Rising methane emissions in response to climate change in Northern Eurasia during the 21st century

We used a biogeochemistry model, the Terrestrial Ecosystem Model (TEM), to examine the methane (CH4) exchanges between terrestrial ecosystems and the atmosphere in Northern Eurasia from 1971 to 2100. Multiple model simulations using various wetland extent datasets and climate change scenarios were conducted to assess the uncertainty of CH4 fluxes, including emissions and consumption. On the basis of these simulations we estimate the current net emissions in the region to be 20–24 Tg CH4 yr − 1 (1 Tg = 1012 g), two-thirds of which are emitted during the summer. In response to climate change over the 21st century, the annual CH4 emissions in the region are projected to increase at a rate of 0.06 Tg CH4 yr − 1, which is an order of magnitude greater than that of annual CH4 consumption. Further, the annual net CH4 emissions are projected to increase by 6–51% under various wetland extent datasets and climate scenarios by the end of the 21st century, relative to present conditions. Spatial patterns of net CH4 emissions were determined by wetland extent. Net CH4 emissions were dominated by wetlands within boreal forests, grasslands and wet tundra areas in the region. Correlation analyses indicated that water table depth and soil temperature were the two most important environmental controls on both CH4 emissions and consumption in the region. Our uncertainty analyses indicated that the uncertainty in wetland extent had a larger effect on future CH4 emissions than the uncertainty in future climate. This study suggests that better characterization of the spatial distribution and the natural diversity of wetlands should be a research priority for quantifying CH4 fluxes in this region.


Introduction
Next to carbon dioxide (CO 2 ), methane (CH 4 ) is the most important greenhouse gas contributing to global climate change. While CO 2 has been responsible for a majority of the increases in radiative forcing, the steep rate of increase in atmospheric CH 4 is also of great concern because it is 25 times more effective on a per unit mass basis than CO 2 in absorbing long-wave radiation on a 100 yr time scale (IPCC 2007). The contribution of CH 4 to radiative forcing from pre-industrial to present time is estimated to be about 20% of that from all greenhouse gases (Le Mer and Roger 2001). Atmospheric CH 4 has a wide variety of natural and anthropogenic sources, and the most important natural sources are wetlands (Bartlett andHarriss 1993, Wuebbles andHayhoe 2002). The total global CH 4 source is relatively well constrained from atmospheric observations, but the strength and trends of the contributing sources are considerably less known due to high spatial and temporal variability (IPCC 2007).
Net methane emissions from terrestrial ecosystems are determined by two different microbial processes: methanogenesis, which produces methane, and methanotrophy, which consumes methane. Methanogenesis occurs under anaerobic conditions such as typically found in wetlands and depends mainly on soil organic matter and vegetation. In contrast, methanotrophy occurs in aerobic soils or aerated surface waters and is strongly controlled by availability of oxygen. Both processes are influenced by water table depth, soil temperature and pH Heimann 2000, MacDonald et al 1998).
Northern Eurasia accounts for 60% of the terrestrial land cover north of 40 • N and contains vast areas of wetlands, especially peatlands, which contain a large amount of organic carbon and are often underlain by continuous and discontinuous permafrost (NEESPI 2004). Compared with low latitudes, the region, especially its northern areas, has been experiencing more dramatic environmental changes, including increasing temperatures, melting of permafrost, changes in precipitation and prolonged growing seasons (Fedorov 1996, Romanovsky et al 2000, IPCC 2007. These changes in climate, plant, soil thermal and hydrological conditions have resulted in changes in the magnitude and timing of CH 4 emissions and consumption (e.g., Friborg et al 1997, West and Schmidt 1998, Zimov et al 2006. To quantify CH 4 fluxes at regional and global levels, many large-spatial-scale biogeochemical models have been developed to estimate current and future CH 4 exchanges between terrestrial ecosystems and the atmosphere (Cao et al 1996, Christensen et al 1996, Walter et al 2001. Large uncertainties exist in the regional and global CH 4 budgets from these model studies. Cao et al (1996) estimated global CH 4 emissions to be 145 Tg CH 4 yr −1 (1 Tg = 10 12 g), of which 24 Tg CH 4 yr −1 was from natural wetlands. Christensen et al (1996) gave a CH 4 flux of 20 Tg CH 4 yr −1 from northern wetlands (>50 • N). Walter et al (2001) put the estimate of global annual CH 4 emissions from wetlands to be 260 Tg CH 4 yr −1 , of which 25% originated from wetlands north of 30 • N. One uncertainty in regional and global CH 4 budgets arises from uncertainty in the extent of wetlands, for it is difficult to characterize inundated areas and their dynamics in a broad range of environmental conditions. For instance, a recent model study conducted by Petrescu et al (2010) put the estimate of current CH 4 emissions from circum-arctic wetlands (<5 • C for mean annual air temperature) in a broad range (between 38 and 157 Tg CH 4 yr −1 ) based on multiple existing or modeled wetland extent datasets. Uncertainty in future climate is another source of uncertainty for estimates of future CH 4 exchanges between terrestrial ecosystems and the atmosphere. This climate uncertainty among estimates from global climate models is largely the result of different assumptions about effective climate sensitivity of the earth system, the strength of aerosol forcing, and the rate at which heat is mixed into the deep ocean (Meehl et al 2007).
To date, there is a lack of comprehensive estimates of CH 4 fluxes for Northern Eurasia that consider both emissions and consumption of CH 4 within different wetland types. Furthermore, little information is currently available about how regional CH 4 dynamics in Northern Eurasia will respond to transient changes in future climate. Here, we apply a processbased biogeochemistry model to examine how uncertainty in wetland extent and future climate influence projected CH 4 fluxes between terrestrial ecosystems and the atmosphere in Northern Eurasia over the 21st century.

Overview
In this study, we conducted simulations of CH 4 fluxes in Northern Eurasia during 1971-2100 with the Terrestrial Ecosystem Model (TEM, (Zhuang et al 2004)) based on three wetland extent datasets and six future climate scenarios. First, we examined the responses of CH 4 fluxes simulated by TEM with three wetland extent datasets to both historical and six future climate scenarios. Then, we assessed the spatial distribution and temporal variability of CH 4 fluxes among different simulations in both the historical period and the future. Finally, we identified the key controls on CH 4 dynamics for both emission and consumption with correlation analyses.

Model and data
The biogeochemistry model TEM explicitly simulates the processes of CH 4 production (methanogenesis) and CH 4 oxidation (methanotrophy), as well as the transport of the gas between the soil and the atmosphere (Zhuang et al 2004). The net CH 4 emissions from soils to the atmosphere are the total of the CH 4 fluxes at the soil/water-atmosphere boundary via different transport pathways (see supplementary material for detailed model description available at stacks.iop.org/ERL/6/ 045211/mmedia). To make spatially and temporally explicit estimates of CH 4 fluxes over Northern Eurasia with TEM, we used data of climate, wetland extent, vegetation type, soils, elevation and atmospheric CO 2 from a variety of sources. Detailed descriptions of data and processing methods are provided as supplementary material (available at stacks.iop. org/ERL/6/045211/mmedia).

Simulation protocol
TEM was applied to simulate both CH 4 emissions and consumption from both wetland and upland ecosystems in Northern Eurasia from 1971 to 2100 at a 0.5 • × 0.5 • spatial resolution. Both wetland and upland ecosystems may occur in each grid cell. The ecosystem-specific CH 4 flux estimates were then area-weighted for each grid cell, as defined by the corresponding wetland extent datasets. We defined the regional net CH 4 emissions as the difference between CH 4 emissions from wetland ecosystems and CH 4 consumption in upland ecosystems. The model parameterization and validation were conducted in a previous study (Zhuang et al 2004), and the same parameter sets were used in this study.
In order to assess the estimate uncertainty of CH 4 emissions and consumption over Northern Eurasia, associated with wetland extent and climate change, we conducted three sets of simulations with TEM based on three wetland extent datasets: WET1 (Matthews and Fung 1987), WET2 (Lehner and Döll 2004), and WET3 (Papa et al 2010) (see supplementary material for explanations of each dataset available at stacks.iop.org/ERL/6/045211/mmedia).
For the historical period , each set of simulations contained one model run driven by historical climate data. For the future period (2001-100), each set of simulations was driven by one of six climate change scenarios developed by Sokolov et al (2005) that represented high and median climate responses to anthropogenic emissions based on either a 'business-as-usual' scenario or a level 1 stabilization scenario. See supplementary material (available at stacks.iop.org/ERL/ 6/045211/mmedia) for more details of climate scenarios.

Statistical analysis
We used the Mann-Kendall trend analysis method (Hamed and Ramachandra Rao 1998) to check the trend of the time-series data. Pearson's correlation coefficient was used to calculate the correlation coefficients between environmental factors and CH 4 fluxes. All the statistical analyses were conducted in MATLAB (The MathWorks, Inc., MA, USA).

Spatial distribution of CH 4 fluxes
The net CH 4 emissions, the difference between emissions and consumption, showed substantial spatial variation over Northern Eurasia, with different spatial patterns visible among the three simulations with WET1, WET2 and WET3 (figure 1). During the 1990s, some ecosystems acted as a source of atmospheric CH 4 , contributing up to 10 g CH 4 m −2 yr −1 , and some dry ecosystems consumed up to 2 g CH 4 m −2 yr −1 . A major source of atmospheric CH 4 was found in western Siberia where wetland ecosystems were identified in all three wetland extent datasets; strong sinks of atmospheric CH 4 were found mainly in the western and southern parts of Northern Eurasia, while other areas acted as weak sinks of atmospheric CH 4 . During the 2090s, the magnitude of net CH 4 emissions is projected to increase for each of the simulations with an unchanged general spatial pattern (figure 1).
Northern Eurasia as a whole acted as a CH 4 source of 21 ± 2.5 Tg CH 4 yr −1 during the 1990s. This source was represented by net CH 4 emissions, which were the difference between the CH 4 emissions at 26 ± 2.4 Tg CH 4 yr −1 , and the CH 4 consumption at 5 ± 0.1 Tg CH 4 yr −1 over the total area of 28.87 million km 2 . The three sets of simulations gave different magnitudes of CH 4 emissions and consumption due to the differences in the spatial distribution of wetland extent in the various datasets (figure S3, table S2 (available at stacks.iop.org/ERL/6/045211/mmedia)). Our estimates of net CH 4 emissions in Northern Eurasia were comparable to those modeled in previous studies (18-33 Tg CH 4 yr −1 , Anisimov 2007, Denisov et al 2010. The source of CH 4 emissions was dominated by wetlands within boreal forests, grasslands and wet tundra areas in the region. Among all ecosystem types, boreal forest accounted for the largest amount of CH 4 fluxes in both emissions (9 ± 3.7 Tg CH 4 yr −1 , figure 2(a)) and consumption (2 ± 0.1 Tg CH 4 yr −1 , figure 2(b)) because of its vast wetland areas (table S2 available at stacks.iop.org/ERL/6/ 045211/mmedia).
On a per unit area basis, there were significant differences in variation patterns across ecosystems both in CH 4 emissions and consumption, compared with those for the regional total fluxes. For CH 4 emissions, the magnitude varied from 20 ± 0.9 mg CH 4 m −2 d −1 in alpine tundra to 198 ± 68.3 mg CH 4 m −2 d −1 in xeric woodlands, with a mean of 93 ± 14.6 mg CH 4 m −2 d −1 for all ecosystem types (figure 2(c)). For CH 4 consumption, the magnitude varied from 0.3 ± 0.0 mg CH 4 m −2 d −1 in alpine tundra to 0.7 ± 0.0 mg CH 4 m −2 d −1 in temperate coniferous forests and xeric shrublands, with a mean of 0.6 ± 0.0 mg CH 4 m −2 d −1 for all ecosystem types (figure 2(d)). Overall, ecosystems located in the warmer southern and western regions had higher CH 4 fluxes than the colder northern regions (figure S2a available at stacks.iop.org/ERL/6/045211/mmedia).
Apart from the different magnitudes of CH 4 fluxes per unit area across ecosystems, discrepancies were found among the three simulations although model parameters and climate inputs for a specific ecosystem were the same during the 1990s. These discrepancies were caused by different spatial distributions of wetland extent for each dataset: the spatial overlay of wetland distribution and vegetation types for the three simulations presented different spatial patterns for wetland types ('wetland-vegetation' combination), which gave different averaged climate conditions, affecting averaged CH 4 fluxes for a specific wetland type. For example, wetlands within xeric shrublands, which presented very different spatial patterns among the three wetland extent datasets (figure S3 available at stacks.iop.org/ERL/6/045211/mmedia), showed large differences in CH 4 emissions among the three simulations ( figure 2(c)). In general, it seemed the differences in wetland extent in the warmer southern ecosystems had more influence on CH 4 emissions than in the colder northern ecosystems, which implied that it was more important to get accurate wetland extent data for modeling CH 4 emissions in warmer southern ecosystems in this region. We compared our simulated daily CH 4 emissions and consumption against site measurements in Northern Eurasia for recent years. We found that, for tundra ecosystems, the mean modeled estimates of net methane emissions during the growing season (60 mg CH 4 m −2 d −1 ) were within the range of measurements (4-195 mg CH 4 m −2 d −1 , Heyer et al 2002). For boreal forest ecosystems, the mean estimates of net methane emissions during the growing season (60-228 mg CH 4 m −2 d −1 ) were well within the range of measured values (21-233 mg CH 4 m −2 d −1 (Takeuchi et al 2003); 124 and 209 mg CH 4 m −2 d −1 -median for eutrophic and oligotrophic wetlands, respectively (Glagolev et al 2008). However, for grassland ecosystems, the mean modeled estimates of net methane emissions during the growing season (230 mg CH 4 m −2 d −1 ) was higher than the observed rates from 73 to 166 mg CH 4 m −2 d −1 (Tsuyuzaki et al 2001). The mean modeled estimates of net methane consumption during the growing season (0.6-1.4 mg CH 4 m −2 d −1 ) was at the low end of the observed rates, from 0.3 to 5.3 mg CH 4 m −2 d −1 (Gal'chenko et al 2001), while the simulated annual consumption (0.12-0.27 g CH 4 m −2 yr −1 ) was above the high end of the range of measurements (0.04-0.12 g CH 4 m −2 yr −1 , Flessa et al 2008).

Temporal variations of CH 4 fluxes
The annual CH 4 fluxes over Northern Eurasia exhibited a significant interannual variability from 1971 to 2100 (figure 3). In all three sets of simulations, CH 4 emissions increased much more rapidly than consumption in both the historical period and the future. For CH 4 emissions, there were notable differences among the three sets of simulations during the historical period. The WET1 simulation gave an obviously higher flux of annual CH 4 emissions (∼29 Tg CH 4 yr −1 during the 1990s) than the other two simulations (∼25 Tg CH 4 yr −1 during the 1990s). From 1971 to 2100, the annual CH 4 emissions for all the three sets of simulations showed significant increasing trends at the rates of 0.056±0.027 Tg CH 4 yr −1 for the WET1 simulations, 0.055± 0.025 Tg CH 4 yr −1 for WET2, and 0.059 ± 0.033 Tg CH 4 yr −1 for WET3. For CH 4 consumption, the three sets of simulations gave similar results during the historical period, while the magnitudes grew increasingly different in the future under different climatic scenarios. The long-term increasing rate of annual CH 4 consumption for the three sets of simulations was 0.006 ± 0.003 Tg CH 4 yr −1 .
Our simulations projected that the annual net CH 4 emissions will increase by 6-51% under various wetland extent datasets and future climate scenarios by the end of the 21st century in comparison with the present conditions. The annual net CH 4 emissions increased from 23 Tg CH 4 yr −1 in the 1970s to 29 ± 2.6 Tg CH 4 yr −1 in the 2090s in the WET1 simulation, from 19 to 25 ± 2.6 Tg CH 4 yr −1 in WET2, and from 19 to 25 ± 3.7 Tg CH 4 yr −1 in WET3 (table 1). The averaged 25% increase of net methane emissions in this century in our multiple model simulations was within the same range as other studies. For example, Anisimov (2007) reported a 15-25% increase by the middle of this century, and an increase of 30% and more by 2080 for most fragile and methane 'rich' Arctic zone. After comparing estimated annual net CH 4 emissions from the three sets of simulations, we found the differences in CH 4 emissions between WE1 and WET2/WET3 simulations to be comparable with the differences we found between changing climate scenarios. Taking one standard error as the measure of uncertainty, the estimate uncertainty of CH 4 emissions from wetland extent (>2 Tg CH 4 yr −1 ) was greater   1, figure 3). The temporal dynamics of CH 4 fluxes showed substantial seasonal variation for the three sets of simulations, with weak fluxes in the winter and strong fluxes in the summer (figure 4). The monthly CH 4 emissions over the region were quite different in magnitude, particularly for peak fluxes (figure 4(a)), while monthly CH 4 consumption had only small differences ( figure 4(b)). The monthly CH 4 emissions were one order of magnitude greater than that of CH 4 consumption. Net CH 4 emissions had a maximum monthly CH 4 source of 5.0 Tg CH 4 mon −1 in July (averaged for all simulations). Summer (May, June and July) was responsible for two-thirds of the annual net CH 4 emissions.

Impact of environmental factors on CH 4 fluxes
To identify the key controls on CH 4 fluxes in Northern Eurasia, we examined correlations between environmental factors and CH 4 fluxes. Correlation analyses were performed between regional total CH 4 fluxes (emissions and consumption) and regional averaged environmental factors over all of Northern Eurasia. Our analysis indicated that increases in water table depth, soil temperature and labile carbon availability associated with climate change were the major factors that caused an increase in CH 4 emissions on an annual basis (table 2). Specifically, methane emissions were strongly correlated (N = 30 years, P < 0.01) with water table depth (r = 0.80), soil temperature (r = 0.79), and NPP (r = 0.50).
The significant influences of soil temperature ( figure 5(b)) and labile carbon availability (figure 5(e)) on methane emissions (figure 3(a)) were consistent with the conclusion that soil temperature and labile carbon availability were key factors in determining methanogenesis (e.g., Christensen et al 2003, Bellisario et al 1999. For CH 4 consumption in Northern Eurasia, our analysis indicated that annual CH 4 consumption was strongly related to water table depth, soil temperature and precipitation (table 2). The lowering of the water table due to increases in soil temperature ( figure 5(b)) and evapotranspiration resulted in an increase in CH 4 consumption ( figure 3(b)). A negative trend in annual mean soil moisture (figure 5(c)) was noticeable, while precipitation tended to increase (figure 5(a)) over our projected study period, suggesting that any potential rise in the water table caused by additional precipitation was more than counteracted by the drop in the water table due to the increase in soil temperature. The correlation between water table depth and CH 4 consumption was consistent with field experiments (e.g., Nykänen et al 1998, Heikkinen et al 2002. The importance of soil temperature to the consumption rate was also consistent with the laboratory studies of Whalen and Reeburgh (1996) for soils with high CH 4 concentrations.

Conclusions and future work
The CH 4 fluxes including emissions and consumption in Northern Eurasia were estimated with TEM for the period 1971-2100. Current net CH 4 emissions were estimated to be 20-24 Tg CH 4 yr −1 , and the magnitude is projected to increase by 6-51% under various wetland extent datasets and future climate scenarios by the end of the 21st century. Sources of CH 4 emissions were dominated by wetlands within boreal forests, grasslands and wet tundra areas in the region, and spatial patterns of emissions were determined by wetland extent. Water table depth and soil temperature were the two most important environmental controls on both CH 4 emissions and consumption in this region. Our uncertainty analyses indicated the uncertainty in wetland extent had a larger effect on future terrestrial CH 4 emissions than the uncertainty in future climate. This highlighted the importance of accurate wetland extent on modeling CH 4 dynamics at the regional scale.
One of the uncertainties in regional estimates of CH 4 arises from the uncertainty of the model parameterization derived from limited field measurements. Another source of uncertainty is the high variability of wetland extent data, which determines the magnitude of CH 4 fluxes. All three wetland extent datasets we used in the simulations represent the extraction of the global datasets, which are not aimed directly at Northern Eurasia, and do not consider specific conditions of wetlands distribution and diversity in the region, which means that they could seriously underestimate the real extent of wetland areas in the region (see Global Peatland Database www.imcg.net, Minayeva et al 2009, Vompersky et al 2005. A large part of wetlands may not be detected by satellite sensors, and more efforts are required to develop more accurate wetlands distribution data for the region. In addition, a number of additional factors should be considered in future analyses. One is consideration of the effects of the deep carbon substrate for methane production (Sirin et al 1998). Second, humaninduced disturbances can both inhibit (e.g., drainage) and enhance (e.g., ditching, damming) CH 4 fluxes from wetlands (Sirin and Laine 2008, Chistotin et al 2006, Glagolev et al 2008, but these effects were not considered in our study. Large areas of wetlands in Europe, central and western parts of European Russia, southern regions of West Siberia and some in the Far East have been disturbed by human activities. To improve future assessment of CH 4 dynamics in this region, research priorities should be directed at better characterizing the spatial distribution and natural diversity of wetlands in the region, associated with human-and climate-induced land use and land cover changes, as well as more detailed and precise descriptions of CH 4 cycling in these biogeochemically complex and spatially uneven ecosystems.