SEARCH

SEARCH BY CITATION

abstract

  1. Top of page
  2. 1. Introduction
  3. 2. Model
  4. 3. Historical anthropogenic forcing and model calibration
  5. 4. Future carbon cycle and climate simulation
  6. 5. Conclusion
  7. 6. Acknowledgement
  8. references

A simple Earth system model, the Four-Spheres Cycle of Energy and Mass (4-SCEM) model, has been developed to simulate global warming due to anthropogenic CO2 emission. The model consists of the Atmosphere–Earth Heat Cycle (AEHC) model, the Four Spheres Carbon Cycle (4-SCC) model, and their feedback processes. The AEHC model is a one-dimensional radiative convective model, which includes the greenhouse effect of CO2 and H2O, and one cloud layer. The 4-SCC model is a box-type carbon cycle model, which includes biospheric CO2 fertilization, vegetation area variation, the vegetation light saturation effect and the HILDA oceanic carbon cycle model. The feedback processes between carbon cycle and climate considered in the model are temperature dependencies of water vapor content, soil decomposition and ocean surface chemistry. The future status of the global carbon cycle and climate was simulated up to the year 2100 based on the “business as usual” (IS92a) emission scenario, followed by a linear decline in emissions to zero in the year 2200. The atmospheric CO2 concentration reaches 645 ppmv in 2100 and a peak of 760 ppmv approximately in the year 2170, and becomes a steady state with 600 ppmv. The projected CO2 concentration was lower than those of the past carbon cycle studies, because we included the light saturation effect of vegetation. The sensitivity analysis showed that uncertainties derived from the light saturation effect of vegetation and land use CO2 emissions were the primary cause of uncertainties in projecting future CO2 concentrations. The climate feedback effects showed rather small sensitivities compared with the impacts of those two effects. Satellite-based net primary production trends analyses can somewhat decrease the uncertainty in quantifying CO2 emissions due to land use changes. On the other hand, as the estimated parameter in vegetation light saturation was poorly constrained, we have to quantify and constrain the effect more accurately.


1. Introduction

  1. Top of page
  2. 1. Introduction
  3. 2. Model
  4. 3. Historical anthropogenic forcing and model calibration
  5. 4. Future carbon cycle and climate simulation
  6. 5. Conclusion
  7. 6. Acknowledgement
  8. references

Global warming, resulting from anthropogenic greenhouse gas emissions such as CO2, is one of the most significant of the Earth's environmental problems. Anthropogenic CO2 has been emitted into the atmosphere since the 18th century, and it was estimated to be approximately 7.0 GtC yr−1 during the 1980s (Prentice et al., 2001). Approximately one half of the emissions remain in the atmosphere, and the other half are absorbed by the ocean and terrestrial biosphere (Prentice et al., 2001). Since global warming is caused by the coupled feedback cycles of energy and carbon (Charlson 2000; Prentice et al., 2001), it is essential to assess the impact of various feedback processes on CO2 uptake by the land and oceans to enable projection of the carbon cycle and climate of the future.

Most of the previous studies on projection of atmospheric CO2 and climate have not considered these feedback processes between the carbon cycle and climate systems. For instance, simple carbon cycle models were used to predict atmospheric CO2 concentrations under a given scenario of anthropogenic CO2 emissions, and then the predicted atmospheric CO2 variations were input into the General Circulation Model (GCM) for the future climate simulation. Then, dynamic carbon cycle models were applied under simulated climate scenarios for the analysis of future biosphere and ocean carbon cycle conditions (Cao and Woodward, 1998; Sarmiento et al., 1998).

Earth system models that integrate models of climate and carbon cycle were only recently developed. Cox et al. (2000) showed that the feedback processes could accelerate climate change based on the fully coupled, three-dimensional carbon–climate model. However, changes in CO2 emission levels due to land use change are not removed directly from the terrestrial biosphere in their model. Another problem is that their model generates higher atmospheric CO2 levels and temperatures compared with current levels, which may likewise be amplified in projections and may result in an exaggerated climate feedback effect.

Simple Earth system models were also developed for analysis of climate and carbon cycle feedback mechanisms (Lenton, 2000; Prentice et al., 2001). Lenton (2000) developed a simple Earth system model that consists of atmospheric, terrestrial and ocean carbon cycle models and an atmospheric climate model. In the third assessment report of the Intergovernmental Panel on Climate Change (IPCC), the Bern model and the Integrated Science Assessment Model (ISAM) were used as the simplified carbon–climate coupled models. Due to a complicated process-based dynamic global vegetation model (the LPJ model; Sitch, 2000), the Bern model cannot be regarded as a simple model. The simple models did not consider variation in vegetation area in a physically based manner or the impact of light availability on vegetation growth saturation (if vegetation density becomes high, vegetation growth is limited due to light availability), both of which have an important effect on the biospheric CO2 uptake.

In the present study, we developed a simple Earth system model, the Four Spheres Cycle of Energy and Mass (4-SCEM) model, for analysis of the coupled cycles of energy and carbon. The feedback processes included are (1) water vapor feedback, (2) the effect of CO2 fertilization on vegetation photosynthesis and (3) temperature effects on soil decomposition, and on the atmosphere–ocean carbon exchange. The predictions of future carbon cycle and climate conditions were analyzed based on the given scenario of anthropogenic CO2 emission levels. Sensitivities of the model parameters were also tested in this study. Details of the 4-SCEM model are described in section 2. Section 3 describes model calibrations, and the projection of future carbon cycle and climate conditions. The results of sensitivity studies are shown in section 4. Finally, section 5 presents the study's conclusions.

2. Model

  1. Top of page
  2. 1. Introduction
  3. 2. Model
  4. 3. Historical anthropogenic forcing and model calibration
  5. 4. Future carbon cycle and climate simulation
  6. 5. Conclusion
  7. 6. Acknowledgement
  8. references

The 4-SCEM model consists of two sub-models, one that is a carbon cycle model (Four Spheres Carbon Cycle model; 4-SCC) and an energy cycle model (Atmosphere–Earth Heat Cycle model; AEHC). The outline of 4-SCEM is shown in Fig. 1. Feedback processes included in the model are (1) CO2 fertilization effects on the biosphere, (2) biospheric productivity saturation effect due to increases in vegetation density, (3) temperature dependence of soil decomposition, (4) temperature dependence of ocean surface chemistry and (5) water vapor feedback in the atmosphere. A detailed description of each model and feedback process is provided below.

image

Figure 1. Outline of four spheres cycle of energy and mass model. It consists of atmospheric one-dimensional radiative convective model and terrestrial and ocean carbon cycle model. Each model is connected with feedback processes as described in the text.

Download figure to PowerPoint

2.1. Atmosphere–Earth Heat Cycle (AEHC) model

The AEHC model is a one-dimensional radiative convective model based on Hayashi (1991), shown in Fig. 1. The modeled atmosphere consists of 13 layers from surface to tropopause (assumed to be 12 km in height). The convective adjustment scheme (e.g. Manabe and Strickler, 1964) was used to make a realistic temperature profile in the troposphere.

The time variation of temperature in an atmospheric layer at height z can be written as

  • image(1)

where T(z), ρ(z), Cp, S(z, ν) and L(z, ν) are temperature, density, specific heat and net flux of shortwave and longwave at height z and wavenumber ν, respectively. The shortwave and longwave flux were obtained from the transmittance of each layer, incident solar flux at the top of the atmosphere and longwave flux from each atmospheric layer and the Earth's surface.

Transmittance of an atmospheric layer at height z and wavenumber ν is based on the Random model, and band equivalent widths of H2O and CO2 are derived from the absorption line data of Houghton (1977). H2O was expressed with the temperature feedback effect based on the Clausius–Clapeyron equation:

  • image(2)

where ρw, HR, P0, le and Rw are the atmospheric water density, relative humidity, a constant for the water vapor saturation curve (1.4 × 1011 Pa; Nakajima et al., 1992), latent heat of water (43655 J mol−1) and gas constant (8.314 J mol−1 K−1), respectively. A value of relative humidity HR was determined iteratively, as discussed later. We have also included the Rayliegh scattering (Manabe and Strickler, 1964) and ozone absorption (Lacis and Hansen, 1974) to estimate the incoming shortwave flux at the top of troposphere.

We have set one homogeneous cloud layer in the AEHC model. The cloud data, cloud cover ratio, height and optical thickness are shown in Table 1 derived from remote sensing data sets (Rossow et al., 1996). We have obtained the cloud radiative properties based on Stephens's parameterization (Stephens, 1978), which estimates the cloud optical properties from the cloud liquid water content. Temperature dependency of cloud liquid water content was derived from Somerville and Remer (1984):

  • image(3)

where LWC(t) and T(t) are cloud liquid water content and temperature at time t (t= 0 indicates its pre-industrial value), and f is the cloud liquid water feedback parameter. In this study, we set f= 0 (no cloud–temperature feedback), and its sensitivity will be analyzed in a future study.

Table 1. Parameters used in AEHC model
ParameterDescriptionValueNotice
HRRelative humidity0.68 
CpSpecific heat1004.0 J kg−1 K−1 
P0Water vapor saturation constant1.4 × 1011 PaNakajima et al., 1992
FCloud feedback parameter0 
 Cloud cover ratio0.623Rossow et al., 1996
 Cloud height4 kmRossow et al., 1996
ASurface albedo0.132Rossow et al., 1996
LWC(0)Pre-industrial cloud liquid water content23.5 g m−2Derived from ISCCP data and Stephens (1978)
ChHeat capacity of Earth's surface4.69 × 1023 J K−1 
AeSurface area5.101 × 1014 m2 
HfSensible and latent heat flux from surface38 W m−2 

Upwelling shortwave and longwave fluxes from the Earth's surface were calculated from surface albedo and temperature. Surface albedo was derived using the 8 yr averaged value from the ISCCP C2 data set (Rossow et al., 1996), and set at 0.12. Time variation of temperature on the Earth's surface (Ts) can be expressed as

  • image(4)

where Ch, Fd, σ and ae are the heat capacity of the Earth's surface (4.69 × 1023 J K−1; Huntingford and Cox, 2000), the net incoming surface flux (shortwave, longwave and sensible and latent heat flux), the Stefan–Boltzmann constant (5.67 × 10−8 W m−2 K−4) and surface area (5.101 × 1014 m2), respectively. Ocean heat uptake may have a significant effect on the climate, and we can include the effect in the parameter of heat capacity of earth surface by tuning its value.

Finally, we determined the value of relative humidity and latent and sensible heat transport. These values were obtained by iterative procedure to satisfy the following conditions: (1) the Earth's surface temperature of 288 K (average surface temperature) and (2) continuous temperature between the Earth's surface and the lowest atmospheric layer. We set the relative humidity parameter to reproduce atmospheric temperature. A relative humidity value of 0.67 was assumed, which is less than the average observed surface value of ∼0.77. Then, the latent and sensible heat transport value was determined to make the temperature profile continuous in the boundary between the Earth's surface and atmosphere. A smaller value of latent and sensible heat transport (this sum is 38 W m−2, which is smaller than the IPCC report of 102 W m−2, Kiehl and Trenberth, 1997) was used to make the surface temperature reasonable.

2.2. Four Spheres Carbon Cycle (4-SCC) model

The 4-SCC model is a box-type simple carbon cycle model based on Ichii et al. (2001), utilized for past and future carbon cycle simulation. It deals with carbon flows among the four spheres (atmosphere, biosphere, lithosphere and hydrosphere). The biosphere consists of two boxes, the terrestrial living biosphere and soil organic matter. The ocean model is a high-latitude exchange/interior diffusion–advection (HILDA) model (Siegenthaler and Joos, 1992, Shaffer and Sarmiento, 1995). Anthropogenic effects included in the model are industrial emissions (fossil fuel and cement production) and land-use changes. Details of this model are described below.

2.2.1. Terrestrial biospheric component The terrestrial biospheric model consists of two carbon reservoirs (vegetation and soil) and five carbon flows (gross primary production, respiration, litterfall, soil decomposition and land use change). Mass balances in the carbon reservoirs are as follows:

  • image(5)
  • image(6)

where Cveg(t) and Csoil(t) are carbon masses of the vegetation and soil reservoirs, and GPP(t), RES(t), LIT(t), LU(t) and DEC(t) are carbon fluxes of gross primary production (GPP), dark respiration, litterfall, CO2 emission due to land use change and soil decomposition, respectively. The anthropogenic carbon emission due to land-use changes is directly removed from the vegetation and soil organic matter box using a fractionation factor of removing carbon from biosphere and soil organic matter (Lp). In an approach that is different from other similar simple models (e.g. Bacastow and Keeling, 1973; Wigley, 1993; Kwon and Schnoor, 1994; Craig and Holmén, 1995; Jain et al., 1995; Kheshgi et al., 1996; Lenton, 2000), we included (1) vegetation area variation and (2) photosynthesis limitation due to light saturation in the process of GPP.

Based on the Production Efficiency model (NPP or GPP = PAR · FPAR · ɛ) PAR is photosynthetically active radiation, FPAR is the fraction of PAR absorbed by the canopy, and ɛ is the light utilization coefficient) (e.g. Kumar and Monteith, 1981), GPP(t) becomes

  • image(7)

where SOL is incoming photosynthetic active radiation (PAR), dt is integrated time, S(t) is the vegetation area at time t, keff is light extinction coefficient, klai is the vegetation density conversion factor from density to leaf area index (LAI) (LAI kg−1 m2), β is the CO2 fertilization factor and Catm(t) is atmospheric carbon mass at time t. The first term, SOL dtS(t), indicates the total available incoming PAR over the vegetation area, the second term, 1 − exp[−keffklaiCveg(t)/S(t)], is FPAR, which follows the Lambert–Beer law. In this study, we assumed that the leaf area index (LAI) is proportional to the vegetation density, Cveg(t)/S(t). The last term, ɛ{1 + β[Catm(t)/Catm(0)]}, shows light utilization efficiency, and includes the CO2 fertilization factor in logarithmic form (e.g. Bacastow and Keeling, 1973).

The carbon flux due to respiration and litterfall is proportional to the mass of living biosphere, and we included the temperature dependence of soil decomposition based on the Q10 equation (e.g. Raich and Schlesinger, 1992). These equations follow:

  • image(8)
  • image(9)
  • image(10)

where Kres, Klit and Kdec are constant and Q10 is the Q10 factor.

Although there are many feedback effects included in more detailed terrestrial models such as geographic distribution in vegetation and vegetation structural variation and precipitation, it is very difficult to include in the simplified model and is not well validated even in the detailed terrestrial models. The simplified model must include these features as for as possible, and these effects in the terrestrial biosphere will be included in a future study.

The reference model parameters were defined as follows and are shown in Table 2. The pre-industrial carbon contents of vegetation and soil organic matter were derived from Siegenthaler and Sarmiento (1993). The initial terrestrial net primary production (NPP) was assumed to be 55 GtC yr−1 based on the average of values estimated using terrestrial carbon cycle models (Cramer et al., 1999). The carbon flux in terrestrial GPP was assumed to be double that of NPP, 110 GtC yr−1, and respiration is 55 GtC yr−1. The fluxes of litter flow and soil decomposition are also 55 GtC yr−1 for each. The Lp value was the same as that in our previous study (Ichii et al., 2001) and set to be 0.7. The value of Q10= 2.0 was derived from a commonly used value (e.g. Cox et al., 2000; Potter et al., 1993).

Table 2. Parameters used in the 4-SCC model
ParameterDescriptionValueNotice
Catm(0)Atmospheric in pre-industrial era588 GtCEtheridge et al., 1999
Cveg(0)Vegetation in pre-industrial era610 GtCSiegenthaler and Sarmiento, 1993
Csoil(0)Soil C in pre-industrial era1600 GtCSiegenthaler and Sarmiento, 1993
GPP(0)GPP in pre-industrial era120 GtC yr−1Prentice et al., 2001
RES(0)Vegetation respiration in pre-industrial era60 GtC yr−1Prentice et al., 2001
LIT(0)Litterfall flux in pre-industrial era60 GtC yr−1Prentice et al., 2001
DEC(0)Soil decomposition in pre-industrial era60 GtC yr−1Prentice et al., 2001
LpFractionation factor of CO2 emission due to land use change0.7Ichii et al., 2001
KresPlant respiration rate constant60/610 yr−1RES(0)/Cveg(0)
KlitPlant litterfall rate constant60/610 yr−1LIT(0)/Cveg(0)
KdecSoil decomposition rate constant60/1600 yr−1DEC(0)/Csoil(0)
Q10Q10 factor2.0Potter et al., 1993; Cox et al., 2000
KVertical diffusion constant4700.0 m2 yr−1Siegenthaler and Joos, 1992
WUpwelling velocity in the interior0.73 m yr−1Siegenthaler and Joos, 1992
QLateral exchange0.002 381 m yr−1Siegenthaler and Joos, 1992
UExchange between surface and deep polar ocean53.0 m yr−1Siegenthaler and Joos, 1992
DeltaThe part of polar ocean0.16Shaffer and Sarmiento, 1995
DeltasThe part of surface layer free from ice0.10Shaffer and Sarmiento, 1995
DDepth of interior3800 mShaffer and Sarmiento, 1995
DsDepth of surface layer50 mShaffer and Sarmiento, 1995

The product of SOL dt S(0)ɛ, which we denote as Fgpp= SOL dtS(0)ɛ, was estimated as follows: SOL = 198 (W m−2) (incoming solar radiation at the Earth's surface; Kiehl and Trenberth, 1997) × 0.5 (half of solar radiation is PAR; McCree, 1981) × dt= 86 400 × 365 (s yr−1), S(0) = 8.7722 × 1013 m2 (the pre-industrial forest area from Houghton and Hackler, 2001), and ɛ= 1.25 × 10−6 gC J−1 (Heimann and Keeling, 1989). The result was roughly 350 GtC yr−1. Although there is a large uncertainty in each parameter, the values were used in the reference simulation. The product of kext and klai can be estimated based on the pre-industrial GPP and other parameters. The remaining parameter, β, was determined iteratively to fit the observed time variation of atmospheric CO2.

2.2.2. Ocean carbon cycle model We used the HILDA model (Siegenthaler and Joos, 1992; Shaffer and Sarmiento, 1995) as an ocean carbon cycle model. The ocean is divided into low and high latitudinal zones. The low latitude zone consists of a well mixed surface layer and a one-dimensional advective–diffusive deep ocean layer. The high latitude zone consists of a well mixed polar surface layer and a deep ocean layer. We show the HILDA model with tracer time variation only briefly, because Shaffer and Sarmiento (1995) did not describe the model equation in the time-variable version.

The time variations of tracer concentrations at the low latitude surface (ΦLS), low latitude interior [ΦI(z)], high latitude surface [ΦHS(z)] and high latitude deep [ΦHD] regions are as follows:

  • image(11)
  • image(12)
  • image(13)
  • image(14)

Table 2 shows detailed explanations of each model parameter, and detailed explanations of the model and its parameters are given in Shaffer and Sarmiento (1995). The boundary conditions are as follows:

  • image(15)
  • image(16)

In the case of ΣCO2, we set the source/sink terms as SI(z) = 0, SHD= 0 and SB= 0. The surface CO2 exchange was modeled based on Fujii et al. (2000), and was added to the source/sink term (SLS and SHS) for the ocean surface at low and high latitudes, instead of removing the gLS and gHS terms. CO2 piston velocity and alkalinity of low and high latitude ocean were set to 3.04 × 10−5 m s−1, 2360 μeq kg−1 and 2290 μeq kg−1 (Fujii, M. personal communication), respectively. Solubility (Weiss, 1974) and equilibrium constants of CO2 dissolution (Goyet and Poisson, 1989) were calculated as a function of temperature, assuming that salinity is 35‰. We assumed that surface temperatures in low and high latitude ocean are TS + 7 K and TS−13 K, respectively, where TS is the Earth's surface temperature calculated by the AEHC model.

2.2.3. Atmosphere carbon balance The carbon balance equation in the atmosphere was established by considering the anthropogenic industrial carbon emissions (e.g. fossil fuel combustion and cement production), and it becomes

  • image(17)

where Catm(t), FOS(t), ASL(t) and ASH(t) are carbon mass in atmosphere, industrial carbon emissions, air to ocean surface flux in low latitude areas and air to ocean surface flux in high latitude areas. GPP(t), RES(t) and DEC(t) are determined from the terrestrial carbon cycle model, and ASL(t) and ASH(t) are from the ocean carbon cycle model.

3. Historical anthropogenic forcing and model calibration

  1. Top of page
  2. 1. Introduction
  3. 2. Model
  4. 3. Historical anthropogenic forcing and model calibration
  5. 4. Future carbon cycle and climate simulation
  6. 5. Conclusion
  7. 6. Acknowledgement
  8. references

3.1. Anthropogenic effects

We assumed that the global carbon cycle was at steady state in the pre-industrial era, before 1750. The model was forced with anthropogenic CO2 emissions due to industrial activities such as fossil fuel combustion and cement production (Marland et al., 2000), land use change (Houghton and Hackler, 2001) and vegetation area variation. Emissions due to land use change between 1750 and 1850 were assumed to increase linearly from zero in 1750, since Houghton and Hackler did not include data before 1850. Historical anthropogenic CO2 emissions are shown in Fig. 2(a).

image

Figure 2. Historical changes in (a) anthropogenic CO2 emission from 1750 to 1990 owing to industrial activities such as fossil fuel combustion (thick line) (Marland et al., 1999) and land use change (thin line) (Houghton and Hackler, 2001), and (b) vegetation area relative to pre-industrial era (1750).

Download figure to PowerPoint

Time variation in vegetation area was estimated based on the world population variation (United Nations, 1999) by assuming that a decrease in vegetation area is linearly related to the increase in world population. Based on the world total forest area of 8.7772 × 1013 m2 in 1750 and 6.2482 × 1013 m2 in 1990 (Houghton and Hackler, 2001) and a world population of 0.79 billion in 1750 and 5.27 billion in 1990 (United Nations, 1999), vegetation area variation relative to 1750 [Sr(t)] becomes

  • image(18)

where Pop(t) is world population at year t. The coefficient 0.0536 was obtained from total forest area and world population in 1750 and 1990. In eq. (18), we introduced a new parameter of Vdec (actual vegetation decrease parameter), because Houghton and Hackler's estimation did not consider all vegetation areas, only forested areas. Many forested areas were not only converted to non-vegetation areas, but also to vegetation areas such as agricultural land and pastures. We used Vdec= 0.4 based on the assumption of an actual vegetation area decrease of 10% in the reference simulation. Consequently, time variation in vegetation area is shown in Fig. 2(b).

Lastly, we assumed that the CO2 fertilization factor is the most unknown parameter, and it was adjusted to minimize the root mean square error between observed and simulated atmospheric CO2 from 1750 to 1990 by fixing other parameters.

3.2. Model calibration and validation

The CO2 fertilization factor (β) was determined through iterative analysis to reproduce past atmospheric CO2 concentrations (1750–1964, Etheridge et al., 1998; and 1965–1990, Keeling and Whorf, 1999) and was used to calibrate model parameters, because it was impossible to determine β from existing data. The obtained β value is 0.16, which is smaller than the experimental value of 0.35 (Harrison et al., 1993). However, we consider that this value is acceptable, because the real β value should be smaller than the experimental value because nutrient and water limitation reduces the fertilization effect.

The past trend of atmospheric CO2 was generally well reproduced by our model, as shown in Fig. 3(a). However, there is a tendency to underestimate CO2 during the period from 1860 to 1960, with the maximum deviation of −13 ppmv in 1940. The root mean square error of CO2 concentration from 1750 to 1990 was approximately 6 ppmv. The tendency of underestimation in these years was also found in the carbon cycle model of Lenton (2000). Time variation of simulated temperature showed a slight small increase trend as shown in Fig. 3(b), and its general tendency is coincident with the observed one (Jones et al., 2000). Again, the simulated temperature is lower than the observed one, and the deviation is as much as 0.2 K.

image

Figure 3. Simulated and observed time variation of carbon cycle and climate. (a) Simulated (thick line) and observed (thin line) variation of atmospheric CO2 (ppmv). (b) Simulated (thick line) and observed (thin line) variation of land temperature. (c) Simulated net biospheric (thick line) and ocean carbon uptake (thin line). (d) Simulated terrestrial net primary production.

Download figure to PowerPoint

Simulated carbon budgets during the 1980s, shown in Fig. 3(c), were compared with the recent IPCC estimation based on atmospheric O2 and CO2 observation. The simulated oceanic carbon uptake of 1.7 GtC yr−1 and biospheric carbon uptake of 0.0 GtC yr−1 are near the central value of the IPCC estimation (Prentice et al., 2001). In addition, we can compare these values with the trend of terrestrial NPP [Fig. 3(d)] obtained from remotely sensed observations. The simulated NPP trend shows a 1.2% increase in the 1980s, close to the satellite-based estimations of 1.8–4.4% 10 yr−1 increase during the 1980s (Ichii et al., 2001).

Lastly, CO2 uptake projected by the terrestrial and ocean carbon cycle model in the 4-SCC model was compared with the output of the other process-based models. The 4-SCC model was forced with the projected atmospheric CO2 concentrations and temperature values derived from the IS92a emissions scenario, as it was in Prentice et al. (2001). We have tested two cases, one under the projected CO2 and constant climate, and the other under the projected CO2 and variable climate. The ocean CO2 uptake ability is 5.3 GtC yr−1 (constant climate) and 4.1 GtC yr−1 (variable climate), near the central values of the other ocean carbon cycle model studies (Orr et al., 2000). The terrestrial model response shows that the CO2 uptakes of 7.9 GtC yr−1 (constant climate) and 2.4 GtC yr−1 (variable climate) in 2100 are within the possible range provided by the model intercomparison studies (Cramer et al., 2001).

4. Future carbon cycle and climate simulation

  1. Top of page
  2. 1. Introduction
  3. 2. Model
  4. 3. Historical anthropogenic forcing and model calibration
  5. 4. Future carbon cycle and climate simulation
  6. 5. Conclusion
  7. 6. Acknowledgement
  8. references

4.1. Future projection in the reference simulation

The IPCC IS92-a “business as usual” scenario (Leggett et al., 1992) was used as an anthropogenic forcing for the years 1990–2100. Beyond 2100, industrial CO2 emissions were assumed to decline linearly and reach 0 in 2200, following Houghton et al. (1996) and Lenton (2000), as shown in Fig. 4.

image

Figure 4. Future carbon emission scenario used in the simulation. IPCC IS92-a “business as usual” scenarios were applied for 1990–2100 with the linear decline to zero at 2200.

Download figure to PowerPoint

Figure 5 and Table 3 show the projected carbon cycle and climate based on the reference simulation. The projected atmospheric CO2 becomes 645 ppmv in 2100 and reaches a peak of 760 ppmv in approximately 2170. Then, it declines and stabilizes at 600 ppmv. The temperature becomes 1.5 K higher in 2100 than at the pre-industrial level, and the peak level becomes 1.8 K higher than the pre-industrial level. The net biospheric carbon uptake (NPP − soil decomposition − land use change) becomes positive at the end of the 20th century, and reaches a peak of 6.6 GtC yr−1 in 2100; it then decreases to become negative in 2200. The net oceanic uptake reaches the maximum 4.7 GtC yr−1 in 2110, and then decreases.

image

Figure 5. Simulated future carbon cycle and climate variation forced with prescribed anthropogenic emission scenario. (a) Future atmospheric CO2 variation based on carbon cycle and climate coupled cycle model (thick line) and carbon cycle and constant climate model (thin line). (b) Future variation of land temperature. (c) Future variation of net biosphere and ocean carbon uptake based on carbon cycle and climate coupled cycle model (thick line) and carbon cycle and constant climate model (thin line). (d) Future variation of terrestrial net primary production based on carbon cycle and climate coupled cycle model (thick line) and carbon cycle and constant climate model (thin line).

Download figure to PowerPoint

Table 3. Test cases and results of sensitivity studies
Test caseΔT (K)VdecFgpp (GtC yr−1)Q10βLu (GtC yr−1)Ocean uptake (GtC yr−1)CO2 in 2100 (ppmv)Max CO2 (year) (ppmv)Bio uptake (GtC yr−1)NPP trend (% 10 yr−1)
Reference1.30.4 3502.00.162.01.7645760 (2170)0.01.2
Const. climate0.00.4 3501.00.142.01.8630730 (2160)0.00.7
AEHC T1.50.4 3502.00.172.01.7639756 (2170)0.11.3
 response2.50.4 3502.00.192.01.6647778 (2170)0.20.7
 4.50.4 3502.00.242.01.3654813 (2180)0.42.7
Vegetation1.30.0 3502.00.082.01.7661803 (2180)0.01.1
 area1.30.8 3502.00.262.01.7659781 (2170)0.01.0
Light1.30.4 1102.00.582.01.7705879 (2180)0.01.1
 saturation1.30.410502.00.102.01.7556588 (2130)0.11.3
Q10 factor1.30.4 3501.00.142.01.7638745 (2170)0.10.8
 1.30.4 3503.00.182.01.7640761 (2170)0.11.5
Land use1.30.4 3502.00.110.61.7720874 (2170)−0.20.6
 change1.30.4 3502.00.182.51.7621728 (2170)0.11.4
Ocean C1.30.4 3502.00.182.01.3644769 (2170)−0.81.5
 uptake1.30.4 3502.00.132.02.5627721 (2160)0.50.6

The climate feedback effect was evaluated in the reference simulation. Figure 5 and Table 3 show a comparison of the two simulation results, with and without the climate feedback effect. The climate feedback effect increased the atmospheric CO2 throughout the simulation, and decreased the ocean and biosphere CO2 uptake. The general mechanism of the climate feedback effect on the carbon cycle can be characterized by decreases in (1) net terrestrial uptake due to an increase in soil decomposition, and (2) net ocean uptake due to ocean surface warming. On the other hand, the terrestrial NPP increase was due to an increase in light availability and CO2 fertilization as shown in Fig. 5(d). The small impact of climate feedback on atmospheric CO2 concentrations is due to the balance of net effects described above. Atmospheric CO2 levels become 630 ppmv in 2100, and reach a peak of 730 ppmv in 2160. As shown below, climate feedback has a smaller effect than the other, unknown parameters such as the light saturation effect and land use change-effected CO2 emissions.

4.2. Sensitivity analysis

We tested sensitivity of the modeled projection based on the future emission scenario of an extended IS92a scenario. In each simulation, we have tuned the CO2 fertilization factor (β) to minimize the root mean square error of observed and simulated CO2 concentrations from 1750 to 1990. The test cases analyzed the sensitivities of (1) temperature response in the AEHC model, (2) light saturation factor (Fgpp), (3) vegetation area decrease factor (Vdec), (4) temperature dependency of soil decomposition (Q10), (5) uncertainty in emissions due to land use changes (L) and (6) uncertainty in the ocean uptake ability, as shown in Table 3.

We did not test the model structure uncertainties of the biosphere and ocean. Biosphere model (subdividing soil reservoir) sensitivities were tested by Lenton (2000), who showed that there is little effect on atmospheric CO2 concentration, but a large effect on the soil reservoir. Ocean model sensitivity, for example the ocean biospheric effect and changes in ocean circulation, must be tested. Although simplified, these models have not been well established yet, and these effects must be included in future studies.

4.2.1. Temperature responses in the AEHC model When CO2 concentrations were doubled from 350 ppmv, the temperature response in our AEHC model was an 1.3 K increase, which is smaller than the temperature change of 1.5–4.5 K at the doubled CO2 concentration predicted by the general circulation model (GCM) studies (Houghton et al., 1996). We assume that the upper and lower limits of temperature response to doubled CO2 are 1.5 and 4.5 K. We have tested three cases of temperature responses to doubled CO2, i.e. 1.5, 2.5 and 4.5 K, by amplifying AEHC model responses. The simulated atmospheric CO2 varies from 639 ppmv (the 1.5 K case) to 654 ppmv (the 4.5 K case) in 2100 and from 756 ppmv (the 1.5 K case) to 813 ppmv (the 4.5 K case) in the year of peak CO2[Fig. 6(a) and Table 3]. The sensitivity of temperature responses in the atmospheric model is weak compared with that of other parameters shown in Table 3.

image

Figure 6. Results of sensitivity analysis. (a) Climate model sensitivity to doubled CO2 (350 to 750 ppmv), (b) vegetation growth limitation factor (Fgpp), (c) actual vegetation area variation (Vdec), (d) Q10 factor, (e) land use emission (L) and (f) ocean uptake capability in the 1980s.

Download figure to PowerPoint

4.2.2. Light saturation factor (Fgpp) Although we used the prescribed value in the estimation of Fgpp, the value of Fgpp has a large uncertainty because of the difficulty in estimation of globally averaged light utilization efficiencies, and total vegetation areas in the pre-industrial era. We tested two extreme cases. One is Fgpp= 110 GtC yr−1, which is a minimum value of Fgpp and makes the GPP equation equal to that of past simple carbon cycle models (e.g. Craig and Holmén, 1995; Wigley, 1993) as discussed later. The other is with a Fgpp value that is three times larger (Fgpp= 1050 GtC yr−1), because we cannot determine an optimal maximum Fgpp value. Although it is desirable to test the sensitivity by using a larger Fgpp to evaluate the model response of the Bacastow and Keeling (1973) type (section 4.3.2), Fgpp= 1050 GtC yr−1 should be sufficient to test the sensitivity as described below.

The simulated atmospheric CO2 varies from 556 ppmv (Fgpp= 1050 GtC yr−1) to 705 ppmv (Fgpp= 110 GtC yr−1) in 2100, and from 588 ppmv (Fgpp= 1050 GtC yr−1) to 879 ppmv (Fgpp= 110 GtC yr−1) in the year of peak CO2[Fig. 6(b) and Table 3]. These results show that the sensitivity of this parameter is the highest of all parameters. The uncertainties derived from Fgpp are almost same as the ISAM model (Prentice et al., 2001), and the parameter must be constrained primarily.

4.2.3. Vegetation area decrease factor (Vdec) We used the forest area at pre-industrial and present times (Houghton and Hackler, 2001) to estimate the vegetation area variations. However, the conversion of forested area to cultivated land does not always decrease the vegetation area. Therefore, the parameter, Vdec, was introduced to convert the forested area variation to vegetation area variation. Since a suitable data set of vegetation area variation is not available, we have set the uncertainty of Vdec as 0.0–0.8. The value of 0.0 indicates that there has not been vegetation area variation since the pre-industrial era, and that of 0.8 means that 80% of the deforested area was converted to a non-vegetation area.

The simulated atmospheric CO2 varies from 659 ppmv (Vdec= 0.8) to 661 ppmv (Vdec= 0.0) in 2100, and from 781 ppmv (Vdec= 0.8) to 803 ppmv (Vdec= 0.4) in the year of peak CO2[Fig. 6(c) and Table 3]. These values are higher than those of the reference simulation are, but the overall sensitivity of this parameter is low. We have tested other Vdec values in the simulation, and found that a Vdec value of approximately 0.4 resulted in the lowest CO2 concentration. A lower Vdec value makes GPP increase due to the vegetation area decrease, but decrease due to a tuned fertilization factor (lower β value). The GPP is determined by a balance of positive and negative effects, and Vdec= 0.4 results in the highest GPP.

4.2.4. Soil decomposition (Q10) This parameter also has large uncertainty, therefore we tested the value of Q10= 1.0 and Q10= 3.0. The simulated atmospheric CO2 varies from 638 ppmv (Q10= 1.0) to 640 ppmv (Q10= 3.0) in 2100 and from 745 ppmv (Q10= 1.0) to 761 ppmv (Q10= 3.0 case) in the year of peak CO2[Fig. 6(d) and Table 3]. Sensitivity of this parameter is also low.

4.2.5. Land use change emissions (L) Following the IPCC's estimation of land use change of 0.6–2.5 GtC yr−1 in the 1980s (Prentice et al., 2001), we tested the sensitivity of the land use change emissions of 0.6 and 2.5 GtC yr−1 in the 1980s. Historical land use change emissions were multiplied by appropriate constant factors, and tuned to have the above values in the 1980s. The simulated atmospheric CO2 varies from 621 ppmv (L= 2.5 GtC yr−1) to 720 ppmv (L= 0.6 GtC yr−1) in 2100 and from 728 ppmv (L= 2.5 GtC yr−1) to 824 ppmv (L= 0.6 GtC yr−1) in the year of peak CO2[Fig. 6(e) and Table 3]. The sensitivity of this parameter is high.

4.2.6. Ocean uptake ability Following the IPCC's ocean uptake estimation of 1.9 ± 0.6 GtC yr−1 during the 1980s (Prentice et al., 2001), we tested an ocean uptake sensitivity of 1.3 (minimum ocean uptake) and 2.5 GtC yr−1 (maximum ocean uptake). The simulated atmospheric CO2 varies from 627 ppmv (2.5 GtC yr−1) to 644 ppmv (1.3 GtC yr−1) in 2100 and from 721 ppmv (2.5 GtC yr−1) to 769 ppmv (1.3 GtC yr−1) in the year of peak CO2[Fig. 6(f) and Table 3]. The sensitivity of this parameter is low.

4.3. Discussion

The light saturation factor (Fgpp) and land use emissions produced large uncertainties in projections of future carbon cycle and climate. Therefore, the following section focuses on the uncertainty in these models and model parameters, and on an approach for decreasing the uncertainties in the model projections.

4.3.1. Model constraints determined by a satellite-based NPP trend The sensitive parameters must be determined more precisely for the projection of carbon cycle and climate. One of the ways to constrain the range of the uncertainty, especially in the land use emission, is to use a satellite-based NPP trend. If land use emissions in the 1980s are assumed to have been 0.6 GtC yr−1, the NPP increase can be estimated as 0.6% 10 yr−1 during the 1980s (Table 3). Ichii et al. (2001) estimated the NPP increase to be 1.8-4.4% 10 yr−1 during the 1980s based on satellite data. However, the satellite-based NPP trend seems to have a large uncertainty due to its short time coverage, large calibration uncertainty and orbital drift effects (e.g. Malmström et al., 1997). We assume that the modeled NPP trend should be more than one-third of the 3% 10 yr−1 increase in global NPP, which is approximately the central value estimated by Ichii et al. (2001). A 1% 10 yr−1 increase in global NPP corresponds with the land use change emissions of 1.2 GtC yr−1 during the 1980s. Simulated atmospheric CO2 concentrations in 2100 are 672 ppmv. Based on the satellite-derived NPP trend, we could show the lower uncertainty in land use CO2 emission values and resulting future atmospheric CO2 concentrations with higher confidence. The uncertainty in the light saturation parameter, unfortunately, could not be constrained by the satellite-based NPP trend.

4.3.2. Comparison with other carbon cycle model studies The previous similar studies of carbon cycle models (Enting et al., 1994; Schimel et al., 1995; Houghton et al., 1996; Prentice et al., 2001) showed higher estimations of future atmospheric CO2 concentrations. The model intercomparison study showed that year 2100 projected CO2 concentrations (without climate feedback) range from 668 to 734 ppmv (Schimel et al., 1996). With land use emissions ranging from 0.4 to 1.8 GtC yr−1, Wigley (1997) estimated the uncertain range of 667 to 766 ppmv in 2100. The third assessment report of IPCC showed that the reference simulations of future carbon cycles reach 723 ppmv (ISAM model), based on the carbon and climate coupled model, and 682 ppmv (ISAM model), based on a carbon cycle model without climate feedback (Prentice et al., 2001). There are few simulation results beyond 2100. Lenton (2000) and Houghton et al. (1996) estimated peak CO2 concentrations of 985 and 850 ppmv, respectively, based on the same emission scenario used in this study.

The simulated future CO2 concentrations in 2100 and beyond 2100, presented in the current study, did not reach such high CO2 levels as shown in Table 3. Although the highest simulation results (720 ppmv for land use change = 0.6 GtC yr−1, and 705 ppmv for Fgpp = 110 GtC yr−1) are almost coincident with the referenced simulation in the ISAM models, the present study's reference simulation of CO2 concentration is not within the range of the other studies. Moreover, the reference simulation in the current study does not fall within the range 668–734 ppmv, based on the constant climate simulation (Schimel et al., 1996), although climate feedback works as a positive feedback process.

The major difference between our model and the previous models exists in the GPP models as shown in eq. (7). We can simplify our GPP models and categorize them into the following three types:

Type 1: Constant S(t), and constant Cveg(t)/S(t) and/or large k to approximate {1 − exp[−kCveg(t)/S(t)]}≃ 1.0, which means constant vegetation area and already saturated vegetation or constant vegetation density:

  • image(19)

Type 2: kextklai[Cveg(t)/S(t)]≃ 0 and the first-order terms in the Taylor expansion of {1 − exp[−kextklaiCveg(t)/S(t)]}, which means no vegetation growth limitation due to light availability:

  • image(20)

Type 3: Constant Cveg(t)/S(t) and/or large k to approximate {1−exp(−kCveg(t)/S(t)]}≃ 1.0, which means variable vegetation area and already saturated vegetation production in terms of light availability:

  • image(21)

Equation (19) is the most commonly used form (e.g. Wigley 1993; Craig and Holmén, 1995). Equation (20) was used in the pioneer carbon cycle models (e.g. Bacastow and Keeling, 1973). Equation (21) is similar to Lenton (2000), although they use different CO2 fertilization formulas and vegetation area variations.

Although the vegetation production model in ISAM incorporates vegetation growth limitation based on the logistic function (Jain et al., 1995, Kheshgi et al., 1996), the results are coincident with those of other carbon cycle models, and our highest projection. Therefore, ISAM model closely resembles our model when Fgpp= 110 GtC yr−1 (vegetation is already saturated in terms of light availability).

Our analysis showed that the choice of the GPP model (choice of Fgpp value) is one of the simplified carbon cycle model's most sensitive factors. Therefore, we have to quantify the light saturation effect of vegetation. However, we conclude that our model is more accurate, because the model includes the light saturation effect on vegetation growth, one of the most important model components for the projection of the future global carbon cycle and climate.

5. Conclusion

  1. Top of page
  2. 1. Introduction
  3. 2. Model
  4. 3. Historical anthropogenic forcing and model calibration
  5. 4. Future carbon cycle and climate simulation
  6. 5. Conclusion
  7. 6. Acknowledgement
  8. references

A simple climate–carbon cycle interactive model, the Four-Spheres Cycle of Energy and Mass (4-SCEM) model, was developed to project future variations in the global carbon cycle and climate. The following processes were included and compared with those of other simple carbon cycle model studies: (1) the vegetation light saturation effect and (2) vegetation area variation. The model was constrained with past variations in atmospheric CO2, and could simulate past atmospheric CO2 and biosphere and ocean carbon uptake during the 1980s. Moreover, the biosphere and ocean carbon cycle model showed similar responses through the year 2100 using a process-based biosphere and an ocean carbon cycle model. The satellite-based NPP trend in the 1980s was used to constrain the CO2 emissions due to land use change, and showed the implications of decreasing the uncertainty of its estimation.

The reference simulation showed that the CO2 concentration becomes 645 ppmv in 2100 and reaches a peak of 760 ppmv in 2170. These simulated concentrations were much lower than those of previous carbon cycle simulations. Our model included the past terrestrial production model (Bacastow and Keeling, 1973; Wigley, 1993; Jain et al., 1995; Kheshgi et al., 1996) as a special case, and the difference between CO2 concentrations was primarily derived from the use of different GPP models. Since our estimated vegetation light saturation parameter has a large uncertainty, we have to quantify and constrain the vegetation light saturation effect more accurately.

Sensitivity analysis successfully quantified the relative importance of estimating the poorly constrained parameters and processes. The climate–carbon cycle feedback effects, which had very large positive feedback effects in Cox et al. (2000), showed a relatively low sensitivity in our model. On the other hand, uncertainties in CO2 emissions due to land use change and the light saturation effect resulted in high sensitivities.

As found in the current study, the future projection of carbon cycle and climate still has a large uncertainty derived from model structure and parameters. Further constraints of the aforementioned parameters and processes are indispensable to improve the confidence with which the future carbon cycle and climate are projected.

6. Acknowledgement

  1. Top of page
  2. 1. Introduction
  3. 2. Model
  4. 3. Historical anthropogenic forcing and model calibration
  5. 4. Future carbon cycle and climate simulation
  6. 5. Conclusion
  7. 6. Acknowledgement
  8. references

This work is financially supported by Research Fellowships of the Japan Society for the Promotion of Science for Young Scientists. We thank Dr. Masahiko Fujii for helpful discussion on ocean carbon cycle models.

references

  1. Top of page
  2. 1. Introduction
  3. 2. Model
  4. 3. Historical anthropogenic forcing and model calibration
  5. 4. Future carbon cycle and climate simulation
  6. 5. Conclusion
  7. 6. Acknowledgement
  8. references
  • Bacastow, R. B. and Keeling, C. D. 1973. Atmospheric carbon dioxide and radiocarbon in the natural carbon cycle: II Changes from A.D. 1700 to 2070 as deduced from a geochemical model. In: Carbon and the biosphere (eds. G. M. Woodwell and E. V. Pecan). Rep. CONF-720510. US Atomic Energy Commission , Washington, DC , 86135.
  • Cao, M. and Woodward, F. I. 1998. Dynamic responses of terrestrial ecosystem carbon cycling to global climate change. Nature 393, 249252.
  • Charlson, R. J. 2000. The coupling of biogeochemical cycles and climate: forcings, feedbacks, and responses. In: Earth system sciences (eds. M. C. Jacobson, R. J. Charlson, H. Rodhe and G. H. Orians). Academic Press, London , 439458.
  • Cox, P. M., Betts, R. A., Jones, C. D., Spall, S. A. and Totterdell, I. J. 2000. Acceleration of global warming due to carbon-cycle feedbacks in a coupled climate model. Nature 408, 184187.
  • Craig, S. G. and Holmén, K. J. 1995. Uncertainties in future CO2 projections. Global Biogeochem. Cycles 9, 139152.
  • Cramer, W., Bondeau, A., Woodward, F. I., Prentice, I. C., Betts, R. A., Brovkin, V., Cox, P. M., Fisher, V., Foley, J. A., Friend, A. D., Kucharik, C., Lomas, M. R., Ramankutty, N., Sitch, S., Smith, B., White, A. and Yong-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, 357373.
  • Cramer, W., Kicklighter D. W., Bondeau, A., Moore III, B., Churkina, G., Nemry, B., Ruimy, A., Schloss, A. L. and the participants of the Potsdam NPP model intercomparison. 1999. Comparing global models of terrestrial net primary productivity (NPP): Overview and key results. Global Change Biol. 5, 115.
  • Enting, I. G., Wigley, T. M. L. and Heimann, M. 1994. Future emissions and concentrations of carbon dioxide: Key ocean/atmosphere/land analysis. CSIRO Division of Atmospheric Res. Tech. Pap. 31 , Victoria , Australia .
  • Etheridge, D. M., Steele, L. P., Langenfelds, R. L., Francey, R. J., Barnola, J. M. and Morgan, V. I., 1998. Historical CO2 records from the Law Dome DE08, DE08–02, and DSS ice cores. In: Trends: A compendium of data on global change, Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory , Oak Ridge , Tennessee , USA .
  • Fujii, M., Ikeda, M. and Yamanaka, Y. 2000. Roles of physical processes in the carbon cycle using a simplified physical model. J. Oceanogr. 56, 655666.
  • Goyet, C. and Poisson, A. 1989. New determination of carbonic acid dissociation constants in sea water as a function of temperature and salinity. Deep Sea Res. Part A 36, 16351654.
  • Harrison, K., Broecker, W. and Bonani, G. 1993. A strategy for estimating the impact of CO2 fertilization on soil carbon storage. Global Biogeochem. Cycles 7, 6980.
  • Hayashi, M. 1991. One dimensional radiative model. In: Earth environmental simulation (ed. O. Yokoyama), Hakua-Shobo (in Japanese), 85109.
  • Heimann, M. and Keeling, C. D. 1989. A three-dimensional model of atmospheric CO2 transport based on observed winds. 2. Model description and simulated tracer experiments. In: Aspects of climate variability in the Pacific and Western Americas. Geophysical Monograph Ser. 55, 237275.
  • Houghton, J. T. 1977. The physics of atmospheres. Cambridge University Press, Cambridge .
  • Houghton, J. T., Filho, L. G. M., Callander, B. A., Harris, N., Kattenberg, A. and Maskell, K. 1996. Climate Change 1995: the science of climate change, Cambridge University Press, Cambridge .
  • Houghton, R. A. and Hackler, J. A. 2001. Carbon flux to the atmosphere from land-use changes: 1850 to 1990. ORNL/CDIAC-131, NDP-050/R1. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory , Oak Ridge , Tennessee , USA .
  • Huntingford, C. and Cox, P. M. 2000. An analogue model to derive additional climate change scenarios from existing GCM simulations. Climate Dynam. 16, 575586.
  • Ichii, K., Matsui, Y., Yamaguchi, Y. and Ogawa, K. 2001. Comparison of global net primary production trend obtained from satellite based normalized difference vegetation index and a carbon cycle model. Global Biogeochem. Cycles 15, 351364.
  • Jain, A. K., Kheshgi, H. S., Hoffert, M. I. and Wuebbles, D. J. 1995. Distribution of radiocarbon as a test of global carbon cycle models. Global Biogeochem. Cycles 9, 153166.
  • Jones, P. D., Parker, D. E., Osborn, T. J. and Briffa, K. R. 2000. Global and hemispheric temperature anomalies-land and marine instrumental records. In: Trends: A compendium of data on global change. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, US Department of Energy , Oak Ridge , Tennessee , USA .
  • Keeling, C. D. and Whorf, T. P. 1999. 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 , Oak Ridge , Tennessee , USA .
  • Kheshgi, H. S., Jain, A. K. and Wuebbles, D. J. 1996. Accounting for the missing carbon-sink with the CO2 fertilization effect. Climatic Change 33, 3162.
  • Kiehl, J. T. and Trenberth, K. E. 1997. Earth's annual global mean energy budget. Bull. Am. Meteorol. Soc. 78, 197208.
  • Kumar, M. and Monteith, J. L. 1981. Remote sensing of crop growth. In: Plants and the daylight spectrum (ed. H. Smith). Academic, San Diego , California , 133144.
  • Kwon, O. Y. and Schnoor, J. L. 1994. Simple global carbon model: The atmosphere-terrestrial biosphere-ocean interaction. Global Biogeochem. Cycles 8, 295305.
  • Lacis, A. A. and Hansen, J. E. 1974. A parameterization for the absorption of solar radiation in the Earth's atmosphere. J. Atmos. Sci. 31, 118133.
  • Leggett, J., Pepper, W. J. and Swart, R. J. 1992. Emissions scenarios of the IPCC: an update. In: Climate Change: The supplementary report to the IPCC scientific assessment (eds. J. T. Houghton, B. A. Callander and S. K. Varney). Cambridge University Press, Cambridge .
  • Lenton, T. M. 2000. Land and ocean carbon cycle feedback effects on global warming in a simple Earth system model. Tellus 52B, 11591188.
  • Malmström, C. M., Thompson, M. V., Juday, G. P., Los, S. O., Randerson, J. T. and Field, C. B. 1997. Interannual variation in global-scale net primary production: Testing model estimates. Global Biogeochem. Cycles 11, 367392.
  • Manabe, S. and Strickler, R. F. 1964. Thermal equilibrium of the atmosphere with a convective adjustment. J. Atmos. Sci. 21, 361385.
  • Marland, G., Boden, T. A. and Andres, R. J. 2000. Global, regional, and national fossil fuel CO2 emissions. In: Trends: A compendium of data on global change. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory , Oak Ridge , Tennessee , USA .
  • McCree, K. J. 1981. Photosynthetically active radiation. In: Physiological plant ecology. I. Responses to the physical environment Vol.12A (eds. O. L. Lange et al.). Springer-Verlag, New York , 4155.
  • Nakajima, S., Hayashi, Y. Y. and Abe, Y. 1992. A study on the “runaway greenhouse effect” with a one-dimensional radiative–convective equilibrium model. J. Atmos. Sci. 49, 22562266.
  • Orr, J. C., Maier-Reimer, E., Mikolajewica, U., Monfray, P., Sarmiento, J. L., Toggweiler, J. R., Taylor, N. K., Palmer, J., Gruber, N., Sabine, C. L., Quere, C. L., Key, R. M. and Boutin, J. 2001. Estimates of anthropogenic carbon uptake from four 3-D global ocean models. Global Biogeochem. Cycles 15, 4360.
  • Potter, C. S., Randerson, J. T., Field, C. B., Matson, P. A., Vitousek, P. M., Money, H. A. and Klooster, S. A. 1993. Terrestrial ecosystem production: A process model based on global satellite and surface data. Global Biogeochem. Cycles 7, 811841.
  • Prentice, I. C., Farquhar, G. D., Fasham, M. J. R., Goulden, M. L., Heimann, M., Jaramillo, V. J., Kheshgi, H. S., Quere, C. L., Scholes, R. J. and Wallace, D. W. R. 2001. The carbon cycle and atmospheric carbon dioxide. In: Climate Change 2001: The scientific basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change (eds. J. T. Houghton, Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell and C. A. Johnson). Cambridge University Press, Cambridge , United Kingdom and New York , USA , 183237.
  • Raich, J. W. and Schlesinger, W. H. 1992. The global carbon dioxide flux in soil respiration and its relationship to vegetation and climate. Tellus 44B, 8199.
  • Rossow, W. B., Walker, A. W., Beuschel, D. E. and Roiter, M. D. 1996. International Satellite Cloud Climatology Project (ISCCP): documentation of new cloud datasets. WMO/TD-No. 737, World Meteorological Organization , p. 115.
  • Sarmiento, J. L., Hughes, T. M. C., Stouffer, R. J. and Manabe, S. 1998. Simulated response of the ocean carbon cycle to anthropogenic climate warming. Nature 393, 245249.
  • Schimel, D., Enting, I. G., Heimann, M., Wigley, T. M. L. and Raynaud, R. 1995. CO2 and carbon cycle. In: Climate change 1994 (eds. J. T. Houghton, L. G. Meira Filho, J. Bruce, H. Lee, B. A. Callander, E. Haites, E. Harris and K. Maskell). Cambridge University Press, New York , 3571.
  • Schimel, D., Alves, D., Enting, I. G., Heimann, M., Joos, F., Raynaud, D. and Wigley, T. M. L. 1996. CO2 and the carbon cycle. In: Climate change 1995: The science of climate change: contribution of WGI to the second assessment report of the IPCC (eds. J. T. Houghton, J. G. Meira Filho, B. A. Callander, N. Harris, A. Kattenberg and K. Maskell). Cambridge University Press, Cambridge , 6586.
  • Shaffer, G. and Sarmiento, J. L. 1995. Biogeochemical cycling in the global ocean. 1. A new, analytical model with continuous vertical resolution and high-latitude dynamics. J. Geophys. Res. 100, 26592672.
  • Siegenthaler, U. and Joos, F. 1992. Use of a simple model for studying oceanic tracer distributions and the global carbon cycle. Tellus 44B, 186207.
  • Siegenthaler, U. and Sarmiento, J. L. 1993. Atmospheric carbon dioxide and ocean. Nature 365, 119125.
  • Sitch, S. 2000. The role of vegetation dynamics in the control of atmospheric CO2 content. Ph.D. Thesis, University of Lund , Sweden , 213 pp.
  • Somerville, R. C. and Remer, L. A. 1984. Cloud optical thickness feedbacks in the CO2 climate problem. J. Geophys. Res. 89, 96689672.
  • Stephens, G. L. 1978. Radiative profiles in extended water clouds. 2: Parameterization scheme, J. Atmos. Sci. 35, 21232132.
  • United Nations 1999. The world at six billion. Population Division, Department of Economic and Social Affairs, United Nations Secretariat .
  • Weiss, R. F. 1974. Carbon dioxide in water and seawater: The solubility of a non-ideal gas. Marine Chem. 2, 203215.
  • Wigley, T. M. L. 1993. Balancing the carbon budget: Implications for projections of future carbon dioxide concentration changes. Tellus 45B, 409425.
  • Wigley, T. M. L., Jain, A. K., Joos, F., Nyenzi, B. S. and Shukla, P. R. 1997. Implications of proposed CO2 emissions limitations. IPCC Technical Paper 4. (eds. J. T. Houghton, L. G. Meira Filho, D. J. Griggs, and M. Noguer). Intergovernmental Panel on Climate Change , Bracknell , UK .