Interactive comment on “ Is Shale Gas a Major Driver of Recent Increase in Global Atmospheric Methane ? ”

4) for the δ13C data I used, I chose the studies reviewed in Golding et al. (2013) because these were highlighted by Schwietzke et al. (2016) (one of the key earlier papers on changes in the global δ13C value of methane) as representative for shale gas. Note that Schwietzke et al. (2016) did not include these shale gas data (either explicitly as distinct from conventional natural gas or as part of their overall estimate for fossil fuels), stating in their supplemental materials: “Note that δ13CFF in this analysis excludes shale gas methane because the share of these sources to global total NG production increased from only 3% to 9% between 2007 and 201316.” Reference #16 is to Golding et al. (2013). As an aside, I believe the important context for looking at changes in methane over the past decade is not the increase from 3% to 9% of total natural gas production, but rather the fact that “shale gas accounted for 63% of the global increase in all natural gas production between 2005 and 2015 (EIA 2016, IEA 2017)” as stated in my manuscript (lines 29-30, p. 4).

, an increase of 7% in global human-caused methane emissions. The change in the stable carbonδ 13 C ratio of methane in the atmosphere over the past 35 years is striking and seems clearly related to the change in the methane concentration ( Fig. 1-B). For the final 20 years of the 20 th Century as atmospheric methane concentrations rose, the isotopic composition became more enriched in the heavier stable isotope of carbon, 13 C, relative to the lighter and more abundant isotope 12 C, resulting in a less negative δ 13 C signal. The isotopic composition remained constant from 1998 to 2008 when the 5 atmospheric concentration was constant. And the isotopic composition has become lighter (depleted in 13 C, more negativeδ 13 C) since 2009 as atmospheric methane concentrations have been rising again (Schaefer et al. 2016). Since biogenic sources of methane tend to be lighter than the methane released from fossil-fuel emissions, Schaefer et al. (2016) concluded that the increase in atmospheric methane in the late 20 th Century was due to increasing emissions from fossil fuels, but that the increase in methane since 2006 is due to biogenic sources, most likely tropical wetlands, rice culture, or animal agriculture. 10 Their model results indicated that fossil fuel sources have remained flat or decreased globally since 2006, playing no major role in the recent atmospheric rise of methane. Schaefer et al. (2016) noted that their conclusion contradicts many reports of increased emissions from fossil fuel sources over this time, and stated that their conclusion "is unexpected, given the recent boom in unconventional gas production and reported resurgence in coal mining and the Asian economy." Six months after the Schaefer et al. (2016) study was published in Science, Schwietzke et al. (2016) presented a similar analysis in Nature that 15 used a larger and more comprehensive data set for the δ 13 C values of methane emissions sources. They too concluded that fossil fuel emissions have likely decreased during this century, and that biogenic emissions are the probable cause of any recent increase in global methane emissions.

20
Model analyses that useδ 13 C methane data to infer emission sources are highly sensitive to changes in the rate of biomass burning: although biomass burning is a relatively small contributor to global methane emissions, those emissions are quite enriched in 13 C relative to the atmospheric methane signal (Rice et al. 2016). Both Schaefer et al. (2016) and Schwietzke et al. (2016) assumed that biomass burning had been constant in recent years. However, Worden et al. (2017) Schwietzke et al (2016) for δ 13 C values of methane emission sources, but including changes in biomass burning over time, Worden et al. (2017) concluded that the recent increase in methane emissions is likely driven more by fossil fuels than by biogenic sources, with an increase of 15.5 Tg per year from fossil fuels (± 3.5 Tg per year) compared to an increase of 12 Tg per year from biogenic sources (± 2.5 Tg per year) when comparing 2007-2014 vs 2001-2006. 30 Clearly global models for partioning methane sources based on theδ 13 C approach are sensitive to assumptions about seemingly small terms such as decreases in biomass burning. In this paper, we explore for the first time another assumption: that the global increase in shale gas development may have caused some of the depletion of 13 C in the global average methane observed over the past decade. Shale gas emissions were not explicitly considered in the models presented by Schaefer et al. (2016) and Worden et al. (2017) and were explicitly excluded in the analysis of Schwietzke et al. (2016).

What is shale gas? 5
Shale gas is a form of unconventional natural gas (mostly methane) held tightly in shale rock formations.
Conventional natural gas, the dominant form of natural gas produced during the 20 th Century, is composed largely of methane that migrated upward from the underlying sources such as shale rock over geological time, becoming trapped under a geological seal (Fig. 2-A). Until this century, shale gas was not commercially developable. The use of a new combination of 10 technologies in the 21 st century -high precision directional drilling, high-volume hydraulic fracturing, and clustered multiwell drilling pads --has changed this. In recent years, global shale gas production has exploded 14-fold, from 31 billion m 3 per year in 2005 to 435 billion m 3 per year in 2015 ( Fig. 2-B), with 89% of this production in the United States and 10% in Canada (EIA 2016). Shale gas accounted for 63% of the total increase in natural gas production globally over the past decade (EIA 2016, IEA 2017. The US Department of Energy predicts rapid further growth in shale gas production globally, reaching 15 1,500 billion m 3 per year by 2040 (EIA 2016; Fig. 2

-B).
Several studies have shown that theδ 13 C signal of methane from shale gas is often lighter (more depleted in 13 C) than that from conventional natural gas (Golding et al. 2013;Botner et al. 2018). Here, we use the data from Figure 1 in the review by Golding et al. (2013) that were explicitly identified as shale gas. The samples are from the New Albany shale (Martini et 20 al. 1998), the Antrim shale (McIntosh et al. 2002), and an organic-rich shale in the northern Appalachian basin (Osborn and McIntosh 2010). Note that these studies appear to be the only ones included in theδ 13 C methane data repository published by Sherwood et al. (2017), which is the data set underlying the analysis by Schwietzke et al.(2016). Out of 61 data points for shale gas in the Golding et al. (2013) figure, only 5 had δ 13 C values similar to those for conventional natural gas, while many samples more closely resembled the signal for biogenic gas. From the 61 values, we calculate a mean value δ 13 C for shale gas 25 of -51.4 o /oo , with a 95% confidence limit of ± 1.2 o /oo. Thus, emissions of methane from shale gas are on average depleted in 13 C relative to atmospheric methane, while methane from conventional natural gas is more 13 C-enriched than atmospheric methane.
It should perhaps not be surprising that the δ 13 C of methane from shale gas tends to be lighter than for conventional 30 natural gas. In the case of conventional gas, the methane has migrated over geological time frames from the shale and other source rocks through permeable rocks until trapped below a seal ( Fig. 2-A). During this migration, some of the methane is likely oxidized by bacteria, perhaps using iron (III) or sulfate as the source of the oxidizing power (Whelan et al. 1986;Rooze et al. 2016). Partial consumption of methane by bacteria would fractionate the methane by preferentially consuming the lighter 12 C isotope and so, gradually enriching the remaining methane in 13 C (Baldassare et al. 2014), resulting in a δ 13 C signal that is less negative. The methane in shales, on the other hand, is tightly held in the rock formation and therefore less likely to have been subject to bacterial oxidation and the resulting fractionation. The expectation, therefore, is that methane in conventional natural gas should be heavier and less depleted in 13 C than is the methane in shale gas.

Calculating the effect of 13 C signal of shale gas on emission sources
To explore the contribution of methane emissions from shale gas, we build on the analysis of Worden et al. (2017). axis) and the δ 13 C values of those emissions (x-axis) by individual sources. Our addition is to separately consider shale gas emissions, recognizing that methane emissions from shale gas are more depleted in 13 C than for conventional natural gas or all other fossil fuels as considered by Worden et al. (2017). For this analysis, we accept that net total emissions increased by 24.7 Tg per year (± 14. Tg per year) since 2008, driven by an increase of ~28.4 Tg per year for the sum of biogenic emissions and emissions from fossil fuels and a decrease of ~3.7 Tg per year for emissions from biomass burning (Worden et al. 2017). 15 We start with the Eq. (1) which reweights the information in Figure 3 (1) 20 where 12 Tg per year is the mean estimate from Worden et al. (2017) for the increase in biogenic emissions, DB-A is the difference in the δ 13 C value for biogenic emission sources and atmospheric methane in 2005, B is the estimate for the increase in biogenic emissions, SG is the estimate for the increase in methane emissions from shale gas, DSG-CG is the difference in the δ 13 C value for shale gas and conventional natural gas, and DA-CG is the difference in the δ 13 C value for atmospheric methane 25 in 2005 and for emissions from conventional natural gas. The x-axis of Figure 3-B shows the δ 13 C for each source; note that the y-axis is the estimate of the change in emissions for each of these sources that we derive below. where TG is total increase in emissions from all natural gas. Note that we test this assumption later in our sensitivity analyses, 5 since some research indicates emissions from shale gas are higher than for conventional gas as a percentage of gas production.
Next, we estimate the likely contributions from coal and oil to the increased methane emissions over the past decade.  (Table 1). From Eq. (4), increased emissions from conventional natural gas are then estimated as 6.5 Tg per year, and from all natural gas (shale plus conventional) as 17.5 Tg per year. From Eq. (7), increased emissions from biogenic sources 10 are estimated as 8.0 Tg per year. The confidence bounds on these estimates, calculated using Eq. (9) and the upper and lower 95% confidence limits for the δ 13 C ratio terms ( Fig. 3-B), are relatively small (Table 1).

15
Our best estimate for the increase in methane emissions from all fossil fuels since 2008 (shale gas, conventional natural gas, coal, and oil) is 20.4 Tg per year (Table 1)

10
Our analysis contain two major assumptions: 1) that methane emissions as a percentage of gas produced are the same for shale gas and conventional natural gas (Eq. (2) and Eq. (3)); and 2) that emissions from oil have remained proportional to the global rate of oil production. Here we explore the sensitivity of our analysis to these assumptions. With regard to the first assumption, some evidence suggests that percent emissions may be higher from shale gas than from conventional natural gas, perhaps due to venting at the time of flow-back following high-volume hydraulic fracturing of shale-gas wells (Howarth et al. With this change in assumptions, estimated shale gas emissions increase by 25% (13.9 instead of 11 Tg per year). Biogenic 20 emissions decrease by 23% (6.2 instead of 8 Tg per year), while total fossil fuel emissions increase (22.2 instead of 20.4 Tg per year). The fossil fuel emission estimate is now 3.6-fold larger than the biogenic emission estimate (Table 2).
Our second major assumption in the base analysis is that methane emission factors for oil production have remained constant over time as a function of production. This may not be true, since 60% of the increase in global oil production between 25 2005 and 2015 was due to tight oil production from shales using the same technologies that allowed shale gas development, high-precision directional drilling and high-volume hydraulic fracturing (calculated from data in EIA 2015 and EIA 2018).
Large quantities of methane are often co-produced with this tight shale oil, and because oil is a much more valuable product than natural gas, for shale-oil fields removed from easy access to natural gas markets, much of the methane may be vented or flared rather than delivered to market. This may be part of the reason for the large increase in methane emissions in recent 30 years in the Bakken shale fields of North Dakota (Schneising et al. 2014).
For this sensitivity scenario #2, we modify equations Eq. (1) through Eq. (9)  this, we follow the approach of Schneising et al. (2014) in combining shale gas and shale oil, scaling the increase in production since 2005 by the energy value of the two products. As in our baseline analysis developed in equations Eq. (1) through Eq.
(9), we assume that conventional natural gas and shale gas have the same percentage methane emission per unit of produced gas. Here we further assume that shale oil has the same emission rate as well, scaled to the energy content of oil compared to natural gas. This sensitivity analysis increases total emissions from fossil fuels by 18% ( Table 2). The contribution from shale gas falls somewhat (from 11 to 9.9 Tg per year), as does that from conventional natural gas (from 6.5 to 5.4 Tg per year), while shale oil becomes an important emission source (5.5 Tg per year). Overall in this scenario, increased emissions from fossil fuels extracted from shales (gas plus oil) are 15.4 Tg per year, two-thirds of the total increase due to fossil fuels.

Conclusions
We conclude that increased emissions from fossil fuels are far more likely than biogenic emissions to have driven the observed global increase in methane over the past decade (since 2008). The increase in emissions from shale gas (perhaps in combination with those from shale oil) makes up more than half of the total increased fossil fuel emissions. That is, the 15 commercialization of shale gas and oil in the 21 st Century has dramatically increased global methane emissions. However, we note an important caveat: our analysis of emissions with explicit consideration of the δ 13 C value for methane in shale gas is based on a small data set, only 61 samples in 3 studies. A clear priority should be to gather more data on the 13 C content of shale-gas methane.

20
Note that while methane emissions are often referred to as "leaks," emissions include purposeful venting, including the release of gas during the flowback period immediately following hydraulic fracturing, the rapid release of gas from blowdowns during emergencies but also for routine maintenance on pipelines and compressor stations (Fig. 4-A), and the steadier but more subtle release of gas from storage tanks (Fig. 4-B) and compressor stations to safely maintain pressures (Howarth et al. 2011). This suggests large opportunities for reducing emissions, but at what cost? Do large capital investments 25 for rebuilding natural gas infrastructure make economic sense, or would it be better to move to phase natural gas out as fuel and instead invest in a 21 st Century energy infrastructure that embraces renewable energy and much more efficient heat and transportation through electrification (Jacobson et al. 2013)?
In October 2018, the Intergovernmental Panel on Climate Change issued a special report, responding to the call of 30 the United Nations COP21 negotiations to keep the planet well below 2 o C from the preindustrial baseline (IPCC 2018). They noted the need to reduce both carbon dioxide and methane emissions, and they recognized that the climate system responds more quickly to methane: reducing methane emissions offers one of the best routes to immediately slowing the rate of global warming (Shindell et al. 2012). Nonetheless, the model scenarios presented in the IPCC report emphasize reducing carbon dioxide emissions first, and these scenarios begin to reduce methane emissions only after 2030. This may reflect the belief of the IPCC authors that methane emissions are dominated by biogenic sources, which are difficult to reduce. Given our conclusion that the oil and gas industry is more likely responsible for recent increases in these emissions, we suggest that the best strategy is to move as quickly as possible away from natural gas, reducing both carbon dioxide and methane emissions.
Doing so will in fact make it easier to reach the COP21 target than predicted by the IPCC (2018). 5 Finally, in addition to contributing to climate change, methane emissions lead to increased ground-level ozone levels, with significant damage to public health and agriculture. Based on the social cost of methane emissions of $2,700 to $6,000 per ton (Shindell 2015), our baseline estimate for increased emissions from shale gas of 11 Tg per year has resulted in damage to public health, agriculture, and the climate of $30 to $65 billion USD per year for each of the past several years. This exceeds 10 the wholesale value for this shale gas over these years.
Appendix A. Sensitivity case #1: emissions per unit of gas produced assumed to be 50% greater for shale gas than for conventional gas. If we use mean values for the differences in the δ 13 C terms in Eq. (A7) (as we did previously for Eq. (9)

20
For the base analysis presented in the main text using equations Eq. (1) through Eq. (9), we assumed that increased emissions from the additional oil development over the past decade were proportional to the increase in that rate of development. That is, the oil produced in recent years had the same emission factor as for oil produced a decade or more ago.
However, 60% of the increase in oil production globally between 2005 and 2015 was for tight oil from shale formations (calculated from data in EIA 2015 and EIA 2018), and methane emissions from this shale oil may be greater than for 25 conventional oil. In this sensitivity case #2, we consider increased emissions from conventional oil and from tight shale oil separately. For conventional oil, the increase in emissions is 40% of the total oil emissions from the base analysis (40% of 1.6 Tg per year, or 0.65 Tg per year, rounded to 0.7 in Table 2), reflecting that conventional oil contributed 40% to the growth in oil production between 2005 and 2015.

30
For the tight shale oil, we follow the approach used by Schneising et al. (2014): the increase in methane emissions from shale gas and shale oil are considered together, normalized to the energy content of the two fuels. Therefore, the sum of the increase in production for shale gas, shale oil, and conventional natural gas is 33.6 trillion MJ per year. Shale gas represents 48% of this, shale oil 26%, and conventional natural gas represents 26%. The sum of shale gas and shale oil represents 74% of the total. 5 For this sensitivity analysis, we further assume that shale gas and conventional natural gas have the same percentage emissions, as in our base case analysis in the main text, and that the 13 C content of methane from shale oil is the same as for shale gas. Using these assumptions, we modify Eq. (2)  Robert Howarth is the sole author, responsible for all aspects of this work. The author declares that he has no conflicts of interest.

5
Financial support was provided by the Park Foundation and an endowment given by David R. Atkinson to support the professorship at Cornell University held by RWH. We thank Tony Ingraffea, Amy Townsend-Small, Euan Nisbet, Martin Manning, Dennis Swaney, and Roxanne Marino for comments on earlier versions of this manuscript. We particularly thank Dennis Swaney for helpful discussion and review of the analyses we report. We thank Gretchen Halpert for the art work in  Coal Geolog., 120, 24-40, doi:10.1016/j.coal.2013.09.001, 2013 Howarth, R.W., Santoro, R., and Ingraffea, A., Methane and the greenhouse gas footprint of natural gas from shale formations, 10 Climatic Change Letters, 106, 679-690, doi:10.1007Letters, 106, 679-690, doi:10. /s10584-011-0061-5, 2011.     (9) and is also presented in Table 1. Assumptions include equivalent percentage emissions as a function of production for shale gas and conventional natural gas, and no contribution of 13 C-depleted methane from tight shale oil production.
b Same assumptions as for the base analysis, except shale gas emissions are assumed to be 50% greater than those from 35 conventional natural gas, expressed as a percentage of production. c Same assumptions as for the base analysis, except emission of 13 C-depleted methane from shale oil is explicitly considered.