Fossil-fueled power plants as a source of atmospheric carbon monoxide

Abstract

Elevated carbon monoxide (CO) mixing ratios in excess of those derived from emissions inventories have been observed in plumes from one gas- and coal-fired power plant and three of four lignite coal-fired electric utility power plants observed in east and central Texas. Observations of elevated CO on days characterized by differing wind directions show that CO emissions from the lignite plants were relatively constant over time and cannot be ascribed to separate sources adjacent to the power plants. These three plants were found to be emitting CO at rates 22 to 34 times those tabulated in State and Federal emissions inventories. Elevated CO emissions from the gas- and coal-fired plant were highly variable on time scales of hours to days, in one case changing by a factor of 8 within an hour. Three other fossil-fueled power plants, including one lignite-fired plant observed during this study, did not emit substantial amounts of CO, suggesting that a combination of plant operating conditions and the use of lignite coal may contribute to the enhanced emissions. Observed elevated CO emissions from the three lignite plants, if representative of average operating conditions, represent an additional 30% of the annual total CO emissions from point sources for the state of Texas.

---

Introduction

Carbon monoxide (CO) is formed from incomplete fuel oxidation during both biomass burning and anthropogenic combustion processes; these two sources are estimated to contribute roughly equally to the CO budget on the global scale.^1-3 In the United States, transportation sources account for about 80%, industrial sources for 7%, and electrical utilities for only 0.5% of the total tabulated anthropogenic CO emissions.⁴

Electric utility power plant operating conditions are carefully monitored to optimize combustion efficiency, and typically operate at efficiencies exceeding 99.9% based on tabulated emissions of CO and CO₂.^5,6 Power plants are not expected to be significant sources of CO; however, as they are large sources of CO₂, small changes in fuel combustion efficiency could result in large percentage increases in CO emissions. Carbon monoxide emissions are monitored periodically by direct measurement of flue gases, but more typically are calculated using emissions factors appropriate to the burner technology and type and amount of fuel consumed.^7-9

We present the results of airborne measurements in plumes downwind of seven different power plants in eastern and central Texas. Substantial enhancements in CO mixing ratios were observed in plumes from three of four lignite coal-fired power plants. Multiple transects at different distances downwind and on different days suggest the elevated CO emissions were reasonably constant over time scales of hours to days. Another power plant with both sub-bituminous coal and natural gas-fired units was also found to be emitting substantial amounts of CO, but these emissions were highly variable over time. Carbon monoxide emissions from other coal- and gas-fired power plants were found to be in good agreement with tabulated emissions inventories.

We analyze data from near-field (2–50 km) transects downwind of each source and establish that the power plants are the sources of the observed elevated CO. We then calculate annual power plant CO emission rates from the slopes of bivariate linear least-squares regression fits¹⁰ to correlated plume NO_y and CO₂ data.¹¹ Finally, we compare CO emission rates derived from the measured plume data to those tabulated in State and Federal emissions inventories for these power plant sources.

Experiment

We report measurements taken downwind of fossil-fueled power plants in August and September 2000 from 8 research flights of the National Center for Atmospheric Research L-188 Electra aircraft, leased by the National Oceanic and Atmospheric Administration (NOAA) for the Texas Air Quality Study (TexAQS). These flights included plume transects, flown perpendicular to the prevailing wind direction in the mixed layer, at multiple distances downwind of seven gas- and coal-fired power plants in eastern and central Texas. Instrumentation aboard the Electra provided meteorological data as well as measurements of NO_y, SO₂, CO₂ and CO, which represent the major primary emissions from these plants. Precision of the 1 Hz NO_y data was limited by counting statistics and ranged from ± 0.04 to ±0.35 ppbv for the plume data presented here; accuracy is conservatively estimated at ±10%.^12,13 CO was measured aboard the aircraft by two independent methods. 1 Hz measurements were provided by a vacuum UV resonance fluorescence instrument for which measurement precision and accuracy have been previously established as ±1.5 ppbv and ±5% respectively.¹⁴ CO measurements were also provided by gas chromatographic analysis of whole-air canister samples.^15-17 A comparison of the data from the independent CO measurements made during TexAQS 2000 is shown in Fig. 1. A linear-least-squares fit to the two independently calibrated CO data sets resulted in a slope of 1.03 (r² = 0.989), demonstrating that the measurements agree within the stated experimental uncertainties. 1 Hz CO₂ measurements were made using a modified Licor non-dispersive infrared absorption instrument, which was calibrated in-flight to reference gases tied to NOAA Climate Monitoring and Diagnostics Laboratory (CMDL) standards; CO₂ measurement precision is estimated as ±0.08 ppmv and accuracy as ±0.2 ppmv. Similar plume analyses could be based on SO₂, but those data were not finalized at the time of this writing.

Fig. 1 Comparison of the independent CO measurements on-board the NCAR Electra during TexAQS 2000. The average NOAA vacuum UV resonance fluorescence measurement was averaged over the whole-air canister sample (WAS) collection time. A linear regression through the 496 coincident samples of the two data sets yields a slope of 1.03 and an r² = 0.989.

Results

The magnitude of CO concentration enhancements in power plant plumes is dependent on source CO emissions rate, atmospheric dispersion after emission, and the natural variability in levels of CO in the boundary layer outside the plume. An example of plume transect data taken downwind and south of two power plants located 18 km apart in East Texas is shown in the time series in Fig. 2a from the Electra flight of September 3, 2000. Enhancements of 160 ppbv CO were very highly correlated with plume NO_y (r² = 0.966) and plume CO₂ (r² = 0.914) downwind of the Monticello plant, while no clearly discernable elevation in CO levels was observed downwind of the Welsh plant. A graph of the relationship of CO to both NO_y and CO₂ in all of the Monticello plume transects, including repeated passes 22 and 43 km downwind, conducted on the same day is shown in Fig. 2b. A measure of reproducibility is provided by comparison of data from repeated transects of the Monticello plume 22 km downwind of the plant at constant altitude within the boundary layer over a span of 30 min on the September 3 flight (Fig. 2b). The variability between transect determinations of derived CO emission rates is ±8% using the fitted CO/NO_y slopes and ±12% using the CO/CO₂ slopes. The ratios of CO to NO_y and CO to CO₂ in the Monticello plume further downwind (Fig. 2b) indicate that the elevated CO emissions from the Monticello plant were constant over a period of at least 4 h. Enhancements of CO approaching 120 ppbv observed in the Martin Lake power plant plume later in the flight were similarly correlated with enhancements in plume NO_y (r² = 0.957) and CO₂ (r² = 0.990).

Fig. 2 (a) Time series of 1 Hz data taken in an aircraft transect 21 km downwind of the Welsh and Monticello power plants on September 3, 2000 shows enhancements of CO of over 150 ppbv in the Monticello plume, compared to background levels. (b) Linear regressions through transect data taken at 22, 43 and 65 km downwind demonstrate the consistency of the CO emissions from the Monticello plant over time. The plots of CO to NO_y and to CO₂ have similar slopes for all transects conducted on that day. These data are also shown in Fig. 3d.

We rule out the possibility of separate sources of CO immediately adjacent to the Monticello and Martin Lake plants by considering transect data from a flight 10 days later taken due west of the plants under easterly winds. Similarly high correlations of enhancements in plume CO with plume NO_y and CO₂ observed on the two flights rule out separate, nearby emitters and implicate these two power plants as consistent and substantial sources of atmospheric CO. Elevated CO emissions from two other power plants were also observed during the TexAQS field study. CO enhancements were observed in the plume from the Big Brown power plant in central Texas, and occasionally in the plumes from the W.A. Parish plant, west of the Houston metropolitan area. Visual observations of the plant sites from the air and examination of State and Federal inventories of emissions source locations^5,6 confirm that none of these power plants is co-located with other point sources of CO. The Big Brown, Monticello and Martin Lake plants are similar in that they all burn lignite coal.¹⁸ The W.A. Parish plant differs in that it has both natural gas-fired and sub-bituminous coal-fired units.

Emissions inventories suggest a CO/CO₂ of ∼0.1% or less. Ratios higher than this indicate less than optimum combustion, fuel use and heat output. Measured relationships between CO and the primary emissions of NO_y and CO₂ in near-field power plant plume transects are shown in Fig. 3. Atmospheric dispersion after emission is accounted for by interpreting fitted slopes to represent the ratios of CO to each of the co-emitted species for which hourly emissions estimates are available. Using fitted slopes of the CO and CO₂ data shown in Fig. 3, we observe average molecular ratios of CO to CO₂ for Monticello, Martin Lake and Big Brown power plants of 0.6%, 0.5% and 0.3% respectively, and a maximum CO to CO₂ ratio of 0.6% for the W. A. Parish plume. These relatively small changes in operating efficiency (for instance from 99.99% to 99.40% for Monticello) result in CO emissions many times greater than indicated in emissions inventories.

Fig. 3 Observed relationships of CO to NO_y and to CO₂ in power plant plumes. Aircraft transect data in plumes from (a) W.A. Parish; (b) Big Brown; (c) Martin Lake; (d) Monticello; (e) Limestone; (f) Welsh (circles) and Tradinghouse (crosses) are shown. Data from transects conducted on different days are denoted by different colors. Slopes of the solid lines indicate expected ratios of CO to CO₂ and CO to NO_y based on tabulated emissions inventories.

Continuous monitoring and reporting of CO emissions by electric utility power plants is not routinely performed, so a direct comparison of observed plume CO enhancements with hourly plant emissions estimates is not possible. Transect data were therefore used to derive annual CO emissions by assuming the observed CO/NO_y and CO/CO₂ emissions ratios to be representative of the yearly average emissions ratios from the plants, then multiplying by the tabulated emissions rate of NO_x or CO₂, respectively. Uncertainties in the derived CO emissions from the plants are estimated as the standard deviation of the NO_y and CO₂ derived emissions estimates and are generally less than ±15%, with Monticello the exception at ±25%. The results for the W.A. Parish plant are presented as a range of values because the CO emissions estimates varied substantially from day-to-day.

Transects of the W. A. Parish power plant plume were conducted on six different days. Derived CO emissions were indistinguishable from tabulated emissions data for three of the flights and were substantially higher than tabulated for the other three flights. An example of the variability of the CO emissions from the W. A. Parish plant is shown in Fig. 4. The figure contains two time series from a flight conducted on August 27th, when the highest CO emissions were measured in the W. A. Parish plume. The transect flown 3 km downwind of the plant showed plume CO enhancements of up to 300 ppbv which were highly correlated (r² > 0.95) with NO_y and CO₂. A subsequent transect flown 15 min later and approximately 23 km downwind of the plant showed no discernable CO enhancement. Trajectory calculations indicate a one-hour difference in plume age, suggesting rapid variation in W.A. Parish plant CO emissions over time. Examination of hourly emissions reported by the W.A. Parish plant operators indicate constant emissions from coal-fired units. Emissions of CO₂ and NO_x from the gas-fired units showed diurnal cycles approaching 100%, presumably the result of cycling the operation of the gas-fired units to meet electricity demands. We speculate that rapidly changing loads in the gas-fired units may have resulted in the observed changes in CO emissions.

Fig. 4 Time series of two transects of the W.A. Parish plume on August 27, 2000 demonstrate the variability in time of CO emissions from the plant. (A) Data from a transect 3 km downwind suggests large CO emissions, correlated with NO_y and CO₂ (r² > 0.95). (B) Data from another transect at 23 km downwind, representing emissions from an hour earlier, contained no discernable enhancement in CO and little correlation (r² < 0.50) with the other species. Based on the ratios of CO to CO₂ and CO to NO_y observed in the 3 km transect, the expected CO plume enhancement in the 23 km transect was 18 ppbv above the background level.

Comparison to emissions inventories

Tabulated emissions in Table 1 were taken from and cross-checked between existing Federal point source inventories for the 1998 and 1999 reporting years.^5,6 The data in Table 1 show that three of four lignite-fueled electric utility power plants observed in this study emitted substantially more CO than reported by the plant operators. Emissions of CO from three other plants, including one lignite-fueled plant, were not significantly different than the tabulated values. According to power plant operators, the plants were operating under normal conditions on the days that the plumes were intercepted. Assuming that the aircraft observations are representative of annual average conditions, the three lignite-burning electric utility power plants observed in this study exceeded tabulated emissions of CO by factors of 22 to 34. These emissions represent over 0.13 Tg, or 140,000 short tons (1 Tg = 10¹² g; 1 short ton = 2000 lbs), of CO in excess of inventory values emitted to the atmosphere every year.

Table 1 Tabulated and measurement-derived annual power plant CO emission rates

Flight date	Source	Primary fuel type	Tabulated annual emissions^a			Derived annual CO emissions^d	Derived/tabulated^e
			CO2^b	NOx^c	CO^c
a Emissions obtained from EPA and Texas Natural Resource Conservation Commision inventories for 1998 and 1999 reporting year. b Units of 10^12 g yr^−1. c Units of 10^9 g yr^−1. d Units of 10^9 g yr^−1; average of slopes of measured CO/CO2 and CO/NOy transect data, multiplied by the ratio of molecular weights and tabulated annual average emissions of CO2 or NOx, respectively. e The ratio of derived to tabulated CO is a measure of the excess CO emissions found in this study. f W. A. Parish transect days: August 25, 27 and 30 and September 1, 6 and 10.
09/03	Martin Lake	Lignite coal	23	31	2.0	58 ± 5	29
09/03	Monticello	Lignite coal	18	22	1.8	62 ± 16	34
09/10	Big Brown	Lignite coal	8.7	14	0.77	14 ± 1	22
09/10	Limestone	Lignite coal	14	26	1.8	1.0 ± 0.80	0.56
09/03	Welsh	Sub-bituminous coal	14	21	1.9	6.5 ± 0.91	3.4
09/10	Tradinghouse	Natural gas	3.9	17	2.6	2.0 ± 0.1	0.77
^f	W. A. Parish	Sub-bituminous coal/natural gas	25	39	5.1	3.3–73	0.64–14

---

These aircraft-derived annual CO emissions estimates suggest that the Monticello and Martin Lake power plants are the two largest CO point sources in Texas, exceeding emissions from other known point sources (e.g., aluminum smelters and carbon black plants) by nearly a factor of two. Measurements further suggest that the Big Brown power plant is the sixth-largest CO point source in the State. Elevated CO emissions from the Monticello, Martin Lake, and Big Brown lignite-fueled plants account for roughly 30% of the CO point source inventory total for the state of Texas. The increased emissions from these three plants are similar in magnitude to the on-road mobile source CO emissions for a city with a population of 1 million people.¹⁹ Not every lignite-fueled plant emitted unexpectedly high CO, however, as illustrated by the Limestone datum in Table 1, which is in quantitative agreement with the tabulated value. We conclude that the use of lignite coal is correlated with but not sufficient to produce elevated CO emissions. Comparison of differences in burner types, operating conditions, and emissions control technology between the Limestone plant and the Big Brown, Monticello and Martin Lake plants would be useful in determining the cause of high CO emitted from lignite-fueled plants.

Enhanced CO emissions from the W. A. Parish plant were found to be 14, 6 and 4 times the tabulated values on three flights, and in agreement with tabulated values observed on three other flights. Successive transects of the W. A. Parish plume on August 27th indicated that the enhanced CO emissions were highly variable on time scales of an hour or less. The variability in the CO emissions from the W. A. Parish plant appears to be related to the diurnally varying use of gas powered units.

Summary

Carbon monoxide emissions from individual electric utility power plants may be substantially larger than tabulated emissions data suggest. Elevated CO emissions were observed from three out of four lignite-fueled power plants in eastern and central Texas. These plants were found to be emitting CO at rates 22 to 34 times those tabulated in State and Federal emissions inventories. The excess annual emissions for the three plants of 0.13 Tg is equivalent to 30% of the total point source CO budget for the state of Texas. These results are in stark contrast to tabulated emissions estimates for these plants, which suggest these plants account for only about 1% of the point source emissions total. The presence of elevated CO, perhaps associated with changing loads in natural-gas-fired units, at the W.A. Parish plant is quite unexpected and more measurements are needed to better define the magnitude and time scales of emissions. Finally, the observation of high CO emissions from lignite-fueled power plants in Texas may have implications for regions where consumption of lignite coal constitutes a large fraction of total fuel usage, such as parts of Eastern and Western Europe.^20,21

References

1. M. A. K. Khalil and R. A. Rasmussen, Nature, 1988, 332, 242.
2. Global Emissions Inventory Activity (GEIA), http://weather.engin.umich.edu/geia/ (assessed July 2001).
3. Emission Database for Global Atmospheric Research (EDGAR), http://www.rivm.nl/env/int/coredata/edgar/ (assessed July 2001).
4. National Air Quality and Emissions Trends Report: 1998, United States Environmental Protection Agency, Research Triangle Park, 2000.
5. EPA, Emissions & Generation Resource Integrated Database (E-GRID), http://www.epa.gov/airmarkets/egrid/ (assessed July 2001).
6. Texas Natural Resource Conservation Commission Point Source Air Emissions Inventory, http://www.tnrcc.state.tx.us/air/aqp/psei.html (assessed July 2001).
7. Compilation of Air Pollutant Emissions Factors, United States Environmental Protection Agency, Research Triangle Park, NC, 1998.
8. Emission Factor Documentation for AP-42 Section 1.7, Lignite Combustion, United States Environmental Protection Agency, Research Triangle Park, NC, 1998.
9. Emission Factor Documentation for AP-42 Section 1.1, Bituminous and Sub-bituminous Coal Combustion, United States Environmental Protection Agency, Research Triangle Park, NC, 1998.
10. W. H. Press, S. A. Teukolsky, W. T. Vetterling and B. P. Flannery, Numerical Recipes in C, Cambridge University Press, 1992.
11. T. B. Ryerson, M. Trainer, J. S. Holloway, D. D. Parrish, L. G. Huey, D. T. Sueper, G. J. Frost, G. Hübler, J. M. Meagher and F. C. Fehsenfeld, Science, 2001, 292, 719.
12. T. B. Ryerson, L. G. Huey, K. Knapp, J. A. Neuman, D. D. Parrish, D. T. Sueper and F. C. Fehsenfeld, J. Geophys. Res., 1999, 104, 5483.
13. T. B. Ryerson, E. J. Williams and F. C. Fehsenfeld, J. Geophys. Res., 2000, 105, 26–447.
14. J. S. Holloway, R. O. Jakoubek, D. D. Parrish, C. Gerbig, A. Volz-Thomas, S. Schmitgen, A. Fried, B. Wert, B. Henry and J. R. Drummond, J. Geophys. Res., 2000, 105, 24–251.
15. F. Flocke, E. L. Atlas, S. Madronich, S. M. Schauffler, K. Aikin, J. J. Margitan and T. P. Bui, Geophys. Res. Lett., 1998, 25, 1891.
16. L. E. Heidt, J. F. Vedder, W. H. Pollock, R. A Lueb and B. E. Henry, J. Geophys. Res., 1989, 94, 11–599.
17. S. M. Schauffler, E. L. Atlas, D. R. Blake, F. Flocke, R. A. Lueb, J. M. Lee-Taylor, V. Stroud and W. Travicek, J. Geophys. Res., 1999, 104, 21–513.
18. September 2000 Fuel Reports, Public Utility Commission of Texas, Austin, TX, 2000.
19. Carbon Monoxide State Implementation Plan, Las Vegas Valley Nonattainment Area, Clark County Board of Commissioners, Clark County, NV, 2000, Appendix A, Section 7.
20. Electrical Power Generation Sector in Poland, US Department of Commerce, Central and Eastern Europe Business Information Center, 1999.
21. International Energy Outlook, US Department of Energy, Energy Information Administration, DOE/EIA-0484, 2000.

---

Footnotes

† Also at Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, USA.

‡ Now at Ball Aerospace, Boulder, Colorado, USA.

---