1. Introduction
High-latitude terrestrial ecosystems contain approximately 30% of the world’s vegetation carbon (McGuire et al. 1995) and about 40% of the world’s soil carbon (Melillo et al. 1995). Much of the soil carbon in these ecosystems is highly labile and has accumulated simply because of cold and/or anaerobic soil conditions (McGuire et al. 2006). This labile carbon pool is highly vulnerable to warming (Goulden et al. 1996), and a large release of carbon from high-latitude soils could substantially affect atmospheric concentrations of CO2 (McGuire et al. 2006). However, increased vegetation growth in response to warming has the potential to mitigate the release of carbon from high-latitude soils (McGuire et al. 1992; Shaver et al. 1992). Analyses based on remote sensing approaches that use a 20-yr record of satellite data indicate that the tundra is greening in the Arctic, suggesting an increase in photosynthetic activity and net primary production (Sitch et al. 2007). In contrast, a recent remote sensing study has found that photosynthetic activity in the boreal forest of North America unaffected by fire has declined in the last two decades (Goetz et al. 2005). While process-based biogeochemical models generally simulate a small net carbon sink in the recent past for the distribution of Arctic tundra (Sitch et al. 2007), there is much spatial variability in estimated net carbon balance by these models (e.g., see McGuire et al. 2000a; Euskirchen et al. 2006; Thompson et al. 2006).
The Western Arctic Linkage Experiment (WALE) was set up to evaluate uncertainties in regional hydrology and carbon estimates in Alaska and the adjacent Yukon Territory associated with 1) alternative driving datasets and 2) alternative simulation models. As part of the WALE, Kimball et al. (Kimball et al. 2007) conducted a study of carbon balance in the WALE region that compared the simulations of remote sensing and process-based models during recent decades. These simulations indicate that plant production and the ratio of vegetation to soil carbon are increasing throughout the region. These results suggest that the regional carbon cycle in the western Arctic is accelerating under a warming climate. The climate data used to drive the simulations presented in Kimball et al. (Kimball et al. 2007), which were drawn from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kistler et al. 2001), were just one of several alternative climate datasets that could have been used in the study. Because there is much uncertainty among the alternative climate datasets in the western Arctic (Drobot et al. 2006; Rawlins et al. 2006; Rupp et al. 2007), it is not clear the degree to which the findings of Kimball et al. (Kimball et al. 2007) depend on the alternative climate datasets. In this study, we assessed the sensitivity of carbon dynamics by the Terrestrial Ecosystem Model (TEM; Euskirchen et al. 2006) in recent decades to alternative driving datasets.
2. Methods
2.1. Overview
In this study, we conducted simulations of carbon dynamics in the WALE region during two recent decades by driving TEM with three alternative climate datasets: 1) NCEP–NCAR reanalysis (Kistler et al. 2001; referred to as NCEP1 in this study); 2) the Climate Research Unit (CRU; Mitchell and Jones 2005); and 3) the fifth-generation Penn State–NCAR Mesoscale Model (MM5; Wu et al. 2007). Among the three simulations, we compared the mean monthly and interannual variability of three carbon fluxes simulated by TEM: 1) net primary production (NPP), 2) heterotrophic respiration (Rh), and 3) net ecosystem production (NEP). We also compared cumulative changes in vegetation, soil, and total carbon storage among the three simulations. To evaluate one of the conclusions of Kimball et al. (Kimball et al. 2007), we compared how the ratio of vegetation to soil carbon is changing through time among the three simulations.
2.2. The Terrestrial Ecosystem Model
The TEM is a monthly, process-based, global-scale ecosystem model that uses spatially referenced information of climate, elevation, soils, and vegetation to simulate monthly estimates of terrestrial carbon, nitrogen, and water dynamics at different spatial scales. The TEM has been used to investigate several issues influencing carbon dynamics in high latitudes including studies focusing on boreal forests (Amthor et al. 2001; Potter et al. 2001; Clein et al. 2002; Zhuang et al. 2002), tundra ecosystems (Clein et al. 2000; McGuire et al. 2000a), and high-latitude ecosystems in general (McGuire et al. 2000b; Zhuang et al. 2003; Zhuang et al. 2004; Zhuang et al. 2006; Euskirchen et al. 2006). In this study, we used the version of the model described in Euskirchen et al. (Euskirchen et al. 2006), which is an update from the version used by Zhuang et al. (Zhuang et al. 2003). This version of the model has an integrated soil thermal model (with an updated soil freeze–thaw algorithm) and considers the effects of freeze–thaw dynamics on gross primary production. The most detailed descriptions can be found in Raich et al. (Raich et al. 1991), McGuire et al. (McGuire et al. 1992), and Tian et al. (Tian et al. 1999). The following is a short summary of model components related to the simulated fluxes analyzed in this study.
2.3. Model parameterization
Although many of the parameters in the model are defined from published information (Raich et al. 1991; McGuire et al. 1992), some are determined by calibrating the model to fluxes and pools of an intensely studied field site. For the application of TEM to the WALE region, we created vegetation-specific parameterizations for wet boreal coniferous forest, dry boreal coniferous forest, boreal deciduous forest, moist tundra, and coastal maritime coniferous forest. The pools and fluxes used to develop these parameterizations (Table 1) are based on intensive field studies at the Bonanza Creek, Toolik Lake, and H. J. Andrews Long Term Ecological Research sites in Alaska and Oregon, and at the Boreal Ecosystem Atmosphere Study in Canada (Tables 1 and 2). We used the estimates of pools and fluxes in Table 1 to calibrate the rate limiting parameters in the flux equations for gross primary production, autotrophic respiration, heterotrophic respiration, litterfall carbon, litterfall nitrogen, plant nitrogen uptake, and soil nitrogen immobilization. Calibration of these parameters in TEM requires the use of climate data, and we used the mean climate averaged over the period from 1961 to 1990 from the CRU climate dataset for the parameterizations as this is the dataset that is generally used to calibrate TEM. We held atmospheric CO2 at 280 ppmv during the calibration procedure as this is approximately the concentration of CO2 that is used to initialize the model simulations in this study.
2.4. Alternative climate datasets
In this study, we drove TEM to estimate carbon dynamics of the WALE region with temperature, precipitation, and net irradiance (NIRR; downwelling shortwave radiation) from three different input climate datasets. One of the datasets we used is from the NCEP–NCAR reanalysis project (Kistler et al. 2001) and is the same NCEP1 dataset from 1980–2000 that was used to drive the TEM simulations reported in Kimball et al. (Kimball et al. 2007). These data, which were generated using an assimilation scheme to ingest data into a climate model in order to fill in data gaps in both space and time, were sampled using Geographical Information System techniques to develop datasets at 25 km × 25 km resolution for use in the WALE simulation (see McGuire et al. 2007a, manuscript submitted to Earth Interactions, hereafter M07). We also drove TEM over the WALE region from 1980 to 2000 with data from the CRU (Jones et al. 2001; Mitchell and Jones 2005). The global CRU dataset is a monthly 0.5° × 0.5° (latitude by longitude) climate dataset that has been developed based on the use of station anomalies in a variance-adjusted interpolation scheme to damp overestimates of variance in regions where data are sparse (Jones et al. 2001). For use in this study, we sampled the CRU datasets at the 0.5° latitude × 0.5° longitude grid cell and disaggregated them based on their areas of intersection to develop datasets at 25 km × 25 km resolution for use in the WALE simulation. Our CRU estimates of NIRR were developed from the CRU mean monthly cloudiness data fields similar to the algorithm used in Pan et al. (Pan et al. 1996). Finally, we drove TEM with data from the MM5 simulation of Wu et al. (Wu et al. 2007), which produced estimates of surface climate over the WALE region at a spatial resolution of 50 km from 1990 to 2000. Thus, in this study, we only drove TEM with data from the MM5 simulation from 1990 to 2000. See Drobot et al. (Drobot et al. 2006) and Herzfeld et al. (Herzfeld et al. 2007) for more information on these datasets.
2.5. Other input datasets
Besides climate datasets, the TEM simulations required data on soil texture, elevation, vegetation, and atmospheric CO2. The soil texture data are based on the Food and Agriculture Organization/United Nations Educational, Scientific and Cultural Organization (FAO/UNESCO 1974) soil map of the world. The input elevation map is based on 10-min digital elevation data (NCAR/Navy 1984). The input vegetation map (see M07), which is an extension of the dataset described in Calef et al. (Calef et al. 2005), is composed of moist tundra (63%), wet coniferous boreal forest (16%), dry coniferous boreal forest (10%), boreal deciduous forest (10%), and temperate maritime coniferous forest (1%). This 1-km-resolution land cover classification was developed by aggregating the 23 land cover classes in the commonly used Advanced Very High Resolution Radiometer (AVHRR)-based classification for Alaska (Fleming 1997) using topography, climate, and geographic location. In Fleming (Fleming 1997), land cover classes corresponding to tundra were delineated from classes corresponding to forest based on growing season temperature and geographic location. Classes corresponding to deciduous forest were delineated from classes corresponding to conifer forest based on summer spectral signatures. The conifer forest class derived from Fleming (Fleming 1997) was separated into black and white spruce classes based on aspect and slope, with black spruce located primarily on northern aspects or low slope angles (Calef et al. 2005; Rupp et al. 2007). We used inputs of atmospheric CO2 from observations averaged from the Mauna Loa and South Pole monitoring stations (Keeling et al. 1995, updated).
2.6. Simulation protocol
The TEM simulations were initialized by running the model to equilibrium (NEP = 0 ± 1 g C m−2 yr−1) using the mean monthly climate for each climate driver dataset (e.g., mean monthly NCEP1 temperature data for 1980–2000), and then running through five cycles of the climate dataset (i.e., five cycles of 1980–2000 for the simulation driven by NCEP1 data) before encountering the final transient period of simulation. We simulated C and N pools and fluxes for each vegetation type for all 25 km × 25 km grid cells in the WALE region. Model outputs were then spatially aggregated according to the relative proportions of individual land cover classes identified within each 25-km grid cell. Spatial aggregation of model outputs within each grid cell was based on linear weighting of dominant and subdominant land cover classes, with no lateral transfers of mass or energy within a specified grid cell or between adjacent grid cells.
3. Results
3.1. Differences in input climate data
Our analysis of absolute differences in mean annual temperature over the entire WALE region indicates that CRU is warmer than NCEP1 on the order of about 1°C (Figure 1a). During the 1990s, mean annual temperature from the MM5 simulations is approximately 3°C cooler than CRU and 2°C cooler than NCEP1. In agreement with Drobot et al. (Drobot et al. 2006), comparison of mean monthly temperature indicates that CRU is warmest in summer and coolest in winter, MM5 is coolest in summer and warmest in winter, and NCEP1 is intermediate (Figure 1b).
Our analysis of absolute differences in annual precipitation indicates that both NCEP1 and MM5 estimate approximately 300 mm more precipitation than CRU, with MM5 being slightly higher than NCEP1 during the 1990s (Figure 1c). In the summer, NCEP1 has the highest estimates of precipitation, a pattern that in the Arctic is generally considered an overestimation because of excessive convective precipitation in the model (Serreze and Hurst 2000), while CRU and MM5 have similar estimates from June to August (Figure 1d). In the winter, MM5 has the highest estimates of precipitation, CRU the lowest, and NCEP1 is intermediate (Figure 1d).
Our comparison of absolute differences in annual NIRR indicates that CRU and MM5 have similar estimates in the 1990s, but that NCEP1 is approximately 80 W m−2 higher than CRU and MM5 (Figure 1e). The higher NIRR estimates of NCEP1, which occur in all months of the year (Figure 1f), are known to be too high because of insufficient cloud cover (Serreze et al. 1998; Serreze et al. 2003). The NIRR estimates of CRU are higher than those of MM5 in the summer, but this pattern is reversed in the winter (Figure 1f).
3.2. Differences in simulated annual and seasonal carbon fluxes
The annual carbon fluxes of NPP estimated by TEM were quite different in magnitude among the three simulations, with the highest fluxes for the CRU simulation and the lowest for the MM5 simulation (Figures 2a). The differences in NPP among simulations are primarily caused by differences that occurred in the growing season (Figures 3a) and are the result of the temperature differences among the simulations driven by CRU, NCEP1, and MM5. The differences in temperature can be seen to affect the seasonality of carbon uptake in the spring, as NPP in June is much higher in the CRU simulation in comparison with the other two simulations. This pattern is likely associated with differences in soil thaw during May as photosynthetic carbon uptake in TEM depends on the freeze–thaw status of the soil (Zhuang et al. 2003); mean May air temperature over the WALE region is 3°C in the CRU dataset, while it is −0.3°C in the NCEP1 and −1.6°C in the MM5 datasets. The earlier start to the growing season and the warmer temperatures leads to high carbon uptake in both June and July in the CRU simulation, while the NCEP1 and MM5 do not achieve peak uptake until July.
The annual carbon fluxes of Rh estimated by TEM were also quite different in magnitude among the simulations and showed a similar pattern to NPP with the highest fluxes for the CRU simulation and the lowest for the MM5 simulation (Figure 2b). In contrast to NPP, the differences in Rh are maintained throughout the year (Figure 3b) and are the result of differences in summer temperature as well as differences in soil carbon among the simulations as Rh depends on both temperature and the size of the soil carbon pool; estimated soil carbon over the entire WALE region at the beginning of 1990 is 1.4 × 1015 g C higher in the CRU simulation than in the NCEP1 simulation, which is 2.7 × 1015 g C higher than in the MM5 simulation.
The annual fluxes of NEP, which is the difference between annual fluxes of NPP and Rh, are generally correlated and similar in magnitude from 1980 to the mid-1990s, but the correlation appears to be breaking down in the late 1990s (Figure 2c) primarily because of different interannual variability in NPP responses among the simulations (Figure 2a). The generally similar magnitudes throughout most of the simulation period are caused by net carbon uptake patterns in summer playing off against net carbon release patterns in winter (Figure 3c). For example, CRU has the highest uptake in the summer and the greatest release in winter while MM5 has the lowest uptake in summer and the lowest release in winter (Figure 3c); the NCEP1 simulation has intermediate uptake in summer and intermediate release in winter.
3.3. Differences in temporal trends and sensitivities of annual carbon fluxes
Spatially averaged annual NPP does not significantly increase or decrease at the 0.05 level in any of the simulations (Figures 4a and 4b; Table 3), but it is marginally significant for the time period 1980–2000 in the simulation driven by NCEP1 climate (P = 0.08). The trend of increasing NPP from 1980 to 2000 in the NCEP1 simulation of TEM is associated with increases in summer air temperature and potential evapotranspiration and decreases in summer radiation. Annual NPP from 1980 to 2000 is significantly positively related to summer air temperature and summer potential evapotranspiration in both the NCEP1 and CRU simulations (Table 4) but is significantly negatively related to summer radiation in only the NCEP1 simulation (Table 4). Across all simulations, the variable that seems to be most strongly related to annual NPP is summer air temperature (Table 4). The spatial pattern of trends in NPP during the 1990s is quite different among the three simulations (Figure 5). For example, at the U.S.–Canada border in the WALE region, the NCEP1 simulation (Figure 5a) indicates large increases in NPP, the CRU simulation (Figure 5b) indicates large decreases in NPP, and the MM5 simulation (Figure 5c) indicates moderate decreases in NPP.
Spatially averaged annual Rh does not increase or decrease in the NCEP1 and MM5 simulations, but it decreases significantly in the CRU simulations (Figures 4c and 4d; Table 3). The Rh decreases in the CRU simulations are associated with significant decreases in annual precipitation (Table 3). From 1980 to 2000, the Rh decreases in the CRU simulation are also associated with increases in summer air temperature and summer potential evapotranpsiration, and decreases in annual precipitation minus potential evapotranspiration (Table 3). However, of these variables, annual Rh from 1980 to 2000 in the CRU simulation is only significantly positively related to annual precipitation and annual precipitation minus potential evapotranspiration (Table 5), which suggests that interannual variability in Rh of the CRU simulation is primarily controlled by variability in simulated soil moisture. However, temperature is also an important control as annual Rh across all simulations is strongly related to annual air temperature in addition to annual precipitation (Table 5). As with NPP, the spatial pattern of trends in Rh during the 1990s is quite different among the three simulations (Figure 5). For example, at the U.S.–Canada border in the WALE region, the NCEP1 simulation (Figure 5d) indicates large increases in Rh, the CRU simulation (Figure 5e) indicates large decreases in Rh, and the MM5 simulation (Figure 5f) indicates moderate increases for the northern part of the border and moderate decreases for the southern part of the border.
Spatially averaged annual NEP from 1980 to 2000 increases significantly in the CRU simulation, and the increase in the NCEP1 simulation is marginally significant (Figure 4e; Table 3). The increases in NEP in the CRU simulation is associated with a nonsignificant increase in NPP coupled with a significant decrease in simulated Rh over the time period (Table 3). In contrast, the increase in NEP in the NCEP1 simulation is associated with a marginally significant increase in NPP that is much stronger than the nonsignificant increase in Rh.
3.4. Comparison of changes in carbon stocks
Vegetation carbon accumulates in the NCEP1 and the CRU simulation but increases most rapidly in the CRU simulation (Figures 6a and 6b); changes in vegetation carbon are negligible in the MM5 simulation (Figure 6b). Changes in vegetation carbon storage depend on both the input of carbon to the vegetation (NPP) and the loss of carbon from the vegetation to the soil (i.e., litterfall carbon, Lc). The accumulation of vegetation carbon in the CRU simulation is greater because NPP is substantially higher than Lc throughout the simulation period (Figure 7b) in comparison with the NCEP1 and MM5 simulations (Figures 7a and 7b).
In contrast to vegetation carbon, soil carbon increases substantially only in the CRU simulation during the 1990s (Figures 6c and 6d). Changes in soil carbon storage depend on both the input of carbon to the soil (Lc) and losses of carbon from the soil (Rh). In the CRU simulation, soil carbon in the 1980s decreases because Rh is greater than Lc, while in the 1990s CRU soil carbon increases because Rh is less than Lc (Figure 7b). The loss and accumulation of soil carbon in the CRU simulation from 1980 to 2000 is largely driven by the decrease in Rh over the time period (Figure 4c; Table 3).
The increase in total carbon storage is greater in the CRU simulations than in the NCEP1 simulation (Figures 6e and 6f), and the increase in both simulations is primarily driven by changes in vegetation carbon compared to changes in soil carbon. The increase in ratio of vegetation to soil carbon is most prominent in the CRU simulation during the 1980s (Figure 8a), with very modest increases in the ratio of vegetation to soil carbon for the NCEP1 simulation in the 1980s (Figure 8a) and in the CRU and NCEP1 simulations in the 1990s (Figure 8b). The substantial increase in the ratio of vegetation to soil carbon storage of the CRU simulation in the 1980s is driven by both increases in vegetation carbon and decreases in soil carbon during the 1980s. In contrast to the NCEP1 and CRU simulations during the 1990s, the MM5 simulations show little change in total carbon storage (Figure 6f) and a negligible change in the ratio of vegetation to soil carbon storage (Figure 8b).
4. Discussion
One of the objectives of the WALE project was to assess the ability of process-based ecosystem models to simulate CO2 exchange of terrestrial ecosystems in the western Arctic during the 1980s and 1990s. Uncertainty in the estimates made by process-based models has many sources and includes uncertainties in parameters, formulations of processes, conceptual issues about the interaction of processes, and driving datasets. We have conducted numerous studies to explore uncertainties in parameters, formulations of processes, and conceptual issues of TEM, the ecosystem model used in this study (e.g., McGuire et al. 1992; McGuire et al. 1993; McGuire et al. 1995; McGuire et al. 1997; Clein et al. 2000; McGuire et al. 2000b; Clein et al. 2002; Zhuang et al. 2003; Zhuang et al. 2006). A companion paper by Kimball et al. (Kimball et al. 2007) reports on a study that evaluated conceptual issues among two process-based ecosystem models used in the WALE project, TEM and Biome–Biogeochemical Cycles (BIOME–BGC), in simulating terrestrial CO2 exchange in the western Arctic during the 1980s and 1990s. That study used the NCEP1 to drive both models. Other studies in the WALE project have documented substantial differences among alternative climate datasets for the WALE region (e.g., Drobot et al. 2006; Herzfeld et al. 2007; Wu et al. 2007) and that these differences have consequences for estimating the dynamics of hydrology (Rawlins et al. 2006) and fire (Rupp et al. 2007) in the WALE region. Similar to the analyses of Rawlins et al. (Rawlins et al. 2006) and Rupp et al. (Rupp et al. 2007), this study extends the study of Kimball et al. (Kimball et al. 2007) to evaluate the consequences of alternative climate datasets for estimating the dynamics of terrestrial CO2 exchange in the WALE region. First, we discuss the consequences of alternative climate datasets for estimating carbon exchange and then we discuss the consequences for estimating changes in carbon storage.
Our simulations revealed the magnitude of estimates of NPP and Rh varied by a factor of 2.5 among the different climate datasets (Figure 2), with the estimates of NPP and Rh of the NCEP1 simulation intermediate between the lowest estimates of the MM5 simulation and the highest estimates of the CRU simulation. It appears that these differences are primarily associated with differences in nonwinter air temperature among the datasets (Figures 1a and 1b), as the differences in magnitude among the simulations are not consistent with the differences in precipitation and radiation among the different climate datasets (Figures 1c–f). For example, the low NPP of the MM5 simulation appears to be associated with a later thaw, earlier freeze up, and lower growing season temperature in comparison with the CRU and NCEP1 climates. Because Rh depends on both soil carbon storage, which is lowest in the MM5 simulation, and annual air temperature (Table 5), which is lowest in the MM5 climate, it is lowest in the simulation driven by MM5 climate.
Our evaluation suggests that differences in temperature can affect seasonality of photosynthetic carbon uptake in the spring, a result that is consistent with the analysis by Kimball et al. (Kimball et al. 2006) that shows that annual anomalies in NPP of the WALE region are highly correlated with the timing in spring thaw. Other studies (e.g., see Euskirchen et al. 2006) have also identified that the timing of spring thaw is an important factor in the uptake of carbon by vegetation across the pan-arctic/pan-boreal region. In contrast to the effects of different datasets on the magnitude of annual vegetation and soil CO2 exchange, the general magnitude of annual net CO2 exchange (i.e., NEP, NPP – Rh) was similar among the simulations driven by different climate sets. However, there were substantial differences in monthly net CO2 exchange among the datasets. The seasonal differences are important in that they can affect evaluations of how well the models are able to reproduce seasonal anomalies in atmospheric concentration of CO2 at high-latitude CO2 monitoring stations (e.g., see McGuire et al. 2000b; Dargaville et al. 2002a; Dargaville et al. 2002b; Zhuang et al. 2003).
Kimball et al. (Kimball et al. 2007) reported a small positive trend in NPP for the region from 1982 to 2000 for simulations of BIOME–BGC and TEM driven by the NCEP1 climate dataset. The results of this study indicate that the trend of NPP from 1980 to 2000 is stronger in the TEM simulation driven by NCEP1 (marginally significant) than in CRU (not significant). These trends appear to be caused by increases in summer air temperature (Table 3) in addition to changes in the timing of thaw (Kimball et al. 2007). Our simulations with TEM from 1980 to 2000 indicated that Rh did not vary substantially in the NCEP1 simulation because of a weak nonsignificant trend in mean annual air temperature and annual precipitation, but that substantial decreases in the CRU simulation were largely associated with decreasing precipitation, which may have resulted in decreasing summer soil moisture to reduce Rh across the two decades.
Trends in NEP of the WALE region from 1980 to 2000 in the CRU simulation were associated with increases in NPP coupled with decreases in Rh. In contrast, in the NCEP1 simulation the marginally significant increase in NEP was primarily associated with an increase in regional NPP. There is much spatial variability in the trends of NPP and Rh across the region in both the NCEP1 and the CRU simulations, and the spatial pattern of trends is quite different between the simulations. Thompson et al. (Thompson et al. 2006) applied TEM in the tundra–forest transition area of northern Alaska between 1981 and 2000 and found that trends in NEP depended on differential responses of NPP and Rh to spatial variability in whether the climate is getting warmer or colder and wetter or drier. Thus, our study and that of Thompson et al. (Thompson et al. 2006) indicate that the overall response of NEP in simulations with TEM across the WALE region depends substantially on the accuracy of temperature and precipitation temporal trends in the climate dataset used to drive the model.
Kimball et al. (Kimball et al. 2007) documented that vegetation and soil carbon storage increased in both TEM and BIOME–BGC simulations that were driven by NCEP1 climate from 1981 to 2000 such that the ratio of vegetation to soil carbon storage increased. We found that this pattern was even stronger for the CRU climate because of stronger increases in vegetation carbon across the two decades and a decline in soil carbon in the 1980s. In contrast, the MM5 simulation showed little change in carbon storage and little change in the ratio of vegetation to soil carbon during the 1990s. Kimball et al. (Kimball et al. 2007) concluded that the simulations of TEM and BIOME–BGC both indicate that the terrestrial carbon cycle is accelerating in the WALE region with a more rapid turnover of carbon. Our study supports this conclusion for simulations driven by both NCEP1 and CRU climates, but not for simulations driven by the MM5 climate.
5. Conclusions
The response of carbon dynamics of high-latitude ecosystems to climate change is of concern because it has the potential to act as a positive feedback to warming if large stores of soil carbon are released in response to warming (McGuire and Chapin 2006; McGuire et al. 2006; McGuire et al. 2007b). Ecological models of carbon dynamics are tools that can be used to evaluate the potential for high-latitude ecosystems to act as a positive or negative feedback to climate (e.g., Zhuang et al. 2006). Our retrospective study identified that simulated exchanges of CO2 and changes in carbon storage over the WALE region from 1980 to 2000 are very sensitive to alternative climate datasets used to the drive the simulations. Rawlins et al. (Rawlins et al. 2006) found that simulated water fluxes over the WALE region were substantially different among simulations driven by alternative climate datasets. Similarly, Rupp et al. (Rupp et al. 2007) found that simulated fire dynamics over the WALE region were substantially different among simulations driven by alternative datasets. All of these results have similar implications for coupling models of terrestrial dynamics with prognostic climate models. Both Drobot et al. (Drobot et al. 2006) and Wu et al. (Wu et al. 2007) have noted that while the mean state of climate variables often differ among alternative climate datasets for the WALE region, the temporal anomalies are often highly correlated. We agree with the recommendation of Rupp et al. (Rupp et al. 2007) that coupling methodologies should use anomalies of future climate model simulations to alter the climate of more trusted datasets. This approach has the potential to properly represent the sensitivity of carbon dynamics simulated by ecosystem models to alternate scenarios of future climatic change.
Acknowledgments
This work was supported by the NSF Arctic System Science Program as part of the Western Arctic Linkage Experiment and Fire Mediated Changes in the Arctic Project (OPP-0095024 and OPP-0328282).
REFERENCES
Amthor, J. S. Coauthors 2001. Boreal forest CO2 exchange and evapotranspiration predicted by nine ecosystem process models: Intermodel comparisons and relationships to field measurements. J. Geophys. Res. 106:33623–33648.
Calef, M. P., A. D. McGuire, H. E. Epstein, T. S. Rupp, and H. H. Shugart. 2005. Analysis of vegetation distribution in Interior Alaska and sensitivity to climate change using a logistic regression approach. J. Biogeogr. 32:863–878.
Clein, J., B. Kwiatkowski, A. D. McGuire, J. E. Hobbie, E. B. Rastetter, J. M. Melillo, and D. W. Kicklighter. 2000. Modeling carbon responses of tundra ecosystems to historical and projected climate: A comparison of a plot- and a global-scale ecosystem model to identify process-based uncertainties. Global Change Biol. 6:(Suppl. 1). 127–140.
Clein, J., A. D. McGuire, X. Zhang, D. W. Kicklighter, J. M. Melillo, S. C. Wofsy, P. G. Jarvis, and J. M. Massheder. 2002. Historical and projected carbon balance of mature black spruce ecosystems across North America: The role of carbon-nitrogen interactions. Plant Soil 242:15–32.
Dargaville, R., A. D. McGuire, and P. Rayner. 2002a. Estimates of large-scale fluxes in high latitudes from terrestrial biosphere models and an inversion of atmospheric CO2 measurements. Climatic Change 55:273–285.
Dargaville, R. J. Coauthors 2002b. Evaluation of terrestrial carbon cycle models with atmospheric CO2 measurements: Results from transient simulations considering increasing CO2, climate, and land-use effects. Global Biogeochem. Cycles 16.1092, doi:10.1029/2001GB001426.
Drobot, S., J. Maslanik, U. C. Herzfeld, C. Fowler, and W. Wu. 2006. Uncertainty in temperature and precipitation datasets over terrestrial regions of the Western Arctic. Earth Interactions 10.[Available online at http://EarthInteractions.org; see WALE Special Theme.].
Euskirchen, S. E. Coauthors 2006. Importance of recent shifts in soil thermal dynamics on growing season length, productivity, and carbon sequestration in terrestrial high-latitude ecosystems. Global Change Biol. 12:731–750.
FAO/UNESCO 1974. Soil Map of the World. Maps I–X.
Fleming, M. D. 1997. A statewide vegetation map of Alaska using phenological classification of AVHRR data. Proc. Second Circumpolar Arctic Vegetation Mapping Workshop and the CAVM-North America Workshop, Arendal, Norway, Colorado University, Institute of Arctic and Alpine Research, 25–26.
Goetz, S. J., A. G. Bunn, G. J. Fiske, and R. A. Houghton. 2005. Satellite-observed photosynthetic trends across boreal North America associated with climate and fire disturbance. Proc. Natl. Acad. Sci. USA 102:13521–13525.
Goulden, M. L., J. W. Munger, S-M. Fan, B. C. Daube, and S. C. Wofsy. 1996. CO2 exchange by a deciduous forest: Response to interannual climate variability. Science 271:1576–1578.
Herzfeld, U. C., S. Drobot, W. Wu, C. Fowler, and J. Maslanik. 2007. Spatiotemporal climate model validation: Case studies for MM5 over northwestern Canada and Alaska. Earth Interactions in press; see WALE Special Theme.
Jones, P. D., T. J. Osborn, K. R. Briffa, C. K. Folland, B. Horton, L. V. Alexander, D. E. Parker, and N. A. Rayner. 2001. Adjusting for sampling density in grid-box land and ocean surface temperature time series. J. Geophys. Res. 106:3371–3380.
Keeling, C. D., T. P. Whorf, M. Wahlen, and M. Pilcht. 1995. Interannual extremes in the rate of rise of atmospheric carbon dioxide since 1980. Nature 375:666–670.
Kimball, J. S., K. C. McDonald, and M. Zhao. 2006. Spring thaw and its effect on terrestrial vegetation productivity in the western Arctic observed from satellite microwave and optical remote sensing. Earth Interactions 10.[Available online at http://EarthInteractions.org; see WALE Special Theme.].
Kimball, J. S. Coauthors 2007. Recent climate-driven increases in vegetation productivity for the western Arctic: Evidence of an acceleration of the northern terrestrial carbon cycle. Earth Interactions 11.[Available online at http://EarthInteractions.org; see WALE Special Theme.].
Kistler, R. Coauthors 2001. The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc. 82:247–267.
McGuire, A. D. and F. S. Chapin III. 2006. Climate feedbacks in the Alaskan boreal forest. Alaska’s Changing Boreal Forest, F. S. Chapin III et al., Eds., Oxford University Press, 309–322.
McGuire, A. D., J. M. Melillo, L. A. Joyce, D. W. Kicklighter, A. L. Grace, B. I. Moore, and C. J. Vorosmarty. 1992. Interactions between carbon and nitrogen dynamics in estimating net primary productivity for potential vegetation in North America. Global Biogeochem. Cycles 6:101–124.
McGuire, A. D., L. A. Joyce, D. W. Kicklighter, J. M. Melillo, G. Esser, and C. J. Vorosmarty. 1993. Productivity response of climax temperate forests to elevated temperature and carbon dioxide: A North American comparison between two global models. Climatic Change 24:287–310.
McGuire, A. D., J. M. Melillo, D. W. Kicklighter, and L. A. Joyce. 1995. Equilibrium responses of soil carbon to climate change: Empirical and process-based estimates. J. Biogeogr. 22:785–796.
McGuire, A. D. Coauthors 1997. Equilibrium responses of global net primary production and carbon storage to doubled atmospheric carbon dioxide: Sensitivity to changes in vegetation nitrogen concentration. Global Biogeochem. Cycles 11:173–189.
McGuire, A. D., J. Clein, J. M. Melillo, D. W. Kicklighter, R. A. Meier, C. J. Vorosmarty, and M. C. Serreze. 2000a. Modeling carbon responses of tundra ecosystems to historical and projected climate: The sensitivity of pan-arctic carbon storage to temporal and spatial variation in climate. Global Change Biol. 6:(Suppl. 1). 141–159.
McGuire, A. D. Coauthors 2000b. Modeling the effects of snowpack on heterotrophic respiration across northern temperate and high latitude regions: Comparison with measurements of atmospheric carbon dioxide in high latitudes. Biogeochemistry 48:91–114.
McGuire, A. D., F. S. Chapin III, J. E. Walsh, and C. Wirth. 2006. Integrated regional changes in arctic climate feedbacks: Implications for the global climate system. Annu. Rev. Environ. Resour. 31:61–91.
McGuire, A. D. Coauthors 2007a. The Western Arctic Linkage Experiment: Overview and synthesis. Earth Interactions submitted; see WALE Special Theme.
McGuire, A. D. Coauthors 2007b. Responses of high latitude ecosystems to global change: Potential consequences for the climate system. Terrestrial Ecosystems in a Changing World, IGBP Series, Springer-Verlag, 297–310.
Melillo, J. M., D. W. Kicklighter, A. D. McGuire, W. T. Peterjohn, and K. M. Newkirk. 1995. Global change and its effects on soil organic carbon stocks. Role of Nonliving Organic Matter in the Earth’s Carbon Cycle, R. G. Zepp and C. Sonntag, Eds., John Wiley & Sons, 175–189.
Mitchell, T. D. and P. D. Jones. 2005. An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int. J. Climatol. 25:693–712.
Pan, Y., A. D. McGuire, D. W. Kicklighter, and J. M. Melillo. 1996. The effects of climate and soil data on estimates of net primary production: A sensitivity analysis with the Terrestrial Ecosystem Mode. Global Change Biol. 2:5–23.
Pan, Y. Coauthors 1998. Modeled responses of terrestrial ecosystems to elevated atmospheric CO2: A comparison of simulations by the biogeochemistry models of the Vegetation/Ecosystem Modeling and Analysis Project (VEMAP). Oecolgia 114:389–404.
Potter, C. Coauthors 2001. Comparison of boreal ecosystem model sensitivity to variability in climate and forest site parameters. J. Geophys. Res. 106:33671–33688.
Raich, J. W. Coauthors 1991. Potential net primary productivity in South America: Application of a global model. Ecol. Appl. 1:399–429.
Rawlins, M. A., S. Frolking, R. B. Lammers, and C. J. Vorosmarty. 2006. Effects of uncertainty in climate inputs on simulated evapotranspiration and runoff in the Western Arctic. Earth Interactions 10.[Available online at http://EarthInteractions.org; see WALE Special Theme.].
Rupp, T. S., X. Chen, M. Olson, and A. D. McGuire. 2007. Sensitivity of simulated boreal fire dynamics to uncertainties in climate drivers. Earth Interactions 11.[Available online at http://EarthInteractions.org; see WALE Special Theme.].
Ryan, M. G., M. B. Lavigne, and S. T. Gower. 1997. Annual carbon cost of autotrophic respiration in boreal forest ecosystems in relation to species and climate. J. Geophys. Res. 102:D24. 28871–28883.
Serreze, M. C. and C. M. Hurst. 2000. Representation of mean Arctic precipitation from NCEP–NCAR and ERA reanalyses. J. Climate 13:182–201.
Serreze, M. C., J. R. Key, J. C. Box, and J. A. Maslanik. 1998. A new monthly climatology of global radiation for the Arctic and comparisons with NCEP–NCAR reanalysis and ISCCP-C2 fields. J. Climate 11:121–136.
Serreze, M. C., M. P. Clark, and D. H. Bromwich. 2003. Monitoring precipitation over the Arctic terrestrial drainage system: Data requirements, shortcomings, and applications of atmospheric reanalysis. J. Hydrometeor. 4:387–407.
Shaver, G. R., W. D. Billings, F. S. Chapin, A. E. Giblin, K. J. Nadlehoffer, W. C. Oechel, and E. B. Rastetter. 1992. Global change and the carbon balance of arctic ecosystems. Bioscience 42:433–441.
Sitch, S. Coauthors 2007. Assessing the carbon balance of circumpolar arctic tundra using remote sensing and process modeling. Ecol. Apps 17:213–234.
Thompson, C. D., A. D. McGuire, J. S. Clein, F. S. Chapin III, and J. Beringer. 2006. Net carbon exchange across the arctic tundra-boreal forest transition in Alaska 1981–2000. Mitig. Adapt. Strat. Global Change 11:805–827.
Tian, H., J. M. Melillo, D. W. Kicklighter, A. D. McGuire, and J. Helfrich. 1999. The sensitivity of terrestrial carbon storage to historical climate variability and atmospheric CO2 in the United States. Tellus 51B:414–452.
Van Cleve, K., L. K. Oliver, B. Schlentner, L. A. Viereck, and C. T. Dyrness. 1983. Productivity and nutrient cycling in taiga forest ecosystems. Can. J. For. Res. 13:747–766.
Wu, W., A. H. Lynch, S. Drobot, J. Maslanik, A. D. McGuire, and U. Herzfeld. 2007. Comparative analysis of the Western Arctic surface climate among observations and model simulations. Earth Interactions 11.[Available online at http://EarthInteractions.org; see WALE Special Theme.].
Zhuang, Q., A. D. McGuire, K. P. O’Neill, J. W. Harden, V. E. Romanovsky, and J. Yarie. 2002. Modeling soil thermal and carbon dynamics of a fire chronosequence in interior Alaska. J. Geophys. Res. 108.8147, doi:10.1029/2001JD001244.
Zhuang, Q. Coauthors 2003. Carbon cycling in extratropical terrestrial ecosystems of the Northern Hemisphere during the 20th century: A modeling analysis of the influences of soil thermal dynamics. Tellus 55B:751–776.
Zhuang, Q., J. M. Melillo, D. W. Kicklighter, R. G. Prinn, A. D. McGuire, P. A. Steudler, B. S. Felzer, and S. Hu. 2004. Methane fluxes between terrestrial ecosystems and the atmosphere at northern high latitudes during the past century: A retrospective analysis with a process-based biogeochemistry model. Global Biogeochem. Cycles 18.GB3010, doi:10.1029/2004GB002239.
Zhuang, Q. Coauthors 2006. CO2 and CH4 exchanges between land ecosystems and the atmosphere in northern high latitudes over the 21st Century. Geophys. Res. Lett. 33.L17403, doi:10.1029/2006GL026972.
Annual and monthly climate inputs used to drive the Terrestrial Ecosystem Model in this study. (a) Mean annual air temperature (TAIR), (b) mean monthly TAIR, (c) total annual precipitation (PREC), (d) total monthly PREC, (e) mean annual downwelling radiation (NIRR) from 1980 to 2000 for NCEP1 and CRU, and from 1990 to 2000 for MM5, and (f) mean monthly NIRR from 1990 to 2000 for all climates.
Citation: Earth Interactions 11, 12; 10.1175/EI229.1
Comparison of annual carbon fluxes (a) NPP, (b) Rh, and (c) NEP over the WALE region in simulations with different climates.
Citation: Earth Interactions 11, 12; 10.1175/EI229.1
Seasonality of carbon fluxes (a) NPP, (b) Rh, and (c) NEP from 1990 to 2000 over the WALE region using different climates.
Citation: Earth Interactions 11, 12; 10.1175/EI229.1
Comparison of annual anomalies in TEM simulations for different climates from 1980 to 2000 for (a) NPP, (c) Rh, and (e) NEP and from 1990 to 2000 for (b) NPP, (d) Rh, and (f) NEP.
Citation: Earth Interactions 11, 12; 10.1175/EI229.1
Linear trends in NPP and Rh from 1990 to 2000 over the western Arctic for simulations with TEM driven by different climates: (a) NCEP1 NPP, (b) CRU NPP, (c) MM5 NPP, (d) NCEP1 Rh, (e) CRU Rh, and (f) MM5 Rh.
Citation: Earth Interactions 11, 12; 10.1175/EI229.1
Cumulative change in carbon stocks simulated by TEM for different climates from 1980 to 2000 for (a) vegetation carbon (vegC), (c) soil organic carbon (soilC), and (e) total ecosystem carbon (totalC) and from 1990 to 2000 for (b) vegC, (d) soilC, and (f) totalC.
Citation: Earth Interactions 11, 12; 10.1175/EI229.1
Comparison of NPP, Rh, and litterfall carbon (Lc) for TEM simulations driven by (a) NCEP1, (b) CRU, and (c) MM5 climate inputs.
Citation: Earth Interactions 11, 12; 10.1175/EI229.1
Comparison of vegetation/soilC fraction anomaly as percent for TEM simulations (a) from 1980 to 2000 driven by NCEP1 and CRU, and (b) from 1990 to 2000 driven by NCEP1, CRU, and MM5 climates.
Citation: Earth Interactions 11, 12; 10.1175/EI229.1
Pools and fluxes used to calibrate the rate-limiting parameters of the application of the Terrestrial Ecosystem Model in this study.
Source and comments for the pools and fluxes of dry boreal coniferous forest and boreal deciduous forest in Table 1.
Spatially averaged temporal trends (g C m−2 yr−1 yr−1) in annual NPP, Rh, and NEP estimated for the WALE region in simulations driven by different climate datasets (NCEP1, CRU, and MM5); trends for variables are reported for mean annual and mean summer air temperature (TAIR; °C yr−1), annual and summer precipitation (PREC; mm yr−1), mean annual and mean summer downwelling shortwave radiation at the top of the canopy (NIRR; W m−2 yr−1), annual and summer potential evapotranspiration (PET; mm yr−1), and annual and summer precipitation minus potential evapotranspiration (P-PET; mm yr−1). Probability that no trend exists is reported in parentheses. Bolded values are significant at p < 0.05.
Spatially averaged sensitivity in annual NPP to different climate variables from simulations for the WALE region driven by climate variables from different climate datasets (NCEP1, CRU, and MM5); sensitivity to variables are reported for mean annual and mean summer air temperature [TAIR; g C m−2 yr−1 (°C)−1], annual and summer precipitation [PREC; g C m−2 yr−1 (mm)−1], mean annual and mean summer downwelling shortwave radiation at the top of the canopy [NIRR; g C m−2 yr−1 (W m−2)−1], annual and summer potential evapotranspiration [PET; g C m−2 yr−1 (mm)−1], and annual and summer precipitation minus potential evapotranspiration [P-PET; g C m−2 yr−1 (mm)−1]. Probability that no trend exists is reported in parentheses. Bolded values are significant at p < 0.05.
