High-resolution emissions of CO2 from power generation in the USA


  • Garielle Pétron,

    1. Global Monitoring Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA
    2. Also at Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, Colorado, USA.
    Search for more papers by this author
  • Pieter Tans,

    1. Global Monitoring Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA
    Search for more papers by this author
  • Gregory Frost,

    1. Chemical Sciences Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA
    2. Also at Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, Colorado, USA.
    Search for more papers by this author
  • Danlei Chao,

    1. Global Monitoring Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA
    2. Also at Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, Colorado, USA.
    Search for more papers by this author
  • Michael Trainer

    1. Chemical Sciences Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA
    Search for more papers by this author


[1] Electricity generation accounts for close to 40% of the U.S. CO2 emissions from fossil fuel burning, making it the economic sector with the largest source of CO2. Since the late 1990s, the Environmental Protection Agency Clean Air Markets Division (EPA CAMD) has kept a repository of hourly CO2 emission data for most power plants in the conterminous United States. In this study, the CAMD CO2 data are used to derive a high spatiotemporal resolution CO2 emissions inventory for the electricity generation sector (inventory available on request). Data from 1998 to 2006 have been processed. This unique inventory can be used to improve the understanding of the carbon cycle at fine temporal and spatial scales. The CAMD data set provides the first quantitative estimates of the diurnal and seasonal cycles of the emissions as well as the year to year variability. Emissions peak in the summertime owing to the widespread use of air conditioning. Summertime emissions are in fact highly correlated with the daily average temperature. In conjunction with the EPA Emissions and Generation Resource Integrated Database (eGRID), we have derived high-resolution maps of CO2 emissions by fossil fuel burned (coal, gas, oil) for the year 2004. The CAMD data set also reflects regional anomalies in power generation such as the August 2003 blackout in the northeastern United States and the 2000–2001 increase in production in California. We recommend that all sectors of the economy report similar high-resolution CO2 emissions because of their great usefulness both for carbon cycle science and for greenhouse gases emissions mitigation and regulation.

1. Introduction

[2] Two major federal agencies in the United States provide information on CO2 emissions from facilities providing power to the electrical distribution grid: the Department of Energy's Energy Information Agency (EIA, http://www.eia.doe.gov) and the Environmental Protection Agency (EPA, http://www.epa.gov).

[3] EIA is in charge of gathering, organizing and projecting energy statistics for the United States. For the electricity sector, the EIA emissions estimates are based on annual fuel consumption reported by each facility multiplied by the heat content estimate, emission factor and combustion efficiency factor for each fuel (EIA Electric Power Annual, available at http://www.eia.doe.gov/cneaf/electricity/epa/epa_sum.html, October 2007).

[4] The main mission of EPA is to verify and ensure compliance of all kind of activities with federal environmental legislation and rules. Under the 1990 Clean Air Act, fossil fuel burning power plants in the United States above a threshold capacity are mandated to report hourly emissions of SO2, NOx and CO2 to the EPA Clean Air Markets Division. Records are available to the public at the unit level (boiler + stack) starting in 1996 (EPA CAMD Data Sets and Published Reports, available at http://camddataandmaps.epa.gov/gdm/index.cfm, 2007). Reported CO2 emissions rely on either (1) mass balance estimation using the amount of fuel burned and emission factors, (2) CO2 continuous emission monitoring (CEM) system, or (3) O2 CEM system. Every two to four years, EPA also releases an extensive database with annual statistics on power generation in the United States. The latest release of the Emissions and Generation Integrated Resource Database eGRID2006_v2 (eGRID, available at http://www.epa.gov/cleanenergy/egrid, 2006) was published online in 2006 and contains statistics on 4841 power generation facilities for the year 2004. The previous release of eGRID in 2002 covered the time period from 1996 to 2000. eGRID contains information including annual emissions estimates for all plants providing power to the grid, not only fossil fuel burning ones.

[5] Detailed bottom-up estimates of greenhouse gases emissions are needed by both the scientific community for local and regional carbon budget analyses and by local and national governments as a policy support tool for emissions mitigation and regulation. As part of a multiagency collaborative effort to better account for and describe fossil fuel burning emissions of CO2 in the United States (CO2 Fossil Fuel Emissions Effort (CO2FFEE) [Gurney et al., 2007]), we are developing a high-resolution inventory for each activity sector of the economy.

[6] This article focuses on U.S. emissions from power generation and useful thermal output. It is organized as follows. First we present some past, contemporary and projected statistics on U.S. electricity production and related CO2 emissions based on a data set published online and updated every year by the EIA. Then we describe and analyze the EPA CAMD CO2 data which we use to derive a high-resolution emissions inventory for CO2 from power generation.

2. Some Statistics on the U.S. Electricity Generation

2.1. Statistics on Energy Source and CO2 Emissions

[7] Energy related activities in the United States (including Alaska and Hawaii) emitted 1.6 × 1015 grams of carbon equivalent CO2 (1.6 PgC) in 2005 which represent 21% of the world emissions for that year (EIA International Energy Annual, 2005, 2007). Electricity generation and useful thermal output at combined heat and power (CHP) plants in the United States are the largest contributing economic sector with 685.1012 grams of carbon equivalent CO2 (685 TgC) emitted in 2005 (EIA Electric Power Annual, available at http://www.eia.doe.gov/cneaf/electricity/epa/epa_sum.html, 2005, 2007). Coal is the most widely used fuel with half of the electricity production share (measured in kilowatt-hours) (eGRID, available at http://www.epa.gov/cleanenergy/egrid, 2006). Then come natural gas or other gases burning plants and nuclear power plants with 19% each. Smaller fractions of electricity come from hydroelectric plants (6.5%) and petroleum/oil burning plants (3%).

[8] Coal has a smaller energy content than oil and natural gas. Therefore coal burning plants are responsible for most of the electricity sector CO2 emissions. In 2004, coal, gas and oil burning plants were responsible for 82.3%, 13.5%, and 4.2%, respectively, of the sector CO2 emissions. For comparison, the percentage contributions of each fuel to the total energy produced from power generation by fossil fuel burning were 71.2%, 24.5% and 4.3%, respectively (eGRID, available at http://www.epa.gov/cleanenergy/egrid, 2006). Nuclear and hydroelectric sources produce negligible CO2 emissions. Emissions from biomass and municipal waste burning plants are also small and come from a renewable resource such that they are not included here.

2.2. Evolution of the Emissions Since 1990

[9] Between 1990 and 2005, the U.S. population increased by 20% and the amount of electricity produced in the United States increased by 34% (U.S. Census, The 2008 Statistical Abstract, Tables 12 and 909, available at http://www.census.gov/compendia/statab/, 2007). Most of the new power was produced by coal burning plants. Emissions of CO2 from the electricity sector increased by 31% during this time frame which indicates no major changes in technology and efficiency (EIA, Emissions of Greenhouse Gases in the United States 2005, ftp://ftp.eia.doe.gov/pub/oiaf/1605/cdrom/pdf/ggrpt/057305.pdf).

[10] In 1990, close to 90% of the electricity sector CO2 emissions were produced by utility company plants. Since then independent producers have become a key player. In 2005, two-thirds of the electricity sector emissions of CO2 came from utility company plants (EIA, Electric Power Annual, 2006: Table U.S. Electric Power Industry Estimated Emissions by State). All types of producers though do report emissions to the EIA and EPA providing a unique data set to build a national emission inventory of point sources with high temporal resolution.

[11] Another way to look at the electricity sector emissions is to allocate those emissions by end use sector proportionally to the annual electricity sales. In 2005, electricity and useful thermal output from the grid were fairly evenly shared between the industrial, residential and commercial sectors. Interestingly, the commercial and residential sectors used 40% more power in 2005 compared to 1990 while the industrial sector used only 5% more (EIA, Emissions of Greenhouse Gases in the United States, 2005).

2.3. Projections to 2030

[12] In its 2007 Annual Energy Outlook report (AEO 2007), the EIA projected the generation of electricity to reach 5,478 billion kilowatt-hours in 2030 (EIA Annual Energy Outlook, available at http://www.eia.doe.gov/oiaf/aeo/index.html, 2007). This estimate is based among other considerations on a population increase of 23% from 2005 to 2030. In the absence of new technology and regulations (such as carbon sequestration and storage), the CO2 emissions from the energy generation sector are projected to reach 910 TgC in 2030, 800 Tg alone coming from coal fired power plants (EIA Annual Energy Outlook, Figure 92, available at http://www.eia.doe.gov/oiaf/aeo/index.html, 2007). In anticipation of future greenhouse gases emissions regulations, large power plants have become primary targets for a national study of carbon sequestration lead by the Department of Energy [National Energy and Technology Laboratory, 2007]. In the AEO 2008, the EIA projections are slightly lower and CO2 emissions from the electricity and useful thermal output sector are projected to reach 830 TgC in 2030. This updated projection illustrates the relatively large uncertainty attached to the many parameters needed to forecast energy demand, supply and price.

3. Building a High-Resolution Emission Inventory Based on the EPA CAMD Data Set

[13] Under the 1990 Clean Air Act, each fossil fuel burning facility reports hourly emissions of SO2, NOx and CO2, fuel flow, unit operation and monitoring performance. All coal burning power plants over 25 megawatt-hours (1 MW-h = 106 watt-hour) and new units under 25 MW-h and burning coal with high sulfur content are required to continuously monitor their SO2, NOx and CO2 emissions (EPA CEM, available at http://www.epa.gov/airmarkets/emissions/continuous_factsheet.html, 2007). Other power plants (including oil or gas burning plants) can use mass balance estimation methods. Record keeping and reporting have been mandatory since 1 January 1996.

[14] EPA does no provide any quantitative estimates of the uncertainties in the emissions reported for each facility. It is clear however that for power plants using mass balance calculations, uncertainties in emission factors will lead to correlated errors in the emissions. Ackerman and Sundquist [2008] have conducted a detailed comparison of reported CO2 emissions from conventional fuel (coal, oil or gas) burning plants in the EIA and eGRID data sets for 1998, 1999, 2000 and 2004. These authors have calculated between 3.5% and 5.8% difference in the annual totals reported by the two agencies for the conterminous United States. Most of the differences are due to plants using stack monitoring for the eGRID estimates and fuel consumption data for the EIA estimates. From the comparison of the EPA eGRID and EIA national totals (Table 1), it appears that the national total uncertainty on CO2 from power generation is quite small compared to uncertainties attached to CO2 biospheric fluxes.

Table 1. U.S. Electricity Generation From Coal, Natural Gas, and Oil Burning and Two Estimates of the Associated CO2 Emissionsa
YearTWhCAMD CO2 (TgC/a)EIA CO2 (TgC/a)
  • a

    Electricity generation in terawatt-hour (1012 Wh) (EIA Electric Power Annual, available at http://www.eia.doe.gov/cneaf/electricity/epa/epa_sum.html, October 2007). The first emissions estimate is derived from the EPA CAMD CO2 hourly data, while the second emissions estimate was calculated by Ackerman and Sundquist [2008]. The CO2 emission estimates are for fossil fuel burning conventional power plants and combined heat and power plants. The power generated by commercial and industrial generators (providing power to the grid) is included here, and the associated emissions account for less than 5% of the total emissions.


[15] The EPA Clean Air Markets Division centralizes all the facility level information and provides online access to prepackaged emissions data files (EPA CAMD Data Sets and Published Reports, available at http://camddataandmaps.epa.gov/gdm/index.cfm, 2007). Another query on the EPA Web site (EPA CAMD Facilities and Contacts, available at http://camddataandmaps.epa.gov/gdm/index.cfm, 2007) provides a text file with the latitude and longitude of each facility, which are not provided in the emissions data files.

[16] Before presenting an analysis of the CAMD data, we show two comparisons. For the conterminous United States, we have compared the CAMD annual total emissions from 1998 to 2006 with the emissions reported to the EIA and processed by Ackerman and Sundquist [2008] (see Table 1). The CAMD and EIA annual estimates are within 10%. Given the fairly close agreement, we have used the CAMD data throughout the rest of this analysis, in order to have access to CO2 hourly fluxes at individual plants for the conterminous United States.

[17] It is also interesting to compare the 2004 CAMD emissions with the ones reported in the latest release of eGRID. In theory, emissions reported for facilities listed in these two EPA products should be identical. The CAMD data set covers only fossil fuel burning power plants and CHP plants in the conterminous United States which is a subset of the facilities listed in eGRID. In 2004, we have 1123 plants with nonzero CO2 emissions in CAMD. All except nine facilities have the same 6-digit EIA identification number (called ORISPL) in both data sets. We found that this identification number is a better identifier than the name of the plant, which can change when ownership changes. Only three power plants in CAMD (ORISPL = 7254, 8000, 55683) are not in the eGRID database and their 2004 emissions total 27 GgC.

[18] The eGRID 2004 conterminous U.S. CO2 emissions estimates from fossil fuel burning plants is 678.5 TgC, of which less than 10% come from CHP plants (57.2 TgC) [Ackerman and Sundquist, 2008]. Yet if we only consider the plants that are also in the CAMD database, this total becomes 624.6 TgC, to be compared with 613.2TgC derived from the EPA CAMD data set (see Table 1). At the facility level and for power and CHP plants that are in both data sets, the CAMD and eGRID 2004 emissions are within 1% at 783 plants which cover 70% of the national total emissions. The difference is larger than 10% at 136 plants which represent 2.7% of the national total emissions. For comparison the EIA reports 655.8 TgC in 2004 (based strictly on fuel burned [Ackerman and Sundquist, 2008]).

4. CAMD CO2 Emission Data Analysis

4.1. Conterminous U.S. Emissions by Fossil Fuel Burned

[19] Coal, oil and gas have different 13C isotopic signatures. For the purpose of serving a wider community of emission inventory users, it is useful to derive CO2 emissions from power and CHP plants for each fossil fuel type. The EPA CAMD Unit Characteristics Report available for each year only mentions the primary and secondary fuel types burned at each facility but does not quantify the shares of each fuel burned (EPA CAMD Facilities and Contacts, available at http://camddataandmaps.epa.gov/gdm/index.cfm, 2007). This information is available however in the eGRID database.

[20] We have used the annual amount of electricity produced at each plant by fuel type burned (fuel share) reported in eGRID to apportion the CAMD CO2 emission data for coal, gas and oil for each facility. We have also used averaged CO2 uncontrolled emission factors (EF) reported by the EIA to convert fuel shares from Btu to CO2 emissions. The averaged EF are 210 pounds of CO2 per million Btu for coal, 128 for gas and 160 for oil (EIA Fuel and Energy Source Codes and Emission Coefficients, available at http://www.eia.doe.gov/oiaf/1605/coefficients.html, 2007). Three plants in CAMD are not listed in eGRID and 26 plants are listed in eGRID with nonzero emissions from conventional fuel burning but with no information on the power generation by fossil fuel burned. For these 29 plants, we have used the “primary fuel” listed in the CAMD Unit Characteristics Report for 2004 (EPA CAMD Facilities and Contacts, available at http://camddataandmaps.epa.gov/gdm/index.cfm, 2007) and their CAMD CO2 emissions total 0.34 TgC in 2004. All but one of these units use natural gas. Five facilities have nonzero generation from fossil fuel burning with zero emissions reported in eGRID (ORISPL = 10697, 1897, 56192, 50771, 589). Overall 60% of the facilities in the CAMD database use a combination of two or three fossil fuels.

[21] Table 2 lists the number of plants in the CAMD database with nonzero emissions that burn a single fossil fuel or a mix of fossil fuels. For each single fuel or fuel mix category, we provide the annual total electricity produced reported in eGRID and the related CAMD CO2 emissions. The total number of plants with the appropriate information in eGRID is 1094. It is interesting to notice that most coal plants have a small contribution from oil and gas burning probably as a back-up for peak demand. We obtain similar numbers as the EIA for the partitioning of the CO2 emissions by fuel type: 516.3 TgC (84%) for coal, 73.6 TgC (12%) for gas and 23.3 TgC (4%) for oil (EIA Electric Power Annual, available at http://www.eia.doe.gov/cneaf/electricity/epa/epa_sum.html, October 2007).

Table 2. The 2004 U.S. Electricity Generation by Fossil Fuel Source for Seven Different Categories of Facilities and the Corresponding 2004 EPA CAMD CO2 Emissions Dataa
Fossil Fuel UsedNumber of PlantsElectricity Produced in TWhCO2 Emitted (TgC/a)
  • a

    Categories of facilities include single fuel, two fuels, and three fuels (based on EPA eGRID, 2007). Twenty-nine plants have no information on their net generation in eGRID, and their total CO2 CAMD emissions amount to 0.34 TgC in 2004.

Coal only1020.0--5.8
Gas only415-367.5-47.7
Oil only17- 0.51.5
Coal + gas88320.26.0-90.4
Coal + oil2181267.1-15.2340.7
Gas + oil247-195.464.536.2
Coal + gas + oil99308.313.815.290.4
CO2 emitted (TgC/a)-516.373.323.3612.8

[22] In Figure 1, we show a diagram of the monthly U.S. CAMD CO2 emissions by fossil fuel type. The temporal pattern is based solely on the 2004 CAMD CO2 data while the partitioning by fuel type is constant through the year for plants using a mix of fuels. The seasonality of gas and oil emissions may therefore be underestimated here. Figure 2 shows a map of the United States with the location of the facilities listed in CAMD and their respective CO2 emissions from coal, gas or oil burning. Notice that there are no coal burning plants in California. To comply with state-level air quality emission regulation, the local generation of electricity in California relies mostly on natural gas burning power plants (52% of the net generation), as well as hydro (17%) and nuclear (16%) plants. As gas is more expensive than coal, electricity rates are 3 to 4 cents per kWh more expensive in California compared to the national average.

Figure 1.

Seasonal cycle of total U.S. EPA CAMD CO2 emissions for year 2004 by fossil fuel type.

Figure 2.

U.S. map of EPA CAMD 2004 CO2 emissions from coal, gas, and oil burning power plants.

4.2. Diurnal Cycle, Weekly Cycle, and Seasonality of the Emissions

[23] The EPA CAMD data also provide a unique insight into the amplitude of the diurnal cycle of the CO2 emissions from power plants. The oscillating lines in Figure 3 show the monthly mean diurnal cycle of the CAMD CO2 data for each month for 2004 (gray) and 2005 (black). During the cold months, the data show a clear “double shoulder” peak for the morning and evening hours due to increased heating and lighting demand. During the summer months, the CAMD CO2 emissions data peak between 1 and 4 P.M. when the temperature is highest and air conditioning demand is greatest.

Figure 3.

Monthly averaged U.S. EPA CAMD CO2 emissions (symbols) and the mean diurnal cycle for each month (lines) for 2004 and 2005.

[24] The square and round symbols in Figure 3 show the monthly CAMD CO2 emissions in GgC/h for 2004 and 2005, respectively. The monthly mean emissions exhibit a main peak during the summer months and a secondary peak during the winter months. 90% of the coal used in the USA is burned for electricity generation. It is therefore quite natural that the EPA CAMD CO2 emissions show annual and semi-annual cycles (Figure 1 and Figure 3) similar to the ones described by Blasing et al. [2005] for the national CO2 emissions from coal burning.

[25] It is important to notice that summer CO2 emissions were larger in 2005 than in 2004, while winter emissions were greater in 2004 than in 2005. Warmer mean winter (December–February) and summer (June–August) temperatures in 2005 (1.41/22.8°C) compared to 2004 (0.56/21.6°C) are certainly the cause of the summer emissions difference and a partial contributor to the difference in winter emissions between 2004 and 2005.

[26] Electricity demand also exhibits a clear weekly cycle with a typical minimum on Sunday and Saturday. Figure 4 shows the average weekly cycle of CAMD CO2 emissions calculated for each month (January to December) using data from 1998 to 2006. Low emissions on national holidays such as Thanksgiving (the last Thursday of November) and Memorial Day (the last Monday of May) also appear on this graph. The lower electricity usage reflects the slowdown in the economy activity as most people have those days off.

Figure 4.

Mean weekly cycle of the total U.S. EPA CAMD CO2 emissions averaged for each month using data from January 1998 to December 2006. We have used the code provided at http://wcarchive.cdrom.com/pub/simtelnet/msdos/fortran/weekday.for to compute the day of the week.

4.3. EPA CAMD CO2 Versus Temperature

[27] We have shown above that the U.S. CAMD CO2 emissions peak during the hottest months of the year. In Figure 5, we have plotted national monthly total CAMD CO2 emission (in TgC/d) as a function of the national monthly mean temperature (NOAA NCDC, available at http://www.ncdc.noaa.gov/oa/documentlibrary/hcs/hcs.html, 2007) from January 1998 to December 2006. There are strong correlations at low (<10°C) and high (>15°C) temperatures. In the wintertime, cold temperatures and shorter daylight hours both contribute to the increase in electricity demand. We later show that in the summertime, emissions are highly correlated with the average (or maximum) daily temperature.

Figure 5.

Monthly total U.S. EPA CAMD CO2 emissions versus the monthly U.S. average temperature (reported online at http://www.ncdc.noaa.gov/oa/documentlibrary/hcs/hcs.html) from January 1998 to December 2006.

[28] To better isolate the temperature dependence of electricity production, we have analyzed the CAMD CO2 data first for eight U.S. regions and later by latitude. The North American Electric Reliability Corporation (NERC) works with eight regions covering the continental USA and part of Canada to “improve the reliability of the bulk power system” (NERC, available at http://www.nerc.com/regional/, 2007). The electricity generated in each region may either be used within the region or transmitted to another region primarily within the same interconnection (see map, Figure 6). In the case of the WECC and ERCOT regions, most power generated is used within the same region. Table 3 shows the eight U.S. subregions we are working with and their 2004 total CAMD CO2 emissions (for power generated in the United States only). As the NERC regions do not follow state boundaries, the partition we have built is not exactly the same as NERC. Nevertheless, we choose to use the NERC names in the rest of this study to designate our regions. Only a small fraction of the power used in the United States is produced in Canada (<2% [National Energy Board, 2007]) or Mexico and the related emissions are not reported to the EPA. Table 3 also displays for each “NERC region” the per capita CO2 emissions from fossil fuel burning plants and the average CO2 emission rate.

Figure 6.

NERC regions and interconnections.

Table 3. The 2004 EPA CAMD CO2 Emissions and Power Generation in Eight U.S. Subregionsa
States/RegionPopulationCO2 Emissions (TgC/a)Generation (TWh) FromTotal Generation From Fossil Fuel (TWh)Emission Rate (MWh/tons of CO2)Tons of CO2 per Capita
  • a

    Each state has been assigned to its “closest” NERC region quoted in parenthesis. The 2004 population numbers are an interpolation between the 2000 and 2005 U.S. Census data. The power generation numbers come from eGRID. To put the numbers in the last column in perspective: in 2004 the per capita total CO2 emissions from fossil fuel burning and flaring was 20.26 tons of CO2 in the United States, while the world average was 4.27 tons of CO2 (EIA International Emissions Data, available at http://www.eia.doe.gov/environment.html, 2007).

WA, OR, CA, NV, AZ, UT, ID, MT, WY, CO, NM (“WECC”)65,353,70681.8236.9175.83.1415.70.874.6
ND, SD, NE, MN, IA, WI (“MRO”)12,547,78248.8165.05.11.7171.81.2714.3
KS, OK (“SPP”)6,261,87122.768.422.50.991.81.1013.3
TX (“ERCOT”)22,458,33861.4146.7179.31.5327.60.8410.0
MO, IL, AR, LA, MS, AL, GA, SC, NC, TN (“SERC”)60,732,515160.4563.088.611.1662.81.089.7
FL (“FRCC”)17,428,36732.662.076.637.2175.80.836.8
MI, IN, OH, PA, WV, KY, VA, NJ, DE, MD (“RFC”)68,620,081179.9687.952.020.0760.01.069.6
NY, VT, ME, NH, MA, RI, CT (“NPCC”)33,375,27825.542.175.333.6150.90.752.8

[29] For each NERC region, we have plotted the total monthly CAMD emission data for each month with an average regional temperature higher than 20°C versus the cumulative monthly regional Cooling Degree Days (for a specific day, the cooling degree day CDD is calculated as the departure of the daily average outdoor temperature from 65°F) from 1998 to 2006 (not shown). The latter are based on population weighted state CDD reported for each month and each state by the NOAA National Climatic Data Center (http://www.ncdc.noaa.gov/oa/documentlibrary/hcs/hcs.html). There is a strong linear relationship for most regions (r2 > 0.82 for ERCOT, FRCC, RFC and SERC, r2 = 0.71 for MRO, 0.66 for NPCC, 0.61 for SPP and 0.45 for WECC). Using the derived slopes and the 29.5 extra CDD projected by Hadley et al. [2006] for 2010, we compute an increase in the U.S. emissions of 1.45 TgC/year between 2004 and 2010. This value is slightly lower than the 2 TgC/year derived by Hadley et al. [2006] with an energy use model.

[30] In Figure 7, we illustrate how the emissions vary with latitude and season for year 2004. The CAMD CO2 data have been binned in one degree latitudinal bands for each month of 2004 (top plot). The middle plot shows the relative spread of the monthly emissions around the annual mean. South of 35°N, emissions peak during the summer months, while north of 36°N, the winter and summer emissions maxima are very close. The bottom plot shows the relative spread of the monthly latitudinal mean (daytime maximum) temperature around the annual mean. The temperature data come from the European Centre for Medium Range Weather Predictions (ECMWF) temperature forecast and have been weighted with population density. For each latitude, the temperature corresponding to the month with the smallest emissions is plotted as a diamond symbol and is referred to as the “optimal temperature” (Figure 7, bottom plot). The month of optimal temperature varies from November–December in southernmost areas to April–May in northerly areas. The winter minimum and summer maximum in the southern latitudes reflect the fact that summertime air conditioning dominates the energy use. In the northern latitudes, the winter and summer energy needs are almost equally high for air conditioning in the summer and lighting and electric heat in the winter. The optimal mean daily maximum temperature is close to 22°C south of 36°N and closer to 17°C north of 36°N. Nationally the optimal monthly average temperature is located between 10 and 15°C (see Figure 5). Table 3 illustrates some major regional differences in energy usage.

Figure 7.

(a) Latitudinal distribution of the EPA CAMD CO2 monthly emissions in 2004 (unit is TgC/d), (b) ratio of the monthly mean daily emissions to the annual mean daily emissions and latitudinal distribution of the U.S. population density (diamond symbols, no unit), and (c) latitudinal and monthly mean daily maximum temperature (degrees Celsius) weighted by population density. Here, the diamond symbols highlight the month with the minimum emissions. In Figures 7a–7c, different symbol colors correspond to different months.

4.4. Regional Anomalies

[31] In a further example of the robust nature of the CAMD data, we present two cases of regional-scale emissions anomalies that are apparent in the data set. These anomalies can be explained by known events in regional power generation and distribution.

[32] On 14 August 2003, the northeastern United States and part of Canada experienced a major power blackout for several hours. The cascade failure of several power lines brought over a hundred power plants in the Unites States off-line. Over 50 million people in New York, New Jersey, Maryland, Connecticut, Massachusetts, Michigan, Ohio and Ontario were affected by the power outage. The CAMD CO2 data for this time period reflect the regional-scale shutdown of power plants. The emissions in New York State started decreasing at 2 P.M. on 14 August and reached practically zero between 3 and 5 P.M. (see Figure 8). It took a few days for the power generation to go back to “normal” in New York and Michigan. The cascading blackout is believed to have followed the failure of the largest unit (# 5) of the Eastlake plant in Ohio which went off-line between 1 and 2 P.M. on 14 August. The extra pressure put on other plants led to higher voltages on some high voltage lines. Some of these lines eventually went out of service when they sagged into “overgrown” trees. Unlike Michigan and New York state, Ohio only lost 30% of its power production for a couple of hours. The plants going off line were relayed by other available or underused units.

Figure 8.

EPA CAMD CO2 data for New York State, Michigan, and Ohio during the northeastern U.S. 2003 blackout.

[33] Another example of a regional anomaly in power production comes from California during the so-called Western Energy Crisis of 2000 and 2001 when rolling blackouts affected several areas. Monthly total CAMD CO2 emissions data for California are shown on Figure 9 for year 2000 and 2001 and for a 9-year average (1998 to 2006). In an attempt to reduce energy prices in the state, the governor of California signed new legislation in 1996 to deregulate the energy market. However, starting in May 2000, energy prices started soaring. Interestingly, emissions are also unusually high from May 2000 to October 2001. The increase in power generation was completely covered by independent producers rather than by utility companies (see insert in Figure 9 and EIA U.S. Electric Power Industry Estimated Emissions by State (available at http://www.eia.doe.gov/cneaf/electricity/epa/epat5p1.html, 2007)). A combination of structural and natural causes led to the Western Energy Crisis: high dependence on electricity imports (∼25%), rupture of a major gas pipeline, increase in unplanned outages, drought in the Pacific Northwest, and hotter summer. The increase in local production during the summer of 2000 compensated the reduction of hydroelectric power coming from Oregon and Washington which were experiencing a drought.

Figure 9.

California EPA CAMD CO2 monthly emissions for 2000 and 2001 and the average over 9 years (1998 to 2006). Inserted at the bottom is a plot of the annual CO2 emissions from California electric utilities and independent power producers (IPP) noncogenerators and cogenerators from 1989 to 2005 as reported by the EIA.

5. Discussion and Conclusions

[34] The EPA CAMD data provide a unique objective and detailed estimation of the CO2 emissions from power plants. There is no such equivalent for any other sector of the economy. The CAMD data agree very well with the emission estimates derived from fossil fuel burned at power plants by the EIA.

[35] By making use of CAMD data, carbon cycle modelers can take into account the strong daily, weekly and seasonal cycles of CO2 emissions from coal, gas and oil burning for electricity generation. Furthermore, since the location of each power plant is known, the associated emissions should be modeled as resulting from a point source. Until now, most studies on fossil fuel CO2 in the USA have relied on the Andres et al. [1996] 1 × 1 degree inventory, which used population density as a spatial proxy and gave only annual average emissions.

[36] The analysis of the EPA CAMD data together with surface temperature and summertime CDD records provides some insight on the climate dependence of electricity use in the United States. A recent analysis by Lynn et al. [2007] suggests that the mean summer surface temperature in the eastern United States could be 5.5°C higher by 2080 compared to the 1990s. How much more CO2 would result from this? Let us assume that the linear relationship between cumulative CDD and CO2 emissions mentioned previously still holds at higher temperatures. For the three NERC regions covered by Lynn et al.'s [2007] study, SERC (61.3 million inhabitants in July 2005), RFC (69 million inhabitants) and the U.S. part of NPCC (33.5 million inhabitants), a 5.5°C increase in the mean summer temperature would result in 0.3 TgC/d extra CO2 emitted to meet the increased electricity needs. This corresponds to approximately 18% of today's average summer emissions for the conterminous United States. This simple analysis is a clear example of a positive feedback between climate and anthropogenic CO2 emissions.

[37] As is required for NOx and SO2 emissions based on various Federal regulations, CO2 emissions by all major point and distributed sources should be estimated and made public. Emission inventories also should be constructed regularly using consistent and verified estimation techniques. Monitoring CO2 emissions at the local scale and for all activity sectors is essential for formulating emission reduction efforts, testing the effectiveness of such efforts, measuring the feedbacks between factors influencing energy use and carbon emissions, and properly identifying and quantifying regional anomalies.


[38] The authors wish to thank the CO2FFEE members for the valuable knowledge, critical experience, and articulated vision on where carbon cycle science is and should be regarding CO2 emissions from fossil fuel use. We also want to deeply acknowledge Puneet Pasrich for helping us download the EPA CAMD data sets which are the core of this study. EPA recently made their data available via an ftp server (ftp://ftp.epa.gov/dmdnload/emissions). We also thank Kate Visser Ackerman and an anonymous reviewer; their constructive comments were key to improving the original manuscript. We thank David Fanning for providing online powerful IDL tips and the Coyote Library (http://www.dfanning.com/). The CO2FFEE members who attended the first “High resolution fossil fuel CO2 emissions” workshop at Purdue in April 2007 are as follows: Kate Ackerman (USGS), Bob Andres (DOE CDIAC), William Ansley (Purdue University), T. J. Blasing (DOE CDIAC), Kathy Corbin (CSU), Jay Gregg (UMD), Marc Fischer (LBNL), Gregory Frost (NOAA ESRL-CU CIRES), Kevin Gurney (Purdue University), Sander Houweling (SRON), Greg Marland (DOE CDIAC), Daniel Mendoza (Purdue University), Diane Pataki (UCI), and Gabrielle Pétron (NOAA ESRL-CU CIRES).