Article pubs.acs.org/est
Atmospheric Mercury in the Barnett Shale Area, Texas: Implications for Emissions from Oil and Gas Processing Xin Lan,* Robert Talbot, Patrick Laine,† Azucena Torres, Barry Lefer, and James Flynn Institute for Climate and Atmospheric Science, Department of Earth and Atmospheric Sciences, University of Houston, Houston, Texas 77004, United States S Supporting Information *
ABSTRACT: Atmospheric mercury emissions in the Barnett Shale area were studied by employing both stationary measurements and mobile laboratory surveys. Stationary measurements near the Engle Mountain Lake showed that the median mixing ratio of total gaseous mercury (THg) was 138 ppqv (140 ± 29 ppqv for mean ± S.D.) during the June 2011 study period. A distinct diurnal variation pattern was observed in which the highest THg levels appeared near midnight, followed by a monotonic decrease until midafternoon. The influence of oil and gas (ONG) emissions was substantial in this area, as inferred from the i-pentane/npentane ratio (1.17). However, few THg plumes were captured by our mobile laboratory during a ∼3700 km survey with detailed downwind measurements from 50 ONG facilities. One compressor station and one natural gas condensate processing facility were found to have significant THg emissions, with maximum THg levels of 963 and 392 ppqv, respectively, and the emissions rates were estimated to be 7.9 kg/yr and 0.3 kg/yr, respectively. Our results suggest that the majority of ONG facilities in this area are not significant sources of THg; however, it is highly likely that a small number of these facilities contribute a relatively large amount of emissions in the ONG sector.
■
tons per year.10,11 It is reported that the averaged mercury concentration in crude oil and natural gas condensate is in the ppbm (μg/kg) level, and normally falls in the range of 1−10 ppbm, although mercury concentration is greatly variable at different locations.10,12,13 The dominant chemical forms of mercury in crude oil and natural gas condensate are reported to be elemental mercury and ionic mercury.14 Mercury in crude oil and natural gas condensate can cause equipment corrosion, catalysis poisoning, and even catastrophic equipment failures when high levels accumulate in the processing facility.14,15 However, a large uncertainty exists in estimating atmospheric mercury emissions from ONG industrial activities due to limited measurements and a large diversity in emission factor estimations.12 Additional work is needed to better quantify mercury emissions from ONG industries. The Barnett Shale area is a vast rock formation that sits in the 13 000 km2 area to the west, south, and north of the Dallas−Fort Worth metroplex. It is one of the largest natural gas fields in U.S., with sufficient amounts of oil storage for commercial production. In recent years, advanced drilling techniques have allowed gas to be extracted from underground shale rock formations. The
INTRODUCTION Mercury is well-established as a toxic global pollutant.1,2 It is introduced into the atmosphere mainly in the form of elemental mercury (Hg0)(>95%), which has a lifetime of ∼1 year and can thus be transported over long distances prior to deposition.3,4 Elemental mercury can be oxidized to form gaseous oxidized mercury (GOM, possibly consisting of HgCl2 + HgBr2 + HgOBr + ..., the exact chemical composition is still unknown) and particulate-bound mercury (PBM). GOM and PBM typically represent less than 5% of total atmospheric mercury, and these forms of mercury have much shorter lifetimes, from hours to days for GOM and days to weeks for PBM.5,6 In this study, we report measurements of total gaseous mercury, which is the sum of Hg0 and GOM. Mercury is a trace component in all fossil fuel sources including coal, petroleum, and natural gas condensate. The use of fossil fuel provides the major pathway to mobilize mercury from its lithospheric reservoir to the atmosphere. The total annual mercury emissions in United States is 47 t, of which 23 t is contributed from coal-fired power plants emissions, according to the 2011 EPA National Emission Inventory (NEI). Mercury levels in coal vary between 0.01 and 1.5 ppmm (mg/kg) from different geographic regions.7,8 Albeit mercury emitted from the oil and gas (ONG) industry is estimated to be significantly less than that from coal fired power plants,9 the total mercury emissions from ONG industry is still estimated to be a few metric © 2015 American Chemical Society
Received: Revised: Accepted: Published: 10692
September 29, 2014 July 8, 2015 July 28, 2015 July 28, 2015 DOI: 10.1021/acs.est.5b02287 Environ. Sci. Technol. 2015, 49, 10692−10700
Article
Environmental Science & Technology
Figure 1. Location of EML sites and mobile laboratory tracks (red solid line).
located just northwest of Fort Worth, TX. This site is located in the geographical center of Barnett Shale natural gas production region.22 The sampling site is situated in a rural, residential, and agricultural setting immediately surrounded by prairie land and well pads. It is approximately 40 km northwest of downtown Fort Worth and 75 km west/northwest of Dallas, TX. Figure 1 depicts the location of the EML and all tracks of the mobile laboratory survey, with the locations denoted of reported mercury point sources from the U.S. Environmental Protection Agency (EPA) NEI, and midstream ONG facilities, landfills, and electricity generators from the EPA Greenhouse Gas Reporting Program (GHGRP). In the 2011 EML campaign, we conducted continuous measurements of important air pollutants for one month, including total gaseous mercury (THg), CO, SO2, O3, NO, and NOx. Total gaseous mercury data were obtained with 2.5-min time resolution by employing a Tekran 2537A instrument, which measures Hg0 and GOM with a cold vapor atomic fluorescence (CVAF) spectrometer. The instrument has a detection limit of ∼10 ppqv (1 ng m−3 = 112 ppqv). Calibration was verified every 72 h using the permeation tube housed inside the instrument. The internal standard was verified at the beginning of this campaign using a Tekran 2505 unit. The accuracy of the mercury measurement was estimated to be ±5%. The THg inlet, consisting of approximately 5 m of Silico Steel Sulfinert tubing (1/4 in. OD, see Supporting Information (SI) for more details on the tubing), was situated about 10 m above ground level. An inline PTFE filter (1 μm size) was used to remove aerosols from the sampled air stream (refer to Lan et al.23 for more details concerning the Tekran instrument). Measurement details for CO, SO2, O3, NO, NOx and meteorological factors, including the instrument setup, calibration, and uncertainty, are provided in the SI. In October 2013, we conducted a 15-day (10/15/2013−10/ 31/2013) mercury measurement campaign in the Barnett Shale area by employing a mobile laboratory. This mobile laboratory is a four-door Chevrolet Silverado truck with a camper shell, which is outfitted with a suite of trace gas instruments, a GPS device,
number of natural gas wells and oil wells in this area has increased rapidly. To date there are 17 351 gas wells and 1018 oil wells reported in the Barnett Shale formation,16 together with compressor stations and natural gas plants, forming a large ONG industrial network. The fugitive emissions from these industrial components have led to public concern. A preliminary study of natural gas operations in the Denver−Julesburg Basin reported 4−8% CH4 loss from processing.17 Additionally, Howarth et al.18,19 have conducted a comprehensive analysis of greenhouse gases (CH4, in particular) emitted from shale gas as a result of hydraulic fracturing and they estimated 3.6−7.9% of all natural gas mined from shale formations leaks to the atmosphere. A recent study in the Haynesville, Fayetteville, and northeastern Marcellus shale gas production regions has reported loss rates from production operations in the range of 0.18−2.8%.20 It is important to investigate the possible emissions of additional (and potentially harmful) gases, i.e., atmospheric Hg, from the ONG extraction and processing systems. To the best of our knowledge, this study represents the first report of atmospheric mercury measurements in the Barnett Shale region. This study aims to provide important information on ambient mercury in a heavy industrial environment with large ONG operations. Because mercury emissions from several sources are poorly quantified in current emissions inventories, the U.S. National Academy of Sciences (NAS) 21 has recommended efforts be initiated immediately to improve the emission inventory for mercury sources to the atmosphere. Our study supports the NAS initiative directly and contributes to enriching the emission inventory database for atmospheric mercury. It will help estimate the regional mercury budget, and further assist regional modeling and policy-making processes.
■
MATERIALS AND METHODS Trace gases and meteorological data in this study were from two major sampling campaigns: 2011 stationary site measurements and 2013 mobile laboratory measurements. In June 2011 (5/28/ 2011−6/30/2011), air was monitored continuously at Eagle Mountain Lake (EML) (32.989° N, 97.471° W, 226 m elevation) 10693
DOI: 10.1021/acs.est.5b02287 Environ. Sci. Technol. 2015, 49, 10692−10700
Article
Environmental Science & Technology and five video cameras (located on the left/right/front/back/top of the truck). During this campaign, we made online measurements of THg and meteorological factors using the same instruments listed above. In addition, we used two Picarro cavity ring down instruments (G2201-i and G2132-i) to measure CO2 and CH4. Measurement details are further discussed in the SI. We made ∼3700 km of on-road measurements of THg during this campaign (see Figure 1 for mobile lab track). The number of facilities with detailed mercury measurements was 50, with 34 well pads, 13 compressor stations, 2 gas plants, and 1crude oil storage facility. During the mobile lab campaign, only two ONG facilities were found to have plumes with significant increases of THg mixing ratios over ambient levels. The American Meteorological Society−Environmental Protection Agency model AERMOD was employed to estimate the emission rates of THg from these facilities. Details of this method were described in Lan et al.24 In this study, the AERMOD View (Lakes Environmental) software with graphical user interface was utilized. The uncertainties of model results were evaluated using the Monte Carlo probabilistic uncertainty methods, which include simultaneously random and independent variations of all significant meteorological inputs.21 One hundred Monte Carlo runs were conducted for each AERMOD simulation to generate the 95% significant interval of the emission rate estimates. More details concerning the model setting and uncertainty estimates are described in the SI. Please note that the 2013 mobile laboratory survey was part of the Barnett Coordinated Campaign, which focused on methane emissions. An analysis of the 2013 mobile laboratory survey CH4 data was reported separately by Lan et al.24
pollutants and thus lowered the THg levels. The averaged wind speed was 7.9 m/s (scalar average), while the wind directions were dominated by south and southeast winds (Figure S1). The 10th percentile and 90th percentile majority of THg mixing ratios were 107 and 169 ppqv, respectively. The maximum mixing ratio occurred during the night of June 4 (at 11:10 pm), reaching up to 756 ppqv; the minimum observed level was 73 ppqv. Elevated THg levels higher than 300 ppqv were detected on 4 days in the first half of study period at about 11:00 pm to 1:00 am local time (LT). A clear diurnal variation pattern (Figure 3) was observed over the entire study period at EML site, where THg exhibited minimum mixing ratios in the afternoon (∼3:00 PM LT) and peaked near midnight. The amplitudes of the monthly average diurnal variations were large, with peak-to-valley difference exceeding 40 ppqv. This pattern was anticorrelated with the diurnal variations of wind speed (R = −0.81, P < 0.05) and PBL height (R = −0.56, P < 0.05). Low wind speed and low PBL height at nighttime facilitated the accumulation of air pollutants and thus increased the mixing ratios of THg. The THg levels started decreasing gradually after midnight, suggesting a nighttime deposition process. Because wet deposition was not frequent due to the limited amount of dew and rainfall during our studied period, dry deposition of elemental mercury was likely responsible for the monotonic decrease in THg mixing ratios after midnight. In the period between midnight and 7:00 am LT, when the PBL height and wind speed were relatively constant, the decrease rate of THg is calculated to be ∼3 ppqv/h. Thus, the deposition velocity is estimated to be 0.3 cm s−1 (deposition velocity = deposition rate × height/average mixing ratios of THg = 3 ppqv/h × 600 m/150 ppqv) under the stable nocturnal boundary layer (monthly averaged height was ∼600 m). If we assume the decrease rate is due to dry deposition only, it is comparable to the dry deposition rate (0.4 cm s−1) reported by Lindberg et al.26 over dry forest canopies and the 0.2 cm s−1 reported for the New England area.27 Besides direct dry deposition, it is possible that the decrease rate also incorporated THg loss in chemical oxidation, resulting from THg transformation to other chemical forms, such as GOM and PBM. However, chemical loss was not likely an important contributor since GOM and PBM levels at inland−rural sites are relatively low.27 Nighttime depletion in THg mixing ratios was observed at many different locations, such as several rural sites in the U.S. Atmospheric Mercury Network28 and the Canadian Atmospheric Mercury Measurement Network.29 Box model simulations by Kim30 and Mao and Talbot31 demonstrated that the dissolution of mercury by dew was likely responsible for the low mercury levels at night in New Hampshire. The dissolved mercury revolatilized after sunrise, increasing mercury levels rapidly. However, in the EML site, THg levels only start increasing at ∼15:00 LT, which may be due to the less efficient re-emission of THg from dry surface than revolatilization of dew. Further study is needed to support this hypothesis. There were slight diurnal variations in wind directions, which shifted from SE in early nighttime (7:00 to 11:00 PM LT) to SSE or S for the rest of the day (Figure 3). Note that the downtown Fort Worth site with several known mercury sources (from the 2011 EPA NEI) is located in the southeast of the EML site (Figure1). Winds from the SE can bring in pollution from urban Fort Worth and contribute to higher THg levels in early nighttime. The correlations between CO, SO 2 , and THg were investigated to obtain source information. Carbon monoxide mainly comes from combustion sources. Sulfur dioxide mainly
■
RESULTS AND DISCUSSION Stationary Site Measurements. The complete time series of atmospheric THg measured at EML is presented in Figure 2.
Figure 2. Time series of THg measured at EML site.
During the study period, the median mixing ratio of THg was 138 ppqv (140 ± 29 ppqv for mean ± SD), slightly lower than the global background levels of 168−190 ppqv reported in a recent review.25 It is also lower when compared to the summer measurement in the urban Houston, Texas area (172 ppqv).23 This characteristic can be explained by the rural nature of the sampling site. In addition, the daytime planetary boundary layer (PBL) height in this area was higher than that in Houston, with a monthly averaged daily maximum of ∼2000 m (Figure 3a) compared to ∼1700 m in the Houston area in summer.23 High PBL heights facilitated the dilution of air pollutants and thus contributed to lower THg levels. The especially high wind speed in our sampling period also favored the dispersion and dilution of 10694
DOI: 10.1021/acs.est.5b02287 Environ. Sci. Technol. 2015, 49, 10692−10700
Article
Environmental Science & Technology
Figure 3. Monthly average diurnal variations of THg, PBL, wind direction, and wind speed (a), and CO, NO, NO2, and SO2 (b).
Figure 4. Correlation of THg with CO and SO2. (a) Time series of THg, CO, SO2, and PBL, where yellow circles show days with especially low nocturnal PBL heights. (b) Scatter plot of CO versus THg; the blue data points represent data from 8 nights with especially high THg levels under low nocturnal PBL condition; the red line is the linear regression line for black data only. (c) Scatter plot of SO2 versus THg.
comes from power plants; coal-fired power plants (CFPP) generally have more SO2 emissions than oil-fired power plants.
Figure 4a shows that a few THg peaks coincided with CO and SO2 peaks, suggesting the influence of combustion sources and 10695
DOI: 10.1021/acs.est.5b02287 Environ. Sci. Technol. 2015, 49, 10692−10700
Article
Environmental Science & Technology
wind directions during studied period, we suspect that emissions from power plants episodically affected our data. The comparisons of diurnal variations among CO, SO2, NO, NO2, and THg reveal different features in their diurnal patterns, albeit all species showed higher mixing ratios at night due to lower PBL height and wind speed (Figure 3b). A well-defined peak was observed at ∼6:00 to 8:00 LT in CO, SO2, NO, and NO2 diurnal patterns, which is likely resulted from vehicle emissions advected from the Fort Worth area. However, no significant peak of THg was observed in this time period, indicating that vehicle emissions may not be an important source of THg in this area. This result is similar to our previous study in the urban Houston area with significantly more traffic.23 VOCs Signatures. Elevated VOCs emissions were reported from areas with large ONG operations.17,22,34,35 During our study period, the median levels of ethane, propane, and n-butane at EML were 3.38, 1.38, and 0.4 ppbv, respectively, which were comparable to the elevated levels reported from other ONG production areas.17 Particularly, we examined the i-pentane (iC5) and n-pentane (nC5) data, whose median levels were 0.21 and 0.15 ppbv, respectively, during study period. The i-pentane/npentane enhancement ratio (iC5/nC5), determined as the slope of the linear fit of iC5 to nC5 observations, is a good indicator for ONG emissions. By employing this ratio, we can minimize the influence of air mass dispersion and dilution as both species should be equally affected. The effect of photochemical loss can also be minimized due to the fact that these two species have similar reaction rates with hydroxyl radicals.36 Previous research found that the iC5/nC5 ratio appears to be similar for different ONG reservoirs and normally fell into the range of 0.8−0.9,37,38 while the values for vehicular emissions are generally higher than 2.39−41 The VOCs measurements from EML show that iC5/nC5 was 1.17 ± 0.01 (±SD), with R = 0.96 (see Figure S3). This value is very close to the reported value from Boulder, Colorado (iC5/ nC5 = 1.10) where iC5 and nC5 emissions are dominated by ONG operations.34 Although no THg plumes can be directly linked to ONG emissions from the one month measurements at EML, it is obvious that this site captured ONG pollution. Further discussion of EML site and its VOCs signature was provided by Zavala-Araiza et al.22 Mobile Laboratory Measurements. An extensive on-road survey was conducted in October 2013 to identify hotspot and spatial variations of several gaseous species, including THg, in ONG fields in the Barnett Shale area. More than 3700 km of THg measurements were conducted, with some parts sampled several times to capture variability in emissions. During the on-road measurements, the high-resolution Picarro instruments provided “first response” of ONG plumes with elevated CH4 levels, and then detailed THg measurements were made at these downwind hotspots. Numerous plumes with elevated CH4 levels were captured in this on-road survey, including plumes from all of the 50 ONG facilities with detailed THg measurements. However, plumes with significant increases of THg mixing ratios were only found from two ONG facilities, which were a compressor station and a natural gas processing plant. Because of the unfavorable wind direction and limited road access, we only made THg measurements once for each of these two facilities during this campaign. Details concerning these two emitters will be discussed in the following section. With large spatial coverage from mobile laboratory measurements, we were able to analyze the geographic distribution of THg levels. To attain this information, first, the influences of wind direction and wind speed needed to be minimized. The
possibly CFPP emissions for those plumes. It is interesting to notice that higher THg levels, including large THg peaks, appeared at several nights with nocturnal PBL heights lower than 100 m (Figure 4a), compared to >300 m for normal nocturnal PBL heights. Low nocturnal PBL resulted from surface-based radiation inversion. Clear sky at night and the open canopy environment of the EML site allowed efficient surface cooling to occur, which led to a stable nocturnal boundary layer. The levels of THg increased when it was trapped inside a shallow surface layer. Figure 4b shows that the correlation between THg and CO was good (R = 0.65, p < 0.005) when excluding some nighttime data with high THg levels under low nocturnal PBL condition. The THg/CO emission ratio was estimated to be 0.34 ppqv/ ppbv, which may represent the THg/CO ratio in the well-mixed background air in the Barnett Shale region. This ratio is comparable to the level of 0.21−0.29 ppqv/ppbv reported in the background air in northeast U.S.27,32 Eight THg plumes are presented in detail in Figure S2. The THg and CO signals from four plumes were not well correlated, which reflects the complexity of emission sources. The overall correlation between THg and SO2 was not significant (Figure 4c), suggesting that the CFPP were not the dominant emission source of THg. The THg and SO2 were also poorly correlated for those time periods even when both species were elevated (see Figure S2), which suggests that other sources may also contribute to the same THg spikes and further obscured the correlation between THg and SO2. The NOAA HYSPLIT Back Trajectory Model was used to trace the sources of large THg spikes appearing at June 2, 3, 4, 5, and 8. This model can take into account the relatively small variability in wind direction. Model results demonstrated that all of the 24-h back trajectories had passed through the south and southeast areas of downtown Fort Worth and a number of large power plants up to 400 km away (Figure 5). This again suggests the possibility that power plant emissions can contribute to large THg spikes in the Barnett Shale area. Considering the low frequency of large THg and SO2 spikes and the small variation in
Figure 5. Back Trajectories (24-h) of air masses with high THg mixing ratios measured at the EML site on June 2, 3, 4, 5, and 8. The red dots and yellow dots stand for the oil-fired and coal-fired units, respectively.33 The pink numbers show mercury emissions in kg/yr from 2011 NEI. Map data copyright 2014 Google, Landsat, SIO, NOAA, U.S.Navy, NGA, GEBCO. 10696
DOI: 10.1021/acs.est.5b02287 Environ. Sci. Technol. 2015, 49, 10692−10700
Article
Environmental Science & Technology
Figure 6. Spatial distributions of THg. (a) THg levels when winds were from the north (315° to 45°). (b) THg levels when winds were from the east (45° to 135°). (c) THg levels when winds were from the south (135° to 225°). (d) THg levels when winds were from the west (225° to 315°). Red rectangle in (a) shows the urban Fort Worth (FW) and Dallas area.
northeast) was the compressor station with high THg emissions. Note that some ONG facilities (whose annual emissions are estimated to be less than 25 000 t of CO2e, including most well pads) were not reported as CH4 point sources in the EPA GHGRP, and thus are not depicted in Figure 6. It is possible that some of these facilities also contributed to the spatial variations of THg levels in this area. Mercury from ONG Facilities. The compressor station with significant THg emissions was located in Wise County of the Barnett Shale area. It was equipped with seven compressor engines, which were likely the sources of THg for this facility. The mixing ratios of THg from pollution plumes were strikingly high, up to 963 ppqv (Figure 7a). In addition, CO2 and CH4 levels as high as 460 ppmv and 15.1 ppmv, respectively, were observed. However, this facility is not listed in the EPA 2011 NEI as a mercury source, or 2012 GHGRP as CO2 and CH4 source. In fact, a very limited number of compressor stations were listed as mercury sources in the EPA 2011 NEI. In the Barnett Shale area, no compressor stations were reported as a mercury source in this inventory. This compressor station is classified as a crude oil and
whole data set was separated into 4 subsets of data according to 4 major wind directions, which are north (315° to 45°), east (45° to 135°), south (135° to 225°), and west (225° to 315°), after removing data from low wind conditions (wind speed
