Reply to comment on “Hydrocarbon emissions characterization in the Colorado Front Range—A pilot study” by Michael A. Levi


Corresponding author: G. Pétron, NOAA Earth System Research Laboratory, R/GMD1, 325 Broadway, Boulder, CO 80305, USA. (


[1] The Comment by Levi (2012) on our paper, Pétron et al. (2012), presents a different interpretation of the atmospheric data and inventory estimates we used to derive our conclusions about methane emissions from oil and natural gas development in the Denver-Julesburg Basin (DJB) in Weld County, Colorado. Levi's (2012) Comment brings up new issues that point to the need for additional information. We maintain the value of the results derived in Pétron et al. (2012), particularly that vented and fugitive methane emissions from Weld County's fossil fuel exploration and production in 2008 were likely larger and more uncertain than values reported by emission inventories. Our findings rely on the interpretation of high-quality atmospheric observations using existing inventory data provided by the industry and regulatory agencies and on reasonable assumptions about the average vented raw gas composition. However, Levi (2012) has caused us to extend our analysis and to better characterize the uncertainties associated with his and with our approaches. In this Reply, we examine some critical limitations of the Pétron et al. (2012) and Levi (2012) interpretations of the atmospheric data using simple, two-source emission models that incorporate inventory data sets of unknown reliability. We present new evidence that the regulatory estimates of flashing emission and regulatory modeled composition profiles for a limited number of condensate tanks, the starting point for the calculations of Pétron et al. (2012) and Levi (2012), probably do not represent the true range of these parameters for the thousands of such sources across the DJB in 2008. The results of Levi (2012) suggest that leakage in Weld County in 2008 was biased toward dry gas wells, which disagrees with current inventories of venting and fugitive emissions in U.S. oil and gas fields, including the DJB. Most importantly, the indirect flux derivations undertaken by Levi (2012) and Pétron et al. (2012) highlight two inherent shortcomings common to most emissions inventories: their reliance on the extrapolation of very limited information and the difficulty in carrying out a full uncertainty analysis of such datasets. We agree with Levi (2012) that there is an urgent need to statistically document the composition profiles and magnitudes of significant sources in oil- and gas-producing fields. Observations-based methods with established uncertainties and that are completely independent of inventory information could directly quantify emission strengths and compositions of both point and aggregated area sources, providing an objective assessment of inventory methodology and estimates.

1 Introduction

[2] In 2008, Weld County in northern Colorado was home to 248,852 inhabitants, around 550,000 head of cattle, 14,517 producing oil and natural gas wells [Colorado Oil and Gas Conservation Commission (COGCC), 2012], 4813 petroleum liquid storage tanks, and 279 natural-gas-burning engines associated with oil and gas operations, including engines at compressor stations (Colorado Department of Public Health and Environment (CDPHE), personal communication). Weld County produced 204.3 billion cubic feet of natural gas and 17.5 million barrels of condensate in 2008 [Bar-Ilan and Morris, 2012]. There are many potential sources of emissions from Weld County's fossil fuel exploration, production, processing, and transmission system, which encompasses most of the Denver-Julesburg fossil fuel Basin (DJB). According to the Western Regional Air Partnership (WRAP) Phase III effort [Bar-Ilan et al., 2008a,2008b], the most comprehensive examination to date of emissions from oil and natural gas exploration and production (E&P) practices in the DJB and other basins across the Rocky Mountain region, 95% of volatile organic compounds (VOCs) emitted in the DJB in 2008 came from either “flashing” (i.e., emissions of volatile components from oil and condensate storage tanks) or “venting” and “fugitives” (these are releases or leaks of natural gas from wells and adjacent equipment such as pneumatic devices and pumps). Based on the WRAP Phase III inventories [Bar-Ilan et al., 2008a, 2008b], Pétron et al. [2012] and then Levi [2012] used a two-source model to understand methane and VOC emissions from oil and gas E&P in Weld County in 2008. In Pétron et al. [2012], we used the term “venting” to indicate emissions from both known releases and unpermitted fugitive sources of raw natural gas. In this Reply, we instead use the term “fugitives” to be more general.

[3] We first summarize the assumptions and results from Pétron et al. [2012] and Levi [2012]. We then present a new in-depth analysis of these data sets motivated by some of the inconsistencies brought to light by Levi's results. Specifically, we address the following issues:

  1. Why the two-source models used by Pétron et al. [2012] and Levi [2012], and the flashing composition profiles and emission estimates on which such models depend, may not be representative of the actual emissions situation in the DJB in 2008.
  2. Why Levi's [2012] findings that suggest leakage is biased towards dry gas wells in Weld County in 2008 disagree with our current understanding of fugitive emissions in that region and at the national level.

[4] We conclude the Reply by highlighting the need for observations-based approaches to estimate emissions from fossil fuel E&P. In the auxiliary material, we present two minor corrections to our original paper.1

2 Summary of Pétron et al. [2012] and Levi [2012] Approaches and Main Findings

[5] In Pétron et al. [2012], we reported atmospheric observations of several VOCs, including methane, propane, and n-butane, collected at a NOAA tall tower site (the Boulder Atmospheric Observatory, or BAO) at the southwestern edge of the DJB and throughout the DJB using an instrumented vehicle (called the Mobile Lab, or ML). From (1) the strong correlations between methane and light alkanes in the DJB samples, (2) the strong similarities in the atmospheric enhancement ratios with those of either flashing or fugitive profiles, and (3) the lack of propane in other methane sources encountered in the region (feedlots, a landfill, and a wastewater treatment plant), we concluded that the majority of the atmospheric enhancements of methane and light alkanes in the DJB were a result of oil and natural gas E&P activities.

[6] In Pétron et al. [2012], we interpreted the atmospheric methane, light alkanes, and benzene correlations we observed in the DJB using a simple two-source model, in which flashing and fugitive emissions were assumed to be the only sources from oil and natural gas E&P (this model is examined further in section 3). To derive the absolute emissions of methane and other VOCs from these sources, we relied on the observed correlations of the atmospheric mixing ratio enhancements and several inventory datasets. From the only inventory and regulatory raw gas composition data that were publicly available at the time of our analysis, we picked three scenarios for the composition profile of fugitive emissions: (1) the mean raw gas composition profile employed by Bar-Ilan et al. [2008a,2008b] in the WRAP Phase III inventory for the DJB, and the (2) median and (3) mean raw gas profiles from a database of 77 natural gas wells sampled across the Wattenberg field of the DJB in 2006 and reported by COGCC [2007]. In 2008, 91.3% of Weld County's natural gas production came from the Wattenberg field. Figure 1 shows the distribution of the methane-to-propane ratios in the 77 raw gas samples analyzed by COGCC [2007]. The three “middle of the road” scenarios picked by Pétron et al. [2012] assume that oil and natural gas wells from all geological formations producing in the DJB potentially leak. We then used a bottom-up estimate of total VOC flashing emissions for 2008 that was an average of the WRAP Phase III 2006 and 2010 inventories [Bar-Ilan et al., 2008a,2008b], as well as flashing composition profiles modeled for 16 condensate tanks in the DJB as part of a regulatory study (CDPHE, personal communication). From this information, we derived three scenarios for fugitive methane emissions in Weld County in 2008 that ranged from 67.6 to 228.9 Gg/yr.

Figure 1.

Histogram of molar methane-to-propane raw gas composition data collected at 77 gas wells in the Denver-Julesburg Basin in 2006 [COGCC, 2007], colored by the geological formations producing in the Basin at the time. The mean and median values of this distribution are indicated by the dotted red lines. The star indicates the mean ratio used in the WRAP III inventory.

[7] In his Comment, Levi [2012] proposes another two-source interpretation of the data collected by NOAA that uses the same inventory estimates as Pétron et al. [2012]. However, instead of assuming a “mean” methane-to-propane ratio for the fugitive gas, as in the Pétron et al. [2012] scenarios, Levi uses the additional constraints of (1) a strong correlation between propane and n-butane in the atmospheric observations (slope = 2.24 ± 0.07 mol/mol (R2 = 1) for flasks from the BAO northeast wind sector and slope = 2.04 ± 0.05 mol/mol (R2 = 1) for Mobile Lab flasks), and (2) a loose correlation between propane and n-butane (slope b = 4.15 with 95% confidence interval of [2.5–6.6] mol/mol) in the COGCC raw gas data set. In his main Comment, Levi presents the set of equations used to derive absolute fugitive emissions for propane, n-butane, and methane. He also explains how he propagates uncertainties in the measurements and in the slope b to compute a range of emission estimates. The quantitative derivation of the results from his model is presented in the supplementary material to Levi [2012], and it will be assessed in detail in section 3. We note here that to yield non-negative fugitive emissions, Levi's model had to exclude all but one of the 16 regulatory flashing emission profiles. The resulting limited set of non-negative results also suggests that leakage in Weld County is biased toward dry gas wells, i.e., wells that produce predominantly methane with relatively little additional hydrocarbons. His estimated range of the fugitive methane source is similar to the range based on the WRAP Phase III VOC inventory (42–63 Gg/yr) and is inconsistent with our top-down scenarios in Pétron et al. [2012].

3 Evaluation of Two-source Model and CDPHE Flashing Emission Data

[8] The two-source models assumed by Levi [2012] and by Pétron et al. [2012] do not consider significant emissions from sources other than fugitives (which are assumed to be described by either the COGCC's or the WRAP's raw gas composition profiles) and flashing (assumed to be represented by WRAP's total VOC emissions and the CDPHE's modeled condensate tank flashing composition profiles). The limitation of this approach appears clearly when the fluxes of n-butane and propane from Levi's two-source model are plotted (Figure 2). In Figure 2, the observed atmospheric correlation slope between propane and n-butane from the NOAA Mobile Lab measurements is shown by the black dotted line. The set of flashing emission estimates of propane and n-butane based on the WRAP estimate of total VOC flashing emissions and the flashing chemical speciation modeled by CDPHE for 16 tanks is given by the light gray squares; these data represent the starting point for the Pétron et al. [2012] and Levi [2012] calculations. Levi's calculations derive the fugitive fluxes of propane and n-butane that are needed so that the total fluxes of propane and n-butane (in moles, shown by the black triangles in Figure 2) match the observed atmospheric correlation slope. We represent the effect of Levi's calculated fugitive emissions as blue or red segments in Figure 2 connecting the starting point of the calculations (gray squares) to the final result of the calculations (black triangles). Red segments indicate that fugitive sources of propane and n-butane must be positive to achieve the atmospheric ratios, while blue segments indicate that propane and n-butane must be removed from the atmosphere to achieve the observed atmospheric ratios. As described in the supplementary material of Levi [2012] and explicitly reported in Tables 1 and 2, Levi's calculation implies that only one condensate tank flashing composition profile (tank #14) out of the set of 16 in the CDPHE dataset results in non-negative propane and butane fugitive fluxes (the values for this tank are highlighted in Figure 2). For the other 15 modeled flashing composition profiles, Levi's calculation requires that fugitive emissions must remove propane and n-butane to match the atmospheric ratios.

Figure 2.

Graphic showing the assumed propane and n-butane fluxes for flashing sources (gray squares) and the calculated fugitive propane and n-butane fluxes using Levi's two-source model (shown by blue lines when these fugitive sources are negative and red lines when positive). Only the n-butane and propane flashing emissions for one tank (#14) fall below the observed atmospheric slope and therefore yield non-negative fugitive propane and n-butane fluxes. Levi's two-source model ensures that the final total propane and n-butane sources (black triangles) fall on the observed atmospheric slope (dashed black line, shown here for Mobile Lab samples).

Table 1. Fugitive Propane (Xp), n-Butane (Xb), and Methane (Xm) Emissions in Gmol/yr Derived with Levi's Equations for Three Values of the Propane-to-n-butane Molar Slope b of the Raw Gas (2.5, 4.15 and 6.6) and for the NOAA Mobile Lab (ML) Atmospheric Propane-to-n-butane Molar Slope of 2.04a
MLb = 2.5b = 4.15b = 6.6
  1. a

    Note that all three fugitive emissions Xp, Xb, and Xm are positive only for tank #14. The methane-to-propane fugitive molar emission ratios (Xm/Xp) for tank #14 are 36.5 for b = 2.5, 83 for b = 4.15, and 110 for b = 6.6.

Table 2. Fugitive Propane (Xp), n-Butane (Xb), and Methane (Xm) Emissions in Gmol/yr Derived with Levi's Equations for Three Values of the Propane-to-n-butane Molar Slope b of the Raw Gas (2.5, 4.15 and 6.6) and for the NOAA BAO Atmospheric Propane-to-n-butane Molar Slope of 2.24a
BAOb = 2.5b = 4.15b = 6.6
  1. a

    Note that all three fugitive emissions Xp, Xb, and Xm are positive only for tank #14. The methane-to-propane fugitive molar emission ratios (Xm/Xp) for tank #14 are 15.5 for b = 2.5, 34 for b = 4.15, and 45 for b = 6.6.


[9] The maximum fugitive methane emission estimates reported by Levi in his Comment correspond to the solutions of his system of equations using only the flashing emissions associated with the composition profile of tank #14. Levi's minimum fugitive methane emission estimates rely on a combination of the tank #14 positive estimate and a negative contribution from the estimate associated with tank #8. Given that we are dealing with source processes (leaks) that must be zero or positive (emitting to the atmosphere, not removing mass from the atmosphere), any single fugitive emission has to be positive to be considered part of the solution. Using Levi's notations for the propane flashing emission associated with tank L flashing composition profile, we maintain that XpL ≥ 0 with L = 1, 16 is the correct formulation, instead of the weaker necessary condition placed on the calculation by Levi [2012] that a linear combination of emissions from various flashing tanks must lead to a non-negative total fugitive source. As a result, the solution space for Levi's derivation should be limited to the total emissions calculated with the composition profile of tank #14. Levi's minimum scenarios using negative methane, propane, and n-butane emissions associated with the profile of tank #8 (see Levi's supplementary material) should not be considered as part of his solution.

[10] At this point in the evaluation of Levi's results, it is important to step back and reassess the assumptions and data sets used in both of our calculations. Figure 3 shows that the current regulatory data sets we used for the flashing and fugitive two-source composition profiles are incompatible with the ambient levels of propane and n-butane we measured. By plotting the n-butane and propane molar fractions (in %) assumed for the two source components (flashing and fugitive emissions) and the molar fraction enhancements (in ppb) observed in the atmosphere, it is clear that the atmosphere is richer in n-butane than any of the potential sources (represented by the squares) considered by either Pétron et al. [2012] or Levi [2012], except for the one flashing source associated with tank #14.

Figure 3.

Molar fractions of n-butane and propane measured in the raw gas (in %, green squares; [COGCC, 2007]), modeled for the flashing emissions from 16 condensate tanks (in %, gray squares; CDPHE, personal communication) and measured in air samples collected in the DJB in the summer by the Mobile Lab (blue circles) and at the NOAA BAO tower (red circles, midday northeast wind sector samples only; see Pétron et al. [2012] for more details on the wind sector filter). The propane-to-n-butane atmospheric slopes from these two datasets are shown by the black dotted and dashed lines. The minimum, maximum, and average propane-to-n-butane slopes calculated by Levi's for the COGCC 77 raw gas samples measurements are shown by the green lines.

[11] The lack of overlap between the atmospheric propane-to-n-butane ratio and all fugitive and flashing composition profiles except one suggests that either the available measured raw gas profiles and modeled flashing emissions profiles are not enriched enough in n-butane relative to propane or that there are n-butane-rich (relative to propane) sources missing in our two-source interpretations. The tightness of the observed atmospheric molar fractions correlations suggests that if there are missing n-butane-rich sources, they must be colocated with, and proportional to, existing sources of propane. Levi discusses none of these possibilities in his Comment, and instead draws his conclusions from results derived with a single flashing emission composition profile for the entire Basin. However, our analysis of his results reveals that the simple two-source models and the existing regulatory composition data sets used by both Levi [2012] and Pétron et al. [2012] likely cannot fully describe the observed propane and n-butane atmospheric ratios.

[12] As shown above, Pétron et al.'s [2012] and Levi's [2012] hybrid top-down/bottom-up approaches rely heavily on inventory estimates from self-reported data sets or regulatory model calculations that use a very limited set of measurements for different types of sources and control technology. The WRAP total VOC emission estimate for flashing from condensate tanks in the DJB in 2006 is based on monthly emission estimates reported by operators to CDPHE. The 16 flashing composition profiles, including the propane-to-n-butane ratio, are model results provided by CDPHE. The reliability or representativeness of these estimates is not generally known, and the true uncertainty of these inventories is likely to be underestimated. As we describe in detail below, changes in inventory methodological assumptions can also have a significant impact on predicted emission magnitudes. Recent developments in regulatory emission inventory methodology have shown that methane and VOC emission levels can be quite sensitive to changes in mitigation efficiency assumptions and process-specific emission factors [EPA, 2012; Shires and Lev-On, 2012; Bar-Ilan and Morris, 2012].

[13] Specifically, both Pétron et al. [2012] and Levi assume that the total VOC flashing source in 2008 is fixed (41.3 Gg) based on averaging the sum of CDPHE-reported total VOC emission estimates for small and large condensate tanks for 2006 [Bar-Ilan et al., 2008a] and projections by WRAP Phase III for 2010 [Bar-Ilan et al., 2008b]. At this time, there are still no reported uncertainty estimates attached to the WRAP Phase III [Bar-Ilan et al., 2008a,2008b] and the more recently developed WestJump [Bar-Ilan and Morris, 2012] total VOC emission inventories for upstream operations in oil and natural gas fields. In 2012, CDPHE introduced a capture efficiency estimate of 0.75 for VOC emissions from control devices (typically flares) installed on condensate tanks [Bar-Ilan and Morris, 2012]; the previous assumption for capture efficiency on these sources was 100%. This single change (see [Bar-Ilan and Morris, 2012] for details on the implementation), made after Pétron et al. [2012] was published, resulted in a substantial increase of the total VOC flashing source estimate for Weld County in 2008, from 41.3 Gg to 67.0 Gg/yr. We conclude that the full range of possible flashing emission profiles and the total VOC flashing source are likely not captured by the set of 16 flashing emission estimates used by Pétron et al. [2012] and Levi [2012].

4 Could Dry Gas Wells Be the Only Ones Leaking in 2008?

[14] Our analysis of Levi [2012] suggests that natural gas leakage in Weld County is biased towards dry gas wells. According to Schlumberger's online oilfield glossary [Schlumberger, 2012], dry gas is defined as “natural gas that occurs in the absence of condensate or liquid hydrocarbons, or gas that has had condensable hydrocarbons removed.” This definition also mentions that “dry gas typically has a gas-to-oil ratio exceeding 100,000 standard cubic feet/stored tank barrel” or an oil-to-gas ratio of less than 10 barrels of oil/million cubic feet (bbl/MMcf) of gas produced.

[15] The methane-to-propane molar ratio in the fugitive sources inferred by Levi is 83 (ranging from 36.5 (slope b = 2.5) to 110 (b = 6.6)) for the Mobile Lab set of solutions and 34 (ranging from 15.5 (b = 2.5) to 45 (b = 6.6)) for the BAO set of solutions. Since Levi did not present his results for propane and n-butane fugitive emission estimates, we have included them here in Tables 1 and 2. Results in these tables are based on the flashing emissions for the 16 reference tanks used in both of our analyses and shown in Table 3 of Levi [2012]. Levi's estimates of “modest” fugitive methane emission rates correspond to methane-to-propane molar ratios above 35 in the leaked and vented gas. From Figure 1, such ratios are only observed in a subset of the producing gas wells in the Wattenberg field, specifically wells tapping the dry gas J Sandstone formation alone or the J Sandstone together with other reservoirs. Higley and Cox [2007] report a median oil-to-gas ratio of 10.43 bbl/MMcf for Muddy “J” Sandstone production across the Wattenberg field. Levi's findings suggest that more liquid-rich wells from the Sussex Sandstone, Codell Sandstone, and Niobrara Shale are not leaking substantial amounts of gas.

[16] We further assess the meaning of Levi's finding using reported 2008 production statistics for all 14,517 producing wells in Weld County [COGCC, 2012]. In 2008, 60% of the natural gas produced in Weld County came from wet gas wells, leaning oil, and oil wells (Figure 4). Only 7.6% of the natural gas produced in Weld County came from dry gas wells (Figure 4), defined as wells with an oil-to-gas ratio of less than 10 bbl/MMcf. Our analysis of Levi's results, that the fugitives are emitted mostly from dry gas wells, suggests that most of the fugitive gas in Weld County in 2008 came from a small subset of the producing wells (1043 dry gas wells out of a total of 14,517 producing oil and gas wells in 2008). We think that Levi's finding ought to be tested in the field by measuring emissions at a “representative” subset of dry gas wells and wet gas wells.

Figure 4.

Fraction of wells, gas production, and oil production covered by various types of wells in Weld County in 2008, based on COGCC data retrieved on 29 September 2012. Oil-to-gas ratios are between 0 and 10 bbl/MMcf for dry gas wells, 10 and 40 bbl/MMcf for leaning dry gas wells, 40 to 67 bbl/MMcf for wet gas wells, 67–300 bbl/MMcf for leaning oil wells, and >300 bbl/MMcf for oil wells. The classification in oil and gas types used here is inspired by categories used in Nelson and Santus [2011].

[17] The current understanding of fugitive and vented emissions from oil and gas fields does not discriminate between dry and wet gas wells to calculate emissions but instead takes into account the equipment and types of activities which are common to most well pads. The only available and updated inventory of methane emissions from U.S. natural gas systems is the national greenhouse gas emissions inventory developed by the U.S. Environmental Protection Agency annually for the United Nations Framework Convention on Climate Change [EPA, 2012]. Although aggregated into large geographic regions, emissions from various source categories in the national inventory provide some knowledge about our current understanding of major contributors to the total methane source. Taking into account the fossil fuel industry's revisions to emissions from well liquid unloading and unconventional well refracturing [Shires and Lev-On, 2012], the inventory reports that major methane contributors are pneumatic devices (27%), liquid unloading (14%), completion (14%), offshore (7%), field separation (6%), compressor exhaust (6%), Kimray Pumps (6%), gathering compressors (6%), and well recompletion (4%). Each of these sources can occur for both dry and wet gas producing well pads and gathering systems.

[18] Similarly, in the WRAP Phase III inventory, based on operators' surveys for 2006 activities in the Denver-Julesburg Basin, [Bar-Ilan et al., 2008a] and in the 2008 WestJump inventory [Bar-Ilan and Morris et al., 2012], 49% of the total VOC fugitive emissions were attributed to pneumatic devices and 32% to unpermitted fugitive emissions from various pieces of equipment at well pads. Effective May 2009, the Colorado Air Quality Control Commission adopted revisions to Regulation No. 7 (adopted in December 2008) that required low-bleed controllers on all valves used in the Denver/Colorado Front Range ozone non-attainment area, which includes Weld County. During the summer of 2008, when Pétron et al. [2012] first deployed their Mobile Lab to study methane sources in Weld County, high-bleed valves were likely used across the oil and gas field. Unpermitted fugitive emissions were also likely ubiquitous in the Denver-Julesburg Basin in 2008. Therefore, it seems unlikely that less than 8% of producing wells (i.e., dry gas wells) could have contributed all or most of the fugitive emission signal we detected with our observations in Weld County in 2008.

5 Conclusions

[19] To complement emission inventory efforts and support air quality and climate assessments, field studies should be directed at better characterization and quantification of methane and VOC emissions in oil and gas basins. So far most emission estimates [EPA, 2012; Bar-Ilan et al., 2008a] have relied on the extrapolation of limited information regarding emissions composition, source activity, emission factors, mitigation practices, and control effectiveness. Pétron et al. [2012] and Levi [2012] have shown how atmospheric enhancement ratios can be used to evaluate the mix of flashing and venting emissions in the DJB. Using the same input data sets and different assumptions to derive fugitive emissions of natural gas, they derive very different results for the amount of natural gas that is leaked or vented to the atmosphere. The main conclusion of this Reply is that both approaches have major limitations and may not cover the true range of uncertainties associated with fugitive and venting methane sources in the DJB.

[20] We agree with Levi [2012] that more direct point source measurements should be gathered to build statistics on methane and VOC emission factors, composition profiles, and source magnitudes representative of today's oil and gas exploration, production, and processing activities in different regions [EPA/GRI, 1996; Mellqvist et al., 2010; Amin et al., 2011; Cuclis, 2012] and to compare emission reductions achieved with various mitigation practices or pieces of equipment. Concurrently, direct area source measurements similar to Turnbull et al. [2011] and Ryerson et al. [2011] could provide a truly independent constraint to evaluate the overall methane and VOC aggregated sources for an entire oil and gas field. These independent measurements are necessary to assess how well the cataloging and extrapolation approaches used in inventories agree with regional fluxes derived from atmospheric observations

6 Supplementary Material: Corrections to Pétron et al. [2012] manuscript

[21] The first correction is the name of the model used by the Colorado Department of Public Health and Environment (CDPHE) to derive flashing emission profiles for the data set of 16 condensate tanks in the DJB sampled in 2002. The manuscript cites the EPA TANK model. Instead, as mentioned in the Western Regional Air Partnership Phase III (WRAP III) report [Bar-Ilan et al., 2008a], the E&P TANK model was in fact used by CDPHE to calculate the flashing emissions profiles.

[22] The second correction, as mentioned in Levi's [2012] Comment, is the reversal of the columns referring to scenarios 2 and 3 in Table 4 of Pétron et al. [2012]. Scenario 2 corresponds to a methane-to-propane molar ratio for the vented gas of 24.83, while scenario 3 corresponds to a ratio of 15.43. The data in all columns of Table 4 referring to these two scenarios should be swapped.