, (2) U.S. Energy Information Administration Natural Gas Navigator: U.S. Natural Gas Consumption by End Use 2010, (3) U.S. Energy Information Administration Natural Gas Navigator: U.S. Natural Gas Gross Withdrawals and Production 2010, (4) Intergovernmental Panel on Climate Change 2006 IPCC Guidelines for National Greenhouse Gas Inventories; Japan, (5) U.S. Environmental Protection Agency Inventory of U.S. Greenhouse Gas Emissions and Sinks: ; Washington, D.C. (6) U.S. Environmental Protection Agency Technical Support Documents for the Proposed Rule for Greenhouse Gas Mandatory Reporting Rule. Subpart under proposed 40 CFR part 98 W Oil and Natural Gas Systems; 2010; p (7) U.S. Energy Information Administration Natural Gas Navigator: Natural Gas Processed 2010, (8) U.S. Environmental Protection Agency State Inventory and Projection Tool (9) U.S. Energy Information Administration Natural Gas Processing: The Crucial Link Between Natural Gas Production and Its Transportation to Market; Washington, D.C., 2006; p. 11. (10) U.S. Energy Information Administration Natural Gas Navigator: Definitions, Sources and Explanatory Notes (11) U.S. Energy Information Administration State to State Natural Gas Pipeline Capacities and Flows 2009, (12) MATLAB version The MathWorks, Inc. 2010: Natick, Massachusetts. (13) Rextag Strategies Interstate Natural Gas Infrastructure Map Book: ; 2010; p (14) Wilkinson, P.; Pierson, P. Gas Facts 2007 Data; Washington DC, (15) PRé Consultants SimaPro Version , Amersfoort, The Netherlands, S20

21 (16) Arteconi, A.; Brandoni, C.; Evangelista, D.; Polonara, F. Life-cycle greenhouse gas analysis of LNG as a heavy vehicle fuel in Europe. Applied Energy 2010, 87, (17) Advanced Resources International Inc. Life-Cycle Emissions Study: Fuel Life-Cycle of U.S. Natural Gas Supplies and International LNG; (18) Okamura, T.; Furukawa, M.; Ishitani, H. Future forecast for life-cycle greenhouse gas emissions of LNG and city gas 13A. Applied Energy 2007, 84, (19) Tamura, I.; Tanaka, T.; Kagjao, T.; Kuwabara, S.; Yoshioka, T.; Nagata, T.; Kurahashi, K.; Ishitani, H. M. S. Life cycle CO2 analysis of LNG and city gas. Applied Energy 2001, 68, (20) Heede, R. LNG Supply Chain Greenhouse Gas Emissions for the Cabrillo Deepwater Port: Natural Gas from Australia to California; Snowmass, CO, (21) World Ports Distances (22) PortWorld Distance Calculator - Ship Voyage Distance Calculator (23) U.S. Energy Information Administration U.S. Liquefied Natural Gas Imports by Point of Entry (24) Ruether, J.; Grol, E.; Ramezan, M. In Second International Conference on Clean Coal Technologies for our Future; Sardinia, Italy, (25) U.S. Environmental Protection Agency Emissions & Generation Resource Integrated Database (egrid 2010); S21