𝗘𝗺𝗲𝗿𝗴𝗲𝗻𝘁 𝗯𝗶𝗼𝗴𝗲𝗼𝗰𝗵𝗲𝗺𝗶𝗰𝗮𝗹 𝗿𝗶𝘀𝗸𝘀 𝗳𝗿𝗼𝗺 𝗔𝗿𝗰𝘁𝗶𝗰 𝗽𝗲𝗿𝗺𝗮𝗳𝗿𝗼𝘀𝘁 𝗱𝗲𝗴𝗿𝗮𝗱𝗮𝘁𝗶𝗼𝗻 - 𝗡𝗮𝘁𝘂𝗿𝗲 𝗖𝗹𝗶𝗺𝗮𝘁𝗲 𝗖𝗵𝗮𝗻𝗴𝗲

Water column

The world's ocean C inventory is dominated by high concentrations of inorganic C (>2000 μmol/kg seawater) in the water column. Inorganic C originates from air–sea exchange, river runoff, and to a lesser degree, dissolution of carbonate-containing minerals, decay of organic matter, and biological respiration. A relatively small, but detectable, fraction of the ocean's inorganic C inventory has accumulated from the increase of atmospheric CO₂ associated with the burning of fossil fuels (Anderson et al. 1998a, 2000). The world's oceans also contain substantial concentrations of organic C (35–150 μmol/kg seawater). Organic C in the ocean originates from both terrestrial sources (river runoff, coastal erosion, atmospheric deposition) and marine primary production. Within the ocean, inorganic and organic C can occur in both particulate (PIC and POC) and dissolved (DIC and DOC) phases. The DOC, POC, and PIC fractions in the water column vary seasonally, especially in the surface water. This strong seasonal signal creates a wide variance in any estimate of DOC, POC, or PIC. However, because the C inventory is dominated by DIC, which has much lower seasonal variability, the overall C inventory varies only by about 5% seasonally.

Seawater entering the Arctic Ocean from both the Pacific Ocean through Bering Strait and from the Atlantic Ocean through Fram Strait and the Barents Sea is diluted in summer by both sea ice meltwater and by river runoff. Even though DIC varies substantially with salinity, with open ocean concentration being ∼2150 μmol/kg at salinity 35‰, a linear relationship between salinity and DIC cannot be applied in the Arctic Ocean as the two sources of freshwater have different DIC concentrations. Based on the observed concentrations of DIC and DOC of water masses with different salinities (Anderson et al. 1998b, Anderson 2002), we estimate DIC stocks of 310 Pg C and DOC stocks of 9 Pg C for the Arctic Ocean, including the shelf seas but not including the Nordic Seas or the Bering Sea (Table 1).

Table 1. Content of dissolved inorganic carbon (DIC) and dissolved organic carbon (DOC) in the Arctic shelf seas and the central basins computed from volumes and concentrations.

The particulate C content of the Arctic Ocean is characterized by substantial seasonal variability. There are a few plankton species that produce PIC, i.e., calcium carbonate shells, in the Arctic Ocean. Calcifying plankton like coccolithophorids are common in the Norwegian Sea and during the last few decades blooms have also been observed both in the Bering Strait and in the Barents Sea (Smyth et al. 2004). However, PIC content is very low within the water column of the Arctic Ocean and appears also to be a small component of settling particles (O'Brien et al. 2006). Marine POC, on the other hand, builds up to significant levels in the shelf seas. Thus, PIC is negligible in the Arctic Ocean C budget but POC may not be. However, the POC content of the water column is very difficult to estimate and it is thus not possible to include it in the general budget we have developed for this review.

The first published measurements of CH₄ in the Arctic Ocean were made during open water in the Beaufort Sea (Macdonald 1976) where it was found to be near equilibrium with the atmosphere (∼3.5 nmol/L) in surface waters but considerably above saturation near the sea bottom (up to 50 nmol/L), which implies that sediments are a major source of CH₄ to shelf waters. Based on these data, we estimate an inventory of approximately 0.4 Gg CH₄ for Canadian Beaufort Shelf water (64 000 km²; mean depth 42 m). Kvenvolden et al. (1993b) observed generally higher concentrations (11–112 nmol/L) in late winter over the Alaskan shelf, which they proposed were due to buildup during the winter when air–sea exchange was prohibited by the ice cover. More recently, Shakhova et al. (2005) measured CH₄ for the East Siberian and Laptev Seas in the summers of 2003 and 2004, showing both areal and vertical patchiness in concentration, with ranges in surface concentration from 2 to 110 nmol/L and an average of about 12 nmol/L, with higher concentrations at depth (up to 154 nmol/L in 2004). From these data, Shakhova et al. (2005) estimated inventories of 5.7 and 1.6 Gg CH₄ for 2003 and 2004, respectively, in their study area (∼160 × 10³ km²). The inventory of CH₄ in the shelf seas clearly varies with season, depending on ice cover and plume distribution, but we can embrace most measurements by taking saturation (∼4 nmol/L) as the lowest value and 240 nmol/L as an upper value (e.g., Macdonald 1976, Damm et al. 2005, Shakova et al. 2005). Based on Jakobsson's (2002) volume estimate for redefined constituent seas (0.65 × 10⁶ km³), we estimate a shelf inventory of 0.1–2.5 Tg CH₄.

There are fewer measurements for CH₄ in the deep Arctic Ocean away from seeps or shelves, and the available data indicate that water column CH₄ of the Arctic Ocean Basin is at or less than atmospheric equilibrium (Macdonald 1976, Damm et al. 2005), which agrees with measurements from other oceans (Conrad and Seiler 1988, Holmes et al. 2000). Elevated CH₄ concentrations may be observed in the upper ocean (∼300 m) associated with particles and bacterial methanogenesis (Karl and Tilbrook 1994, Holmes et al. 2000), although we assume the interior Arctic Ocean (under the permanent ice pack) to be a poor location for such CH₄ production due to the low productivity and associated particle flux (O'Brien et al. 2006). Accordingly, a basin CH₄ concentration range of 1–4 nmol/L together with basin-water volume of 12.3 × 10⁶ km³ implies an inventory of 0.2–0.8 Tg CH₄. Sea ice, which excludes dissolved gases during freezing, likely contributes negligibly to the CH₄ inventory. We combine our estimates of the shelf inventory and the basin inventory to estimate that the Arctic Ocean and associated shelf seas have a CH₄ inventory of between 0.3 and 3.3 Tg CH₄.

Sediment

Enormous amounts of particulate C are stored in Arctic Ocean sediments, mainly in the shelf seas. Where sediments have accumulated for a sufficient time, some of the organic matter has been transformed to oil and gas (e.g., see Bakke et al. 1998: Fig. 10.1) and in a number of areas into CH₄ hydrates (e.g., see Kvenvolden 1988). Estimating the amount of C in Arctic sediments is somewhat arbitrary depending on the depth/age range of sediments used. We estimate the inventory of organic C residing in the sediment surface mixed layer (∼3 cm in the basins and 10 cm over the shelves and slopes) to be ∼7.8 Pg C (shelves) and 1.6 Pg C (basins). This inventory of 9.4 Pg C includes recent deposition and organic matter potentially in contact with biota, and therefore is potentially available for biological processing.

Atmospheric analyses of carbon dioxide exchange

Overall, current atmospheric inversion studies indicate that during the 1990s the Arctic region was on average a modest CO₂ sink that is less than 0.5 Pg C/yr (Table 2). These estimates do not include fluxes from the European sector of the Arctic region (northern Europe and European Russia), which cannot be separately extracted from the global region layout defined by the TransCom inter-comparison experiment (Gurney et al. 2003). The atmospheric inversions indicate that these average surface–atmosphere CO₂ fluxes are subject to substantial interannual variations (up to ±0.5 Pg C/yr) caused primarily by climate fluctuations and associated variation in wildfires on land CO₂ exchange (Bousquet et al. 2000, Rödenbeck et al. 2003, Baker et al. 2006). The listed uncertainties in Table 2 are 1 standard deviation as estimated by the inversion procedure. In the case of multi-model averages, the listed uncertainty also includes the transport model error computed from the spread of the different model flux estimates (Baker et al. 2006). The uncertainty estimate for the flux from the entire Arctic has been computed as the geometric mean of the flux uncertainties of the three component regions. This yields a conservative uncertainty estimate, since potential uncertainty covariances among the three regions are neglected.

Table 2. Recent, representative top-down inversion-based estimates of surface-atmosphere CO₂ fluxes in the Arctic region given in Pg C/yr.

There are several important limitations associated with the inversion methodology, including the large computational resources required for the inversion of the atmospheric transport model and the imperfection in the numerical representation of sub-grid scale mixing processes (e.g., surface and planetary boundary layer dynamics, vertical transport in convective clouds). Furthermore, significant uncertainty in the model results stems from a sparse observational network (see Gurney et al. 2003), where individual measurements are of limited accuracy and precision and thus may not be representative of the appropriate temporal and spatial scale of the transport model. Furthermore, observations are often not readily comparable across monitoring networks due to differences in measurement techniques and the use of different standards. The most serious limitation of the top-down approach follows from the limited number of observations and the need to adequately represent sources and sinks with a relatively high spatial and temporal resolution. In general, there exists a very large number of possible surface source-sink configurations that are in principle consistent with the atmospheric observations. As such, the atmospheric observations alone are not sufficient to uniquely determine the sinks and sources of CO₂ at the surface of the earth.

A recent evaluation of the atmospheric transport models employed in the TransCom experiment has indicated systematic biases in how most of the models represent the vertical dilution of the seasonal cycle of the CO₂ concentration over the northern extratropical hemisphere (Stephens et al. 2007). These biases imply an overestimation on the order of 30% of the magnitude of the northern extratropical CO₂ sink as inferred by the models when run in the inverse mode. Because the northern extratropical region represents area north of approximately 20° N, it is not clear if these biases also imply an overestimation of the atmospheric inversion estimates for the Arctic. We note that the high-resolution, time dependent inversion conducted by Rödenbeck et al. (2003) listed in Table 2 was performed with one of the models that showed a relatively small bias. For purposes of this review, we adopt the 0.0–0.8 Pg C/yr sink estimate for the Arctic from Baker et al. (2006), which has an uncertainty that is intermediate between Gurney et al. (2003) and Rodenbeck et al. (2003).

Estimates of terrestrial carbon dioxide exchange

The net exchange of CO₂ of terrestrial ecosystems is generally determined by the net production of vegetation C by plants (photosynthesis minus plant respiration), the decomposition of organic matter, and the release of CO₂ in fires that burn organic C stored in vegetation and soils. Global analyses indicate that vegetation of terrestrial ecosystems fix approximately 100 Pg C/yr through the process of photosynthesis and release about half of that to the atmosphere in plant respiration and about half in the decomposition of soil organic matter (McGuire et al. 1997). Tundra ecosystems are responsible for approximately 2% of this exchange, while boreal forest ecosystems are responsible for approximately 8% (McGuire et al. 1997). Freshwater systems serve to cycle C from terrestrial systems into marine systems and the atmosphere. In this section we summarize research that has evaluated the net exchange of CO₂ from (1) tundra and boreal forest ecosystems, and (2) freshwater lake and river ecosystems during the last two decades.

1. Tundra and boreal forest ecosystems

Multi-year ground-based C budget analyses at arctic tundra and boreal forest sites show a high degree of spatial and temporal variability, making it difficult to ascertain whether these ecosystems as a whole are currently acting as longer-term sinks or sources for atmospheric CO₂ (e.g., Schulze et al. 1999, Aurela et al. 2001, Hobbie et al. 2002). Studies have shown rather dramatic fluctuations in annual budgets over decadal time scales at both tundra and forest sites (Oechel et al. 2000, Barr et al. 2007, Dunn et al. 2007). Variability is high across tundra sites from the European Arctic (Heikkinen et al. 2004), Siberia (Corradi et al. 2005), Alaska (Kwon et al. 2006, Schuur et al. 2009), Greenland (Soegaard et al. 2000, Groendahl et al. 2007), Svalbard (Lloyd 2001), and northern Scandinavia (Aurela et al. 2004, Johansson et al. 2006).

The interannual and across-site variability of CO₂ exchange in arctic tundra ecosystems is driven primarily by growing-season dynamics and moisture conditions. Growing-season rates of CO₂ uptake by tundra ecosystems have been shown in several studies to be closely related to the timing of snow melt, with earlier snowmelt resulting in greater uptake of atmospheric CO₂ (Aurela et al. 2004, Groendahl et al. 2007). The annual C budget is not only controlled by growing-season exchange, but to a large extent by the losses during the shoulder (snow melt/soil thaw and senescence/soil freeze) and winter seasons (Johansson et al. 2006). Pulse emissions during the thaw period (Nordstroem et al. 2001) and long warm autumns after senescence that result in respiratory C losses (Johansson et al. 2006) are examples of shoulder season processes that have been shown important for determining annual budgets. Overall, the studies conducted to date suggest that tundra regions in the Arctic are sources of C to the atmosphere under dry and mesic conditions (e.g., in well-drained settings or in warm and dry years) and are C sinks under wet conditions (e.g., in poorly-drained settings or in cold and wet years).

Given both the spatial and temporal variability and the scarcity of ground-based observations of C exchange in tundra ecosystems of the Arctic, Sitch et al. (2007) evaluated the use of remote-sensing and process-based approaches to estimate whether these ecosystems are acting as sources or sinks of atmospheric CO₂. Analyses based on remote-sensing approaches that use a 20-year data record of satellite data in the late 20th century indicate that tundra is greening in the Arctic, suggesting an increase in photosynthetic activity and net primary production. Process-based modeling studies generally simulate a small net C sink for the distribution of Arctic tundra (e.g., between 3 and 4 Tg C/yr over the last 25 years of the 20th century in McGuire et al. 2000), a result that is within the uncertainty range of field-based estimates of net C exchange across all tundra ecosystems in the Arctic.

The current and future role of boreal forests in the global C budget is also uncertain. The substantial sink in boreal forests, as estimated from the “top-down” approaches described in Atmospheric analyses of carbon dioxide exchange, has been supported by inventory-based studies (e.g., Myneni et al. 2001), with C sequestration in these regions likely due to increased production resulting from enhanced CO₂ uptake, increased nutrient availability and longer growing seasons. On the other hand, some analyses of boreal forest C budgets suggest decreasing production in recent decades as a result of drought stress, nutrient limitation and increased disturbance (see Goetz et al. 2005). “Bottom-up” regional estimates of net CO₂ exchange between boreal forest ecosystems and the atmosphere have been developed by both inventory- and process-based models. While each of these approaches has different strengths and shortcomings, a comparison of estimates (Table 3) does provide some sense for the range of estimates.

Table 3. Comparison of carbon balance estimates for northern high-latitude terrestrial regions among different modeling approaches.

Inventory-based analyses indicate that, in 1990, the Northern European region (Finland, Norway, and Sweden) stored approximately 5 Tg C/yr in soils and 19 Tg C/yr in trees (Liski et al. 2002). Inventory-based analyses also generally show that forests in Russia have been storing C in the late 20th century, with estimates ranging between approximately 300 and 500 Tg C/yr depending on the model and the time period analyzed (Myneni et al. 2001, Shvidenko and Nilsson 2003). Process-based analyses that consider fire as the only disturbance estimate a range of between neutral C exchange to a sink of around 300 Tg C/yr (Balshi et al. 2007), with the range depending on the time period analyzed and whether or not the model applications consider CO₂ fertilization. Temporal analysis of Russian forest inventory data on leaves, wood, and roots during the last half of the 20th century indicates that the fraction of leaves is generally increasing (Lapenis et al. 2005), consistent with the satellite record in the late 20th century (Myneni et al. 2001). However, the fraction of wood and roots in Russia generally decreased in the last half of the 20th century, which suggests that analyses of changes in C storage based on the detection of green biomass by satellites may overestimate the sink of C in Russia (Lapenis et al. 2005).

Models based on forest inventory data indicate that, near the end of the 20th century, Canada was a C sink of between 50 and 100 Tg C/yr (Kurz and Apps 1999, Myneni et al. 2001, Liski et al. 2003). Process-based modeling approaches suggest that the range of estimates is slightly lower (from near neutral to a sink of 80 Tg C/yr; Chen et al. 2000, Balshi et al. 2007). For Alaska, process-based models estimate that terrestrial ecosystems stored between 5 and 12 Tg C/yr in the 1980s and between 0 and 9 Tg C/yr in the 1990s (Balshi et al. 2007). The range of uncertainty using estimated changes in C storage from process-based models depends in part on whether the models implement a response of photosynthesis to increasing atmospheric CO₂ (i.e., CO₂ fertilization; Zhuang et al. 2006, Balshi et al. 2007). While models generally estimate that boreal North America was a sink in the late 20th century, several studies suggest that the sink strength is decreasing and becoming a source as we move from the 20th to the 21st century because of increased disturbance (fire and insects) and drier summers (Kurz and Apps 1999, Goetz et al. 2005, Balshi et al. 2007, Kurz et al. 2008).

The combination of bottom-up estimates for tundra and boreal forest ecosystems of Canada, Alaska, Northern Europe, and Russia suggests that terrestrial ecosystems across the Arctic acted as a sink of 0–600 Tg C/yr near the end of the 20th century. A consideration of only the inventory-based estimates would put the range at between 300 and 600 Tg C/yr. The neutral exchange of the process-based estimates for the Arctic region is associated with an assumption that CO₂ fertilization is not playing a role in C sequestration (Balshi et al. 2007). We accept the range of the inventory estimates (300 and 600 Tg C/yr) as our synthetic estimate of net uptake of CO₂ in the late 20th century for terrestrial ecosystems of the Arctic region and conclude that CO₂ fertilization is likely playing a role in C sequestration of the region.

2. Freshwater lake and river ecosystems

Inland waters are a major landscape feature of the Arctic, with lakes occupying as much as 48% of the land surface in some high latitude regions (Riordan et al. 2006), and with rivers of the pan-arctic drainage basin accounting for approximately 10% of global discharge (Forman et al. 2000). About 36% of the global surface area of lakes is contained in the region north of 45° N latitude, according to the Global Land Cover Characterization database (Loveland et al. 2000). Lakes and rivers play an important and active role in the global C budget as conduits of terrestrial organic matter to the ocean, as sources of both terrestrially and aquatically produced CO₂ to the atmosphere, and as sinks of C stored in sediments and sequestered in aquatic production. While aquatic primary production can contribute to varying degrees, the C balance of freshwater ecosystems is dominated by inputs from land (Kling et al. 1991, Cole et al. 1994). By extrapolating per-area efflux measurements from their study on the North Slope of Alaska, Kling et al. (1991) estimate that CO₂ evasion from arctic lakes and rivers is approximately 20 Tg C/yr for tundra in the Arctic.

A recent analysis by Cole et al. (2007) estimates that, globally, about half of the C input from terrestrial ecosystems is discharged to the ocean, with about 40% lost to CO₂ efflux from the surface and 10% stored in sediments of inland waters. Algesten et al. (2003) found similar trends amongst a series of catchment studies in Scandinavian boreal lakes. With few data available for compiling a quantitative estimate of the evasion of CO₂ from lakes and rivers throughout the Arctic, we down-scaled the global estimates based on the proportion of the global totals for lake area and river discharge represented in the pan-arctic. Of the global CO₂ efflux estimates given in Cole et al. (2007), lakes contributed 70–150 Tg C/yr and rivers an additional 150–300 Tg C/yr. Our first-order estimates for pan-arctic CO₂ evasion from inland waters, then, are between 25 and 54 Tg C/yr from lakes (based on 36% of the global lake area) and between 15 and 30 Tg C/yr from rivers (based on 10% of the global discharge).

There is considerable variability in C cycling processes within the inland waters of the pan-arctic that contributes to the uncertainty of our lake and river CO₂ efflux estimates. In comparison to large lakes, the per-area flux of CO₂ to the atmosphere is generally higher from smaller lakes (Christensen et al. 2007). In general, substantial losses of terrestrial C have been observed from small and medium sized lakes in both the subarctic (Jonsson and Karlsson 2003) and arctic environments (Kling et al. 1991, Eugster et al. 2003). These emissions have been found to have terrestrial organic C as substrates and many of the small lakes are therefore considered net heterotrophic and their C cycling is intimately coupled to terrestrial processes in the areas surrounding them (Kling et al. 1991). When measuring the CO₂ balance of lakes there is a challenge in capturing large emissions that can occur in connection with ice break-up and the release of trapped gases in the spring. In comparison to tundra lakes, per-area surface emissions from lakes in the boreal forest are generally higher (Algesten et al. 2003). Reducing the uncertainty in our estimates of CO₂ efflux from inland waters, then, requires increased sampling and field study of aquatic C cycling processes across lakes and rivers with diverse attributes, in addition to improved mapping of these attributes for extrapolating estimates over the pan-arctic.

Estimates based on ocean observations

While the Arctic Ocean is small compared to the global oceans, the per-area primary production and air–sea fluxes of CO₂ are many times the global average, particularly in inflowing shelf seas like the Barents Sea and Bering-Chukchi Seas. The Pacific waters flowing through Bering Strait follow varied routes within the Arctic Ocean before entering the North Atlantic, making the Arctic Ocean an important transport path between these two major global oceans (Fig. 1). Another process specific to the Arctic Ocean is deep water formation, which transports dissolved organic and inorganic C from the surface ocean to the intermediate and deep waters. This circulation is globally important in sequestering anthropogenic CO₂. Subsurface waters formed in this way constitute the headwaters of the Meridional Overturning Circulation, spreading water to all of the global oceans.

At present, the Arctic Ocean is undergoing rapid changes, including decreasing sea ice coverage (Stroeve et al. 2007), which have great potential to alter processes important to C exchange with the atmosphere. Both cooling and primary production promote a flux of CO₂ from the atmosphere to the surface ocean. However, as much of the central Arctic Ocean surface is covered by sea ice, at least during the winter season, gas exchange is hampered even if it does not necessarily provide a completely impermeable barrier. Also, sea ice melt ponds and open brine channels might act as spring/summer CO₂ sinks (Semiletov et al. 2004). One consequence of the seasonal ice cover is that the surface waters of the central Arctic Ocean are undersaturated with respect to CO₂, which suggests that the loss of CO₂ from surface waters through fixation and associated flux of organic C to deeper water exceeds the net physical flux of CO₂ from air to sea.

Estimates of the air to sea CO₂ flux have been made using budget computations and direct measurements of the difference in partial pressure of CO₂ between the atmosphere and ocean surface and indicate that the uptake of CO₂ by the Arctic Ocean is approximately 24 Tg C/yr (Anderson et al. 1998b). It should be noted, however, that our estimate for CO₂ exchange of the Arctic Ocean is subject to bias because it has not accounted for interannual variability in the sub-domain estimates. Also, the sub-domain estimates are based on data that are incomplete with respect to sampling the variability of exchange in both season and space. The most comprehensive measurement campaign of the surface water pCO₂ field was performed in 2002 and 2004 within the Shelf-Basin Initiative project (Bates 2006, Bates et al. 2006). The data show large variability in surface water pCO₂ in both time and space. The mean computed air to sea fluxes were in the range 36 to 61 mmol·m⁻²·d⁻¹ for the summer and fall investigations. The net annual air–sea flux estimated for the Chukchi Sea shelf from these data is 31–45 Tg C/yr (Bates 2006). It should be noted that spatial coverage of the Shelf-Basin Initiative project is limited with much of the data derived from the Alaska coastal region and the continental slope, and thus the uptake of deeper water may be overestimated. Nevertheless, the estimate of Bates (2006) suggests that the estimate of 24 Tg C/yr for the mean annual uptake of CO₂ by the Arctic Ocean by Anderson et al. (1998b) could be substantially low because of seasonal, interannual, and spatial sampling/scaling issues. Given the estimates of Anderson et al. (1998b) and Bates (2006), we infer that the mean annual sink for atmospheric CO₂ of the Arctic Ocean and its associated shelf seas lies between 24 and ∼100 Tg C/yr.

Part of the CO₂ taken up by the surface ocean is transported to intermediate and deep waters by density increase due to cooling and/or the addition of salt from brine drainage during sea ice production. However, the steady state vertical transport of DIC or DOC is not as remarkable as the non-steady state anthropogenic contribution, which is equivalent to the differential amount that DIC has increased through equilibrating with an atmosphere of higher pCO₂ from the burning of fossil fuel. The sink of anthropogenic CO₂ was computed for the Arctic Ocean and its adjacent subarctic seas in 1991 to be 26 Tg C/yr by applying a plume entrainment model to measured data, with the inventory being 1350 Tg C (Anderson et al. 1998a). The Greenland Sea (one of the major deep water formation areas) sink of anthropogenic CO₂ was computed to be 2.4 Tg C/yr for the year 1994 by a mixing box model constrained by CFCs (Anderson et al. 2000). None of these numbers are significant in a global context, but they are significantly larger when expressed per unit area. Furthermore, the processes behind the deep water formation are sensitive to environmental change, which potentially could considerably change these sinks. For example, the uptake of CO₂ is potentially sensitive to changes in sea ice because the addition of brine from growing sea ice to the seawater provides an efficient means to transport CO₂ enriched surface water to depth (Anderson et al. 2004).

Atmospheric analyses

Estimates of CH₄ emissions can be calculated with the top-down method from atmospheric measurements obtained by the global observation network (see Atmospheric analyses of carbon dioxide exchange), subject to the same limitations as discussed above for CO₂. Furthermore, most inversion studies involving CH₄ require specified a priori wetland emissions estimates derived from bottom-up approaches (e.g., Matthews and Fung 1991), which must be scaled down substantially in order to be compatible with the atmospheric concentrations (e.g., Bergamaschi et al. 2007). In addition to an a priori emission pattern, the inversion modeling system also needs to include the atmospheric sink of CH₄ by oxidation with the OH radical. This is done by prescribing OH fields from atmospheric chemistry model simulations (e.g., Spivakovsky et al. 2000) that may be also optimized in the inversion. Alternatively, in a pre-inversion step, the OH fields may be independently optimized to match the global sink of methylchloroform (Bousquet et al. 2005). Space-borne remote-sensing methods have recently become a promising tool for improved mapping of the global atmospheric CH₄ distribution (Frankenberg et al. 2005). Even though the satellite instruments measure only the column-integrated CH₄ content in cloud-free areas with less accuracy than in situ measurements, the large amount of observations and the global coverage are distinct advantages, and the observations can be used in a top-down inversion system (Bergamaschi et al. 2007).

Recent inversion studies (Mikaloff Fletcher et al. 2004a, b, Bergamaschi et al. 2005, 2007, Chen and Prinn 2006) indicate that the Arctic is a source of CH₄ to the atmosphere of between 15 and 50 Tg CH₄/yr (Table 4). Atmospheric inversion studies indicate that there is substantial inter-annual variability of CH₄ sources (Bousquet et al. 2006), with the interannual variability in the Arctic associated with climate variability. Bousquet et al. (2006) attribute a significant fraction of the recently observed slow-down in the global growth rate of CH₄ in northern latitudes to a reduction of CH₄ emissions from wetlands caused by a drying trend after 1999.

Table 4. CH₄ emissions (Tg CH₄/yr) from the Arctic region as estimated by recent atmospheric inversion studies.

Estimates based on terrestrial observations excluding gas hydrates

Soils of terrestrial ecosystems have the capacity to both produce and consume CH₄. Recent studies estimate that CH₄ emissions from the world's soils range from 150 to 250 Tg CH₄/yr (Prather et al. 2001), with one-quarter to one-third of the total emitted from the wet soils of high latitudes (Walter et al. 2001a, b). Estimates of global CH₄ consumption by soil microbes are in the range of 10–30 Tg CH₄/yr (Prather et al. 2001), an order of magnitude lower than the emission estimates. Most of the CH₄ consumption occurs in the well-drained soils of temperate and tropical areas (Ridgwell et al. 1999). A wide range of observational studies has been conducted to quantify the rates of and controls on CH₄ emissions from wet and mesic tundra ecosystems (e.g., Whalen and Reeburgh 1992, Christensen 1993, Christensen et al. 1995, 2000, Friborg et al. 2000, Corradi et al. 2005) as well as uptake by dry tundra ecosystems (Whalen and Reeburgh 1990a, Christensen et al. 1999). Earlier studies, only a few of which have been conducted in tundra regions, were reviewed by Bartlett and Harriss (1993). The process-level understanding gained through these studies has lead to a range of modeling efforts to investigate the effects of changing climate on CH₄ emissions from tundra ecosystems (Cao et al. 1996, Walter and Heimann 2000, Zhuang et al. 2004, Sitch et al. 2007).

Emission estimates of CH₄ from northern high latitude terrestrial ecosystems in the late 20th century range from 31 to 106 Tg CH₄/yr (Table 5), but estimates published since 1998 range between 31 and 65 Tg CH₄/yr. Consumption estimates are much less, and range from 0 to 15 Tg CH₄/yr. The combination of the consumption and emissions estimates published since 1998 results in a range of 16–65 Tg CH₄/yr for net CH₄ emissions. A process-modeling study by Zhuang et al. (2004) places the annual net emission rate at the end of the century for the region north of 45° N at 51 Tg CH₄/yr.

Table 5. Estimates of emissions and consumption of methane from soils in northern high latitudes during the late 20th century.

The atmospheric fluxes of CH₄ from lakes may have been under-represented in early attempts to estimate global CH₄ emissions (Matthews and Fung 1987). It has recently been estimated that thermokarst lakes are responsible for substantial CH₄ emissions (Walter et al. 2006). These subarctic and arctic lake systems in Siberia and Alaska are surrounded by permafrost soils that are rich in C accumulated during the Pleistocene. A broad lake survey by Bastviken et al. (2004) has also reported significant CH₄ emissions from boreal, subarctic, and arctic lakes, with smaller lakes generally having much higher emissions than larger lakes. Thus, the scaling-up of CH₄ measurements from lakes requires differentiation between (1) small and large lakes and (2) the presence or absence of permafrost (Bastviken et al. 2004, Walter et al. 2006). Few studies have attempted to scale-up lake CH₄ fluxes for northern high latitudes, but recently Walter et al. (2007), using data from Siberia, Alaska, and the literature, estimated that lakes in this region (excluding large lakes with likely small emissions) emit 15–35 Tg CH₄/yr. If there has been no double-accounting between the emissions estimates of Walter et al. (2007) and the emissions estimates in Table 5 that have been published since 1998, the bottom-up estimates of net CH₄ emissions for the Arctic range between 31 and 100 Tg CH₄/yr.

Emissions from CH₄ exchange associated with CH₄ hydrate

CH₄ hydrate can occur both below and within permanently frozen soil. Permafrost has low permeability due to the presence of ice within the sediment pores and thus impedes the vertical migration of free gas. However, this permeability is temperature dependent and increases up to five orders of magnitude as permafrost warms to 0°C (Williams 1982). Thermal dissociation of gas hydrate associated with permafrost is thought to be widespread along the Arctic coastline where the continental shelf has been warmed by ocean transgression (Kvenvolden et al. 1993a, Paull et al. 2007). In this case there has been significant time to dissociate some gas hydrate and increase the permeability of the permafrost interval by warming. Thermal modeling has illustrated that, in comparison to continental slope hydrates, gas hydrate in terrestrial permafrost responds much slower to climatic warming. Some exceptions do occur where thawed zones, or taliks, exist beneath large lakes or river channels. In some of these settings, e.g., in the Mackenzie Delta of Canada, vigorous gas seepage is known to occur. It is speculated that these features are connected to deeper gas sources via faults or gas-saturated groundwater advection. To our knowledge, there is no estimate of CH₄ emissions associated with the dissociation of gas hydrate in or below subterranean permafrost.

Wuebbles and Hayhoe (2002) have estimated that global CH₄ emissions associated with the dissociation of gas hydrate are 5 Tg CH₄/yr, with an uncertainty range between 0.4 and 12.2 Tg CH₄/yr. We use the ratio of the area of continuous permafrost (approximately 10 × 10⁶ km²; Zhang et al. 1999) to the area of global ocean plus continuous permafrost ([375 + 10] × 10⁶ km²) to develop the first-order estimate of between 0 and 0.3 Tg CH₄/yr emitted from the dissociation of gas hydrate associated with subterranean permafrost in the Arctic. This range does not affect our overall estimate of between 31 and 100 Tg CH₄/yr emitted from terrestrial regions of the Arctic. Similarly, a prorating of the global 14 Tg CH₄/yr emissions of Wuebbles and Hayhoe (2002) from other natural geological sources, such as volcanoes, to the Arctic would not affect our overall estimate of terrestrial CH₄ emissions.

Emissions based on ocean observations

Establishing CH₄ fluxes for the Arctic Ocean is difficult because the very few available observations are characterized by high seasonal and spatial variability. Sources of CH₄ to the Arctic Ocean include inflow from the Pacific Ocean, diffusion of microbially produced CH₄ in near-surface sediments, bubble seeps from deep sediments, riverine inflow, and in situ production in particulate organic matter. Methane leaves the Arctic Ocean system via the archipelago outflow through net exchange with the atmosphere and by oxidation within the ocean. Exchange with the Atlantic Ocean via Fram Strait and the Barents Sea may provide a net source or sink for Arctic Ocean CH₄. Many of these processes are presently difficult to quantify due to a lack of data; for example, oxidation within the ocean is presently un-quantified but may be significant (Kvenvolden et al. 1993a, Damm et al. 2005), and the bubbling of CH₄ proves an elusive process to quantify.

Several estimates have been made for CH₄ flux from Arctic shelf waters to the atmosphere, including for the Alaskan shelf (∼0.016 g/m²; Kvenvolden et al. 1993a) and for the Laptev and East Siberian shelves (1.0–2.6 × 10⁴ g/km²; Shakhova et al. 2005, Shakhova and Semiletov 2007). Prorating these fluxes to the entire Arctic shelf (5 × 10⁶ km²) implies a source of 0.08–0.13 Tg CH₄/yr. The flux to the atmosphere, which does not include potentially higher fluxes associated with river plumes or polynyas, must at least be supported by a corresponding seabed flux into the bottom water. Damm et al. (2007) recently provided an Arctic-wide estimate for CH₄ evasion from polynyas of 0.005–0.020 Tg/yr, which does not affect the estimated range that we present here. Methane bubbles, which bypass flux estimates based on concentration gradients, are likely an important vertical transport mechanism in the shelf water of the Arctic Ocean (e.g., Shakhova et al. 2005, Paull et al. 2007), but perhaps less so in deeper water where dissolution feeds the deep-ocean reservoir (Damm and Budéus 2003). Hovland et al. (1993) have set the global seepage of CH₄ from continental shelf sediments (27.4 × 10⁶ km²) at 1–50 Tg CH₄/yr, which would imply an Arctic shelf flux to the atmosphere of 1–12 Tg CH₄/yr. It is possible that the discrepancy between this number and measured flux estimates can be explained by bubble transport. We assume that this estimate encompasses CH₄ emissions associated with the dissociation of gas hydrate throughout the Arctic Ocean and associated shelf seas.

Exchange between the Arctic and adjacent oceanic basins

In estimating the exchange of DIC, DOC, POC, and CH₄ of the Arctic Ocean with the Atlantic and Pacific Oceans, we assume that the exchange is approximately balanced. That is, the (1) inflow from the North Atlantic through the Nordic Seas is offset by southward flows through North Atlantic Deep Water and the surface water of the East Greenland Current and (2) inflow through the Bering Strait is balanced by the southward flow through the Canadian Arctic Archipelago. On the order of 8 Sv (1 Sv = 10⁶ m³/s) of warm, saline water flows into the Nordic Seas from the North Atlantic over the Greenland Scotland Ridge (Hansen and Østerhus 2000). A similar volume is transported south into the North Atlantic through a bottom current of about 6 Sv and through a surface flow of about 2 Sv in the East Greenland Current. The Central Nordic Seas act as a mixing point for waters of Atlantic and Arctic origin where some of the warm Atlantic water mixes with the south flowing East Greenland Current before making it into the Arctic Ocean. Of the water entering the Nordic Seas, about 2 Sv continues into the Arctic Ocean over the Barents Sea, while a large volume flux passes north through Fram Strait. This latter flux has been reported to be on the order of 10 Sv (Schauer et al. 2004), which means that there must be a substantial recirculation within the Nordic Seas. Water also enters the Arctic Ocean from the Pacific Ocean through the Bering Strait, at a mean volume flux of 0.8 Sv (Roach et al. 1995, Woodgate et al. 2005). Roughly an equal amount has been suggested to exit the Arctic Ocean through the Canadian Arctic Archipelago (Prinsenberg and Hamilton 2005).

The DIC concentration in seawater is fairly high (typically above 2000 μmol/kg H₂O). The approximately 8 Sv that enters the Nordic Seas from the Atlantic Ocean is responsible for a DIC transport of about 1 Pg C/yr. The in-flux through Bering Strait is estimated to be approximately an order of magnitude lower (0.1 Pg C/yr). DOC concentrations from inflowing Atlantic water have been measured in the range of 52–75 μmol/L (Børsheim and Mycklestad 1997, Wheeler et al. 1997, Opsahl et al. 1999, Fransson et al. 2001). Multiplying this average concentration with the 8-Sv flux gives a DOC inflow of between 21 and 30 Tg C/yr. The DOC concentration in the waters passing the Bering Strait is more variable, with values reported in the range of 34–134 μmol/L H₂O (Walsh et al. 1997, Wheeler et al. 1997). Based on a water transport of 0.8 Sv, the influx of DOC through the Bering Strait is estimated to be between 1 and 5 Tg C/yr. The exchange of POC between the Arctic and adjacent ocean basins is likely several orders of magnitude less than the exchange of DOC, and can be considered negligible.

A reasonable range of estimates for CH₄ concentration of Pacific inflowing water through the Bering Strait is 4 to 20 nmol/L, which implies a flux into the Arctic of 2–8 Gg CH₄/yr using a volume transport of 0.8 Sv (Roach et al. 1995). There are no data for quantifying the exchange of CH₄ between the North Atlantic and Arctic Ocean, but assuming that it is approximately an order of magnitude higher than the Bering Strait exchange implies an exchange of between 20 and 80 Gg CH₄/yr.

Riverine fluxes

There has been substantial progress in the last few years to improve the database for estimating the export fluxes of DOC, DIC, and POC to the Arctic Ocean (Guo and Macdonald 2006). The current understanding of export fluxes for Arctic rivers is summarized in Table 6. The total freshwater discharge of the pan-arctic drainage basin (16 × 10⁶ km²) is approximately 3500 km³/yr. Based on this water flux and a discharge-weighted average DOC concentration of 9.4 ppm, we estimate the total export fluxes of C from rivers to the Arctic Ocean to be 43 Tg C/yr as DIC, approximately 33 Tg C/yr as DOC, and about 6 Tg C/yr as POC. Black carbon, a result of the incomplete combustion of either fossil fuel or biomass, is 0.03–0.10% of POC in the seven largest rivers of the Arctic (Guo et al. 2004, Elmquist et al. 2008). This implies that the flux of black carbon, which is very difficult for microbes to decompose, to the Arctic Ocean is between 0.2 and 0.6 Tg C/yr. These fluxes are likely overestimated for DIC and underestimated for DOC and POC because of the disproportional partitioning in discharge and DOC concentration (Finlay et al. 2006, Raymond et al. 2007, Cai et al. 2008). Arctic rivers have a high seasonal variability in discharge, with the concentration of DIC decreasing and DOC and POC increasing during high flows (e.g., from melt water in the spring). Thus, inadequate sampling during peak flows of the spring river ice breakup have tended to overestimate DIC flux and to underestimate discharge-weighed DOC and POC flux (by between 20% and 60%; Finlay et al. 2006, Guo et al. 2007, Raymond et al. 2007, Cai et al. 2008).

Table 6. Carbon fluxes from rivers to the Arctic Ocean.

Carbon can also enter the Arctic Ocean as particulate inorganic carbon (PIC) and CH₄, but there are very few data on either PIC or CH₄ concentrations in river water. Recent studies of the Yukon River have documented a PIC/DIC flux ratio of 0.10 and a PIC/POC ratio of 0.67 (Striegl et al. 2007), which would imply a PIC flux of 3 to 4 Tg C/yr if these ratios were applied throughout the pan-arctic drainage basin. However, it is likely that PIC export is much lower since the PIC/POC ratio of the Yukon River is probably only representative of the MacKenzie River, which dominates the sediment budget of the Arctic Ocean with a sediment load that is largely derived from limestone rich mountains. With respect to CH₄, it is likely that rivers are at or above equilibrium with the atmosphere and that higher concentrations are obtained during ice cover. Assuming river concentrations of 10–200 nmol/L (Semiletov et al. 1996, Galimiov et al. 2006), together with a total river inflow of 3253 km³/yr (Serreze et al. 2006), implies an input of 1 to 11 Gg CH₄/yr to the Arctic Ocean from rivers. River plumes may be locally important for CH₄ surface concentrations (e.g., see Shakhova et al. 2005, Shakhova and Semiletov 2007), but they contribute little to the Arctic Ocean budget.

Coastal and atmospheric POC inputs

Coastal erosion in the Arctic is an important source of POC to the Arctic Ocean, and this flux is likely to increase with warming because of enhanced erosion associated with the loss of a protective sea ice buffer, increasing storm activity, and thawing of coastal permafrost (e.g., Rachold et al. 2000, 2004, Jorgenson and Brown 2005). Based on recent estimates by Rachold et al. (2004), POC transport across the Arctic land/ocean interface through coastal erosion is in the order of 6–7 Tg C/yr (Table 7). POC fluxes from coastal erosion are equivalent to riverine POC fluxes, but much smaller than riverine DOC fluxes. The total supply of POC to Arctic shelves from wind erosion, which has been estimated to be approximately 2 Tg C/yr by Stein and Macdonald (2004), is highly uncertain. Similarly, fluxes of atmospheric black carbon soot from boreal wildfires and fossil fuel combustion are highly uncertain. In this review we also do not explicitly consider the wind driven exchanges of POC like black carbon between the Arctic and regions outside the Arctic. Although black carbon deposition is most likely a quantitatively minor component of total C transport across the land/ocean interface (Kim et al. 2005, Elmquist et al. 2008), its deposition on high albedo surfaces like snow and sea ice has potentially important consequences for energy exchange in the Arctic (Hansen and Nazarenko 2004).

Table 7. Comparisons of carbon fluxes (Tg C/yr) from terrestrial to marine regions in the Arctic through rivers emptying into oceans or through coastal erosion.

Transport and fate of terrestrially derived carbon

The fate of terrestrial organic C during its transport in arctic rivers and within the Arctic Ocean remains uncertain and largely unknown (Stein and Macdonald 2004). One potential pathway could be the direct mineralization of river DOC to DIC, resulting in a CO₂ flux across the river/atmosphere interface (e.g., Kling et al. 1991, Mayorga et al. 2005). Laboratory incubation studies have identified that a significant fraction of soil DOC can be degraded (Neff and Hooper 2002, Cleveland et al. 2004). However, the contribution of riverine DOC to CO₂ out-gassing in river systems of the Arctic has not been well quantified, and our estimate of 15–30 Tg C/yr is based on some very simple assumptions (see Freshwater lake and river ecosystems).

Recent measurements of ¹⁴C ages for organic C in rivers collectively show that DOC tends to be contemporary, while POC is older (Guo and Macdonald 2006). These observations suggest that changes in climate will have different effects on the sources of DOC and POC, with DOC responding to how changes affect the release of soil organic C of recent origin and POC responding to effects on the release of older soil organic C that is currently stored in permafrost (Guo et al. 2007). Based on the assumption that old organic C is mostly derived from river bank soils, Guo and Macdonald (2006) estimated that 78% of Yukon River POC came from old soil organic C, whereas all high molecular weight DOC, which comprises about 70% of bulk DOC, was derived from modern terrestrial organic C.

DOC exhibits conservative mixing in Arctic estuaries and coastal regions (Cauwet and Sidorov 1996, Dittmar and Kattner 2003, Guéguen et al. 2005, 2007), which suggests little removal during initial transport. However, DOC is not likely to be conserved during its transport through the Arctic Ocean. Hansel et al. (2004) estimate a half-life of 7.1 years for terrestrial DOC in the Arctic Ocean, with microbial activity and reactions with light the most probable causes of degradation.

The fate of terrestrially derived POC in the Arctic Ocean is less clear. Although the total terrestrial supply of POC to the Arctic Ocean is much lower than total marine production, this source can be important near points of entry (river mouths and shorelines; Telang et al. 1991, Naidu et al. 2000, Macdonald et al. 2004). For example, at the Beaufort Shelf edge, a location in proximity to terrestrial POC sources, settling particles contained 15–35% C of terrestrial origin (O'Brien et al. 2006). Stein and Macdonald (2004) estimate that shelf and basin sediments capture 8.7 Tg C/yr and 2.4 Tg C/yr, respectively, of ocean water column POC. A further breakdown assigns 36% of the shelf burial as marine-derived POC compared to 20% in the basin, which suggests that the sediments of the Arctic Ocean have a stronger imprint of terrestrial POC than sediments of other oceans.

Response of feedbacks pathways in terrestrial ecosystems of the Arctic

Processes in terrestrial regions of the Arctic that are sensitive to change on a 10–20 year time frame are those that are primarily sensitive to changes in atmospheric variables (e.g., temperature, precipitation, CO₂ concentration), and include photosynthesis (feedback pathways 2, 5, 6, and 7 in Fig. 4) and fire (feedback pathway 8 of Fig. 4). The net direction of the photosynthesis and fire feedbacks depends substantially on landscape wetness and dryness. For example, dry conditions may decrease photosynthesis more so than it is promoted by longer growing seasons, and may also result in the release of C by triggering an increase in fire activity across boreal forest regions. The analysis of Thompson et al. (2006) indicates that the trends in C storage in Alaskan tundra from 1981–2000 were positive under colder and wetter conditions and negative under drier or warmer conditions. The studies by Balshi et al. (2009a, b) suggest that climate warming in boreal North America is likely to promote higher fire frequency, resulting in the release of C to the atmosphere from terrestrial ecosystems of the Arctic. However, whether the land acts as source or sink of C to the atmosphere depends on whether photosynthesis in undisturbed boreal forests responds to increases in atmospheric CO₂ (Balshi et al. 2009b). The uncertainty about the response of vegetation photosynthesis to increases in atmospheric CO₂ is the largest single factor contributing to the uncertainty in process-based model analyses of how C storage in Arctic ecosystems will respond to global change. Another large source of uncertainty stems from unknown potential responses of disturbance regimes (e.g., insects, forest harvest, and forest degradation; feedback pathways 11 and 13 in Fig. 4) to changes in climate. Evidence suggests possible increases in the frequency of fires (Kasischke and Turetsky 2006, Soja et al. 2007) and insect outbreaks (Kurz et al. 2008), which could create a large source of CO₂ to the atmosphere from boreal forest regions.

The 50–100-year time frame involves processes that respond more slowly to climate than the short-term processes described above. For terrestrial ecosystems, these include slow ecological processes and responses associated with the thawing of permafrost. Ecological processes that might respond slowly to warming include an increase in shrub tundra, changes in tree species, treeline advance, and forest degradation (feedback pathways 9 and 10 in Fig. 4). The expansion of shrubs is best documented for the North Slope of Alaska, where shrub cover has increased by about 16% in land area since 1950 (Tape et al. 2006), and shrub growth appears to be increasing throughout the tundra of much of the Arctic (Kaplan et al. 2003). While treeline advances into tundra ecosystems have been documented during recent decades in Russia, Canada, and Alaska (McGuire et al. 2007), the rate of advance is very slow and most likely to affect landscape C storage on the order of 500–1000 years (Starfield and Chapin 1996, Chapin and Starfield 1997, Lloyd et al. 2002).

Changes in the distribution of tree species within the boreal forest are also possible in the next 50–100 years, with boreal broadleaf deciduous species likely to become more widely distributed if disturbance frequency increases (Chapin et al. 2000). While a number of dynamic global vegetation models indicate that temperate species are likely to move northward in response to a warming climate during the next 50–100 years (Cramer et al. 2001), these models generally do not include processes that would be responsible for migration of temperate species into boreal regions. Similar to the northward expansion of treeline, the latitudinal migration of temperate forest species into the southern boreal forest may be a dynamic operating at longer time scales (i.e., 500–1000 years). This process has been found to occur on shorter time scales along elevational gradients, however, and may be facilitated by more frequent or severe disturbance (Beckage et al. 2008). Regardless of the rate at which migration occurs, these changes in species distribution are expected to decrease C storage as boreal broadleaf deciduous and temperate forest ecosystems generally have smaller ecosystem C stocks than the boreal conifer forest ecosystems that they would replace (e.g., see Smith and Shugart 1993).

There is a great deal of uncertainty on the future rate of permafrost degradation in response to climate warming. One argument is that thaw is likely to proceed at a slow pace during the next century because the thermal regime of permafrost is buffered by the overlying, protective layer of organic matter (Yi et al. 2007). The simulations of Lawrence and Slater (2005), which did not take into consideration the protective layer of organic matter, indicate that substantial permafrost thaw could happen in the first half of the 21st century. However, when the protective layer of organic matter is taken into consideration, the degradation of permafrost is somewhat slower with most of the loss occurring by around 2070 (Lawrence et al. 2008). The simulations of Euskirchen et al. (2006), which also considered the protective layer of organic matter, estimated that permafrost loss in the first 40–50 years of the 21st century would be minimal, and primarily located in regions characterized by a discontinuous permafrost regime at the transition zone between boreal and temperate forest ecosystems (Fig. 6). Although Euskirchen et al. (2006), Lawrence and Slater (2008), and Lawrence and Slater (2005) disagree on the temporal pattern, the analyses each estimate substantial loss of near surface permafrost by 2100 (between 6 and 11 × 10⁶ km² among the analyses). Thus, while these analyses identify the potential for the mobilization of C stocks associated with permafrost thaw during this century, it is not clear at what time scale this dynamic will operate. Furthermore, none of these studies considered the potential effects of fire, which is a disturbance agent that has the potential for the rapid release large stocks of C to the atmosphere. This release can occur not only in the direct and immediate combustion of organic matter, but also by exposing the remaining C stocks in the soil to substantial warming and decomposition through removal of the protective organic layer above the permafrost.

Once permafrost thaws, the direction of feedbacks to the climate system depends largely on landscape wetness and dryness. An increase in landscape wetness will tend to promote increased CH₄ emissions through enhanced methanogenesis (feedback 3 in Fig. 4) while slowing CO₂ release because of anaerobic conditions that hinder decomposition. In contrast, an increase in landscape dryness can promote greater release of CO₂ through enhanced decomposition (feedback 1 in Fig. 4). A number of analyses are indicating that landscape dryness is increasing over the last several decades in the boreal forest, particularly in areas of discontinuous permafrost (Smith et al. 2005, Riordan et al. 2006, Bunn et al. 2007, Goetz et al. 2007). Large releases of CO₂ are possible if landscape drying were to become pervasive in areas of continuous permafrost as ponds and wetlands comprise a substantial part of the landscape (Smol and Douglas 2007).

Response of feedbacks pathways in marine ecosystems of the Arctic

For marine systems, processes sensitive to changes in surface conditions, such as sea ice cover and near surface water temperature, could have substantial responses to climate change on the 10–20 year time frame. Similarly, changes in river discharge and exchanges with surrounding oceans have a direct, near-term impact on the marine ecosystem through its effect on CO₂ fixation. The loss of sea ice will support an increasing air–sea flux of CO₂ that is otherwise impeded by the ice cover, along with triggering an increase in the input of light and atmospheric mixing energy to the marine system. On the other hand, the melting of sea ice will also add freshwater to the system, which will lower salinity and tend to stabilize the surface layer. As such, a decline in sea ice cover as a result of rising temperature can have both positive and negative feedbacks with respect to the C cycle of the Arctic: ocean CO₂ sequestration can be increased by amplifying the physical transfer of DIC to the surface layer (feedback 1 in Fig. 5) and biological uptake of CO₂ through added light and nutrient availability (feedback 5 in Fig. 5), while CO₂ uptake can decrease through changes in surface layer chemistry that lead to the creation of a stable photic zone (feedback 6 in Fig. 5). The balance of these competing feedbacks and their overall effect on the uptake of CO₂ into the marine system is not clearly understood, as the seasonal timing of these processes is crucial to the net result. Also, ocean acidification associated with increases in atmospheric CO₂ may modify these feedbacks by affecting inorganic and biotic C dynamics in the ocean (Guinotte and Fabry 2008).

The key process for enhancing nutrient availability in the Arctic Ocean is an increase in nutrient input from the surrounding seas. This process will have a direct impact on primary production and the draw down of CO₂ from the atmosphere. The uptake of CO₂ by marine systems is also driven by decreases in both water temperature and salinity. With a lack of sea ice in the summer, increased water temperatures would decrease the equilibrium concentration of DIC because of decreased CO₂ solubility (feedback 2 in Fig. 5). While a lower salinity with a decrease in sea ice will tend to counteract this effect (feedback 1 in Fig. 5), the effect of increasing temperature on decreasing solubility is likely to be stronger. Additionally, an increase in water temperature has the potential to promote the release of CO₂ and CH₄ through enhanced decomposition and methanogenesis of organic C in the water column (feedback 7 in Fig. 5).

The discharge of freshwater from the terrestrial surface to the marine system has increased in the pan-arctic drainage basin during the 20th century (Peterson et al. 2002), and the projected increase in precipitation in the Arctic suggests that this trend will continue or accelerate through the 21st century. It is generally thought that C delivery to the ocean will be correlated to discharge, although it is not clear whether the distribution of C as DIC, PIC, DOC, and POC will change (Striegl et al. 2007). In drainage systems dominated by discontinuous permafrost, it is expected that ground water flow will increase along with in-stream processing of DOC, resulting in greater riverine CO₂ efflux as well as DIC delivery to the ocean (Striegl et al. 2007). Responses of riverine biogeochemistry to warming will occur as a result of changes in organic composition, chemical/phase speciation, and C yields from rivers, and these changes will manifest differently in the particulate and dissolved components undergoing transport (Guo et al. 2007). Changes in permafrost dynamics, hydrological cycles, vegetation, and microbial processes have the potential to play an important role in responses of riverine C dynamics. If the delivery of organic C from land to ocean increases in response to a warming Arctic climate, then there is the potential to increase C sequestration through burial in coastal sediments (feedback 3 in Fig. 5). However, increased delivery of organic C to the ocean is likely to lead to potential C release through enhanced decomposition and methanogenesis in the water column (feedback 7 in Fig. 5).

Release of CH₄ from hydrates in terrestrial and marine regions of the Arctic

The release of free inorganic CO₂ and CH₄ frozen in terrestrial soils and marine sediments, in addition to the dissolution of CH₄ from gas hydrates (feedback pathways 14 in Fig. 4 and 4 in Fig. 5) as a result of permafrost thaw, are expected to proceed at a very slow pace. While there is uncertainty about the degree to which near surface permafrost will thaw, the thawing at greater depths due to the transfer of heat from the overlying atmosphere is likely to be a millennial-scale response. This disappearance of thick permafrost is most likely to occur in settings where the ice content is high and its vertical structure is exposed to the atmosphere and erosional runoff (e.g., along river banks).

We have estimated that between 5 and 195 Pg CH₄ exists as gas hydrate within and below subterranean and submerged permafrost of the Arctic. A doubling of atmospheric CH₄ concentration would require the release of an additional 50 Pg CH₄ from all sources over the next 100 years (Archer 2007). Thus, it is conceivable that substantial release of CH₄ from the melting of gas hydrate could influence the radiative forcing of climate. The northern coast of Siberia has been eroding for thousands of years and it is thought that this erosion has exposed hydrates to melting (Romankevich 1984). However, the erosion mechanism is not sufficient to support the release of a large amount of CH₄ from the melting of gas hydrate, which would require increased melting at depth (Corell et al. 2008). The mechanism for this enhanced release has not been elucidated and it is unlikely that CH₄ would reach the atmosphere through 0.2–1.2 km thick sections of sediment without significant oxidation (Corell et al. 2008).