Global geological methane emissions: An update of top-down and bottom-up estimates

A wide body of literature suggests that geological gas emissions from Earth’s degassing are a major methane (CH4) source to the atmosphere. These emissions are from gas-oil seeps, mud volcanoes, microseepage and submarine seepage in sedimentary (petroleum-bearing) basins, and geothermal and volcanic manifestations. Global bottom-up emission estimates, ranging from 30 to 76 Tg CH4 yr–1, evolved in the last twenty years thanks to the increasing number of flux measurements, and improved knowledge of emission factors and area distribution (activity). Based on recent global grid maps and updated evaluations of mud volcano and microseepage emissions, the global geo-CH4 source is now (bottom-up) estimated to be 45 (27–63) Tg yr–1, i.e., ~8% of total CH4 sources. Top-down verifications, based on independent approaches (including ethane and isotopic observations) from different authors, are consistent with the range of the bottom-up estimate. However, a recent top-down study, based on radiocarbon analyses in polar ice cores, suggests that geological, fossil (14C-free) CH4 emissions about 11,600 years ago were much lower (<15 Tg yr–1, 95% CI) and that this source strength could also be valid today. Here, we show that (i) this geo-CH4 downward revision implies a fossil fuel industry CH4 upward revision of at least 24–35%. (ii) The 95% CI estimates of the recent radiocarbon analysis do not overlap with those of 5 out of 6 other bottom-up and top-down studies (no overlap for the 90% CI estimates). (iii) The contrasting lines of evidence require further discussion, and research opportunities exist to help explain this gap.

The global CH 4 emissions from these sources have been mainly estimated through bottom-up procedures by various authors (see Table 1), based on process-based modelling, statistical evaluations of experimentally determined emission factors and activity data (number of emission points or emission area) and inventories. Global geo-CH 4 emission estimates (including all five geo-CH 4 categories) range from 30 to 76 Tg yr -1 , with a typical mean around 50 Tg yr -1 ( Table 1). As shown in detail below, top-down emission estimates, based on present-day atmospheric data of isotopic ( 14 C and 13 C/ 12 C) CH 4 composition or ethane (C 2 H 6 ) emissions (also derived from polar ice cores), are consistent with the order of magnitude of the bottom-up estimates (Etiope et al., 2008;Schwietzke et al., 2016;Nicewonger et al., 2016;Dalsøren et al., 2018). Differently, Petrenko et al. (2017) proposed a substantially lower global estimate, ranging from 0 (zero) to 15.4 Tg yr -1 (95% CI). This range was derived from radiocarbon ( 14 C) measurements in CH 4 trapped in ice cores in Antarctica and referring to the atmosphere of 11,000-12,000 years ago, between the Younger Dryas and Preboreal intervals. Assuming that geological emissions today are not higher (or even lower) than in the analysed period, Petrenko et al. (2017) concluded that previous present-day geo-CH 4 estimates of ~5 0 Tg yr -1 are overestimated. As stated in Petrenko et al. (2017), this geo-CH 4 downward revision also implies an upward revision of the present-day fossil fuel industry CH 4 emission estimates.
The objectives of this paper are: (1) Combine and reasses the most recent global bottom-up geo-CH 4 emission estimates based on recently published global grid maps and updated inventories discussed in Etiope et al. (2019).
(2) Summarize top-down emission estimates by combining data from Schwietzke et al. (2016), Nicewonger et al. (2016, Saunois et al. (2016) and Dalsøren et al. (2018), followed by a calculation of the average ethane/methane (C 2 /C 1 ) ratio for the five geological sources (having specific C 2 /C 1 ratios, as reported in Etiope and Ciccioli, 2009). Based on this overview of geo-CH 4 top-down emission estimates using multiple species, datasets, and methods, the discrepancy between these estimates and the range proposed by Petrenko et al. (2017) is addressed. (3) Discuss and quantify in more detail the fossil fuel industry CH 4 upward revision proposed by Petrenko et al. (2017).
(4) Offer suggestions for further research activities to reconcile the above discrepancies, and to better constrain geological emission estimates.

Bottom-up geo-CH 4 emission estimates
The objective of the recently published globally gridded dataset of geo-CH 4 emissions (Etiope et al., 2019) was to develop the first comprehensive a priori emission grid for atmospheric modelling. The gridding work allowed refining the CH 4 emission estimates for mud volcanoes and microseepage, thanks to a better assessment of their activity (global area) and emission factors (Etiope et al., 2019). These new estimates are reported in Table 2. However, these grid maps do not represent the entire global geo-CH 4 source because for some categories of geo-CH 4 sources, namely onshore gas-oil seeps, submarine seepage and geothermal emissions, the datasets used for the spatial gridding (developed for modelling purposes) are incomplete or do not contain the information necessary for improving all previous estimates.
The following provides a brief summary of the gridded geo-CH 4 estimates (Etiope et al., 2019) as well as the other previous global geo-CH 4 total estimates from the literature ( Table 1). Note that the approaches used to quantify emission uncertainties vary among the different bottomup studies given the different spatial scales (global total vs. grid-level uncertainties) as described below. In most cases, however, best estimates of lower and upper bounds were reported, and these ranges are summarized here.

Updates based on gridded geo-CH 4 estimates
The CH 4 emission range of mud volcanoes, 3.9-8.3 (mean 6.1) Tg yr -1 , combines uncertainties of the mud volcano areas and emission factors (Etiope et al., 2019). The mud volcano areas were estimated using image (Google Earth) analysis, photos and published literature (uncertainty of 6%) and the emission factors were based on regression analysis between area and seepage flux for 16 mud volcanoes measured in Europe and Asia (uncertainty of about 42%; Etiope et al., 2019).
The emission range of microseepage, 15-33 (mean 24) Tg yr -1 , reflects an uncertainty of about 38% estimated through analysis of sensitivity of the microseepage model (based on geospatial and statistical analyses) used to derive the global microseepage emission (Etiope et al., 2019). Briefly, the microseepage emission factors are based on a statistical analysis of a dataset of 1509 flux measurements (acquired by accumulation chamber method) from 19   (2015) Total~4 5~27~63 different petroliferous basins, associated with geological factors (macro-seeps, faults, seismicity) that can influence the microseepage intensity (Etiope et al., 2019); the microseepage area (activity) was estimated using the global area of petroleum fields and macro-seeps (which may occur also outside a petroleum field) and knowing, from the global microseepage dataset, that microseepage (positive CH 4 fluxes) occurs in about 57% of the petroleum field area (Etiope et al., 2019). The sensitivity of the model was tested by combining different emission factors (median, geometric mean, upper and lower 95% confidence limit) and activity (microseepage area 20% smaller or higher).
Concerning the global submarine emissions, which only refers to waters shallower than 500 m where emitted methane can reach the atmosphere, a new range is suggested combining data from Etiope et al. (2019) and previous estimates (Kvenvolden et al. 2001;Judd, 2004). Etiope et al. (2019) report a partial dataset (15 areas) of local and regional emission estimates, totally resulting in 1.8-6.0 (mean 3.9) Tg yr -1 . They estimate that other 16 areas may release an additional ~1 Tg yr -1 . In these areas, gas was observed to reach the sea surface via bubble plumes, but the output to the atmosphere was not provided. This yields a global submarine emission range of ~3 -7 Tg yr -1 . Previous estimates are those reported by Kvenvolden et al. (2001), where a range of 10-30 Tg yr -1 was proposed (see also Judd, 2004). The partial emission dataset reported in Etiope et al. (2019) includes major submarine seepage areas investigated so far, but may be a conservative (low) estimate. Nevertheless, it suggests that global submarine emissions may not exceed the minimum value of 10 Tg yr -1 by Kvenvolden et al. (2001). Therefore, until further data emerge, the range 3-10 Tg yr -1 is proposed here as a best guess.

Literature estimates for other geo-CH 4 categories
Gridding work in Etiope et al. (2019) did not result in new estimates from onshore gas-oil seeps and geothermal manifestations. The most detailed global total emission estimates from these two geo-CH 4 categories are still those from previous statistical and process-based modeling ( Etiope et al., 2008a;Etiope, 2015; Table 2).
Global CH 4 emission estimates from gas-oil seeps, 3-4 (mean 3.5) Tg yr -1 , were based on a database of fluxes that were measured directly (typically by accumulation chamber method) from 66 gas seeps in 12 countries, assuming that their flux and size distributions were representative of the global gas-oil seep population (Etiope et al., 2008a). The global emission range reflects an uncertainty of about 15% estimated combining two different extrapolations of emission factors over the global number of seeps (Etiope et al., 2008a). Global CH 4 emission estimates from geothermal manifestations, 2.2-7.3 (mean 4.7) Tg yr -1 , were derived on the basis of the most updated estimates of global CO 2 emissions from volcanic areas (540 Tg yr -1 ), from non-volcanic areas (300-1,000 Tg yr -1 ), and a wide dataset on CO 2 /CH 4 compositional ratios in both areas (Etiope 2015 and references therein). The emission range reflects an uncertainty of about 53% derived from the uncertainty of non-volcanic CO 2 emissions (Etiope 2015; the uncertainty of volcanic CO 2 degassing was not quantified in the original work; Burton et al., 2013).
In summary, considering the updates based on gridded geo-CH 4 estimates and literature estimates for the other geo-CH 4 categories, the global bottom-up geological CH 4 emission is now estimated at ~4 5 (27-63) Tg yr -1 ( Table 2). Note that this estimate uses the best individual estimates presently available for the five geo-CH 4 source categories, while the global estimate proposed in Etiope et al. (2019) refers to the global emissions "extrapolated" from gridded maps ( Table 1). As explained above, the extrapolation from gridded maps contains incomplete information on oil-gas seeps, submarine and geothermal emissions. Although derived through different approaches the two estimates are however similar.

Top-down geo-CH 4 emission estimates
Top-down geo-CH 4 emission estimates can be derived via multiple approaches, based on the present-day fraction of radiocarbon ( 14 C) free CH 4 in the atmosphere (Lassey et al., 2007), pre-industrial ethane (C 2 H 6 ) and CH 4 isotopic composition in polar ice cores and box modelling ( Nicewonger et al., 2016;Schwietzke et al., 2016;Dalsøren et al., 2018). Similar to the bottom-up approaches described above, the way the reporting of emission uncertainties varies among studies given the different methodologies (e.g., atmospheric forward modelling of set emission scenarios vs. inverse or box modelling with explicit posterior uncertainties). When only lower and upper bounds were reported, then these ranges are summarized here. Lassey et al. (2007) deduced that 30% ± 2.3% (1 SD) of the global CH 4 source for 1986-2000 is 14 C-free (and thus fossil), although they considered this "a plausible re-estimate rather than a definitive revision" of previous lower fossil CH 4 estimates (on average 20%). Taking into account the average top-down total CH 4 source of 560 Tg yr -1 (reported by Saunois et al., 2016 and valid for the period 1986-2000, as indicated by Lassey et al., 2007) yields 168 Tg yr -1 total fossil CH 4 emissions, i.e., natural (geological) plus anthropogenic (fossil fuel industry sources including CH 4 venting and leaks). Using the range of the fossil fuel industry fraction reported in the average bottom-up and top-down estimates by Saunois et al. (2016), i.e., 101-134 Tg yr -1 , yields geological emission range of 34-67 (mean 50.5) Tg yr -1 .

Pre-industrial 13 C/ 12 C approach
The box modelling by Schwietzke et al. (2016), based on CH 4 concentration and isotopic data from ice-core records suggest geo-CH 4 emissions of 31-71 Tg yr -1 (1 SD, mean 51 Tg yr -1 ). These estimates consider a new database of present-day 13 C/ 12 C signatures for all CH 4 sources, and assume that these signatures are also representative of the preindustrial era. The relatively wide geo-CH 4 range is partly due to a wide range of prescribed biomass burning CH 4 emissions from present-day estimates. The central value, 51 Tg yr -1 , is consistent with the above estimate combining Lassey et al. (2007) and Saunois et al. (2016) data. Note that this wide geo-CH 4 uncertainty range is largely due to uncertainties in pre-industrial biomass burning CH 4 baseline emission estimates (i.e., long-term averages), but it is unrelated to current-day uncertainties the trend of burning CH 4 emissions (Worden et al., 2017).

Present day 13 C/ 12 C and inventory approach
With the same box model plus 3-D forward modelling, but using present-day atmospheric CH 4 and isotopic data, Schwietzke et al. (2016) suggested a total fossil (geological plus fossil fuel industry) CH 4 source of 150-200 Tg yr -1 . Considering the fossil fuel industry emission estimates of Saunois et al. (2016) above, i.e., 101-134 Tg yr -1 , requires geo-CH 4 emissions of 16-99 Tg yr -1 (minimum-maximum range). Note that the global total CH 4 source budget in Saunois et al. (2016) is ~5 % less than in Schwietzke et al. (2016) due to slightly different assumptions in the CH 4 sink magnitude. As a result, the above geo-CH 4 emission estimate of 16-99 Tg yr -1 may be a slight underestimate.

Pre-industrial C 2 H 6 approach
Based on pre-industrial concentrations of C 2 H 6 in polar ice cores, Nicewonger et al. (2016) estimated total geological C 2 H 6 emissions of 2.2-3.5 Tg yr -1 , which is consistent with previous estimates of 2-4 Tg yr -1 proposed by Etiope and Ciccioli (2009). Additional ethane ice core data and modelling suggests that geo-C 2 H 6 emissions could be even higher (5-6 Tg yr -1 ; Nicewonger et al., 2018). Considering here the conservative range of Nicewonger et al. (2016), we estimate the total geo-CH 4 emission estimate using global averages of ethane/methane (C 2 /C 1 ) ratios from geological sources (reported in Etiope and Ciccioli, 2009). An average, emission-weighted geological C 2 /C 1 ratio can be derived taking into account the emission fraction of the five geological sources identified in Table 2. A C 2 H 6 source of 2.2-3.5 Tg yr -1 (Nicewonger et al., 2016) requires a CH 4 source in the range of 29-46 (mean 37.5) Tg yr -1 (Table 3).

Present day C 2 H 6 approach
Based on C 2 H 6 observations and simulations with a detailed atmospheric-chemistry transport model, Dalsøren et al. (2018) estimated total geological C 2 H 6 emissions of 2-4 Tg yr -1 , similar to Nicewonger et al. (2016) and Etiope and Ciccioli (2009). In this case, using the same C 2 /C 1 ratio estimated above, yields global CH 4 emissions in the range of 27-52 (mean 39.5) Tg yr -1 ( Table 3).

Summary of all approaches above
All geo-CH 4 emission estimates are then summarized in Figure 1, which categorizes studies into Petrenko et al. (2017), bottom-up, top-down C 2 H 6 -based, and the other top-down studies discussed above. For Petrenko et al. (2017), the 95% confidence interval and range of mean values in Figure 1 are as reported in that study. For bottom-up, we used the minimum and maximum values from Table 2 as described in Section 2. We then assumed a uniform distribution around the minimum and maximum values to calculate 95% confidence intervals and the standard deviation. Note that assuming a uniform distribution is conservative in the sense that the 95% confidence intervals extend closer to the underlying minimum and maximum values compared to, e.g., Gaussian or triangular distributions. In other words, the lack of overlap in 95% confidence intervals between Petrenko et al. (2017) and the bottom-up approach would be even more pronounced if instead a Gaussian or triangular distribution was assumed.
Top-down C 2 H 6 -based 95% confidence intervals were calculated from the joint probability distribution of Nicewonger et al. (2016) and Dalsøren et al. (2018) by performing a Monte Carlo simulation (N = 10,000) giving equal weight to each study, and also assuming a uniform distribution using each study's minimum and maximum values. The colored dots show the 95% confidence intervals of Nicewonger et al. (2016) and Dalsøren et al. (2018) individually. The same approach was used for the top-down "other" category representing Lassey et al. (2007), Saunois et al. (2016), and Schwietzke et al. (2016) as described above. Note that the 95% confidence intervals of Petrenko et al. (2017) only overlap with those of Schwietzke et al. (2016), not with those of the other studies or the three categories. The overlap in the 95% confidence intervals represents the intersection of the extreme ends of both distributions (90% confidence intervals do not overlap; the range is 18-83 Tg yr -1 in Schwietzke et al., 2016). Further, the right side of the 95% confidence interval in Petrenko et al. (2017) is a factor of 3 to 6 lower than all other studies. For illustration, averaging the mean values of the bottom-up and top-down estimates and using their full range of 95% confidence interval uncertainties in Figure 1 (excluding those in Petrenko et al., 2017) results in a global geo-CH 4 emission range of ~2 8-75 (mean 45) Tg yr -1 . In Figure 2, this value is compared with other natural and anthropogenic CH 4 sources. Only combining the lowest estimates from different authors, the global geo-CH 4 emission would be 18 Tg yr -1 (see Table 6 in Etiope et al., 2019) and compatible with the upper limit of Petrenko et al. (2017).

Implications of a geo-CH 4 downward revision on fossil fuel industry CH 4 estimates
Present-day total fossil CH 4 emissions have been estimated top-down using 13 C/ 12 C and 14 C data as described above, but these approaches alone do not attribute emissions between the geo-CH 4 source and the fossil fuel industry CH 4 source. Separate geo-CH 4 estimates (bottom-up and top-down, Sections 2 and 3) have been used to then infer the fossil fuel industry CH 4 source. As noted in Petrenko et al. (2017), a geo-CH 4 downward revision thus necessitates a fossil fuel industry CH 4 upward revision to satisfy the present-day 13 C/ 12 C and 14 C constraints. Consider a geo-CH 4 source of 46 Tg yr -1 (mean of bottom-up and top-down estimates, Section 3), and a total fossil fuel CH 4 source of 172 Tg yr -1 (mean of 13 C/ 12 C and 14 C estimates, Section 3). This implies a fossil fuel industry CH 4 source of 126 Tg yr -1 (consistent with the upper end of recently revised bottom-up estimates of Saunois et al., 2016;Section 3). A geo-CH 4 downward revision to 0-15 Tg yr -1 (estimate by Petrenko et al., 2017) thus requires a fossil fuel industry CH 4 upward revision to 157-172 Tg yr -1 , which is 24-35% larger than previous estimates. Note that the percentage range would increase if lower fossil fuel industry CH 4 emission estimates were used as a baseline, e.g., the lower bottom-up range in Saunois et al. (2016).

Temporal variation of geological methane sources
While all bottom-up and top-down estimates, following independent techniques from different authors, have similar ranges, suggesting a global geo-CH 4 emission source in the order of 40-50 Tg yr -1 , the radiocarbon ( 14 C-CH 4 ) data in ice cores reported by Petrenko et al. (2017) revise these estimates downward, with a range of 0 (zero) to 18.1 Tg yr -1 (<15.4 Tg yr -1 , 95% CI) at least for the atmosphere between 11,000 and 12,000 years ago (Younger-Dryas Preboreal transition). Petrenko et al. (2017) assumed that those past geological emissions are not lower than today, claiming therefore that the previous present-day geo-CH 4 estimates are too high.
The assumption that geological emissions are constant over the Holocene is not necessarily correct. Earth's degassing, which is a process mainly driven by gas advection, is basically controlled by gas pressure gradients in the subsurface and permeability of fractured rocks. These factors can vary considerably on short time scales, in relation, for example, to cycles of gas pressure discharges and buildup in reservoirs, seismic activity, mud volcano eruptions, hydraulic pressure of aquifers and neotectonic stresses (e.g., Quigley et al., 1999;Yang et al., 2006;Delisle et al., 2010;Etiope, 2015). Episodes of enhanced mud volcanism, for example, were recognised in the Upper Quaternary (Etiope et al., 2008b). Modern formation of new mud volcanoes and seeps is documented in several countries (Etiope, 2015). Combinations of such temporal variations at regional scale may lead to a significant variation of the global emission.
Similarly, it is not clear whether during the Younger-Dryas Preboreal transition a wider cover of ice and permafrost may have "capped" and lowered a significant portion of seepage in the boreal hemisphere (as a potential explanation for the discrepancy in geo-CH 4 estimates).

How geological CH 4 emission estimates can be better constrained
We identify two major lines of research that can better constrain the geo-CH 4 emission estimates: (a) CH 4 flux derivation of large active seeps and mud volcanoes based on atmospheric in-situ measurements or remote sensing; (b) improving the definition of global microseepage area.
(a) The existence of mud volcanoes, which alone are estimated to emit ~6 Tg yr -1 ( Table 2), provides a unique opportunity for direct measurements that could test the hypothesis of near-zero global geo-CH 4 emissions. Twentyfive of the world's largest mud volcanoes are located in a relatively small region surrounding the Caspian Sea, and they emit an estimated 1.5 Tg yr -1 in aggregate (Etiope et al., 2019). The individual mud volcanoes can be considered point sources with spatial dimensions comparable to oil and gas production and processing facilities. Their emission estimates can thus be empirically verified using "fence-line" downwind measurement methodologies employed during oil and gas methane field measurements over the last decade in the US and internationally (Alvarez et al., 2018). Similarly, the advancement of spacebased remote sensing instruments could lead to the flux quantification of the entire region described above. This, however, also requires source attribution of geo-CH 4 and other CH 4 sources. The 1.5 Tg yr -1 represents only a relatively small fraction of the bottom-up estimated global geo-CH 4 source. However, verifying emission estimates from these mud volcanoes represents a check on the overall bottom-up method that is applied similarly for other seepage categories, and simultaneously a check on the near-zero emission hypothesis of the radiocarbon-based geo-CH 4 downward revision.
(b) Based on statistical treatment of emission factor and activity (area) data, microseepage is considered the largest geo-CH 4 source (10-25 Tg yr -1 ; Etiope and Klusman, 2010; 24 ± 9 Tg yr -1 ; Etiope et al., 2019). Its main uncertainty is due to the limited knowledge of the actual global area where microseepage occurs. Ground-based measurements (soil-gas and flux data) and remote sensing surveys (multispectral imagery) demonstrated that every petroleum field investigated so far is characterized by areas of microseepage, especially at the boundary of the field (Klusman et al., 1998;Etiope and Klusman, 2010;Asadzadeh and de Souza Filho, 2017 and references therein). Flux measurements (1509 data from 19 petroleum fields) showed that microseepage occurs in at least half of the petroleum field area (Etiope et al., 2019). A review of remote sensing investigations, preferably with new data, could allow to better