Evaluation of the terrestrial carbon cycle, future plant geography and

Global Change Biology (2008) 14, 1–25, doi: 10.1111/j.1365-2486.2008.01626.x
Evaluation of the terrestrial carbon cycle, future plant
geography and climate-carbon cycle feedbacks using five
Dynamic Global Vegetation Models (DGVMs)
S . S I T C H *, C . H U N T I N G F O R D w , N . G E D N E Y *, P. E . L E V Y z, M . L O M A S § , S . L . P I A O } ,
R . B E T T S k, P. C I A I S } , P. C O X **, P. F R I E D L I N G S T E I N } , C . D . J O N E S k, I . C . P R E N T I C E w w
and F . I . W O O D W A R D §
*Met Office Hadley Centre, JCHMR, Maclean Building, Wallingford OX10 8BB, UK, wCentre for Ecology and Hydrology
Wallingford, Maclean Building, Wallingford OX10 8BB, UK, zCentre for Ecology and Hydrology Bush Estate, Penicuik, Midlothian
EH26 0QB, UK, §Department of Animal & Plant Sciences, University of Sheffield, Sheffield S10 2TN, UK, }IPSL/LSCE, Unite
mixte 1572 CEA-CNRS, CE-Saclay, Bat 701, 91191 Gif sur Yvette, France, kMet Office Hadley Centre, Fitzroy Road, Exeter EX1
3PB, UK, **School of Engineering, Computer Science and Mathematics, University of Exeter, Exeter ES4 4QF, UK, wwQUEST,
Department of Earth Sciences, University of Bristol, Wills Memorial Building, Queens Road, Bristol BS8 1RJ, UK
Abstract
This study tests the ability of five Dynamic Global Vegetation Models (DGVMs), forced
with observed climatology and atmospheric CO2, to model the contemporary global
carbon cycle. The DGVMs are also coupled to a fast ‘climate analogue model’, based on
the Hadley Centre General Circulation Model (GCM), and run into the future for four
Special Report Emission Scenarios (SRES): A1FI, A2, B1, B2. Results show that all
DGVMs are consistent with the contemporary global land carbon budget. Under the
more extreme projections of future environmental change, the responses of the DGVMs
diverge markedly. In particular, large uncertainties are associated with the response of
tropical vegetation to drought and boreal ecosystems to elevated temperatures and
changing soil moisture status. The DGVMs show more divergence in their response to
regional changes in climate than to increases in atmospheric CO2 content. All models
simulate a release of land carbon in response to climate, when physiological effects of
elevated atmospheric CO2 on plant production are not considered, implying a positive
terrestrial climate-carbon cycle feedback. All DGVMs simulate a reduction in global net
primary production (NPP) and a decrease in soil residence time in the tropics and extratropics in response to future climate. When both counteracting effects of climate and
atmospheric CO2 on ecosystem function are considered, all the DGVMs simulate
cumulative net land carbon uptake over the 21st century for the four SRES emission
scenarios. However, for the most extreme A1FI emissions scenario, three out of five
DGVMs simulate an annual net source of CO2 from the land to the atmosphere in the
final decades of the 21st century. For this scenario, cumulative land uptake differs by
494 Pg C among DGVMs over the 21st century. This uncertainty is equivalent to over 50
years of anthropogenic emissions at current levels.
Keywords: carbon cycle feedbacks, biogeography, DGVM
Received 16 March 2007; revised version received 20 December 2007 and accepted 17 January 2008
Introduction
In recent years, much attention has been placed on the
role of terrestrial biosphere dynamics in the climate
Correspondence: S. Sitch, e-mail: [email protected]
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd
system (Cramer et al., 2001), and the possibility of
anthropogenic climate change inducing major alterations in terrestrial ecosystems. Terrestrial ecosystems
may become a source of CO2 under imposed climate
change and, thus, act to accelerate the build-up of
atmospheric CO2 concentrations [see Cox et al. (2000)].
1
2 S . S I T C H et al.
This has major policy implications for climate change
mitigation and reduces ‘permissible’ emissions to
achieve stabilization (Jones et al., 2006).
Cox et al. (2000) ran the TRIFFID Dynamic Global
Vegetation Model (DGVM) coupled with the low ocean
resolution Hadley Centre General Circulation Model
(GCM), HadCM3LC, in a fully interactive carbon cycle
experiment for one future emission scenario (IS92a).
They found a very large climate-carbon feedback,
caused in particular by enhanced midlatitude soil decomposition in response to future warming, but also
from ‘dieback’ of Amazon forests (Betts et al., 2004; Cox
et al., 2004) in response to both future warming and
drying. Dufresne et al. (2002) used the IPSL GCM and
the simple land carbon cycle model SLAVE to perform a
similar analysis and found a much smaller climatecarbon feedback.
The C4MIP model intercomparison (Friedlingstein
et al., 2006) has extended this work by quantifying the
uncertainty in future climate-carbon cycle feedbacks
among a large set of 11 climate-carbon cycle models
for the Special Report Emission Scenarios (SRES) A2
emissions scenario. All models predict a reduction in
the combined efficiency of the ocean and land carbon
cycles to absorb anthropogenic carbon emissions due to
future climate change. By 2100, this translates to an
extra 20–200 ppmv of anthropogenic CO2 remaining in
the atmosphere, as compared with CO2 scenarios based
on the assumption that the current fraction of emitted
CO2 drawn down naturally into the oceans and land
biosphere remains constant into the future (Friedlingstein et al., 2006). The result corresponds to an additional climate warming of 0.1–1.5 1C (Friedlingstein
et al., 2006). The majority of models attribute these
changes predominantly to the land carbon cycle, and
in particular to reductions in land carbon uptake in
the tropics. However, there was no consensus among
models on the relative roles of changes in net primary
productivity (NPP) and heterotrophic respiration (RH)
(Friedlingstein et al., 2006).
In this study, we make a more controlled comparison
between the responses of five different DGVMs, by
exposing each of them to the same set of climate change
scenarios. It is unfeasible to couple and run multiple
DGVMs within a single GCM, and so we coupled the
five DGVMs to a computationally efficient ‘GCM analogue model’ (AM) and a simple ocean carbon cycle
model, both calibrated against the climate change
simulated by HadCM3LC (Huntingford & Cox, 2000;
Huntingford et al., 2004). Initially, the DGVMs are run
over the historical period 1901–2002 forced with
observed monthly climatology and atmospheric CO2
content (hereafter referred to as ‘Offline simulations’).
Then, using the AM system, a second set of simulations
is conducted over the period, 1860–2100 using four
SRES emission scenarios and a common set of patterns
of climate change from HadCM3LC GCM (hereafter
referred to as ‘Coupled simulations’). This study accounts for biogeochemical feedbacks. Biogeophysical
feedbacks associated with individual DGVMs, although
important, are beyond the scope of the present study.
We address the following questions. Are DGVMs able
to simulate the contemporary global land carbon cycle?
To what extent do the DGVMs agree on their global and
regional responses to future changes in climate and
atmospheric composition? How uncertain is the climate-carbon cycle feedback? Can specific ecological
processes be identified as the source for the overall
uncertainties in DGVM response? What are the relative
uncertainties in future atmospheric CO2 associated with
different choices of DGVM and anthropogenic emission
scenario?
Methods
The IMOGEN climate-carbon cycle system
The AM consists of a global thermal two-box model
which calculates both global mean temperature rise
over land and surface oceans, in response to increases
in atmospheric radiative forcing associated with changing atmospheric greenhouse gas concentrations. The
land value then multiplies a set of patterns across each
land grid-box and each month, for the key variables
determining ‘weather’ and associated land surface response (e.g. temperature, humidity, windspeed, shortwave and longwave radiative fluxes). The AM
capitalizes on the analysis of Hadley GCM output
that, to a good approximation, reveals that many aspects of surface climatology vary linearly to changes in
global mean temperature response over land (Huntingford & Cox, 2000) – it is this observation that allows
the possibility to extrapolate existing Hadley GCM
simulations to a range of different pathways in atmospheric greenhouse gas concentrations. For this reason,
the spatial patterns capturing such linearity can be
defined from a small number of HadCM3 simulations
(Huntingford & Cox, 2000); the AM defines ‘patterns of
change per degree of global warming over land’ for
temperature at 1.5 m (K K1), relative humidity at
1.5 m (%K1), windspeed at 10.0 m (m s1 K1), downward longwave radiation (W m2 K1), downward
shortwave radiation (W m2 K1), precipitation rate
(mm day1 K1), diurnal temperature range (K) and
surface pressure (hPa K1). The scaling factor for these
patterns (i.e. the change in global mean temperature
over land, calculated using the thermal two-box climate
model) is also calibrated to HadCM3 output. Hourly
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
U N C E RTA I N T Y I N L A N D C A R B O N C Y C L E F E E D B A C K S
surface climate is derived by temporal disaggregation
of the monthly means (including conversion of precipitation into either rainfall or snow fall) based on the
diurnal temperature range and observed fraction of wet
days. Interannual variability of the GCM climate is not
presently included in the AM.
Predicted spatial and seasonal changes in surface
climate [for a range of different future trajectories in
atmospheric greenhouse gases (GHGs)] are important
for impacts assessment. For this reason, the AM was
extended by coupling it to the Met Office land surface
model that includes the TRIFFID DGVM for an identical
land grid structure as HadCM3. In addition, an extra
global box model describes the oceanic uptake of atmospheric CO2. The flux is linear in the gradient between
atmospheric and surface oceanic CO2 concentrations;
the latter related to both the global mean oceanic mixedlayer temperature and concentration of dissolved inorganic carbon in the surface water, itself a function of the
history of CO2 drawdown. The dependence on previous
fluxes of atmospheric-ocean carbon dioxide is based on
the model of Joos et al. (1996), with modelled dependence on oceanic temperature changes given by Takahashi et al. (1993); the equations are described in full
in the Appendix of Huntingford et al. (2004). The
resultant model structure is called IMOGEN (Integrated
Model Of Global Effects of climatic aNomalies); see
Huntingford et al. (2004).
IMOGEN is forced by a prescribed emissions scenario
of CO2. Annual atmospheric CO2 concentrations are
updated each year accounting for annual anthropogenic
CO2 emissions and changes in global land and ocean
carbon storage as calculated by TRIFFID and the ocean
box model, respectively. The concentration of non-CO2
GHGs are prescribed as a function of time for each
emission scenario.
IMOGEN provides an impacts modelling system for a
broad range of different emission trajectories, based on
changes in surface climate predicted by HadGCM3LC,
but without the need to rerun the full GCM. Here, we
use this system to investigate the influence of different
land carbon cycle descriptions on the global carbon
cycle by inserting alternative DGVMs into the IMOGEN
structure.
Land carbon cycle models (DGVMs)
There is now a range of well-established DGVMs operated by different ecosystem research groups, but with
alternative parameterizations and diverse inclusion of
processes (Prentice et al., 2007). Five DGVMs are applied
here: the HyLand (HYL) model is based on the Hybrid
DGVM (Friend et al., 1997; Friend & White, 2000) with
modifications as documented in Levy et al. (2004);
3
the Lund–Potsdam–Jena DGVM (LPJ) (Sitch et al.,
2003), with the updated hydrology of Gerten et al.
(2004); ORCHIDEE (ORC) as described in Krinner
et al. (2005); Sheffield-DGVM (SHE) (Woodward et al.,
1995; Woodward & Lomas, 2004) and TRIFFID (TRI)
(Cox, 2001). A description of the DGVMs used in this
intercomparison is given in Table 1. In this study,
we focus on two aspects of land surface modelling:
vegetation dynamics and the carbon cycle. However,
these models have also been developed to simulate
soil hydrological processes and the exchange of water
between the land and the atmosphere. In the case
of land-surface models coupled to GCMs, energy
exchange between the land surface and atmosphere is
also simulated.
Datasets
Atmospheric composition and climate. In the offline
historical simulations (i.e. forcing the DGVMs with
specified surface conditions), we use annual global
atmospheric CO2 concentrations for the period 1901–
2002 based on data from ice-core records and
atmospheric observations (Keeling & Whorf, 2005).
These simulations use monthly climatology for the
period 1901–2002 from the University of East Anglia
Climate Research Unit (CRU) gridded dataset (New et al.,
2000), based on global collection of measurements. These
measurements are aggregated to a resolution of 3.751
longitude 2.51 latitude, in keeping with output from
HadCM3LC and associated patterns of the GCM AM.
For our coupled climate-carbon cycle simulations,
IMOGEN instead requires prescribed fossil fuel and
land-use emissions. Such emissions of CO2 are based on
historical records of fossil fuel burning and land-use
emissions from Marland et al. (2003) and Houghton
(2003), respectively, for the period 1860–1999. The
Intergovernmental Panel on Climate Change (IPCC)
emission scenarios of A1FI, A2, B1, B2 (Nakicenovic
et al., 2000) are used for the period 2000–2100. Radiative
forcing from non-CO2 GHGs, as defined for the A1FI, A2,
B1, B2 scenarios (Nakicenovic et al., 2000) are added to the
forcing due to raised atmospheric carbon dioxide
concentrations within IMOGEN. At present, the effect
of historical and future sulphate aerosols on climate is
not considered.
In these coupled climate-carbon cycle simulations, the
AM climate uses the ‘monthly persistent anomaly patterns’
multiplied by the calculation of DTl, and then such
anomalies added to an initial climatology. The initial
climatology for these simulations is based on an updated
version of the Leemans & Cramer (1991) monthly means
for period 1931–1960, as modified by Friend (1998) (known
hereafter as the ‘CL dataset’).
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
f(Assimilation) Gifford
(1995)
f(Assimilation)
Penman–Monteith
transpiration (Monteith
& Unsworth, 1990)
One soil layer
Bucket model (dynamic
water holding capacity)
Canopy energy balance
(Friend, 1995)
n/a
Beer’s Law (applied to
PFTs)
Sapwood respiration
Fine root respiration
Evapotranspiration
Water balance
Canopy temperature
Radiation
n/a
n/a
Ecosystem structure
Phenology
Cold deciduous
Dry deciduous
Aerodynamics
Canopy scaling
Jarvis (1976)/Stewart
(1988)
Optimum Nleaf
distribution
Farquhar et al. (1980)
1 day
Stomatal conductance
Shortest time step
Physiology
Photosynthesis
HyLand (HYL)
GDD requirement
Temperature threshold
Soil moisture threshold
Beer’s Law (applied to
vegetation fraction)
n/a
Two soil layers
Modified bucket model
from Neilson (1993)
Surface runoff 1 drainage
Snow pack
n/a
GDD requirement
Temperature threshold
Soil moisture threshold
Beer’s Law (applied to
vegetation fractions)
Log-wind profile
Dependent on
temperature, sapwood
mass and C : N ratio
f(T,Croot)
Transpiration,
interception loss, bare
ground evaporation
and snow sublimation
are computed using
Monteith–type
formulations
(Ducoudré et al., 1993)
Two soil layers (deep
bucket layer and upper
layer of variable depth)
Surface runoff 1 drainage
Snow pack
n/a
Optimum Nleaf
distribution
Optimum Nleaf distribution
Dependent on sapwood
mass and C : N ratio
(Lloyd & Taylor, 1994)
f(T,Croot)
Total evapotranspiration
(Monteith, 1965)
Farquhar et al. (1980)/
Collatz et al. (1992)
Ball et al. (1987)
0.5 h
ORCHIDEE (ORC)
Farquhar et al. (1980)/
Collatz et al. (1992)
Haxeltine & Prentice (1996)
1 day
Lund–Potsdam–Jena (LPJ)
Table 1 Characteristics of the Dynamic Global Vegetation Models (DGVMs)
Soil moisture threshold
Temperature threshold
Beer’s Law (applied to
total vegetation)
Log-wind profile
Three soil 1 one litter
layer Modified Bucket
model
Drainage
Snow pack
n/a
f(T,Croot)
Penman–Monteith
transpiration
(Monteith,
1981) 1 interception 1
evaporation from soil
surface
Annual sapwood
increment, C : N f(T)
Optimum Nleaf
distribution
Farquhar et al. (1980)/
Collatz et al. (1992)
Leuning (1995)
1 day
Sheffield-DGVM (SHE)
Temperature sum with
threshold
n/a
Diagnosed from energy
balance
Neutral transfer
coefficients using z0
proportional to height
Beer’s Law (applied to
vegetation fractions)
Four soil layer
Darcy’s law
Optimum Nleaf
distribution. Sellers
et al. (1992)
Pipe model to diagnose
sapwood volume, then
Q10 relationship
f(T,Nroot)
Penman–Monteith
transpiration
(Monteith,
1981) 1 interception
(fixed fraction)
Collatz et al. (1991)/
Collatz et al. (1992)
Cox et al. (1998)
1/2 h
TRIFFID (TRI)
4 S . S I T C H et al.
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
Daily litter carbon balance
CENTURY (Parton et al.,
1993), modified by
Comins & McMurtrie
(1993)
Allometric relationships
n/a
Fixed C : N
Litter fall
Decomposition
N uptake
N allocation
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
Shrubs
Grasses/forbs
Trees deciduous
n/a
C3 herbaceous
Needleleaf evergreen
Plant functional types (PFTs)
Trees evergreen
Broadleaf evergreen
C allocation
n/a
Grass
n/a
C3 herbaceous
C4 herbaceous
Temperate needleleaf
evergreen
Boreal needleleaf
evergreen
Tropical raingreen
Temperate summergreen
Boreal summergreen
Tropical evergreen
Temperate broadleaf
evergreen
Implicit, dependent on
demand
n/a
Annual allometric
relationship for
individuals
f(T,ytop,tissue type)
Soil moisture and carbon
balance threshold
Annual litter carbon balance
Tropical broadleaf
raingreen
Temperate broadleaf
summergreen
Boreal broadleaf
summergreen
Boreal needleleaf
summergreen
n/a
C3 herbaceous
C4 herbaceous
Tropical broadleaf
evergreen
Temperate broadleaf
evergreen
Temperate needleleaf
evergreen
Boreal needleleaf
evergreen
n/a
n/a
Based on resource
optimization
(Friedlingstein et al.,
1998)
Based on Parton et al.
(1988)
Dependent on climate
zone. Botta et al. (2000)
Shrubs
C3 herbaceous
C4 herbaceous
Broadleaf deciduous
Needleleaf deciduous
Needleleaf evergreen
Broadleaf evergreen
Based on soil C and N
decomposition also
dependent on soil T
and moisture
Variable N with light
Daily allocation by
demand in order
of priority
LAI4roots4wood
Monthly litter carbon
balance
Similar to CENTURY
(Parton et al., 1993)
Growth threshold
Shrubs
C3 herbaceous
C4 herbaceous
Needleleaf
Broadleaf
Fixed C : N
Continued
Partitioning into
‘spreading’ and
‘growth’ based on LAI
leaf : root : wood
partitioning from
allometric relationships
n/a
f(T,y,Csoil) McGuire et al.
(1992)
Monthly litter
n/a
U N C E RTA I N T Y I N L A N D C A R B O N C Y C L E F E E D B A C K S
5
Prescribed disturbance
rate for each PFT
Minimum ‘seed’ fraction
for all PFTs
All PFTs establish
uniformly as small
individuals
Dependent on carbon
pools
Establishment
Mortality
Deterministic baseline selfthinning carbon balance
Fire
Extreme temperatures
Nonhomogeneous areabased competition for
light (1-layer), H2O
(3 layers)
Climatically favoured
PFTs establish in
proportion to area
available, as small
individuals
Carbon balance, Age
Wind throw
Fire
Extreme temperatures
Nonhomogeneous areabased competition for
light (1-layer), H2O
(2 layers)
Climatically favoured
PFTs establish in
proportion to area
available, as small
individuals
Deterministic baseline
self-thinning carbon
balance
Fire
Extreme temperatures
Nonhomogeneous areabased competition for
light (1-layer), H2O
(2 layers)
Climatically favoured PFTs
establish in proportion to
area available, as small
individuals
Competition between
PFTs for light
Lokta-Volterra in
fractional cover
Sheffield-DGVM (SHE)
ORCHIDEE (ORC)
Lund–Potsdam–Jena (LPJ)
HyLand (HYL)
Vegetation dynamics
Competition
Table 1. (Contd.)
TRIFFID (TRI)
6 S . S I T C H et al.
The AM climate patterns are derived from the
HadCM3LC coupled ocean–atmosphere GCM [see
Gordon et al. (2000) for summary details], with interactive
ocean and land carbon cycles (Cox et al., 2000).
The AM patterns of seasonal temperature and
precipitation change are shown in Figs 1 and 2. For a unit
change in future global mean temperature over land DTl,
HadCM3LC simulates large temperature increases: yearround across Amazonia; during the nongrowing season
across northern hemisphere tundra ecosystems and during
the growing season across the North American and Asian
boreal forests, temperate North America and northern
hemisphere Mediterranean ecosystems. HadCM3LC
simulates a large year-round decrease in rainfall rate
across the Amazonian rainforest and seasonal forests of
North-East Brazil (Fig. 2) in the future (Cox et al., 2004).
HadCM3LC simulates decreases in summer rainfall across
temperate, boreal and Mediterranean ecosystems in North
America and Eurasia. The rainfall rate increases across
many of these ecosystems during the rest of the year. Yearround decreases in rainfall are simulated across the waterlimited ecosystems of Australia and Southern Africa.
Rainfall decreases across the western Sahel during the
northern-hemisphere summer, and year-round increases
are simulated for the tropical rainforests of Central Africa.
Many of these changes can be expected to alter present day
terrestrial ecosystem structure and function.
Nino-3 and ocean flux data. Simulated interannual
variation (IAV) in CO2 by DGVMs is correlated
against the observed Niño-3 index, a measure of the
ENSO cycle. The Niño-3 index is the mean sea surface
temperature (SST) anomaly in the region 51N to 51S,
150–901W derived from a climate dataset of SST (Rayner
et al., 2000). To estimate the IAV in ‘natural CO2’ in
response to ENSO, a dataset of modelled monthly ocean
CO2 fluxes (Buitenhuis et al., 2006) for the period 1955–
2003 were added to the land fluxes from the DGVMs. In
the absence of available data, IAV in ocean fluxes was
set to zero between 1901 and 1954.
Experimental design
Model initialization. For initialization of the forced
contemporary carbon cycle simulation, we used the
mean monthly fields over the first decade from the
CRU dataset. The mean observed climate (the CL
dataset) was used to initialize terrestrial carbon pools
and vegetation structure at their preindustrial
equilibrium states for the coupled simulations. In both
sets of simulations, LPJ used monthly climatology
selected from a random sequence of years between
1901 and 1930 from the CRU dataset (New et al., 2000)
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
U N C E RTA I N T Y I N L A N D C A R B O N C Y C L E F E E D B A C K S
DJF 1.5 m temperature change patterns
MAM 1.5 m temperature change patterns
90°N
90°N
60°N
60°N
30°N
30°N
0°
0°
30°S
30°S
180°
90°W
0.6
0.8
0°
0.95
90°E
1.05
1.2
180°
1.4
90°N
60°N
60°N
30°N
30°N
0°
0°
30°S
30°S
0.6
0.8
0°
0.95
90°E
1.05
1.2
180°
1.4
0.8
0°
0.95
90°E
1.05
1.2
1.4
SON 1.5 m temperature change patterns
90°N
90°W
90°W
0.6
JJA 1.5 m temperature change patterns
180°
7
90°W
0.6
0.8
0°
0.95
90°E
1.05
1.2
1.4
Fig. 1 Mean seasonal patterns of 1.5 m land temperature change per unit increase in global land temperature (winter – DJF, spring –
MAM, summer – JJA, autumn – SON).
for model initialization. Interannual varying climate is
required by LPJ to simulate realistic fire dynamics. Each
DGVM is allowed to calculate its own vegetation
distribution. SHE adopted the Global Land Cover
map to describe PFT fractions (GLC 2000, Bartholome
et al., 2002) and assumed fixed vegetation throughout
the transient simulations. Three experiments were
conducted.
Offline historical carbon cycle. In the first set of
experiments, each DGVM is run from its preindustrial
equilibrium at 1901 over the historical period 1901–2002
using observed fields of monthly climatology and
annual global atmospheric CO2 concentration, at the
GCM grid resolution of 3.751 longitude 2.51 latitude.
No land or ocean carbon cycle feedbacks are included.
Coupled climate-carbon cycle. Each DGVM is run from its
preindustrial equilibrium at 1860 over the historical and
future period 1860–2100 at the GCM grid resolution.
Once in their equilibrium state, the DGVMs are then
driven within the IMOGEN framework using climate
anomalies consistent with HadCM3LC. This is
undertaken for four IPCC SRES fossil fuel and landuse emission scenarios (A1FI, A2, B1, B2) and radiative
forcing from non-CO2 GHGs. For LPJ, climate
anomalies were added to a random sequence of 30
years baseline climatology throughout the transient
simulation. Land-use emissions are treated as external,
and do not affect directly vegetation area and carbon
pools. Although land-cover change is important for
both climate and the global carbon cycle (Brovkin
et al., 1999; Betts, 2000; Gitz & Ciais, 2003; Brovkin
et al., 2004; Sitch et al., 2005), inclusion of explicit landuse and land cover changes are beyond the scope of the
current study. However, we go beyond the groundbreaking intercomparison of Cramer et al. (2001) by
including and diagnosing climate-carbon cycle
feedbacks, and by spanning a wider range of emission
scenarios.
Prescribed climate. In order to quantify future climatecarbon cycle feedbacks an additional ‘prescribed-climate’
experiment is needed. Here, the ‘Coupled ClimateCarbon Cycle’ simulations are repeated assuming a
prescribed climate. The observed climate dataset used
in the spin-up is prescribed throughout the transient
period, 1860–2099 (i.e. radiative forcing of GHGs, both
CO2 and non-CO2, are kept constant at 1860 levels), but
the vegetation does respond directly to CO2 increases,
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
8 S . S I T C H et al.
90°N
90°N
60°N
60°N
30°N
30°N
0°
0°
30°S
30°S
180°
90°W
−0.2
−0.1
0°
−0.02
0.02
90°E
0.1
180°
−0.2
0.2
90°N
90°N
60°N
60°N
30°N
30°N
0°
0°
30°S
30°S
180°
90°W
−0.2
−0.1
0°
−0.02
0.02
180°
90°E
0.1
90°W
0.2
0°
−0.1
−0.02
90°W
−0.2
90°E
0.02
0°
−0.1
−0.02
0.02
0.1
0.2
90°E
0.1
0.2
Fig. 2 Mean seasonal patterns of rainfall change on land, units are in mm day1 K1 (winter – DJF, spring – MAM, summer – JJA,
autumn – SON).
through ‘fertilization’ effects. The difference in CO2
concentrations between the ‘Coupled’ and ‘Prescribed
Climate’ experiments gives the magnitude of the climatecarbon cycle feedback.
Climate-carbon cycle feedback analysis
Following the methodology of Friedlingstein et al.
(2003), the carbon cycle feedback gain, g, can be defined
as follows:
DCOp2
g¼ 1
;
ð1Þ
DCOc2
change, thus
DCcL ¼ bL DCOc2 þ gL DT c ;
where DCcL (Pg C) is the change in land carbon storage
due to an increase in atmospheric CO2 concentration of
DCOc2 (ppmv) in the coupled simulation and a temperature increase of DTc (K), bL is the global land carbon
sensitivity to atmospheric CO2 and gL is the global land
carbon sensitivity to climate change.
For the ‘prescribed climate’ simulation, it follows
that:
p
p
DCL ¼ bL DCO2 ;
p
where DCOc2 and DCO2 are the changes in atmospheric
CO2 mixing ratios between 2099 and 1860 for the
coupled climate-carbon cycle and the ‘prescribed
climate’ simulations, respectively. Hence, a positive
value of g indicates a positive feedback of the climate
system, i.e. the coupled system results in more atmospheric CO2.
A second metric of climate feedback strength can also
be defined following Friedlingstein et al. (2003, 2006),
whereby the change in land carbon storage can be
defined as a dependence on direct CO2 forcing and
climate change, here taken as global temperature
ð2Þ
ð3Þ
p
where DCL (Pg C) is the change in land carbon storage
due to an increase in atmospheric CO2 concentration of
p
DCO2 (ppmv) in the ‘prescribed climate’ simulation.
From Eqns (1) & (2),
gL ¼
p
p
DCcL DCL DCOc2 =DCO2
;
DT c
ð4Þ
where the numerator represents the ‘climate alone’
impact on land carbon uptake (Friedlingstein et al.,
2006).
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
U N C E RTA I N T Y I N L A N D C A R B O N C Y C L E F E E D B A C K S
Results
Contemporary carbon cycle
The results show all DGVMs to be broadly consistent
with decadal budgets of the global land carbon cycle
(Prentice et al., 2001) when forced with observed
monthly climatology (Table 2).
Over the 1980s, DGVMs simulate global mean land–
atmosphere fluxes, also known as net ecosystem exchange, NEE (RHNPP; a negative sign indicates a
land uptake of carbon), of between 1.32 and
1.80 Pg C yr1, both close to the IPCC mean value of
1.9 and within the range of 3.8 to 0.3 Pg C yr1
(Prentice et al., 2001).
Likewise for the 1990s, simulated land–atmosphere
fluxes of between 1.52 and 2.75 Pg C yr1 are close
to the IPCC mean value of 2.6 and range of 4.3 to
1.0 Pg C yr1 (Prentice et al., 2001). Also, DGVMs simulate a greater land carbon uptake in the 1990s than
during the 1980s, in agreement with IPCC estimates,
with global land uptake shared between tropical and
extra-tropical regions. DGVM estimates of land–atmosphere fluxes do not span the whole IPCC range, and are
generally less negative compared with IPCC. This may
be due to sinks related to northern forest regrowth and
nitrogen deposition that are not included in this study.
The DGVMs are also able to simulate the correct
global response to ENSO-driven interannual climate
variability (Fig. 3; lower right panel), in agreement with
earlier studies (Tian et al., 1998; Jones et al., 2001; Peylin
et al., 2005). Years with anomalous increases in the
atmospheric CO2 growth rate are synonymous with
the El Niño phenomenon (e.g. 1983, 1987, early 1990s,
1998), and correspond to peaks in global land–atmosphere exchange, and visa versa during La Niña years
(e.g. 1985, 1989, 1996, 1999) (Fig. 3; lower right panel).
Table 2 Global land carbon budgets for the 1980s and 1990s,
expressed as decadal mean land–atmosphere exchange (RhNPP), units are Pg C yr1, and the simulated cumulative land
uptake from 1958 to 2002 in Pg C
IPCC Residual Land Sink
Prentice et al. (2001)
Model
HyLand (HYL)
Lund–Potsdam–Jena (LPJ)
ORCHIDEE (ORC)
Sheffield-DGVM (SHE)
TRIFFID (TRI)
1980s
1990s
1.9 (3.8
to 0.3)
2.6 (4.3
to 1.0)
1.67
1.32
1.58
1.80
1.62
2.39
1.52
2.21
2.75
2.47
IPCC, Intergovernmental Panel on Climate Change.
1958–
2002
71.5
67.7
81.4
85.3
110.1
9
For each DGVM, the regression of interannual
anomalies in ‘natural CO2’ against annual mean Niño3 index are plotted in Fig. 4, where anomalies in the
‘natural CO2’ flux are calculated as the sum of the
annual land (from DGVMs) and ocean carbon fluxes
(Buitenhuis et al., 2006) subtracting the mean annual
flux over the previous decade.
Correlations are significant at the 95% confidence
level. The slope of the regression represents the sensitivity of the biosphere to IAV in climate (see Table 2).
Gradients range from 0.27 0.06 ppm yr1 1C1 for
HYL to 0.81 0.14 ppm yr1 1C1 for SHE, with intermediate values of 0.32 0.09, 0.42 0.09 and
0.56 0.10 ppm yr1 1C1 for LPJ, ORC and TRI, respectively. The error is calculated as the 95% confidence
interval of the regression and is shown in the figure as
dotted lines. For the period 1966–1996, excluding years
1983, 1992 and 1993 (years strongly affected by volcanic
eruption), Jones et al. (2001), estimated an observed
slope in the regression between observed CO2 anomalies at Mauna Loa and Niño-3 index of
0.51 0.09 ppm yr1 1C1. However, model slopes do
not account for atmospheric transport and, therefore,
may not be comparable with the observed slope as there
is evidence that IAV in winds is non-negligible (Dargaville et al., 2000).
Future atmospheric CO2
Results indicate large variations in projected future
atmospheric CO2 concentration associated with uncertainties in the terrestrial biosphere response to changing
climatic conditions (Fig. 5, Table 3). By 2100, atmospheric CO2 concentrations differ by up to 246 ppmv
among DGVMs for the coupled simulations with the
A1FI scenario (Table 3). The LPJ and TRI simulate the
highest future CO2 concentrations across all four SRES
scenarios.
With prescribed climate (Fig. 5; top panel), the intermodel spread is relatively small, and smaller than the
differences between SRES scenarios. This indicates a
robust behaviour of DGVMs in the way they depict CO2
fertilization and turnover rates. CO2 is lower for each
scenario than in the coupled simulations. In the coupled
climate-carbon cycle simulation (Fig. 5; bottom panel),
a larger spread among DGVMs is seen compared with
the prescribed climate simulations (Fig. 5; top panel),
illustrating that DGVMs are less robust in the way
they respond to climate. However, although the intermodel spread increases, the CO2 range is still dominated by the scenario differences (i.e. there is little
overlap between the spread bars on the right of the
bottom panel).
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
10 S . S I T C H et al.
850
Mean land precipitation
(mm yr )
Global mean land
temperature (°C)
14.0
13.5
13.0
12.5
750
700
1890 1910 1930 1950 1970 1990 2010
Year
5
380
Land-atmosphere
exchange (Pg C yr )
CO concentration (p.p.m.v)
12.0
1890 1910 1930 1950 1970 1990 2010
Year
800
360
340
320
300
0
−5
ORC
SHE
TRI
HYL
LPJ
−10
1890 1910 1930 1950 1970 1990 2010
Year
280
1890 1910 1930 1950 1970 1990 2010
Year
−2
−1
0
1
~
Mean Nino-3
(°C)
2
3
ORC
−3
−2
−1
0
1
~
Mean Nino-3
(°C)
2
3
−1
0
1
~
Mean Nino-3
(°C)
2
3
∆ CO (p.p.m.v. yr )
2
1
0
−1
−2
−3
HYL
3
2
1
0
−1
−2
−3
−3
∆ CO (p.p.m.v. yr )
∆ CO (p.p.m.v. yr )
3
∆ CO (p.p.m.v. yr )
3
2
1
0
−1
−2
−3
−3
∆ CO (p.p.m.v. yr )
Fig. 3 Global mean land climatology (temperature, 1C, red; precipitation, mm yr1, blue), atmospheric CO2 content (black) and
simulated land–atmosphere exchange over the 20th century by HyLand (HYL, black), Lund–Potsdam–Jena (LPJ, yellow), ORCHIDEE
(ORC, blue), Sheffield (SHE, green), and TRIFFID (TRI, red). Red and blue dashes represent periods of strong El Niño (red) and La Niña
(blue), respectively. Linear regressions are also plotted through the temperature and precipitation data.
3
2
1
0
−1
−2
−3
−3
LPJ
−2
−1
0
1
~
Mean Nino-3 (°C)
2
3
−1
2
3
SHE
−2
0
1
~
Mean Nino-3 (°C)
3
2
1
0
−1
−2
−3
−3
TRI
−2
Fig. 4 Regression of simulated interannual variability (IAV) in ‘Natural CO2’ (ppm) against annual mean Niño-3 temperature anomaly
( 1C) for HyLand (HYL), Lund–Potsdam–Jena (LPJ), ORCHIDEE (ORC), Sheffield (SHE) and TRIFFID (TRI).
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
U N C E R TA I N T Y I N L A N D C A R B O N C Y C L E F E E D B A C K S
11
1400
CO2 (p.p.m)
1200
Prescribed climate
1000
800
600
400
1900
1950
2000
2050
2100
Year
1400
CO2 (p.p.m)
1200
All DGVM
all SRES envelop
(light grey)
Coupled
1000
DGVM mean
all SRES envelop
(dark grey)
800
600
400
1900
1950
2000
Year
2050
Bars shows the
range in 2100
2100 produced by
several DGVM
Fig. 5 Global atmospheric CO2 mixing ratios (ppmv) for the ‘Prescribed Climate’ (top panel) and ‘Coupled’ (bottom panel) simulation,
respectively. Coloured lines represent the mean across all five Dynamic Global Vegetation Models (DGVMs) for each Special Report Emission
Scenarios (SRES) scenario (yellow – A1FI, red – A2, green – B1, blue – B2). The bars show the range among five DGVMs for each scenario.
HYL simulates the lowest CO2 concentrations for SRES
B1 (Table 3); however, the model simulates the median
concentration for the extreme emissions scenario, A1FI.
This points to the potential for a highly nonlinear response, and possible tipping points in terrestrial biosphere function to extreme climate change.
Future global land carbon cycle
The magnitude of future land uptake varies markedly
among DGVMs (Fig. 6). Note, only LPJ is run with IAVs
in climate (see ‘Methods’).
All DGVMs simulate a positive cumulative net carbon uptake by 2099 in response to changes in future
climate and atmospheric CO2 composition. All the
models simulate peak annual carbon uptake in the
mid-2050s and drop thereafter. This general shape
seems common to all DGVMs and scenarios.
Two models, ORC and SHE, simulate large increases
in vegetation biomass and moderate increases in soil
stocks (Fig. 6; middle and lower panels), whereas HYL
and TRI simulate increases in only vegetation and soil,
respectively. HYL simulates the largest gain in vegetation carbon in all SRES scenarios (367 Pg C for A1FI
scenario; Table 3), but alongside LPJ, the lowest soil
carbon gains among DGVMs. Indeed, LPJ simulates net
losses in soil carbon under all scenarios, whereas HYL
simulates net losses only in the two more extreme SRES
scenarios, A1FI and A2. This compares with TRI that
simulates lowest gains in vegetation carbon (and net
losses for A1FI and A2) and only moderate gains in soil
carbon over the period.
LPJ and SHE provided fields of simulated natural
biomass burning. Emissions from wildfires are simulated to increase from 1.6 and 4.2 Pg C yr1 for SHE and
LPJ, respectively, to 6.3 and 7.5 Pg C yr1 by 2100 in
response to changing environmental conditions for the
A1FI scenario. These increases are largely attributed to
an increase in standing biomass and an increase in
wildfire frequency over Amazonia in response to future
warming and drought.
Regional land carbon cycle and vegetation dynamics
There is a general consensus among the DGVMs in
terms of the qualitative regional response of vegetation
stocks to changing climate and atmospheric composition (Fig. 7).
All models simulate a decrease in vegetation carbon
over Amazonia (Fig. 7), in response to the reduction in
precipitation predicted by HadCM3LC. TRI simulates
the strongest Amazon dieback, with woody vegetation
replaced by herbaceous plants (Fig. 8).
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
12 S . S I T C H et al.
Table 3 Simulated atmospheric CO2 mixing ratio in 2099 for each Dynamic Global Vegetation Model (DGVM) and Special Report
Emission Scenario (SRES) combination (units are in ppmv) for the S2 simulation
Atmospheric CO2 content 2099 (ppmv) (S2)
HyLand (HYL)
A1FI
A2
B1
B2
A1FI
A2
B1
B2
A1FI
A2
B1
B2
A1FI
A2
B1
B2
Lund–Potsdam–
Jena (LPJ)
1019
1184
894
1050
535
669
611
754
Cumulative land uptake, 2000–2099 (Pg C)
320
11
302
9
320
53
315
33
Change in vegetation carbon (Pg C)
367
69
344
60
277
66
296
60
Change in soil carbon (Pg C)
48
58
41
52
43
13
20
27
ORCHIDEE
(ORC)
Sheffield-DGVM
(SHE)
TRIFFID (TRI)
994
878
553
627
938
836
571
628
1162
1031
656
739
413
374
308
309
505
438
257
290
63
53
85
69
306
278
217
223
336
294
168
194
8
8
7
1
107
97
91
85
169
144
89
97
70
62
78
68
Change in terrestrial carbon stocks between 2000 and 2099 for each combination of DGVM and SRES scenario (units are in Pg C).
All DGVMs simulate increases in vegetation carbon
over tundra ecosystems, in response to climate warming, with longer growing seasons and elevated ambient
CO2 levels all of which stimulate plant production.
ORC, TRI and LPJ simulate increasing woody coverage
in the tundra, in agreement with observational trends in
Alaska (Silapaswan et al., 2001; Sturm et al., 2001; Stow
et al., 2004; Sitch et al., 2007). LPJ simulates a marked
decrease in vegetation cover over boreal regions, with
boreal evergreen forest replaced by deciduous woody
and herbaceous plants by 2099 in the A1FI SRES scenario (Fig. 8). There is less agreement in simulated
changes in soil carbon stocks (Fig. 7). ORC and TRI
simulate large increases in soil carbon storage in highnorthern latitudes, whereas SHE and HYL simulate
only moderate increases, and LPJ a strong decrease.
HYL, LPJ and TRI simulate decreases in soil carbon
across Amazonia, whereas ORC and SHE simulate
small increases.
Although the global responses of TRI and LPJ in
terms of land uptake are similar (Fig. 6), the underlying
regional responses are markedly different. The TRI
global response is due to large decreases in vegetation
and soil carbon in the tropics, counter-balanced by large
carbon uptake in high-latitude ecosystems. Midlatitudes see a reduction in soil stocks.
LPJ simulates only a moderate Amazon dieback, and
a large reduction in boreal forest coverage and large
high-latitude losses in soil carbon. The high initial
estimates of boreal forest carbon stocks in LPJ can partly
explain the strong reduction in storage under very
strong warming accompanied by severe summer
drought. ORC simulates a reduction in vegetation carbon in the temperate–boreal ecotone in Europe with
replacement of evergreen forests by deciduous vegetation. HYL simulates large carbon uptake in all ecosystems except over Amazonia, where, similar to TRI, the
DGVM simulates a reduction in both vegetation and
soil stocks. ORC and SHE both simulate only moderate
decreases in vegetation biomass across Amazonia and
small increases in soil carbon, the latter being a qualitatively different response to TRI, HYL and LPJ. Note,
the SHE model has fixed vegetation, and does not
simulate changes in the coverage of plant functional
types (PFTs).
Terrestrial climate-carbon cycle feedbacks
Terrestrial climate-carbon cycle feedbacks are all positive and range between 40 and 319 ppmv for all DGVMs
and four SRES emission scenario combinations (Table
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
6
3
0
−3
−6
2000 2020 2040 2060 2080 2100
Year
A1F1
200
0
−200
2000 2020 2040 2060 2080 2100
Year
Change in Land uptake (Pg C yr−1)
400
9
HYL A1FI
LPJ
ORC
SHE
TRI
12
Change in vegetation carbon (Pg C)
Change in land uptake (Pg C yr−1)
12
Change in vegetation carbon (Pg C)
U N C E R TA I N T Y I N L A N D C A R B O N C Y C L E F E E D B A C K S
400
HYL B1
LPJ
ORC
SHE
TRI
9
6
3
0
−3
−6
2000 2020 2040 2060 2080 2100
Year
B1
200
0
−200
2000 2020 2040 2060 2080 2100
Year
400
A1F1
200
0
−200
2000 2020 2040 2060 2080 2100
Year
Change in soil carbon (Pg C)
400
Change in soil carbon (Pg C)
13
B1
200
0
−200
2000 2020 2040 2060 2080 2100
Year
Fig. 6 Change in land carbon uptake, Pg C yr1, (top panels) relative to the present day (mean 1980–1999) for five Dynamic Global
Vegetation Models (DGVMs) from coupled climate-carbon cycle simulations with two Special Report Emission Scenarios (SRES)
emission scenarios, A1FI (solid lines), B1 (dashed lines), bracketing the range in emissions. Change in global vegetation (middle panels)
and soil carbon (top panels), Pg C, between 2100 and 2000 under scenarios A1FI (solid lines) and B1 (dashed lines) for HyLand (HYL,
black), Lund–Potsdam–Jena (LPJ, yellow), ORCHIDEE (ORC, blue), Sheffield (SHE, green), and TRIFFID (TRI, red). Note: only LPJ is run
with interannual variations in climate (see ‘Methods’).
4). The maximum range associated with the choice of a
DGVM is 227 ppmv. LPJ and TRI have the largest
climate-carbon cycle feedbacks, SHE the lowest with
HYL and ORC being intermediate.
From Table 4, feedback gains for the DGVMs range
between 0.14 and 0.36 for the A1FI SRES scenario
and between 0.16 and 0.43 for the B1 SRES scenario.
LPJ and TRI have the highest feedback gains at 0.36
(0.43) and 0.35 (0.39) for the A1FI (B1) SRES scenario,
respectively, ORC moderate at 0.25 (0.31). SHE and
HYL have the lowest average feedback gains at 0.14
(0.20) and 0.23 (0.16) for the A1FI (B1) SRES scenario,
respectively. For a given DGVM, bL varies a great deal
among SRES scenarios, whereas gL is more robust.
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
14 S . S I T C H et al.
HYL CV
HYL CS
90°N
90°N
60°N
60°N
60°N
30°N
30°N
30°N
0°
0°
0°
30°S
30°S
30°S
180°
90°W
−6
0°
−3 −0.1 0.1
90°E
3
180°
90°W
−6
6
0°
−3 −0.1 0.1
ORC CV
90°E
3
180°
60°N
60°N
30°N
30°N
30°N
0°
0°
0°
30°S
30°S
30°S
−6
−3 −0.1 0.1
90°E
3
180°
−6
6
SHE CV
90°N
90°W
0°
−3 −0.1 0.1
90°E
3
180°
60°N
60°N
30°N
30°N
30°N
0°
0°
0°
30°S
30°S
30°S
90°W
−6
0°
−3 −0.1 0.1
90°E
3
180°
90°W
−6
6
0°
−3 −0.1 0.1
TRI CV
90°E
3
60°N
60°N
60°N
30°N
30°N
30°N
0°
0°
0°
30°S
30°S
30°S
−6
−3 −0.1 0.1
90°E
3
180°
−6
6
0°
−3 −0.1 0.1
90°E
3
180°
90°N
60°N
60°N
60°N
30°N
30°N
30°N
0°
0°
0°
30°S
30°S
30°S
90°W
−6
0°
−3 −0.1 0.1
90°W
−6
6
90°N
180°
90°E
3
180°
6
90°W
−6
0°
−3 −0.1 0.1
6
90°E
3
3
6
0°
90°E
3
6
0°
−3 −0.1 0.1
90°E
3
6
LPJ TotC
180°
6
90°E
TRI Totc
LPJ CS
LPJ CV
90°N
90°W
0°
−3 −0.1 0.1
TRI CS
90°N
0°
90°W
−6
6
90°N
90°W
3
SHE TotC
180°
90°N
18°0
0.1
−3 −0.1 0.1
90°N
60°N
180°
90°W
−6
6
SHE CS
90°N
−3 −0.1
90°E
90°N
60°N
0°
0°
ORC TotC
ORC CS
90°N
90°W
90°W
−6
6
90°N
180°
HYL TotC
90°N
90°W
−6
0°
−3 −0.1 0.1
90°E
3
6
Fig. 7 Change in land carbon storage (TotC) and component vegetation (CV) and soils (CS) carbon stocks between 1860 and 2099 from
the coupled climate-carbon cycle simulation under Special Report Emission Scenarios (SRES) emission scenario A1F1 (units are Pg C) for
HyLand (HYL), Lund–Potsdam–Jena (LPJ), ORCHIDEE (ORC), Sheffield (SHE) and TRIFFID (TRI).
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
U N C E R TA I N T Y I N L A N D C A R B O N C Y C L E F E E D B A C K S
HYL TREE
HYL HER
90°N
90°N
60°N
60°N
30°N
30°N
0°
0°
30°S
30°S
180°
90°W
−50
−20
0°
−1
90°E
1
20
180°
90°W
−50
50
−20
90°N
60°N
60°N
30°N
30°N
0°
0°
30°S
30°S
90°W
−50
−20
0°
−1
90°E
1
20
180°
−50
50
90°N
60°N
60°N
30°N
30°N
0°
0°
30°S
30°S
−50
−20
0°
−1
90°E
1
−20
20
180°
−50
50
90°N
60°N
60°N
30°N
30°N
0°
0°
30°S
30°S
−50
−20
0°
−1
90°E
1
20
50
−1
90°E
1
20
50
−20
0°
−1
90°E
1
20
50
LPJ HER
90°N
90°W
1
0°
90°W
LPJ TREE
180°
90°E
TRI HER
90°N
90°W
−1
90°W
TRI TREE
180°
0°
ORC HER
ORC TREE
90°N
180°
15
20
180°
50
90°W
−50
−20
0°
−1
90°E
1
20
50
Fig. 8 Change in vegetation coverage (%) for aggregated plant functional types, tree (TREE) and herbaceous (HER) between 1860 and
2099 for the five Dynamic Global Vegetation Models (DGVMs) from the coupled climate-carbon cycle simulation under Special Report
Emission Scenarios (SRES) emission scenario A1FI.
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
16 S . S I T C H et al.
Table 4
Model
Carbon cycle gain, g, along with component sensitivities of land carbon storage to CO2 (bL) and to climate (gL)
bL (Pg C ppm1)
Norby et al. (2005)
C4MIP, A2 Friedlingstein et al. (2006)
HadCM3LC (TRI)
1.3
IPSL-CM4-LOOP (ORC)
1.3
CLIMBER2-LPJ
1.1
C4MIP Model Range
0.2–2.8
C4MIP Model Avg
1.35
This study, A2
HyLand (HYL)
1.58
Lund–Potsdam–Jena (LPJ)
1.48
ORCHIDEE (ORC)
1.94
Sheffield-DGVM (SHE)
1.50
TRIFFID (TRI)
1.49
This study, A1FI
HYL
1.45
LPJ
1.36
ORC
1.75
SHE
1.41
TRI
1.40
This study, B1
HYL
2.64
LPJ
2.4
ORC
3.36
SHE
2.13
TRI
2.21
This Study, B2
HYL
2.16
LPJ
1.98
ORC
2.70
SHE
1.87
TRI
1.90
b550 (%)
gL (Pg C K1)
Gain, g
Climate-C-cycle
feedback (ppmv)
23
177
20
57
20 to 177
79
22
18
34
23
31
103
198
137
60
188
0.22
0.37
0.27
0.15
0.36
136
282
158
81
265
22
18
34
24
31
112
203
138
61
195
0.23
0.36
0.25
0.14
0.35
165
319
180
92
303
–
–
–
–
–
62
229
161
79
194
0.16
0.43
0.31
0.20
0.39
40
166
81
58
144
23
15
–
23
31
67
208
143
69
185
0.17
0.41
0.29
0.18
0.38
56
190
98
62
170
Calculations are made for the year 2099 relative to 1860 for the A1FI and B1 SRES scenarios. b550 represents the percentage increase
in global NPP from present day (year 2000) to future atmospheric CO2 concentrations of 550 ppm (taken from the prescribed climate
simulation). Magnitude of the climate-carbon cycle feedback between 2000 and 2099 for each combination of DGVM and SRES
scenario, coupled-prescribed climate simulations (units are in ppmv).
However, all the models produce significant positive
feedbacks, implying an acceleration of the rate of CO2
increase via the response of the land carbon cycle to
climate change.
Figure 9 shows the ‘climate alone’ changes in Land
Uptake (Pg C), NPP (Pg C yr1) and soil carbon residence
time (year) plotted against global temperature change
(K) from the coupled simulation. Values for bL, gL and
gain, g, for the A1FI and B1 model simulations are given
in Table 4, and compared against literature data (Norby
et al., 2005; Friedlingstein et al., 2006).
All DGVMs agree on a reduction in land uptake with
climate change, which implies a consensus on a positive
land climate-carbon cycle feedback. LPJ and TRI have
the largest climate-land carbon storage sensitivity (i.e.
the largest gL values). For moderate changes in global
temperature, HYL is least sensitive, although at global
temperature changes exceeding 3 1C, HYL exhibits
the strongest sensitivity to further warming. All
DGVMs agree on a decrease in global NPP, RH and
soil carbon residence time (Cs/RH) with climate warming. Changes in RH are a composite response of decomposition rate and litter inputs to climate warming.
Although the decomposition rates increase with climate
change, seen here as a decrease in soil carbon residence
time, soil carbon stocks decline, in particular due to
declining litter input, via reductions in NPP.
The extra-tropical response of land carbon to climate
warming differs among models, with LPJ, TRI and ORC
simulating reductions in uptake (Fig. 10), and simulated
uptake by SHE and HYL remaining unchanged. The
latter is a result of the counterbalancing effects of an
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
U N C E R TA I N T Y I N L A N D C A R B O N C Y C L E F E E D B A C K S
Change in soil C residence time (year)
40
250
TRI
SHE
ORC
LPJ
HYL
−10
−600
−1000
0
1
2
3
4
5
6
Surface temperature change (K)
5
−5
−200
−1400
400 550 700 850 1000
Atmospheric CO2 (p.p.m)
0
Change in NPP (Pg C yr−1)
80
Change in land uptake (Pg C)
100
0
−20
−40
−60
−80
−100
0
1
2
3
4
5
6
Surface temperature change (K)
20
Change in RH (Pg C yr −1)
Net primary productivity (Pg C yr −1)
120
60
20
200
140
17
0
−20
−40
TRI
−60
SHE
ORC
−80
LPJ
HYL
−15
0
1
2
3
4
5
6
Surface temperature change (K)
−100
0
1
2
3
4
5
6
Surface temperature change (K)
Fig. 9 Simulated net primary productivity (NPP) sensitivity to atmospheric CO2 (prescribed climate simulation). Simulated land
uptake sensitivity, net primary productivity (NPP), heterotrophic respiration (RH) (coupled-prescribed climate) and soil residence time
(from the coupled simulation) to global mean temperature change for two Special Report Emission Scenarios (SRES) emission scenarios,
A1FI (solid line) and B1 (dashed line) for five Dynamic Global Vegetation Models (DGVMs), HyLand (HYL, black), Lund–Potsdam–Jena
(LPJ, yellow), ORCHIDEE (ORC, blue), Sheffield (SHE, green) and TRIFFID (TRI, red).
increase in extra-tropical NPP and a decrease in soil
carbon residence time with warming. For NPP, ORC
and TRI are fairly insensitive to warming in the extratropics. Despite a reduction in boreal evergreen forests
in LPJ, caused by a heat and summer drought induced
reduction in boreal forest NPP, the deciduous and
herbaceous PFTs, which are better suited to this new
environment, have high NPP. Hence, the reductions in
boreal vegetation carbon simulated by LPJ may be a
transitory effect, and a new equilibrium may be approached after 2100 in which vegetation stocks recover
(Smith & Shugart, 1993).
As a test, the LPJ A1FI coupled simulation was
extended after 2100, with fixed environmental conditions of 2100, until a new equilibrium in terms of
vegetation distribution and land carbon pools was
reached. Equilibrium land carbon for year 2100 was
simulated at 2263 Pg C compared with 2412 Pg C at the
end of the transient simulation at 2100. Vegetation
biomass was higher in the equilibrium for year 2100
at 1092 Pg C compared with 956 Pg C from the transient
simulation, although this hides large regional differences in both sign and magnitude (e.g. the ‘Boreal’ and
Amazonian forests continue to loose biomass as the
new equilibrium at 2100 conditions is approached).
Nevertheless, in time boreal evergreen forests are replaced by open deciduous woodland, with lower woody coverage and biomass than the original forest. The
soils are far from equilibrium by the end of the transient
simulation at 2100. LPJ simulates an equilibrium global
soil carbon stock of 1171 Pg C for year 2100 compared
with 1456 Pg C at the end of the transient. In general,
the disequilibrium in carbon stocks north of 301N are
controlled by soil carbon (whose decomposition is slow
under ambient conditions), whereas in the tropics,
vegetation carbon is more important.
All DGVMs simulate large reductions in NPP over
the tropics with climate warming. NPP of TRI is most
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
−200
−600
−1000
−1400
0
1
2
3
4
5
6
Surface temperature change (K)
10
0
−10
−20
−30
−40
−50
0
1
2
3
4
5
6
Surface temperature change (K)
Change in NPP (Pg C yr−1)
200
Change in RH (Pg C yr−1)
Change in soil C residence
time (year)
Change in land uptake
(Pg C)
18 S . S I T C H et al.
20
0
−20
−40
−60
−80
−100
0
1
2
3
4
5
6
Surface temperature change (K)
20
0
−20
−40
−60
−80
−100
TRI
SHE
ORC
LPJ
HYL
0
1
2
3
4
5
6
Surface temperature change (K)
Fig. 10 Simulated regional sensitivity of land uptake, net primary productivity (NPP), heterotrophic respiration (RH) (coupledprescribed simulations) and soil residence time (from the coupled simulation) to global mean temperature change for the A1FI Special
Report Emission Scenarios (SRES) emission scenario for five Dynamic Global Vegetation Models (DGVMs), HyLand (HYL, black), Lund–
Potsdam–Jena (LPJ, yellow), ORCHIDEE (ORC, blue), Sheffield (SHE, green) and TRIFFID (TRI, red). Extra-tropics are defined as land
north of 301N and south of 301S, solid lines; and tropics as land between 301S and 301N, dashed lines.
sensitive to climate change. DGVMs agree on an increase in extra-tropical RH and decreases in tropical RH
with global warming. The RH response is a composite
of changes in litter input (NPP and vegetation mortality) and changes in the soil C decomposition rate with
climate change. Despite little change in extra-tropical
NPP in LPJ, RH increases due to a large decrease in soil
residence time with warming. All DGVMs simulate
reductions in soil carbon residence time across the extra-tropics. In the tropics, soil carbon residence time is
insensitive to global climate change with large reductions in NPP matched by equally large reductions in RH,
implying a possible substrate limitation to tropical RH.
Discussion
Major features of the modelled carbon cycle
Results show that all DGVMs are consistent with global
land carbon budgets for the 1980s and 1990s, and are in
agreement with other modelling studies on cumulative
land uptake over the last 50 years (McGuire et al., 2001).
DGVMs are also able to simulate the correct sign of the
global land carbon response to ENSO, but with differing magnitudes of response.
Although all five DGVMs simulate cumulative net
carbon uptake by 2099 in response to changes in future
climate and atmospheric composition, the magnitude of
land uptake varies markedly among DGVMs. Results
indicate large uncertainties in future atmospheric CO2
concentration associated with uncertainties in the terrestrial biosphere response to changing climatic
conditions.
All five DGVMs have similar response of productivity to elevated atmospheric CO2. NPP is stimulated by
between 18% and 34% when atmospheric CO2 (from the
prescribed climate simulation) is elevated from ambient
concentrations to 550 ppmv. This is in good agreement
with a median stimulation of 23% for the forest sites in
the Free-Air-CO2 Enrichment experiments (Norby et al.,
2005). DGVMs agree much less in the way they respond
regionally to changing climate, confirmed by the large
range in gL (the sensitivity of land carbon storage to
climate) among DGVMs (Table 4).
Dependence on modelled climate change
The DGVM response is very much linked to the GCM
climatology applied (Berthelot et al., 2005; Schaphoff
et al., 2006; Scholze et al., 2006). LPJ run with HadCM2
simulates large land carbon uptake, whereas with
HadCM3, from Schaphoff et al. (2006), the model simulates a land source of carbon after 2050. Only LPJ run
with CGCM and CSIRO GCM climatology project a
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
U N C E R TA I N T Y I N L A N D C A R B O N C Y C L E F E E D B A C K S
greater land carbon source and sink, respectively
(Schaphoff et al., 2006). LPJ had a very moderate response from Cramer et al. (2001) study using the more
‘moderate’ HadCM2 climatology, as well as from C4MIP
study (Friedlingstein et al., 2006) where it was coupled to
the CLIMBER-2 Earth System Model of Intermediate
Complexity (EMIC) (Petoukhov et al., 2000).
A warming over land in CLIMBER-2 is much weaker
than in the HadCM3. In the C4MIP run, land temperatures in the latitudinal zone 30–601N increase by 4 1C in
CLIMBER-LPJ, while in the HadCM3 run warming in
this region is 6–9 1C for IS92. In the A1FI scenario
applied in this study, the warming is even stronger.
Also, the precipitation increase with warming (dP/dT)
in CLIMBER is stronger than in the HadCM3.
Indeed, CLIMBER-LPJ in C4MIP has a gL of
57 Pg C K1 compared with 198 Pg C K1 for IMOGEN-LPJ. As expected, the gL from HadCM3LC-TRI in
C4MIP (177 Pg C K1) is similar to IMOGEN-TRI
(188 Pg C K1) in this study. However, this is clearly
not the case for ORC and LPJ, with a gL for IPSL-ORC in
C4MIP of 20 Pg C K1 compared with 137 Pg C K1
for IMOGEN-ORC in this study. The implication is that
the climate-carbon cycle feedback is also highly dependent on the nature of the simulated climate change.
The IMOGEN climate simulation did not include the
cooling effect of sulphate aerosols, and as a result, the
rate of warming over the historical period in the
coupled-climate cycle experiment is greater than observed (Jones et al., 2003). For the original Cox et al.
(2000) runs, the land temperature for 1991–2000 is about
1.8 1C warmer than the 1860–1890 average (global
mean is about 1.2 1C warmer), whereas observations
for land, indicate a 1 1C warming with a global mean
of 0.7–0.8 1C. In the ‘sulphates 1 natural’ runs of
Jones et al. (2003), the land temperature for the period
1991–2000 was 0.8 1C warmer than the preindustrial,
with a global mean of 0.6 1C. Unsurprisingly, the
DGVMs with the largest climate-carbon cycle feedbacks
(LPJ and TRI) also simulate the smallest contemporary
land uptake under the excessive historical climate
warming simulated in our coupled climate-carbon
cycle experiment (results not shown). LPJ driven
with anomalies from a HadCM3 climatology which
includes sulphate aerosols simulated a larger contemporary land carbon uptake (Schaphoff et al., 2006;
Scholze et al., 2006), just as HadCM3LC did when
aerosol effects were included (Jones et al., 2003). This
does indicate, however, that according to the more
‘pessimistic’ DGVMs, terrestrial ecosystems have the
potential to become a net global source of carbon in the
coming decades if the cooling effect of sulphates has
been underestimated, and drops off as anticipated
(Andreae et al., 2005).
19
Responses of ecosystem processes to heat and drought
All DGVMs simulate a reduction in soil carbon in
response to climate forcing. Three DGVMs simulate
reductions in soil carbon across tropical ecosystems,
and four models simulate reductions across northern
midlatitudes, the latter in broad agreement with Cox
et al. (2000). This is despite rather different soil model
formulations. Nevertheless, there remains a large ‘process’ uncertainty among models due to differential
decomposition-moisture responses (Peylin et al., 2005).
There has been much discussion in literature about
the magnitude of soil decomposition sensitivity to
temperature and whether this response would be sustained over the coming decades or if it is a short-lived
phenomenon (Davidson et al., 2000; Giardina & Ryan,
2000; Knorr et al., 2005). Based on experimental data
synthesis, Giardina & Ryan (2000) argue that readily
decomposable soil organic matter (SOM) is mainly
responsible for the observed temperature sensitivity,
implying long-term soil respiration to be governed by
substrate availability and litter quality. Davidson et al.
(2000) refuted these conclusions, and argued that temperature sensitivity is just one of the many uncertain
factors difficult to ascertain in isolation. Knorr et al.
(2005) have shown how these conflicting opinions are
compatible with long-term temperature sensitivity of
soil respiration, with the experimental findings explained by a rapid depletion of labile SOM with negligible response of nonlabile SOM on experimental
timescales. The review of Davidson & Janssens (2006)
identified the need for decomposition to be seen as
dependent on many factors simultaneously, such as soil
temperature, moisture structure and litter quality.
The quantitative response of DGVMs to drought
differs among DGVMs, with TRI and HYL most sensitive to reduced rainfall and elevated temperatures
across Amazonia. LPJ and ORC simulate moderate
forest-dieback. Drought-induced plant mortality results
from decreased photosynthesis, leading to resource
limitations, and/or to plant-hydraulic failure (Van
Nieuwstadt & Sheil, 2005).
In a drought experiment in an east-central Amazonian rainforest at Tapajós, a 50% reduction in precipitation led to a 25% reduction in NPP over the
first 2 years of the experiment (Nepstad et al., 2002).
Despite reductions in NPP and leaf area, there was no
immediate drought response of trees (e.g. leaf senescence). With deep roots, trees can access soil moisture
reserves and are able to withstand several years of
drought. Nevertheless, the ecosystem response to persistent, prolonged drought may lead to increased forest
mortality, as appears to be the case at Tapajós (Saleska
et al., 2003).
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
20 S . S I T C H et al.
Several studies have recorded increased rates of tree
mortality during severe El Niño years across neo-tropical forests (Condit et al., 1995; Williamson et al., 2000).
During the 1997 drought, plots in the central Amazon
near Manaus, received only 32% of the normal dry
season rainfall, leading to a 70% increase in tree mortality to 1.91% yr1 (Williamson et al., 2000). However,
mortality rates returned to near-normal levels in subsequent years, and therefore a single drought appears to
have only a modest impact of ecosystem structure
(Williamson et al., 2000).
In addition to the direct effect of drought on plant
mortality, an indirect effect via increased fire risk is
likely to exacerbate matters (Nepstad et al., 1999, 2002;
Van Nieuwstadt & Sheil, 2005). The resilience of tropical
forests to more frequent and severe droughts, both in
terms of the direct drought effect and indirect, via
biomass burning, is key to understanding the potential
for large-scale tropical forest dieback and its implications for the global carbon cycle. Existing DGVMs show
a range of responses to reductions in precipitation, from
a resilient forest (LPJ, ORC) to a vulnerable forest (TRI,
HYL). Several studies (Cox et al., 2004; Huntingford
et al., 2004) indicate a degree of uncertainty in the onset
of Amazonian forest-dieback with HadCM3 climatology, relating to the choice of DGVM parameterization
and driving climatology.
Also important for the carbon balance of tropical
forests is deforestation (Cramer et al., 2004). Further,
the impoverished secondary ecosystems and primary
forests bordering deforested land are likely to be more
susceptible to wildfire (Cochrane et al., 1999; Nepstad
et al., 1999), and frequent disturbance is likely to affect
the ability of forests to regenerate. In general, a greater
process-based understanding of large-scale plantdrought responses and interaction with wild-fire and
land-use, is needed, and this should filter into the next
generation of DGVMs. Indeed, although the effects of
land-use and land cover changes are very important for
future biogeography and biogeochemistry, inclusion is
beyond the scope of the present study. This will be a
major focus of development in the next generation of
DGVMs.
Another interesting finding is the differential response of LPJ and ORC vegetation dynamics in the
boreal forests. Despite ORC vegetation dynamics being
closely related to that of LPJ, their response is qualitatively different. LPJ simulates a boreal forest dieback in
response to strong climate warming (Joos et al., 2001;
Lucht et al., 2006), a combined result of suboptimal
photosynthesis at high temperatures (related to the
PFT-specific photosynthesis–temperature response
curves), and plant response to a reduction in summer
precipitation (i.e. summer drought). For LPJ, temperate
trees and herbaceous vegetation are favoured in the
future climatic conditions of HadCM3. The temperature
optima and high temperature limits for photosynthesis
of evergreen conifers used in LPJ range between 10 and
25 1C, and 35 and 42 1C, respectively (Larcher, 1983).
Temperate deciduous trees, on the other hand, have
optima at 15–25 1C, and high temperature limits at 40–
45 1C. Because the temperate PFTs are assigned higher
temperature ranges, they gradually replace the boreal
types and hence, LPJ has a seemingly paradoxical overall increase in NPP over the ‘boreal’ zone, but a reduction in boreal forest coverage. The optimal temperature
ranges for photosynthesis among PFTs in ORC acclimate to recent climate conditions, and also ORC employs a different photosynthesis scheme. Given the
importance of NPP in driving vegetation dynamics
among DGVMs, it is not surprising the response
of the two DGVMs diverge. In a sensitivity study,
Matthews et al. (2007) show the importance of the representation of the photosynthesis–temperature response for
the strength of the climate-carbon cycle feedback.
The prospect of Europe’s climate becoming more
Mediterranean, with warmer summers, reduced rainfall
and more frequent and severe droughts, will likely
impact vegetation production, carbon sequestration,
vegetation structure and disturbance regimes, favouring more high-temperature tolerant and drought-resistant species. The potential for such changes in
biogeochemistry is evident from the 2003 summer
drought (Ciais et al., 2005) and the recent drier summers
in mid- and high northern latitudes (Angert et al., 2005).
Together, this points to a critical element in modelling
dynamic global vegetation; the number of PFTs defined,
their respective optimal ranges, and ability of plant
species within PFT groups to adapt and plant processes
to acclimate to new environmental conditions.
Role of nutrient constraints
DGVMs have been criticized for disregarding the potential effects of nutrient (especially nitrogen) limitation
on the ability of ecosystems to sequester CO2. Hungate
et al. (2003) suggested that the CO2 responses projected
by Cramer et al. (2001) were much too large and that
future modelling work must include interactive nitrogen cycling. This is indeed a focus of much current
DGVM development (Prentice et al., 2007). However,
the issue is more complex than indicated by Hungate
et al. (2003) for several reasons. First, multiyear free-air
carbon dioxide enrichment (FACE) experiments in temperate forests have not supported the preliminary indications (Oren et al., 2001) of a rapid decline in the
stimulation of NPP due to nitrogen shortage [see e.g.
Moore et al. (2006)]. Second, biogeochemistry models
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
U N C E R TA I N T Y I N L A N D C A R B O N C Y C L E F E E D B A C K S
with interactive nitrogen cycling have shown an additional stimulation of NPP due to increased rates of
mineralization in a warmer climate (Melillo et al.,
1993; VEMAP, 1995). In this study, SHE includes interactive nitrogen cycling. This model shows overall one of
the smallest reductions of NPP in response to warming
among DGVMs, and shows a strong positive response
of NPP to CO2. Thus, it is not clear what impact a
realistic representation of carbon–nutrient interactions
would have on the tendency of the land to act as a
carbon source or sink.
Benchmarking
Benchmarking global models is a key procedure, to
enable confidence in their future projections (Dargaville
et al., 2002; Morales et al., 2005). This study has shown
the ability of models to satisfy contemporary global
carbon cycle constraints, while future projections diverge markedly. Jones et al. (2006) noted a similar
phenomenon with a simple carbon cycle model. Many
different parameter combinations were able to recreate
the historical record, but their behaviour diverged in the
future. Process-based, local observations are required to
constrain models, as well as large-scale observations
that hide cancellation of processes. An expanded set of
data for evaluation of short-timescale dynamics for
benchmarking (e.g. Morales et al., 2005; Friend et al.,
2006) is useful. However, DGVMs should also be evaluated for longer time-scale dynamics, e.g. future
drought-induced and extreme heat stress mortality.
Information is needed from drought experiments (e.g.
in tropical rainforests, Nepstad et al., 2002; Asner et al.,
2004; Meir & Grace, 2004) and extra-tropical ecosystems
(Hanson et al., 2001; Moorcroft et al., 2004), from actual
large-scale regional droughts, (e.g. the European
drought 2003, Ciais et al., 2005), from Paleo data (Scheffer et al., 2006), including tree-rings, and from warming
experiments in boreal forests (Smith & Shugart, 1993;
Marchand et al., 2005).
Conclusion
This study indicates large uncertainties in future atmospheric CO2 concentrations associated with uncertainties in the terrestrial biosphere response to changing
climatic conditions. All DGVMs simulate cumulative
net carbon uptake by 2099 in response to changes in
future climate and atmospheric composition for all
SRES scenarios; however, the magnitude of this uptake
varies markedly among DGVMs. All five DGVMs have
similar response of productivity to elevated atmospheric CO2 in agreement with field observations (Norby et al., 2005).
21
The DGVMs are in less agreement in the way they
respond to changing climate. However, consistent
among DGVMs is a release of land carbon in response
to climate, implying a significant positive climate-carbon cycle feedback in each case. This response is mainly
due to a reduction in NPP and a decrease in soil
residence time in the tropics and extra-tropics, respectively. Major DGVM uncertainties include the following: NPP response to climate in the tropics; soil
respiration response to climate in the extra-tropics.
Uncertainty in future cumulative land uptake
(494 Pg C) associated with land processes is equivalent
to over 50 years of anthropogenic emissions at current
levels. Therefore, improving our ability to model terrestrial biosphere processes (e.g. plant response to
drought/heat stress) is paramount if we are to enhance
our ability to predict future climate change.
Acknowledgements
The authors wish to thank the following for their contribution to
this study: Martin Best and Ben Booth for advice on IMOGEN,
Werner von Bloh and Sibyll Schaphoff for technical advice on
LPJ, Olivier Boucher for comments on the manuscript, Victor
Brovkin for helpful comments on LPJ results, Andrew Everitt for
computational support at CEH Wallingford and Andrew Friend
for assistance with the climate data. The contribution of R. A. B.,
C. D. J., N. G., S. S. was supported by the UK DEFRA Climate
Prediction Programme under Contract PECD 7/12/37. This
study was also supported by the ENSEMBLES FP6 and the
NERC QUEST programmes.
References
Andreae MO, Jones CD, Cox P (2005) Strong present-day aerosol
cooling implies a hot future. Nature, 435, 1187–1190, doi:
10.1038/nature03671.
Angert A, Biraud S, Bonfils C et al. (2005) Drier summers cancel
out the CO2 uptake enhancement induced by warmer springs.
Proceedings of the National Academy of Sciences of the United
States of America, 102, 10823–10827.
Asner GP, Nepstad D, Cardinot G, Ray D (2004) Drought stress
and carbon uptake in an Amazon forest measured with spaceborne imaging spectroscopy. Proceedings of the National Academy of Sciences of the United States of America, 101, 6039–6044.
Ball JT, Woodrow IE, Berry JA (1987) A model predicting
stomatal conductance and its to the control of photosynthesis
under different environmental conditions. In: Progress in Photosynthesis (ed. Biggins I), pp. 221–224. Martinus Nijhoff Publishers, the Netherlands.
Bartholome E, Belward AS, Achard F et al. (2002) GLC 2000 Global
land cover mapping for the year 2000. Project status November
2002. European Commission, Joint Research Centre.
Berthelot M, Friedlingstein P, Ciais P, Dufresne J-L, Monfray P
(2005) How uncertainties in future climate change predictions
translate into future terrestrial carbon fluxes. Global Change
Biology, 11, 959–970, doi: 10.1111/j.1365-2486.2005.00957.x.
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
22 S . S I T C H et al.
Betts RA (2000) Offset of the potential carbon sink from boreal
forestation by decreases in surface albedo. Nature, 408,
187–190.
Betts RA, Cox PM, Collins M, Harris PP, Huntingford C, Jones
CD (2004) The role of ecosystem-atmosphere interactions in
simulated Amazonian precipitation decrease and forest dieback under global climate warming. Theoretical Applied Climatology, 78, 157–175.
Botta A, Viovy N, Ciais P, Friedlingstein P, Monfray P (2000) A
global prognostic scheme of leaf onset using satellite data.
Global Change Biology, 6, 704–725.
Brovkin V, Ganopoloski A, Claussen M, Kubatzki C, Petoukhov
V (1999) Modelling climate response to historical land cover
change. Global Ecology and Biogeography, 8, 509–517.
Brovkin V, Sitch S, von Bloh W, Claussen M, Bauer E, Cramer W
(2004) Role of land cover changes for atmospheric CO2 increase and climate change during the last 150 years. Global
Change Biology, 10, 1–14, doi: 10.1111/j.1365-2486.2004.00812.x.
Buitenhuis ET, Le Quéré C, Aumont O et al. (2006) Biogeochemical fluxes mediated by mesozooplankton, in revision. Global
Biogeochemical Cycles, 20, GB2003, doi: 10.1029/2005GB002511.
Ciais P, Reichstein M, Viovy N et al. (2005) Europe-wide reduction in primary production caused by the heat and drought in
2003. Nature, 437, 529–533.
Cochrane MA, Alencar A, Schulze MD, Souza CM, Nepstad DC,
Lefebvre P, Davidson EA (1999) Positive feedbacks in the fire
dynamic of closed canopy tropical forests. Science, 284, 1832–
1835.
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. Agricultural and Forest Meteorology, 54, 107–136.
Collatz GJ, Ribas-Carbo M, Berry JA (1992) Coupled photosynthesis-stomatal conductance model for leaves of C4 plants.
Australian Journal of Plant Physiology, 19, 519–538.
Comins HN, McMurtrie RE (1993) Long-term response of nutrient-limited forests to CO2 enrichment – equilibrium behavior of plant-soil models. Ecological Applications, 3, 666–681.
Condit R, Hubbell SP, Foster RB (1995) Mortality rates of 205
neotropical tree and shrub species and the impact of a severe
drought. Ecological Monographs, 65, 419–439.
Cox PM (2001) Description of the ‘‘TRIFFID’’ dynamic global
vegetation model. Hadley Centre Technical Note 24.
Cox PM, Betts RA, Collins M (2004) Amazonian forest dieback
under climate-carbon cycle projections for the 21st century.
Theoretical and Applied Climatology, 78, 137–156, doi: 10.1007/
s00704-004-0049-4.
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, Huntingford C, Harding RJ (1998) A canopy conductance and photosynthesis model for use in a GCM land surface
scheme. Journal of Hydrology, 212–213, 79–94.
Cramer W, Bondeau A, Schaphoff S, Lucht W, Smith B, Sitch S
(2004) Tropical forests and the global carbon cycle: impacts of
atmospheric CO2, climate change and rate of deforestation.
Philosophical transactions of the Royal Society of London. Series B,
Biological Sciences, 359, 331–343, doi: 10.1098/rstb.2003.1428.
Cramer W, Bondeau A, Woodward FI et al. (2001) Global response of terrestrial ecosystem structure and function to CO2
and climate change: results from six dynamic global vegetation models. Global Change Biology, 7, 357–373.
Dargaville RJ, Heimann M, McGuire AD et al. (2002) Evaluation
of terrestrial carbon cycle models with atmospheric CO2
measurements: results from transient simulations considering
increasing CO2, climate, and land-use effects. Global Biogeochemical Cycles, 16, 1092, doi: 10.1029/2001GB001426.
Dargaville RJ, Law RM, Pribac F (2000) Implications of interannual variability in atmospheric circulation on modelled CO2
concentrations and source estimates. Global Biogeochemical
Cycles, 14, 931–944.
Davidson E, Janssens I (2006) Temperature sensitivity of soil
carbon decomposition and feedbacks to climate change. Nature, 440, 165–173.
Davidson EA, Trumbore SE, Amundson R (2000) Soil warming
and organic carbon content. Nature, 408, 789–790.
Ducoudré NI, Laval K, Perrier A (1993) SECHIBA, a new set of
parameterizations of the hydrologic exchanges at the land–
atmosphere interface within the LMD atmospheric general
circulation model. Journal of Climate, 6, 248–273.
Dufresne J-L, Friedlingstein P, Berthelot M et al. (2002) Effects of
climate change due to CO2 increase on land and ocean carbon
uptake. Geophysical Research Letters, 29, 1–1, doi: 10.1029/
2001GL013777.
Farquhar GD, von Caemmerer S, Berry JA (1980) A biochemical
model of photosynthetic CO2 assimilation in leaves of C3
species. Planta, 149, 78–90.
Friedlingstein P, Cox PM, Betts R et al. (2006) Climate-carbon
cycle feedback analysis, results from the C4MIP model intercomparison. Journal of Climate, 19, 3337–3353.
Friedlingstein P, Dufresne J-L, Cox PM, Rayner P (2003) How
positive is the feedback between climate change and the
carbon cycle? Tellus B, 55B, 692–700.
Friedlingstein P, Joel G, Field CB, Fung IY (1998) Toward an
allocation scheme for global terrestrial carbon models. Global
Change Biology, 5, 755–770.
Friend AD (1995) PGEN – an integrated model of leaf photosynthesis, transpiration, and conductance. Ecological Modelling,
77, 233–255.
Friend AD (1998) Parameterisation of a global daily weather
generator for terrestrial ecosystem modeling. Ecosystem Modelling, 109, 121–140.
Friend AD, Arneth A, Kiang NY et al. (2006) FLUXNET and
modelling the global carbon cycle. Global Change Biology, 12,
1–24, doi: 10.1111/j.1365-2486.2006.01223.x.
Friend AD, Stevens AK, Knox RG, Cannell MGR (1997) A
process-based, terrestrial biosphere model of ecosystem dynamics (Hybrid v3.0). Ecological Modelling, 95, 249–287.
Friend AD, White A (2000) Evaluation and analysis of a
dynamic terrestrial ecosystem model under preindustrial conditions at the global scale. Global Biogeochemical Cycles, 14,
1173–1190.
Gerten D, Schaphoff S, Haberlandt U, Lucht W, Sitch S (2004)
Terrestrial vegetation and water balance – hydrological evaluation of a dynamic global vegetation model. Journal of
Hydrology, 286, 249–270.
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
U N C E R TA I N T Y I N L A N D C A R B O N C Y C L E F E E D B A C K S
Giardina C, Ryan M (2000) Evidence that decomposition rates of
organic carbon in mineral soil do not vary with temperature.
Nature, 404, 858–861.
Gifford RM (1995) Whole plant respiration and photosynthesis of
wheat under increasing CO2 concentration and temperature:
long-term vs. short-term distinctions for modelling. Global
Change Biology, 1, 385–396.
Gitz V, Ciais P (2003) Amplifying effects of land-use change on
future atmospheric CO2 levels. Global Biogeochemical Cycles, 17,
1024, doi: 10.1029/2002GB001963.
Gordon C, Cooper C, Senior C et al. (2000) The simulation of SST,
sea ice extents and ocean heat transport in a version of the
Hadley Centre coupled model without flux adjustments. Climate Dynamics, 16, 147–168.
Hanson PJ, Todd DE Jr, Amthor JS (2001) A six-year study of
sapling and large-tree growth and mortality response to
natural and induced variability in precipitation and throughfall. Tree Physiology, 21, 345–358.
Haxeltine A, Prentice IC (1996) BIOME3: an equilibrium
terrestrial biosphere model based on ecophysiological
constraints, resource availability, and competition among
plant functional types. Global Biogeochemical Cycles, 10,
693–709.
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.
Hungate B, Dukes JS, Shaw MR, Luo Y, Field CB (2003) Nitrogen
and climate change. Science, 302, 1512–1513.
Huntingford C, Cox PM (2000) An analogue model to derive
additional climate change scenarios from existing GCM simulations. Climate Dynamics, 16, 575–586.
Huntingford C, Harris PP, Gedney N, Cox PM, Betts RA,
Marenso J, Gash JHC (2004) Using a GCM analogue model
to investigate the potential for Amazonian forest dieback.
Theoretical and Applied Climatology, 78, 177–185.
Jarvis PG (1976) The interpretation of the variations in leaf water
potential and stomatal conductance found in canopies in the
field. Philosophical Transactions of the Royal Society of London
Series B, 273, 593–610.
Jones CD, Collins M, Cox PM, Spall SA (2001) The carbon cycle
response to ENSO: a coupled climate-carbon cycle model
study. Journal of Climate, 14, 4113–4129.
Jones CD, Cox PM, Essery RLH, Roberts DL, Woodage MJ (2003)
Strong carbon cycle feedbacks in a climate model with interactive CO2 and sulphate aerosols. Geophysical Research Letters,
30, 1479, doi: 10.1029/2003GL016867.
Jones CD, Cox PM, Huntingford C (2006) Climate-carbon cycle
feedbacks under stabilisation: uncertainty and observational
constraints. Tellus, 58B, 603–613, doi: 10.1111/j.1600-0889.2006.
00215.x.
Joos F, Bruno M, Fink R, Siegenthaler U, Stocker TF, LeQuéré C,
Sarmiento JL (1996) An efficient and accurate representation of
complex oceanic and biospheric models of anthropogenic
carbon uptake. Tellus, 48B, 397–417.
Joos F, Prentice IC, Sitch S et al. (2001) Global warming feedbacks
on terrestrial carbon uptake under the Intergovernmental
Panel on Climate Change (IPCC) emission scenarios. Global
Biogeochemical Cycles, 15, 891–907.
23
Keeling CD, Whorf TP (2005) Atmospheric CO2 records from
sites in the SIO sampling network. In: Trends: A Compendium of
Data on Global Change. Carbon Dioxide Information Analysis
Center, Oak Ridge National Laboratory, US Department of
Energy, Oak Ridge, TN, USA. http://cdiac.ornl.gov/trends/
co2/sio-mlo.htm
Knorr W, Prentice IC, House JC, Holland EA (2005) Long-term
sensitivity of soil carbon turnover to warming. Nature, 433,
298–301.
Krinner G, Viovy N, de Noblet-Ducoudré N et al. (2005) A
dynamic global vegetation model for studies of the coupled
atmosphere-biosphere system. Global Biogeochemical Cycles, 19,
GB1015, doi: 10.1029/2003GB002199.
Larcher W (1983) Physiological Plant Ecology. Springer, Heidelberg.
Leemans R, Cramer W (1991) The IIASA Databased for Mean
Monthly Values of Temperature, Precipitation, and Cloudiness of a
Global Terrestrial Grid. International Institute for Applied Systems Analysis (IIASA) RR-91-18, Vienna, Austria.
Leuning R (1995) A critical appraisal of a combined stomatalphotosynthesis model for C3 plants. Plant, Cell and Environment, 18, 339–355.
Levy PE, Cannell MGR, Friend AD (2004) Modelling the impact
of future changes in climate, CO2 concentration and land use
on natural ecosystems and the terrestrial carbon sink. Global
Environmental Change, 14, 21–30.
Lloyd J, Taylor JA (1994) On the temperature dependence of soil
respiration. Functional Ecology, 8, 315–323.
Lucht W, Schaphoff S, Erbrecht T, Heyder U, Cramer W (2006)
Terrestrial vegetation redistribution and carbon balance under
climate change. Carbon Balance and Management, 1, 1–7, doi:
10.1186/1750-0680-1-6.
Marchand FL, Mertens S, Kockelbergh F, Beyens L, Nijs I (2005)
Performance of high Arctic tundra plants improved during
but deteriorated after exposure to a simulated extreme temperature event. Global Change Biology, 11, 2078–2089.
Marland G, Boden TA, Andres RJ (2003) 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, US Department of
Energy, Oak Ridge, TN, USA. http://cdiac.ornl.gov/trends/
emis/tre_glob.htm
Matthews HD, Eby M, Ewen T, Friedlingstein P, Hawkins BJ
(2007) What determines the magnitude of carbon cycle-climate
feedbacks. Global Biogeochemical Cycles, 21, GB2012, doi:
10.1029/2006GB002733.
McGuire AD, Melillo JM, Joyce LA, Kicklighter DW, Grace AL,
Moore B III, Vorosmarty CJ (1992) Interactions between carbon
and nitrogen dynamics in estimating net primary productivity
for potential vegetation in North America. Global Biogeochemical Cycles, 6, 101–124.
McGuire AD, Sitch S, Clein JS et al. (2001) Carbon balance of
the terrestrial biosphere in the twentieth century: analyses
of CO2, climate and land use effects with four processbased ecosystem models. Global Biogeochemical Cycles, 15,
183–206.
Meir P, Grace J (2004) The effects of drought on tropical forest
ecosystems. In: Tropical Forests & Global Atmospheric Change
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
24 S . S I T C H et al.
(eds Malhi Y, Phillips O), pp. 75–84. Oxford University Press,
Oxford, UK.
Melillo JM, McGuire AD, Kicklighter DW, Moore B III, Vorosmarty CJ, Schloss AL (1993) Global climate change and terrestrial net primary production. Nature, 363, 234–240.
Monteith JL (1981) Evaporation and environment. In: The State
and Movement of Water in Living Organisms (ed. Fogg CE),
pp. 205–234. Cambridge University Press, Cambridge, UK.
Monteith JL (1965) Accommodation between transpiring vegetation and the convective boundary layer. Journal of Hydrology,
166, 251–263.
Monteith JL, Unsworth MH (1990) Principles of Environmental
Physics. Edward Arnold, London.
Moorcroft MD, Masters GJ, Brown VK, Clarke IP, Taylor ME,
Whitehouse AT (2004) Changing precipitation patterns alter plant
community dynamics and succession in an ex-arable grassland.
Functional Ecology, 18, 648–655.
Moore DJP, Aref S, Ho RM, Pippen JS, Hamilton JG, De Lucia EM
(2006) Annual basal area increment and growth duration of
Pinus taeda in response to eight years of free-air carbon
dioxide enrichment. Global Change Biology, 12, 1367–1377.
Morales P, Sykes MT, Prentice IC et al. (2005) Comparing and
evaluating process-based ecosystem model predictions of carbon and water fluxes in major European forest biomes. Global
Change Biology, 13, 108–122.
Nakicenovic N, Alcamo J, Davis G et al. (2000) IPCC Special
Report on Emission Scenarios. Cambridge University Press,
Cambridge, UK.
Neilson RP (1993) Vegetation redistribution: a possible biosphere
source of CO2 during climatic change. Water, Air and Soil
Pollution, 70, 659–673.
Nepstad DC, Moutinho P, Dias-Filho MB et al. (2002) The effects
of partial throughfall exclusion on canopy processes, aboveground production, and biogeochemistry of an Amazon forest.
Journal of Geophysical Research, 107, D20, 8085, doi: 10.1029/
2001JD000360.
Nepstad DC, Verissimo A, Alencar A et al. (1999) Large-scale
impoverishment of Amazonian forests by logging and fire.
Nature, 398, 505–508.
New MG, Hulme M, Jones PD (2000) Representing twentiethcentury space-climate variability, Part II, Development of
1901–1996 monthly grids of terrestrial surface climate. Journal
of Climate, 13, 2217–2238.
Norby RJ, DeLucia EH, Gielen B et al. (2005) Forest response to
elevated CO2 is conserved across a broad range of productivity. Proceedings of the National Academy of Sciences of the United
States of America, 102, 18052–18056.
Oren R, Ellsworth DS, Johnsen KH et al. (2001) Soil fertility limits
carbon sequestration by forest ecosystems in a CO2-enriched
atmosphere. Nature, 411, 469–472.
Parton WJ, Scurlock JMO, Ojima DS et al. (1993) Observations
and modeling of biomass and soil organic matter dynamics for
the grassland biome worldwide. Global Biogeochemical Cycles, 7,
785–809.
Parton W, Stewart J, Cole C (1988) Dynamics of C, N, P, and S in
grassland soil: a model. Biogeochemistry, 5, 109–131.
Petoukhov V, Ganopolski A, Brovkin V, Claussen M, Eliseev A,
Kubatzki C, Rahmstorf S (2000) CLIMBER-2: a climate system
model of intermediate complexity: I. Model description and
performance for present climate. Climate Dynamics, 16, 1–17.
Peylin P, Bousquet P, Le Quéré C et al. (2005) Multiple constraints
on regional CO2 flux variations over land and oceans.
Global Biogeochemical Cycles, 19, GB1011, doi: 10.1029/
2003GB002214.
Prentice IC, Bondeau A, Cramer W et al. (2007) In: Terrestrial
Ecosystems in a Changing World. IGBP Book Series (eds Canadell
J, Pitelka L, Pataki D), pp. 336. Springer, Heidelberg, Germany.
Prentice IC, Farquhar GD, Fasham MJR et al. (2001) The carbon
cycle and atmospheric carbon dioxide. In: Climatic Change
2001: The Scientific Basis. Contribution of Working Group I to the
Third Assessment Report of the Intergovernmental Panel on Climate
Change, pp. 185–225. Cambridge University Press, Cambridge.
Rayner NA, Parker DE, Frich P, Horton EB, Folland CK,
Alexander LV (2000) SST and sea-ice fields for ERA40. Proceedings of the second WCRP International Conference on Reanalyses. Wokefield Park, Reading, UK, World Climate Research
Program, pp. 18–21.
Saleska SR, Miller SD, Matross DM et al. (2003) Carbon in
Amazon forests: unexpected seasonal fluxes and disturbance-induced losses. Science, 302, 1554–1557.
Schaphoff S, Lucht W, Gerten D, Sitch S, Cramer W, Prentice IC
(2006) Terrestrial biosphere carbon storage under alternative
climate projections. Climatic Change, 74, 97–122.
Scheffer M, Brovkin V, Cox PM (2006) Positive feedback between
global warming and atmospheric CO2 concentration inferred
from past climate change. Geophysical Research Letters, 33,
L10702, doi: 10.1029/2005GL025044.
Scholze M, Knorr W, Arnell NW, Prentice IC (2006) A climatechange risk analysis for world ecosystems. Proceedings of the
National Academy of Sciences of the United States of America, 103,
13116–13120.
Silapaswan CS, Verbyla DL, McGuire AD (2001) Land cover
change on the Seward Peninsula: the use of remote sensing to
evaluate the potential influences of climate warming on historical vegetation dynamics. Canadian Journal of Remote Sensing, 27, 542–554.
Sitch S, Brovkin V, von Bloh W et al. (2005) Impacts of future land
cover changes on atmospheric CO2 and climate. Global Biogeochemical Cycles, 19, GB2013, doi: 10.1029/2004GB002311.
Sitch S, McGuire AD, Kimball J et al. (2007) Assessing the carbon
balance of circumpolar Arctic tundra using remote sensing
and process modelling. Ecological Applications, 17, 213–234.
Sitch S, Smith B, Prentice IC et al. (2003) Evaluation of ecosystem
dynamics, plant geography and terrestrial carbon cycling in
the LPJ dynamic vegetation model. Global Change Biology, 9,
161–185.
Smith TM, Shugart HH (1993) The transient response of terrestrial carbon storage to a perturbed climate. Nature, 361,
523–526.
Stewart JB (1988) Modelling surface conductance of pine forest.
Agricultural and Forest Meteorology, 43, 19–35.
Stow DA, Hope A, McGuire AD et al. (2004) Remote sensing of
vegetation and land- cover change in Arctic tundra ecosystems. Remote Sensing of Environment, 89, 281–308.
Sturm M, Racine C, Tape K (2001) Increasing shrub abundance in
the Arctic. Nature, 411, 546–547.
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x
U N C E R TA I N T Y I N L A N D C A R B O N C Y C L E F E E D B A C K S
Takahashi T, Olafsson J, Goddard JG, Chipman D, Sutherland S
(1993) Seasonal-variation of CO2 and nutrients in the highlatitude surface oceans – a comparative study. Global Biogeochemical Cycles, 7, 843–878.
Tian H, Melillo JM, Kicklighter DW, McGuire AD, Helfrich JVK,
Moore B, Vorosmarty CJ (1998) Effect of interannual climate
variability on carbon storage in Amazonian ecosystems. Nature, 396, 664–667.
Van Nieuwstadt MGL, Sheil D (2005) Drought, fire and tree survival
in a Borneo rain forest, East Kalimantan, Indonesia. Journal of
Ecology, 93, 191–201, doi: 10.1111/j.1365-2745.2004.00954.x.
VEMAP Member (1995) Vegetation/ecosystem modeling and
analysis project: comparing biogeography and biogeochemis-
25
try models in a continental-scale study of terrestrial ecosystem
responses to climate change and CO2 doubling. Global Biogeochemical Cycles, 9, 407–437.
Williamson GB, Laurance WF, Oliveria AA, Gascon C,
Lovejoy TE, Pohl L (2000) Amazonia tree mortality
during the 1997 El Niño drought. Conservation Biology, 14,
1538–1542.
Woodward FI, Lomas MR (2004) Vegetation-dynamics –
simulating responses to climate change. Biological Reviews, 79,
643–670.
Woodward FI, Smith TM, Emanuel WR (1995) A global land
primary productivity and phytogeography model. Global Biogeochemical Cycles, 9, 471–490.
r 2008 The Authors
Journal compilation r 2008 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2008.01626.x