Recent work has led to a growing recognition of the role of inland waters in the global carbon cycle, particularly as hotspots for production of carbon dioxide (CO₂) and/or methane (CH₄) (Cole et al. 2007; Tranvik et al. 2009; Bastviken et al. 2011). Of particular concern is CH₄ production in aquatic ecosystems, as CH₄ is a greenhouse gas about 34 times more powerful than CO₂ on a 100-yr time scale (Myhre et al. 2013). Thawing of permafrost in Arctic and boreal regions is implicated in an increase in CH₄ emissions from lakes (Walter et al. 2006; Wik et al. 2013). Impoundment of rivers into flood control, hydropower, or drinking water reservoirs can increase the proportion of C that is released as CH₄ vs. CO₂ due to decreased O₂ availability in lakes vs. streams (St. Louis et al. 2000; Bastviken et al. 2004, 2008; Barros et al. 2011). Recent work indicates that CH₄ emissions may be enhanced in eutrophic reservoirs such as those found in mid-latitude agricultural regions of the United States, likely due to increased autochthonous production fed by fertilizer runoff followed by stratification and hypolimnetic hypoxia (Jacinthe et al. 2012; Beaulieu et al. 2014). Nearly all of the previous studies of CH₄ dynamics in lakes have been in small lakes and reservoirs, although Lakes Tanganyika and Kivu in the African Rift Valley have high CH₄ production in anoxic waters and sediments at depth but are not a significant source of atmospheric CH₄ (Durich-Kaiser et al. 2011; Zigah et al. 2015). However, large lakes that are shallow, eutrophic, and seasonally stratified may be a positive source of atmospheric CH₄, including Lake Erie, one of the Laurentian Great Lakes.

The carbon cycle in the North American Great Lakes is largely uncharacterized (Alin and Johnson 2007). Lake Erie is the shallowest of the Great Lakes, with seasonal algae blooms triggered by nutrient runoff followed by stratification and hypoxia in sediments and the hypolimnion. Investigations of the CO₂ balance of the Great Lakes are still preliminary, but indicate that Lake Erie may be a small sink for CO₂ via burial of terrestrial or aquatic organic matter (McKinley et al. 2011). Earlier work demonstrated that sediments in Lake Erie were a source of CH₄ (Howard et al. 1971; as reported in Weaver and Dugan 1972 and Bastviken et al. 2004), and confirmed the presence of methanogenic bacteria in the lake (Frea et al. 1977; Ward and Frea 1980). The highest CH₄ concentrations in surface waters were found in the postturnover season, indicating that CH₄ was produced in the hypolimnion and released to the atmosphere during fall turnover. Another group measured CH₄ efflux from sediments in the central basin of Lake Erie, and showed that CH₄ production and efflux can account for up to 30% of sediment oxygen demand in the lake (Adams et al. 1982). Despite these previous efforts, CH₄ emissions from Lake Erie are not well characterized, even as algal blooms have increased in recent years with increasing surface runoff linked to climate warming (Michalak et al. 2013). If increasing algal blooms lead to greater organic carbon loading to bottom waters and sediments, this may increase CH₄ evasion from Lake Erie in a positive feedback to climate warming. Increasing hypoxia in Lake Erie due to changes in meteorological patterns or increasing temperatures may also lead to a positive feedback if CH₄ emissions increase (O'Reilly et al. 2015; Zhou et al. 2015).

Extraction and distribution of natural gas is one of the largest anthropogenic CH₄ sources globally and nationally (Myhre et al. 2013; United States Environmental Protection Agency [USEPA] 2014). CH₄ is emitted throughout the natural gas supply chain, including at wells and in pipelines (USEPA 2014; Lamb et al. 2015). These emissions undermine the ability of natural gas to lower total greenhouse gas emissions when used to replace coal or oil for power generation (Howarth et al. 2011; Howarth 2014), and recent studies indicate that emissions of CH₄ from natural gas may be underestimated regionally and globally (Townsend-Small et al. 2012; Miller et al. 2013; Brandt et al. 2014). A recent synthesis also indicated that global CH₄ emissions from both fossil fuel extraction and wetlands/lakes might be on the rise (Kirschke et al. 2013).

There are two stable isotopes that can help illuminate CH₄ sources: carbon-13 (¹³C) and hydrogen-2 (²H). Ratios of these rare isotopes to the more common isotope are generally expressed in delta (δ) notation. Fossil fuel CH₄ produced by high pressure and temperature alteration of organic matter retains isotopic signatures similar to the original plant material, with δ¹³C and δ²Η values ranging from −50‰ to −30‰ and −300‰ to −150‰, respectively. Conversely, isotopic discrimination during anaerobic organic matter degradation leads to depleted isotope ratios in biogenic CH₄ relative to the substrate, with δ¹³C and δ²Η values from −63‰ to −53‰ and −350‰ to −250‰, respectively (Whiticar 1999; Townsend-Small et al. 2012, 2015), and δ¹³C of CH₄ derived from biogenic carbonate reduction can be even more depleted (Smith and Pallasser 1996; Kirk et al. 2015). Microbial oxidation of CH₄ in oxygenated waters results in enrichment of both ¹³C and ²Η in remaining CH₄ (Whiticar 1999).

Here, we present a preliminary dataset of CH₄ emission rates and isotopic composition from Lake Erie in pursuit of the following research questions: (1), do hypoxia and CH₄ generation in sediments and the water column in Lake Erie lead to efflux of CH₄ to the atmosphere, and (2), is offshore natural gas drilling an additional source of atmospheric CH₄ from the lake?

Materials and methods

Study area

Lake Erie is the smallest (by volume) and shallowest (average depth ∼19 m) of the Laurentian Great Lakes, and is divided by the international border between the United States and Canada. Lake Erie is both urbanized and agricultural, with an intensively cultivated watershed in both countries as well as 17 metropolitan areas (Waples et al. 2008). Phosphorus loading as a result of these activities results in eutrophication including seasonal algal blooms followed by hypoxia. Reductions in phosphorus loading since signing of the international Great Lakes Water Quality Agreement in 1972 helped mitigate these problems, but algal blooms have increased in recent years, perhaps due to increased temperatures and surface water runoff (Michalak et al. 2013), and hypoxia persists in bottom waters and sediments throughout the lake in summer (Smith and Matisoff 2008; Scavia et al. 2014). In 2014, a large cyanobacteria bloom in Lake Erie led to a drinking water shutdown in the city of Toledo, Ohio (Otten and Paerl 2015).

There are 2104 offshore natural gas wells in Lake Erie, all in Canadian waters (Fig. 1). These are conventional gas wells drilled mostly since the 1970s into Silurian or Cambrian targets, including the Clinton Sandstone (Ontario Oil, Gas, and Salt Resources Library [http://www.ogsrlibrary.com]). Oil extraction is currently not permitted in Canadian waters, and both oil and gas drilling are banned in United States waters of the Great Lakes. Of these wells, 1543 (73%) are plugged and abandoned, and 441 are actively producing gas. Other categories include lost (34 wells total), suspended (68), and unknown (10). Underwater gathering pipelines link wells to natural gas processing and transmission systems in onshore Ontario.

Results

Lake Erie was a positive flux of CH₄ from nearly all surface water sites visited in both years (Fig. 2; Table 1). The average (± SE) flux from both years and both United States and Canadian waters was 8.0 ± 2.5 mg C m⁻² d⁻¹. Fluxes ranged from 0 47.3 mg C m⁻² d⁻¹ to 47.3 mg C m⁻² d⁻¹, and fluxes were not statistically different between the 2 yr (10.7 ± 14.7 mg C m⁻² d⁻¹ in 2012 vs. 3.5 ± 1.6 mg C m⁻² d⁻¹ in 2013) (Table 1), although we did not necessarily expect fluxes to be identical between sampling years. At sites where fluxes were measured using both methods and during both years, fluxes in 2012 were on average 10.7 times higher than in 2013 (Table 1). Some previous studies have shown that the flux chamber method may overestimate gas exchange vs. wind-speed based estimates (St. Louis et al. 2000; Vachon et al. 2010), whereas other studies have shown that flux chamber, wind-based modeling approaches, and eddy covariance methods are consistent in low wind conditions (< 3 m s⁻¹) (Guérin et al. 2007; Cole et al. 2010; Eugster et al. 2011; Gålfalk et al. 2013). At least one previous study has shown that lake size may have a significant impact on gas transfer velocity, particularly at wind speeds higher than 5 m s⁻¹, indicating that gas emissions from large lakes may have a different relationship with wind speed than smaller lakes (Vachon and Prairie 2013). However, the wind speed model is more easily applied than flux chambers in Lake Erie, because consistently high winds (average wind speed at the NOAA Cleveland Central Buoy in September 2015 = 6.2 m s⁻¹) leading to whitecap conditions that cause chambers to be vented or flipped over. In cases in our 2012 field season where the chamber was destabilized, either a smaller number of time points were used to calculate the flux (i.e., the time points before the chamber was vented), or the flux was considered not measured.

Dissolved CH₄ concentrations measured in 2013 are shown in Table 2, and two depth profiles taken in 2013 are shown in Fig. 3. In most sampling sites with depths greater than ∼ 11 m, the water column was stratified with depleted O₂ in bottom waters (Table 2; Fig. 3A). Shallower sites were well-mixed with near-atmospheric levels of O₂ in bottom waters (Table 2). An exception is site 452, which is in the deepest part of Lake Erie, and where O₂ is depleted to ∼ 50% of atmospheric levels at the thermocline, but recovers to 10 mg L⁻¹ in the hypolimnion (Fig. 3B). In well-mixed sites, CH₄ concentrations are similar at the surface and in bottom waters, but are supersatured with CH₄ throughout (Table 2). Previous studies in small stratified lakes have shown that CH₄ concentrations are higher in bottom waters than near the surface, consistent with hypolimnetic accumulation and diffusion upwards through the thermocline, where CH₄ oxidation and ventilation occurs in the surface well-mixed layer (i.e., Beaulieu et al. 2014). However, in a large lake such as Lake Erie, vertical and horizontal mixing may also exerts significant control over water column profiles of CH₄, so hydrodynamic modeling may be needed to distinguish between vertical diffusion and lateral advection of dissolved gases.

Stable isotopic composition of CH₄ samples from bottom water reflects the contribution of both biogenic and thermogenic processes as well as microbial CH₄ oxidation to CH₄ present in Lake Erie (Fig. 4). Samples in Fig. 4 were taken from the sediment-water interface in winter, spring, summer and fall in 2012–2014. Most samples taken in areas without natural gas infrastructure had δ¹³C and δ²Η values in the range of previous measurements of CH₄ derived from anaerobic organic matter respiration (Townsend-Small et al. 2012, 2015), with some samples exhibiting very enriched isotopic compositions consistent with CH₄ oxidation (Whiticar 1999) (Fig. 4). Some samples taken in Canada above natural gas wells (Fig. 1) have isotopic signatures consistent with biogenic CH₄, while others are similar to thermogenic natural gas (Fig. 4). Many of the samples from natural gas extraction areas also have enriched isotopic signatures affected by oxidation (Fig. 4).

Figure 5 further shows the effects of CH₄ source, vertical mixing, and oxidation on CH₄ concentrations and isotopic composition. At both sites, CH₄ concentrations are highest in the hypolimnion and decline closer to the surface (see also Fig. 3). At site 15, the isotopic composition of CH₄ in the epilimnion reflects the mixing of CH₄ diffusing from below the thermocline, CH₄ oxidation which enriches ¹³C, and mixing of CH₄ produced in situ with atmospheric CH₄ (∼ 47‰; Townsend-Small et al. 2012, 2015). If CH₄ in the hypolimnion at site 15 is produced by anaerobic organic matter respiration it has an enriched isotopic composition consistent with production in sediments and oxidation in the water column (Fig. 5A). Conversely, CH₄ in hypolimnetic waters at site 452, in the Canadian offshore natural gas field, reflects a clear natural gas source (see Fig. 4) while surface waters have a depleted isotopic signature consistent with biogenic CH₄ transported laterally (Fig. 5B).

Discussion

Here, we present data demonstrating that Lake Erie is a positive source of atmospheric CH₄ in late summer (8.0 ± 2.5 mg C m⁻² d⁻¹). Our stable isotopic measurements indicate that most CH₄ in the lake is generated through anaerobic organic matter respiration, but that fugitive natural gas from offshore wells also contributes to CH₄ emissions. Most sites we visited in late summer 2012 and 2013 were a positive source of atmospheric CH₄ (Fig. 2). Higher fluxes were observed in natural gas extraction regions (p = 0.09), with an average (± standard error) flux from natural gas fields of 12.4 ± 4.8 mg C m⁻² d⁻¹ vs. 4.0 ± 1.1 mg C m⁻² d⁻¹ from areas without natural gas infrastructure (Table 1). This trend is driven by the relatively high CH₄ fluxes measured in 2012 via flux chambers: in 2013, there was no significant difference between fluxes in gas fields vs. areas without natural gas extraction (Table 1). Isotopic measurements of CH₄ taken from bottom waters from throughout the lake indicate that some natural gas well and/or pipelines contribute thermogenic CH₄ to the lake, while other sampling sites in natural gas extraction areas have largely biogenic CH₄ (Fig. 4). We also see some evidence for a small amount of thermogenic CH₄ present at sites without natural gas infrastructure, perhaps due to natural seepage (e.g., Quigley et al. 1999).

About 40% of Lake Erie has active or abandoned natural gas wells on the lake floor (Fig. 1). Of the 35 sites where we sampled bottom water above natural gas wells, 10 had an isotopic composition indicative of a predominant natural gas source (Fig. 4), so we assumed that 30% of the area of Lake Erie overlying natural gas wells contributed thermogenic CH₄ to the lake: in other words, 12% of the lake area was a source of thermogenic CH₄. Using a flux rate of 12.4 ± 4.8 mg C m⁻² d⁻¹ for the 12% of the lake contributing thermogenic CH₄, and a flux rate of 4.0 ± 1.1 mg C m⁻² d⁻¹ for the rest of the lake, we estimate that fugitive natural gas is the source of about 30% of the total daily CH₄ emissions from Lake Erie, 1.3 × 10⁵ kg C d⁻¹ in late summer during stratified conditions. These results imply that the lake may be a source of CH₄ year-round, as fugitive emissions from natural gas wells and pipelines are not likely to be dependent on temperature and oxygen availability, as is biogenic CH₄ production. Active natural gas wells are a well-known source of atmospheric CH₄ (USEPA 2014), and about 29% of the gas wells in Lake Erie are active. The underwater pipelines connecting these wells to onshore gathering compressor stations are also a likely source of thermogenic CH₄ in Lake Erie (Lamb et al. 2015). The majority of wells in the lake are abandoned, and these can also be a source of CH₄, although they likely emit less CH₄ than active wells (Etiope et al. 2013; Kang et al. 2014), particularly if they are properly plugged and closed.

Our results (8.0 ± 2.5 mg C m⁻² d⁻¹) fall within the range of CH₄ emissions observed in hydroelectric reservoirs around the world, which range from ∼0 to 50 mg C m⁻² d⁻¹ at similar latitudes to Lake Erie (Barros et al. 2011). The diffusive CH₄ flux from Lake Erie in 1971 was estimated to be 0.5 mg C m⁻² d⁻¹, or about 15 times lower than our estimated flux (Howard et al. 1971). CH₄ fluxes from Lake Erie were lower than observed in a eutrophic reservoir in southwestern Ohio (176 mg C m⁻² d⁻¹), although Lake Erie has a more oxygenated water column (and presumably more CH₄ oxidation) and relatively less nutrient loading per area than a typical reservoir in this region (Beaulieu et al. 2014).

Positive efflux of CH₄ was also found in shallower parts of the lake without water column stratification, such as the western basin and Sandusky Bay (Fig. 2), indicating that CH₄ produced in sediments diffuses to surface waters without being completely oxidized, as shown in smaller lakes (Bastviken et al. 2008; Wik et al. 2013; Beaulieu et al. 2014). The western basin is traditionally the site of the largest algal blooms in Lake Erie, and is the location of the largest phosphorus sources to the lake, the Maumee and Sandusky Rivers (Baker et al. 2014; Conroy et al. 2014; Kane et al. 2014). In other words, organic carbon loading to sediments is highest in this part of the lake, and a portion of organic matter in sediments is respired as CH₄ and vented to the atmosphere. This is of particular concern as river discharge and/or algal biomass increase in Lake Erie (Michalak et al. 2013). Future studies may consider measuring primary productivity to see if C loading to sediments results in an increase in CH₄ evasion to the atmosphere. These results also indicate that Lake Erie may be a source of biogenic CH₄ in seasons other than late summer, as sediment hypoxia likely persists throughout spring and summer, if not fall and winter as well.

Our results also indicate that CH₄ produced in sediments and bottom waters of Lake Erie is partially or completely (in the case of sites with a CH₄ flux of 0) oxidized as CH₄ diffuses toward surface waters (Figs. 4, 5). Previous studies have used stable isotopes to calculate CH₄ oxidation rates in stratified lakes using the difference between surface and bottom water δ¹³C-CH₄ and the fractionation factor for CH₄ oxidation (e.g., Beaulieu et al. 2014). This is complicated in Lake Erie due to the large surface area of the lake and consistently high winds (e.g., Vachon and Prairie 2013). Subsurface and wind-driven currents also likely influence water column patterns of CH₄ concentration, and the presence of two different CH₄ sources with distinct isotope ratios makes constraining oxidation rates difficult. Future studies may be able to calculate CH₄ oxidation rates with more depth profiles of CH₄ concentrations and isotopic composition at stratified sites.

Our data in Fig. 5 illustrate the difficulty in using depth profiles of CH₄ concentration and isotopic composition to estimate CH₄ sources, vertical diffusion rates, and oxidation rates in large lakes. In the current study, we interpret stable isotopic composition of CH₄ in bottom waters as indicative of CH₄ production pathway. However, our depth profiles (Fig. 5) and isotopic data (Fig. 4) show that oxidation of CH₄ exert significant influence over isotopic composition of our samples, including in the hypolimnion and/or sediment-water interface. Our depth profile (e.g., Fig. 5A) may show that biogenic CH₄ is oxidized in the hypolimnion ([O₂] ∼1 mg L⁻¹), leading to enriched isotopic signatures, and that mixing of hypolimnetic CH₄ with atmospheric CH₄ (δ¹³C ∼−47‰) controls δ¹³C-CH₄ in the epilimnion. Our isotopic measurements of CH₄ from bottom waters indicate that most of the enriched CH₄ signatures originate from sites above natural gas wells, whereas oxidation would be expected to affect both types of sites: we interpret this as evidence that thermogenic CH₄ is indeed emitted into the lake. Additional work is needed on constraining emission rates from offshore natural gas wells through direct measurements, as well as measurements of CH₄ oxidation rates for a better understanding of CH₄ sources in Lake Erie. If natural gas leakage from these offshore wells is indeed contributing to CH₄ efflux from Lake Erie, this flux may increase as water column temperatures rise and O₂ concentrations decrease (e.g., O'Reilly et al. 2015). An additional complication to using depth profiles in Lake Erie is the possibility of lateral advection and mixing, much more likely in such a large lake. Hydrodynamic modeling of Lake Erie would also help in interpreting depth profiles.

To scale our preliminary measurements to a lake-wide estimate of CH₄ emissions, we used historical data from the NOAA Cleveland Central Buoy (http://www.glerl.noaa.gov/res/recon/station-clv.html), which is located at our sampling site 1326 (Table 1) to show that the lake is typically stratified in late summer (August and September) with dissolved oxygen levels at the sediment-water interface ranging from 0 mg L⁻¹ to 3 mg L⁻¹. Previous studies have also presented evidence that hypoxia is common in the Central Basin of Lake Erie in late summer (Zhou et al. 2013, 2015), so we assumed that our results from 2012 to 2013 were representative of normal late summer conditions. We used a surface area for Lake Erie of 25,655 km² and assumed that 12% of the lake had CH₄ fluxes similar to our average flux over natural gas wells (12.4 ± 4.8 mg C m⁻² d⁻¹) and the rest of the lake had a flux of 4.8 ± 1.1 mg C m⁻² d⁻¹ (Table 1). This leads to an estimated daily CH₄ flux from Lake Erie of 1.3 × 10⁵ kg C d⁻¹ for late summer. Propagating the standard deviation of our flux measurements leads to uncertainties of about 50%, but this could be reduced with greater intensity of flux measurements in future studies. Additional work, including year-round measurements, is needed to calculate an annual flux of CH₄ from Lake Erie, which could be used to compare emissions from Lake Erie to global lake CH₄ emissions (54 × 10¹² g C yr⁻¹) (Barros et al. 2011; Bastviken et al. 2011). Putting Lake Erie's CH₄ emissions into a regional context is also complicated. Neither Ohio nor Ontario have produced estimates of state-wide CH₄ emissions, and Lake Erie would likely constitute a “natural” rather than anthropogenic CH₄ source in such greenhouse gas budgets. Future work would benefit from collaboration with local and regional governments to assess the importance of this source in regional greenhouse gas emissions budgets.

Implications

Emissions of CH₄ from lakes and wetlands is one of the dominant sources globally, and there is growing concern that this flux may be increasing in response to increasing global temperatures. More work is needed in the Great Lakes to determine even basic terms in the C budget, and long-term monitoring can help to determine how changing climate and hydrology may affect Great Lakes biogeochemistry. Other open questions remaining for Lake Erie include determining how much organic matter in the lake is autochthonous vs. allochthonous, how much of buried organic carbon is respired as CO₂ and CH₄, and whether these emissions are large enough to overcome the impact of long-term carbon burial in sediments.

As Lake Erie has a large surface area (∼25,000 km², the thirteenth largest lake on Earth by surface area) and is the first of the Laurentian Great Lakes to host offshore natural gas wells, these results have potentially significant implications for future emissions scenarios. If increasing intensity of precipitation in the Lake Erie watershed and subsequent nutrient runoff supports greater algal biomass (Michalak et al. 2013), hypoxia may intensify, possibly exacerbating greater emissions of CH₄ in a positive feedback to climate warming (i.e., Beaulieu et al. 2014; O'Reilly et al. 2015). Furthermore, fugitive emissions of CH₄ from the natural gas supply chain may also increase as the United States and Canada increasingly exploit conventional and shale gas resources (Howarth et al. 2011; Townsend-Small et al. 2012; Howarth 2014). Our work also highlights the difficulty in scaling up CH₄ emission rate studies from small lakes to larger lakes, such as Lake Erie, as the larger surface area and higher winds make comparisons to previous studies difficult. Larger scale studies of the C cycle in all of the Laurentian Great Lakes are needed to assess the role of this system in regional and global C budgets.