Atmospheric implications of large C₂-C₅ alkane emissions from the U.S. oil and gas industry

Abstract

Emissions of C₂-C₅ alkanes from the U.S. oil and gas sector have changed rapidly over the last decade. We use a nested GEOS-Chem simulation driven by updated 2011NEI emissions with aircraft, surface and column observations to 1) examine spatial patterns in the emissions and observed atmospheric abundances of C₂-C₅ alkanes over the U.S., and 2) estimate the contribution of emissions from the U.S. oil and gas industry to these patterns. The oil and gas sector in the updated 2011NEI contributes over 80% of the total U.S. emissions of ethane (C₂H₆) and propane (C₃H₈), and emissions of these species are largest in the central U.S. Observed mixing ratios of C₂-C₅ alkanes show enhancements over the central U.S. below 2 km. A nested GEOS-Chem simulation underpredicts observed C₃H₈ mixing ratios in the boundary layer over several U.S. regions and the relative underprediction is not consistent, suggesting C₃H₈ emissions should receive more attention moving forward. Our decision to consider only C₄-C₅ alkane emissions as a single lumped species produces a geographic distribution similar to observations. Due to the increasing importance of oil and gas emissions in the U.S., we recommend continued support of existing long-term measurements of C₂-C₅ alkanes. We suggest additional monitoring of C₂-C₅ alkanes downwind of northeastern Colorado, Wyoming and western North Dakota to capture changes in these regions. The atmospheric chemistry modeling community should also evaluate whether chemical mechanisms that lump larger alkanes are sufficient to understand air quality issues in regions with large emissions of these species.

1. Introduction

The rise in oil prices caused domestic production of oil and gas to experience a rapid growth in the U.S. since 2005 [U.S. EIA, 2017], increasing emission rates of many trace gases over oil and gas-producing basins [de Gouw et al., 2014; Kort et al., 2016]. Development of, active production from, and abandonment of oil and gas wells emit volatile organic compounds (VOCs). These emissions impact climate [Brandt et al., 2014; Brantley et al., 2014; Franco et al., 2016; Mitchell et al., 2015; Roscioli et al., 2015], the formation of ozone and aerosols [Field et al., 2015; Guo, 2012; Koss et al., 2015; Pacsi et al., 2015; Phillips-Smith et al., 2017; Pusede and Cohen, 2012; Rappenglück et al., 2014], and human exposure to air toxics [Brantley et al., 2015; Halliday et al., 2016; Zielinska et al., 2014]. Observations suggest that depending on the lifetime and emission rate of each species, the impact on atmospheric abundances of VOCs emitted by oil and gas sources can be substantial at local, regional, and global scales. For example, inside the Denver-Julesburg Basin Gilman et al. [2013] estimated that oil and gas sources are the dominant source (72–96 %) of regional C₂ to C₇ alkane emissions. Similarly, in the Uintah Basin, oil and gas leakage contributes 43–82% of observed abundances of C₂-C₅ alkanes [Helmig et al., 2014; Swarthout et al., 2015]. In the Marcellus shale region, multiple studies show that unconventional oil and gas production is responsible for recent positive trends in the observed abundances of methane (CH₄) and ethane (C₂H₆) [Goetz et al., 2017; Peischl et al., 2015; Vinciguerra et al., 2015]. In the Northern Hemisphere annual growth rates of C₂H₆ abundances of 3–5% yr⁻¹ between 2009–2014 have been attributed to the recent increase of oil and gas extraction in North America [Franco et al., 2016; Helmig et al., 2016].

In the context of rapidly changing industrial activities and the fact that production is often driven by transitory economics, updating emission inventories for the U.S. oil and gas sector represents a challenge. In addition to the rapid growth of the oil and gas industry, there are a number of factors that make constraining VOC emissions from this industry difficult: 1) Natural gas composition varies with the type of reservoir (e.g., tight gas vs. shale gas) [Kort et al., 2016; Tzompa-Sosa et al., 2017; Warneke et al., 2014]; 2) Emissions depend on the stage (e.g., development, production or abandoned) of a well. Most of the VOC emissions occur during production [Pacsi et al., 2015], but emissions can continue for decades even after the well has been abandoned [Kang et al., 2014]; 3) Emission inventories rely on activity factors and emission factors that represent typical emission rates for oil and gas wells. However, Brandt et al. [2016] found that in the U.S. 5% of the wells contribute over 50% of the total leakage volume of CH₄. These emission outliers (so-called “super-emitters”) are poorly understood and not represented in emission inventories; 4) National and state regulations vary with respect to in situ emission control technologies [U.S. EPA, 2016a].

In this work, we examine C₂-C₅ alkane emissions from the most recently updated 2011 National Emission Inventory (NEI), which includes updates over important oil-and-gas-producing basins and revised speciation profiles. We use those emissions to estimate the contribution to atmospheric abundances of C₂-C₅ alkanes over the U.S. from this industry. There have been several modeling studies that have begun to explore this issue [Kort et al., 2016; T M Thompson et al., 2017]. Also, we compare abundances of C₂-C₅ to a suite of surface observations, column measurements, and aircraft profiles.

2. Methods

In order to investigate the oil and gas contribution to atmospheric abundances of C₂-C₅ alkanes over the U.S., we use emission fluxes from the model-ready version of the 2011v6.3 emissions modeling platform (more specifically the 2011ek modeling case) and incorporate them into the Goddard Earth Observing System global chemical transport model (GEOS-Chem). In this section, we explain the regridding and unit conversion process of the 2011v6.3 emissions modeling platform fluxes, the creation of year-round daily emission fluxes for the year 2011, and the implementation of year-round daily emission fluxes into GEOS-Chem.

2.1. Updated 2011NEI emission fluxes over the U.S.

In the U.S., the NEI is released every three years. It is based on activity data from state and local agencies. Here we use an updated version of the 2011NEI that is part of the EPA 2011v6.3 emissions modeling platform [U.S. EPA, 2016b] (https://www.epa.gov/air-emissions-modeling/2011-version-63-technical-support-document). Specifically, the modeling case used for the emissions is from the initial version of the 2011v6.3 platform and is also known as “2011ek”. This platform uses the Carbon Bond Mechanism version 6 (CB06) to compute emissions for use as inputs to chemical transport models that require hourly and gridded emissions of chemical species. Relevant to this study, CB06 includes chemical reactions to treat explicit VOC species, such as C₃H₈, benzene, and acetone. In previous model versions, these explicit species were lumped in the paraffin (PAR) species. This work uses specific emissions of propane, benzene and acetone for all model simulations (see Section 2.2). Also, the handling of PAR species in our GEOS-Chem model simulations is explained in Section 2.2.

In the 2011v6.3 modeling platform, oil and gas emission sources are divided into point and non-point sources. Oil and gas point sources include extraction and distribution of oil and natural gas, pipeline transportation, and support activities for oil and gas operations. Non-point oil and gas sources include drill rigs, workover rigs, artificial lifts, hydraulic fracturing engines, pneumatic pumps and other devices, storage tanks, flares, truck loading, compressor engines, and dehydrators. The 2011v6.3 platform is expected to better represent the spatial distribution, amount, and type of species emitted from oil and gas sources due to the incorporation of updates over important oil-and-gas-producing basins and speciation profiles based on the Western Regional Air Partnership (WRAP, www.wrapair2.org). WRAP is a voluntary partnership of states, tribes, federal land managers, local air agencies, and the U.S. Environmental Protection Agency (EPA) within the contiguous U.S. West plus North and South Dakota. The WRAP region encompasses several major U.S. natural gas production basins. Incorporating WRAP data into the 2011v6.3 platform is part of multiple efforts by the EPA to revisit and understand the dynamic nature of oil and gas emissions.

2.1.1. Regridding and unit conversion process of emission fluxes

The air quality model-ready emissions in the platform dataset contain daily files with hourly primary emission fluxes in moles per second (mol/s) on a curvilinear grid at 12 km x 12 km horizontal resolution for all states inside the contiguous U.S. [U.S. EPA, 2017]. These model-ready emission files were created using the Sparse Matrix Operator Kernel Emissions modeling system (SMOKE, http://www.smoke-model.org/) version 3.7. We converted the emission fluxes using mass conservative interpolation into kilograms per square meter per second (kg/m²/s) and regridded them onto a rectilinear grid at a 0.1° longitude by 0.1° latitude resolution (equivalent to approximately 8 km x 11 km over the U.S.). We used the Earth System Modeling Framework (ESMF) software to interpolate from the curvilinear SMOKE data grid to the rectilinear GEOS-Chem grid. Table 1 shows a list of the 2011v6.3 platform anthropogenic emission sources considered in this study.

Table 1.

Characteristics of emission sources from the 2011v6.3 platform emissions dataset included in this work.

Source category	Emission sector	Temporal resolution	Number of files
Point	Electric generating units	daily	365
	Point oil and gas	day-of-week	64
	Other point sources	day-of-week	64
Non-point	Agricultural ammonia	daily	365
	Commercial marine vessels	monthly	12
	Non-point oil and gas	weekly	100
	Other non-point sources	weekly	100
	Railroads	monthly	12
	Residential wood combustion	daily	365
On-road		daily	365
Non-road		day-of-week	64

Note: day-of-week and weekly temporal resolutions include emission files for holidays and the consecutive day after each holiday.

2.1.2. Creation of year-round daily emission fluxes

Each emission sector in the 2011v6.3 platform has daily emission flux files, presented in one out of four different temporal resolutions: daily, according to the day-of-week, weekly, and monthly. Sectors with daily temporal resolution have hourly emissions computed for every day of 2011. Sectors with a temporal approach according to the day-of-week have hourly emissions for four representative days per month: a Saturday, Sunday, Monday and weekday (representing Tuesday through Friday). For sectors with a weekly temporal approach, hourly emissions are computed for all seven days of one representative week in each month. Additionally, the day-of-week and weekly temporal resolutions include emission files for holidays and the consecutive day after each holiday. Table 1 summarizes the temporal resolution approach for each of the emission sectors considered. For the seven sectors without a daily temporal resolution, year-round daily emission files were created by reproducing the emission flux files according to the temporal resolution of each sector. For example, each emission sector with a monthly temporal resolution had twelve emission flux files; thus, each monthly file was reproduced according to the number of days of the month it represented.

The complete emissions dataset in 2011v6.3 platform contains additional emission sources and species than the ones considered in this work. Thus, hereafter we will refer to the implemented emissions from 2011v6.3 platform as updated 2011NEI emissions.

2.2. GEOS-Chem simulations

We conducted two nested simulations (0.5°x0.6°) over North America (40° to 140° W, 10° to 70° N) using the 3-D chemical transport model GEOS-Chem version 10–01 (Bey et al. [2001], http://www.geos-chem.org) for the year 2011. The GEOS-Chem model was driven by off-line GEOS-5 assimilated meteorological fields (https://gmao.gsfc.nasa.gov/GEOS/) with 47 vertical levels. Global simulations at 2°x2.5° resolution with a spin-up of 18 months were used as boundary conditions for the nested simulations. The emissions and injection timesteps were set to 20 minutes; the transport timestep was set to 10 minutes.

In our baseline simulation, we implemented the updated 2011NEI emission fluxes into GEOS-Chem using the stand-alone software component for computing emissions, Harvard-NASA Emissions Component (HEMCO) version 1.1.005 [Keller et al., 2014]. Over the U.S. (CONUS), all anthropogenic and biofuel emissions were derived from the updated 2011NEI. Outside the CONUS, we used anthropogenic emissions from the Emissions Database for Global Atmospheric Research (EDGAR v4.2) and VOC emissions from the Reanalysis of the Tropospheric chemical composition (RETRO) emission inventory, except for C₂H₆ and C₃H₈ for which we used the Tzompa-Sosa et al. [2017] and Xiao et al. [2008] emission inventories, respectively. We also include regional anthropogenic emission inventories for northern Mexico [Kuhns et al., 2003], Canada (https://www.canada.ca/en/environment-climate-change/services/national-pollutant-release-inventory/tools-resources-data.html), Europe (http://www.ceip.at/), and Asia (http://meicmodel.org/dataset-mix.html). Non-U.S. biofuel emissions were from the Yevich and Logan [2003] emission inventory, with two exceptions. 1) We used ammonia (NH₃) emissions from the Global Emissions InitiAtive (GEIA), and 2) C₂H₆ emissions from Tzompa-Sosa et al. [2017]. Shipping, aviation and natural sources are expected to make minor contributions to the emissions of C₂-C₅ alkanes; the default global datasets incorporated into GEOS-Chem were used for these sectors. Table 2 shows the summary of the emission inventories that we used in the baseline simulation.

Table 2.

Configuration of emission inventories in our baseline simulation.

Emission inventory	Region	Base year	Species in GEOS-Chem
Anthropogenic
Updated 2011NEI	CONUS	2011	ACET, ALD2, ALK4, BCPI, BCPO, BENZ, C2H4, C2H6, C3H8, CO, EOH, FORM, HONO, MACR, MEK, MOH, NH3, NO, NO2, OCPI, OCPO, PRPE, RCHO, SO2, SO4, TOLU, XYLE
BRAVO	Northern Mexico	1999	CO, NO, SO2, SO4
CANADA*	Canada	2002	CO, NO, SO2, SO4
		2008	NH3
EMEP	Europe	2011	CO, NH3, NO, SO2
		2000	ALD2, ALK4, MEK, PRPE
MIX*	Asia	2010	ALD2, ALK4, CH2O, CO, NH3, NO, SO2, SO4, MEK, PRPE
EDGAR*	global	2008	CO, NAP, NH3, NO, SO2, SO4
Tzompa-Sosa et al. [2017]	global	2010	C2H6
Xiao et al. [2008]	global	1985	C3H8
RETRO	global	2000	ACET, ALD2, ALK4, BENZ, CH2O, C2H2, C2H4, MEK, PRPE, TOLU, XYLE
Biofuel
Updated 2011NEI	CONUS	2011	ACET, ALD2, ALK4, BCPI, BCPO, BENZ, C2H4, C2H6, C3H8, CO, EOH, FORM, HONO, MACR, MEK, MOH, NH3, NO, NO2, OCPI, OCPO, PRPE, RCHO, SO2, SO4, TOLU, XYLE
Yevich and Logan [2003]	global, except the CONUS	1985	ACET, ALD2, ALK4, BENZ, CH2O, C2H2, C2H4, C3H8, CO, GLYC, GLYX, HAC, MEK, MGLY, NAP, NO, PRPE, SO2, TOLY, XYLE
GEIA	global	1998	NH3
Tzompa-Sosa et al. [2017]	global	2010	C2H6
Shipping
EMEP	Europe	2011	CO, NO, SO2
ARCTAS	global	2008	SO2
ICOADS	global	2002	CO, NO
Aviation
AEIC	global	2005	ACET, ALD2, ALK4, BC, CH2O, C2H6, C3H8, CO, MACR, NO, SO2, SO4, OC, PRPE, RCHO
Natural sources
GEOS-Chem default	global	2000	NO
		1985	DMS
		2009	SO2
GEIA	global	1990	NH3

Notes:

^1.Over the CONUS, all anthropogenic and biofuel emissions in the baseline simulation come from the updated 2011NEI.

^2.Unless otherwise noted, the simulation uses the same year as the base year for the 2011 simulation.

^*Projected to 2010 using GEOS-Chem default annual scaling factors.

For the second simulation (hereafter updated 2011NEI: OG off), we maintained the same configuration of emission inventories as the baseline simulation, but we turned off the updated 2011NEI emissions of ≥ C₂ alkanes from the oil and gas sector. We used this simulation to investigate the contribution of oil and gas related activities to the abundance of the ≥ C₂ alkanes over the CONUS.

The alkane speciation used in GEOS-Chem was originally based on the Lurmann et al. [1986] condensed gas phase chemical mechanism. The current mechanism in GEOS-Chem treats C₃H₈ and C₂H₆ as explicit species. ≥ C₄ alkanes are lumped into one tracer, originally named ALKA in Lurmann et al. [1986], and currently named ALK4 in GEOS-Chem. The rate constant in GEOS-Chem associated with the reaction of OH with ALK4 is based on the absolute rate coefficient of butane (9.1×10⁻¹²e^(−405/T) cm³ molecule⁻¹ s⁻¹, Atkinson et al. [2006]), but is used to represent the chemistry of all ≥ C₄ alkanes. Both of our simulations have a global annual mean tropospheric mass-weighted OH concentration of 1.3×10⁶ molecule cm⁻³, which is very close to the upper bound of previous model studies in the literature [Naik et al., 2013; Voulgarakis et al., 2013]. In this study, we consider ALK4 specifically as n-butane, i-butane, n-pentane, and i-pentane (hereafter referred to as C₄-C₅ alkanes), rather than a more inclusive ≥ C₄ alkanes. Thus, one obvious challenge with the model-observation comparison that we present later in the paper is the use of a single reaction rate for C₄-C₅ alkanes. As explained in Section 2.1, 2011v6.3 platform emissions of PAR species include alkanes. We assigned a fraction of PAR to C₄-C₅ alkane species based on Simon et al. [2010]. Simon et al. [2010] summarize the 50 VOCs with the largest emissions over the U.S.; for an example day, C₄-C₅ alkanes correspond to 36% of these emissions over the U.S. Thus, we set ALK4 emissions as 36% of the total PAR emitted species. The remaining reactive carbon in the PAR species is not considered here. Omitting such a large fraction of reactive carbon limits our ability to provide a full view of the impact of oil and gas operations and urban activities on atmospheric composition. This is a known limitation to our approach. However, we investigated the impact of attributing 100% of PAR to ALK4 versus 36% of PAR to ALK4 on ALK4 lifetime. Omitting such a large fraction of carbon changes the lifetime of ALK4 over the U.S. by < 5%. As discussed in the conclusions, we suggest the addition of a new GEOS-Chem tracer for C₆-C₈ alkanes, which based on the 50 VOCs with largest emissions over the U.S. [Simon et al., 2010], account for ~40% of PAR. Thus, adding such fraction of the remaining reactive carbon from the PAR species could provide a better estimate of the full impact of the emissions from this sector.

The GEOS-Chem mechanism does not include other paraffin compounds, such as alkynes, and higher aromatic VOCs that have also been found in high abundances (compared to background values) over oil and gas basins [Abeleira et al., 2017; Gilman et al., 2013; Helmig et al., 2014; Pétron et al., 2012; Pétron et al., 2014; Swarthout et al., 2013; Swarthout et al., 2015; C R Thompson et al., 2014; Zielinska et al., 2014]. Additionally, our model simulations do not include tropospheric chlorine chemistry; thus reaction with OH is the only tropospheric sink of C₂-C₅ alkane species. Sherwen et al. [2016] used the same initial GEOS-Chem version, but they added tropospheric halogen chemistry (Cl, Br, I). In their study, adding a chlorine sink term led to decreases in the tropospheric global burdens of C₂H₆, C₃H₈, and ≥ C₄ alkanes of 19%, 14% and 12%, respectively. However, global tropospheric burden changes are heterogeneous, and in general they are lower over land compared to oceans. At the surface over the U.S., the annual average changes are smaller (typically <10% for C₂H₆ and less for C₃H₈, and ≥ C₄ alkanes; see Sherwen et al. [2016] Figure 19). The inclusion of halogen chemistry would decrease the O₃ burden, and thus the OH burden as well. This would increase the lifetimes of C₂H₆, C₃H₈, and ≥ C₄ alkanes against OH oxidation. Given that the inclusion of updated 2011NEI emissions in our model produces significant increases in C₂-C₅ alkane fluxes over the U.S. compared to the emission inventory used by Sherwen et al. [2016], a comprehensive understanding of these two model developments would require additional model simulations.

3. Results and discussion

3.1. Contribution of the oil and gas sector to emissions of C₂-C₅ alkanes

3.1.1. Ethane and Propane

The oil and gas sector emits C₂H₆ and C₃H₈ primarily due to leakage during the production, processing, and transportation of natural gas [Gilman et al., 2013; Kort et al., 2016; Pétron et al., 2012; Roest and Schade, 2017]. Trace amounts of C₂H₆ and C₃H₈ can also be produced during hydrocarbon combustion processes [Basevich et al., 2012; Gomer and Kistiakowsky, 1951; Sangwan et al., 2015; Thynne, 1962]. In the updated 2011NEI, total emissions of C₂H₆ and C₃H₈ are dominated by oil and gas sources (point sources - e.g. oil and gas extraction, distribution, pipelines - and non-point sources - e.g. flares, drill and workover rigs) with an estimated contribution to total anthropogenic emissions for the U.S. of 89% and 82%, respectively (Figure 1). The remaining percentage of C₂H₆ and C₃H₈ emissions is distributed among other sources such as vehicles, and residential wood combustion. We note that from the two oil and gas sectors considered, non-point sources are the biggest contributors, accounting for 95% and 97% of the total oil and gas contribution estimate.

Figure 1.

Updated 2011NEI emissions of C₂H₆, C₃H₈, and C₄-C₅ alkanes by sector. C₄-C₅ alkanes are presented as 36% of PAR emissions. Units for C₂H₆ and C₃H₈, are in Gg yr⁻¹; and units for C₄-C₅ alkanes are presented in Gg C yr⁻¹.

The total updated 2011NEI emissions of C₂H₆ from the present study are 15% lower than the Tzompa-Sosa et al. [2017] C₂H₆ emission inventory estimate, which was calculated by scaling C₂H₆ emissions of 2011NEI version 1 (2011NEIv1) by a factor of 1.4 based on a comparison to existing observations. It is important to note that the 2011NEIv1 did not contain updates over oil and gas basins based on the WRAP (the updated 2011NEI used here includes the WRAP data), causing oil and gas regions like the Uintah basin to have minimal C₂H₆ emission fluxes. In this region, where studies have found important C₂H₆ emission enhancements [Helmig et al., 2014; Koss et al., 2015; Warneke et al., 2014], emission fluxes from oil and gas sources in 2011NEIv1 are close to zero, thus upward scaling by 1.4 still results in a small flux. Thus, if the scaling in Tzompa-Sosa et al. [2017] were applied in this study, the result would be a different spatial distribution and amount of emission fluxes compared to the updated 2011NEI used here. Lastly, we notice that between 2011NEIv1 and the updated 2011NEI used in this work, emissions of C₃H₈ have higher emission flux increases compared to C₂H₆. The highest emission fluxes occur over oil and gas regions in the central U.S., with increases of up to 300 ng m⁻² s^-1.

3.1.2. C₄-C₅ alkanes

Over the last decade, leakage from oil and gas sources has become an important contributor to the emissions of C₄-C₅ alkanes [Gilman et al., 2013; Johansson et al., 2014; Roest and Schade, 2017; Swarthout et al., 2013; Swarthout et al., 2015], which historically were dominated by automobile combustion, and fugitive emissions from gasoline and diesel distribution [Lee et al., 2006; Schauer et al., 2002]. Thus, urban areas are the locations where enhancements of C₄-C₅ alkanes are commonly observed [Aceves and Grimalt, 1993; Bi et al., 2005; Lee et al., 2006]. Rossabi and Helmig [2018] recently used data collected between 2001and 2015 over the U.S. to show a predominantly decreasing trend in C₄-C₅ alkanes surface mixing ratios, but they found a relative increase in the predominance of the n-isomers. They attributed this pattern to changes in isomeric ratios in gasoline sector emissions, and emissions from the oil and gas industry. The emergence of U.S. oil and gas development as a larger source of C₄-C₅ alkanes has increased their atmospheric abundances in areas with low population density [Gilman et al., 2013; Pétron et al., 2014; Warneke et al., 2014]. Gilman et al. [2013] estimated that the mean oil and gas contribution to C₄-C₅ alkane emissions in northeastern Colorado is 93%−96%. Based on the updated 2011NEI emissions (Figure 1), we estimate that oil and gas sources (including both point and non-point sources) over the CONUS are the third most important emission source of C₄-C₅ alkanes with an annual contribution of 26% of the total emissions.

3.2. Geographical distribution of oil and gas C₂-C₅ alkane emissions and its contribution to U.S. total anthropogenic emissions

In the U.S., emissions of C₂H₆ and C₃H₈ are mainly clustered inside oil and gas basins, where the contribution of the oil and gas sector to total anthropogenic emissions is > 90% (Figure 2, panels a and b) For C₄-C₅ alkanes, the emissions not only occur inside oil and gas basins, but also in urban areas due to the importance of other fossil fuel sources. The contribution of urban sources to total emissions of C₄-C₅ alkanes over oil-and-gas-producing regions reduces the overall percentage contribution of these sources (Figure 2, panel c).

Figure 2.

Left column: spatial distribution of anthropogenic emissions of C₂H₆, C₃H₈, and C₄-C₅ alkanes. Right column: spatial distribution of the contribution of oil and gas emissions to total anthropogenic emissions of C₂H₆, C₃H₈, and C₄-C₅ alkanes. C₂-C₅ alkane emissions data from the updated 2011NEI. C₄-C₅ alkanes are presented as 36% of PAR emissions.

A comparison between regional emissions of C₂H₆, C₃H₈, and C₄-C₅ alkanes shows that the central region of the U.S. is the most important contributor to total CONUS C₂-C₅ alkane emissions in 2011, contributing ~70% of C₂H₆ and C₃H₈ total CONUS emissions, and ~40% of the emissions of C₄-C₅ alkanes (Figure 3). The central region fully encompasses four U.S. oil and gas basins: Eagle Ford (Texas), Permian (Texas), Niobrara (Colorado and Wyoming), and Bakken (North Dakota). This estimate is likely to be higher for years later than 2011 for C₂H₆ due to the massive increase in oil and gas exploitation in the Bakken basin [Kort et al., 2016; Peischl et al., 2016].

Figure 3.

Regional contributions (as %) to U.S. total anthropogenic emissions of C₂H₆, C₃H₈, and C₄-C₅ alkanes. C₂-C₅ alkane emissions data from the updated 2011NEI. C₄-C₅ alkanes are presented as 36% of PAR emissions.

3.3. Model comparison to observations and oil and gas contribution to atmospheric abundances of C₂-C₅ alkanes

We compare 2011 abundances of C₂H₆, C₃H₈, and C₄-C₅ alkanes from a GEOS-Chem simulation to a suite of observations over North America (Table 3 and Figure 4). To the best of our knowledge, this constitutes the largest compendium of C₂-C₅ alkane observations compared to model output for this region. Also, we estimate the contribution of oil and gas to atmospheric abundances of C₂H₆, C₃H₈, and C₄-C₅ alkanes by turning off the emissions of these species from the oil and gas sector in a separate GEOS-Chem simulation (updated 2011NEI: OG off).

Table 3.

Observations from surface sites and airborne campaigns, ordered by type and date.

2011 FTIR column measurements
Species	Site	Location	Period	Reference
C2H6	Toronto, Ontario, Canada	79.4° W, 43.6° N	Jan-Dec 2011	Wiacek et al. [2007]
C2H6	Boulder, Colorado, USA	105.3° W, 40.4° N	Jan-Dec 2011	Hannigan et al. [2009]
Aircraft campaigns
Species	Field campaign	Region	Period	Reference
C2–5	ARCTAS	110° to 126° W, 30° to 50° N	Apr, Jun-Jul 2008	Simpson et al. [2010] Simpson et al. [2011]
C2–5	HIPPO	90° to 116° W, 25° to 50° N	Jun-Sep 2011	Wofsy et al. [2012]
C2–5	SEACR4S	80° to 126° W, 25° to 50° N	Aug-Sep 2013	Blake et al. [2014] Schauffler [2014]
C2–5	FRAPPÉ	101° to 109° W, 38° to 46° N	Jul-Aug 2014	Richter et al. [2015]
2011 Surface flask measurements from National Oceanic and Atmospheric Administration (NOAA) / Institute of Arctic and Alpine Research (INSTAAR) Global VOC Monitoring Program
Species	Site	Location	Period	Website
C2–5	Key Biscayne, Florida (KEY), USA	80.16°W, 25.67°N	Jan-Dec 2011	https://www.esrl.noaa.gov/gmd/dv/data/
C2–5	Park Falls, Wisconsin (LEF), USA	90.27° W, 45.95° N	Jan-Dec 2011
C2–5	Southern Great Plains, Oklahoma, (SGP), USA	97.5° W, 36.8° N	Jan-Dec 2011
C2–5	Trinidad Head, California (THD), USA	124.15° W, 41.05° N	Jan-Dec 2011
C2–5	Wendover, Utah (UTA), USA	113.72° W, 39.9° N	Jan-Dec 2011
2011 Surface flask measurements from Photochemical Assessment Monitoring Stations (PAMS)
Species	Site	Location	Period	Website
C2–5	Baltimore, Maryland (BAL), USA	76.6° W, 39.3° N	Jun-Aug 2011	https://www.airnowtech.org
C2–5	Boston, Massachusetts (BOS), USA	71.1° W, 42.4° N	Jun-Aug 2012
C2–5	El Paso, Texas (ELP), USA	106.4° W, 31.8° N	Jan-Dec 2011
C2–5	Gary, Indiana (GAR), USA	87.3° W, 41.6° N	Jun-Dec 2011
C2–5	Houston, Texas (HOU), USA	95.4° W, 29.8° N	Jan-Dec 2011
C2–5	Los Angeles, California (LAX), USA	118.3° W, 34.1° N	Jan-Dec 2011
C2–5	Philadelphia, Pennsylvania (PHI), USA	75.2° W, 40° N	May-Oct 2011
C2–5	Atlanta, Georgia (SDK), USA	84.4° W, 33.8° N	Jun-Aug 2011
C2–5	Springfield, Massachusetts (SPR), USA	72.5° W, 42.1° N	Jun-Aug 2011
Surface observations
Species	Site	Location	Period	Reference
C4–5	Houston Ship Channel, Texas (HSC), USA	95.03° W, 29.65° N	Sep 2006	Johansson et al. [2014]
C3–5	San Francisco, California (STR), USA	122.45° W, 37.76° N	Jun-Aug 2007–2010	Pétron et al. [2012]1
C3–5	Walnut Grove, California (WGC), USA	121.49° W, 38.26° N	Jun-Aug 2007–2010	Pétron et al. [2012] 1
C3–5	Moody, Texas (WKT), USA	97.33° W, 31.32° N	Jun-Aug 2007–2010	Pétron et al. [2012] 1
C3–5	Park Falls, Wisconsin (LEF), USA	90.27° W, 45.93° N	Jun-Aug 2007–2010	Pétron et al. [2012] 1
C5	Barnett Shale, Texas (BST), USA	97.42° W, 33.27° N	May 2010	Zielinska et al. [2014]
C3–5	Boulder Atmospheric Observatory, Colorado (BAO), USA	105.01° W, 40.05° N	Aug 2007-Apr 2010	Pétron et al. [2012]
C2–5	Boulder Atmospheric Observatory, Colorado (BAO), USA	105.01° W, 40.05° N	Feb-Mar 2011	Gilman et al. [2013]
C2–5	Boulder Atmospheric Observatory, Colorado (BAO), USA	105.01° W, 40.05° N	Feb-Mar 2011	Swarthout et al. [2013]
C2–5	Beaumont Downtown, Texas (BDT), USA	94.07° W, 30.04° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Cesar Chavez HS, Texas (CCH), USA	95.25° W, 29.68° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Channelview, Texas (CNV), USA	95.13° W, 29.8° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Clinton, Texas (CLT), USA	95.26° W, 29.73° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Corpus Christi Oak Park, Texas (CCO), USA	97.43° W, 27.8° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Corpus Christi Palm, Texas (CCP), USA	97.42° W, 27.8° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Corpus Christi Solar Estates, Texas (CCS), USA	97.54° W, 27.83° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Dallas Hinton, Texas (DHT), USA	96.86° W, 32.82° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Danciger, Texas (DNG), USA	95.76° W, 29.14° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Decatur Thompson, Texas (DTS), USA	97.58° W, 33.22° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Deer Park, Texas (DPK), USA	95.13° W, 29.67° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Dish Airfield, Texas (DAF), USA	97.3° W, 33.13° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Eagle Mtn Lake, Texas (EML), USA	97.48° W, 32.99° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	El Paso Chamizal, Texas (EPC), USA	106.46° W, 31.77° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	El Paso Delta, Texas (EPD), USA	106.41° W, 31.76° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Everman Johnson Park, Texas (WJP), USA	97.29° W, 32.62° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Flower Mound, Texas (FWM), USA	97.13° W, 33.05° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Fort Worth NW, Texas (FWN), USA	97.36° W, 32.81° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	HRM3, Texas (HRM), USA	95.18° W, 29.76° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Lake Jackson, Texas (LJK), USA	95.47° W, 29.04° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Lynchburg Ferry, Texas (LBF), USA	95.08° W, 29.76° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Milby Park, Texas (MPK), USA	95.26° W, 29.71° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Nederland HS, Texas (NDL), USA	94.01° W, 29.98° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Odessa Hays, Texas (OHY), USA	102.34° W, 31.84° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Texas City 34th St., Texas (TXC), USA	94.95° W, 29.41° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Wallisville Rd, Texas (WVR), USA	94.99° W, 29.82° N	Jan-Dec 2011	TCEQ [2012] 2
C2–5	Hickory, Pennsylvania (HKY), USA	80.30° W, 40.30° N	July 2012	Swarthout et al. [2015]
C2–5	Racoon Creek State Park, Pennsylvania (RCS), USA	80.50° W, 40.50° N	July 2012	Swarthout et al. [2015]
C2–5	Erie, Colorado (ERC), USA	105.05° W, 40.05° N	Mar-Jun 2013	C R Thompson et al. [2014]
C2–5	Boulder Atmospheric Observatory, Colorado (BAO), USA	105.01° W, 40.05° N	Mar-May, Jul-Sep, 2015	Abeleira et al. [2017]

Notes:

^1.C₄ observations only include n-C₄H₁₀.

^2.C₄-C₅ observations only include n-C₄H₁₀ and n-C₅H₁₂.

Figure 4.

Summary of observations listed in Table 3. Labels of overlapping surface observations are not shown. Locations of active wells come from FracTracker (accessed Nov 2015, www.fractracker.org). In order to provide a sense for well spatial distribution over states with missing data, shale and tight gas plays (Energy Information Administration, accessed Dec 2014, www.eia.gov/dnav/ng/ng_sum_lsum_a_EPG0_xdg_count_a.htm) are shown.

3.3.1. Comparison to ground-based FTIR C₂H₆ column observations

In this section, we compare 2011 C₂H₆ total columns derived from ground-based Fourier Transform Infrared (FTIR) solar observations at the Boulder and Toronto stations to GEOS-Chem simulated C₂H₆ total columns for our two emission scenarios (Figure 5). The C₂H₆ total columns were consistently determined at both sites following the methodology described in Franco et al. [2015]. The latter paper further provides information on the typical systematic and random uncertainties affecting the column measurements. The first emission scenario considers all emissions and sectors from the updated 2011NEI. In the second emission scenario, C₂H₆ emissions from the oil and gas industry are turned off (updated 2011NEI: OG off). Finally, the oil and gas contribution to C₂H₆ total columns is calculated by subtracting the results of the second scenario from those of the first scenario (updated 2011NEI - updated 2011NEI: OG off).

Figure 5.

Comparison of 2011 FTIR C₂H₆ total columns to GEOS-Chem C₂H₆ columns using a simulation with and without oil and gas sources from the updated 2011NEI. Black dots represent FTIR monthly mean C₂H₆ total columns, and the grey shading denotes their associated 1σ standard deviation. Monthly means are displayed proportionally to the observations available in each month. The blue line represents modeled C₂H₆ total columns using all sectors from the updated 2011NEI. The red line represents modeled C₂H₆ total columns with C₂H₆ emissions from oil and gas sector turned off (updated 2011NEI: OG off). The blue and red lines are running mean fits to the daily-averaged model columns (with a 6-week wide integration time and a 15-day time step).

At the Boulder station, C₂H₆ emissions from the updated 2011NEI reproduce observed C₂H₆ total columns outside of winter months. A difference of ~0.2–1×10¹⁶ molecules cm⁻² is observed during the winter season (including November). At the Toronto station, modeled updated 2011NEI C₂H₆ emissions underestimate (on average by ~0.5×10¹⁶ molecules cm⁻²) the observed C₂H₆ total column throughout the year 2011. We note that as shown in Table 2, our simulation does not include recent updates to C₂H₆ emission fluxes made by Environment and Climate Change Canada. The difference in observed and modeled C₂H₆ total columns might be due to a combination of underestimated urban C₂H₆ leakage from natural gas delivery and end use, residential wood combustion, and the higher resolution (0.5°x0.6°) analysis made in this study using 2°x2.5° C₂H₆ emissions derived by Tzompa-Sosa et al. [2017]. The coarser resolution of the C₂H₆ emissions over Toronto limits the ability of our higher resolution model simulation to capture local enhancements. The total C₂H₆ columns observed and produced by both emission scenarios over Toronto are larger than over Boulder. Considering that the column measurements are sensitive to the whole troposphere and lower stratosphere, the column difference between Toronto and Boulder can be explained by the altitude difference between both stations (~1.5 km). Another possible explanation of the column difference is the latitudinal gradient in C₂H₆, with higher abundances towards the Artic [Helmig et al., 2016; Simpson et al., 2012].

There is a greater contribution to modeled total C₂H₆ columns from emissions from the oil and gas sector (~0.7×10¹⁶ molecules cm⁻²) at the Boulder station compared to the Toronto station (~0.4×10¹⁶ molecules cm⁻²). This finding is consistent with results presented in Franco et al. [2016] for the 2009 – 2014 period. Among six FTIR stations (including Toronto) located at different latitudes across the Northern Hemisphere, they showed that the Boulder station had the highest rate of change in the C₂H₆ total column over this time period, presumably associated with the oil and gas development in the central U.S. The high contribution of the oil and gas sector over Boulder is also shown in results from this study (Figure 2). Furthermore, from the four regions analyzed in Section 3.3.3 (Figure 10), the region where the Boulder station is located shows the highest percentage contribution from the oil and gas sector to total abundances of C₂-C₅ alkanes throughout the troposphere.

Figure 10.

2011 simulated percentage contribution from the oil and gas sector to total abundances of C₂H₆.

3.3.2. Comparison to surface flask observations

In this section, we compare 2011 simulated alkane mixing ratios with and without oil and gas sources (blue and red lines, respectively in Figure 6) to measured C₂-C₅ surface mixing ratios from samples collected at selected U.S. stations (Table 3) from the NOAA Global Greenhouse Gas Reference Network (GGGRN) Stations are ordered from higher to lower latitudes due to the observed strong latitudinal gradient of C₂-C₅ alkane abundances [Helmig et al., 2016; Simpson et al., 2012]. Differences between our simulations with and without emissions of C₂H₆ and C₃H₈ from the oil and gas sector, suggest that the SGP station was more impacted by emissions from this industry compared to the rest of the stations throughout 2011. The estimated annual oil and gas source contributions to surface C₂H₆ and C₃H₈ mixing ratios at the SGP station is 86% for both species. The higher oil and gas impact at the SGP station is expected because it is located inside an oil and gas region. Typical i-pentane/n-pentane ratio for regions dominated by emissions from the oil and gas sector range from 0.89 – 1.10 [Gilman et al., 2013]. The calculated 2011 i-pentane/n-pentane ratio at SGP is 0.97, corroborating that air masses in this area are highly impacted by oil and gas sources. LEF and KEY are two other stations where the model predicts that oil and gas activities make a large contribution to atmospheric abundances of C₂-C₅ alkanes (~25% for C₂H₆ and C₃H₈). The relatively high oil and gas contributions are consistent with Helmig et al. [2016]; their analysis shows higher rates of changes between 2009–2014 in C₂H₆ and C₃H₈ occurring in sites downwind the central and eastern U.S. We note that only a few of long-term stations are ideally located to capture changes related to the major oil and gas source regions that have the highest emissions in the updated 2011NEI. Long-term monitoring stations located in northeastern Colorado, Wyoming, and North Dakota should be considered in order to capture emission changes in the oil and gas sector.

Figure 6.

Comparison of 2011 surface mixing ratios for C₂H₆, C₃H₈ and C₄-C₅ alkanes (from top to bottom) to modeled 2011 emissions from the updated 2011NEI with and without oil and gas sources. Black dots represent monthly mean observations from NOAA GGGRN global surface flask network (Table 3), and the grey areas denote their associated 90^th percentile. The blue line represents monthly mean simulated surface mixing ratios using emissions from all sectors of the updated 2011NEI. The red line represents mixing ratios from the updated 2011NEI: OG off simulation. The stations are ordered from higher (left) to lower (right) latitudes. Note the various vertical scales.

For C₄-C₅ alkanes, the model shows a stronger seasonality compared to observations and overestimates monthly mean mixing ratios at all the selected stations. The stronger seasonality in the model output compared to observations could be a function of either an incorrect representation of the seasonality of C₄-C₅ alkane emissions in the updated 2011NEI, and/or the use of a single reaction rate for C₄-C₅ alkanes. Absolute differences between observed and modeled C₄-C₅ alkane surface mixing ratios are highest at SGP and LEF stations, where the monthly overestimations are as high as 22 ppbC (bottom row of Figure 6). At the THD and UTA stations, the absolute overestimation is not nearly as dramatic as for the other stations; the average annual overestimation for both sites is ~1.2 ppbC. However, on an annual average, the model overestimates C₄-C₅ alkanes at all sites by a factor of ~4. The consistent model bias of C₄-C₅ alkane surface mixing ratios suggests that our choice of assigning C₄-C₅ alkanes as a continuous fraction of total PAR emitted species across the CONUS (see Section 2.2), should be revisited. Another possible cause is the overestimation of total PAR emissions in the U.S.

3.3.3. Seasonal comparison to averaged observational datasets

GEOS-Chem averaged seasonal model output for the year 2011 and observed abundances of C₂H₆, C₃H₈, and C₄-C₅ alkanes are shown in Figures 7, 8, and 9, respectively. The bottom row in the upper panels of Figures 7–9, show filled circles that represent seasonal averages of daily surface flask measurements and other surface observations as averages of their specific sampling period in the corresponding season when they occurred. Time periods and locations for each dataset are presented in Table 3. As can be noted in Table 3, there are very few aircraft observations for 2011. The upper panels of Figures 7–9 present aircraft observations from other years. Given high rates of change in C₂H₆ and C₃H₈ since 2009, particularly over the central U.S. [Franco et al., 2016; Helmig et al., 2016], directly comparing the model to these observations is a challenge. We use the observations and model to show seasonal, horizontal and vertical gradients in these species across the U.S. We provide comparisons where we can do so conservatively. In order to provide a direct comparison to surface observations, the lower panels of Figures 7–9 show modeled versus surface flask observations from the year-round TCEQ and INSTARR datasets.

Figure 7.

Upper panel: Mean distribution of C₂H₆ abundances for different seasons and altitude ranges compared to observations from aircraft campaigns and surface measurements (Table 3). The background contours are model outputs for 2011. The filled circles represent seasonally averaged observations. Aircraft measurements (0–2, 2–6, and 6–10 km) are averaged vertically for each altitude range and horizontally every 1° × 1°. Lower panel: Scatter plot of seasonal surface observations from TCEQ (filled circles) and INSTARR (transparent triangles) datasets.

Figure 8.

Upper panel: Mean distribution of C₃H₈ abundances for different seasons and altitude ranges compared to observations from aircraft campaigns and surface measurements (Table 3). The background contours are model outputs for 2011. The filled circles represent seasonally averaged observations. Aircraft measurements (0–2, 2–6, and 6–10 km) are averaged vertically for each altitude range and horizontally every 1° × 1°. Lower panel: Scatter plot of seasonal surface observations from TCEQ (filled circles) and INSTARR (transparent triangles) datasets.

Figure 9.

Upper panel: Mean distribution of C₄-C₅ alkane abundances for different seasons and altitude ranges compared to observations from aircraft campaigns and surface measurements (Table 3). The background contours are model outputs for 2011. The filled circles represent seasonally averaged observations. Aircraft measurements (0–2, 2–6, and 6–10 km) are averaged vertically for each altitude range and horizontally every 1° × 1°. Lower panel: Scatter plot of seasonal surface observations from TCEQ (filled circles) and INSTARR (transparent triangles) datasets

Across the U.S., there is a seasonal gradient in C₂-C₅ alkane mixing ratios due to the seasonal variations in OH concentrations (upper panels of Figures 7–9); there are higher C₂-C₅ alkane mixing ratios during fall and winter compared to spring and summer (more on this in Section 3.3.2). Most of the aircraft observations (0–10 km) presented here were collected during summer months. In this season, the observations cover most of the CONUS. Although the aircraft campaigns occurred during different years (2008–2014), the almost full coverage of the CONUS provides an overview of the spatial distribution of C₂H₆, C₃H₈, and C₄-C₅ mixing ratios. C₂-C₅ alkane abundances are more homogenous above 2 km compared to the boundary layer. From 2 to 10 km, mixing ratios primarily reflect northern hemisphere background abundances, while in the boundary layer enhancements mirror the spatial distribution of emissions. The largest boundary layer enhancements for these species occur over Colorado, Texas and Oklahoma. The model output shown in Figures 7–9 corresponds to seasonal averages of monthly means. The spatial distribution of tropospheric abundances is determined by the atmospheric lifetime of each C₂-C₅ alkane. Consequently in our model simulations, tropospheric abundances of C₂H₆ and C₃H₈ are more homogeneous across the CONUS compared to abundances of C₄-C₅, which show stronger local enhancements below 2 km.

In some regions (e.g., the Colorado Front Range), most of the observed abundances of C₄-C₅ alkanes have been attributed to oil and gas activities [Abeleira et al., 2017; Gilman et al., 2013]. This is often diagnosed using the ratio of the isomers of butane and pentane [Pétron et al., 2014; C R Thompson et al., 2014]. Enhanced abundances of the C₄-C₅ alkanes compared to background values are well documented over oil and gas regions [Abeleira et al., 2017; Johansson et al., 2014; Swarthout et al., 2013]. For example in the Colorado Front Range, ratios of n-butane, i-pentane, and n-pentane to C₃H₈ in air masses impacted by oil and gas emissions have ranges of 0.43–0.56, 0.13–0.16, and 0.13–0.19, respectively [Gilman et al., 2013; Pétron et al., 2014; Swarthout et al., 2013]. In addition to oil and gas sources, typical urban sources of C₄-C₅ alkanes are landfills [U.S. EPA, 2009] and traffic [Abeleira et al., 2017; Kirchstetter et al., 1996].

A portion of the HIPPO flights did occur in 2011, but only two HIPPO flights (9 and 11 August 2011) cross our region of interest during this period. However, these two HIPPO flights allow us to make a direct comparison between aircraft observations and model output. We sampled the model at the coincident time and location of the observations from the HIPPO flights on 9 and 11 August 2011. Model bias is largest for C₃H₈ and is only apparent at altitudes below 700 hPa. The normalized mean model bias for C₃H₈ (NMB = sum(model – obs)/(sum(obs)) is −44% for the 9 August 2011 flight and −33% for the 11 August 2011 flight. This supports the conclusion drawn from the surface observation comparison for 2011 (Figure 8 lowest row) that the model underpredicts C₃H₈ mixing ratios.

The SEAC⁴RS observations cover a region of the U.S. where we would expect a large influence from emissions from the oil and gas sector. The lower panels of Figures 7 and 8 show that the model simulation underpredicts average TCEQ observations of C₂H₆ and C₃H₈ over Texas (filled circles) by ~50% throughout the year. Though there are few long-term monitors in this region, it is likely that there were changes in the average abundance of C₂-C₅ in this region, even over the two-year period between 2011 (model output) and 2013 (SEAC⁴RS observations). Helmig et al. [2016] summarize observed trends in C₂H₆ and C₃H₈ over the period 2009 – 2014 for long-term surface sites. The largest trend in C₃H₈ over the period 2009 – 2014 in this region in Helmig et al. [2016] is for Moody, Texas. The change is 286 pptv/year. The model, based on 2011 emissions and meteorology, underpredicts the observed 2013 SEACR⁴S C₃H₈ mixing ratios over Texas below 2 km by ~ 1ppbv. Even a trend of 286 pptv/year applied to the SEAC⁴RS data is unlikely to close the model-measurement gap. When we compare average model output to the SEAC⁴RS observations, hypothetical de-trending would still result in an underprediction of the observations by >400 pptv. This rough calculation also supports the conclusion that C₃H₈ is underpredicted by the model.

As discussed in Section 2.2, the mass-weighted OH concentration for our model simulation is at the upper end of what is reported in other modeling studies. If the OH in the model were higher, this would cause the model lifetime of propane to be shorter than reality and this would cause simulated propane to decay more quickly downwind of sources. However, it seems highly unlikely that our hypothesis that propane emissions are too low is a product of unreasonably high OH concentrations for these reasons. 1) The model bias in propane is only below 700 hPa over the U.S. when compared to the HIPPO flights. A problem with OH should produce bias outside of the boundary layer over the U.S. 2) The model underprediction of the propane at the surface is very large (as shown in Figure 8). In many locations, the model under-predicts propane by almost an order of magnitude. There is no indication that OH is incorrect by an order of magnitude. 3) The model is missing an additional sink of propane (halogens). Adding an additional sink, would only serve to push the simulated propane down. This would further support the hypothesis that emissions are too low for this species.

There is a strong diurnal cycle in the mixing ratios of alkanes within the boundary layer [Abeleira et al., 2017; Vinciguerra et al., 2015]. The model output in the upper panels of Figures 7–9 represents seasonal means, thus the model represents an average of the entire diurnal cycle over these seasons. In contrast, the majority of the aircraft observations were collected during the day when local emissions are mixed into a larger volume and reacting with OH. Despite this, the simulated abundances of C₂H₆ and C₃H₈ at altitudes below 2 km are on average 5 and 3 ppb lower, respectively (both modeled and observed abundances are horizontally averaged every 1° × 1°). The discrepancy between the model and the observations is largest for the FRAPPÉ aircraft campaign, which is also the most recent field campaign presented in this study (2014) and encompasses the region with higher annual rates of change of C₂H₆ total column from 2009 to 2014 as estimated by Franco et al. [2016].

Figures 10–12 present the simulated percentage contribution from the oil and gas sector to total abundances of C₂-C₅ alkanes. We use the updated 2011NEI: OG off simulation to estimate the percentage contribution of emissions from this sector to total C₂-C₅ alkane mixing ratios. Of the regions examined here, the lowest contribution of U.S. oil and gas activity to surface mixing ratios of C₂-C₅ alkanes is over California, which has relatively little local oil and gas development compared to the other regions of the U.S. Gentner et al. [2009] reported i-pentane/n-pentane ratios for California during summertime ranging from 2.9 for liquid gasoline to 3.8 for gasoline vapors. The June-July mean i-pentane/n-pentane ratios over this area for the 2008 ARCTAS aircraft campaign and 2011 surface flask observations at the LAX station were 2.0 and 2.2, respectively, suggesting that air masses are dominated by urban sources (Figure 11).

Figure 12.

2011 simulated percentage contribution from the oil and gas sector to total abundances of C₄-C₅ alkanes.

Figure 11.

2011 simulated percentage contribution from the oil and gas sector to total abundances of C₃H₈

Over the central and southeastern U.S. the model attributes a higher percentage of the near-surface C₂H₆ and C₃H₈ to oil and gas related activities, compared to C₄-C₅ alkanes. Figure 11 shows that the model attributes most of the C₃H₈ at the surface to emissions from the oil and gas sector. The estimated oil and gas contribution to near surface C₂H₆ and C₃H₈ mixing ratios over Colorado is consistent with results from Gilman et al. [2013], who estimated mean percentage contributions of 72% and 90%, respectively. However, our estimated contribution from oil and gas sources to C₄-C₅ alkanes is ~10% lower compared to the Gilman et al. [2013] calculation of 93–96%. This difference suggests that our choice to assign C₄-C₅ alkanes as 36% of the total emitted PAR species over the U.S. for all emissions sources, should be revisited over regions where oil and gas activities abut urban areas, like Colorado. For the Colorado Front Range, the percentage of C₄-C₅ alkanes to total emitted PAR species is likely higher than 36%.

Figures 10, 11 and 12 suggest that emissions over oil and gas regions can impact atmospheric abundances over much of the U.S. lower-to-mid free troposphere. This does not imply that the atmosphere is well mixed over a given area from the surface to 10 km on a given day. Rather, it reflects typical characteristic time scales for vertical transport which are ~1 week for mixing in the lower free troposphere and ~1 month for mixing throughout the troposphere. The lifetimes of C₂H₆ and C₃H₈ are on average sufficiently long such that these species can be mixed vertically. However, we note that alkane lifetimes due to OH reactions can have significant seasonality [Miller et al., 2010]. Figures 10 and 11 reflect more vigorous mixing in summer months.

4. Conclusions

We use a GEOS-Chem nested simulation driven by updated 2011NEI emissions in combination with a collection of observations over the U.S. to 1) examine the spatial patterns in observed atmospheric abundances of C₂-C₅ alkanes, and 2) estimate the contribution of the U.S. oil and gas industry to the observed patterns.

The updated 2011NEI, indicates that the oil and gas sector dominated U.S. emissions of C₂H₆ and C₃H₈ with a contribution to total emissions of 89% and 82%, respectively [U.S. EPA, 2017]. Emissions of these two species are largely located inside U.S. oil and gas basins. As implemented in GEOS-Chem, oil and gas sources represent the third most important emission source for C₄-C₅ alkanes. Other fossil fuel sources contribute significantly to the emissions of these larger alkanes, thus their emissions are located not only inside oil-and-gas-producing basins, but also within urban and industrial areas.

Aircraft observations over the period 2008–2014 show that the highest mixing ratios of C₂-C₅ alkanes were encountered over the central U.S. boundary layer (mainly over Colorado, Texas and Oklahoma) during this period. Observations were much more homogenous above 2 km for all the species considered here. Both, the suite of observations and modeled C₂-C₅ alkane abundances, show that U.S. oil and gas emissions impact large regions of the lower troposphere especially over the central and eastern U.S. The surface and limited aircraft observation-model comparisons for C₃H₈ suggest that the emissions of C₃H₈ in the updated 2011NEI may continue to be too low.

Given that increases in C₂-C₅ alkane abundances driven by emissions from the U.S. oil and gas industry began in 2009, we do not recommend using the updated 2011NEI for prior years. There are many locations where oil and gas development is relatively recent. Similarly, the updated 2011NEI precedes much of the extraction of oil and gas in the Bakken. Thus if simple scaling factors were to be applied to this inventory for simulations after 2011, we would not expect that the resulting emissions would represent this area well. Furthermore, the reported increasing trends in atmospheric concentrations of oil and natural gas related emissions during 2010–2015 [Franco et al., 2016; Helmig et al., 2016; Vinciguerra et al., 2015], suggest that the C₂-C₅ alkane emission estimates in this paper are likely a low estimate for years following 2011.

Due to the increasing importance of oil and gas emissions in the U.S., long-term measurements of C₂-C₅ alkanes are needed in order to document how the emissions of these species are changing. We recommend continued support of existing long-term measurements of C₂-C₅ alkanes. We also suggest continuous consistent monitoring of surface mixing ratios in northeastern Colorado, Wyoming and North Dakota. Further, we suggest that the community evaluate whether chemical mechanisms that lump larger alkanes are sufficient to understand air quality issues in regions with large emissions of these species.

Key points:

• Oil and gas development is the largest source of ethane and propane in the U.S.; this sector is the third largest source of C₄-C₅ alkanes.

• Propane is underpredicted over several U.S. regions.

• Boundary layer enhancements of C₂-C₅ alkanes mixing ratios are largest over the central U.S.