Article pubs.acs.org/est
Characterizing Fugitive Methane Emissions in the Barnett Shale Area Using a Mobile Laboratory Xin Lan,* Robert Talbot, Patrick Laine,† and Azucena Torres Institute for Climate and Atmospheric Science, Department of Earth & Atmospheric Sciences, University of Houston, Houston, Texas 77204, United States S Supporting Information *
ABSTRACT: Atmospheric methane (CH4) was measured using a mobile laboratory to quantify fugitive CH4 emissions from Oil and Natural Gas (ONG) operations in the Barnett Shale area. During this Barnett Coordinated Campaign we sampled more than 152 facilities, including well pads, compressor stations, gas processing plants, and landfills. Emission rates from several ONG facilities and landfills were estimated using an Inverse Gaussian Dispersion Model and the Environmental Protection Agency (EPA) Model AERMOD. Model results show that well pads emissions rates had a fat-tailed distribution, with the emissions linearly correlated with gas production. Using this correlation, we estimated a total well pad emission rate of 1.5 × 105 kg/h in the Barnett Shale area. It was found that CH4 emissions from compressor stations and gas processing plants were substantially higher, with some “super emitters” having emission rates up to 3447 kg/h, more then 36,000-fold higher than reported by the Environmental Protection Agency (EPA) Greenhouse Gas Reporting Program (GHGRP). Landfills are also a significant source of CH4 in the Barnett Shale area, and they should be accounted for in the regional budget of CH4.
■
INTRODUCTION Methane is an important trace gas in the atmosphere. It is a strong greenhouse gas with global warming potential as high as 28−34 times that of carbon dioxide (CO2) using a 100-year integration time scale.1,2 Currently CH4 contributes approximately one-third of the total radiative forcing of carbon dioxide.3 It plays a significant role in air quality by acting as a major sink of the hydroxyl radical (OH),4−7 the dominant oxidant that defines the atmospheric oxidation capacity, which has significant impacts on the chemical transform of carbon monoxide (CO), sulfur dioxide (SO2), nitrogen oxides (NOx), ammonia (NH3), and tropospheric ozone (O3).8 Atmospheric CH4 can be released from both natural sources, such as wetlands and methane hydrates, and anthropogenic sources, such as natural gas systems and rice cultivation.4,9 Anthropogenic emissions account for 50−60% of the total global CH4 budget.10 Methane is the primary component of natural gas, and it accounts for ∼77% to 99% of total natural gas as a mole fraction.11,12 Natural gas is a relatively clean burning energy source because it produces more energy per carbon dioxide molecule formed than coal and oil (177% and 140%, respectively) combustion.12 In the past decade, the horizontal-drilling and hydraulic-fracturing techniques have led to a boom in natural gas production. However, CH4 emissions associated with the production and transmission of natural gas have raised concerns from several parties. © 2015 American Chemical Society
The U.S. Environmental Protection Agency (EPA) estimated that 25% of CH4 emissions in the U.S. are attributed to natural gas and petroleum systems.13 However, the EPA estimates are primarily based on self-reported emission rates, which frequently utilize emission factors and assume consistency throughout different regions of the U.S. Thus, these estimates are highly uncertain and probably do not reflect fugitive CH4 emissions from current practices. Several studies reported that current inventories, including the EPA Greenhouse Gas Reporting Program (GHGRP) and the Emissions Database for Global Atmospheric Research (EDGAR), are underestimating methane emissions from the ONG industry.14−21 It was further reported that the current EPA database underestimates methane emissions in the south-central U.S., including the Barnett Shale area, by a factor of ∼2.7.19 Studies by the Environmental Defense Fund (EDF) suggested that if the total CH4 emissions are greater than ∼2.7% of production, the immediate net radiative forcing for natural gas use is worse than that for coal use as a source for electricity (note that the 2.7% breakeven point is sensitive to the calculation methods and measurements of emission factors of fuels).22,23 Further Received: Revised: Accepted: Published: 8139
December 31, 2014 March 23, 2015 April 13, 2015 July 7, 2015 DOI: 10.1021/es5063055 Environ. Sci. Technol. 2015, 49, 8139−8146
Article
Environmental Science & Technology
Stationary samplings were conducted 450 ppmv) in both plumes, a good correlation between CO2 and CH4 was observed from facility a but not facility b. The coincident CO2 and CH4 signals suggest a significant influence of combustion, which was probably related to engine exhaust. For those ONG facilities with good correlations between CO2 and CH4, the enhancement ratios of CO2 and CH4 (ΔCO2/ ΔCH4) were greatly variable, ranging from 5 to 25 for large gas facilities. The lack of correlation between CO2 and CH4, on the other hand, may suggest fugitive emissions from other components/process, instead of only compressor engines, within these facilities. Emission Rate Estimates. The emission rates of well pads were estimated using the Gaussian Dispersion Model. Model results showed a large range of CH4 emission rates, from 0.009 to 58 kg/h. A histogram of well pads emission rates (Figure 4)
uncertainties, sensitivity tests were conducted. More details concerning the uncertainty estimates and sensitivity test results are described in the SI.
■
RESULTS AND DISCUSSION General Characteristics. Figure 1 shows the entire 3,700 km of mobile laboratory paths, covering 9 counties in the Barnett Shale area. Albeit the majority of our measurements were taken in daytime, nighttime measurements were also conducted in the first 3 days of this campaign. It was found that the influence of diurnal PBL height and stability variation was substantial on the regional background CH4 levels. We made two transits at different times of a day on a highway, which has a large amount of well pads on both sides. During these two transits, no significant differences were observed for wind direction, wind speed, or mobile lab speed (∼90 km/h); however, the background CH4 levels were up to 0.1 ppmv higher during the transit before sunrise than midmorning (see Figure S1, Supporting Information). The low PBL heights and increased stability at night contributed to a higher CH4 level by suppressing the mixing of CH4. Overall, the daytime (10:00 to 17:00 LST) CH4 measurements covered a range of 1.89 ppmv (10th percentile) to 2.92 ppmv (90th percentile) for the 1−1.5 s data. The lower bound of this range is slightly higher than the concurrent background CH4 levels of the Northern Hemisphere, based on data collected at the Mauna Loa site (∼1.836 ppmv, the monthly average for October, 2013) (http://www.esrl.noaa.gov/gmd/ dv/iadv/graph.php?code=MLO&program=ccgg&type=ts). The highest CH4 measured during this campaign was 90 ppmv (instantaneous reading with original time resolution), which was when our mobile laboratory was located in the pollution plume of a compressor station. Figure 2 shows all the large CH4 spikes (defined as CH4 >2.2 ppmv) that we observed during the campaign. It is clear that different meteorological conditions, such as wind and PBL height, can influence CH4 concentration; however, the large CH4 peaks obviously resulted from significant nearby emission sources. The widely scattered CH4 peaks demonstrated the large spatial distribution of ONGs facilities with high CH4 emissions. CO2 versus CH4. Besides CH4, CO2 was also measured during this campaign. Significant elevation of CO2 levels were observed from some selected ONG facilities. The maximum observed CO2 level was 524 ppmv during the mobile lab survey, from a location downwind of a gas processing plant (see Figure 3a). For a few compressor stations and gas plants, the increase in CO2 mixing ratio was less than 10 ppmv. Overall,
Figure 4. Histogram of all estimated well pad emission rates.
showed a fat-tailed distribution. In total, 70% of well pads had emission rates less than 5 kg/h. The calculated uncertainties of these emission rate estimates were in the range of −28% to 81% for different well pads. We were able to obtain detailed information, including oil/condensate production, gas production, water production, and the age of the wells (http://www. didesktop.com/), for these well pads. We found weak linearcorrelations between well pad emission rates based on dispersion model inversions and gas production (Figure 5). This relationship is statistically significant (p = 0.05) with a correlation coefficient R equal to 0.57, suggesting that the more gas a well produces, the more CH4 it releases, which is consistent with a previous study.29 The total gas production in 8142
DOI: 10.1021/es5063055 Environ. Sci. Technol. 2015, 49, 8139−8146
Article
Environmental Science & Technology
models. The emission rates estimated by the Gaussian Dispersion Model ranged from 19 to 3447 kg/h (Table 1). The minimum and maximum emission rates from AERMOD were 21 kg/h for a two engines compressor station (#3) and 2119 kg/h for a ten engine compressor station (#4). The uncertainties for both models are about a factor of 3, which is comparable for uncertainty estimates reported for several dispersion models.23,24 Although no experiment was conducted during this campaign to evaluate the application of dispersion models and mobile laboratory measurements, a recent study using similar approaches to access CH4 emissions has conducted controlled release experiments for validation and reported accuracy values ranging from −87% to 184% of actual emissions.29 On-site measurements were also conducted for the two engines compressor station (#3) by the University of West Virginia group during this campaign. They found that the CH4 emission rate of this facility from direct measurements was 20.7 kg/h,31 which was very close (within ∼10%) to our estimates from both Gaussian method (19 kg/h) and AERMOD (21 kg/ h). Figure 6 shows an example of an AERMOD simulation for a compression station (#4). This facility was measured twice, on Oct. 22 and Oct. 31. Both model results showed much larger emission rates for the first day (Table 1). A burning flare was observed during our first visit to this facility but not in the second visit. A similar situation also occurred in the gas processing plant, where a few flares were observed during our second visit to the site but not on the first visit. These flares corresponded to a much higher emission rate (Table 1). Although flaring is considered as an efficient approach to remove CH4 and VOCs (if properly operated), burning flares can be an indication of different operational status, which may involve a change in engine loading. This gas processing plant was also measured by other groups during this campaign,32,33 and a comparison is provided in Lavoie at al.32 It is possible that the temporal variability of emission rates for a large ONG facility can be as much as a factor of 2, although better constrains on the overall uncertainty from dispersion models are needed to further validate this point.
Figure 5. Well pad estimated emission rate versus gas production.
the Barnett Shale area was reported to be 5,343 million cubic feet per day (mmcf/day) in 2013 (http://www.rrc.state.tx.us/ media/22204/barnettshale_totalnaturalgas_day.pdf). Using this linear relationship we can estimate the total well pad emissions to be 1.5 × 105 kg/h (uncertainty range: 1.16 × 105− 2.04 × 105 kg/h). Please note that other factors, such as the operation and maintenance status, can also influence CH4 emissions, which unfortunately cannot be evaluated in this study due to a lack of necessary information. We also estimated the proportional loss rates for these well pads, which equals to the emission rate divided by the gas production. Most results fell in the range of 0.01% to 47.8% (see Figure S2, Supporting Information), and the median value and average value were 2.1% and 7.9%, respectively. The proportional loss rates and “super emitters” of production sites measured during the EDF coordinated campaign are discussed further in a companion paper by Zavala-Araiza et al.30 For compressor stations and gas processing plants, AERMOD was also applied to estimate emission rates, in addition to the Gaussian Dispersion Model. These compressor stations used mostly reciprocating engines, with the number ranging from two to ten (information provided in facility Title V files). Table 1 shows the emission rates estimated from both
Table 1. Estimated Emission Rates for Six Compressor Stations and One Gas Processing Plantb
no. of engine
max. CH4 level in plume (ppmv)
max. 10 min average CH4 (ppmv)
Gaussian estimate (kg/h)
Gaussian uncertainty (%)
AERMOD estimate (kg/h)
AERMOD uncertainty (%)
2013 GHGRP (kg/h)
AERMOD/ GHGRP ratio
CH4 emissions estimated from VOCs composition (kg/h)
AERMOD/CH4 emissions estimated from VOCs composition
date
sourcea
17Oct 23Oct 31Oct 22Oct 31Oct 24Oct 24Oct 24Oct 29Oct
CS#1
7
22.68
12.57
323
−32/44
374
−15/70
0.15
2520
71
5.3
CS#2
6
89.6
43.0
3447
−67/249
1860
−29/86
0.05
36208
44
42.3
CS#3
2
3.25
2.25
19
−15/197
21
−37/273
N/A
N/A
93
0.2
CS#4
10
52.3
26.89
809
−63/38
2119
−18/39
0.13
16283
36
58.9
CS#4
10
15.38
7.52
484
−83/151
226
−21/109
0.13
1737
36
6.3
CS#5
6
25.20
13.59
352
−77/167
238
−8/101
0.06
4170
51
4.7
CS#6
4
2.51
2.37
42
−66/295
49
−37/169
N/A
N/A
15
3.3
GPP
∼38
3.66
2.7
287
−17/25
746
−7/52
115.4
6.5
320
2.3
GPP
∼38
6.71
5.89
986
−33/153
1723
−9/105
115.4
14.9
320
5.4
a CS represents compressor stations; GPP represents gas processing plants. bCompressor stations without GHGRP emission rates (indicated as N/ A) presumably have CH4 emission less than 25,000 metric tons of CO2 equivalents.
8143
DOI: 10.1021/es5063055 Environ. Sci. Technol. 2015, 49, 8139−8146
Article
Environmental Science & Technology
Figure 6. AERMOD simulation for compressor station #4. The color scale shows the concentration of CH4 in μg/m3. The blue lines depict building structures associated with possible downwash effects. Red crosses represent engines. The blue square mark locates our mobile laboratory during stationary measurements.
(reported in the Title V files) by the average CH4 composition to VOC composition in the natural gas stream of the Bend Arch-Fort Worth Basin (https://www.tceq.texas.gov/assets/ public/implementation/air/am/contracts/reports/ei/ 5821199776FY1211-20121031-ergi-condensate_tank.pdf). The mass ratio of CH4/VOCs was calculated to be 6.734, excluding the C8 + Heavy VOCs (0.03% in volume) because the exact chemical compositions were not documented. The resulting CH4 emission rates are presented in Table 1. The ratios of AERMOD estimates versus to these estimates were in the range of 0.2 to 58.9, much smaller than the AERMOD/GHGRP ratios. This could be an indication that both these estimates and AERMOD results are more credible to represent actual emissions. This rationing method assumes that the released CH4 has the same composition as the original gas stream. It may cause errors in estimating CH4 emissions from some components, such as dehydrators and storage tanks, since the CH4 composition in the associated leaking gas may change greatly after processing. However, it is possibly a valid assumption for CH4 emissions from engines, which accounts for the majority of CH4 emissions.31 Note that the VOC emissions reported in the Title V files were under controlled conditions for engines and glycol dehydrator emissions (for flash tanks and regenerators). If the VOC control was not well maintained, then the estimated CH4 emissions should be higher and even closer to the AERMOD estimates. Methane plumes from landfills have large spatial dimensions, due to the large size (on the order of 1 km) of the emission areas. On Oct. 19, nine landfills were visited. The minimum CH4 mixing ratio (from 1 to 1.5 s data) observed at these nine landfills was 3.25 ppmv, while the maximum was up to 14.76 ppmv. Four landfill signals were chosen for emission rate estimation using the multipoint Gaussian Dispersion Model. Model results yield relatively large CH4 emission rates, ranging from 86 to 2087 kg/h (Table 2). The uncertainty in estimating
Considering the complexity of source term (emission rate) estimates using inversion methods, it is encouraging to find that model results from the Gaussian Dispersion Model and AERMOD agreed with each other at a reasonable level. Both modeled source emission estimates were bound or close to the factor-of-two lines in the scatter plot (see Figure S3, Supporting Information). The differences of two model results are mostly within model uncertainty ranges. However, emission rates estimated by AERMOD may be more credible since it is a stateof-art dispersion model currently recommended by EPA for these types of industrial sources. It is also noticeable that both model results are much higher (at a magnitude of 104, see Table 1) than the emission rates reported in the 2013 GHGRP. This result is in rough agreement with Lavoie et al.32 using an aircraft-based mass balance approach, which reported the ratio of measurements to GHGRP to be 9,280−11,195 for a compressor station. The GHGRP data is self-reported annual emission rates by facilities under regular operations as requested by the U.S. EPA. Our results suggest the possibility that the 2013 GHGRP has greatly underestimated emission rates (up to a magnitude of 104 times) from individual compressor stations and gas processing plants. However, this study and the study by Lavoie et al.32 represent an estimate of emissions based on a very short sample collection interval, while the GHGRP represents an average annual estimate of emissions. There is a possibility that our measurements are higher due to our accounting for interim or episodic emissions, in addition to long-term emissions. As a result, our estimates may not be directly comparable to the GHGRP. The changing gas demand, machines malfunctioning, periodic leaking, and liquids unloading can also cause large temporal variability in CH4 emissions, which also may not be fully included in the GHGRP. To provide a better comparison, we also estimated CH4 emission rates by rationing the hourly VOC emission rates 8144
DOI: 10.1021/es5063055 Environ. Sci. Technol. 2015, 49, 8139−8146
Article
Environmental Science & Technology
Table 2. Estimated Emission Rates (Using the Gaussian Dispersion Model) versus Emission Rates from 2013 GHGRP for Landfills date 19-Oct 19-Oct 19-Oct 25-Oct
source
max. CH4 level in plume (ppmv)
max. 10 min averaged CH4 (ppmv)
Gaussian estimate (kg/h)
2013 GHGRP (kg/h)
4.16
2.69
562
658
landfill #1 landfill #2 landfill #3 landfill #4
uncertainty (%)
Gaussian/GHGRP ratio
−48/55
0.9
3.25
2.39
86
296
−57/−31
0.3
14.76
10.82
2087
1045
−22/19
2.0
4.51
3.27
217
153
−57/13
1.4
a
a
This emission rate was reported by the 2012 GHGRP instead of 2013 GHGRP, because no emission rate was reported in the 2013 GHGRP for this facility. Averyt, K., Tignor, M. M. B., Miller, H. L., Chen, Z. L., Eds.; Cambridge Univ. Press: New York, 2007; pp 129−234. (2) Shindell, D. T.; Faluvegi, G.; Koch, D. M.; Schmidt, G. A.; Unger, N.; Bauer, S. E. Improved Attribution of Climate Forcing to Emissions. Science 2009, 326 (5953), 716−718. (3) The NOAA annual greenhouse gas index. http://www.esrl.noaa. gov/gmd/aggi/ (accessed Apr 13, 2015). (4) Fung, I.; John, J.; Lerner, J.; Matthews, E.; Prather, M.; Steele, L. P.; Fraser, P. J. 3-Dimensional model synthesis of the global methane cycle. J. Geophys. Res.: Atmos. 1991, 96 (D7), 13033−13065. (5) Hein, R.; Crutzen, P. J.; Heimann, M. An inverse modeling approach to investigate the global atmospheric methane cycle. Global Biogeochem. Cycles 1997, 11 (1), 43−76. (6) Lelieveld, J.; Crutzen, P. J.; Dentener, F. J. Changing concentration, lifetime and climate forcing of atmospheric methane. Tellus, Ser. B 1998, 50 (2), 128−150. (7) Spivakovsky, C. M.; Logan, J. A.; Montzka, S. A.; Balkanski, Y. J.; Foreman-Fowler, M.; Jones, D. B. A.; Horowitz, L. W.; Fusco, A. C.; Brenninkmeijer, C. A. M.; Prather, M. J.; Wofsy, S. C.; McElroy, M. B. Three-dimensional climatological distribution of tropospheric OH: Update and evaluation. J. Geophys. Res.: Atmos. 2000, 105 (D7), 8931− 8980. (8) Atmospheric chemistry and physics, from air pollution to climate change; Seinfeld, J. H., Pandis, S. N., 2nd Eds.; Wiley-Interscience: U.S., 2006; pp 21−55. (9) Wuebbles, D. J.; Hayhoe, K. Atmospheric methane and global change. Earth-Sci. Rev. 2002, 57 (3−4), 177−210. (10) Kirschke, S.; Bousquet, P.; Ciais, P.; Saunois, M.; Canadell, J. G.; Dlugokencky, E. J.; Bergamaschi, P.; Bergmann, D.; Blake, D. R.; Bruhwiler, L.; Cameron-Smith, P.; Castaldi, S.; Chevallier, F.; Feng, L.; Fraser, A.; Heimann, M.; Hodson, E. L.; Houweling, S.; Josse, B.; Fraser, P. J.; Krummel, P. B.; Lamarque, J. F.; Langenfelds, R. L.; Le Quere, C.; Naik, V.; O’Doherty, S.; Palmer, P. I.; Pison, I.; Plummer, D.; Poulter, B.; Prinn, R. G.; Rigby, M.; Ringeval, B.; Santini, M.; Schmidt, M.; Shindell, D. T.; Simpson, I. J.; Spahni, R.; Steele, L. P.; Strode, S. A.; Sudo, K.; Szopa, S.; van der Werf, G. R.; Voulgarakis, A.; van Weele, M.; Weiss, R. F.; Williams, J. E.; Zeng, G. Three decades of global methane sources and sinks. Nat. Geosci. 2013, 6 (10), 813−823. (11) Zumberge, J.; Ferworn, K.; Brown, S. Isotopic reversal (‘rollover’) in shale gases produced from the Mississippian Barnett and Fayetteville formations. Mar. Petrol. Geol. 2012, 31, 43−52. (12) North American Energy Standards Board. https://www.naesb. org//pdf2/wgq_bps100605w2.pdf (accessed Apr 13, 2015). (13) Inventory of U.S. greenhouse gas emissions and sinks: 1990−2011; U.S. Environmental Protection Agency. http://epa.gov/ climatechange/Downloads/ghgemissions/US-GHG-Inventory-2013Main-Text.pdf (accessed Apr 13, 2015). (14) Petron, G.; Frost, G.; Miller, B. R.; Hirsch, A. I.; Montzka, S. A.; Karion, A.; Trainer, M.; Sweeney, C.; Andrews, A. E.; Miller, L.; Kofler, J.; Bar-Ilan, A.; Dlugokencky, E. J.; Patrick, L.; Moore, C. T., Jr.; Ryerson, T. B.; Siso, C.; Kolodzey, W.; Lang, P. M.; Conway, T.; Novelli, P.; Masarie, K.; Hall, B.; Guenther, D.; Kitzis, D.; Miller, J.; Welsh, D.; Wolfe, D.; Neff, W.; Tans, P. Hydrocarbon emissions
landfill emissions, based on Monte Carlo analysis, is
