Article Cite This: Environ. Sci. Technol. XXXX, XXX, XXX−XXX
pubs.acs.org/est
Development of a Fuel-Based Oil and Gas Inventory of Nitrogen Oxides Emissions Alan M. Gorchov Negron,†,⊥ Brian C. McDonald,*,‡,§ Stuart A. McKeen,‡,§ Jeff Peischl,‡,§ Ravan Ahmadov,‡,∥ Joost A. de Gouw,‡,§ Gregory J. Frost,§ Meredith G. Hastings,† Ilana B. Pollack,‡,§,# Thomas B. Ryerson,§ Chelsea Thompson,‡,§ Carsten Warneke,‡,§ and Michael Trainer§ †
Department of Earth, Environmental, and Planetary Sciences, Brown University, Providence, Rhode Island 02912, United States Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, Colorado 80309, United States § Chemical Sciences Division, NOAA Earth System Research Laboratory, Boulder, Colorado 80305, United States ∥ Global Systems Division, NOAA Earth System Research Laboratory, Boulder, Colorado 80305, United States Environ. Sci. Technol. Downloaded from pubs.acs.org by DURHAM UNIV on 08/31/18. For personal use only.
‡
S Supporting Information *
ABSTRACT: In this study, we develop an alternative Fuelbased Oil and Gas inventory (FOG) of nitrogen oxides (NOx) from oil and gas production using publicly available fuel use records and emission factors reported in the literature. FOG is compared with the Environmental Protection Agency’s 2014 National Emissions Inventory (NEI) and with new top-down estimates of NOx emissions derived from aircraft and groundbased field measurement campaigns. Compared to our topdown estimates derived in four oil and gas basins (Uinta, UT, Haynesville, TX/LA, Marcellus, PA, and Fayetteville, AR), the NEI overestimates NOx by over a factor of 2 in three out of four basins, while FOG is generally consistent with atmospheric observations. Challenges in estimating oil and gas engine activity, rather than uncertainties in NOx emission factors, may explain gaps between the NEI and top-down emission estimates. Lastly, we find a consistent relationship between reactive odd nitrogen species (NOy) and ambient methane (CH4) across basins with different geological characteristics and in different stages of production. Future work could leverage this relationship as an additional constraint on CH4 emissions from oil and gas basins.
■
Pennsylvania (PA),16−19 Haynesville, Texas (TX)/Louisiana (LA),18,20 Permian, TX/New Mexico (NM),21 Barnett, TX,22−24 Eagle Ford, TX,25,26 San Juan, CO,27 and Bakken, North Dakota (ND).28,29 Previous work has identified discrepancies between emission inventories of oil and gas operations and atmospheric measurements for CH4, VOCs, and NOx. Most studies have focused on fugitive leaks of CH410,30−37 and VOCs,5,10,33 with many suggesting that emission inventories underestimate both pollutants. Relatively few studies have examined NO x emissions from oil and gas activity. Ahmadov et al.5 found that the 2011 National Emissions Inventory (NEI), reported by the U.S. Environmental Protection Agency (EPA), overestimated NOx emissions by over a factor of 4 in the Uinta Basin, UT. The high NOx emissions contributed to significant model under-predictions of O3 levels under stagnant air conditions, due to excess titration of O3 by nitric oxide
INTRODUCTION In the U.S., oil and natural gas production is a growing source of ozone (O3) precursors and greenhouse gases (GHGs), including nitrogen oxides (NOx = NO + NO2), volatile organic compounds (VOCs), carbon dioxide (CO2), and methane (CH4). The rapid rise in U.S. oil and natural gas production over the past 10 years has motivated new attempts to quantify emissions from these sources to evaluate their impacts on air quality and climate.1,2 Several oil and gas producing regions are in or adjacent to areas that often violate the National Ambient Air Quality Standard (NAAQS) for 8-h O3. However, monitoring networks of criteria air pollutants in oil and gas producing regions are generally sparse as compared to urbanized regions, and the impacts of oil and gas production on regional air quality potentially understated.3 Surprisingly, high O3 concentrations have been found during wintertime in sparsely populated oil and gas production regions in the Uinta, Utah (UT), and the Upper Green River, Wyoming (WY), basins.4−8 Oil and gas production has also been reported to be an important source of O3 precursor emissions in Edmonton, Alberta,9 Denver-Julesburg, Colorado (CO),10−15 Marcellus, © XXXX American Chemical Society
Received: Revised: Accepted: Published: A
April 26, 2018 August 1, 2018 August 3, 2018 August 3, 2018 DOI: 10.1021/acs.est.8b02245 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
from each state and asked to report sales by end use categories. Samples are then scaled up to account for the whole population of sellers. Mapping Oil and Gas CO2 Emissions. In Figure S1, we show that the number of drilling rigs,50 consistently correlates with off-road diesel fuel consumption across various states. To downscale state-level reports of off-road diesel fuel use, we use the location of drill rigs as a spatial proxy. Spatial drilling information is gathered from state databases, such as the location of spuds (the start of the well drilling process). Similarly, the amount of natural gas consumed by the oil and gas industry consistently scales with the amount of natural gas sold to industrial, commercial, and residential establishments (Figure S2). Therefore, we allocate natural gas fuel use based on natural gas production per well using state databases. In addition, we account for variability in the partitioning of fuel between different engine sources (e.g., dehydrators, compressor engines, etc.) by calculating the fraction of fuel used by each engine type as described below. The types of natural gas engines needed varies by basin. Dehydrators and lateral compressors are present across all basins delivering natural gas. Wellhead compressors are more common in older basins past peak production, but scarcer in newer fields with high productivity per well (e.g., Marcellus). We used information on the fraction of wells equipped with wellhead compressors in each basin from the EPA Oil & Gas Tool51 (O&G Tool) (Table S1). Heaters are used as (1) heater treaters to provide heat for separators to remove NGLs (e.g., condensate) from the gas stream, (2) natural gas line heaters to avert hydrate formation during changes in pressure, and (3) tank heaters.52 Therefore, we expect basins dominated by natural gas production (i.e., dry gas basins) to use less natural gas fuel for heaters compared with basins producing significant NGLs and oil (i.e., wet gas basins). Artificial lifts depend on oil production and may be powered by electricity, summarized in Table S1. We develop CO2 emission factors (in kg CO2/L natural gas or kg CO2/L oil) that capture this variability (Figure S3a), which are then multiplied with welllevel production data. Fuel use calculated by these CO2 emission factors is then normalized to equal total local natural gas fuel reported by EIA for each state. Additional details on how CO2 emissions factors were compiled are included in the SI. Annual fuel use records are distributed evenly across each day per year. Seasonal variations of oil and gas production were not included in FOG. However, we do not expect seasonal variations to change our results. Monthly natural gas production statistics for the states included in our inventory indicate that seasonal variation is small (±15% relative to the annual average).53 Fuel-Based NOx Emission Factors. Figure S3b summarizes emission factors of NOx normalized to CO2. We use the ratio of NOx emissions to CO2 emissions by engine type to generate fuel-based NOx emission factors. Gridded NOx emissions are created by multiplying NOx emission factors with maps of CO2 emissions. Below we describe emission factors for drilling and compressor engines in greater detail. The EPA O&G Tool is used for emission factors of dehydrators, heaters, and artificial lifts. The natural gas processing plant emission factor is based on Continuous Emission Monitoring System (CEMS) measurements of oil refineries. No CEMS data are available for natural gas
(NO). This lack of agreement for CH4, VOCs, and NOx suggests that basin specific characteristics and/or emissions activity information in some basins are not described well by existing U.S. inventories. Two approaches are generally used to estimate emissions. Bottom-up studies construct emissions from first-principles, using information on emissions activity at an engine or facility level, and multiply with emission factors that quantify the mass of an emitted pollutant per unit of activity. Uncertainties are associated with data sources used to quantify oil and gas production activities and a lack of updated source testing of emission factors, especially of superemitters.38 Top-down studies use atmospheric measurements to derive basin-wide emissions, providing a means to evaluate uncertainties in bottom-up inventories.32,39−41 Top-down estimates usually do not specify the partitioning of emissions among different source sectors. In addition, they are often limited to a few hours of measurements, and potentially subject to sampling biases related to variable emissions activity,42 as well as to uncertainties associated with meteorology on any given day. In this study, we estimate NOx emissions from combustion associated with oil and gas development. We construct an alternative fuel-based oil and gas (FOG) inventory of NOx emissions. Similar fuel-based approaches have been used to estimate emissions from on-road transportation and off-road equipment,43−46 some of which have been found to be consistent with atmospheric observations.45,46 We evaluate FOG with top-down NOx emissions derived from ambient measurements made during three NOAA-led field studies: the (i) Uinta Basin Wintertime Ozone Study (UBWOS) in 2012− 13, (ii) Southeast Nexus (SENEX) Study in 2013, and (iii) Shale Oil and Natural Gas Nexus (SONGNEX) Study in 2015.
■
METHODS Oil and Gas Development Fuel-Use. Off-road diesel consumption is associated with drilling rigs and hydraulic fracturing equipment. The Energy Information Administration (EIA) provides annual reports of off-road diesel fuel use by state, including end uses by the oil and gas industry.47 The EIA also annually reports the amount of natural gas consumed locally during energy production, known as lease fuel.48 Natural gas is utilized on-site in oil and gas producing regions by 5 engine categories: heaters, dehydrators, wellhead compressors, lateral compressors, and artificial lifts (see the Supporting Information, SI, for glossary). Dehydrators, heaters, and wellhead and lateral compressors are used during the production of natural gas and natural gas liquids (NGLs). Artificial lifts are used to improve wellhead oil production. For natural gas processing plants, we use direct reporting of CO2 emissions as a proxy for fuel-use. Location and annual CO2 emissions of processing plants are compiled from the Facility Level Information on Greenhouse Gases Tool (FLIGHT) from EPA’s Greenhouse Gas Reporting Program (GHGRP).49 The EIA uses two separate methodologies to estimate natural gas fuel use and off-road diesel fuel use. From 2011 to the present, EIA has gathered information on natural gas production and local consumption directly from state agencies, state databases, and commercial records (e.g., DrillingInfo). If information on local natural gas use is not available, then an average historical ratio of the state’s local natural gas use to production is used. The EIA estimates off-road fuel use through the EIA-821 survey of major energy sellers with a response rate of ∼90%. Samples of energy sellers are drawn B
DOI: 10.1021/acs.est.8b02245 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
Figure 1. Correlation between enhancements of NOy and CH4 across oil and gas basins measured by the NOAA P-3 aircraft during the SENEX 2013 and SONGNEX 2015 field campaigns. Circular markers (colored by basin) represent enhancements associated with oil and natural gas production above regional/global background concentrations of NOy and CH4. Two linear fits (black and red lines), the 95% confidence intervals of the slopes, and corresponding R2 values are shown. One fit is performed across all basins, and the other across all basins except the Permian, Texas. The regressions are calculated using two-sided fits. Aircraft observations influenced by point sources, such as natural gas processing plants, and cities are not included in the regression.
controlled lateral compressors, we also utilize in situ mobile laboratory measurements made in Pennsylvania.60 Monte Carlo Uncertainty Analysis. We first perform a sensitivity analysis to identify the most important variables for estimating NOx emissions from oil and gas production. Emissions are most sensitive to both fuel use and NOx emission factors for drill rigs and wellhead compressors. We quantify the overall uncertainty in our basin-level NOx emission estimates using a Monte Carlo analysis. For each variable in our emissions inventory, we assign uncertainty ranges. Two distinct Monte Carlo analyses are conducted. Each scenario differs by the source of NOx emission factors for drill rigs and wellhead compressors, which are either from (1) the EPA O&G Tool, or (2) revised emission factors reported in the literature.19,58 See the SI for more details on the uncertainty analysis. Top-Down NOx Emissions. To evaluate our fuel-based inventory, we estimate top-down NOx emissions derived from aircraft and ground-based observations. Following Ahmadov et al.,5 estimates of CH4 emissions from oil and gas activity are correlated with reactive odd nitrogen species (NOy) measurements and used to estimate basin-wide NOx emissions from oil and gas operations. Here, we treat NOy as a more conserved atmospheric tracer of NOx emissions, and we assume that enhancements above regional background levels in both NOy and CH4 come from oil and gas operations. We assert that this is a reasonable assumption because combustion and fugitive emissions tend to be adjacent to one another (e.g., drilling rigs are colocated with well completions). We utilize ambient NOy/CH4 enhancement molar mixing ratios measured by aircraft for the northeast Marcellus (0.012 ± 0.003 mol/mol), east Fayetteville (0.007 ± 0.002 mol/mol), and Haynesville (0.006 ± 0.003 mol/mol) basins. We then use these enhancement ratios to estimate NOx emissions relative to previously reported CH4 emission fluxes attributed to oil and gas production in the three basins in 2013.61 To estimate uncertainty we propagate errors in mixing ratios and reported uncertainties in CH4 fluxes (between 35 and 50% of basin-wide CH4). For the Uinta basin, we utilize previously reported top-
processing plants, so we assume their emission factors are comparable to those of oil refineries (0.42 g NOx kg−1 CO2). Considerable range in drilling rig NOx emission factors exists in the literature, with upper estimates at 26 g NOx kg−1 CO2.54−56 This value assumes Tier 0 (i.e., uncontrolled) engines, and does not account for lower emission factors of engines certified as Tier 1 and newer.19 The EPA O&G Tool reports an average emission factor of 9 g NOx kg−1 CO2 in 2014, and the EPA NONROAD model reports an emission factor for oil field diesel equipment of 7.4 g NOx kg−1 CO2 in 2013.57 While more recent models of drill rigs have lower NOx emission factors, drill rigs have long life spans. Federal regulations apply to new engines, and do not require retrofits of older engines. We use an emission factor distribution adapted from Roy et al.,19 which accounts for drilling rig fleet changes, and expand it to include tier 0 diesel engines (Figure S4a). For compressor engines, NOx emission factors are highly dependent on exhaust after-treatment devices (e.g., catalysts). Table S2 summarizes NOx regulations in each basin considered in this study. We use wellhead and lateral compressor emission factors based on an equipment survey of NOx emission factors from the Texas Commission of Environmental Quality (TCEQ).58 Lateral compressor engines are considerably larger than wellhead compressors. Because of their size, lateral compressor engines tend to be more strictly regulated for NOx than wellhead compressors. We consider lateral compressor engines to be larger than 373 kW (500 Horsepower) and wellhead compressors to be smaller, based on equipment surveys.58 There is wide variability in wellhead compressor NOx emission factors (Figure S4b). High, noncatalyst engine emission factors are generally found in regions in attainment for national ozone standards. These emission factors exhibit a trimodal distribution that reflects differences in emission standards implemented on nonroad spark-ignition engines over time (Tier 0−2).59 By contrast, low emission factors are generally found in nonattainment areas where compressor engines are more strictly regulated (Table S2).58 Similar differences are shown for lateral compressors in attainment and nonattainment areas (Figure S4c). For C
DOI: 10.1021/acs.est.8b02245 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
Figure 2. NOx emissions mapped from the Fuel-Based Oil and Gas (FOG) inventory in metric tons per day for 2013. Emissions are projected onto a 4 × 4 km2 horizontal grid for: (a) Wyoming, Utah, and Colorado, (b) Texas and Louisiana, (c) Pennsylvania, and (d) Arkansas. Major oil and gas basins are mapped out and labeled (dashed black lines). NOAA P-3 aircraft flight tracks (black lines) are shown for the SONGNEX and SENEX field campaigns in panels over (b) Haynesville, (c) northeast Marcellus, and (d) east Fayetteville. Panel (a) shows the location of a ground site located at Horsepool, Utah (blue star), which was used to derive top-down NOx emissions by Ahmadov et al. (2015).
Subtracting the background mixing ratio controls for the influence of global and upwind emission sources on the intercepts for NOy/CH4 ratios, and enhancements in oil and gas basins are interpreted as from oil and gas activity. Regression slopes for individual basins are listed in Table S3. It is unlikely that our NOy/CH4 ratios are influenced by emissions from agriculture. Most of the data points shown in Figure 1 are from the SONGNEX field campaign, which consists of spring time measurements, during which fertilizer use is expected to be relatively lower than during summer. The similarity of NOy/CH4 mixing ratios for SONGNEX and SENEX (Table S3) suggests that SENEX measurements, made during the summer, are also not strongly influenced by agriculture.
down NOx emissions derived from surface monitoring and aircraft measurements.5 While more recent measurements of oil and gas emissions have been made in other basins, we want to first demonstrate the feasibility of FOG with the UBWOS and SENEX field studies. We primarily evaluate FOG with field measurements made in regions dominated by natural gas production. In the future, FOG can be more thoroughly evaluated in oil dominated basins (e.g., Bakken), whose emissions are captured in other field campaigns. For individual basins, screening techniques are employed to include measurements made within the planetary boundary layer and exclude data points over power plants and cities. Figure S5 shows an example of NOy/CH4 correlation plots made for the Haynesville basin, including two flights made in 2013 and two flights in 2015. We restrict the use of aircraft measurements to those within the planetary boundary layer as reported by Peischl et al.61 Spikes in sulfur dioxide (SO2) and the location of large point sources are used to identify and remove plumes from power plants. High NOx concentrations (>3 standard deviations from the mean) signifying spikes from power plants or large industrial point sources in the city are removed. The potential role of urban emissions was informed by wind direction relative to urban locations. Urban emissions are most important in Haynesville within the northeastern quadrant of the basin, where Shreveport, Louisiana, is located. Similar regression analyses are performed for the northeast Marcellus and east Fayetteville basins. Figure 1 illustrates the relationship between NOy and CH4 across 7 basins, which includes 14 days of measurements across two NOAA-led field campaigns (SENEX in 2013 and SONGNEX in 2015). A correlation is shown between NOy and CH4 enhancements above background levels. The background level is determined as the lowest values found within the planetary boundary layer.
■
RESULTS AND DISCUSSION Fuel-Based Oil and Gas (FOG) Inventory. Figure 2 shows our gridded NOx emission maps from all oil and gas source categories discussed in the Methods section. Note that the color scale is logarithmic and the same across all four panels. A key feature of the four basins is that they are located in different regions of the U.S., and they are characterized by different oil versus natural gas production levels. The Uinta Basin, UT, is a revitalized old wet gas basin composed of stacks of gas and oil plays found in both conventional and unconventional formations. Recent drilling has primarily been in vertical wells, with most activity between depths of 400−550 m. The Haynesville, TX/LA, basin is an overpressurized dry gas basin located between depths of 3000− 4300 m. Considerable horizontal drilling has occurred in the Louisiana side of the basin, with more recent trends showing declines in drilling and production. Just north of Haynesville is the NGL rich Bossier shale, which is often grouped with Haynesville. The northeastern Marcellus, PA, is a new D
DOI: 10.1021/acs.est.8b02245 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
Figure 3. Total basin-wide NOx emission estimates for area sources of oil and gas drilling and production activities. Four bottom-up inventories are displayed as follows: (i) the 2014 National Emission Inventory (dark gray bars), (ii) the 2014 EPA Oil and Gas Tool (light gray bars), (iii) FOG estimated using default NOx emission factors from the EPA Oil and Gas Tool (dark blue bars), and (iv) FOG with revised NOx emission factors for drilling and compressor engines reported in the literature (light blue bars). Note that only area sources of emissions are shown, and do not include trucking and natural gas processing plant emissions. Top-down emissions of NOx (red bars) derived in this study are shown for Haynesville, northeast Marcellus, and east Fayetteville, as well as previously reported top-down NOx emissions for Uinta.5 Above each bottom-up inventory bar, relative differences with top-down emissions are shown in each basin.
fuel demand and a moderate emission factor, usually comprising of 20−30% of basin wide NOx emissions. While processing plants comprise a significant fraction of CO2 emissions, they usually constitute a small fraction of NOx emissions since large point sources are heavily regulated, and new facilities are often required to install NOx emission control equipment. Wellhead compressors and artificial lifts are the most variable sources of NOx. Although these engines consume relatively small amounts of fuel (