Methane bubbling from northern lakes: present and future contributions to the global methane budget

Katey M Walter, Laurence C Smith, F Stuart Chapin


Large uncertainties in the budget of atmospheric methane (CH4) limit the accuracy of climate change projections. Here we describe and quantify an important source of CH4—point-source ebullition (bubbling) from northern lakes—that has not been incorporated in previous regional or global methane budgets. Employing a method recently introduced to measure ebullition more accurately by taking into account its spatial patchiness in lakes, we estimate point-source ebullition for 16 lakes in Alaska and Siberia that represent several common northern lake types: glacial, alluvial floodplain, peatland and thermokarst (thaw) lakes. Extrapolation of measured fluxes from these 16 sites to all lakes north of 45° N using circumpolar databases of lake and permafrost distributions suggests that northern lakes are a globally significant source of atmospheric CH4, emitting approximately 24.2±10.5 Tg CH4 yr−1. Thermokarst lakes have particularly high emissions because they release CH4 produced from organic matter previously sequestered in permafrost. A carbon mass balance calculation of CH4 release from thermokarst lakes on the Siberian yedoma ice complex suggests that these lakes alone would emit as much as approximately 49 000 Tg CH4 if this ice complex was to thaw completely. Using a space-for-time substitution based on the current lake distributions in permafrost-dominated and permafrost-free terrains, we estimate that lake emissions would be reduced by approximately 12% in a more probable transitional permafrost scenario and by approximately 53% in a ‘permafrost-free’ Northern Hemisphere. Long-term decline in CH4 ebullition from lakes due to lake area loss and permafrost thaw would occur only after the large release of CH4 associated thermokarst lake development in the zone of continuous permafrost.


1. Introduction

Global climate change is one of the most important issues facing modern society. Global mean surface air temperatures have increased approximately 0.6°C over the past century resulting in the retreat of glaciers, thawing of permafrost and sea ice, increase in river discharge and alteration of terrestrial and aquatic ecosystems in ways that demand both immediate and long-term societal response (Overpeck et al. 1997; IPCC 2001; Peterson et al. 2002; Romanovsky et al. 2002; ACIA 2004; Hinzman et al. 2005; Stern Review 2006). Projections suggest even greater warming resulting from rising greenhouse gas concentrations over the twenty-first century (IPCC 2001, forthcoming).

Methane (CH4) is the third most important greenhouse gas in the atmosphere after carbon dioxide (CO2) and water vapour, and it is arguably the most dynamic. During the last glacial period, the concentration of atmospheric CH4 rose and fell by 50% in association with rapid climate warming (Brook et al. 2000; Dallenbach et al. 2000). It has increased by approximately 250% since the pre-industrial era, exceeding the rate of CO2 increase by 120% (IPCC 2001). Recent decades saw a rapid rise in atmospheric CH4 concentration, 1% annually in the 1980s, followed by a slowing of the growth rate during the 1990s that has been attributed to a decline in fossil fuel emissions following the breakdown of the Soviet Union (Dlugokencky et al. 2003). Since 1999, anthropogenic emissions have increased in northern Asia, possibly reflecting acceleration of the Chinese economy (Bosquet et al. 2006).

Non-fossil fuel emissions associated with human activities include rice agriculture, waste treatment, animal husbandry and biomass burning. After years of study of atmospheric CH4, natural sources were thought to have been largely identified as wetlands, oceans, termites and hydrates. However, recent work suggests that significant new sources of atmospheric CH4 are still being identified. Keppler et al. (2006) proposed that methane production in terrestrial plants through an unknown mechanism can account for up to 50% of the modern methane sources. Walter et al. (2006) identified enhanced CH4 emissions associated with the permafrost degradation and the arctic thermokarst (thaw)-lake expansion as a positive feedback to climate warming. It seems that the more CH4 is studied, the more surprises we discover.

Understanding the atmospheric CH4 budget remains limited by large uncertainties in the individual magnitudes and spatial–temporal variability of sources at both regional and global scales (Chen & Prinn 2006). Variation in hydroxyl radical photochemistry, the primary sink of atmospheric CH4, also contributes to uncertainties in atmospheric CH4 dynamics (Bosquet et al. 2006). Nonetheless, the concentration of atmospheric CH4 is projected to increase significantly given the large pools of carbon (C) stored in permafrost that can be converted to CH4 upon thaw, the susceptibility of CH4 hydrates to release from the ocean floor with rising seawater temperature and the acceleration of CH4-producing anthropogenic activities. Since regulation efforts are focused primarily on limiting CO2 production, CH4 may become a proportionally larger agent of climate forcing in the future. It is therefore imperative that we strive to understand sources and sinks of atmospheric CH4 and improve constraints on their flux estimates so that we can better predict, prepare for, and perhaps even act to mitigate future changes.

The aim of this paper is to examine one particular source of atmospheric CH4 that may be significantly larger than previously recognized: ebullition (bubbling) from northern lakes. Careful measurements of the spatial and temporal patchiness of ebullition in Siberian thermokarst lakes revealed that total emissions from lakes were five times greater than earlier estimates that did not account for the patchiness of ebullition (Zimov et al. 1997; Walter et al. 2006). Furthermore, thaw of permafrost along lake margins releases labile organic matter previously sequestered in permafrost for centuries to millennia into anaerobic lake sediments, enhancing CH4 production and ebullition emission and serving as a positive feedback to climate warming. Here we propose that consideration of point-source ebullition, a newly recognized mode of emission that dominated total lake CH4 emissions from lakes where it was studied, will reveal that lakes are a significant source of atmospheric CH4. Given the prominence of CH4-producing lakes in the relatively C-rich, permafrost-dominated landscapes of the north, the fate of lakes is a critical source of uncertainty in climate change studies. In this paper, we estimate current CH4 ebullition emissions from all northern lakes, based on measured fluxes that are extrapolated to the circumpolar zone north of 45° N. Estimates of thermokarst-lake emissions associated with degradation of a particular type of extremely ice-rich permafrost, known as the ‘yedoma ice complex’ (Czudek & Demek 1970; Zimov et al. 2006), throughout the Holocene are used as a basis for a C-mass balance calculation to project the magnitude of future CH4 release from yedoma permafrost through lakes if permafrost continues to thaw. Finally, using a space-for-time substitution, we speculate on changes in lake emissions in two future scenarios for the north: (i) transitional permafrost (discontinuous, sporadic and isolated) disappears and lake area increases by 10% in the zone of continuous permafrost and (ii) a hypothetical permafrost-free state where surface permafrost is altogether absent.

2. The significance of methane ebullition from lakes

Much of atmospheric CH4 originates in high northern latitudes, where its atmospheric concentration is highest (Steele et al. 1987; Fung et al. 1991). The relative contribution of various northern CH4 sources is, however, still poorly understood (Nisbet 1989; Mikaloff Fletcher et al. 2004). Northern wetlands, which have been extensive through most of the Holocene (Mathews & Fung 1987; Aselmann & Crutzen 1989) were an important pre-industrial source of atmospheric CH4 (Rasmussen & Khalil 1984; Severinghaus & Brook 1999; Smith et al. 2004; MacDonald et al. 2006). Since arctic warming is predicted to enhance wetland CH4 emissions during the next century (IPCC 2001), improving the estimate of northern wetland contribution to atmospheric CH4 is an important research objective (Mathews & Fung 1987; Nisbet 1989; Reeburgh et al. 1998; Dlugokencky et al. 2001; Mikaloff Fletcher et al. 2004).

Wetlands are not the only natural source of atmospheric CH4 at high latitudes. Roughly 40% of the world's lakes occur north of 45° N. In some northern regions, lakes occupy as much as 22–48% of the land surface (Zimov et al. 1997; Hinkel et al. 2003; Riordan et al. 2006). However, despite their prominence on the landscape, lakes have been largely ignored in models that simulate the global atmospheric CH4 budget because flux estimates based on CH4 diffusion from lakes seemed too low to warrant inclusion in global models (Rudd & Hamilton 1978; Fallon et al. 1980; Whalen & Reeburgh 1990; Kling et al. 1992; Rudd et al. 1993; Bastviken et al. 2004). Although ebullition was also recognized as an important process, it was not well quantified due to its high temporal and spatial variability. When included in whole-lake flux estimates, ebullition was often extrapolated from short-term measurements with randomly placed bubble traps or floating chambers (Bartlett et al. 1988; Hamilton et al. 1994; Keller & Stallard 1994; Zimov 1997; Casper et al. 2000; Grant & Roulet 2002; Bastviken et al. 2004). These techniques systematically underestimate the magnitude of lake emissions because they fail to capture a particular type of ebullition, which, although spatially rare, can dominate whole-lake emissions (Walter et al. 2006). In Siberia, the probability of capturing point source or hotspot-bubbling points (3.7% of lake surface area) using random placement of 14 bubble traps (less than 0.03% of lake area) was only 0.001%. By walking on the surface of early winter lake ice, Walter et al. mapped the distribution and abundance of point sources and hotspots of ebullition, which are identified as specific classes of bubble clusters or open holes in lake ice distinct from background ebullition (figure 1). Daily measurements of ebullition associated with point sources and hotspots were conducted, so the whole-lake CH4 emissions could be calculated based on maps of the distribution of different source types across lake surfaces. Together, point-source and hotspot ebullition accounted for 70% of total emissions from Siberian thermokarst lakes, while molecular diffusion was only 5% (figure 2). Application of this technique to glacial lakes on the north slope of Alaska's Brooks Range suggested that ebullition accounted for approximately 95% of the CH4 flux from these lakes (Kling et al. 1992; Walter et al. submitted). Including ebullition fluxes from point sources and hotspots increased CH4 emission estimates for lakes fivefold in Siberia (from 6.8 (Zimov et al. 1997) to 34 g m−2 yr−1 (Walter et al. 2006)) and 2.5- to 14-fold in Alaska (from 0.6 g m−2 yr−1 based on molecular diffusion (Kling et al. 1992) to 1.6–9.3 g CH4 m−2 yr−1 based on point-source ebullition (Walter et al. submitted).

Figure 1

Cross-section of a North Siberian thermokarst lake showing dynamics of (a) background, (b) point-source and (c) hotspot ebullition. Background ebullition, defined as the ebullition measured by random placement of bubble traps in lakes, is typically a low-magnitude flux that covers the majority of the lake surface area. Background ebullition, characterized by lower CH4 content and younger radiocarbon age, is often produced in surface lake sediments, typically Holocene-aged peat (d) from where the buoyancy of individual bubbles is sufficient to release bubbles from sediments where they are formed. In contrast, CH4 emitted through point sources produced deeper in lake sediments is characterized by higher concentrations of CH4 and older radiocarbon ages than background ebullition (Walter et al. in press a). Methane bubbles originating from greater depths beneath lakes flow into common bubble tubes and are emitted through single small holes at the sediment surface. Rising to the surface of the lake from these fixed points, bubbles become trapped in lake ice throughout winter, forming distinct patterns in ice depending on the magnitude of CH4 emissions. Fixed points from which extremely high ebullition rates are sufficient to maintain open holes in lake ice throughout winter are called hotspots. Hotspots emit bubbles with a high CH4 concentration (up to 90%) and ancient radiocarbon age (up to 43 000 years), suggesting that CH4 is produced at great depth in the lake's thaw bulb (e) from an ancient organic matter source (Pleistocene-aged organics in yedoma) and discharged via a network of bubble paths that come together as a single bubble tube at the sediment surface. Stable isotope signatures suggest that lake CH4 is of biogenic origin (Zimov et al. 1997; Walter et al. 2006). Thermokarst lakes studied occurred in the zone of continuous permafrost in the yedoma (f; icy, Pleistocene-aged, organic-rich permafrost) region of Siberia.

Figure 2

Pie chart depicting the relative contributions of different measured modes of CH4 emissions from North Siberian thermokarst lakes during winter (ice-cover period, dark slices) and summer (ice-free period, light slices; after Walter et al. 2006). Together, point-source and hotspot ebullition dominated annual lake emissions.

We estimated point-source ebullition sampled for 16 lakes in Alaska and Siberia (figure 3) that represent common types of pan-arctic lakes including glacial, alluvial floodplain, peatland and thermokarst lakes. Point-source ebullition occurred in all lakes sampled (except one large glacial lake in Alaska) and was greatest in thermokarst lakes, where thermokarst erosion actively deposits organic materials released from thawing permafrost. Average point-source emissions from non-thermokarst lakes were variable (17.9±12.1 g CH4 m−2 yr−1) and did not differ by lake type. Areas of thermokarst lakes influenced by thermokarst erosion, on the other hand, had 7.5-fold higher average CH4 point-source emissions (135±51.8 g CH4 m−2 yr−1; SAS ANOVA. F=17.983,12, p<0.0001).

Figure 3

Point-source ebullition measured from 16 lakes in Alaska and Siberia representing several common northern lake types: glacial, alluvial-floodplain, peatland and thermokarst lakes. Error bars represent standard deviation of ebullition based on 2–3 transects of point-source surveys per lake. Data sources were previously described by Walter et al. (2006) and Walter et al. (submitted), except data for two peatland lakes, noted by asterisks, in Alaska which are presented here for the first time.

3. Estimating CH4 ebullition from all northern lakes

The current spatial distribution of lakes and wetlands in the Northern Hemisphere is influenced by numerous factors including climate (precipitation minus evapotranspiration, P−E), geomorphology, substrate permeability, glacial history, the presence or absence of permafrost and permafrost properties (Wetzel 2001; Yoshikawa & Hinzman 2003; Klein et al. 2005; Smith et al. in press). By combining global databases on the location of large lakes (sized 0.1–50 km2; Lehner & Doll 2004), topography, permafrost (Brown et al. 1997, 2001), peatlands (MacDonald et al. 2006) and extent of Last Glacial Maximum glaciation (Ray & Adams 2001), Smith et al. (in press) identified glaciation history and the presence of permafrost to be the greatest first-order controls on the current distribution of lakes in the Northern Hemisphere (north of 45° N). Lake densities and lake area fractions were 300–350% greater in glaciated versus non-glaciated terrain, and approximately 100–170% greater in permafrost influenced versus permafrost-free terrain. Geomorphological processes associated with glaciation promote lake formation by reducing topographic relief, carving depressions in bedrock and depositing dead ice that forms kettle lakes upon thaw. The Canadian Shield is particularly rich in lakes formed following deglaciation (figure 4). Excluding Greenland, nearly two-thirds (27.3×106 km2) of the total land area (approx. 41.3×106 km2) was classified as lowlands or previously ice covered and 95% of northern lakes occur in these lowland and postglacial terrains. Of those lakes, 75% occur in permafrost. Permafrost promotes lakes by reducing infiltration of surface water into the subsurface and through its role in thermokarst-lake cycles (Carson & Hussey 1962; Sellmann et al. 1975; Jorgenson & Osterkamp 2005). Organic-rich soils, such as peat and gytija, and lacustrine or glacio-marine clays with low hydraulic conductivities can also promote lakes in peatland areas by reducing infiltration. Climate (P−E), a critical factor that influences lake area and distributions on shorter time-scales (Klein et al. 2005; Riordan et al. 2006), was not evaluated by Smith et al.

Figure 4

Pan-arctic distribution of lakes using GLWD data northward of 45° N overlaying permafrost zones (continuous, discontinuous, sporadic and isolated) depicted by colour. Lakes are particularly abundant in North Siberia and on the Canadian Shield.

The amount of CH4 emitted from northern lakes annually is unknown. The few studies that have actually measured CH4 emissions (Rudd & Hamilton 1978; Whalen & Reeburgh 1990; Naiman et al. 1991; Kling et al. 1992; Rudd et al. 1993; Hamilton et al. 1994; Nakayama et al. 1994; Roulet et al. 1994, 1997; Zimov 1997; Bastviken et al. 2004) probably missed the large flux associated with point-source bubbling. To determine whether or not such emissions are of sufficient magnitude to be included as important sources in the global atmospheric CH4 budget, we applied the average CH4 point-source emission measured across the non-thermokarst lake types (17.9±12.1 g CH4 m−2 yr−1) to the area of lakes north of 45° N (table 1) except in thermokarst lakes of the Russian zone of continuous permafrost, where we applied CH4 emission estimates from detailed field-based measurements of Siberian thermokarst lakes (34.5±9.5 g CH4 m−2 yr−1; Walter et al. 2006). We calculated the area of Russian thermokarst lakes based on Geographical Information System (GIS) analyses showing that 65.5% of lakes north of the Arctic Circle occur in Russia (Holmes & Lammers 2006) and estimate (Mostakhov 1973) that 90% of Russian lakes in the permafrost zone are thermokarst lakes. We excluded all large lakes (more than 50 km2) from our estimate owing to the possibility of low rates of point-source ebullition (table 2). Altogether, we estimate point-source ebullition from northern lakes to be approximately 24.2±10.5 Tg CH4 yr−1 (table 1). In contrast, the diffusive flux is 1.1±0.2 Tg CH4 yr−1 assuming a constant flux for all lake area (1.0±0.2 g CH4 m−2 d−1), which is the average diffusive flux measured for glacial lakes on the north slope of Alaska's Brooks Range (Kling et al. 1992) and for Siberian thermokarst lakes (Walter et al. 2006), assuming 120 days of open water annually (Walter et al. 2006).

View this table:
Table 1

Estimating CH4 emissions from northern lakes based on GIS calculations of lake area in permafrost-dominated and permafrost-free regions for the (a) present day and for future scenarios where (b) transitional permafrost disappears (scenario 1) and where (c) surface permafrost disappears altogether (scenario 2).

View this table:
Table 2

Estimating CH4 emissions from northern lakes and global lakes in the context of the global atmospheric CH4 budget.

Our estimates of the ebullition flux from northern lakes are conservative for several reasons. (i) Although thermokarst lakes are common in the discontinuous permafrost zone of Russia as well as in North America and Europe, their areal extent is unknown; therefore, we assigned the lower, non-thermokarst lake CH4-emission value to all lakes in these regions in our calculations. (ii) Preliminary measurements in several boreal thermokarst ponds in Alaska suggest that fluxes from thermokarst lakes and ponds could be higher than we assume in our calculations. For example, the point-source ebullition from a boreal thermokarst pond in Alaska was 5.6-fold higher (195±8 g CH4 m−2 yr−1) than the thermokarst-lake emission value used in our calculation (34.5±9.5 g CH4 m−2 yr−1; Walter et al. submitted). (iii) The fine-scale lake database from which we estimated lake area, which is the best available (Lehner & Doll 2004), excludes lakes smaller than 0.1 km2 and therefore significantly underestimates thermokarst-lake numbers and area (Frey & Smith 2007; our observation figure 5). These small lakes are particularly important contributors to CH4 ebullition because they are the most numerous lakes in North Siberia (Grosse et al. 2005). In addition, small lakes have the largest fluxes per unit area because their low area: perimeter ratio causes lake-margin carbon inputs (thermokarst erosion and aquatic plant production) to be relatively more influential. (iv) On several occasions, we observed pulses of CH4 release that were large enough to lift thick carpets of lake-bottom peat to the surface for several days (observed twice) or to produce violent eruptions of CH4 along lake margins lasting seconds to tens of seconds (Walter et al. 2006). These large CH4 release events occurred too infrequently to measure. All these sources of uncertainty suggest that there is still much to learn about CH4 emission from northern lakes and that future research is likely to increase the magnitude of our current estimate of CH4 flux. Improved information on different lake types and sizes could be particularly important. Synthetic Aperture Radar (SAR), which provides a remote-sensing signal of CH4 bubbles in lake ice, may provide a new tool for upscaling field measurements of CH4 ebullition to the pan-arctic scale (Walter et al. in press b).

Figure 5

Quantifying lake area at the pan-arctic scale using the Global Lakes and Wetlands Database (Lehner & Doll 2004) with applied correction factor from ASTER data analysis. Though it is the best pan-arctic GIS lake coverage we know of, the Global Lake and Wetlands Database (bright lake perimeters) significantly underestimated lake area for the Kolyma region of northeastern Siberia. Comparison with finer resolution ASTER data revealed that many small lakes were not captured by the GLWD.

Ebullition also occurs in low-latitude lakes, reservoirs, rivers and wetlands, particularly where there are large organic matter inputs to sediments (Rudd et al. 1993; Keller & Stallard 1994; Casper et al. 2000; St Louis et al. 2000). Ebullition rates could be higher in low-latitude lakes than in the Arctic because higher sediment temperatures should speed methanogenesis (Valentine et al. 1994; Galy-Lacaux et al. 1999; St Louis et al. 2000) and make CH4 less soluble (Yamamoto et al. 1976; Fearnside 2004), forming bubbles more readily in sediments. Nonetheless, if we apply the conservative non-thermokarst lake point-source ebullition rate to low-latitude lakes, where point-source ebullition has not been specifically quantified (1.5×106 km2), then ebullition emissions of small lakes (0.1–50 km2) globally would be approximately 31.7±15.6 Tg CH4 yr−1 or up to 8% of global emissions (table 2).

4. Siberian thermokarst: a time bomb for future CH4 emissions?

Northern lakes are not only important sources of atmospheric CH4 today, but their importance will increase as climate change proceeds. Northern Hemisphere permafrost contains approximately 950 Gt of C (Zimov et al. 2006), an amount that would more than double the current atmospheric CO2 concentration if oxidized upon thaw under aerobic conditions. Roughly half of this C is contained in the yedoma ice complex of North Siberia, whose extent is 106 km2 (Zimov et al. 2006), only 7.6% of permafrost area and 3.7% of the land surface area north of 45° N. The yedoma ice complex formed in the Late Pleistocene primarily on the extensive unglaciated lowlands of Siberia is unique among permafrost substrates given its high ice-wedge content (50% by volume) and large pool of labile organic matter (Zimov et al. 2006; figure 1f). Laboratory incubations and field studies revealed that the C in yedoma is extremely labile, and if thawed under aerobic conditions it is nearly 100% mineralized within a century (Dutta et al. 2006; Zimov et al. 2006). Analyses of permafrost degradation in this ice complex revealed that since the Last Glacial Maximum roughly half of yedoma has thawed, mostly under inundated, anaerobic conditions either by coastal erosion associated with sea-level rise (Romanovskii et al. 2000) or by inland thermokarst-lake formation and expansion (Czudek & Demek 1970). Under anaerobic conditions, CH4, whose relative greenhouse effect is 23 times greater than that of CO2, is produced in equal proportion to CO2 (Conrad et al. 2002). Since the mean annual temperature of lake water exceeds that of surrounding permafrost, thermal erosion of permafrost occurs along lake margins and at the edges of thaw bulbs beneath lakes. Given the high ice content of yedoma, melting ice causes the ground surface to collapse, releasing previously frozen organic C stored in yedoma to microbial decomposition in the lake bottoms. This process, called ‘thermokarst’, continuously exposes new frozen surfaces to thermal erosion, particularly during periods of high summer insolation or years of high precipitation (Bosikov 1991). In large lakes, wave-driven erosion enhances thermokarst along lake margins and, together with thermal erosion (Carson & Hussey 1962; Czudek & Demek 1970), constitutes the mechanism by which thermokarst lakes migrate and degrade ice-rich permafrost.

Using patterns of thermokarst-lake development throughout the Holocene and a C-mass balance calculation based on the amount of C decomposed in yedoma beneath lakes, we predict here the magnitude of CH4 that may escape from lakes to the atmosphere in the future if yedoma thaws completely. Today yedoma contains approximately 450 Gt of C (Zimov et al. 2006), of which approximately 269 Gt C are contained in yedoma that has remained frozen since the Pleistocene. We assume that 50% of the 269 Gt of C will be exposed to anaerobic decomposition conditions beneath lakes in the future as permafrost thaws, consistent with the pattern of lake development throughout the Holocene evidenced by lake scars that cover 50% of the yedoma region (Czudek & Demek 1970; Walter et al. 2006; Zimov et al. 2006). Measurements of the C content of samples collected from exposures of yedoma frozen since the Pleistocene and from cross-sections of the refrozen thaw bulbs of former lakes exposed along the cut banks of Siberian rivers revealed that 33% of the C stored in yedoma is decomposed beneath lakes under anaerobic conditions (Kholodov et al. 2003; Walter et al. 2006; Zimov et al. 2006). Applying a 16.5% conversion factor of yedoma C to CH4 in accordance with the stoichiometry of methanogenesis whereby CO2 and CH4 are produced in equal proportions (Conrad et al. 2002), 135 Gt of C from yedoma that has remained frozen since the Pleistocene would be converted to CH4 as yedoma warms and thaws in the future. This is equivalent to approximately 30 000 Tg of CH4. This estimate does not include CH4 that would also be produced from thermokarst lakes that form in the basins of relict drained lakes, a source which has not yet been quantified, but which we expect would further increase regional lake emissions. We can, however, estimate additional CH4 emissions associated with CH4 that would be produced from allochthonous and autochthonous organic matter sources other than yedoma permafrost. Radiocarbon analysis of CH4 emitted from modern thermokarst lakes in the yedoma region of Siberia revealed that approximately 60% of CH4 emitted from lakes today is fuelled by Pleistocene-aged organic C, while decomposition of contemporary organic detritus contributes to the remaining lake-CH4 emissions (Walter et al. 2006). If this pattern holds in the future, then we could expect a total of approximately 49 000 Tg of CH4 to be emitted from lakes fuelled by both permafrost and contemporary C sources if yedoma warms and thaws as predicted (ACIA 2004; Sazonova et al. 2004).

Variability in global climate models gives rise to uncertainties in predicting rates of yedoma degradation and thermokarst lake expansion; however, the mean annual temperature of Siberian permafrost increased up to 3°C during recent decades (Romanovsky et al. 2001; Sazonova et al. 2004) and is predicted to continue warming and thawing during this century (Romanovsky et al. 2001; ACIA 2004; Sazonova et al. 2004; Lawrence & Slater 2005). In the discontinuous permafrost zone in Alaska, permafrost temperatures are within 1°C of thawing (Osterkamp & Romanovsky 1999). Thus, the large pool of still-frozen Pleistocene-age C in Siberia can be considered a potential CH4 time bomb, with sufficient C stores that thermokarst lake development would release approximately 10 times the current atmospheric CH4 burden (4850 Tg CH4, IPCC) by bubbling at some (as yet unknown) rate as the zero degree temperature threshold of yedoma is crossed. If this CH4 were to be released during the next 500–1000 years, then average CH4 emissions rates from lakes would be approximately 50–100 Tg CH4 yr−1, rates that are approximately 8–50% of those predicted for global anthropogenic CH4 emissions under different scenarios of global warming by 2100 (236–597 Tg CH4 yr−1, IPCC SRES Emissions Scenarios 2000; mean, 184 Tg CH4 yr−1, de la Chesnaye & Weyant 2006). Our estimates of CH4 emission from future yedoma thermokarst lakes also do not include degradation of permafrost in the remainder of the Northern Hemisphere. Another approximately 450 Gt of C is thought to reside in surface permafrost of other northern terrestrial soils (ACIA 2004; Smith et al. 2004), and thermokarst lakes in these regions will also be important sources of CH4 to the atmosphere.

5. Future of northern lake emissions in a permafrost-free Arctic

Widespread permafrost warming and degradation are projected to intensify during the twenty-first century (ACIA 2004). Permafrost in the interior of Alaska has already warmed by 1.5°C since the 1980s (Osterkamp & Romanovsky 1999; Osterkamp 2003), and temperatures in boreholes on the north slope of Alaska rose by 2–4°C during the past 50–100 years (Lachenbruch & Marshall 1986). Warm temperatures from 1989 to 1998 led to the thaw of massive ice wedges that had been stable for thousands of years in northern Alaska (Jorgenson et al. 2006). The occurrence of thermokarst ponds and depressions has long been observed in association with permafrost degradation in Alaska and Canada (Sellmann et al. 1975; Burn & Smith 1990; Jorgenson et al. 2001) as well as in Mongolia (Sharkuu 1998), China (Ding 1998) and Russia (Czudek & Demek 1970; Bosikov 1991; Pavlov 1994). More recently, studies have documented changes in lake area occurring during the past half century associated with permafrost degradation (Yoshikawa & Hinzman 2003; Smith et al. 2005; Riordan et al. 2006; Walter et al. 2006). Other factors suggested to influence lake area dynamics in association with permafrost thaw are climatic changes (P−E, Riordan et al. 2006); wildfires, which lower albedo and increase soil thermal conductivity and ground heat flux (Chambers & Chapin 2002; Yoshikawa et al. 2003); and the interaction between surface and groundwater, particularly when permafrost degradation leads to open thaw bulbs beneath lakes (Kane & Slaughter 1973; Yoshikawa & Hinzman 2003). In areas of discontinuous permafrost, remote-sensing studies revealed a decrease in lake area (Yoshikawa & Hinzman 2003; Riordan 2006), while areas in continuous permafrost have seen a net increase in lakes (Smith et al. 2005; Walter et al. 2006) and water-filled pits (Jorgenson et al. 2006) or no change in lake area (Riordan et al. 2006). Despite the observed changes in particular regions of Siberia and Alaska, Smith et al. (in press) found surprisingly little difference in the abundance of lakes at the pan-arctic scale between the different zones of continuous, discontinuous, sporadic and isolated permafrost. It was the lack of permafrost altogether that had the greatest effect (decrease) on lake abundance, even in the previously glaciated regions where, over large spatial scales, the prevalence of lakes decreases by approximately a factor of two in the absence of permafrost (Smith et al. in press).

Given that lakes are net emitters of CH4, the fate of lakes in the north is an important question to be reconciled in models of climate change. In the short term, permafrost degradation through thermokarst-lake formation and expansion, particularly in the zone of continuous permafrost, is expected to release a large amount of CH4 as explained above. However, widespread loss of lakes over the long term as surface permafrost disappears (Smith et al. in press) should cause a corresponding reduction in Northern Hemisphere CH4 emissions (figure 6), particularly if lakes are replaced by aerobic, non-wetland ecosystems. If permafrost degradation and lake area loss cause wetland area to increase, changes in regional CH4 emissions may be offset by emissions from wetlands (Vourlitis & Oechel 1997; Friborg et al. 2003; Christensen et al. 2004).

Figure 6

Schematic depicting future of CH4 emissions from northern lakes as the north changes from a permafrost-rich landscape to a landscape free of surface permafrost. A large release of CH4 is anticipated in association with thermokarst lake degradation of permafrost in the zone of continuous permafrost, particularly in the region of Siberian yedoma permafrost. As permafrost degradation proceeds, eventually there will be a loss of lake area and thermokarst lakes. In the long-term permafrost-free north, a net loss in lake area is estimated to result in a decrease in northern lake CH4 emissions.

To what extent might lakes disappear and what will be the effect on atmospheric CH4? To address this question, we used the ‘space-for-time’ substitution of lake area and lake area fraction (Lf) for two future scenarios of permafrost degradation after Smith et al. (in press; table 1), applying the 50% correction factor for small lakes missed in the GLWD (figure 5). In a more probable transitional scenario (scenario 1), areas that are today transitional permafrost (discontinuous, sporadic and isolated) become permafrost free, while the prevalence of lakes increases by 10% in the area that is currently continuous permafrost, roughly doubling the lake area increase observed in the region of continuous permafrost in Siberia since 1973 (Smith et al. 2005, in press). To estimate changes in lake area in scenario 1, we replaced the lake area fraction (Lf=total lake area/total land area×100) attributed to transitional zones of permafrost (discontinuous Lf=6.02; sporadic Lf=6.24, and isolated Lf=5.50) with the area of lakes characteristic of modern permafrost-free landscapes (Lf=2.32). The result was a 58–63% loss in lake area depending on the permafrost zone (figure 7). The lake area fraction in the zone of current continuous permafrost (Lf=6.14) was increased by 10% to Lf=6.75. In the more extreme scenario (permafrost-free, scenario 2), replacing the area of lakes attributed to each of the currently permafrost-dominated regions with the area of lakes characteristic of permafrost-free landscapes, we found a 58–63% loss in lake area in the various permafrost zones (figure 7). The area of lakes in the current permafrost-free zone remains the same in both future scenarios. As a result, if permafrost thaws according to scenarios 1 and 2 then the land north of 45° N would experience a net lake area loss of 16% (scenario 1) and eventually 44% (scenario 2) compared with today (figure 7). Applying CH4 ebullition rates associated with non-thermokarst lakes to all lakes in the transitional and permafrost-free scenarios, except Russian thermokarst lakes in the continuous permafrost zone in scenario 1, resulted in a net 12 and 53% decline in CH4 ebullition emissions from northern lakes, respectively (21.4±10.7 Tg CH4 yr−1 transitional scenario; 11.3±7.7 Tg CH4 yr−1 permafrost-free scenario), compared with the estimate of modern emissions (24.2±10.6 Tg CH4 yr−1; table 1). Predicted changes in diffusive emission estimates are also shown in table 1, but are significantly lower than the dominant mode of ebullition emission. This approach provides a rough estimate of changes in lake area and CH4 emissions associated with potential future disappearance of surface permafrost.

Figure 7

Changes in northern lake area predicted for future scenarios in which transitional permafrost disappears (scenario 1) and surface permafrost altogether disappears (scenario 2) based on a space-for-time substitution of lake area fractions from current permafrost-free land surfaces to current permafrost-dominated land areas. Corrected lake area values (y-axis) were calculated from the GIS estimates of GLWD lakes by Smith et al. (in press).

6. Conclusion

In northern high latitudes, where the concentration of atmospheric CH4 is highest and where much of atmospheric CH4 originates (Steele et al. 1987; Fung et al. 1991), lakes are a dominant landscape feature occupying as much as 48% of the land surface in some regions (Riordan et al. 2006). Lakes are important emitters of CH4, particularly when attention is paid to ebullition (Hamilton et al. 1994; Keller & Stallard 1994; Zimov et al. 1997; Walter et al. 2006). However, the role of lakes in the atmospheric CH4 budget had not previously been assessed. The abundance of lakes in the north appears to be governed in part by the presence or absence of permafrost (Yoshikawa & Hinzman 2003; Smith et al. 2005, in press). Landscape processes such as permafrost degradation and thermokarst erosion are known to enhance CH4 production and emissions from expanding thermokarst lakes in permafrost regions (Zimov et al. 1997; Walter et al. 2006). In this study, we extrapolated measurements of point-source ebullition from 16 lakes representing common northern lake types to the extent of current lake distributions in regions with and without permafrost to estimate total northern lake CH4 emissions (approx. 24.2±10.5 Tg CH4 yr−1).

Rather than attempting to provide a precise estimate of lake CH4 emissions, we sought to demonstrate that ebullition from lakes may be a much larger and globally significant source of atmospheric CH4 than that previously thought because (i) point-source ebullition is a dominant (and previously unrecognized) source of CH4 emissions from lakes, and (ii) lakes are a prominent landscape feature in the north that convert organic C sequestered for hundreds to thousands of years in permafrost into the radiatively important CH4 in the atmosphere. Our calculations suggest that tens of thousands of teragrams of CH4 will be released from thermokarst lakes as permafrost warms and thaws in the future, but that, eventually, disappearance of permafrost altogether will result in a net, approximately 50%, loss of lake area and associated CH4 emissions from lakes. These estimates can be refined by research that accounts for variability in CH4 emissions from different types of northern lakes and lake regions. The estimates presented here are conservative in that they do not account for the many high-emission thermokarst lakes in North America, Europe and the discontinuous permafrost zone of Russia. The magnitude of both current and future emission estimates from northern lakes is large and should be accounted for in models of atmospheric CH4 dynamics and ecosystem feedbacks to climate change.


We thank Peter Schlesinger at Woods Hole Research Center for conducting the analysis of correction factor on the Global Lakes and Wetland Database using analysis of Aster data. Field surveys were conducted with the assistance of Dimtri Draluk, Faith Reynolds, Edward Ricter and Catherine Thompson. Justin Mattix assisted with construction of figure 1. Logistical support was provided by the Northeast Science Station in Cherskii, Russia and Toolik Lake Field Station, Alaska. Research funding was provided by the National Science Foundation through the Russian–American Initiative on Shelf-Land Environments of the Arctic (RAISE) of the Arctic System Science Program (ARCSS) and Polar Programs, Environmental Protection Agency STAR Fellowship Program, and NASA Earth System Science Fellowship Program.


  • One contribution of 18 to a Discussion Meeting Issue ‘Trace gas biogeochemistry and global change’.


View Abstract