This study describes an analysis of the contemporary C balance of the Arctic system in which the land and ocean area of the Arctic Basin (Fig. 1a) is treated as a linked system of CO2 and CH4 exchange across terrestrial, marine and atmospheric components. The study area for the terrestrial component of the Arctic Basin is defined as the land area within the watersheds of the major rivers that drain into the Arctic Ocean (Lammers et al., 2001). This hydrologic perspective allows the examination of the linkage between the terrestrial and marine components of the Arctic C cycle. The watersheds of the Arctic Basin contain most of the northern high-latitude land underlain with near-surface permafrost (the cryospheric perspective) and encompass large expanses of arctic tundra and boreal forest ecosystems (the ecological perspective; Fig. 1b). The marine component of the study includes the area north of 65°N, which covers the whole of the Arctic Ocean proper as well as a portion of its marginal seas.
Figure 1. Maps showing (a) the delineation of the terrestrial watersheds of the Arctic Basin and (b) the spatial distribution of the major ecosystem types across these watersheds.
Download figure to PowerPoint
Several process-based tools were used to conduct this analysis of C dynamics across the Arctic Basin between years 1997 and 2006 through simulations of land–atmosphere CO2 and CH4 exchange, the transfer of land-based C to the Arctic Ocean, and ocean–atmosphere CO2 exchange. CO2 and CH4 exchange between the terrestrial ecosystems of the basin and the atmosphere, along with the export of dissolved organic C (DOC) to the Arctic Ocean, were estimated using the Terrestrial Ecosystem Model (TEM), a process-based biogeochemistry model with coupled carbon and nitrogen cycles (Raich et al., 1991). The TEM considers the effects of a number of factors on its simulations of C dynamics including changes in atmospheric CO2 concentration, tropospheric ozone pollution, nitrogen deposition, climate variability and change and disturbance/land use including fire, forest harvest and agricultural establishment and abandonment. TEM also calculates pyrogenic emissions of CO2, CH4 and CO from the combustion of vegetation and soil carbon in wildfires. The DOC leaching dynamics of TEM are a function of soil C decomposition rate, soil DOC concentration and water flux through the soil. We used the methane dynamics module of TEM (MDM-TEM) to estimate the exchange of CH4 with the atmosphere of both wetlands, which generally emit CH4 to the atmosphere and uplands, which generally consume CH4 from the atmosphere. The MDM-TEM considers the effects of a number of factors in its simulations of CH4 dynamics including the area of wetlands, fluctuations in the water table of wetlands, temperature and labile carbon inputs into the soil solution derived from the net primary production (NPP) estimates of TEM. The MIT ocean biogeochemistry model simulated the net exchange of CO2 with the atmosphere as driven by changes in sea ice, water temperature, ocean circulation and DOC inputs from TEM.
The results of these simulations were compared with estimates of CO2 and CH4 exchange from atmospheric inversion models and with observations of terrestrial C export from Arctic watersheds. The simulated transfer of land-based C to the Arctic Ocean was compared against estimates based on a sampling of DOC export from major Arctic rivers (McClelland et al., 2008). The land–atmosphere CO2 exchange estimate was compared with results from the TransCom 3 atmospheric inversion model intercomparison project (Gurney et al., 2008), and CH4 to results from atmospheric inversion-estimated surface emissions (Chen and Prinn, 2006). To compare the ‘bottom-up’ results from our model simulations with the ‘top-down’ estimates from these inversion studies, we summarize our estimates of surface-atmosphere CO2 and CH4 exchange for the land and ocean area matching the three high-latitude regions defined in the TransCom 3 model experiments (Gurney et al., 2002), namely the Boreal North America, Boreal Asia and Northern Ocean regions.
2.2. Terrestrial CO2 fluxes and pyrogenic CH4 and CO emissions
We used the TEM as a ‘bottom-up’ approach to estimate the exchanges of CO2 with the atmosphere from terrestrial ecosystems of the Arctic Basin. For this study, we used a version of the model (TEM6) that has been modified from Felzer et al. (2004), which simulated ozone pollution effects, to also include the influence of permafrost dynamics (Zhuang et al., 2003; Euskirchen et al., 2006), atmospheric nitrogen deposition, biological nitrogen fixation, DOC leaching, wildfire (Balshi et al., 2007), agricultural conversion and abandonment and timber harvest on terrestrial C dynamics. C pools and associated fluxes are simulated at a monthly time-step for individual ‘cohorts’ of unique vegetation types and disturbance history organized within spatially explicit 0.5° latitude × 0.5° longitude grid cells. To initialize the C, N and water pools for the beginning of the analysis period (1997–2006), in each model run we simulated dynamics since the year 1000 for each cohort among the 30 169 half-degree grid cells covering the land region north of 45°N. For the Arctic Basin C budget analysis, C fluxes and stock changes are summarized for the basin watersheds (Fig. 1a) and within the Boreal North America and Boreal Asia regions for comparison with the TransCom 3 estimates of land–atmosphere CO2 flux.
The TEM simulations in this study were driven by temporally- and spatially explicit data sets on atmospheric carbon dioxide concentration ([CO2]), tropospheric ozone (O3), N deposition, climate variability and change and fire, forest harvest and agricultural establishment and abandonment. Global annual atmospheric [CO2] data are from the Mauna Loa station (Keeling and Whorf, 2005). [CO2] data for the time period of years 1000–1900 are held constant at the year 1901 level (296.3 ppm). Monthly air temperature (°C), precipitation (mm) and incident short-wave solar radiation (Wm−2) data derived from observations for the period 1901–2002, gridded at 0.5° resolution, were obtained from the Climate Research Unit (CRU; University of East Anglia, UK; Mitchell and Jones, 2005). The CRU climate variables were extended to 2006 with NCEP/NCAR Reanalysis 1 data sets (NOAA-ESRL Physical Sciences Division, Boulder CO) using a regression procedure based on data anomalies from a 10-yr (1993–2002) mean for each variable (see Drobot et al., 2006). These data sets were hind-casted to year 1000 by a repeating 30-yr cycle of the 1901–1930 monthly data to initialize the carbon pools with climate variability (except for the simulation without climate variability, where 1901–1930 monthly means were used to drive the model for each year). The ozone (O3) pollution data set used in this study, represented by the AOT40 index (a measure of the accumulated hourly ozone levels above a threshold of 40 ppbv), is based on Felzer et al. (2005) and covers the time period from 1860 to 2006. Before 1860, the ozone level in each 0.5° grid cell was assumed to equal the AOT40 of 1860 (which is equal to zero). The atmospheric N deposition data were based on van Drecht et al. (2003), extended from 2000 to 2006 by adding the difference in annual N deposition rate from 1999 to 2000 to succeeding years, for each 0.5° grid cell [e.g. 2001 N deposition rate = 2000 + (2000–1999), etc.]. For years 1000–1859, annual N deposition was assumed to equal the per grid cell rates in 1860.
The distribution of vegetation types in this study (Fig. 1b; see Table 1) was derived from the Global Land Cover Characterization (GLCC; Loveland et al., 2000) version 2 Seasonal Land Cover Regions (SLCR) data set available at 1km (equal-area) resolution for North America and Eurasia. The translated vegetation map was aggregated to the 0.5° grid matching the input climate data sets while retaining the area represented by each unique vegetation type within a grid cell as an individual, non-spatial cohort. Wetland cohort areas were assigned to each grid cell based on a 1°× 1° grid cell fraction inundated database (Matthews and Fung, 1987), where wetland area equals the product of fraction inundated and total cell area. To enable the evaluation of different disturbance and land use change events, we have developed a number of spatially explicit time series data sets to prescribe the timing, area and distribution of historical disturbances and land use change. Historical annual burn areas for North America from 1950 to 2002 were available from the various Alaska and Canada fire databases compiled for the study by Balshi et al. (2007). That study's fire data sets were extended from 2002 to 2006 with updated data from the U.S. Department of the Interior Bureau of Land Management (Alaska) and the Canadian Large Fire Database. The data were extended for Eurasia using the Global Fire Emission Database version 2 (van der Werf et al., 2006). Forest harvest and land use (crops or pasture) cohorts were created in the input data set, derived from 1°× 1° gridded, annual land use transitions data for years 1700–2000, modelled by Hurtt et al. (2006). For Eurasia, the land use transitions data set was hind-casted to the start of the initialization period by linearly ‘ramping-up’ the transitions rates from 0% per year (for each 1°× 1° grid cell) starting in year 1000 to the year 1700 rates. For North America, we assumed land use transition rates of 0% prior to the year 1700. For both regions, the data were extended by simply using the 2000 rates for years 2001–2006.
Table 1. Summary of areas (Mha = 104 km2) of input land cover and disturbance data for the watersheds draining into Arctic Sea Basins, 1997–2006.
|Basin||Total inland||Total upland||Total wetland||Arctic tundra||Boreal forest||Temperate forest||Woodland/Grassland||Burn||Harvest||Agriculture|
|0. Outside Basin||1551||1327||118||122||332||413||578||69.11||40.24||295.35|
|1. Arctic Archipelago||119||70||0||69||1||0||0||0.01||0.00||0.00|
|2. Arctic Subocean||32||1||0||1||0||0||0||0.00||0.00||0.00|
|3. Baffin Bay||59||2||0||2||0||0||0||0.00||0.00||0.00|
|4. Barents Sea||127||96||21||20||67||26||4||0.15||2.03||4.02|
|5. Beaufort Sea||209||171||22||76||102||13||3||7.06||0.91||3.64|
|6. Bering Strait||121||100||17||68||47||0||1||8.29||0.12||0.02|
|7. Chukchi Sea||23||19||3||21||1||0||1||0.08||0.00||0.00|
|8. East Siberian Sea||134||95||33||91||31||0||5||2.46||0.00||0.01|
|9. Foxe Basin||27||15||0||15||0||0||0||0.00||0.00||0.00|
|10. Greenland Sea||59||1||0||1||0||0||0||0.00||0.00||0.00|
|11. Hudson Bay||324||236||59||88||123||28||56||5.90||1.71||58.10|
|12. Hudson Strait||46||31||9||32||8||0||0||0.10||0.04||0.00|
|13. Kara Sea||655||492||136||138||264||61||165||35.56||2.27||31.34|
|14. Laptev Sea||363||333||19||103||226||2||20||14.65||0.28||1.11|
|15. Norwegian Sea||13||11||0||6||4||1||1||0.00||0.13||0.26|
|16. South Greenland||117||13||1||12||1||0||1||0.00||0.00||2.03|
To quantify the effects of the various controlling factors considered in this study on terrestrial C dynamics across the northern high latitude region, we conducted a series of model simulation experiments. A simulation framework was designed to allow an analysis of the relative contribution of the different driving factors to the overall C balance of the system over the recent 10-yr period as compared to the patterns simulated in previous decades. Each simulation builds upon the potential vegetation data set by incorporating an additional transient data set at each successive model run: (1) [CO2], (2) O3, (3) N deposition, (4) climate variability, (5) fire, (6) forest harvest and (7) agricultural establishment and abandonment. Since the transient data sets were individually added in each successive run, the effects of each on C flux were determined by subtracting the results of a simulation from those of the subsequent run. In this study, we report the effects of these factors on both terrestrial CO2 exchange and DOC export.
Information about regional carbon sources and sinks can be derived from a ‘top-down’ approach based on variations in observed atmospheric [CO2] via inverse modeling with atmospheric tracer transport models. The land–atmosphere CO2 exchange estimated by the TEM for this study was compared with model mean and spread from the results of the TransCom 3 project, an intercomparison of atmospheric CO2 inversion models that includes an ensemble of transport models and model variants (Gurney et al., 2002, 2008). The fluxes from the two approaches are compared on the basis of the net ecosystem exchange (NEE, see Chapin et al., 2006) for two high-latitude TransCom land regions (Boreal North America and Boreal Asia). NEE, a negative value of which indicates a surface sink, is the net flux that integrates all vertical exchanges of CO2 between the atmosphere and the land and ocean. The TransCom 3 NEE estimates are based on the ensemble of models run on observation data from the 104-station network (a 1995–2006 monthly time series), with the long-term model mean subtracted from deseasonalized flux estimates to remove the bias in the estimates (see Gurney et al., 2008). The TEM calculates monthly NEE for terrestrial ecosystems as the net difference between photosynthetic uptake and the release of CO2 through plant respiration, decomposition, the decay of harvested products and the CO2 emissions associated with biomass burning. Because the TEM estimates total C emissions associated with biomass burning (see Balshi et al., 2007), we partitioned the total emissions into pyrogenic emissions of CO2, CH4 and CO. The proportion of flaming versus smoldering emissions were determined using ratios for vegetation (80% flaming: 20% smoldering) and soil (20%: 80%) C converted in fire, based on Kasischke and Bruhwiler (2003). The mean emission factors reported in French et al. (2002) were used to calculate the amount of each gas released in fires. Only the emissions of C as CO2 are included in the calculation of NEE, while C emitted as CH4 (fCH4) and CO (fCO) is included in the net ecosystem C balance (NECB; see Chapin et al., 2006). We use the sign conventions for NEE and NECB as defined by Chapin et al. (2006) in which a positive NEE follows the atmospheric sciences sign convention and represents a net flux of CO2 from the surface to the atmosphere (a terrestrial or marine source of CO2). In contrast, NECB follows the ecological sciences sign convention in which a positive NECB represents a net sink of C (i.e. accounting for all forms of C including CO2, CO, CH4, DOC, etc.) in land or ocean ecosystems. Thus, NEE represents the strength of a CO2 source or sink from the land surface, while NECB represents the change in total C storage of an ecosystem. NEE from the TransCom 3 estimates and the model estimates of this study are compared monthly, annually and as deseasonalized fluxes, the latter calculated as the 13-month trapezoidal mean on monthly NEE (Gurney et al., 2008).
2.3. Terrestrial CH4 fluxes
We used the MDM-TEM as a ‘bottom-up’ approach to estimate the biogenic exchanges of CH4 with the atmosphere from terrestrial ecosystems of the Arctic Basin. The MDM-TEM explicitly simulates the processes of CH4 production and CH4 oxidation as well as the transport of the gas between the soil and the atmosphere to estimate net biogenic CH4 emissions (Zhuang et al., 2004, 2007). The model description and parametrizations for both upland and wetland ecosystems are documented in our previous studies (Zhuang et al., 2004, 2006). To simulate net biogenic CH4 exchanges in our study area, which is spatially heterogeneous with respect to land ecosystem types, soils and climate, we apply the module to each 0.5° (latitude × longitude) grid cell within the study area. The regional net CH4 emissions are estimated as the difference between CH4 emissions from wetland ecosystems and CH4 consumption in upland ecosystems. The MDM-TEM in this study was driven with the climate (air temperature, precipitation and incident short-wave solar radiation), vegetation, elevation and soil texture data described earlier for the simulations of CO2 exchange by TEM. The MDM-TEM also used vapour pressure data as input, which were assembled from the CRU data sets and extended to 2006 with NCEP, as described above for the other climate variables used in TEM. Monthly air temperature, precipitation and vapour pressure are interpolated into daily time steps following the method described in Zhuang et al. (2004). MDM-TEM was also driven by spatially explicit data on soil water pH (Carter and Scholes, 2000) and leaf area index (LAI). Monthly LAI for our simulation period is organized following Zhuang et al. (2004) with the existing data for the period 1982–1999 (Myneni et al., 1997, 2001). For the period 2000–2006, monthly LAI data were simulated by TEM. During our simulations, LAI is assumed to remain constant within a month, that is, daily LAI in a particular month is assumed to be the mean monthly value of LAI for that month. The NPP data required for driving MDM-TEM were based on the NPP estimates of TEM, which were aggregated over the cohorts within a grid cell for each month of the simulation.
Similar to CO2, information of regional CH4 sources can be derived from a ‘top-down’ approach based on variations in observed atmospheric CH4 concentrations via inverse modelling with atmospheric tracer transport models. Using an atmospheric inversion approach, Chen and Prinn (2006) estimated methane surface emissions for different methane regional sources and/or processes between 1996 and 2001. Data from 13 high-frequency and 79 low-frequency CH4 observing sites were averaged into monthly mean values with associated errors arising from instrumental precision, mismatch error and sampling frequency. Simulated methane mole fractions were generated using the 3-D global chemical transport model (MATCH), driven by NCEP analysed observed meteorology (T62 resolution; Kalnay et al., 1996), which accounts for the impact of synoptic and interannually varying transport on methane observations. Monthly Arctic emissions were estimated separately for the North American and Eurasian sectors and these results are compared with the sum of the pyrogenic and biogenic CH4 emissions estimated by TEM.
2.4. Terrestrial DOC export
We estimate DOC loading to the river networks of the Arctic Basin by simulating DOC production on land and leaching into rivers in TEM. With this approach, we can examine how climate change and disturbance may affect DOC production and loss from land ecosystems. The production of DOC in TEM is assumed to result from the incomplete decomposition of soil organic matter. As a result, the production of DOC depends upon the same factors that influence decomposition, that is, the amount and quality of soil organic matter, soil temperature and soil moisture (see McGuire et al., 1997). The proportion of DOC produced from decomposition is assumed to vary with vegetation type and is determined from annual NPP estimates of intensively studied field sites and annual DOC export either observed in nearby rivers or estimated from review studies (Table 2). Under equilibrium conditions, NPP would equal decomposition rates and DOC production would equal DOC leaching rates. However, for some ecosystems, no DOC leaching is assumed to occur because no water is transferred between the soil and the river networks at the model calibration site. The TEM assumes DOC is stored in the soil until it is leached from this pool based on the concentration of DOC in soil water and the flux of water from soil to the neighbouring river network. We report the effects of the various controlling factors on DOC from the results of the simulation experiments described in Section 2.2.
Table 2. Determination of the ratio of DOC production to decomposition.
|Vegetation type||DOC export (g C m−2 yr−1)||NPP (g C m−2 yr−1)||DOC: decomposition|
|Dwarf shrub arctic tundra/alpine tundra||1.2 (Peterson et al., 1986)||65 (McGuire et al., 1992)||0.0186|
|Low shrub tundra||2.2 (Peterson et al., 1986)||120 (McGuire et al., 1992)||0.0186|
|Boreal needleleaf evergreen forest||0.9 (MacLean et al., 1999; Petrone, 2005)||152 (Clein et al., 2002)||0.0060|
|Boreal needleleaf deciduous forest||5.0 (estimate)||424||0.0118|
|Boreal broadleaf deciduous forest||0.4 (MacLean et al., 1999; Petrone, 2005)||336||0.0012|
|Temperate needleleaf evergreen forest||4.3 (Aitkenhead and McDowell, 2000)||600||0.0072|
|Temperate broadleaf deciduous forest||4.3 (Aitkenhead and McDowell, 2000)||730||0.0059|
|Xeric woodlands||2.67 (Aitkenhead and McDowell, 2000)||550 (McGuire et al., 1992)||0.0049|
|Xeric Shrubland||0.0 (No river discharge)||110 (McGuire et al., 1992)||0.0000|
|Grasslands||0.0 (No river discharge)||200 (McGuire et al., 1992)||0.0000|
For linkage to the ocean biogeochemistry model, we use the watershed boundaries (Fig. 1a) to determine the land areas of the Arctic Basin that contribute DOC to the Arctic Ocean. This boundary covers 24.2 million km2 of land in which 2276 river systems drain into the Arctic Ocean, Hudson Bay and the northern Bering Sea and is represented by 21 025 grid cells (0.5° latitude × 0.5° longitude). DOC export is estimated for each of watersheds draining into the sixteen sea basins by summing the TEM DOC leaching estimates across the grid cells of the appropriate watersheds associated with the sea basins, which represents a conservative estimate of aquatic freshwater processing of DOC.
To evaluate model performance, we compare DOC export estimated by TEM to those obtained by Manizza et al. (2009) using an empirical approach. Manizza et al. (2009) estimate DOC export from rivers draining into ten sea basins identified by Lammers et al. (2001): (1) Arctic Archipelago, (2) Barents Sea, (3) Beaufort Sea, (4) Bering Strait, (5) Chukchi Sea, (6) East Siberian Sea, (7) Hudson Bay, (8) Hudson Strait, (9) Kara Sea and (10) Laptev Sea. The DOC export into the other six sea basins is assumed to be negligible.
2.5. Ocean CO2 fluxes
The MIT ocean biogeochemistry model used in this study was driven by an ocean general circulation model (OGCM) of the MIT General Circulation Model (Marshall et al., 1997) that includes a coupled sea-ice model. The model is configured on a ‘cubed-sphere’ grid in a limited area Arctic domain with open boundaries at ∼55°N in the Atlantic and Pacific sectors. Prescribed boundary conditions for potential temperature, salinity, flow and sea-surface elevation are provided from previous integrations of a global configuration of the same model (Menemenlis et al., 2005). The grid is locally orthogonal and has a variable horizontal resolution with an average spacing of ∼18 km, which allows the model to represent eddies. The mesh resolves major Arctic straits, including many of the channels of the Canadian Archipelago. This configuration of the MIT ocean model has also been used to assess the freshwater budget of the Arctic Ocean (Condron et al., 2009).
The atmospheric state (10-m surface winds, 2-m air temperature, humidity and downward long and short-wave radiation) is taken from the six-hourly datasets of the NCEP reanalysis (Kalnay et al., 1996). Monthly mean estuarine fluxes of fresh water are based on the Arctic Runoff database (Shiklomanov et al., 2000; Lammers et al., 2001).
We couple our Arctic OGCM to a simplified ocean biogeochemistry model, which now explicitly represents the transport and cycling of dissolved inorganic carbon (DIC), total alkalinity, phosphate, dissolved organic phosphorus and dissolved oxygen (Dutkiewicz et al., 2005). We added an explicit representation of riverine DOC, which has a time-varying riverine source based on empirical or TEM estimates, as well as a simple representation of the sink due to microbial respiration of this riverine DOC after it reaches the Arctic Ocean; DOC is not explicitly respired to DIC in the rivers in this implementation. We first developed parametrizations of the seasonal and regional delivery of terriginous DOC to the Arctic basin based on an empirical data set (Manizza et al., 2009). We implemented this source in the context of an Arctic basin configuration of the MIT ocean circulation model. Using this framework, we demonstrated that the modelled sources and transport of DOC from terrestrial sources was sufficient to accurately capture the observed relationships between DOC and salinity in the Arctic Ocean provided that the timescale for respiration of DOC in the oceans is about 10 yr (Manizza et al., 2009). Hence, we couple the marine and terrestrial carbon cycles by explicitly representing the influence of riverine DOC in estimating the air–sea CO2 fluxes in the Arctic Ocean.