Abstract

Transient simulations are presented of future climate and vegetation associated with continued rising levels of CO2. The model is a fully coupled atmosphere–ocean–land–ice model with dynamic vegetation. The impacts of the radiative and physiological forcing of CO2 are diagnosed, along with the role of vegetation feedbacks. While the radiative effect of rising CO2 produces most of the warming, the physiological effect contributes additional warming by weakening the hydrologic cycle through reduced evapotranspiration. Both effects cause drying over tropical rain forests, while the radiative effect enhances Arctic and Indonesian precipitation.

A global greening trend is simulated primarily due to the physiological effect, with an increase in photosynthesis and total tree cover associated with enhanced water-use efficiency. In particular, tree cover is enhanced by the physiological effect over moisture-limited regions. Over Amazonia, South Africa, and Australia, the radiative forcing produces soil drying and reduced forest cover. A poleward shift of the boreal forest is simulated as both the radiative and physiological effects enhance vegetation growth in the northern tundra and the radiative effect induces drying and summertime heat stress on the central and southern boreal forest. Vegetation feedbacks substantially impact local temperature trends through changes in albedo and evapotranspiration. The physiological effect increases net biomass across most land areas, while the radiative effect results in an increase over the tundra and decrease over tropical forests and portions of the boreal forest.

1. Introduction

As atmospheric concentrations of CO2 continue to rise, substantial changes to the earth system are expected. GCM simulations for the Intergovernmental Panel on Climate Change (IPCC) (Cubasch et al. 2001) are projecting a mean global warming of 1.4°–5.8°C during 1990–2100. These simulations predict enhanced warming over land, particularly over the northern portions of North America and Asia (Cubasch et al. 2001). They generally simulate an intensified hydrologic cycle, diminished Northern Hemispheric (NH) snow cover and sea ice extent, and a weakened oceanic thermohaline circulation (Cubasch et al. 2001). However, the IPCC simulations neglect the role of vegetation feedbacks.

These projected climatic changes will likely impact the biosphere, while vegetation changes will produce feedbacks on the climate. Various observational datasets hint at ongoing ecological changes, potentially related to climatic change. There is evidence of an expanding growing season over the boreal zone and Europe, based on satellite-derived vegetation data (Myneni et al. 1997; Zhou et al. 2001) and phenological data (Menzel and Fabian 1999), respectively. Modeling studies have suggested significant vegetation feedback on climate (e.g., Bonan et al. 1992; Pielke et al. 1998; Bonan 2002). Recent analyses of vegetation and climate variability, derived from remote sensing and observed climate data, have revealed significant vegetation feedbacks on climate, especially over the boreal forest (Liu et al. 2006; Notaro et al. 2006).

The impact of rising levels of CO2 can be divided into two forcings: radiative and physiological. The direct radiative effect on the atmosphere produces warming, which can extend the growing season in colder environments, induce drying and vegetation dieback in water-limited regions, and incur heat stress over boreal forests. The physiological effect of rising CO2 on vegetation (i.e., the CO2 fertilization effect) enhances photosynthesis, particularly among C3 plants (Curtis 1996; Koch and Mooney 1996; Mooney et al. 1999). Changes in plant structure, physiology, competition, and distribution can be expected due to higher temperatures and levels of CO2 (Pritchard et al. 1999). For instance, an expansion of boreal forests into snow-covered tundra would reduce the surface albedo and enhance the regional warming (Foley et al. 1994; Levis et al. 1999; Foley et al. 2000; Notaro et al. 2005, 2006; Liu et al. 2006) due to a positive albedo feedback (Robinson and Kukla 1985; Bonan et al. 1992; Laine and Heikinheimo 1996; Sharratt 1998; Levis et al. 2000). By enhancing vegetation amount through the physiological effect, vegetation feedbacks on the atmosphere can then produce additional climatic changes.

Plants adjust their stomatal openings in order to maximize their intake of CO2 and minimize water loss, both necessary ingredients for photosynthesis. Plant water-use efficiency is defined as the ratio of photosynthetic uptake of CO2 to the canopy latent heat flux. The physiological effect of rising levels of CO2 results in plants narrowing their stomatal openings, thereby reducing water loss through evapotranspiration and increasing the plant water-use efficiency (Polley et al. 1993; Field et al. 1995; Curtis 1996; Sellers et al. 1996; Drake et al. 1997; Farquhar 1997; Körner 2000). The decrease in evapotranspiration and associated increase in sensible heat flux result in a weakened hydrologic cycle, reduced precipitation, and slight warming, due to the physiological effect of rising CO2 (Cox et al. 2004; Levis et al. 2000). Several modeling studies noted a decrease in evapotranspiration over tropical landmasses due to the physiological effect (Betts et al. 1997; Levis et al. 2000; Betts et al. 2004). The physiological forcing produces two opposing effects (Betts et al. 1997). One effect is an increase in leaf area due to enhanced photosynthesis, resulting from higher levels of CO2. The opposing effect is a reduction in leaf area resulting from stomatal closure and a subsequent decrease in evapotranspiration, which reduces precipitation and leads to less plant growth.

Previous modeling studies have investigated the impact of predicted rising levels of CO2 on vegetation and climate (Table 1). Sellers et al. (1996), Betts et al. (1997), and Bounoua et al. (1999) found that the physiological effect reduced stomatal conductance and evapotranspiration, leading to additional warming over land. Bounoua et al. (1999) concluded that the radiative effect produced most of the warming, increased global precipitation, enhanced photosynthesis over the Tropics, and reduced photosynthesis over the northern latitudes from heat stress. Cramer et al. (2001) showed that the physiological effect can produce an expansion of forest into savanna and grassland into arid tropical regions. Levis et al. (1999) found that vegetation feedbacks enhance the warming across the northern high latitudes during spring and summer, by reducing albedo, but can also counter wintertime warming by increasing surface emissivity and infrared heat loss when boreal trees mask snow cover. Bergengren et al. (2001) simulated a poleward migration of the boreal forest into tundra related to the albedo feedback and spread of temperate grasslands into the southern boreal zone due to greater summertime warming.

Table 1.

Studies of future changes in vegetation and climate.

Studies of future changes in vegetation and climate.
Studies of future changes in vegetation and climate.

Three studies in Table 1 focused on Amazonia. Costa and Foley (2000) showed increased evapotranspiration due to the radiative effect and decreased evapotranspiration due to the physiological effect, while both effects contributed separately to a rise in Amazonia precipitation. However, later studies by Betts et al. (2004) and Cox et al. (2004) used the Third Hadley Centre Coupled Climate–Carbon Cycle General Circulation Model (HadCM3LC) Top-Down Representation of Interactive Foliage and Flora including Dynamics (TRIFFID) with dynamic vegetation and an interactive carbon cycle and concluded that both effects and the inclusion of these dynamic feedbacks produced a reduction in precipitation and forest dieback.

The present study is the first to analyze the global impacts of rising levels of CO2, out to 4 × CO2, on climate and vegetation using transient simulations from a fully coupled atmosphere–ocean circulation model with dynamic vegetation. Ensemble transient simulations, which are unique among coupled vegetation–climate studies, are performed using the Fast Ocean Atmosphere Model–Lund–Potsdam–Jena (FOAM–LPJ) (Notaro et al. 2005; Gallimore et al. 2005) to study changes in climate (e.g., temperature, hydrologic cycle) and vegetation (e.g., biome patterns, vegetation cover, biomass, fires). The individual roles of radiative and physiological forcing and the impact of vegetation feedbacks are investigated.

Section 2 describes the model and simulations. A brief overview of simulated vegetation changes is provided in section 3a to offer insight into the climatic changes discussed in sections 3bd. A detailed description of vegetation changes is given in section 3e. The conclusions are in section 4.

2. Methods

a. Model description

FOAM–LPJ is a fully coupled global atmosphere–ocean–land model with dynamic vegetation. Details of the model and its performance are presented by Notaro et al. (2005) and Gallimore et al. (2005). The coupled atmospheric–oceanic component is FOAM version 1.5 (Jacob 1997). The atmospheric component, the Parallel Community Climate Model version 3 (PCCM3) University of Wisconsin—Madison (UW) (Drake et al. 1995; Jacob 1997), uses a horizontal resolution of R15 and 18 vertical levels, while the z-coordinate oceanic component, the Ocean Model version 3 (OM3), uses a horizontal resolution of 1.4° latitude × 2.8° longitude and 24 vertical levels. FOAM runs efficiently on parallel computing systems (Jacob et al. 2001) and produces a steady long-term climate without using flux adjustment.

FOAM was synchronously coupled to a modified version of the Lund–Potsdam–Jena dynamical global vegetation model (LPJ-DGVM) (Sitch 2000; Cramer et al. 2001; McGuire et al. 2001; Sitch et al. 2003). The land grid has a horizontal resolution of 1.4° latitude × 2.8° longitude. Nine plant function types (PFTs) are simulated: two tropical trees, three temperate trees, two boreal trees, and two grasses (Table 1 of Notaro et al. 2005). FOAM–LPJ vegetation processes include plant competition, biomass allocation, establishment, mortality, soil and litter biogeochemistry, natural fire, and successional vegetation changes. Details of the coupling are outlined by Gallimore et al. (2005) and in Fig. 1 of Notaro et al. (2005). No relaxation of climate forcing toward observations is applied to adjust the simulated vegetation. The simulations do not include nutrient limitations (Field et al. 1995) or forcings from land use, volcanic eruptions, aerosols, or solar activity.

FOAM produces reasonable climate variability, including ENSO (Liu et al. 2000; Liu and Wu 2004), Pacific decadal variability (Wu et al. 2003; Wu and Liu 2003), and tropical Atlantic variability (Wu and Liu 2002), although the simulated variability is generally less than observed. The biome distribution simulated by FOAM–LPJ is in reasonable agreement with the expected potential natural vegetation distribution (Notaro et al. 2005; Gallimore et al. 2005).

b. Simulations

Twelve transient simulations are performed, each of which are 144 years in duration, with CO2 increasing 1% yr−1 from its 1975 level (335 ppm) to 4 × CO2 (Table 2). The primary simulations, RP, consist of a three-member ensemble including both the radiative and physiological forcings from rising CO2. An additional sixty years were run at the end of RP1, with the CO2 level set to 4 × CO2, to allow the climate to reach equilibrium. The individual impacts of the radiative and physiological effects are revealed through three-member ensembles, R and P, respectively. In FOAM–LPJ, the CO2 level used for photosynthesis in the vegetation model (physiological value) can be set separately from the value used in the radiation model (radiative value). While ensembles RP, R, and P incorporate dynamic vegetation, ensemble RFIX considers the radiative effect under fixed vegetation cover conditions. All ensemble runs are performed as extensions from an equilibrium control with 1975 CO2 level meridional overturning circulation (MOC) (Notaro et al. 2005), using different initial conditions.

Table 2.

Summary of the FOAM–LPJ simulations used in this study.

Summary of the FOAM–LPJ simulations used in this study.
Summary of the FOAM–LPJ simulations used in this study.

In the RFIX experiments, the daily processes of LPJ coupling are permitted while the annual part of the coupling, which determines PFT fractional coverage, is turned off (Gallimore et al. 2005). The PFT distribution in the initial restart file is replaced by the mean coverage from the MOC control run and held constant throughout the RFIX simulations. Changes in seasonal leaf cover (phenology) are permitted in these runs.

c. Trends and significance

Changes in climatic and ecological variables are computed using linear regression, multiplying the slope of the regression line by 144 years. To test for trend significance, we apply the two-tailed Wigley and Santer (1990) statistical test, NT5. This test evaluates the local difference in time means, grid point by grid point, providing the fraction of grid points with significant differences at the 5% level. Wigley and Santer consider standard statistical tests to be inadequate for testing a change in the mean due to problems of multiplicity, spatial autocorrelation, and unknown sampling distributions. The NT5 test for spatial field comparison evaluates the likelihood that a change does not occur by chance (Narisma and Pitman 2003). In comparing sample fields, d and m, the test statistic is computed as t = (dxmx)/Sx, where the variance is S2x = (s2d,x + s2m,x)/ (nt − 1). The number of data points in time is nt. We compare the mean climate and vegetation during the first and last 30 years of the simulations.

The significance is tested using the resampling, pooled-permutation procedure (PPP) of Preisendorfer and Barnett (1983). Given a sufficient number of randomized resamplings, a null sampling distribution is determined against which the test statistics are compared, resulting in a p value. For each grid point, we use 500 randomizations of the time ordering of d and m, compute a distribution of test statistics, and then compare the actual test statistic at that grid point against the distribution to determine the p value. For assessing overall field significance, the p value is determined as the fractional number of successes, in which the local difference is assessed as statistically significant (Table 3). We apply the NT5 test and PPP rigorously by only identifying regions in figures that are statistically significant (p < 0.05) in all three ensemble members.

Table 3.

Changes in the annual average global, Northern Hemisphere, and Southern Hemisphere area mean of various variables in the RP, R, P, and RFIX ensembles. Land + ocean variables include surface air temperature (T), precipitation (precip), and evaporation (evap). Land variables include top-layer soil moisture (SM), evapotranspiration (ET), percent total cloud cover (CLDTOT), percent low cloud cover (CLDLOW), percent tree cover, and percent grass cover. Ocean variables include SST and sea ice area. Percent changes are based on linear trend for the 144 years, while absolute changes are computed as the difference between the last 30 years and first 30 years. Italics indicate p < 0.05, based on Student’s t test. Bold text indicates which ensemble produced the largest change for each variable.

Changes in the annual average global, Northern Hemisphere, and Southern Hemisphere area mean of various variables in the RP, R, P, and RFIX ensembles. Land + ocean variables include surface air temperature (T), precipitation (precip), and evaporation (evap). Land variables include top-layer soil moisture (SM), evapotranspiration (ET), percent total cloud cover (CLDTOT), percent low cloud cover (CLDLOW), percent tree cover, and percent grass cover. Ocean variables include SST and sea ice area. Percent changes are based on linear trend for the 144 years, while absolute changes are computed as the difference between the last 30 years and first 30 years. Italics indicate p < 0.05, based on Student’s t test. Bold text indicates which ensemble produced the largest change for each variable.
Changes in the annual average global, Northern Hemisphere, and Southern Hemisphere area mean of various variables in the RP, R, P, and RFIX ensembles. Land + ocean variables include surface air temperature (T), precipitation (precip), and evaporation (evap). Land variables include top-layer soil moisture (SM), evapotranspiration (ET), percent total cloud cover (CLDTOT), percent low cloud cover (CLDLOW), percent tree cover, and percent grass cover. Ocean variables include SST and sea ice area. Percent changes are based on linear trend for the 144 years, while absolute changes are computed as the difference between the last 30 years and first 30 years. Italics indicate p < 0.05, based on Student’s t test. Bold text indicates which ensemble produced the largest change for each variable.

The signal-to-noise ratio (SNR) is computed to assess the robustness of predicted trends. The SNR is determined as a ratio between the mean trend among the ensemble members and the standard deviation of these trends. In assessing changes in biome distribution, we compute kappa statistics (Cohen 1960; Monserud 1990; Prentice et al. 1992; Harrison et al. 1998). Here, the kappa statistic quantifies the agreement between two biome maps (modern versus future), determined by the fraction of the total number of grid cells assigned to the same biome category while accounting for the number of grid cells that might be the same by chance.

3. Results

a. Overview of vegetation changes

Statistically significant changes in tree cover and grass cover in ensembles RP, R, and P are shown in Fig. 1. The radiative effect reduces forest cover over the southern/central boreal forest, the Amazon, and southern Africa, while contributing to an expansion of forest into the NH tundra. Vast portions of the Asian and Canadian boreal forests are replaced by grasslands. The physiological effect enhances forest cover over drier regions, such as the tundra, the western United States, the Sahel, and Australia. These two forcings combined, in ensemble RP, lead to a poleward shift of the boreal forest, depletion of Amazon forest cover, and increased forest cover over the Sahel and western United States.

Fig. 1.

Change in (a), (c), (e) tree cover and (b), (d), (f) grass cover fraction in the (a), (b) RP; (c), (d) R; and (e), (f) P ensembles, computed based on the slope of a linear regression fit multiplied by 144 yr. Only those changes that achieve statistical significance (p < 0.05) with the NT5 test in all three ensemble members are shown.

Fig. 1.

Change in (a), (c), (e) tree cover and (b), (d), (f) grass cover fraction in the (a), (b) RP; (c), (d) R; and (e), (f) P ensembles, computed based on the slope of a linear regression fit multiplied by 144 yr. Only those changes that achieve statistical significance (p < 0.05) with the NT5 test in all three ensemble members are shown.

b. Climatic change: Temperature

The global mean surface air temperature increases 2.7°C (p < 0.05) in the RP transient ensemble, with a larger increase in the NH (3.4°C) than in the SH (1.9°C) (Table 3). The majority of this warming is associated with the radiative effect of increasing CO2, with a 2.3°C increase (p < 0.05) in global mean temperature in ensemble R. The physiological effect contributes an additional 0.4°C warming globally, with greater warming in the NH (0.5°C) than in the SH (0.2°C). Among ensemble members, the physiological effect produces a warming of 0.3°–0.4°C across the globe, 0.4°–0.7°C across the NH, and 0.1°–0.3°C across the SH, suggesting a consistent yet weak warming. Vegetation feedbacks, when averaged over the globe or either hemisphere, are not substantial in FOAM–LPJ, as evident by the small 0.1°C reduced global warming in ensemble R than ensemble RFIX. This is in relative agreement with the study by Zhao et al. (2001), in which simulations with natural land cover and present-day land cover are compared, and it is found that land-cover changes produce significant regional climate changes but do not impact the global mean temperature or precipitation.

Laboratory experiments and field observations generally show that an increase in CO2 enhances most plant photosynthesis and water-use efficiency, while reducing water vapor conductance and evapotranspiration through stomata (Polley et al. 1993; Field et al. 1995; Curtis 1996). Modeling studies have likewise shown a decrease in stomatal conductance and evapotranspiration and an increase in sensible heat flux. This results in a weaker hydrologic cycle and slight warming (Henderson-Sellers et al. 1995; Pollard and Thompson 1995; Sellers et al. 1996; Costa and Foley 2000; Levis et al. 2000; Notaro et al. 2005). Sellers et al. (1996) noted a reduction in evapotranspiration and increase in temperature over tropical continents. The present study shows similar impacts from CO2 fertilization. Unlike Sellers et al. (1996), the RP ensemble exhibits a reduction in global evapotranspiration (Table 3), with both radiative and physiological effects contributing to a substantial reduction over tropical landmasses.

The largest rise in temperature in the RP ensemble occurs within 60°–90°N, primarily attributed to the snow albedo feedback, with the zonal mean warming increasing poleward from 5° to 8°C within this zone (Fig. 2a). Warming within the Tropics mostly averages 2°–3°C. Cooling occurs along the edge of the Antarctic ice sheet due to enhanced freshening of the upper ocean, which leads to greater sea ice. For all experiments, the largest uncertainty among the ensemble members is found over the mid to high latitudes (Fig. 2b). While the radiative effect is dominant, the physiological effect contributes an additional 1°C warming in the northern high latitudes. However, the standard deviation of temperature changes among the P ensemble members ranges from 0.4° to 0.6°C at the northern mid to high latitudes (Fig. 2b), which is substantial compared to the mean predicted temperature change and thus bears some uncertainty. Ensemble R produces slightly less warming across the northern mid to high latitudes than ensemble RFIX, related to changes in boreal forest cover.

Fig. 2.

(a) Zonally averaged change in surface air temperature (°C) in 144-yr ensembles. (b) Standard deviation of temperature changes (°C) among ensemble members (“noise”). For both panels, results for RP, R, P, and RFIX are shown in thick solid, thick dashed, thin dashed, and thin solid lines, respectively.

Fig. 2.

(a) Zonally averaged change in surface air temperature (°C) in 144-yr ensembles. (b) Standard deviation of temperature changes (°C) among ensemble members (“noise”). For both panels, results for RP, R, P, and RFIX are shown in thick solid, thick dashed, thin dashed, and thin solid lines, respectively.

The spatial distribution of annual temperature change is shown in Fig. 3 for ensembles RP, R, P, and RFIX. Ensemble RP contains warming of at least 4°C across much of Canada, the United States, Europe, northern Asia, and the Amazon. Regional cooling is found south of Greenland and along the Antarctic ice shelf. The warming in ensemble R is similar in pattern but weaker than in ensemble RP. The predicted warming in ensembles RP, R, and RFIX is statistically significant nearly everywhere. The physiological effect in ensemble P produces warming globally, strongest over land and exceeding 0.8°C across the northern mid to high latitudes; the mid to high latitude warming is attributed to increased forest growth, which results in both increased energy absorption in the canopy and the masking of snow cover. The warming in ensemble P achieves statistical significance primarily over landmasses ranging from the Tropics to midlatitudes. Across the globe, the field significance of annual temperature trends yields p values among ensemble members ranging from 0.02–0.03 for RP, 0.02–0.04 for R, 0.03–0.03 for RFIX, to 0.33–0.50 for P. Ensemble P fails to achieve field significance but contains local areas of significant warming. Reduced warming is noted near Lake Bakal in ensemble R but mostly absent in ensemble RFIX, coinciding with an area of vast boreal tree loss in ensemble R. Meanwhile, increased vegetation in ensemble R leads to greater warming over northeast and northwest Russia in March–May (MAM) and the Canadian tundra in June–August (JJA) than in ensemble RFIX.

Fig. 3.

Change in annual surface air temperature (°C) in 144-yr ensembles. Results are shown for (a) RP, (b) R, (c) P, and (d) RFIX ensembles. The contour interval is 1°C for (a), (b), and (d) and 0.2°C for (c). The less strict criterion for ensemble P allows for easier viewing. Changes in this figure, and in later figures, are computed as the linear trend multiplied by 144 yr. Shading indicates statistical significance (p < 0.05) was achieved with the NT5 test in all three ensemble members.

Fig. 3.

Change in annual surface air temperature (°C) in 144-yr ensembles. Results are shown for (a) RP, (b) R, (c) P, and (d) RFIX ensembles. The contour interval is 1°C for (a), (b), and (d) and 0.2°C for (c). The less strict criterion for ensemble P allows for easier viewing. Changes in this figure, and in later figures, are computed as the linear trend multiplied by 144 yr. Shading indicates statistical significance (p < 0.05) was achieved with the NT5 test in all three ensemble members.

The predicted temperature trends in ensembles RP, R, and RFIX are robust, with high SNRs (Fig. 4), particularly over the Tropics. The largest SNRs in RP are found in the tropical east Pacific. While temperature trends are large over the mid to high latitudes in RP and R, the substantial atmospheric noise results in somewhat lower SNRs. The SNRs are slightly higher over the northern mid to high latitudes in ensemble RFIX than R due to the lack of vegetation-cover changes. Ensemble P exhibits substantially lower SNRs, peaking over tropical and subtropical landmasses. Additional physiological ensemble members could reduce this uncertainty.

Fig. 4.

Signal to noise ratio for surface air temperature trends in ensemble (a) RP, (b) R, and (c) P. Contours are 5, 10, 20, and 50. Right-hand panels show the percentage of grid points with SNR > 5 for each latitudinal band.

Fig. 4.

Signal to noise ratio for surface air temperature trends in ensemble (a) RP, (b) R, and (c) P. Contours are 5, 10, 20, and 50. Right-hand panels show the percentage of grid points with SNR > 5 for each latitudinal band.

The largest seasonal warming in ensemble RP occurs in December–February (DJF), with NH mean temperature increases of 4.0°, 2.9°, 3.2°, and 3.7°C in DJF, MAM, JJA, and SON, respectively (Fig. 5). Outside of JJA, the warming trend is largest over the Arctic Ocean and extreme northern Eurasia and North America. However, during JJA, the greatest warming of 4°–7°C is found over the midcontinents, where the physiological effect substantially enhances the warming trend. The local warming minimum over Siberia is most distinct during MAM and DJF, suggesting a link to snow albedo feedback. The global temperature trends in RP achieve field significance for all ensemble members during DJF, MAM, and JJA and for two members during SON.

Fig. 5.

Change in surface air temperature (°C) in the RP ensemble for (a) DJF, (b) MAM, (c) JJA, and (d) SON. Shading indicates statistical significance (p < 0.05) was achieved with the NT5 test in all three ensemble members.

Fig. 5.

Change in surface air temperature (°C) in the RP ensemble for (a) DJF, (b) MAM, (c) JJA, and (d) SON. Shading indicates statistical significance (p < 0.05) was achieved with the NT5 test in all three ensemble members.

The physiological effect in ensemble P produces statistically significant warming of 1°–2.5°C, related to a rise in Bowen ratio, across most of North America and Eurasia during JJA, contributing toward the inland warming maxima (Fig. 6). This is somewhat comparable to the 3°–4°C warming over the same continents attributed to the radiative effect in ensemble R. Clearly, GCMs need to include the physiological effect of rising CO2 for studies of climate change. Despite achieving statistical significance, the areas of warming over land in ensemble P generally only reach a relatively low, yet substantial, SNR ≈ 5. Statistically significant warming in excess of 0.5°C is found over the tropical landmasses in ensemble P during JJA, resulting from a weakening of the hydrological cycle and evapotranspiration and in turn an increased Bowen ratio. The slight warming over Amazonia attributed to the physiological effect agrees with Costa and Foley’s (2000) study. Levis et al. (1999) concluded that CO2 fertilization amplifies the radiative warming over northern mid to high latitude land, especially in spring, but reduces the warming in winter. However, the present study found that it amplifies the warming in every season, particularly in summer.

Fig. 6.

Change in JJA surface air temperature (°C) for the (a) RP, (b) R, (c) P, and (d) RFIX ensembles. Contour interval is 1°C for (a), (b), and (d) and 0.3°C for (c). Shading indicates statistical significance (p < 0.05) was achieved with the NT5 test in all three ensemble members.

Fig. 6.

Change in JJA surface air temperature (°C) for the (a) RP, (b) R, (c) P, and (d) RFIX ensembles. Contour interval is 1°C for (a), (b), and (d) and 0.3°C for (c). Shading indicates statistical significance (p < 0.05) was achieved with the NT5 test in all three ensemble members.

Over land, the largest difference in annual temperature trend between ensembles R and RFIX is found over Mongolia and southern Russia, representing the southern edge of the simulated boreal forest. Ensemble R produces 0.3°–1.0°C less warming than ensemble RFIX, attributed to a loss of boreal tree cover in ensemble R. In MAM, ensemble R simulates up to 1.5°C less warming (p < 0.05) than ensemble RFIX around Lake Bakal (Fig. 7). The SNR of the difference in MAM temperature trends between R and RFIX peaks over central Asia (SNR ≈ 1–5), representing an area of substantial difference in warming attributed to changing vegetation.

Fig. 7.

Difference in change in MAM surface air temperature (°C) (RFIX − R). Shading indicates areas where the difference in temperature trends is statistically significant (p < 0.05) using the NT5 test among all three ensemble members.

Fig. 7.

Difference in change in MAM surface air temperature (°C) (RFIX − R). Shading indicates areas where the difference in temperature trends is statistically significant (p < 0.05) using the NT5 test among all three ensemble members.

c. Climatic change: Hydrologic cycle

Intensification of the hydrologic cycle is evident in ensembles RP, R, and RFIX associated with the radiative effect of increasing CO2. Global mean precipitation increases 1.8% and 2.0% in RP and R, respectively, with a more substantial percent increase in the NH than SH (Table 3). The physiological effect in ensemble P reduces globally averaged precipitation by a statistically insignificant 0.1% and the percent differences between R and RFIX are minimal; however, regional effects are more distinct. On the global or hemispheric scale, none of the experiments exhibit field significance for precipitation trends (Table 3).

Hydrological changes resulting from the radiative effect should not be viewed as purely local since hydrology has a strong dynamical component. The change in zonal mean precipitation minus evaporation (PME) in ensemble R closely matches the finding of Kutzbach et al. (2005), who analyzed the IPCC Fourth Assessment Report (AR4) model simulations. PME increases in the Tropics and mid to high latitudes, while decreasing in the subtropics. Although the rising global mean temperature produces a net global moistening, there is a dynamical redistribution of this moisture that results in increases of PME across certain zonal band and decreases across others.

The percent change in precipitation for the four ensemble sets are shown in Fig. 8. The radiative effect produces a dramatic increase in Indonesian precipitation (3–9 cm month−1, p < 0.05), particularly in SON (Fig. 9), associated with higher SSTs in the West Pacific warm pool. Tropical Pacific precipitation increases by 5%–20% (p < 0.05) along the equator as the cold tongue weakens. The SNR for annual precipitation trends in ensemble R peaks (SNR > 5) over Indonesia and the U. S. midwest, with statistically significant increases in precipitation. Large Arctic warming, leading to enhanced evaporation, results in a 5%–20% increase in precipitation (p < 0.05) in ensembles RP and R, with a maximum increase during DJF reaching 20%–100% (p < 0.05). SNRs for ensemble RP generally peak around 5–10 over the Arctic in DJF, suggesting a robust signal of increased precipitation. DJF precipitation also significantly increases over central/northern Africa in ensemble RP. The radiative effect in ensemble R produces vast areas of reduced precipitation, soil moisture, and latent heat flux (sensible heat flux increases) in the Tropics and subtropics, including the Amazon, Congo/South Africa, and Australia.

Fig. 8.

Percent change in annual precipitation in the (a) RP, (b) R, (c) P, and (d) RFIX ensembles, based on the difference in mean precipitation between the first and last 30 years. Contours are ±5%, 10%, 20%, 40%, and 80%. Shading indicates statistical significance (p < 0.05) was achieved with the NT5 test in all three ensemble members.

Fig. 8.

Percent change in annual precipitation in the (a) RP, (b) R, (c) P, and (d) RFIX ensembles, based on the difference in mean precipitation between the first and last 30 years. Contours are ±5%, 10%, 20%, 40%, and 80%. Shading indicates statistical significance (p < 0.05) was achieved with the NT5 test in all three ensemble members.

Fig. 9.

Percent change in precipitation in the RP ensemble for (a) DJF, (b) MAM, (c) JJA, and (d) SON, based on the difference in mean precipitation between the first and last 30 years. Contours are ±5%, 10%, 20%, 40%, and 80%. Shading indicates statistical significance (p < 0.05) was achieved with the NT5 test in all three ensemble members.

Fig. 9.

Percent change in precipitation in the RP ensemble for (a) DJF, (b) MAM, (c) JJA, and (d) SON, based on the difference in mean precipitation between the first and last 30 years. Contours are ±5%, 10%, 20%, 40%, and 80%. Shading indicates statistical significance (p < 0.05) was achieved with the NT5 test in all three ensemble members.

The physiological effect of rising CO2 produces drying (about 5%–10%) in the Amazon and Congo in ensemble P, associated with reduced evapotranspiration. However, none of the annual precipitation trends in P achieve statistical significance and SNRs are quite low, generally around 1 over land. Associated with this precipitation decline over the Amazon and Congo is a reduction in latent heat flux (as plants become more water-use efficient) and increase in sensible heat flux, leading to a warming of 0.6°–0.8°C. The decrease in Amazonian precipitation due to both radiative and physiological forcing of rising CO2 agrees with the study of Betts et al. (2004) but is opposite to the conclusions of Costa and Foley (2000), whose simulations neglected long time-scale vegetation dynamics.

The impact of changes in vegetation cover is revealed by comparing precipitation trends between R and RFIX. Ensemble R produces up to 20% greater increase in precipitation over Indonesia than RFIX, related to an increase in tropical tree cover and evapotranspiration in ensemble R. However, the small increase in precipitation over central Russia, Mongolia, and northern China in ensemble RFIX is reduced by 5%–10% in ensemble R due to a loss of boreal tree cover in ensemble R.

Trends in soil moisture are also analyzed to evaluate hydrological changes. The global mean increase in soil moisture of 0.4% in ensemble RP consists of a NH increase of +1.7% and a SH decrease of −1.5%. Both the radiative and physiological effects contribute toward an increase in NH soil moisture of about +1.4% and +0.3% for ensemble R and P, respectively. The radiative effect strengthens the hydrological cycle, while the physiological effect increases plant growth and evapotranspiration. In the SH, the radiatively driven drying of −2.5% in ensemble R exceeds the small increase in soil moisture of +0.4% in ensemble P. Since most SH landmasses are positioned in the Tropics, they are vulnerable to heat stress, vegetation dieback, and drying.

The NH changes in JJA soil moisture average −1.4%, −0.9%, −0.2%, and −1.7% for ensembles RP, R, P, and RFIX, respectively, illustrating the drying impact of radiative forcing from rising CO2 and reduction of soil moisture trends by interactive vegetation cover dynamics. The radiative effect induces a reduction in soil moisture over the Amazon and Australia (reduced forest cover) and a statistically significant increase over the U. S. Midwest (increased forest cover), while the physiological effect produces drying over tropical forests (reduced evapotranspiration), including the Amazon and Congo (p < 0.05 for the Congo). Levis et al. (2000) likewise found that the radiative effect generally increased midhigh latitude soil moisture and reduced tropical soil moisture, while the physiological effect narrowed stomatal openings and reduced tropical soil moisture. In some regions (e.g., Amazonia), the radiative and physiological effects produce the same impacts on soil moisture, while other regions (e.g., Australia) show competing effects. In comparing R and RFIX, the differences are attributed to changes in vegetation cover, including the increase in soil moisture in southwest Russia in ensemble R. The radiative and physiological effects produce net soil moisture drying in the NH in JJA, while increasing net soil moisture in the other seasons. In the SH, the radiative effect induces drying in every season while the physiological effect decreases net soil moisture in JJA and increases it in every other season.

Global evaporation changes by +1.8%, +2.0%, and −0.1% (not significant) in ensembles RP, R, and P, respectively (Table 3). An increase in Arctic evaporation by 5%–25% (p < 0.05) in ensembles RP and R, related to enhanced warming from the radiative effect of rising CO2, contributes to a substantial increase in precipitation. Tropical and subtropical regions show diminished evaporation (p < 0.05) in ensemble R by 5%–25% over the Amazon, Australia, and central/southern Africa. Small areas of increased evaporation (statistically insignificant) are noted in ensemble P in moisture-limited regions where vegetation cover is enhanced. The reduced evaporation (p < 0.05) over the central Russian boreal forest in ensemble R, which is absent in ensemble RFIX, is attributed to a dramatic loss of deciduous trees.

The global reduction in evapotranspiration by 2% in ensemble RP is attributed to the physiological effect (−3.0%) (Table 3). The physiological effect decreases evapotranspiration in both hemispheres, particularly in the tropical rain forests (not statistically significant), while the radiative effect increases evapotranspiration in the NH and decreases it in the SH. The radiative effect generally increases evapotranspiration in the Arctic (p < 0.05), as vegetation expands into the tundra, and decreases it in the Tropics and SH subtropics (p < 0.05 over the Amazon, South Africa, Australia), resulting from forest dieback. Fixed vegetation cover in ensemble RFIX enhances the increase in evapotranspiration in the NH and limits its reduction in the SH. The decrease in tropical evapotranspiration is substantially reduced in ensemble RFIX compared to R.

Both radiative and physiological forcings contribute to a net reduction in total cloud cover globally over land (Table 3). Global total cloud cover diminishes by 2.3%, 1.7%, and 0.8% in ensembles RP, R, and P, respectively. In ensemble RP, cloud cover decreases over almost all land, except for a 3%–9% increase over Indonesia as the warm pool expands. Reductions in total cloud cover over Amazonia, central and southern Africa, and Australia reflect diminished moisture flux from soil and vegetation in ensemble R. The decrease in cloud cover is dominated by a reduction in low clouds, with decreases of 7.0%, 4.0%, and 3.3% in these ensembles. Low cloud amount decreases by at least 10% over the Amazon, the Congo, and Australia in ensemble P due to reduced evapotranspiration.

d. Vegetation changes

FOAM–LPJ simulates a future global greening trend due to rising levels of CO2 (Fig. 10). Total vegetation cover increases 4.9%, 1.0%, and 4.7% in ensembles RP, R, and P, respectively, primarily attributed to the physiological effect of enhancing photosynthesis. The global greening trends are robust among ensemble members, with net increases in vegetation cover ranging from 4.9% to 5.0% for RP, 0.6% to 1.5% for R, and 4.3% to 5.0% for P. Levis et al. (1999) and Betts et al. (1997) likewise simulated an enhancement of vegetation cover due to the combined impacts of radiative and physiological forcings from future rising CO2. Global forest cover increases during the late twentieth century but later becomes steady during the twenty-first century. The radiative and physiological effects on forests oppose one another, with percent changes of −2.8% (ensemble range: −2.6% to −3.1%) and +5.2% (ensemble range: +5.0 to +5.5%) (Fig. 10 and Table 3); however, the physiological effect is still dominant and produces a small net increase in forest cover in ensemble RP. The 4.2% increase in global grass cover is attributed to the radiative effect, with net changes of +3.9% and −0.5% in ensembles R and P, respectively.

Fig. 10.

Time series of global mean percent coverage of (a) vegetation (tree + grass), (b) trees, (c) grass, (d) boreal trees, (e) temperate trees, and (f) tropical trees in runs RP (thick solid), R (thin solid), and P (dashed). Annual mean surface air temperature (°C) over land for the (g) boreal (45°–90°N), (h) temperate (15°–45°N, 15°–45°S), and (i) tropical (15°S–15°N) zones. Annual mean precipitation (cm) over land for the (j) boreal, (k) temperate, and (l) tropical zones.

Fig. 10.

Time series of global mean percent coverage of (a) vegetation (tree + grass), (b) trees, (c) grass, (d) boreal trees, (e) temperate trees, and (f) tropical trees in runs RP (thick solid), R (thin solid), and P (dashed). Annual mean surface air temperature (°C) over land for the (g) boreal (45°–90°N), (h) temperate (15°–45°N, 15°–45°S), and (i) tropical (15°S–15°N) zones. Annual mean precipitation (cm) over land for the (j) boreal, (k) temperate, and (l) tropical zones.

Total boreal tree cover remains nearly constant until 2030 in ensemble RP but then rapidly decreases due to the radiative effect, with associated heat stress and localized drying (Fig. 10). The timing of this sudden drop is quite consistent among ensemble members. The net changes in total boreal tree cover are −5.5%, −3.7%, and +0.9% in ensembles RP, R, and P, respectively. The physiological effect contributes to a slight increase in boreal tree cover by enhancing photosynthesis. More boreal tree cover is lost after 2080 than in the prior 105 years in ensemble RP, revealing nonlinear vegetation responses; late in the simulations, tree cover continues to decrease in the central/southern boreal forest but the expansion of boreal trees/shrubs into the tundra slows, resulting in an acceleration of the net loss of global boreal forest. Temperate tree cover increases by 5.7%, 0.6%, and 4.1% in ensembles RP, R, and P, respectively, with the expansion of temperate tree cover in RP mainly attributed to carbon fertilization. Finally, tropical tree cover only shows slight increases in ensembles RP, R, and P.

Statistically significant changes in tree-cover and grass-cover fractions are shown in Fig. 1 for ensembles RP, R, and P. Tree cover increases in RP over the tundra, western United States, Great Lakes Basin, Sahel, southwest Russia/Eastern Europe, Middle East, and Uruguay/Argentina and decreases over the Asian boreal forest, central Canada, and Amazonia. The physiological effect is responsible for the increase in tree cover over dry regions, such as the Sahel and Australia, as plants reduce their stomatal openings, lower their evapotranspiration, and more efficiently use their limited water supply. The Amazonia forest dieback produced by FOAM–LPJ is also simulated by Betts et al. (2004) and Cox et al. (2004). The field significance of forest and grass cover changes across the globe does not reach p = 0.05, although substantial regionally significant trends are produced.

The radiative effect of rising CO2 produces warming, localized soil drying, and increased summertime heat stress on boreal trees, leading to their loss (p < 0.05) over much of Russia into northern China and also central Canada. Portions of the boreal forest are replaced by temperate forests and grasslands across North America and Eurasia. In LPJ, heat damage mortality for boreal trees is based on the number of annual growing degree days (GDDs) exceeding 23°C (Sitch et al. 2003). Over portions of the Russian boreal forest, as JJA mean temperatures approach 23°C, the tree cover rapidly diminishes. Since boreal trees mask snow-covered surfaces, they lower the surface albedo and induce warming at high latitudes. The widespread loss of central Asian boreal tree cover in ensemble R, due to heat stress and localized soil drying, reduces this positive feedback and thereby leads to a minimum in warming over central Russia, which is absent in ensemble RFIX (Fig. 6). Radiatively induced drying trends (p < 0.05) over Amazonia, South Africa, and Australia in ensemble R lead to forest dieback. However, enhanced warming and evaporation over the tundra support a poleward expansion of boreal deciduous forest (p < 0.05) in ensemble R. Over Indonesia, increasing forest cover (p < 0.05) in ensemble R amplifies the positive precipitation trend in RFIX through enhanced evapotranspiration. Also, over southwest Russia/Eastern Europe, increasing tree cover (p < 0.05) due to the radiative effect supports positive soil moisture trends that are absent in RFIX.

The physiological effect in run P increases tree cover (p < 0.05) over moisture-limited regions through enhanced photosynthesis and water-use efficiency, consisting of an increase in deciduous trees over the tundra and western United States and of evergreen trees over the Sahel, Australia, Middle East, and Uruguay/Argentina. All regions of expanded tree cover in ensemble P show small, localized increases in evapotranspiration due to greater leaf area, while most other areas (e.g., tropical forests) show substantial reductions in evapotranspiration due to decreased stomatal conductance.

The SNR for tree cover trends is presented in Fig. 11. The highest SNRs (SNR > 10) are generally produced over the northern mid to high latitudes in ensembles RP and R, reflecting a consistent signal of poleward boreal forest shift attributed to the radiative effect of rising CO2. No substantial region of forest cover change in ensemble P exhibits a SNR > 10, suggesting less agreement among ensemble members. The Tropics exhibit a particularly low SNR in ensemble P, where tree cover change signals are generally small. This pattern is opposite to the temperature SNRs in Fig. 4, which peak over the Tropics.

Fig. 11.

SNR for tree-cover trends in ensemble (a) RP, (b) R, and (c) P. Areas that are shaded have a SNR > 1. Right-hand panels show the percentage of land grid points with SNR > 1 for each latitudinal band.

Fig. 11.

SNR for tree-cover trends in ensemble (a) RP, (b) R, and (c) P. Areas that are shaded have a SNR > 1. Right-hand panels show the percentage of land grid points with SNR > 1 for each latitudinal band.

Substantial changes in biome distribution are simulated as CO2 levels reach 4 × CO2 (Fig. 12). Much of the tundra and polar desert disappears, being replaced by shrubland and woodland in the equilibrium period of the RP run due to both radiative and physiological forcings. A lack of soil across the tundra will likely slow the poleward expansion of the boreal forest. Vast portions of the Eurasian and Canadian boreal forests are converted to grasslands and temperate forests as summer heat stress and regional soil drying induce boreal tree mortality. The Sahel transition zone pushes northward as the physiological effect enhances vegetation growth in this moisture-limited region. The expansion of temperate forests is primarily attributed to the physiological effect. In comparing the modern and future biome distribution, a kappa value (κ) of 0.54 is computed globally, indicating a fair agreement (Monserud 1990) between the two maps in Fig. 12. The kappa value is substantially lower for the NH (κ = 0.49) than the SH (κ = 0.72). North of 45°N, the biome patterns substantially differ between simulations, representing very poor agreement (κ = 0.22) due to substantial changes in the boreal forests.

Fig. 12.

Mean biome distribution for the (a) “modern” control and (b) 4 × CO2 equilibrium run (RP). Biome classifications are based on vegetation fraction, vegetation height, and surface temperature (from an algorithm outlined in Fig. 11 of Joos et al. 2004).

Fig. 12.

Mean biome distribution for the (a) “modern” control and (b) 4 × CO2 equilibrium run (RP). Biome classifications are based on vegetation fraction, vegetation height, and surface temperature (from an algorithm outlined in Fig. 11 of Joos et al. 2004).

Substantial biomass changes result from both the radiative and physiological effect. Biomass is defined here as the sum of heartwood, sapwood, and leaf carbon mass. Starting with an initial global biomass storage of 1.54 × 1015 kg carbon, the percent changes in global biomass in ensembles RP, R, and P are +31%, −16%, and +53% (Fig. 13). The widespread increase in biomass in RP is primarily attributed to the physiological effect, including over the tropical forests (Levis et al. 1999; Joos et al. 2001), and suggests a growing sink of carbon in the future.

Fig. 13.

(a) Average biomass (kg °C × 1010) in the first 30 years of RP. Biomass is the sum of heartwood, sapwood, and leaf carbon mass. The percent change in biomass, shown for (b) RP, (c) R, and (d) P, is the difference between the first and last 30 years of the ensembles. Left label bar for (a) and right bar for (b)–(d). Only those changes that achieve statistical significance (p < 0.05) with the NT5 test in all three ensemble members are shown.

Fig. 13.

(a) Average biomass (kg °C × 1010) in the first 30 years of RP. Biomass is the sum of heartwood, sapwood, and leaf carbon mass. The percent change in biomass, shown for (b) RP, (c) R, and (d) P, is the difference between the first and last 30 years of the ensembles. Left label bar for (a) and right bar for (b)–(d). Only those changes that achieve statistical significance (p < 0.05) with the NT5 test in all three ensemble members are shown.

Over the Arctic (>60°N), the radiative effect produces a 9% reduction in total biomass (7.4 × 1013 kg to 6.7 × 1013 kg), despite an increase over the tundra; this reduction is attributed to a loss of heartwood mass. The physiological effect, however, produces a substantial 26% increase in total Arctic biomass, with increases in leaf, sapwood, and heartwood mass. Both effects in ensemble RP lead to a 3% increase in total Arctic biomass. Cox et al. (2000) also simulated an increase in vegetation carbon across the northern boreal forest and tundra. Across the central Asian and Canadian boreal forests, the radiative effect enhances summertime heat stress and produces localized areas of soil drying, killing 50%–100% of the biomass (p < 0.05). Also, the radiative effect, through drying, reduces tropical biomass over Amazonia, Africa, and Australia (p < 0.05). The aforementioned changes in biomass have implications on the global carbon budget and the occurrence of fires.

Extensive areas of the Amazon, Congo/Sahel, and Southeast Asia exhibit high SNRs > 10 among positive biomass trends in ensemble RP, providing evidence of a distinct signal. These positive biomass trends are attributed to the physiological effect of rising CO2, with SNRs > 10 in ensemble P over the same regions.

The fraction of photosynthetically active radiation (FPAR) absorbed by the plant canopy serves as an index of plant amount. Annual area-average FPAR is decomposed into four terms, each summed over the nine PFTs and all months (Notaro et al. 2005):

 
formula

Here f is vegetation cover fraction and d is seasonal leaf cover fraction for each PFT. By decomposing FPAR into components and computing the trends of each component, the overall FPAR trend is better understood (Fig. 14). Term 1 represents mean FPAR with no trend. A trend in term 2 represents a change in leaf cover or length of growing season, or GDDs. A trend in Term 3 involves changes in fractional vegetation cover. Finally, trends in Term 4, representing covariance or feedbacks between f and d, are small.

Fig. 14.

Annual FPAR decomposition. (a) Change in FPAR, (b) change in term 2, and (c) change in term 3 for run RP1. Similar plots are shown in (d)–(f) for run R1 and in (g)–(i) for run P1.

Fig. 14.

Annual FPAR decomposition. (a) Change in FPAR, (b) change in term 2, and (c) change in term 3 for run RP1. Similar plots are shown in (d)–(f) for run R1 and in (g)–(i) for run P1.

The change in NH mean FPAR in runs RP, R, and P is +0.023 (+6%), −0.012 (−3%), and +0.042 (+11%), respectively, with the physiological effect responsible for most of the greening trend and the largest overall FPAR trend. The majority of the FPAR increase in RP is associated with a lengthening of the growing season (Term 2). A widespread increase in GDDs is simulated over the northern mid to high latitudes in RP, particularly north of 50°N where low temperatures can limit the growing season. The sharp decrease in FPAR over central Asia and Canada’s southern boreal forests results from decreasing vegetation cover, attributed to radiative forcing of rising CO2. In run R, the negative FPAR trend is primarily due to a decrease in total vegetation cover (Term 3). Positive FPAR trends in the tundra in run R reflect both an increase in GDDs and total vegetation cover. The substantial increase in NH mean FPAR in run P results from increases in both GDDs (Term 2) and total vegetation cover (Term 3), although changes in the latter term are twice as large as the former term. Small increases in FPAR are simulated in run P across most of the northern mid to high latitudes due to a lengthening growing season. Enhanced photosynthesis in P produces large increases in FPAR in excess of 0.1 across the Sahel, western United States, and Middle East due to increased vegetation cover.

LPJ determines natural fire frequency as a function of fuel load and litter moisture (Sitch et al. 2003). It is likely that future changes in moisture and biomass will lead to changes in fire occurrence. There is fairly good agreement between the regions of frequent fires simulated by FOAM–LPJ (Fig. 15) and those simulated by Thonicke et al. (2001) using LPJ driven by observational data. Globally, the percent change in trees burned is −16%, +6%, and −17% in ensembles RP, R, and P, respectively. While it is not surprising that the radiative effect would cause an increase in trees burned, it is unexpected that the physiological effect would cause a substantial decrease and even outweigh the radiative effect. In ensemble R, it is inferred that the radiative effect increases the number of trees burned (p < 0.05) over Australia, South Africa, and Western Europe due to a drying trend and over the tundra due to increased biomass litter in this region of limited fuel load. The trend toward fewer trees burned (p < 0.05) in ensemble P over areas such as the United States, Mexico, South Africa, and Australia reflects an increase in soil moisture attributed to the physiological effect. One of the few areas that show an increase in trees burned in ensemble P is the tundra, due to increasing biomass.

Fig. 15.

(a) Mean number of trees burned (× 107) by fire in the first 30 years of RP. The percent change in trees burned, shown for (b) RP, (c) R, and (d) P, is the difference between the first and last 30 years of the runs. Only those changes that achieve statistical significance (p < 0.05) with the NT5 test in all three ensemble members are shown.

Fig. 15.

(a) Mean number of trees burned (× 107) by fire in the first 30 years of RP. The percent change in trees burned, shown for (b) RP, (c) R, and (d) P, is the difference between the first and last 30 years of the runs. Only those changes that achieve statistical significance (p < 0.05) with the NT5 test in all three ensemble members are shown.

4. Conclusions

The future impact of rising levels of CO2, out to 4 × CO2, is investigated through 12 transient simulations. These simulations are used to decompose the radiative and physiological forcings of CO2 and vegetation feedbacks on climate. We use a fully interactive atmosphere–ocean–land–ice model, FOAM–LPJ, with dynamic vegetation and ocean parameterizations.

The RP ensemble predicts an increase in global mean surface air temperature by 2.7°C. While the majority of this warming is associated with the radiative effect, the physiological effect contributes additional warming. The radiative effect produces peak warming across the tundra and boreal forest due to the snow albedo feedback. The physiological effect enhances photosynthesis and the water-use efficiency of plants, while reducing stomatal conductance and evapotranspiration, which weakens the hydrologic cycle and induces warming.

The radiative effect increases precipitation over the Arctic, due to enhanced warming and evaporation, and Indonesia, due to expansion of the warm pool. The physiological effect reduces precipitation and soil moisture over tropical forests by lowering evapotranspiration. Both forcings induce drying and forest dieback over Amazonia.

The simulated global greening trend in ensemble RP is primarily associated with the physiological effect, which increases tree cover through enhanced photosynthesis associated with higher water-use efficiency. The radiative effect generally produces a small net loss in tree cover, due to soil drying, and a substantial enhancement in grass cover. Reduced soil moisture over Amazonia, South Africa, and Australia from the radiative effect results in a decrease in forest cover. The radiative effect produces poleward advancement of the boreal forest into the tundra and replacement of boreal trees in the southern and central boreal forest with grass and temperate trees. The warming trend over central Russia is weaker in ensemble R than in RFIX due to vegetation feedbacks. The physiological effect increases tree cover across moisture-limited regions, such as the tundra, lower Sahara, western United States, and Australia, by enhancing photosynthesis through higher plant water-use efficiency.

The net increase in NH FPAR in RP is mostly related to an expanded growing season. While the radiative effect produces a widespread increase in GDDs, it also substantially reduces vegetation cover over portions of the boreal forest. The physiological effect induces a small widespread increase in seasonal leaf cover, while enhancing total vegetation cover across moisture-limited regions.

The widespread rise in net biomass in RP is attributed to the physiological effect. While the radiative effect enhances biomass over the tundra through more favorable growing conditions, it also largely reduces biomass over the tropical forests and central/southern boreal forests through drying and boreal heat stress. Related to changes in fuel load and litter moisture, the radiative effect produces a global-average increase in trees burned while the physiological effect substantially reduces the number of trees burned.

Several factors remain unclear regarding the impact of future climate change on vegetation. FOAM–LPJ does not include nutrient limitation, which might limit growth (Field et al. 1995) from physiological forcing. The use of a coarse R15 atmosphere could potentially amplify the variance induced by a land perturbation compared to a model with a fine resolution atmosphere. While FOAM–LPJ simulates substantial loss of boreal forest to summertime heat stress, the actual importance of such stress remains uncertain. Numerous modeling studies have shown the replacement of southern boreal trees by grass and temperate trees due to heat stress and drying (Bounoua et al. 1999; Bergengren et al. 2001; Joos et al. 2001; Notaro et al. 2005). Paleoecological data from BIOME6000 for the relatively mild Holocene suggests conversion to grassland and temperate forest within 40°–55°N (Prentice et al. 1996; Boenisch et al. 2001). A cluster analysis study by Stolbovoi (1999) predicted a poleward shift of the Asian boreal forest due to future warming, with the south side suffering from heat stress. While Berry and Bjorkman (1980) argue for vegetation acclimation to temperature changes, Stolbovoi (1999) concludes that boreal trees will eventually suffer heat stress, even if precipitation is sufficient for adaption. Other studies have observed heat or drought stress on North American spruce trees at high latitudes (Barber et al. 2000; Arain et al. 2002; Lloyd and Fastie 2002; Wilmking et al. 2004). The importance or existence of summertime heat stress on boreal trees remains uncertain. If realistic, it could substantially impact future changes in vegetation and climate.

The relative success of FOAM–LPJ at producing a global warming trend, global greening trend, and poleward shift of the boreal forest during the preindustrial to modern period that are similar to observed, as demonstrated by Notaro et al. (2005), ensues confidence in the climate and vegetation predictions of the current study. The global warming trend for RP nearly doubles in the future simulations as compared to the preindustrial to modern period simulation. In both periods, the boreal forest shifts poleward, although the central boreal dieback and expansion of vegetation into the tundra are much more distinct in the future simulations. The simulated global greening trend continues through both periods.

Acknowledgments

This research was funded by DOE and NSF. The authors thank Drs. S. Sitch, S. Schaphoff, F. Joos, and N. Pederson for their thoughts on boreal heat stress. We also appreciate the suggestions from Professors J. Kutzbach and A. Pitman and advice on statistical testing from Dr. G. Narisma. We are grateful to the anonymous reviewers for their helpful guidance.

REFERENCES

REFERENCES
Arain
,
A.
,
T. A.
Black
,
A. G.
Barr
,
P. G.
Jarvis
,
J. M.
Massheder
,
D. L.
Verseghy
, and
Z.
Nesic
,
2002
:
Effects of seasonal and interannual climate variability on net ecosystem productivity of boreal deciduous and conifer forests.
Can. J. For. Res.
,
32
,
878
891
.
Barber
,
V. A.
,
G. P.
Juday
, and
B. P.
Finney
,
2000
:
Reduced growth of Alaskan white spruce in the twentieth century from temperature-induced drought stress.
Nature
,
405
,
668
673
.
Bergengren
,
J. C.
,
S. L.
Thompson
,
D.
Pollard
, and
R. M.
Deconto
,
2001
:
Modeling global climate–vegetation interactions in a doubled CO2 world.
Climatic Change
,
50
,
31
75
.
Berry
,
J. A.
, and
O.
Bjorkman
,
1980
:
Photosynthetic response and adaptation to temperature in higher plants.
Annu. Rev. Plant Physiol.
,
31
,
491
543
.
Betts
,
R. A.
,
P. M.
Cox
,
S. E.
Lee
, and
F. I.
Woodward
,
1997
:
Contrasting physiological and structural vegetation feedbacks in climate change simulations.
Nature
,
387
,
796
799
.
Betts
,
R. A.
,
P. M.
Cox
,
M.
Collins
,
P.
Harris
,
C.
Huntingford
, and
C. D.
Jones
,
2004
:
The role of ecosystem–atmosphere interactions in simulated Amazonian precipitation decrease and forecast dieback under global warming.
Theor. Appl. Climatol.
,
78
,
157
175
.
Boenisch
,
G.
,
S. P.
Harrison
, and
I. C.
Prentice
, and
Biome 6000 members
,
2001
:
BIOME 6000 data release 1. IGBP PAGES/World Data Center for Paleoclimatology Data Contribution Series 2001-046, NOAA/NGDC Paleoclimatology Program, Boulder CO
.
Bonan
,
G.
,
2002
:
Ecological Climatology: Concepts and Applications.
Cambridge University Press, 678 pp
.
Bonan
,
G. B.
,
D.
Pollard
, and
S. L.
Thompson
,
1992
:
Effects of boreal forest vegetation on global climate.
Nature
,
359
,
716
718
.
Bounoua
,
L.
, and
Coauthors
,
1999
:
Interactions between vegetation and climate: Radiative and physiological effects of doubled atmospheric CO2.
J. Climate
,
12
,
309
324
.
Cohen
,
J.
,
1960
:
A coefficient of agreement for nominal scales.
Educ. Psychol. Meas.
,
20
,
37
46
.
Costa
,
M. H.
, and
J. A.
Foley
,
2000
:
Combined effects of deforestation and doubled atmospheric CO2 concentrations on the climate of Amazonia.
J. Climate
,
13
,
18
34
.
Cox
,
P. M.
,
R. A.
Betts
,
C. D.
Jones
,
S. A.
Spall
, and
I. J.
Totterdell
,
2000
:
Acceleration of global warming due to carbon-cycle feedbacks in a coupled climate model.
Nature
,
408
,
184
187
.
Cox
,
P. M.
,
R. A.
Betts
,
M.
Collins
,
P. P.
Harris
,
C.
Huntingford
, and
C. D.
Jones
,
2004
:
Amazonian forest dieback under climate-carbon cycle projections for the 21st century.
Theor. Appl. Climatol.
,
78
,
137
156
.
Cramer
,
B. A.
, and
Coauthors
,
2001
:
Global response of terrestrial ecosystem structure and function to CO2 and climate change: Results from six dynamic global vegetation models.
Global Change Biol.
,
7
,
357
373
.
Cubasch
,
U.
, and
Coauthors
,
2001
:
Projections of future climate change.
Climate Change 2001: The Scientific Basis, J. T. Houghton et al., Eds., Cambridge University Press, 525–582
.
Curtis
,
P. S.
,
1996
:
A meta-analysis of leaf gas exchange and nitrogen in trees grown under elevated carbon dioxide.
Plant Cell Environ.
,
19
,
127
137
.
Drake
,
B. G.
,
M. A.
Gonzalez-Meler
, and
S. P.
Long
,
1997
:
More efficient plants: A consequence of rising atmospheric CO2?
Annu. Rev. Plant Physiol. Plant Mol. Biol.
,
48
,
609
639
.
Drake
,
J.
,
I.
Foster
,
J.
Michalakes
,
B.
Toonen
, and
P.
Worley
,
1995
:
Design and performance of a scalable parallel community climate model.
Parallel Comput.
,
21
,
10
.
1571
1591
.
Farquhar
,
G. D.
,
1997
:
Carbon dioxide and vegetation.
Science
,
278
,
1411
.
Field
,
C.
,
R.
Jackson
, and
H.
Mooney
,
1995
:
Stomatal responses to increased CO2: Implications from the plant to the global scale.
Plant Cell Environ.
,
18
,
1214
1225
.
Foley
,
J. A.
,
J. E.
Kutzbach
,
M. T.
Coe
, and
S.
Levis
,
1994
:
Feedbacks between climate and boreal forests during the Holocene epoch.
Nature
,
371
,
52
54
.
Foley
,
J. A.
,
S.
Levis
,
M. H.
Costa
,
W.
Cramer
, and
D.
Pollard
,
2000
:
Incorporating dynamic vegetation cover within global climate models.
Ecol. Appl.
,
10
,
6
.
1620
1632
.
Gallimore
,
R.
,
R. L.
Jacob
, and
J. E.
Kutzbach
,
2005
:
Coupled atmosphere–ocean–vegetation simulations for modern and mid-Holocene climates: Role of extratropical vegetation cover feedbacks.
Climate Dyn.
,
25
,
755
776
.
Harrison
,
S. P.
, and
Coauthors
,
1998
:
Intercomparison of simulated global vegetation distributions in response to 6 kyr BP orbital forcing.
J. Climate
,
11
,
2721
2742
.
Henderson-Sellers
,
A.
,
A. J.
Pitman
,
P. K.
Love
,
P.
Irannejad
, and
T. H.
Chen
,
1995
:
The Project for Intercomparison of Land Surface Parameterization Schemes (PILPS): Phases 2 and 3.
Bull. Amer. Meteor. Soc.
,
76
,
489
503
.
Jacob
,
R. L.
,
1997
:
Low frequency variability in a simulated atmosphere ocean system. Ph.D. thesis, University of Wisconsin—Madison, 159 pp
.
Jacob
,
R.
,
C.
Schafer
,
I.
Foster
,
M.
Tobis
, and
J.
Anderson
,
2001
:
Computational design and performance of the Fast Ocean Atmosphere Model, version one.
Proceedings of the 2001 International Conference on Computational Science, V. N. Alexandrov, J. J. Dongarra, and C. J. K. Tan, Eds., Springer-Verlag, 175–184
.
Joos
,
F.
,
I. C.
Prentice
,
S.
Sitch
,
R.
Meyer
,
G.
Hooss
,
G-K.
Plattner
,
S.
Gerber
, and
K.
Hasselmann
,
2001
:
Global warming feedbacks on terrestrial carbon uptake under the Intergovernmental Panel on Climate Change (IPCC) emission scenarios.
Global Biogeochem. Cycles
,
15
,
891
907
.
Joos
,
F.
,
S.
Gerber
,
I. C.
Prentice
,
B. L.
Otto-Bliesner
, and
P. J.
Valdes
,
2004
:
Transient simulations of Holocene atmospheric carbon dioxide and terrestrial carbon since the Last Glacial Maximum.
Global Biogeochem. Cycles
,
18
.
GB2002, doi:10.1029/2003GB002156
.
Koch
,
G. W.
, and
H. A.
Mooney
,
1996
:
Response of terrestrial ecosystems to elevated CO2: A synthesis and summary.
Carbon Dioxide and Terrestrial Ecosystems, G. W. Koch and H. A. Mooney, Eds., Academic Press, 415–429
.
Körner
,
C.
,
2000
:
Biosphere responses to CO2 enrichment.
Ecol. Appl.
,
10
,
1590
1619
.
Kutzbach
,
J. E.
,
J.
Williams
, and
S.
Vavrus
,
2005
:
Simulated 21st century changes in regional water balance of the Great Lakes region and links to changes in global temperature and poleward moisture transport.
Geophys. Res. Lett.
,
32
.
L17707, doi:10.1029/2005GL023506
.
Laine
,
V.
, and
M.
Heikinheimo
,
1996
:
Estimation of surface albedo from NOAA AVHRR data in high latitudes.
Tellus
,
48A
,
424
441
.
Levis
,
S.
,
J. A.
Foley
, and
D.
Pollard
,
1999
:
Potential high-latitude vegetation feedbacks on CO2-induced climate change.
Geophys. Res. Lett.
,
26
,
747
750
.
Levis
,
S.
,
J. A.
Foley
, and
D.
Pollard
,
2000
:
Large-scale vegetation feedbacks on a doubled CO2 climate.
J. Climate
,
13
,
1313
1325
.
Liu
,
Z.
, and
L.
Wu
,
2004
:
Atmospheric response to North Pacific SST anomaly: The role of ocean–atmosphere coupling.
J. Climate
,
17
,
1859
1882
.
Liu
,
Z.
,
J.
Kutzbach
, and
L.
Wu
,
2000
:
Modeling climate shift of El Niño in the Holocene.
Geophys. Res. Lett.
,
27
,
2265
2268
.
Liu
,
Z.
,
M.
Notaro
,
J.
Kutzbach
, and
N.
Liu
,
2006
:
An observational assessment of global vegetation–climate feedbacks.
J. Climate
,
19
,
787
814
.
Lloyd
,
A. H.
, and
C. L.
Fastie
,
2002
:
Spatial and temporal variability in the growth and climate response of treeline trees in Alaska.
Climatic Change
,
52
,
418
509
.
McGuire
,
A. D.
, and
Coauthors
,
2001
:
Carbon balance of the terrestrial biosphere in the twentieth century: Analyses of CO2, climate and land use effects with four process-based ecosystem models.
Global Biogeochem. Cycles
,
15
,
183
206
.
Menzel
,
A.
, and
P.
Fabian
,
1999
:
Growing season extended in Europe.
Nature
,
397
,
659
.
Monserud
,
R. A.
,
1990
:
Methods for comparing global vegetation maps. Working Paper WP-90-40, IIASA, Laxenburg, Austria, 22 pp. [Available from International Institute for Applied Systems Analysis, Schlossplatz 1, Laxenburg A-2361, Austria.]
.
Mooney
,
H. A.
, and
Coauthors
,
1999
:
Ecosystem physiology responses to global change.
Implications of Global Change for Natural and Managed Ecosystems: A Synthesis of GCTE and Related Research, B. H. Walker et al., Eds., IGBP Book Series, Vol. 4, Cambridge University Press, 141–189
.
Myneni
,
R. B.
,
C. D.
Keeling
,
C. J.
Tucker
,
G.
Asrar
, and
R. R.
Nemani
,
1997
:
Increased plant growth in the northern high latitudes from 1981–1991.
Nature
,
386
,
698
701
.
Narisma
,
G. T.
, and
A. J.
Pitman
,
2003
:
The impact of 200 years of land cover change on the Australian near-surface climate.
J. Hydrometeor.
,
4
,
424
436
.
Notaro
,
M.
,
Z.
Liu
,
R.
Gallimore
,
S. J.
Vavrus
,
J. E.
Kutzbach
,
I. C.
Prentice
, and
R. L.
Jacob
,
2005
:
Simulated and observed preindustrial to modern vegetation and climate changes.
J. Climate
,
18
,
3650
3671
.
Notaro
,
M.
,
Z.
Liu
, and
J. W.
Williams
,
2006
:
Observed vegetation–climate feedbacks in the United States.
J. Climate
,
19
,
763
786
.
Pielke
,
R.
,
R.
Avissar
,
M.
Raupach
,
A. J.
Dolman
,
X.
Zhen
, and
A. S.
Denning
,
1998
:
Interactions between the atmosphere and terrestrial ecosystems: Influence of weather and climate.
Global Change Biol.
,
4
,
461
475
.
Pollard
,
D.
, and
S. L.
Thompson
,
1995
:
Use of a land-surface transfer scheme (LSX) in a global climate model: The response to doubling stomatal conductance.
Global Planet. Change
,
10
,
129
161
.
Polley
,
H. W.
,
H. B.
Johnson
,
B. D.
Marino
, and
H. S.
Mayeux
,
1993
:
Increase in C3 plant water use efficiency and biomass over glacial to present CO2 concentrations.
Nature
,
361
,
61
64
.
Preisendorfer
,
R. W.
, and
T. P.
Barnett
,
1983
:
Numerical model–reality intercomparison tests using small-sample statistics.
J. Atmos. Sci.
,
40
,
1884
1896
.
Prentice
,
I. C.
,
W.
Cramer
,
S. P.
Harrison
,
R.
Leemans
,
R. A.
Monserud
, and
A. M.
Solomon
,
1992
:
A global biome model based on plant physiology and dominance, soil properties and climate.
J. Biogeogr.
,
19
,
117
134
.
Prentice
,
C. I.
,
J.
Guiot
,
B.
Huntley
,
D.
Jolly
, and
R.
Cheddadi
,
1996
:
Reconstructing biomes from palaeoecological data: A general method and its application to European pollen data at 0 and 6 ka.
Climate Dyn.
,
12
,
185
194
.
Pritchard
,
S. G.
,
H. H.
Rogers
,
S. A.
Prior
, and
C. M.
Peterson
,
1999
:
Elevated CO2 and plant structure: A review.
Global Change Biol.
,
5
,
807
837
.
Robinson
,
D. A.
, and
G.
Kukla
,
1985
:
Maximum surface albedo of seasonally snow-covered lands in the Northern Hemisphere.
J. Climate Appl. Meteor.
,
24
,
402
411
.
Sellers
,
P. J.
, and
Coauthors
,
1996
:
Comparison of radiative and physiological effects of doubled atmospheric CO2 on climate.
Science
,
271
,
1402
1406
.
Sharratt
,
B. S.
,
1998
:
Radiative exchange, near-surface temperature and soil water of forest and cropland in interior Alaska.
Agric. For. Meteor.
,
89
,
269
280
.
Sitch
,
S.
,
2000
:
The role of vegetation dynamics in the control of atmospheric CO2 content. Ph.D. dissertation, Lund University, Lund, Sweden, 213 pp
.
Sitch
,
S.
, and
Coauthors
,
2003
:
Evaluation of ecosystem dynamics, plant geography and terrestrial carbon cycling in the LPJ dynamic global vegetation model.
Global Change Biol.
,
9
,
161
185
.
Stolbovoi
,
V.
,
1999
:
Classification of Russia’s forests in relation to global climate warming. Interim Rep. IR-99-005, International Institute for Applied Systems Analysis, 24 pp
.
Thonicke
,
K.
,
S.
Venevsky
,
S.
Sitch
, and
W.
Cramer
,
2001
:
The role of fire disturbance for global vegetation dynamics: Coupling fire into a Dynamic Global Vegetation Model.
Global Ecol. Biogeogr.
,
10
,
661
677
.
Wigley
,
T. M. L.
, and
B. D.
Santer
,
1990
:
Statistical comparison of spatial fields in model validation, perturbation, and predictability experiments.
J. Geophys. Res.
,
95
,
851
865
.
Wilmking
,
M.
,
G. P.
Juday
,
V. A.
Barber
, and
H. S. J.
Zald
,
2004
:
Recent climate warming forces contrasting growth responses of white spruce at treeline in Alaska through temperature thresholds.
Global Change Biol.
,
10
,
1724
1736
.
Wu
,
L.
, and
Z.
Liu
,
2002
:
Is tropical Atlantic variability driven by the North Atlantic Oscillation?
Geophys. Res. Lett.
,
29
.
1653, doi:10.1029/2002GL014939
.
Wu
,
L.
, and
Z.
Liu
,
2003
:
Decadal variability in North Pacific: The eastern North Pacific mode.
J. Climate
,
16
,
3111
3131
.
Wu
,
L.
,
Z.
Liu
, and
R.
Gallimore
,
2003
:
Pacific decadal variability: The tropical Pacific mode and North Pacific mode.
J. Climate
,
16
,
1101
1120
.
Zhao
,
M.
,
A. J.
Pitman
, and
T.
Chase
,
2001
:
The impact of land cover change on the atmospheric circulation.
Climate Dyn.
,
17
,
467
477
.
Zhou
,
L.
,
C. J.
Tucker
,
R. K.
Kaufmann
,
D.
Slayback
,
N. V.
Shabanov
, and
R. B.
Myneni
,
2001
:
Variations in northern vegetation activity inferred from satellite data of vegetation index during 1981 to 1999.
J. Geophys. Res.
,
106
,
20069
20083
.

Footnotes

* Center for Climatic Research Contribution Number 913

Corresponding author address: Dr. Michael Notaro, Center for Climatic Research, University of Wisconsin—Madison, 1225 West Dayton St., Madison, WI 53706. Email: mnotaro@wisc.edu