JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116, D09108, doi:10.1029/2010JD015224, 2011
Diagnosing the equilibrium state of a coupled global biosphere‐
atmosphere model
Shanshan Sun1 and Guiling Wang1
Received 22 October 2010; revised 11 February 2011; accepted 21 February 2011; published 12 May 2011.
[1] This study combines a conceptual modeling approach and a physical process–based
modeling approach to diagnose and understand the equilibrium state(s) of coupled
biosphere‐atmosphere models using the NCAR CAM3‐CLM3‐DGVM model as an
example, with a focus on the West Africa and South America regions. First, a
conceptual model is parameterized according to results from offline simulations using
CAM3‐CLM3 and CLM3‐DGVM. The resulting conceptual model is then used to
predict the potential biosphere‐atmosphere equilibrium state(s) for the coupled CAM3‐
CLM3‐DGVM model. Finally, simulations using the coupled CAM3‐CLM3‐DGVM
model itself are carried out to verify the results of the conceptual model. Only one
equilibrium state is found in both the conceptual model and the physically based numerical
model. The CLM3 model features excessively high soil evaporation and very low plant
transpiration, which leads to high bare‐ground ET and low sensitivity of ET to vegetation
cover changes in the land model. Despite the high sensitivity of precipitation to
evapotranspiration changes in the atmospheric model, the land model deficiency leads to a
high amount of precipitation in the absence of vegetation cover and a low sensitivity of
precipitation to vegetation changes in CAM3‐CLM3, two conditions that are found to
anchor a coupled biosphere‐atmosphere model to a single equilibrium state. This statement
also holds for the up‐to‐date version of the models using CLM4. This study provides an
example for the importance of land surface processes in determining the potential
equilibrium states of fully coupled biosphere‐atmosphere models.
Citation: Sun, S., and G. Wang (2011), Diagnosing the equilibrium state of a coupled global biosphere‐atmosphere model,
J. Geophys. Res., 116, D09108, doi:10.1029/2010JD015224.
1. Introduction
[2] Past numerical modeling studies have identified terrestrial biosphere‐atmosphere interaction as one mechanism
that can lead to multiple equilibrium states in the climate
system [e.g., Charney, 1975; Claussen, 1998; Wang and
Eltahir, 2000a; Oyama and Nobre, 2003]. Under current
climate conditions, these modeling studies have identified the
Sahel and the Amazon border areas as potential regions possessing multiple equilibrium states attributable to biosphere‐
atmosphere interactions. In these regions, vegetation growth is
limited by water availability; and vegetation cover changes can
exert significant impact on land‐atmosphere interactions and
modify local climate conditions. Specifically, by increasing
surface albedo and Bowen ratio and decreasing surface
roughness, desertification or deforestation could reduce local
precipitation through their impact on surface water and heat
fluxes, net radiation, and large‐scale circulation as well. The
1
Department of Civil and Environmental Engineering and Center for
Environmental Sciences and Engineering, University of Connecticut,
Storrs, Connecticut, USA.
Copyright 2011 by the American Geophysical Union.
0148‐0227/11/2010JD015224
reduction in precipitation may cause further degradation of
vegetation, leading to a positive feedback. This positive feedback between vegetation and precipitation might give rise to
long‐lasting drought and push the land‐atmosphere system
into a drier equilibrium state.
[3] In the Sahel region and Amazon borders area, simulated precipitation is found to be very sensitive to prescribed
changes in vegetation cover [e.g., Claussen, 1998; Oyama
and Nobre, 2004], a climate characteristic of regions
prone to multiequilibrium states [Wang, 2004; Liu et al.,
2006]. Consistently, multiequilibrium states have been
found to exist in these regions in various climate models,
including models with asynchronous vegetation‐climate
coupling [Claussen, 1997, 1998; Oyama and Nobre, 2003]
and models with synchronous vegetation‐climate coupling
[Wang and Eltahir, 2000a, 2000b; Zeng and Neelin, 2000;
Zeng et al., 2002; Wei et al., 2006].
[4] In addition to physically based numerical climate
models, conceptual models with as few as two variables
(one representing the atmosphere and the other the biosphere) have also been developed and used to study the
multiequilibrium behaviors resulting from biosphere‐
atmosphere interactions [Svirezhev and Bloh, 1996; Brovkin
et al., 1998; Zeng et al., 1999; Da Silveira Lobo Sternberg,
D09108
1 of 14
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
2001; Wang, 2004]. Conceptual models’ advantages lie in
their highly simplified representation of physical mechanisms, which makes results easier to interpret. In a conceptual
model, depending on whether temperature, radiation, or
water supply is the limiting factor for vegetation growth, the
atmospheric state variable can be thermal, radiative, or
hydrological. Although radiation is the limiting factor for
the interior Amazon [e.g., Myneni et al., 2007], water is the
main limiting factor for vegetation growth in the Sahel and
Amazon boundary regions. Precipitation (P) is therefore
used as the atmosphere state variable in conceptual models
pertaining to these regions [Brovkin et al., 1998; Zeng et al.,
1999; Da Silveira Lobo Sternberg, 2001; Wang, 2004].
Using vegetation density or leaf area index (LAI) as the
biosphere variable in conceptual models, precipitation’s
response to vegetation cover changes is represented by a
P*(V) curve and vegetation’s response to precipitation is
represented by a V*(P) curve. The climate equilibrium
states can be identified by the intersections between these
two curves. Regions with multiple equilibrium states are
characterized by a high sensitivity of precipitation to local
vegetation cover changes and a high sensitivity of vegetation to precipitation changes.
[5] Combining the numerical modeling approach with the
conceptual modeling approach, Wang [2004] parameterized
the P*(V) and V*(P) curves of a conceptual model using
outputs from a physically based stand‐alone zonally symmetrical atmospheric model and a dynamic vegetation
model. The so defined conceptual model is then used to
interpret and predict behaviors of the physically based,
coupled zonally symmetrical model. Specifically, to derive
the P*(V) curve, the atmospheric model was driven with
different prescribed vegetation covers to simulate the
response of precipitation to vegetation changes. To derive
the V*(P) curve, the land model with dynamic vegetation
was driven with different prescribed atmospheric forcings to
simulate the response of vegetation to precipitation changes.
The resulting conceptual model predicts two stable equilibrium states in the Sahel region, the wet one corresponding
to the current condition and the arid one corresponding to a
much drier and less productive condition. These two equilibrium states correspond extremely well with the two
equilibrium states derived from the physically based coupled dynamic vegetation‐atmosphere model [Wang and
Eltahir, 2000a; Wang, 2004].
[6] The Wang [2004] study demonstrated the usefulness
of conceptual modeling as a tool to diagnose and understand
behaviors of complex physically based models. However,
the zonally symmetrical atmospheric model used by Wang
[2004] was characterized by a very small magnitude of
atmospheric interannual variability due to the lack of zonal
wave disturbance and other simplifications. It remains
questionable whether the same approach can be applicable
to complex, commonly used global climate models, the
behaviors of which are much more difficult to understand
and to predict than those of a simplified model such as the
one used by Wang [2004]. Using a coupled atmosphere‐
land‐dynamic vegetation model as an example, this study
applies the Wang [2004] to a GCM and investigates how
land surface processes may play a role in determining the
number and state(s) of regional climate equilibrium, a topic
D09108
that is critical for understanding and predicting runaway
behaviors in the climate system.
2. Model Description
[7] The models used in this study include the conceptual
model of Wang [2004] and several physically based
numerical models, including the National Center for Atmospheric Research (NCAR) Community Atmosphere Model
version 3 (CAM3) coupled with the Community Land Model
version 3 (CLM3) including the Dynamic Global Vegetation
Model (DGVM) (CAM3‐CLM3‐DGVM) and its component
models (i.e., CAM3‐CLM3 with static vegetation and
CLM3‐DGVM). In addition, an updated model version
CAM4‐CLM4 is also used to test the sensitivity of our results
to recent implemented model parameterization changes.
2.1. The Conceptual Model
[8] The two‐variable conceptual model includes four
basic equations. The first two describe the dynamics of the
biosphere (dV/dt = (V* − V)/t) and the stochastic nature of
the atmosphere (P = P* + sR(t)). Here t is time, t is the
characteristic time scale of vegetation variation. The two
variables of the model are vegetation V as the biosphere
state and precipitation P as the atmospheric state variable.
V* stands for the vegetation state in equilibrium with precipitation P; P* stands for the precipitation amount in
equilibrium with vegetation V. Precipitation P consists of
two parts, P* as determined by vegetation and a stochastic
component P′ = sR(t), in which s is the standard deviation
of precipitation caused by atmospheric internal variability
and the random time series R(t) follows standard normal
distribution.
[9] The other two equations of the conceptual model
define V* and P* used in the first two equations. They
represent the response of equilibrium vegetation to precipitation changes (the V* (P) curve) and the response of
equilibrium precipitation to vegetation changes (the P* (V)
curve), respectively. While analytical forms of the two
equations can be used in theoretical studies on general
biosphere‐atmosphere interactions, in this study the two
equations will be derived based on output from the
numerical models CAM3‐CLM3 and CLM3‐DGVM as the
conceptual model is developed to describe the behaviors of
the coupled model CAM3‐CLM3‐DGVM. While the first two
equations are needed to describe climate variability, we will be
primarily dealing with the P* (V) and V* (P) relationships in
this study as the focus here is on diagnosing the equilibrium
state(s) of the model.
2.2. CAM3‐CLM3‐DGVM and Its Component Models
[10] The coupled biosphere‐atmosphere model CAM3‐
CLM3‐DGVM model is used to derive the equilibrium state
(s) of the biosphere‐atmosphere system in different regions.
The model simulates the dynamic land‐atmosphere system
including vegetation dynamics under the present‐day solar
irradiance and monthly varying climatological sea surface
temperature (SSTs) [Hurrell et al., 2008]. At each time step,
CAM3 simulates the atmospheric processes and provides
atmospheric forcing to CLM3‐DGVM; driven with atmospheric forcing from CAM3, CLM3‐DGVM simulates the
land surface biogeophysical, physiological, biogeochemical
2 of 14
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
processes and ecosystem dynamics, providing surface fluxes
(sensible heat flux, latent heat flux (or evapotranspiration),
and momentum flux) to CAM3 and updating the vegetation
structure and distribution. For CAM3, the Eulerian spectral
dynamical core is chosen with a T42 horizontal resolution
and a total of 26 levels in the vertical direction. Details of
the CAM3 model are provided by Collins et al. [2004], and
the description for a more recent version CAM4 is provided
by Neale et al. [2010].
[11] The component model CAM3‐CLM3 (or CAM4‐
CLM4) simulates the land‐atmosphere system while specifying a static vegetation condition. It is used in this study to
derive the precipitation amount corresponding to different
levels of prescribed vegetation conditions. Its land surface
scheme CLM3 simulates the land surface biogeophysical
and physiological processes. However, without DGVM,
CAM3‐CLM3 does not include representation of biogeochemical processes and ecosystem dynamics.
[12] The component model CLM3‐DGVM is used to
derive the vegetation structure and distribution corresponding to a given atmospheric forcing. In addition to the land
surface biogeophysical and physiological processes simulated in the land model, CLM3‐DGVM also simulates the
biogeochemical processes and dynamics of natural vegetation,
including growth, mortality, and competition [Levis et al.,
2004]. The dynamic vegetation model is modified from the
Lund‐Potsdam‐Jena (LPJ) DGVM [Sitch et al., 2003]. Biogeophysical and biogeochemical processes are simulated with
a 20 min time step while plant phenology is evaluated daily.
Vegetation structure and distribution are updated yearly based
on the annual accumulation of carbon for each individual
PFT. The following description focuses on land surface
hydrological processes because they significantly influence
the behavior of the coupled CAM3‐CLM3‐DGVM model
pertaining to the focus of this study, as shown in section 3.
[13] The land surface model CLM3 [Dai et al., 2003;
Oleson et al., 2004] represents the land surface with ten soil
layers, up to five snow layers and one vegetation layer. Land
surface heterogeneity is characterized by allocating four
types of patches in each grid cell (glacier, lake, wetland and
vegetated). The vegetated patch may include some bare soil.
Natural vegetation is represented as combinations of different plant function types, and ten plant functional types
(PFTs) are considered: tropical broadleaf evergreen and
deciduous trees, temperate needleleaf evergreen trees, temperate broadleaf evergreen and deciduous trees, boreal
needleleaf evergreen and boreal deciduous trees, C4, C3 and
C3 arctic grasses. In the presence of vegetation, the model
evapotranspiration (ET) consists of three components, the
soil evaporation, vegetation interception loss and transpiration, each of which is estimated using the bulk aerodynamic
equation. Soil moisture in the top soil layer (1.75 cm) is
used to scale the moisture availability for soil evaporation.
Both precipitation characteristics and vegetation density (as
quantified by leaf and stem area index) influence the
amount of precipitation intercepted by vegetation, which
limits the interception loss through its influence on the area
of canopy surface that is wet. Transpiration is regulated by
leaf stomatal conductance, which varies with solar radiation
and rooting zone soil moisture through their impact on
photosynthesis.
D09108
[14] The improper partitioning of ET among its three
components, including overestimation of soil evaporation
and interception loss and underestimation of plant transpiration, is a known problem in CLM3 [e.g., Bonan and Levis,
2006; Lawrence et al., 2007; Wang and Wang, 2007;
Oleson et al., 2008; Lawrence and Chase, 2009]. More
recent versions of the model, CLM3.5 and CLM4, include
modifications of various model parameterizations to address
this problem [e.g., Oleson et al., 2007, 2010]. Among those
considered are a soil resistance term or a litter resistance
term to slow down soil evaporation, a more rapid drop of
soil evaporation with top layer soil wetness, and a reduced
fraction of precipitation that can be intercepted by vegetation canopy. These parameterization changes have led to a
certain degree of model performance improvement in simulating the surface hydrological budget when tested in offline mode [e.g., Oleson et al., 2008; Lawrence et al., 2011].
However, it is not clear how they may influence the model
behaviors pertaining to biosphere‐atmosphere equilibrium.
3. Methodology and Experimental Design
3.1. Main Experiments and Methodology
[15] To derive the equilibrium state(s) of the biosphere‐
atmosphere system using the physically based numerical
model, the coupled model CAM3‐CLM3‐DGVM is run for
200 years. At the end of each 200 year simulation, although
carbon accumulation is not complete, a vegetative equilibrium is obtained. To evaluate the potential existence of
multiple equilibrium states of the coupled model, two different initial vegetation conditions are experimented on, one
with bare ground and the other with forest everywhere. In
the forest initial condition run, vegetation is initialized to be
tropical broadleaf evergreen trees in the tropics and temperate broadleaf cold deciduous trees in the extratropics.
[16] Components of the physically based numerical model
are used to parameterize the conceptual model of biosphere‐
atmosphere system. In order to parameterize the P*(V)
curve in the conceptual model, CAM3‐CLM3 is used to
simulate precipitation under different prescribed vegetation
covers. In the P*(V)_global experiments, the land surface
files are taken directly from the outputs of the 200 years
CAM3‐CLM3‐DGVM run starting from bare ground, representing different stages of global vegetation distribution (which
ranges from bare ground to grass dominance or to tree dominance where climate is suitable for trees to survive). To exclude
the possible nonlocal impacts introduced by global vegetation
cover changes, companion experiments for the P*(V) curve
for two regions (P*(V)_Sahel and P*(V)_SE‐Amz) are also
carried out, in which only vegetation cover in the region of
interest ((15°N–20°N, 16°W–11°E) for the Sahel, and (18°S–
7°S, 47°W–36°W) for the SE Amazon) is modified while
vegetation elsewhere is kept the same as the equilibrium state
derived from the coupled CAM3‐CLM3‐DGVM simulation.
Although these differences in the way vegetation is specified
led to some subtle differences in local vegetation between
the global and regional experiments, the difference is by all
means very small. So the difference between the P*(V) curves
based on the P*(V)_global experiments and those based on
the regional experiments (P*(V)_Sahel or P*(V)_SE‐Amz)
is primarily due to the impact of nonlocal vegetation cover
changes on the regional climates.
3 of 14
D09108
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
Table 1. Summary of Experiments
Model Components
Experiments
Category
Fully coupled
Experiment
DGVM_offline
Experiment
Atmosphere
Land
Vegetation
Experiment
Length
Number of
Simulations
CAM3
CLM3
DGVM
200 years
One
Prescribed atmospheric
forcings; Different
among simulations
CAM3
CLM3
DGVM
200 years
Six
CLM3
20 years
nine
CAM3
CLM3
20 years
six
P*(V)_SE‐Amz
Experiment
CAM3
CLM3
20 years
six
ET_overwrite
Experiment
CAM3, with ET prescribed
instead of simulated
by CLM3
CAM4
CLM3
Static global vegetation covers,
different among simulations
Different static Sahel vegetation
covers among simulations. Other
regions share the same static
vegetation covers
Different static SE‐Amz vegetation
covers among simulations. Other
regions share the same static
vegetation covers
Same static vegetation cover globally
among simulations
20 years
six
CLM4
Default vegetation cover globally
20 years
one
CAM4
CLM4
20 years
one
CAM4
CLM4
Default vegetation cover with LAI
halved globally
Bare ground globally
20 years
one
P*(V)_global
Experiment
P*(V)_Sahel
Experiment
CAM4‐CLM4_ctrl
Experiment
CAM4‐CLM4_05lai
Experiment
CAM4‐CLM4_bare
Experiment
[17] In order to parameterize the V*(P) curve in the
conceptual model, CLM3‐DGVM model is used in
DGVM_offline experiments to simulate the vegetation
structure and distribution under different atmospheric forcings. The hourly atmospheric forcing including surface
temperature, wind, specific humidity, pressure, precipitation
and incident solar radiation is derived from the coupled
CAM3‐CLM3 model outputs with data of each year treated
as a different realization of the mean climate. CLM3‐
DGVM will cycle through 1 year’s atmospheric forcing data
multiple times until a vegetative equilibrium state is
reached. Each experiment is run for 200 years to ensure that
vegetation distribution has come to equilibrium states in the
Sahel and the Amazon southeastern boundary regions where
C4 grass is the dominant vegetation species. Therefore the
model will cycle through 1 year’s atmospheric forcing 200
times to simulate the corresponding vegetation distribution
under this specific climate condition.
[18] Based on the intersection(s) between the P*(V) and
the V*(P) curves, the conceptual model predicts the equilibrium state(s) for the CAM3‐CLM3‐DGVM model.
Comparison with the equilibrium state(s) derived from
CAM3‐CLM3‐DGVM itself will then provide validation for
the conceptual model and for the applicability of the conceptual model in interpreting behaviors of the physically
based numerical model.
3.2. CAM4‐CLM4 Experiments
[19] As shown later in section 4, the lack of precipitation
sensitivity to vegetation changes in CAM3‐CLM3 is an
important characteristic that dominates the behavior of the
modeled biosphere‐atmosphere system. Extensive parameterization changes have been implemented in the latest
version of the NCAR model, and some of them are likely to
cause changes in the model precipitation sensitivity to
vegetation cover changes. To test this, three sensitivity
experiments using CAM4‐CLM4 are conducted, each
specifying a different level of vegetation density. With the
same prescribed SSTs forcing and land surface cover, the
only difference among these three experiments is that LAI
takes the default value in CAM4‐CLM4_ctrl, is reduced by
half in CAM4‐CLM4_05lai, and is set to zero in CAM4‐
CLM4_bare. Such vegetation differences are applied at the
global level.
3.3. ET_overwrite Experiments
[20] An important pathway for vegetation to influence
precipitation is through ET. The sensitivity of precipitation
to vegetation changes therefore depends on both the sensitivity of ET to vegetation changes and the sensitivity of precipitation to ET changes, with the former being determined
by the land surface model and the latter by the atmosphere
model. ET_overwrite experiments using CAM3‐CLM3 are
designed to test the precipitation sensitivity to ET changes,
where the CLM3‐simulated ET (and therefore latent heat)
flux is replaced by a specified flux when passed to CAM3.
First, in a 1 year simulation using CAM3‐CLM3‐DGVM
restarting from its equilibrium state, ET in every time step
of the whole year was written to a data file; using CAM3‐
CLM3 with vegetation specified according to the equilibrium vegetation, six experiments were carried out in which
the model read in the ET flux from the data file in every
time step and specified the ET in CAM3 to be 20%, 60%,
80%, 100%, 120% and 160% of the read‐in values, respectively. That means at each time step, evapotranspiration
provided to the atmosphere in CAM3 is prescribed, which
replaces the actual evapotranspiration predicted by the land
model CLM3. These six experiments are each 20 years long;
in the same experiment, ET is specified the same from year to
year with diurnal and seasonal cycles. The last ten model
4 of 14
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
D09108
Figure 1. Peak leaf area index distribution at the equilibrium state of the fully coupled CAM3‐CLM3‐
DGVM model, with the six study areas labeled using the open boxes.
years’ outputs are used to analyze precipitation’s response to
ET variations.
[21] To summarize, Table 1 lists all experiments used in
this study and how various parameters/parameterizations are
treated.
4. Results
[22] Experiments using the coupled CAM3‐CLM3‐
DGVM model initialized with drastically different initial
vegetation conditions, one with bare ground and one with
full forest cover, produced the same equilibrium state. The
global patterns of precipitation and vegetation distribution at
this equilibrium state are generally similar to the present‐day
observations. Our results analysis here focuses on two
broadly defined regions, West Africa and the Amazon
region, which are further divided as shown in Figure 1.
These two regions are chosen as their climates and circulation patterns are found to be sensitive to land surface
condition changes in previous studies [e.g., Xue and Shukla,
1993; Zhang et al., 1996; Xue et al., 2006; Patricola and
Cook, 2008; Alo and Wang, 2010], and two of the subareas, the Sahel area (15°N–20°N and 16°W–11°E) and the
Amazon southeastern area (SE‐Amz, 7°S–18°S and 47°W–
36°W), were found to be prone to multiple equilibrium
states attributable to vegetation‐climate interactions
[Claussen, 1998; Wang and Eltahir, 2000a; Zeng and
Neelin, 2000; Oyama and Nobre, 2003].
[23] Climate in West Africa is under the influence of the
West African monsoon circulation. The rainy season in
West Africa occurs during boreal summer, and its length
decreases away from the Guinea coast northward. Annual
precipitation and vegetation distribution in West Africa are
characterized by strong gradient in the latitudinal direction,
and the variation in the zonal direction is rather small.
Except for the coastal region, vegetation in most areas of
West Africa is limited by precipitation. On the contrary, in
most areas of the Amazon, precipitation is abundant and
does not show a strong spatial gradient, and vegetation is
more limited by light availability. However, in the Amazon
border areas, water is the limiting factor for vegetation
growth.
[24] Despite the overall agreement with observations in
the general patterns of vegetation and precipitation distribution, discrepancy between the CAM3‐CLM3‐DGVM
equilibrium and observations are found at the regional level.
Compared with the climatic research unit (CRU) data [New
et al., 2000], yearly precipitation is overestimated in West
Africa and underestimated in South America including the
Amazon. Correspondingly, in West Africa, the simulated C4
grass cover extends into the Sahara desert, while in the
Amazon region tropical broadleaf deciduous tree and C4
grass cover is overestimated at the expense of evergreen tree
coverage.
[25] In two of the subareas (Sahel and SE‐Amz) at the
equilibrium state of the coupled CAM3‐CLM3‐DGVM
model, vegetation is dominated by grass, similar to present‐
day observations in the SE‐Amz region but represents an
overestimation of grass coverage in the Sahel region. The
other four subareas are densely forested at the equilibrium
state of the fully coupled CAM3‐CLM3‐DGVM model,
which is also similar to present‐day observations. For all six
regions, spatial averages of peak leaf area index (PLAI) and
yearly precipitation are used as the state variables for the
biosphere and the atmosphere, respectively, in plotting the
P*(V) and V*(P) curves for the conceptual model. The use
of LAI as the state variable for the biosphere may present an
issue when the dominant vegetation type differs within the
precipitation range of the same V*(P) curve. For example,
trees with a PLAI of 2 would require much more precipitation than grass with a PLAI of 2. However, this has not
been a problem for the six subareas examined in this study.
In addition, biomass was also tested as an alternative bio-
5 of 14
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
D09108
Figure 2. Conceptual models for the (a, c) Sahel region and (b, d) SE‐Amz region. Figures 2a and 2b are
based on P*(V)_global experiments; Figures 2c and 2d are based on P*(V)_Sahel experiments and
P*(V)_SE‐Amz experiments. Corresponding to each specific PLAI value, the five red circles represent annual
precipitation in the last 5 years of the P*(V)_global experiment in Figures 2a and 2b, the P*(V)_Sahel experiment in Figure 2c, and the P*(V)_SE‐Amz experiment in Figure 2d, with the 5 year average shown by the
black stars. Red line is the linear least square fit of red circles. Corresponding to each specific precipitation
value, the blue triangle represents PLAI averaged over the last 10 years each DGVM_offline experiment.
Blue line is the linear least square fit of blue triangles. The magenta solid circle in each panel represents
the annual precipitation and PLAI averaged over the last 30 years of the coupled CAM3‐CLM3‐DGVM
simulation.
sphere variable and the results are consistent with results
based on PLAI.
[26] As shown in Figure 2, for both the Sahel region and
the SE‐Amz region, the P*(V) and V*(P) curves have only
one intersection, indicating that the conceptual models
predict only one equilibrium state for the coupled CAM3‐
CLM3‐DGVM model. This is consistent with results of the
CAM3‐CLM3‐DGVM experiments. As a comparison, the
magenta circles in Figure 2 show the equilibrium state
derived directly from CAM3‐CLM3‐DGVM itself. In both
regions, the equilibrium state predicted by the conceptual
model corresponds very well to the equilibrium derived
from the coupled CAM3‐CLM3‐DGVM model. This
agreement validates the methodology of parameterizing
conceptual models based on GCM outputs. The constructed
conceptual model can therefore provide useful indications of
GCM behaviors.
[27] As shown in Figure 2, in the P*(V)_global experiments in which vegetation covers vary globally, the multiyear average of precipitation (black asterisks) increases
slightly and then fluctuates. Such precipitation changes
however are not significant, and are much smaller in magnitude than the interannual variability of precipitation
(Figure 2). The precipitation response to the LAI changes is
therefore considered negligible in the Sahel and SE‐Amz
regions. When vegetation is modified in the specific region
only (Figure 2c for Sahel region and Figure 2d for SE‐Amz
region as examples), precipitation shows a similar lack of
response to vegetation cover changes. For both regions,
P*(V) curves are almost the same between experiments with
globally modified vegetation and those with regionally
6 of 14
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
modified vegetation. The number of intersection between
P*(V) and V*(P) curves remains one, and the location of
this intersection shifts only slightly. Note that while vegetation cover changes do influence nonlocal precipitation,
such impact is in general small compared with the impact
on local precipitation.
[28] As indicated by Wang [2004], for a given V*(P)
relationship, two factors influence the number of equilibrium solutions of the conceptual model: precipitation
amount in the absence of vegetation and sensitivity of precipitation to vegetation changes. Specifically, low precipitation in the absence of vegetation (lower than what it takes
to grow any vegetation) combined with low sensitivity of
precipitation to land cover changes tends to favor one arid
equilibrium state; low precipitation in the absence of vegetation combined with medium‐to‐high sensitivity of precipitation to land cover changes tends to favor two stable
equilibrium states, an arid one and a wet one; high precipitation in the absence of vegetation (higher than what it takes
for some grass to survive) favors one wet equilibrium state
regardless of how sensitive the precipitation is to land cover
changes. In Figure 2 for both regions, the precipitation
amount in the absence of vegetation is very high, higher
than what it takes for grass to survive, and the yearly precipitation shows very low sensitivity to vegetation changes.
Compared with the P*(V) curves in previous studies that
showed multiequilibrium states in the Sahel region [Brovkin
et al., 1998; Wang, 2004], CAM3‐CLM3 seems to overestimate precipitation under bare‐ground condition and
underestimate the sensitivity of precipitation to vegetation
cover changes. As explained above, when precipitation in
the absence of vegetation is high, only one stable equilibrium state can exist regardless of precipitation’s sensitivity
to land surface changes. This same reason is also responsible for the lack of multiequilibrium states in the SE‐Amz
region in the coupled model.
[29] Note that the lack of interannual variability in the
atmospheric forcing used for the V*(P) experiments may
have led to some overestimation of vegetation coverage and
potentially LAI in semiarid regions [Notaro, 2008]. While
this may slightly influence the state of the equilibrium
predicted by the conceptual model, it should have no impact
on the number of equilibrium state(s) as the latter is so
overwhelmingly determined by the location and shape of the
P*(V) curve.
[30] To understand the low sensitivity of precipitation to
vegetation changes in CAM3‐CLM3, we examine how the
most relevant hydrological variables and radiative variables
(averaged for each region) vary with PLAI in the P*(V)
experiments (Figure 3). In the Sahel region (first through
third rows in Figure 3), precipitation increases as PLAI
increases from zero to 0.8 and then fluctuates as PLAI
continues to increase. Evapotranspiration (ET) responds
similarly, leading to cooling as PLAI increases. Atmospheric moisture convergence decreases, probably as a result
of cooler surface favoring divergence. ET is much larger in
magnitude than the atmospheric moisture convergence,
indicating that local evapotranspiration’s impact on precipitation is dominant. Surface runoff decreases slightly as
PLAI increases from zero to 0.8 and does not vary much as
PLAI further increases. Subsurface drainage responds similarly. Soil moisture in the rooting zone (e.g., in the top 7
D09108
soil layers totaling 80 cm for C4 grass) decreases as PLAI
increases from zero to 0.8 as a result of plant uptake, and
shows no detectable decrease as PLAI further increases.
[31] For the three components of evapotranspiration, as
PLAI increases from zero to 0.8, interception loss increases
and vegetation transpiration increases while soil evaporation
decreases. Similar to other variables, the impact of vegetation seems to saturate at PLAI equal to 0.8. The response of
ET partition to vegetation changes is qualitatively consistent
with our expectations. Quantitatively however, compared
with data from Dirmeyer et al. [2006] (which is an observation‐driven multimodel product), the model’s water budget is problematic. Contribution to total evapotranspiration
from vegetation transpiration is much smaller than those
from soil evaporation and/or interception loss. Under well‐
vegetated conditions, soil evaporation is comparable with
that of interception loss and both are about twice the magnitude of vegetation transpiration. This improper water
budget can be caused by overestimation of interception loss,
ground evaporation and underestimation of plant transpiration, all of which are well documented in previous studies
[e.g., Bonan and Levis, 2006; Lawrence et al., 2007; Oleson
et al., 2008]. The underestimation of vegetation transpiration and the low sensitivity of transpiration to PLAI changes
especially when PLAI exceeds a certain threshold, combined with the overestimation of soil evaporation and
interception loss, downplay the impact of vegetation on the
hydrological process and reduces precipitation’s sensitivity
to vegetation cover changes.
[32] In the SE‐Amz region, compared with the Sahel
region, the response of various hydrological and thermal
variables to vegetation changes seems to saturate at a different level, at about PLAI equal to 2 (fourth through sixth
rows of Figure 3). The responses of most variables to LAI
changes in this region follow a pattern similar to those in the
Sahel, with two exceptions. First, the sensitivity of precipitation and total ET to LAI changes is even lower, with
negligible changes from bare soil to vegetated land. Second,
opposite to the Sahel region, albedo in the SE‐Amz region
increases with the increase of PLAI. Land surface albedo is
a combination of vegetation albedo and soil albedo, and the
latter depends on soil color and the top‐layer soil moisture.
As explained by Mei and Wang [2010], in the Amazon
region soil albedo is lower than grass albedo due to the dark
soil color and high soil moisture content. In addition to the
high evaporation from bare ground, this increase of albedo
with LAI further downplays the sensitivity of precipitation
to LAI changes in the SE‐Amz region.
[33] In both regions, the model soil evaporation from bare
ground is at a magnitude similar to the total ET over well
vegetated land. This high soil evaporation leads to large
amount of precipitation in the absence of vegetation, which
favors one wet stable equilibrium state regardless of how
sensitive precipitation might be to vegetation cover changes.
Wallace et al. [1990] compared ET measured over bare
ground and over fallow bush land in semiarid regions in
West Africa, and found that ET over bare ground is much
lower than over bush land except for during the hours
immediately after a rain event. In the more recent versions
of the Community Land model (CLM3.5 and CLM4),
hydrological parameterization changes led to improved ET
partitioning [e.g., Lawrence et al., 2007; Oleson et al., 2008;
7 of 14
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
D09108
Figure 3. Surface variables in the Sahel region (first, second, and third rows) and the SE‐Amz region
(fourth, fifth, and sixth rows) in the P*(V)_global experiments (blue triangles) and P*(V)_Sahel experiments and P*(V)_SE‐Amz experiments (red circles).
Stöckli et al., 2008; Lawrence et al., 2011]. However, even
in CLM4 the ET contrast between bare ground and vegetated land is still low. As a result, the model precipitation
in the absence of vegetation is still high in CAM4‐CLM4,
and precipitation sensitivity to LAI changes is still low
(Figure 6).
[34] The companion P*(V)_Sahel and P*(V)_SE‐Amz
experiments (results shown in red in Figure 3), in which
only the vegetation in the Sahel region or the SE‐Amz
region is modified through gradually increasing PLAI from
zero while the rest of the globe shares the same vegetation
distribution, are designed to exclude potential impacts on
8 of 14
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
D09108
Figure 4. Conceptual models for the (a) Guinean Coast region, (b) C – Amz region, (c) NW – Amz
region, and (d) NE – Amz region, with the P*(V) curve based on the P*(V)_global experiments. Legend
is the same as in Figure 2.
precipitation from nonlocal vegetation changes. In the
P*(V)_Sahel and P*(V)_SE‐Amz experiments, vegetation’s
impacts on radiative processes and hydrological processes in
the Sahel region and the SE‐Amz region are very similar to
those in P*(V)_global experiments.
[35] The same approach was applied to the four other areas
in the Amazon region and West Africa region, including a
densely vegetated inland area in central Amazon (C‐Amz,
4°S–9°S, 59°W–50°W), and other three tropical areas:
NW‐Amz (4°S–9°N, 81°W–61°W), NE‐Amz (4°S–9°N,
64°W–47°W) and Guinean Coast (4°N–9°N, 16°W–11°E).
Same as Sahel and SE‐Amz regions, C‐Amz, NE‐Amz,
NW‐Amz and Guinean Coast regions have only one
equilibrium state derived from the conceptual model,
which corresponds well to equilibrium state in the coupled
CAM3‐CLM3‐DGVM models (magenta dots in Figure 4).
In these regions too, the regional climate is anchored to a
single equilibrium state due to the excessive precipitation
in the absence of vegetation and low sensitivity of precipitation to vegetation changes, characteristics that result
from overestimation of bare soil evaporation and a partially
related underestimation of sensitivity of evapotranspiration
to vegetation changes (Figure 5).
[36] To test whether precipitation’s sensitivity to vegetation cover changes has improved in the latest version of the
models, results from CAM4‐CLM4 are shown in Figure 6
for all six regions. Compared with results in Figure 2, the
definition of the Sahel region as shown in Figure 6 is
extended southward by a few degrees in the CAM4‐CLM4
experiments to include more vegetated land. The P*(V)
curves show that precipitation increases slightly as LAI
increased from zero to default values in all six regions, with
a precipitation sensitivity slightly higher than that shown in
Figure 2 and Figure 4 for CAM3‐CLM3. However, the
difference in the average annual precipitation among the
three experiments is still very small compared with the large
interannual variation of precipitation in the last 5 years of
each experiment. In the offline CLM4, ET also shows low
sensitivity to LAI changes. Therefore recently implemented
model parameterization changes in CAM4 and CLM4 do
not seem to completely address the lack of model precipitation sensitivity to vegetation changes. Compared with
CLM3, CLM4 is generally wetter, and therefore the minimum precipitation required for plant establishment is lower
in CLM4. V*(P) curves in the conceptual model will then
shift slightly to the left. With the lack of sensitivity of
9 of 14
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
D09108
Figure 5. ET versus PLAI for the (a) Guinean Coast region, (b) C – Amz region, (c) NW – Amz region,
and (d) NE – Amz region, based on averages over the last 5 years of the P*(V)_global experiments.
precipitation to vegetation changes, the two curves in the
conceptual model will therefore still have only one intersection point. Results related to biosphere‐atmosphere
equilibrium as documented in this study are therefore not
expected to qualitatively change if the experiments were
carried out with the recently released version4 of the NCAR
models.
5. Role of the Atmosphere Model
[37] An important pathway for vegetation to influence
precipitation is through ET. The sensitivity of precipitation
to vegetation changes therefore depends on both the sensitivity of ET to vegetation changes and the sensitivity of
precipitation to ET changes, with the former being determined by the land surface model and the latter by the
atmosphere model. Therefore, it remains unclear whether
the model deficiency related to surface water budget as
documented in section 4 is the only cause for the lack of
sensitivity of precipitation to vegetation changes, or whether
correcting such model deficiency would make a difference
in the number and/or state of the model equilibrium state(s).
[38] The ET_overwrite experiments, in which ET is prescribed at 20%, 60%, 100%, 120% and 160% of the control
values (see Table 1), are used to study precipitation’s sensitivity to ET changes. In Figure 7, the red line represents
the P*(ET) curve and the magenta line marks the Y = X
relation. In all six regions, precipitation is highly sensitive
to ET changes as ET increases from 20% to 160% of the
control ET, and this increase in most cases is more than
the increase of local moisture supply. Depending on the
hydrological regime, the response of moisture convergence
(as an indicator of large‐scale circulation) can be substantial.
With the exception of the Guinea Coast region, in five of the
six regions examined, the P*(ET) curves are approximately
parallel to the Y = X line when ET is below a certain
threshold, indicating that the moisture convergence (therefore large‐scale circulation) is not sensitive to ET changes
under very dry conditions. The precipitation changes
therefore mainly result from changes of local moisture
supply. As ET further increases beyond the threshold value,
the slope of the P*(ET) curves becomes larger than 1,
indicating that moisture convergence increases with ET.
Therefore, under wetter conditions, in addition to the
enhancement of local moisture supply, the increase of ET
also leads to enhanced moisture supply from surrounding
regions through changes in the large‐scale circulation. This
threshold value is at about 100% of the control ET for the
10 of 14
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
D09108
Figure 6. P*(V) curves for the (a) Sahel region, (b) SE – Amz region, (c) Guinean Coast region,
(d) C – Amz region, (e) NW – Amz region, and (f) NE – Amz region, based on CAM4‐CLM4_ctrl,
CAM4‐CLM4_0.5lai and CAM4‐CLM4_bare experiments. Legend is similar to that in Figure 2.
two dry regions (the Sahel and the SE‐Amz), and is at only
40–60% of the control ET for the three wetter regions (the
C‐Amz, NW‐Amz and NE‐Amz regions). For the Guinean
Coast region, the slope of the P*(ET) curve is less than 1 as
ET increases from 20% to 60% and then becomes larger
than 1 as ET further increases from 100% to 160%. So the
moisture convergence here decreases first and then increases
with the increase of ET.
[39] The red brackets on the x axis of Figure 7 mark the
range of ET in the P*(V)_global experiments (shown in
Figure 3 and Figure 5) in which LAI varies widely, from
zero to local vegetation growth capacities (1.6 for Sahel region;
around 5.0 for NW‐Amz, NE‐Amz, SE‐Amz, Guinean Coast
regions and around 7.0 for C‐Amz region). The low sensitivity of ET to vegetation changes combined with the high
sensitivity of precipitation to ET changes indicates that the
lack of precipitation response to vegetation changes and the
high amount of precipitation in the absence of vegetation in
the CAM model result primarily from the low sensitivity of
ET to vegetation changes in the land model. Therefore, if in
the land model the soil evaporation in the absence of vegetation is reduced and the sensitivity of ET to vegetation
changes is increased, the model precipitation will become
much more responsive to vegetation changes and the precipitation in the absence of vegetation will become much
lower than it currently is. Such changes may lead to a change
in the number of equilibrium states of the coupled biosphere‐
atmosphere system.
6. Conclusions and Discussion
[ 40 ] The current study follows the approach of Wang
[2004] to explore the potential for multiequilibrium states
in a coupled biosphere‐atmosphere model using the NCAR
CAM3‐CLM3‐DGVM model as an example. We parameterize the conceptual model of Wang [2004] using results
from CAM3, CLM3 and its DGVM model, where the
response of vegetation to precipitation changes is defined
using results from CLM3‐DGVM driven with different
atmospheric forcings (including precipitation) and the
response of precipitation to vegetation changes is defined
using results from CAM3‐CLM3 driven with prescribed
vegetation covers. The resulting conceptual model predicts
only one equilibrium state for each of six regions examined
(including the Sahel region and the SE‐Amz region where
two equilibrium states were found to exist in some models),
11 of 14
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
D09108
Figure 7. Precipitation.versus ET for the (a) Sahel region, (b) SE – Amz region, (c) Guinean Coast
region, (d) C – Amz region, (e) NW – Amz region, and (f) NE – Amz region, based on average over
the last 10 years of the ET_overwrite experiments. The blue asterisks mark results from the ET_overwrite
100% experiment, and the red brackets denote the ET ranges simulated in the P*(V)_global experiments
for each region.
which correspond well with the single equilibrium state for
each region derived from the fully coupled CAM3‐CLM3‐
DGVM model.
[41] Based on the conceptual model, the physical processes influencing the equilibrium state(s) of the coupled
system are analyzed to understand why the CAM3‐CLM3‐
DGVM model simulates only one equilibrium state in the
Sahel and SE‐Amz regions, and in several other regions as
well. Different from some previous modeling studies [Brovkin
et al., 1998; Zeng et al., 2002; Wang, 2004], precipitation in
CAM3‐CLM3 is very high under bare‐ground condition and
shows little sensitivity to vegetation cover changes, which
work together to anchor the climate system to a single equilibrium state. The large magnitude of precipitation in the
absence of vegetation and the low precipitation sensitivity
result from the lack of sensitivity of total evapotranspiration
to vegetation changes, whereas the latter is related to a deficiency of the land surface model in simulating the surface
water budget including overestimation of the soil evaporation
and the underestimation of the transpiration. This attribution
is further confirmed by results from a set of ET‐overwriting
experiments that show a high sensitivity of precipitation to
prescribed evapotranspiration changes in CAM3 (which
therefore eliminates the atmospheric model as a possible cause
for the low precipitation sensitivity to vegetation changes).
[42] In all regions examined, ET changes trigger a highly
sensitive response of precipitation, which in some cases
involves changes in large‐scale circulation as indicated by
changes in the moisture convergence. Previous studies have
pointed out that West Africa and South America precipitation is closely linked to interactions between local ET and
large‐scale circulation. In most of West Africa, precipitation
is associated with the summer monsoon circulation, and
increase of local moisture supply (through vegetation and
soil moisture changes) was found to favor a stronger monsoon
circulation that penetrates further inland (thus increasing
moisture convergence in the Sahel region) [e.g., Xue and
Shukla, 1993; Zheng and Eltahir, 1998; Wang and Eltahir,
2000a]. Dynamically, interactions between increased ET
and African easterly jet (AEJ) slow down AEJ and strengthen
the low‐level westerly jet. Reduced moisture exports by AEJ
and increased moisture imports by low‐level westerly jet
12 of 14
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
tend to increase precipitation in the Sahel [e.g., Patricola and
Cook, 2008]. For the South America region, Xue et al. [2006]
found that the partition between sensible and latent heat flux
plays an important role in local monsoon simulation. When
ET is lower, compensating increased sensible heat flux
strengthens ventilation effect and lower moisture static
energy (MSE) of land. Lower land MSE modifies moisture
flux convergence and reduces precipitation. For the SE‐Amz
region located in the Amazon boundary area, numerical
modeling study of Oyama and Nobre [2004] focused on
vegetation’s impacts on precipitation and shown that
decreased roughness length by desertification strengthens
low‐level westward flow. At the same time, decreased diabatic heating caused by ET reduction increases divergence.
Their net effect is the reduction of moisture convergence and
precipitation following desertification.
[43] Various studies have tackled the issue of improper ET
partitioning in CLM3.0 [Bonan and Levis, 2006; Lawrence
et al., 2007; Oleson et al., 2007, 2008; Lawrence and Chase,
2009; Oleson et al., 2010], and the resulting modifications to
hydrological parameterizations have been included in the latest
version CLM4.0. Lawrence et al. [2007] found that suppressing the high soil evaporation and interception loss and
boosting the low vegetation transpiration led to improved
model performance in simulating precipitation in some areas
especially in the southeastern United States. Lawrence and
Chase [2009] shows that a more realistic ET partitioning in
CLM improves precipitation simulation, reducing precipitation in bare regions as soil evaporation decreases and
enhancing precipitation in vegetated boreal forests as vegetation transpiration increases. Our sensitivity experiments
using both the offline CLM4 and CAM4‐CLM4 confirm
previous findings on improved ET partitioning and surface
water budget. However, the impact of such improvement on
precipitation amount in the absence of vegetation and on
precipitation sensitivity to vegetation changes seems small,
and certainly is not enough to change the number of the biosphere‐atmosphere equilibrium in the coupled CAM‐CLM‐
DGVM model.
[44] Conceptual modeling proves to be an effective tool to
diagnose the ability of fully coupled land‐atmosphere
models in simulating the equilibrium state(s) attributed to
biosphere‐atmosphere interactions. For a fully coupled land‐
atmosphere model including vegetation dynamics, it takes
hundreds of model years for the vegetation distribution to
come to an equilibrium state globally, and multiple runs
using the coupled model with different initial conditions are
needed to determine whether the coupled model has one or
multiple equilibrium states. Moreover, understanding and
attributing these results are difficult due to the difficulty of
separating the different feedback processes within a coupled
model. Conceptual modeling provides a computationally
efficient alternative that allows explicit and separate treatment of the land and atmosphere responses, thus providing
better insight regarding the land and atmosphere processes
that influence the number and state(s) of the biosphere‐
atmosphere equilibrium. In the current study, results from
the conceptual model directly led to the identification of
model characteristics (or deficiencies) that anchor the coupled numerical model to a single equilibrium state, which
will guide our future effort to modify model sensitivity and
D09108
behaviors in order to explore the potential for multiple
equilibrium states in coupled biosphere‐atmosphere model.
[45] Acknowledgments. This work was supported by funding from
the National Science Foundation (NSF) Climate and Large Scale Dynamics
Program (ATM 0531485, to G. Wang). The authors thank Mike Notaro and
an anonymous reviewer for their very helpful comments.
References
Alo, C. A., and G. L. Wang (2010), Role of dynamic vegetation in regional
climate predictions over western Africa, Clim. Dyn., 35, 907–922,
doi:10.1007/s00382-010-0744-z.
Bonan, G. B., and S. Levis (2006), Evaluating aspects of the Community
Land and Atmosphere Models (CLM3 and CAM3) using a dynamic
global vegetation model, J. Clim., 19, 2290–2301, doi:10.1175/
JCLI3741.1.
Brovkin, V., M. Claussen, V. Petoukhov, and A. Ganopolski (1998), On
the stability of the atmosphere‐vegetation system in the Sahara/Sahel
region, J. Geophys. Res., 103, 31,613–31,624, doi:10.1029/
1998JD200006.
Charney, J. (1975), Dynamics of deserts and drought in the Sahel, Q. J. R.
Meteorol. Soc., 101, 193–202, doi:10.1002/qj.49710142802.
Claussen, M. (1997), Modeling bio‐geophysical feedback in the African
and Indian monsoon region, Clim. Dyn., 13, 247–257, doi:10.1007/
s003820050164.
Claussen, M. (1998), On multiple solutions of the atmosphere‐vegetation
system in present‐day climate, Global Change Biol., 4, 549–559,
doi:10.1046/j.1365-2486.1998.t01-1-00122.x.
Collins, W. D., et al. (2004), Description of the NCAR Community
Atmosphere Model (CAM3.0), NCAR Tech. Rep. NCAR/TN‐464+STR,
226 pp., Natl. Cent. for Atmos. Res., Boulder, Colo.
Dai, Y., et al. (2003), The Common Land Model (CLM), Bull. Am. Meteorol.
Soc., 84, 1013–1023, doi:10.1175/BAMS-84-8-1013.
Da Silveira Lobo Sternberg, L. S. L. (2001), Savanna‐forest hysteresis in
the tropics, Global Ecol. Biogeogr., 10, 369–378, doi:10.1046/j.1466822X.2001.00243.x.
Dirmeyer, P. A., X. A. Gao, M. Zhao, Z. Guo, T. Oki, and N. Hanasaki
(2006), GSWP‐2: Multimodel analysis and implications for our perception of the land surface, Bull. Am. Meteorol. Soc., 87, 1381–1397,
doi:10.1175/BAMS-87-10-1381.
Hurrell, J. W., J. J. Hack, D. Shea, J. M. Caron, and J. Rosinski (2008), A
new sea surface temperature and sea ice boundary dataset for the Community Atmosphere Model, J. Clim., 21, 5145–5153, doi:10.1175/
2008JCLI2292.1.
Lawrence, D. M., P. E. Thornton, K. W. Oleson, and G. B. Bonan (2007),
The partitioning of evapotranspiration into transpiration, soil evaporation,
and canopy evaporation in a GCM: Impacts on land‐atmosphere interaction, J. Hydrometeorol., 8, 862–880, doi:10.1175/JHM596.1.
Lawrence, D. M., et al. (2011), Parameterization improvements and functional and structural advances in version 4 of the Community Land
Model, J. Adv. Model. Earth Syst., 3, 1, doi:10.1029/2011MS000045.
Lawrence, P. J., and T. N. Chase (2009), Climate impacts of making evapotranspiration in the Community Land Model (CLM3) consistent with the
Simple Biosphere Model (SiB), J. Hydrometeorol., 10, 374–394,
doi:10.1175/2008JHM987.1.
Levis, S., G. B. Bonan, M. Vertenstein, and K. W. Oleson (2004), The Community Land Model’s Dynamic Global Vegetation Model (CLM‐
DGVM): Technical description and user’s guide, NCAR Tech. Rep.
NCAR/TN‐459+IA, 50 pp., Natl. Cent. for Atmos. Res., Boulder, Colo.
Liu, Z., Y. Wang, R. Gallimore, M. Notaro, and I. C. Prentice (2006), On the
cause of abrupt vegetation collapse in North Africa during the Holocene:
Climate variability vs. vegetation feedback, Geophys. Res. Lett., 33,
L22709, doi:10.1029/2006GL028062.
Mei, R., and G. L. Wang (2010), Rain follows logging in the Amazon?
Results from CAM3–CLM3, Clim. Dyn., 34, 983–996, doi:10.1007/
s00382-009-0592-x.
Myneni, R. B., et al. (2007), Large seasonal swings in leaf area of Amazon
rainforests, Proc. Natl. Acad. Sci. U. S. A., 104, 4820–4823, doi:10.1073/
pnas.0611338104.
Neale, R. B., et al. (2010), Description of the NCAR Community Atmosphere Model (CAM 4.0), NCAR Tech. Note NCAR/TN‐XXX+STR, 194
pp., Natl. Cent. for Atmos. Res., Boulder, Colo.
New, M., M. Hulme, and P. Jones (2000), Representing twentieth century
space‐time climate variability. Part 2: Development of 1901–1996
13 of 14
D09108
SUN AND WANG: BIOSPHERE‐ATMOSPHERE EQUILIBRIUM
monthly grids of terrestrial surface climate, J. Clim., 13, 2217–2238,
doi:10.1175/1520-0442(2000)013<2217:RTCSTC>2.0.CO;2.
Notaro, M. (2008), Response of the mean global vegetation distribution to
interannual climate variability, Clim. Dyn., 30, 845–854, doi:10.1007/
s00382-007-0329-7.
Oleson, K. W., et al. (2004), Technical description of the Community Land
Model (CLM), NCAR Tech. Rep. NCAR/TN‐461+STR, 186 pp., Natl.
Cent. for Atmos. Res., Boulder, Colo.
Oleson, K. W., G.‐Y. Niu, Z.‐L. Yang, D. M. Lawrence, P. E. Thornton,
P. J. Lawrence, R. Stockli, R. E. Dickinson, G. B. Bonan, and S. Levis
(2007), Technical description of the Community Land Model (CLM
3.5), report, Natl. Cent. for Atmos. Res., Boulder, Colo.
Oleson, K. W., et al. (2008), Improvements to the Community Land Model
and their impact on the hydrological cycle, J. Geophys. Res., 113,
G01021, doi:10.1029/2007JG000563.
Oleson, K. W., et al. (2010), Technical description of version 4.0 of
Community Land Model (CLM), report, Natl. Cent. for Atmos. Res.,
Boulder, Colo.
Oyama, M. D., and C. A. Nobre (2003), A new climate‐vegetation equilibrium state for Tropical South America, Geophys. Res. Lett., 30(23), 2199,
doi:10.1029/2003GL018600.
Oyama, M. D., and C. A. Nobre (2004), Climatic consequences of a
large‐scale desertification in northeast Brazil: A GCM simulation study,
J. Clim., 17, 3203–3213, doi:10.1175/1520-0442(2004)017<3203:
CCOALD>2.0.CO;2.
Patricola, C. M., and K. H. Cook (2008), Atmosphere/vegetation feedbacks: A mechanism for abrupt climate change over northern Africa,
J. Clim., 113, D18102, doi:10.1029/2007JD009608.
Sitch, S., et al. (2003), Evaluation of ecosystem dynamics, plant geography
and terrestrial carbon cycling in the LPJ dynamic global vegetation
model, Global Change Biol., 9, 161–185, doi:10.1046/j.13652486.2003.00569.x.
Stöckli, R., D. M. Lawrence, G.‐Y. Niu, K. W. Oleson, P. E. Thornton,
Z.‐L. Yang, G. B. Bonan, A. S. Denning, and S. W. Running (2008),
Use of FLUXNET in the Community Land Model development,
J. Geophys. Res., 113, G01025, doi:10.1029/2007JG000562.
Svirezhev, Y., and W. von Bloh (1996), A Minimal Model of Interaction
between climate and vegetation: qualitative approach, Ecol. Modell.,
92, 89–99, doi:10.1016/0304-3800(95)00198-0.
Wallace, J. S., H. H. C. Gash, and M. V. K. Sivakumar (1990), Preliminary
measurements of net radiation and evaporation over bare soil and fallow
bushland in the Sahel, Int. J. Climatol., 10, 203–210, doi:10.1002/
joc.3370100207.
Wang, D. G., and G. L. Wang (2007), Towards a robust canopy hydrology
scheme with precipitation sub‐grid variability, J. Hydrometeorol., 8,
439–446, doi:10.1175/JHM585.1.
D09108
Wang, G. (2004), A conceptual modeling study on biosphere‐atmosphere
interactions and its implications for physically based climate modeling,
J. Clim., 17, 2572–2583, doi:10.1175/1520-0442(2004)017<2572:
ACMSOB>2.0.CO;2.
Wang, G., and E. A. B. Eltahir (2000a), Biosphere‐atmosphere interactions
over West Africa. II: Multiple Climate equilibria, Q. J. R. Meteorol. Soc.,
126, 1261–1280, doi:10.1256/smsqj.56503.
Wang, G., and E. A. B. Eltahir (2000b), The role of ecosystem dynamics in
the enhancing the low‐frequency variability of the Sahel rainfall, Water
Resour. Res., 36, 1013–1021, doi:10.1029/1999WR900361.
Wang, G., E. A. B. Eltahir, J. A. Foley, D. Pollard, and S. Levis (2004),
Decadal variability of rainfall in the Sahel: Results from the coupled
GENESIS‐IBIS atmosphere‐biosphere model, Clim. Dyn., 22, 625–637,
doi:10.1007/s00382-004-0411-3.
Wei, J., R. E. Dickinson, and N. Zeng (2006), Climate variability in a simple model of warm climate land‐atmosphere interaction, J. Geophys.
Res., 111, G03009, doi:10.1029/2005JG000096.
Xue, Y., and J. Shukla (1993), The influence of land surface properties on
Sahel climate. Part I. Desertification, J. Clim., 6, 2232–2245,
doi:10.1175/1520-0442(1993)006<2232:TIOLSP>2.0.CO;2.
Xue, Y., F. de Sales, W.‐P. Li, C. R. Mechoso, C. A. Nobre, and H.‐M.
Juang (2006), Role of land surface processes in South American monsoon development, J. Clim., 19, 741–762, doi:10.1175/JCLI3667.1.
Zeng, N., and J. D. Neelin (2000), The role of vegetation‐climate interaction and interannual variability in shaping the African savanna, J. Clim.,
13, 2665–2670, doi:10.1175/1520-0442(2000)013<2665:TROVCI>2.0.
CO;2.
Zeng, N., J. D. Neelin, K.‐M. Lau, and C. J. Tucker (1999), Enhancement
of interdecadal climate variability in the Sahel by vegetation interaction,
Science, 286, 1537–1540, doi:10.1126/science.286.5444.1537.
Zeng, N., H. Katrina, and J. D. Neelin (2002), Nonlinear dynamics in a
coupled vegetation‐atmosphere system and implications for desert‐forest
gradient, J. Clim., 15, 3474–3487, doi:10.1175/1520-0442(2002)
015<3474:NDIACV>2.0.CO;2.
Zhang, H., A. Henderson‐Sellers, and K. McGuffie (1996), Impacts of
tropical deforestation. Part I: Process analysis of local climatic change,
J. Clim., 9, 1497–1517, doi:10.1175/1520-0442(1996)009<1497:IOTDPI>2.0.
CO;2.
Zheng, X., and E. A. B. Eltahir (1998), The role of vegetation in the
dynamics of West African monsoons, J. Clim., 11, 2078–2096,
doi:10.1175/1520-0442-11.8.2078.
S. Sun and G. Wang, Department of Civil and Environmental
Engineering, University of Connecticut, 261 Glenbrook Rd., Unit 2037,
Storrs, CT 06269‐2037, USA. ([email protected])
14 of 14