Will the tropical land biosphere dominate the climate–carbon cycle feedback during the twenty-first century?

Clim Dyn (2007) 29:565–574 DOI 10.1007/s00382-007-0247-8 Will the tropical land biosphere dominate the climate–carbon cycle feedback during the twenty-first century? T. J. Raddatz Æ C. H. Reick Æ W. Knorr Æ J. Kattge Æ E. Roeckner Æ R. Schnur Æ K.-G. Schnitzler Æ P. Wetzel Æ J. Jungclaus Received: 21 February 2006 / Accepted: 16 March 2007 / Published online: 17 April 2007 Ó Springer-Verlag 2007 Abstract Global warming caused by anthropogenic CO2 emissions is expected to reduce the capability of the ocean and the land biosphere to take up carbon. This will enlarge the fraction of the CO2 emissions remaining in the atmosphere, which in turn will reinforce future climate change. Recent model studies agree in the existence of such a positive climate–carbon cycle feedback, but the estimates of its amplitude differ by an order of magnitude, which considerably increases the uncertainty in future climate projections. Therefore we discuss, in how far a particular process or component of the carbon cycle can be identified, that potentially contributes most to the positive feedback. The discussion is based on simulations with a carbon cycle model, which is embedded in the atmosphere/ocean general circulation model ECHAM5/MPI-OM. Two simulations covering the period 1860–2100 are conducted to determine the impact of global warming on the carbon cycle. Forced by historical and future carbon dioxide emissions (following the scenario A2 of the Intergovernmental Panel on Climate Change), they reveal a noticeable positive climate–carbon cycle feedback, which is mainly driven by the tropical land biosphere. The oceans contribute much less to the positive T. J. Raddatz
C. H. Reick
W. Knorr
J. Kattge Max Planck Institute for Biogeochemistry, Jena, Germany E. Roeckner
R. Schnur
K.-G.Schnitzler
P. Wetzel
J. Jungclaus Max Planck Institute for Meteorology, Hamburg, Germany Present Address: T. J. Raddatz (&)
C. H. Reick Max Planck Institute for Meteorology, Hamburg, Germany e-mail: thomas.raddatz@zmaw.de Present Address: W. Knorr QUEST, University of Bristol, Bristol, UK feedback and the temperate/boreal terrestrial biosphere induces a minor negative feedback. The contrasting behavior of the tropical and temperate/boreal land biosphere is mostly attributed to opposite trends in their net primary productivity (NPP) under global warming conditions. As these findings depend on the model employed they are compared with results derived from other climate–carbon cycle models, which participated in the Coupled Climate– Carbon Cycle Model Intercomparison Project (C4MIP). Keywords Climate
Carbon cycle
Feedback
Global warming
C4MIP
NPP 1 Introduction During the last century about half of the CO2 released to the atmosphere by human activities was taken up by the ocean and the land biosphere (Prentice et al. 2001). Within the present century even larger emissions and a substantial global warming are expected, while there is considerable uncertainty about the continued uptake of CO2 by both the land and the ocean. In particular, the CO2 induced climate change could either enhance or decrease the rate of carbon uptake by the ocean and the land biosphere leading to either a negative or positive climate–carbon cycle feedback. Commonly, the latter is presumed as higher temperatures reduce the solubility of CO2 in seawater and are expected to enhance soil respiration (Knorr et al. 2005a). In fact, the first scenario simulation with a coupled climate–carbon cycle model (Cox et al. 2000) showed a strong positive feedback, which turned the land biosphere from a sink to a source around the year 2050. All subsequent simulations with other models resulted also in a positive but much weaker climate–carbon cycle feedback (Friedlingstein 123 566 T. J. Raddatz et al.: Will the tropical land biosphere dominate the climate–carbon cycle feedback during the twenty-first century? et al. 2006). The differences in the strength of the modeled feedback have been attributed to various processes: Temperature dependence of heterotrophic respiration (Cox et al. 2000; Prentice et al. 2001; Zeng et al. 2004) and primary productivity (Matthews et al. 2005), vertical mixing in the Southern Ocean (Friedlingstein et al. 2003), and cycling of carbon in the living biomass (Friedlingstein et al. 2003), especially in the Amazon forest (Cox et al. 2004). Furthermore, the impact of global warming on the already hot tropical and the temperate/cold extra-tropical land biosphere may be different (Zeng et al. 2004). To reduce uncertainties in future climate projections it is, therefore, of major importance to identify those processes and components of the global carbon cycle that contribute most to the positive climate–carbon cycle feedback now commonly ascertained. In particular, such a rating could provide guidance for the design of carbon cycle observing systems (Global Carbon Project 2003) and is discussed here by means of simulations performed with a coupled climate– carbon cycle model based on a high-resolution atmosphere/ ocean general circulation model. A description of this model is given in Sect. 2 and the setup of the simulations is specified in Sect. 3. In Sect. 4 the model’s performance in reproducing the main features of the present carbon cycle is evaluated. The simulated development of the carbon cycle during the twenty-first century is presented in Sect. 5 and the relative contributions of the major carbon cycle components—ocean, tropical and extra-tropical land biosphere—are analyzed. From this, the hypothesis arises, that the tropical land biosphere dominates the climate–carbon cycle feedback, which is discussed in Sect. 6 including a look at results obtained with other climate–carbon cycle models in the Coupled Climate–Carbon Cycle Model Intercomparison Project (C4MIP). A conclusion is given in Sect. 7. 2 Model description The atmosphere/ocean general circulation model (AOGCM) consists of ECHAM5 (Roeckner et al. 2003) in T63 (approx. 2°) resolution with 31 vertical levels representing the atmosphere and MPI-OM (Marsland et al. 2003) in 1.5° resolution with 40 vertical levels representing the ocean. The coupling of these two physical models is described in Jungclaus et al. (2006) and is implemented without any flux adjustment. This AOGCM is also employed for scenario simulations contributing to the fourth assessment report of the Intergovernmental Panel on Climate Change (IPCC). The carbon cycle model comprises the ocean biogeochemistry model HAMOCC5 (Wetzel et al. 2005) and the modular land surface scheme JSBACH, which is based on the biosphere model BETHY (Knorr 2000) and the 123 ECHAM5 soil scheme. HAMOCC5 simulates inorganic carbon chemistry as well as phyto- and zooplankton dynamics, which depend on temperature, solar radiation, turbulence and nutrients (N, P, Fe, Si). It also considers the build up of detritus, its sinking and remineralisation. JSBACH includes a photosynthesis module following Farquhar et al. (1980) for C3 plants and Collatz et al. (1992) for C4 plants. Besides the photosynthetic pathway, 13 plant functional types (PFT) are distinguished by maximum carboxylation rate, maximum electron transport rate, specific leaf area carbon content, and phenotype. The spatial distribution of the PFT is prescribed on the basis of the global land cover characteristics data base version 2.0 of the U.S. Geological Survey (Loveland et al. 2000). The interplay between the assimilation rate and stomatal conductance is explicitly modeled. Both quantities therefore depend on temperature, soil moisture, water vapour, as well as CO2 concentration of the ambient air, and the absorption of solar visible radiation, which is computed for three canopy layers. Maintenance respiration is strongly increasing with temperature (Ra ~ exp{const(T – T0)/T}; T canopy temperature [K], T0 = 298.16 K) and is inhibited during night. Growth respiration is assumed to be 20% of the resulting net assimilation. The photosynthesis module implicitly includes a temperature acclimation of photosynthesis up to 55°C leaf temperature (Kattge and Knorr 2006). Above 55°C photosynthesis and maintenance respiration are reduced to zero at 60°C (Collatz et al. 1991; Cuntz et al. 2003). Under present climate conditions and present atmospheric CO2 concentration the scheme specifies a global net primary productivity (NPP) of about 66 PgC/ year. This biomass is allocated to a wood pool and a pool representing active plant tissue (leaves, fine roots, etc.). From the latter pool carbon is transferred to the soil at a rate proportional to the leaf shedding rate, whereas wood is decomposed into soil assuming a fixed life time. Soil carbon is partitioned into a pool with a short (about 1 year) and one with a long turnover time (about 100 years). It is released to the atmosphere by heterotrophic respiration, which is depending linearly on soil moisture and exponentially on soil temperature (Q10 = 1.5; Raich and Potter 1995). Altogether the total biomass stored by the land biosphere according to this scheme is about 1,850 PgC under present conditions. Vegetation phenology is differentiated according to five phenotypes: evergreen, summergreen, raingreen forest or shrubland, grassland, and managed (non-forest) areas. It is completely independent of predefined calendar dates and solely driven by temperature, soil moisture and NPP. Up to now JSBACH considers no nutrient limitation. Three-dimensional transport of carbon within the ocean and the atmosphere as well as the exchange with the land biosphere are calculated each time step, so that the daily and the seasonal cycle of the atmospheric CO2 concentra- T. J. Raddatz et al.: Will the tropical land biosphere dominate the climate–carbon cycle feedback during the twenty-first century? 567 tion are resolved. The carbon flux between the ocean and the atmosphere is determined once per day. 3 Setup of the simulations As the carbon cycle includes long-term processes, the coupling of the ocean biogeochemistry and the land biosphere to the AOGCM causes trends in the carbon budgets and possibly also in the climate state for several centuries. To reach a pre-industrial steady state of the coupled climate–carbon cycle system without a huge computational cost, we employ the following procedure. As a starting point we utilize the pre-industrial climate state of the AOGCM, which has been well equilibrated in a former simulation. This approach is plausible, because the introduction of the new land surface scheme JSBACH implies only limited changes in the modeled climate, as the soil scheme and the albedo calculations are adopted from ECHAM5. So, as a first step, the output of the AOGCM is repeatedly used to drive the ocean biogeochemistry, until it reaches equilibrium. In a second step we run the coupled climate–carbon cycle model with a fixed atmospheric CO2 concentration of 285 ppm and repeatedly drive offline simulations of the land carbon pool model with the resulting records of NPP, leaf area index (LAI), soil moisture and temperature. This yields a first estimate of equilibrium values for the land carbon pools. These are introduced into the climate–carbon cycle model, and the second step is repeated with a freely evolving atmospheric CO2 distribution. Thereby the land carbon pools are equilibrated with respect to the daily and seasonal cycles of the atmospheric CO2 concentration. Finally, we perform a 170-year control integration with the coupled model system to assure that both, the simulated climate and the carbon cycle, are in a steady state. For the whole control run period the drift is less than 0.2 K in surface temperature and less than 1 ppm in atmospheric CO2 concentration. In order to separate the effects of increasing atmospheric CO2 and greenhouse gas warming on the vegetation and ocean we perform two transient simulations with the coupled climate–carbon cycle model, which both start from the pre-industrial equilibrium state described above. The setup of these two transient simulations follows the specifications of the Coupled Carbon Cycle Climate Model Intercomparison Project (C4MIP; Friedlingstein et al. 2006). They are conducted by adding anthropogenic CO2 emissions to the lowest layer of the atmosphere as the only forcing: following Marland et al. (2003) for fossil fuel emissions as well as Houghton (2003) for land use change emissions with a total of 418 PgC during the period 1860–1999 and, following the Fig. 1 Temporal course of near surface air temperature (°C) over land (solid line) and the ocean (dotted line) in the coupled simulation (red), the uncoupled simulation (blue), and as observed (black; Jones et al. 1999) SRES A2 scenario, 1,770 PgC during the twenty-first century. Biogeophysical effects of land-use change (e.g. albedo modifications) and the impact of land-use change on vegetation productivity are not considered. In the first transient simulation the complete climate– carbon cycle feedback is taken into account, so that anthropogenic CO2 acts on the carbon cycle and, as a greenhouse gas, on the climate. This model run is hereafter called ‘coupled’. In the second transient simulation greenhouse warming is suppressed by fixing the CO2 concentration in the radiation code, so that anthropogenic CO2 affects only photosynthesis, stomatal conductance and the ocean biogeochemistry (‘uncoupled’).1 Accordingly, the difference in surface CO2 fluxes between both runs reflects the impact of greenhouse gas warming on the carbon cycle. Figure 1 shows the temporal course of global, annual average near surface air temperature in both simulations. The coupled run exhibits the expected global warming with a larger trend over the land than over the ocean. By chance, the warming during the historical period is of similar amplitude as the one observed, despite numerous forcings are neglected (other anthropogenic greenhouse gases (CH4, CFCs, N2O, O3), anthropogenic aerosol, volcanic and solar). Nevertheless, this coincidence facilitates the comparison of the simulated carbon cycle with observations in Sect. 4 as similar global changes in climate act on the modeled and the real carbon cycle. Furthermore, even the 1 We choose to call the simulations ‘coupled’ and ‘uncoupled’ as this designation is commonly used to specify this setup of simulations. Nevertheless, atmosphere/ocean as well as atmosphere/land fluxes of energy, water and CO2 are calculated and exchanged each time step in both runs. 123 568 T. J. Raddatz et al.: Will the tropical land biosphere dominate the climate–carbon cycle feedback during the twenty-first century? spatial pattern of warming during the twentieth century is quite similar in the coupled run and the observations (not shown), despite the regionally confined forcing of anthropogenic aerosol is neglected in the simulation. The uncoupled run shows no apparent trend in air temperature over the ocean and a slight warming trend over land, which is caused by reduced transpiration due to stomatal closure induced by elevated atmospheric CO2. 4 Present carbon cycle Although the sub-models have been extensively validated (Knorr 2000; Knorr and Heimann 2001; Marsland et al. 2003; Jungclaus et al. 2006; Roeckner et al. 2005; Wetzel et al. 2005), their biases may be amplified in the coupled system. Therefore it has to be confirmed, as a prerequisite for a realistic view on the future carbon cycle, that the model reflects the major features of the present carbon cycle. Table 1 Comparison of simulated carbon uptake (PgC/year) for 1980–1989 and 1990–1999 in the coupled run with estimates from IPCC and a recent inversion study 1980s Ocean 1990s Land Ocean Land IPCC TAR 1.9±0.6 1.9±1.1a 1.7±0.5 3.0±1.1a Roedenbeck et al. (2003)b 1.8±0.3 2.2±0.9 2.3±0.2 2.6±0.9 Climate–carbon cycle modelc 1.96±0.07 2.18±0.25 2.24±0.07 2.01±0.25 a Assuming emissions due to land-use change as in IPCC Special Report on Land Use, Land-use Change and Forestry (2000) b Assuming a net riverine land–ocean C transport of 0.6 PgC/year and emissions due to land-use change as in footnote a c Uncertainties are rms of decadal net carbon fluxes of the control simulation Fig. 2 Average monthly deviation from annual mean atmospheric CO2 concentration at Shemya Island (left, 53°N 174°E, 1986–2002) and Ocean Station M (right, 66°N 2°E, 1982–2002) as observed (solid line; Tans and Conway 2005) and simulated in the coupled run (dashed line) 123 Table 1 shows the carbon uptake during the 1980s and 1990s of the ocean and land biosphere in the coupled simulation, which agree within the error estimates with the values inferred from inversion studies (Roedenbeck et al. 2003) and N2/O2 measurements (Prentice et al. 2001). The only exception is the model’s smaller land uptake during the 1990s, which might be explained by the Pinatubo eruption in 1991. During the 3 years after this volcanic event the observed atmospheric CO2 growth rate was exceptionally low. According to inversion studies (Roedenbeck et al. 2003; Peylin et al. 2005) as well as simulations with a climate–carbon cycle model (Jones and Cox 2001) this is attributable to enhanced land carbon uptake. In our model simulations, the aerosol distribution is set constant in time, so that the impact of the Pinatubo eruption on the land biosphere is not included. Another explanation for a part of the lower land uptake during the 1990s is the models considerable variability in decadal land carbon uptake. The one of the 1990s (2.01 PgC/year) seems to be low by chance, i.e. a result of random fluctuations, as the land uptake is considerably higher in the adjacent decades (2.18 and 2.62 PgC/year for the 1980s and 2000–2009, respectively). A comparison of observed and modeled atmospheric CO2 concentration reveals, that the amplitude of the seasonal cycle is underestimated over the North Pacific by about 30% (Fig. 2). There the modeled outgassing of CO2 from the ocean in the season July to September lifts up the late summer minimum in atmospheric CO2 concentration, which disagrees with surface water pCO2 measurements (Takahashi 2002). By contrast, over the North Atlantic the amplitude of the simulated seasonal cycle of atmospheric CO2 and the seasonality of the ocean/atmosphere CO2 flux are in accordance with the observations. As a general bias of the land biosphere model would affect the atmospheric CO2 records over both ocean basins, we infer, that the low amplitude of the seasonal cycle in atmospheric CO2 concentration over the North Pacific is at least partly attributable to the bias of the ocean biogeochemistry model in this region. T. J. Raddatz et al.: Will the tropical land biosphere dominate the climate–carbon cycle feedback during the twenty-first century? 569 Fig. 3 Anomalies of the sea surface temperature (SST) in the Nino3 region (solid line) and anomalies in the rate of carbon uptake (dashed line) as observed (left) and simulated in the coupled run (right). Anomalous C uptake is derived from detrended 12-month increments of monthly atmospheric CO2 concentration at Mauna Loa (Keeling and Whorf 2005) scaled to PgC/year. All curves are smoothed by taking 5-month averages. Observed SST is available at the Climate Prediction Center (CPC) After removing the seasonal cycle and the upward trend, the remaining interannual anomalies in atmospheric CO2 content are known to be tightly correlated with the El Nino Southern Oscillation (ENSO) except during the three years after the Pinatubo eruption and lag this climate mode by 4 months (Jones et al. 2001; Knorr et al. 2005b). Figure 3 demonstrates, that this is well reproduced by the model. As the model overestimates the amplitude of ENSO by about 50% (Jungclaus et al. 2006), the interannual variations in the carbon cycle are also too large. Accounting for this bias in the climate model the carbon cycle model is still a bit too sensitive to ENSO. A linear regression reveals a simulated flux of 0.87 PgC/year per degree change of the Nino3 index in comparison to 0.60 PgC/year per degree in the observational Nino3 record before the disturbances by the major volcanic eruptions of El Chichon (1982) and Pinatubo (Jones and Cox 2001; Jones et al. 2001). The simulated sequence of ENSO events is not related to the observed one, because it is determined by the model’s internal atmosphere/ocean dynamics and is independent of external forcings like the CO2 emissions, which are the only reference to the calendar in the model simulation. Nevertheless, the reproduced relationship between ENSO and variations in atmospheric CO2 concentration indicates, that the carbon cycle model responds reasonably to climate anomalies. The evaluation of atmospheric CO2 variations is summarized in Fig. 4 by a comparison of the modeled concentration at Mauna Loa (Hawaii) with the well-known observational record. Similar to Shemya Island (see Fig. 2, left), the amplitude of the seasonal cycle is underestimated, whereas the simulated interannual variability is too large (see Fig. 3). The accumulation of atmospheric CO2 in the years 1958–2004 is overestimated by about 4 ppm or 8 PgC, which is a small deviation in comparison to the amount of emissions released to the atmosphere during this period (311 PgC). At the present stage of observational evidence it seems to be impossible to decide, whether this small deviation is caused by an underestimation of simulated land/ocean uptake or an overestimation of prescribed land-use change emissions. So, a rigorous validation/calibration of the models long-term carbon uptake is not possible on the basis of the observed trend in atmospheric CO2 alone. Accordingly, most current climate–carbon cycle models exhibit differences between observed and simulated atmospheric CO2 of a similar magnitude (Friedlingstein et al. 2006). Nevertheless, their estimates of atmospheric CO2 concentration in the year 2100 diverge by about 300 ppm. Obviously, it is not sufficient to solely regard the historic evolution of globally averaged atmospheric CO2 concentration to constrain future land/ocean carbon uptake under a given emission scenario. 5 Future carbon cycle Fig. 4 Monthly CO2 concentration at Mauna Loa Hawaii as observed (blue; Keeling and Whorf 2005) and simulated in the coupled model run (red) The storage of anthropogenic carbon in the atmosphere, in the ocean and on land is summarized in Fig. 5. Global ocean uptake is nearly identical in both simulations (about 123 570 T. J. Raddatz et al.: Will the tropical land biosphere dominate the climate–carbon cycle feedback during the twenty-first century? Fig. 5 Accumulation of anthropogenic carbon on land (solid line), in the atmosphere (dashed line) and in the ocean (dotted line) during the coupled simulation (red) and the uncoupled simulation (blue) 520 PgC). Therefore, the difference in the atmospheric CO2 concentration between both runs, and hence the positive climate–carbon cycle feedback is mostly driven by the warming of 5.8 K over land areas (Fig. 1) and its influence on the carbon uptake there. In the year 2100 this amounts to 655 PgC in the uncoupled run and is reduced to 484 PgC in the coupled run. The regional distribution of the difference in net surface carbon flux between the two simulations is illustrated in Fig. 6 for the end of the twenty-first century. The North Atlantic Ocean takes up less carbon in the coupled run, because deep water formation in the Labrador Sea and Greenland Sea decreases and the Thermohaline Circulation (THC) slows down from 21 to 15 Sv (1 Sv = 106 m3/s). This is partly offset by an increasing carbon flux into the ocean in areas where sea ice melts. Globally the impact of the greenhouse warming on the oceanic sink is limited, as the finite buffer capacity of the mixed-layer restricts the Fig. 6 Difference in uptake of anthropogenic C between the coupled simulation and the uncoupled simulation in the period 2070–2100. Regions with negative (positive) values take up less (more) carbon under global warming conditions and contribute to a positive (negative) climate–carbon cycle feedback 123 carbon uptake in both runs and the decline of the THC sets in too late to considerably reduce the carbon uptake during the whole coupled simulation. Interestingly, most C4MIP climate–carbon cycle models with an explicit ocean circulation show a similar sensitivity of the ocean carbon uptake to global warming (Friedlingstein et al. 2006), which may be due to the same fundamental mechanism. Figure 6 also shows, that the land biosphere stores more carbon at high latitudes during the model simulation with greenhouse warming than in the one without. By contrast, the carbon uptake of the tropical land areas is much smaller under climate change conditions, than in the uncoupled case. This opposite response of the cold and the warm terrestrial biosphere to global warming has also been reported by Zeng et al. (2004). It is considered here in more detail by regarding gross primary production (GPP) and net primary production (NPP) in the latitudinal bands 30°S– 30°N (tropics) and 30°N–90°N (northern temperate/boreal) as displayed in Fig. 7. During the uncoupled simulation GPP increases by about 50% in both regions, reflecting the strong CO2 fertilization inherent to the photosynthesis scheme employed. The enhancement of NPP is even more pronounced. It amounts to 75% in the temperate/boreal range and to 100% in the tropics. The different growth rates of GPP and NPP can be explained by the way elevated atmospheric CO2 acts on vegetation. It raises GPP, but it does not significantly affect maintenance respiration (Long et al. 2004), if a possible downregulation of photosynthetic capacity is not considered as in our model. Therefore, the simulated absolute increase in NPP by CO2 fertilization is generally about 80% of that in GPP. The growth of NPP relative to its pre-industrial value thus depends on the pre-industrial ratio of NPP to GPP. This is smaller for tropical than for temperate/boreal vegetation due to high plant respiration costs at high temperature. Consequently, the relative increase in NPP is particularly pronounced in the tropics. Additionally, NPP is enhanced indirectly by a higher soil moisture level and less water stress due to stomatal closure under elevated atmospheric CO2. Again the tropical and subtropical areas profit more as water stress is more frequent and severe there than in the temperate/boreal zone. Overall, these mechanisms amplify the dominant role of the tropics for global NPP under elevated atmospheric CO2 and result in a sizable carbon uptake of northern cold ecosystems (179 PgC) and a vigorous carbon uptake of the tropics (445 PgC) during the simulation without global warming. In the coupled run the vegetation of the temperate/boreal region profits from the warmer climate—GPP and NPP almost double during the simulated period. By contrast, GPP in the tropics is not affected by the change in climate. This is partly a result of the only small increase in GPP T. J. Raddatz et al.: Will the tropical land biosphere dominate the climate–carbon cycle feedback during the twenty-first century? 571 Fig. 7 Annual terrestrial GPP (dashed line) and NPP (solid line) in the coupled simulation (red) and in the uncoupled simulation (blue) for the temperate/boreal region (30°N– 90°N, left) and the tropics (30°S–30°N, right) with temperature in an already hot environment, but also more water stress due to decreased soil moisture in most tropical and subtropical areas keeps GPP low. In turn, NPP in the tropics increases less in the coupled than in the uncoupled run due to enhanced maintenance respiration at higher temperature. This mechanism is still quite uncertain, because plant respiration may acclimate to a warmer environment (King et al. 2006). Nevertheless, the warming advances NPP by about 4 PgC/year in the temperate/boreal range, whereas it decreases NPP by about 12 PgC/year in the tropics. Heterotrophic respiration (not shown) is in turn enhanced especially in high northern latitudes due to the pronounced warming there, but this only partially offsets the latitudinal contrast in the response of NPP to global warming. Therefore, the different sign in the net carbon flux of the temperate/boreal and tropical land areas introduced by greenhouse warming is mostly attributable to opposite trends in NPP. Altogether, the change in climate implies a minor land uptake of carbon in the temperate/ boreal region of 11 PgC and a major loss of 178 PgC in the tropics. 6 Discussion of climate–carbon cycle feedback On the basis of simulations with a climate–carbon cycle model we find a dominant role of the tropical land biosphere for the climate–carbon cycle feedback in the twenty-first century. In view of neglected aspects of the carbon cycle (population dynamics of vegetation, fire, nutrient limitation) and uncertain parameters (Q10) we discuss here, in how far this result may depend on the model employed. The introduction of population dynamics in the treatment of the vegetation would probably amplify the contrast in the response of the temperate/boreal and the tropical land biosphere to global warming. A northward shift of the tree-line (i.e. the conversion of tundra into taiga) would further enhance NPP in the high northern latitudes, increase the carbon storage in living vegetation, and enlarge the carbon sink there. Furthermore, a possible dieback of the Amazon forest (Cox et al. 2004) would cause a pronounced release of CO2 to the atmosphere in this region. During our coupled simulation Amazonian precipitation decreases by about 10% while the temperature rises by about 7 K in this region. In summary, these processes would reinforce the dominant role of the tropics for the climate–carbon cycle feedback in our simulations. Vegetation fires are probably also a major actor of the global carbon cycle, but current estimates of fire emissions are still very uncertain (Boschetti et al. 2004). As vegetation fires are not considered in our simulations, heterotrophic respiration implicitly includes the flux of carbon released to the atmosphere by fire. Heterotrophic respiration is parameterized to increase with soil moisture, but fires are obviously more vigorous under dry conditions. So the omission of vegetation fires may introduce a systematic bias in the carbon cycle during the transient simulations via trends in the hydrological cycle. These are an increase of precipitation in the temperate/boreal and a decrease of precipitation in the sub-tropical range during the coupled run as in most other GCM studies (Wang 2005). Soil moisture increases due to global warming in most parts of the boreal region, India and some other small areas. A pronounced loss of soil moisture occurs in some sub-tropical and the Sahelian region, so that globally averaged soil moisture keeps at the same level during the coupled run, but is reduced with respect to the uncoupled run. Overall, the model results show no obvious indication for an additional massive release of carbon by fires in the temperate/boreal range as a consequence of global warming. Therefore, the introduction of vegetation fires in the model would probably not alter the main conclusion, that the climate–carbon cycle feedback is mostly driven by the tropical land biosphere. The temperature dependence of heterotrophic respiration is another major uncertainty in the simulation of the carbon cycle and an appropriate value of Q10 may as well be larger (Knorr et al. 2005a) or smaller (Rayner et al. 2005) than the one we prescribe in our model (Q10 = 1.5 with respect to soil temperature). For most other current 123 572 T. J. Raddatz et al.: Will the tropical land biosphere dominate the climate–carbon cycle feedback during the twenty-first century? climate–carbon cycle models Q10 = 2 is used (Friedlingstein et al. 2006). In order to estimate the sensitivity of the terrestrial carbon uptake to the choice of Q10 we perform offline calculations of the carbon pool model. These were either driven by the output of the coupled or the uncoupled run and therefore exclude the feedbacks due to a concomitant alteration of the atmospheric CO2 concentration. Nevertheless, they clearly demonstrate a distinct influence of Q10 on the response of soil carbon to global warming. Setting Q10 = 2, the difference in the amount of soil carbon between the coupled and the uncoupled run is about 80 PgC larger in the year 2100 than in the standard simulations. Furthermore, the temperate/boreal land biosphere changes the sign in its response to global warming. For Q10 = 2 the carbon uptake is smaller in the coupled than in the uncoupled run, but this contribution to a positive climate– carbon cycle feedback is still about three times smaller than the one of the tropical land biosphere. It should be mentioned that several hundred PgC stored in northern peat lands and permafrost areas are not considered by the model. On the long-term this carbon may be released due to the strong warming in the boreal regions. The decomposition rate should be very low, because of the still low temperatures there and a lack of oxygen in these soils. Nevertheless, uncertainties are large and a major additional flux of CO2 from these reservoirs during the twenty-first century cannot be excluded. A more comprehensive way to assess the individual contributions of the carbon cycle components to the climate–carbon cycle feedback is a look at the results of C4MIP. This intercomparison project comprises 11 coupled climate–carbon cycle models, which are described by Friedlingstein et al. (2006). They also present a feedback analysis, which is based on globally averaged output of coupled and uncoupled simulations forced by historical and SRES A2 emissions of CO2. Referring to the same set of simulations, Fig. 8 displays the change in carbon storage due to greenhouse warming separately for the ocean, the temperate/boreal and tropical land biosphere. The model described in this paper is denominated ‘MPI’. Apparently, there are large differences among the models, which partly arise from their different climate sensitivity. At the time of CO2 doubling the increase in surface temperature ranges from 1.2 K (CSM-1) to 2.5 K (MPI, HadCM3LC, IPSLCM4-LOOP). Thus the change in climate as the driving force of the signals shown in Fig. 8 varies considerably, although the primary forcing (i.e. the anthropogenic CO2 emissions) is the same for all models. Still this variation may not be spanning model uncertainty, as all land biosphere models considered have a similar structure, based on PFTs and with generic GPP versus temperature functions. But the temperature sensitivity of these functions 123 Fig. 8 Difference in cumulated carbon uptake (1860–2100) between the coupled run and the uncoupled run for the major carbon cycle components–ocean (left), temperate/boreal (30°N–90°N, middle) and tropical land biosphere (30°S–30°N, right). Displayed are results derived from 11 climate–carbon cycle models that contributed to the C4MIP intercomparison (Friedlingstein et al. 2006) differs between models, e.g. some models use a Gaussian distributed temperature dependence of photosynthesis with a globally constant optimum temperature (Cox et al. 1999). Matthews et al. 2005 showed that the value of this optimum temperature has a substantial impact on the simulated carbon-cycle climate feedback. By contrast, JSBACH uses an implicit temperature acclimation (Knorr 2000; Kattge and Knorr 2006), with the consequence that photosynthesis has a broad optimum between 40 and 55°C. This is in contrast to other models, with photosynthesis in the tropics being at/or beyond optimum at current temperatures. Another noticeable feature of some models is the uptake of carbon by the ocean in response to global warming, despite the fact that an increase in temperature should lead to an outgassing of CO2 from the ocean. This paradoxical behavior can be explained by the interaction of the carbon cycle components. Global warming implies a pronounced release of land carbon, which enhances the atmospheric CO2 growth rate. The higher CO2 partial pressure then eventually overcompensates the impact of greenhouse warming on the ocean/atmosphere carbon exchange. In turn, the ocean uptake influences the land uptake via the atmospheric CO2 content as all models exhibit CO2 fertilization. In general, the differences in CO2 fluxes between the coupled and uncoupled run arise from the change in climate and the modified atmospheric CO2 concentration. So, the interrelation of the carbon cycle components in their responds to greenhouse warming could be disentangled properly by repeating the coupled and uncoupled simulations with the same prescribed atmospheric CO2 record. Despite the complex interdependence of the sub-systems it is obvious from the results shown in Fig. 8, that the land T. J. Raddatz et al.: Will the tropical land biosphere dominate the climate–carbon cycle feedback during the twenty-first century? biosphere contributes more to the positive climate–carbon cycle feedback than the ocean. All 11 models simulate a stronger reduction in land carbon uptake than in ocean carbon uptake due to greenhouse warming. Nine models attribute this change in land carbon storage mostly to the tropical range. The two models simulating a stronger decrease in temperate/boreal carbon uptake (BERN-CC and CLIMBER) are both of intermediate complexity and are those, which include the Lund–Potsdam–Jena vegetation model (LPJ, Sitch et al. 2003). Their small signal in tropical land carbon uptake can be traced back to a slight increase of tropical NPP in response to global warming, whereas most of the other models simulate a distinct reduction of NPP in the tropics. As this is also the case for LPJ, if it is driven by, e.g. the output of the Hadley Center climate model (Cramer et al. 2001), the regional climate trends in the coupled run of BERN-CC and CLIMBER are probably the main reason for their exceptional increase of tropical NPP. Overall, the C4MIP results confirm, that the impact of global warming on the net carbon uptake will probably be largest for tropical land areas. Figure 8 also shows, that the magnitude of this contribution to a positive climate–carbon cycle feedback is a bit more uncertain than that of the temperate/boreal region and much more uncertain than that of the ocean. The model results differ most with respect to the tropical land biosphere. 7 Conclusion The model study presented here indicates a substantial positive feedback between the global climate and the carbon cycle. Following the SRES A2 scenario, this feedback is predominantly set up by the response of the tropical land biosphere to global warming during the twenty-first century. This result is mostly a consequence of a decline in tropical NPP with increasing temperature, whereas some arguments are presented, that it may be quite independent of particular assumptions about population dynamics of vegetation, fires and the temperature dependence of heterotrophic respiration. Furthermore, a comparison of current climate–carbon cycle models (C4MIP) reveals, that a climate-driven change in carbon uptake is potentially largest in tropical land areas. There the models also show a bit larger differences in the estimated impact of global warming on carbon uptake than in the temperate/boreal zone. In view of these results the title of this article is formulated as a question, which is crucial to answer with regard to projections of the future development of the carbon cycle. It may also be a question beyond empirical knowledge about ecosystem functioning. If the world’s most productive ecosystem, the Amazon forest, will warm by as much 573 as 7 K as in our coupled simulation, it will undergo climate conditions that are out of the range ecosystems experience today. Projected carbon uptake of such land areas is accordingly uncertain. On the one hand this highlights the ecological challenge to deduce, how already hot ecosystems will respond to a further increase in temperature. On the other hand it challenges the modeling community to constrain the uncertainty in climate trends over tropical land areas. At least it should be an argument to enhance observations, biogeochemical as well as meteorological, in tropical forest areas, where apparently there is still a lack of monitoring stations in comparison to temperate regions. Acknowledgments This research was financed by the German Ministry for Education and Research (BMBF) under two DEKLIM projects, grants 01LD0106 and 01LD0024, and by the European Community under the CYCLOPES project. The simulations were performed on the NEC SX-6 supercomputer installed at the German Climate Computing Centre (DKRZ) in Hamburg. We thank all C4MIP participants for contributing their model results and numerous helpful comments. References Boschetti L, Eva HD, Brivio PA, Gregoire JM (2004) Lessons to be learned from the comparison of three satellite-derived biomass burning products. Geophys Res Lett 31:L21501 doi:10.1029/ 2004GR021229 Collatz GJ, Ball JT, Grivet C, Berry JA (1991) Physiological and environmental regulation of stomatal conductance, photosynthesis and transpiration: a model that includes a laminar boundary layer. Agric For Meteorol 54:107–136 Collatz GJ, Ribas-Carbo M, Berry JA (1992) Coupled photosynthesisstomatal conductance model for leaves of C4 plants. Aust J Plant Physiol 19:519–538 Cox PM, Betts RA, Bunton CB, Essery RL, Rowntree PR, Smith J (1999) The impact of new land surface physics on the GCM simulation of climate and climate sensitivity. Clim Dyn 15:183– 203 Cox PM, Betts RA, Jones CD, Spall SA, Totterdell IJ (2000) Acceleration of global warming due to carbon–cycle feedbacks in a coupled climate model. Nature 408:184–187 Cox PM, Betts RA, Collins M, Harris PP, Huntingforth C, Jones CD (2004) Amazonian forest dieback under climate-carbon cycle projections for the 21st century. Theor Appl Climatol 78:137– 156 doi:10.1007/s00704–004–0049–4 Cramer W, Bondeau A, Woodward FI, Prentice IC, Betts RA, Brovkin V, Cox PM, Fisher V, Foley JA, Friend AD, Kucharik C, Lomas MR, Ramankutty N, Sitch S, Smith B, White A, Young-Molling C (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 Cuntz M, Ciais P, Hoffmann G, Knorr W (2003) A comprehensive global three-dimensional model of d18O in atmospheric CO2: 1. Validation of surface processes. J Geophys Res 108(D17):4527 doi:10.1029/2002JD003153 Farquhar GD, von Caemmerer S, Berry JA, (1980) A biogeochemical model of photosynthesis in leaves of C3 species. Planta 149:78–90 Friedlingstein P, Dufresne J-L, Cox PM, Rayner P (2003) How positive is the feedback between climate change and the carbon cycle? Tellus 55B:692–700 123 574 T. J. Raddatz et al.: Will the tropical land biosphere dominate the climate–carbon cycle feedback during the twenty-first century? Friedlingstein P, Cox P, Betts R, Bopp L, von Bloh W, Brovkin V, Cadule P, Doney S, Eby M, Fung I, Bala G, John J, Jones C, Joos F, Kato T, Kawamiya M, Knorr W, Lindsay K, Matthews HD, Raddatz T, Rayner P, Reick C, Roeckner E, Schnitzler K-G, Schnur R, Strassmann K, Weaver AJ, Yoshikawa C, Zeng N (2006) Climate - carbon cycle feedback analysis, results from the C4MIP model intercomparison. J Climate 19:3337–3353 doi:10.1175JCLI3800.1 Global Carbon Project (2003) Science framework and implementation. In: Canadell JG, Dickinson R, Hibbard K, Raupach M, Young O (eds) Earth system science partnership (IGBP, IHDP, WCRP, DIVERSITAS) Report No. 1, Global Carbon Project Report No. 1, 69 pp, Canberra Houghton RA (2003) Revised estimates of the annual net flux of carbon to the atmosphere from changes in land use and land management 1850–2000. Tellus 55B:378–390 Jones PD, New M, Parker DM, Martin S, Rigor IG (1999) Surface air temperature and its changes over the past 150 years. Rev Geophys 37(2). doi:10.1029/1999RG900002 Jones CD, Cox PM (2001) Modeling the volcanic signal in the atmospheric CO2 record. Global Biogeochem Cycles 15:453– 465 Jones CD, Collins M, Cox PM, Spall SA (2001) The carbon cycle response to ENSO: a coupled climate-carbon cycle model study. J Climate 14:4113–4129 Jungclaus JH, Keenlyside N, Botzet M, Haak H, Luo JJ, Latif M, Marotzke J, Mikolajewicz U, Roeckner E (2006) Ocean circulation and tropical variability in the coupled model ECHAM5/MPI-OM. J Climate 19:3952–3972. doi:10.1175JCLI3827.1 Kattge J, Knorr W (2007) The temperature dependence of photosynthetic capacity in a photosynthesis model acclimates to plant growth temperature: a re-analysis of data from 36 species. Plant Cell Environ (accepted) Keeling CD, Whorf TP (2005) Atmospheric CO2 records from sites in the SIO air sampling network. In: Trends: a compendium of data on global change. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, Oak Ridge King AW, Gunderson CA, Post WM, Weston DJ, Wullschleger SD (2006) Plant respiration in a warmer world. Science 312. doi:10.1126/Science.1114166 Knorr W (2000) Annual and interannual CO2 exchange of the terrestrial biosphere: Process based simulations and uncertainties. Global Ecol Biogeogr 9:225–252 Knorr W, Heimann M (2001) Uncertainties in global terrestrial biosphere modeling. Part II: global constraints for a processedbased vegetation model. Global Biogeoch Cycles 15:227–246 Knorr W, Prentice IC, House JI, Holland EA (2005a) Long-term sensitivity of soil carbon turnover to warming. Nature 433:298– 301 Knorr W, Scholze M, Gobron N, Pinty B, Kaminski T (2005b) Global-scale drought caused atmospheric CO2 increase. Eos Trans AGU 86(18): 178 Long SP, Ainsworth EA, Rogers A, Ort DR (2004) Rising atmospheric carbon dioxide: Plants face the future. Annu Rev Plant Biol 55:591–628. doi: 10.1146/annurev.arplant.55.031903.141610 Loveland TR, Reed BC, Brown JF, Ohlen DO, Zhu Z, Yang L, Merchant JW (2000) Development of a global land cover characteristics data base and IGBP DISCover from 1 km AVHRR data. Int J Remote Sens 21:1303–1330. doi: 10.1080/ 014311600210191 Marland G, Boden TA, Andres RJ (2003) Global, regional and national emissions. In: Trends: a compendium of data on global 123 change. Carbon Dioxide Information Center, Oak Ridge National Laboratory, U. S. Department of Energy, Oak Ridge Marsland SJ, Haak H, Jungclaus JH, Latif M, Roeske F (2003) The Max-Planck-Institute global ocean/sea ice model with orthogonal curvilinear coordinates. Ocean Modell 5:91–127 Matthews HD, Eby M., Weaver AJ, Hawkins BJ (2005) Primary productivity control of simulated carbon cycle–climate feedbacks. Geophys Res Lett 32:L14708. doi:10.1029/ 2005GL022941 Peylin P, Bousquet P, Le Quere C, Sitch S, Friedlingstein P, McKinley G, Gruber N, Rayner P, Ciais P (2005) Multiple constraints on regional CO2 flux variations over land and oceans. Global Biogeochem Cycles 19:GB1011. doi:10.1029/ 2003GB002214 Prentice IC et al (2001) The carbon cycle and atmospheric CO2. In: Houghton JT, Yihui D (eds) The Intergovernmental Panel on Climate Change (IPCC) Third Assessment Report. Cambridge University Press, New York, chap. 3, pp 185–237 Raich JW, Potter CS (1995) Global patterns of carbon dioxide emissions from soils. Global Biogeochem Cycles 9:23–36 Rayner PJ, Scholze M, Knorr W, Kaminski T, Giering R, Widmann H (2005) Two decades of terrestrial carbon fluxes from a carbon cycle data assimilation system (CCDAS). Global Biogeochem Cycles 19:GB2026. doi:10.1029/2004GB002254 Roeckner E, Baeuml G, Bonaventura L, Brokopf R, Esch M, Giorgetta M, Hagemann S, Kirchner I, Kornblueh L, Manzini E, Rhodin A, Schlese U, Schulzweida U., Tompkins A (2003) The general circulation model ECHAM5. Part I: Model description. Report 349, Max-Planck-Institut for Meteorology, Hamburg Roeckner E, Brokopf R, Esch M, Giorgetta M, Hagemann S, Kornblueh L, Manzini E, Schlese U, Schulzweida U (2005) Sensitivity of simulated climate to horizontal and vertical resolution in the ECHAM5 atmosphere model. J Climate 19:3771–3791. doi:10.1175JCLI3824.1 Roedenbeck C, Houweling S, Gloor M, Heimann M (2003) CO2 flux history 1982–2001 inferred from atmospheric data using a global inversion of atmospheric transport. Atmos Chem Phys Discuss 3:2575–2659 Sitch S, Smith B, Prentice IC, Arneth A, Bondeau A, Cramer W, Kaplan JO, Lewis S, Lucht W, Sykes MT, Thonicke K, Venevsky S (2003) Evaluation of ecosystem dynamics, plant geography and terrestrial carbon cycling in the LPJ dynamic global vegetation model. Global Change Biol 9:161–185 Takahashi T, Sutherland SC, Sweeney C, Poisson A, Metzl N, Tilbrook B, Bates N Wanninkhof R, Feely RA, Sabine C, Olafsson J, Nojiri Y (2002) Global sea-air CO2 flux based on climatological surface ocean pCO2 and seasonal biological and temperature effects. Deep-Sea Res 49:1601–1622 Tans PP, Conway TJ (2005) Monthly atmospheric CO2 mixing ratios from the NOAA CMDL carbon cycle cooperative global air sampling network, 1968–2002. In: Trends: a compendium of data on global change. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, Oak Ridge Wang G (2005) Agricultural drought in a future climate: results from 15 global climate models participating in the IPCC 4th assessment. Clim Dyn 25:739–753. doi:10.1007/s00382–0050057-9 Wetzel P, Winguth A, Maier-Reimer E (2005) Sea-to-air CO2 fluxes from 1948 to 2003. Global Biogeochem Cycles 19:GB2005. doi:10.1029/2004GB002339 Zeng N, Qian H, Munoz E, Iacono R (2004) How strong is carbon cycle-climate feedback under global warming? Geophys Res Lett 31:L20203. doi:10.1029/2004GL020904