Recent decreases in fossil-fuel emissions of ethane and methane derived from firn air

Methane and ethane are the most abundant hydrocarbons in the atmosphere and they affect both atmospheric chemistry and climate. Both gases are emitted from fossil fuels and biomass burning, whereas methane (CH4) alone has large sources from wetlands, agriculture, landfills and waste water. Here we use measurements in firn (perennial snowpack) air from Greenland and Antarctica to reconstruct the atmospheric variability of ethane (C2H6) during the twentieth century. Ethane levels rose from early in the century until the 1980s, when the trend reversed, with a period of decline over the next 20 years. We find that this variability was primarily driven by changes in ethane emissions from fossil fuels; these emissions peaked in the 1960s and 1970s at 14–16 teragrams per year (1 Tg = 1012 g) and dropped to 8–10 Tg yr−1 by the turn of the century. The reduction in fossil-fuel sources is probably related to changes in light hydrocarbon emissions associated with petroleum production and use. The ethane-based fossil-fuel emission history is strikingly different from bottom-up estimates of methane emissions from fossil-fuel use, and implies that the fossil-fuel source of methane started to decline in the 1980s and probably caused the late twentieth century slow-down in the growth rate of atmospheric methane.

Methane and ethane are the most abundant hydrocarbons in the atmosphere and they affect both atmospheric chemistry and climate. Both gases are emitted from fossil fuels and biomass burning, whereas methane (CH 4 ) alone has large sources from wetlands, agriculture, landfills and waste water. Here we use measurements in firn (perennial snowpack) air from Greenland and Antarctica to reconstruct the atmospheric variability of ethane (C 2 H 6 ) during the twentieth century. Ethane levels rose from early in the century until the 1980s, when the trend reversed, with a period of decline over the next 20 years. We find that this variability was primarily driven by changes in ethane emissions from fossil fuels; these emissions peaked in the 1960s and 1970s at 14-16 teragrams per year (1 Tg 5 10 12 g) and dropped to 8-10 Tg yr 21 by the turn of the century. The reduction in fossil-fuel sources is probably related to changes in light hydrocarbon emissions associated with petroleum production and use. The ethane-based fossil-fuel emission history is strikingly different from bottom-up estimates of methane emissions from fossilfuel use 1,2 , and implies that the fossil-fuel source of methane started to decline in the 1980s and probably caused the late twentieth century slow-down in the growth rate of atmospheric methane 3,4 .
Ethane is an organic trace gas that is primarily emitted to the atmosphere during mining, processing, transport and consumption of fossil fuels, during use of biofuels, and during biomass burning [4][5][6] . It acts as a precursor of ozone and carbon monoxide in the troposphere. Ethane is oxidized rapidly by the hydroxyl radical (OH . ) and has a seasonally varying lifetime (annual mean ,2 months). The short lifetime, coupled with a north-south asymmetry in its sources, leads to geographic and temporal variability in its atmospheric abundance 5,7 . Ethane abundance displays strong seasonal variability, with peak-to-peak amplitudes comparable to annual-mean levels. Annual-mean ethane levels are highest at high northern latitudes (HNL, that is, 30-90u N; 1.5 parts per billion) and lowest at high southern latitudes (HSL, that is, 30-90u S; 250 parts per trillion) 4 .
We measured ethane in firn air collected at Summit (in Greenland), at the West Antarctic Ice Sheet -Divide, WAIS-D (in Antarctica), and at the South Pole. A synthesis inversion method was used to develop atmospheric histories of ethane for each site, using a one-dimensional firn air diffusion model ( Fig. 1) (see Methods Summary). The mean temperatures and the ice accumulation rates are very similar at Summit (231 uC, 20 cm yr 21 ) and at WAIS-D (231 uC, 22 cm yr 21 ), resulting in similar gas age distributions in the firn and atmospheric histories constrained for the past 50-60 years (refs 8, 9). South Pole is considerably colder (251 uC), with a lower ice accumulation rate (,8 cm yr 21 ) and a deeper firn column. As a result, atmospheric histories based on South Pole firn air measurements are constrained for the past 80-90 years (ref. 10). The ethane atmospheric histories based on the WAIS-D and South Pole firn air measurements display the same trends for the period since 1950 (Fig. 1).
Mean annual ethane levels measured over Summit and South Pole are consistent with those at other high-latitude sites, indicating that the  measurements in polar firn air provide a reasonable sampling of the ethane mixing ratios in high latitudes (see Supplementary Information). Together, the three atmospheric histories show ethane levels peaking in the HNL and HSL atmospheres in the early 1980s, followed by a decline over the 20 years that followed. The South Pole site contains sufficiently old air to constrain the ramp-up period during the early to mid twentieth century ( Fig. 1). On the basis of these atmospheric histories, HSL annual-mean ethane levels increased fivefold between 1910 and 1980 (from about 60 p.p.t. to 280 p.p.t.), and declined more than 10% (to less than 250 p.p.t.) by 2000. In the HNL, ethane increased from ,1.7 p.p.b. in 1950 to ,2.0 p.p.b. in 1980, and declined 25% to ,1.5 p.p.b. in 2000.
To infer the causes of the atmospheric ethane variations in terms of changes in large-scale sources, we must relate high-latitude ethane levels to hemispheric averages. We used the modern atmospheric ethane distribution 4,6 and sensitivity tests with the UCI Chemical Transport Model (UCI-CTM) 11,12 to derive ratios relating the response of the HNL and HSL to changes in hemispheric mean ethane levels (see Supplementary Information). Next, we used a simple two-box model (see Methods Summary) representing the troposphere in the Northern and Southern hemispheres to simulate variations in mean atmospheric ethane levels over the past century resulting from various scenarios for fossil-fuel, biomass-burning, and biofuel emissions. Fossil-fuel and biofuel emissions are concentrated in the northern mid-latitudes, and biomass burning in the tropics. As a result, these sources contribute to hemispheric ethane levels with different efficiencies (see Methods Summary and Supplementary Information).
The two-box model was used to infer the fossil-fuel ethane emission histories needed to achieve agreement with the firn-air-derived atmospheric ethane histories from Summit and South Pole, under various assumptions about biomass-burning emissions (Fig. 2, see Supplementary Information). The results show clearly that ethane variability over decadal timescales during the twentieth century was dominated by changes in the fossil-fuel ethane source. When biomass-burning and fossil-fuel ethane emissions are both allowed to vary after 1950, the model yields a biomass-burning source of less than 1 Tg yr 21 in 1950, rising to ,3 Tg yr 21 by 2000, followed by a drop to 2 Tg yr 21 . The model results show a large change in fossil-fuel ethane emissions in the first half of the twentieth century, peaking at 14-16 Tg yr 21 in the late 1950s. Fossil-fuel emissions are then constant for about 20 years, followed by a 45% drop over the next 30 years. The model yields 8-10 Tg yr 21 for fossil-fuel ethane emissions at the end of the twentieth century, consistent with the modern ethane budget inferred from atmospheric ethane levels 6,13 .
Fossil-fuel emissions are also a significant source of methane into the atmosphere, and the observed decline in the growth rate of atmospheric methane 3,4 parallels the decline in atmospheric ethane levels during 1980-2000. This evidence suggests that the decline in methane growth rates was caused by a gradual reduction in fossil-fuel emissions, which was already underway when continuous direct atmospheric measurements of methane started in 1984 and 1985 3,4 . The stabilization of atmospheric ethane in recent years 4 (also see Supplementary Information) suggests that fossil-fuel emissions are now steady. Thus, the recent increases in atmospheric methane levels are probably not derived from increased fossil-fuel emissions and must be due to other sources, as previously suggested 14,15 .
The ethane-based biomass-burning emission history agrees well with some independent estimates 6,16 . However, our fossil-fuel emissions history is quite different from bottom-up methane emission inventories 1,2 (Fig. 3). Most notably, the ethane-based fossil-fuel emissions display a steep ramp-up after 1920 and a sharp decline after 1980, whereas the bottom-up methane fossil-fuel emission inventories display a generally increasing trend through the entire twentieth century, with the sharpest increases occurring between 1950 and 1980. Hydrocarbons are emitted from a variety of fossil-fuel sources, each exhibiting a different range of methane/ethane emission ratio (MER). For example, coal and natural gas emissions are methane-rich (MER 5 5-100) with respect to emission ratios measured near oil-storage and oil-processing facilities (MER 5 3-5) 6 . It is possible that the global-average MER (the ratio of total methane to total ethane emissions from all fossil-fuel sources combined) changed over time as the relative emission strengths of the different fossil-fuel sources varied during the twentieth century,

LETTER RESEARCH
but the apparent differences between the ethane-based fossil-fuel emission histories and inventory-based methane emissions (Fig. 3) cannot be reconciled with such considerations. Inconsistencies are already apparent in mid-century, when the ratio of bottom-up methane emissions to ethane-based fossil-fuel emissions imply an MER of 2-3 (Fig. 3). It is highly unlikely that the global-average MER during the 1950s could be lower than point sources with the lowest MERs, especially considering that both coal and natural gas were sources of significant hydrocarbon emissions during that period. This suggests the bottom-up methane inventories underestimate fossil-fuel emissions at mid-century. The ethane-based estimates of fossil-fuel emissions show a decline during the 1980s, while the methane-inventory-based estimates of fossil-fuel emissions show an increase (Fig. 3). Opposing trends in the fossil-fuel emissions of ethane and methane are highly unlikely, because this requires ethane to be selectively removed from both newly introduced and existing hydrocarbon sources while the residual methane is released to the atmosphere. Natural gas is used in the production of feedstock ethane in plastics manufacturing. However, it seems implausible that methane, which is economically the most valuable component of natural gas, would simply be vented back into the atmosphere after ethane had been removed from natural gas. In fact, the amount of feedstock ethane produced from natural gas is inversely related to the price of natural gas 17 , and probably declined as natural gas became progressively more valuable during the second half of the twentieth century. We conclude that the discrepancy between the ethane-based fossil-fuel emission histories and the fossil-fuel emission histories from the bottom-up methane inventories cannot be explained by realistic changes in global-average MER during the twentieth century.
The decline in fossil-fuel emissions, as calculated from our firn-air ethane records, coincides with a period of rapid expansion in natural gas production during the second half of the twentieth century 18,19 . We speculate that the rising economic value of natural gas during the late twentieth century 19 and the development of cleaner technologies led to sharp reductions in the release of light hydrocarbons into the atmosphere. Emissions linked specifically to the growing natural gas industry must have been more than offset by large reductions in the venting of light hydrocarbons, including methane and ethane, associated with production and processing of petroleum. These changes appear to be underestimated in the bottom-up methane inventories. We estimate that the total decline in fossil-fuel emissions of ethane was 5-6 Tg yr 21 during 1980-2000. Attributing this decline entirely to decreases in fossil-fuel emission sources with the lowest MER (3-5) implies a 15-30 Tg yr 21 drop in fossil-fuel emissions of methane. Distributing the decline among different fossil-fuel sources would yield a larger change in methane emissions from fossil fuels. An independent inversion analysis of methane observations suggests that a 20 Tg yr 21 drop in fossilfuel methane emissions during the 1990s contributed to the decline in methane growth rates 20 . Our results are consistent with such a decline, but provide additional evidence that the drop in fossil-fuel emissions started sooner, by about a decade. The decline in fossil-fuel emissions during the 1990s accounts for less than 60% of the total reduction during 1980-2000 in the two-box model results for ethane, suggesting that the reduction in fossil-fuel methane emissions could be ,40% larger than the previous estimate 20 of 20 Tg yr 21 .
Given that we do not have comparable observations from lower latitudes, it is possible that a shift in the location of fossil-fuel emissions towards lower latitudes within the Northern Hemisphere contributed to the observed ethane decline during 1980-2000. However, we estimate this effect to be relatively small on the basis of sensitivity tests with the UCI-CTM, which show that possible changes in the location of the emission reductions for ethane could add about 20% to the magnitude of emission reductions calculated above (see Supplementary Information).
It is also possible that the late twentieth century decline in ethane could have been caused by a decrease in the atmospheric lifetime of ethane, but we estimate the likelihood for such changes to be low. Assuming fixed emissions, a 5-6 Tg yr 21 increase in ethane loss (roughly one-third of the peak budget) would be required to account for the observed ethane decline. A change of this magnitude due to increases in OH . concentrations during the late twentieth century is unlikely 21,22 ; however, atmospheric levels of chlorine atoms (Cl . ) might have increased, because of increasing tropospheric NO x , increasing tropospheric ozone and increasing acidity of aerosols 23,24 . A Cl . sink  30 . Our estimate of ethane biomass-burning emissions from the variable biomass-burning case is also in good agreement with some independent estimates of historical ethane emissions from biomass burning 16 (green diamonds, left y-axis) and a total burning emissions estimate (biomass burning1biofuels) for present day 6 (red triangle, left y-axis).

RESEARCH LETTER
of the magnitude required to explain the ethane decline would have a minor effect on the methane budget (,3%), because the relative reactivity of Cl . versus OH . (k Cl /k OH ) is considerably smaller for methane than it is for ethane. Accounting for the ethane decline would require an increase of roughly 1.1 3 10 4 Cl . cm 23 over the entire marine boundary layer, or a much larger increase if just the polluted boundary layer is considered. Estimates for the clean marine boundary layer are in the range 10 3 -10 4 Cl . cm 23 . It is not possible to assess whether a change of the required magnitude is viable with the limited observational data available at present 23,25 .

METHODS SUMMARY
Firn-air measurements, modelling and inversion. At all three sites, multiple flasks (see Supplementary Information) from the same depth were filled sequentially, using established methods 10 . All flasks were analysed at the University of California, Irvine, using ,100 cm 3 STP samples on a gas chromatography-mass spectrometry (GC-MS) system designed for trace gas analysis on small air samples 26 . Summit and South Pole flasks were analysed at least twice. We used a onedimensional firn-air model and a synthesis inversion method to derive atmospheric histories of the annual-mean, high-latitude, tropospheric abundances of ethane from the firn-air measurements (see Supplementary Information). Two-box model. The Northern and Southern hemispheres are represented as equal mass boxes of 2.2 3 10 9 Tg (exchange time 1 year). Ethane is lost through OH . oxidation ( 1 =2:6 months 21 ) and transport into the stratosphere ( 1 =35 yr 21 ), yielding a present day lifetime of 2.3 months. The annual-mean OH . loss frequency varies over the past 100 years based on the methane feedback on tropospheric OH . : 20.32% in OH . for every 11% in methane 27 . The OH . changes differently in Northern and Southern hemispheres: 20.305% and 20.335%, respectively, based on simulations with the UCI-CTM 11,12 . We adopt a methane increase of 900 p.p.b. to 1,790 p.p.b. since 1900 28 . Fossil-fuel, biofuel and biomass-burning sources in the model Northern Hemisphere are adopted from previous work 6 and represent 93%, 81% and 58%, respectively, of the total emissions. The fossil-fuel emission histories for various biomass-burning scenarios are developed by an inverse optimization algorithm (see Supplementary Information). Three-dimensional model. The UCI-CTM's simulation of ethane with realistic distribution of sources 29 is used to relate the northern and southern tropospheric mean abundances to the annual-mean abundance over the ice sheets, and thus provide a correction to the two-box model histories for the firn-air modelling.