Global Change Biology (1998) 4, 561–579
Coupling dynamic models of climate and vegetation
J O N A T H A N A . F O L E Y, * S A M U E L L E V I S , * I . C O L I N P R E N T I C E 2 , † , D AV I D P O L L A R D ‡
and S T A R L E Y L . T H O M P S O N §
*Climate, People, and Environment Program (CPEP), Institute for Environmental Studies, University of Wisconsin, 1225 West
Dayton Street, Madison, WI 53706, USA, †Global Systems Group, Department of Plant Ecology, The Ecology Building, Lund
University, S-223 62 Lund, Sweden, ‡Earth System Science Center, Penn State University, University Park, PA 16802, USA,
§Climate and Global Dynamics Division, National Center for Atmospheric Research (NCAR), PO Box 3000, Boulder, CO
80307, USA
Abstract
Numerous studies have underscored the importance of terrestrial ecosystems as an
integral component of the Earth’s climate system. This realization has already led to
efforts to link simple equilibrium vegetation models with Atmospheric General Circulation Models through iterative coupling procedures. While these linked models have
pointed to several possible climate–vegetation feedback mechanisms, they have been
limited by two shortcomings: (i) they only consider the equilibrium response of
vegetation to shifting climatic conditions and therefore cannot be used to explore
transient interactions between climate and vegetation; and (ii) the representations of
vegetation processes and land-atmosphere exchange processes are still treated by two
separate models and, as a result, may contain physical or ecological inconsistencies.
Here we present, as a proof concept, a more tightly integrated framework for simulating
global climate and vegetation interactions. The prototype coupled model consists of the
GENESIS (version 2) Atmospheric General Circulation Model and the IBIS (version 1)
Dynamic Global Vegetation Model. The two models are directly coupled through a
common treatment of land surface and ecophysiological processes, which is used to
calculate the energy, water, carbon, and momentum fluxes between vegetation, soils,
and the atmosphere. On one side of the interface, GENESIS simulates the physics and
general circulation of the atmosphere. On the other side, IBIS predicts transient changes
in the vegetation structure through changes in the carbon balance and competition
among plants within terrestrial ecosystems.
As an initial test of this modelling framework, we perform a 30 year simulation in
which the coupled model is supplied with modern CO2 concentrations, observed ocean
temperatures, and modern insolation. In this exploratory study, we run the GENESIS
atmospheric model at relatively coarse horizontal resolution (4.5° latitude by 7.5°
longitude) and IBIS at moderate resolution (2° latitude by 2° longitude). We initialize
the models with globally uniform climatic conditions and the modern distribution of
potential vegetation cover. While the simulation does not fully reach equilibrium by
the end of the run, several general features of the coupled model behaviour emerge.
We compare the results of the coupled model against the observed patterns of modern
climate. The model correctly simulates the basic zonal distribution of temperature and
precipitation, but several important regional biases remain. In particular, there is a
significant warm bias in the high northern latitudes, and cooler than observed conditions
over the Himalayas, central South America, and north-central Africa. In terms of
precipitation, the model simulates drier than observed conditions in much of South
America, equatorial Africa and Indonesia, with wetter than observed conditions in
northern Africa and China.
Comparing the model results against observed patterns of vegetation cover shows
that the general placement of forests and grasslands is roughly captured by the model.
In addition, the model simulates a roughly correct separation of evergreen and deciduous
forests in the tropical, temperate and boreal zones. However, the general patterns of
© 1998 Blackwell Science Ltd.
561
562
J . A . F O L E Y et al.
global vegetation cover are only approximately correct: there are still significant regional
biases in the simulation. In particular, forest cover is not simulated correctly in large
portions of central Canada and southern South America, and grasslands extend too far
into northern Africa.
These preliminary results demonstrate the feasibility of coupling climate models with
fully dynamic representations of the terrestrial biosphere. Continued development of
fully coupled climate-vegetation models will facilitate the exploration of a broad range
of global change issues, including the potential role of vegetation feedbacks within the
climate system, and the impact of climate variability and transient climate change on
the terrestrial biosphere.
Keywords: biosphere–atmosphere interactions, climate–vegetation feedback, coupled models
Introduction
The Earth’s physical climate system and biosphere coexist
within a thin spherical shell, extending from the deep
oceans to the upper atmosphere, that is driven by energy
from the sun. The resulting interactions among the atmosphere, oceans, and terrestrial ecosystems give rise to the
biogeochemical, hydrological, and climate systems that
support life on this planet. During the last few centuries,
both the terrestrial biosphere and physical climate system
have been undergoing fundamental changes in response
to human activity. It is therefore of paramount importance
to improve our understanding of global-scale climatic
and biospheric processes.
The Earth’s climate is, in part, controlled by the fluxes
of energy, water, and momentum across the lower boundary of the atmosphere. Over the oceans, the exchange of
energy and water with the atmosphere depends primarily
on the sea-surface temperature. Fluxes over land, on the
other hand, are controlled by a more complex set of
attributes, including the state of the vegetation cover and
soils. Changes in vegetation, whether driven by climate
or human activities, can substantially alter the physical
characteristics of the land surface, potentially leading to
large feedbacks on the atmosphere. In fact, a number
of recent studies have suggested that feedbacks from
changing global vegetation patterns may have played an
important role in determining climatic conditions during
the recent geological past:
d Possible role of vegetation feedbacks in initiating glaciations.
Gallimore & Kutzbach (1996) and de Noblet et al. (1996)
have indicated how changes in the position of the boreal
forest – tundra boundary (and the associated differences
in surface albedo between dark evergreen forests and
snow-covered tundra), may be partially responsible for
initiating ice age conditions during the Quaternary.
According to this hypothesis, orbital variations and lower
CO2 concentrations initiate a trend towards glacial condiCorrespondence: J.A. Foley, fax 11/608 262 5964
tions, which is strongly amplified through the expansion
of highly reflective tundra into the darker boreal forests.
Previous climate modelling studies, which neglected the
potential for vegetation feedbacks, generally failed to
produce the conditions necessary to initiate past glaciations.
d Boreal forest feedbacks on early Holocene warming. In
another set of palaeoclimate sensitivity studies, Foley
et al. (1994) and TEMPO Authors (1996) discussed how
a poleward shift in the boreal forest – tundra boundary
during the early and middle Holocene epoch (roughly
6000–9000 years before present) may have strongly amplified orbitally induced warming of the high northern
latitudes. Once again, the vegetation feedback mechanism
acts to amplify an initial shift in climate through changes
in surface albedo associated with tundra and boreal forest
ecosystems. This vegetation feedback mechanism, which
may strongly amplify high-latitude warming, should be
further examined in terms of future climate change
scenarios.
d A ‘green sahara’ and the African monsoon of the early
Holocene. Finally, Kutzbach et al. (1996) and Claussen &
Gayler (1997) have evaluated how the orbitally enhanced
monsoon rains of northern Africa may have been amplified through vegetation feedbacks. The geological record
indicates that the Sahara of the early and middle Holocene
was a much wetter and more productive environment
than present, with extensive grasslands, savannas and
lakes (e.g. COHMAP Authors 1988). Such changes in the
land surface (and the associated changes in albedo,
leaf area, rooting depth, soil moisture, and evapotranspiration) may have enhanced the strength of the
African monsoon. The same mechanism (but acting in
reverse) was suggested by Charney et al. (1975) to explain
increasing droughts in the Sahel in terms of anthropogenic desertification.
These palaeoclimate sensitivity studies, which have
been partially corroborated by geological evidence, pro© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
C O U P L I N G D Y N A M I C M O D E L S O F C L I M AT E A N D V E G E TAT I O N
vide a very powerful motivation to further examine
climate and vegetation interactions. In particular, future
global change scenarios must be re-evaluated to consider
the potential for vegetation feedback mechanisms. For
example, scenarios of CO2-induced global warming,
already amplified in the high latitudes by snow and seaice feedbacks, may be substantially modified by longterm changes in the boreal forest and tundra boundaries.
Furthermore, several AGCM simulations have indicated
that continental interiors may become much drier in
response to global warming conditions, but this result
has not considered the potential feedbacks caused by
changes in vegetation cover. In addition, it is hypothesized that the physiological effects of increasing atmospheric CO2 concentrations could significantly influence
terrestrial vegetation by decreasing stomatal conductance
and increasing canopy leaf area (Field et al. 1995), which
may also give rise to significant feedbacks on the atmosphere (Pollard & Thompson 1995; Friend & Cox 1995;
Henderson-Sellers et al. 1995; Sellers et al. 1996; Betts et al.
1997). Clearly, models used to simulate future climate
must be improved to consider changes in vegetation
cover, and their consequent feedbacks on the atmosphere.
Linking climate and vegetation models
When considering the potential interactions between
vegetation and the atmosphere, we may distinguish
between: (i) the rapid biophysical processes of energy,
water, carbon, and momentum exchange; and (ii) the
long–term interactions which arise from fundamental
changes in the vegetation cover. For over a decade, many
AGCMs have used land surface parameterizations (e.g.
BATS of Dickinson et al. 1986; SiB of Sellers et al. 1986)
to represent the rapid biophysical exchanges of energy,
momentum, and water between terrestrial ecosystems
and the atmosphere (Sellers et al. 1997). Until recently
these land surface parameterizations have been operated
by fixing the geographical distribution of vegetation types
(Fig. 1a). Typically, each land gridcell of the AGCM is
assigned a single vegetation type (e.g. tropical rainforest,
boreal forest, tundra, desert), which has an associated set
of biophysical characteristics, including leaf area index,
albedo, rooting depth, and roughness length. However,
using fixed patterns of global vegetation within climate
models severely limits their use in studies of global
environmental change.
During the last few years, a few preliminary attempts
at linking models of vegetation cover to AGCMs have
been made. Henderson-Sellers (1993) pioneered this work
by operating the CCM1-Oz AGCM in conjunction with
the Holdridge (1947) equilibrium vegetation scheme. In
this study, the Holdridge scheme is applied annually to
update the geographical distribution of vegetation types
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
563
within the climate model. More specifically, the two
models are iteratively linked through an equilibrium
asynchronous coupling procedure (Fig. 1b), where a
single year of the climate model simulation is used to
drive changes in the equilibrium distribution of vegetation types which, in turn, are fed back into the AGCM
land surface package. In principle, this iterating cycle of
equilibrium climate and equilibrium vegetation simulations is repeated until the results of both models converge
to some final state. In this initial study, Henderson-Sellers
found that the coupled behaviour of the CCM1-Oz and
Holdridge models was stable, with no particular trends
or instabilities in the climate-vegetation system. In addition, she found only regional-scale differences in the
simulation of modern climate between simulations performed with and without interactive vegetation cover.
In another effort to link climate and equilibrium vegetation models, Claussen (1994) performed several simulations with the ECHAM AGCM linked to the BIOME
equilibrium vegetation model of Prentice et al. (1992). Like
Henderson-Sellers, Claussen also employed an iterative
equilibrium asynchronous coupling procedure to link the
climate and equilibrium vegetation models. However,
in Claussen’s study the BIOME model was driven by
multiyear averages of the climate simulation, rather than
the single-year climate simulations used in HendersonSellers (1993). Claussen also investigated the sensitivity
of the coupled climate-vegetation to the specification of
initial conditions. In particular, he compared the results
of two coupled simulations: one initialized with the
observed geographical distribution of vegetation types,
and another initialized with tropical forests replaced with
subtropical deserts and subtropical deserts replaced with
tropical forests. His results showed that the equilibrium
climate was affected by the initial vegetation conditions
to permit savannas to persist in the desert of southwestern Sahara. These results suggest that the coupled
climate-vegetation system, like many other nonlinear
systems, may arrive at different equilibrium states
depending on the choice of initial conditions.
A few other studies have applied asynchronously
linked climate-vegetation models to issues of climate
change. For example, de Noblet et al. (1996) linked the
BIOME equilibrium vegetation model to the LMD AGCM
to examine how changes in the boreal forest and tundra
boundary may have contributed to the initiation of
glaciations during the Quaternary. Claussen (1998),
Claussen (1997), and Claussen & Gayler (1997) used
the coupled ECHAM-BIOME model to investigate the
influence of initial vegetation conditions on changes in
the African and Indian monsoon during the modern and
middle Holocene eras. Finally, Betts et al. (1997) have
used coupled climate-vegetation models to investigate
564
J . A . F O L E Y et al.
Fig. 1 Methods of coupling climate and
vegetation models.
the potential role of vegetation feedbacks on future
climatic change.
These exploratory studies all underscore the importance of incorporating representations of changing vegetation cover within climate models. However, they leave
several important areas open for further model development. In particular, the current set of linked climatevegetation models have two major limitations:
(a) Most existing vegetation models simulate only the
equilibrium response of vegetation cover to changes in
climate. Models, like BIOME and the Holdridge scheme,
are unable to simulate transient changes in vegetation
cover in response to shifting environmental conditions,
such as changing climate and CO2 concentrations. Therefore, current vegetation models cannot address the time-
dependent nature of the atmosphere–biosphere response
to climate variability or future climatic change.
(b) Vegetation models and AGCM land surface parameterizations are not always physically consistent. For
example, in the coupled ECHAM-BIOME model, there
are two independent treatments of the surface water
and energy balance: one for describing land–atmosphere
exchange processes (in ECHAM) and another for estimating the soil moisture requirements of plants (in BIOME).
In ECHAM evapotranspiration is calculated on an 80
minute timestep using a detailed aerodynamic flux
formulation, while BIOME calculates evapotranspiration
once a day using a modified Priestley–Taylor formulation
linked to a soil water balance model (Claussen, pers.
comm.). An additional limitation of most equilibrium
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
C O U P L I N G D Y N A M I C M O D E L S O F C L I M AT E A N D V E G E TAT I O N
vegetation models is that they do not explicitly consider
structural attributes of vegetation, like leaf area index
and vegetation height, which are often required by AGCM
land surface parameterizations. Ideally, land surface
physics, ecological processes, and vegetation structure
should be simulated within a single, physically consistent
model (Foley 1995).
To address these issues, a new group of fully dynamic
biosphere models has recently emerged. These state-ofthe-art models, often called Dynamic Global Vegetation
Models (DGVMs), are designed to provide a more integrated and physically consistent simulation of land surface
physics, ecological processes, and vegetation structure.
Most importantly, DGVMs are also capable of simulating
transient changes in ecosystem processes and vegetation
structure (Steffen et al. 1992; Walker 1994). Some DGVMs,
including the Integrated Biosphere Simulator (IBIS) of
Foley et al. (1996), are also designed to be incorporated
directly within AGCMs, thus providing a more complete
linkage between atmospheric and ecological processes
(Fig. 1c). IBIS has already been used off-line of an AGCM
to investigate global patterns of water balance, primary
productivity, and vegetation dynamics (Foley et al. 1996),
and the potential impact of increasing CO2 concentrations
on the hydrology of the Amazon basin (Costa & Foley,
1997). Here we present a new modelling framework for
simulating atmosphere–biosphere interactions. Specifically, we have directly coupled the GENESIS global climate
model to the IBIS dynamic global vegetation model.
IBIS — a dynamic global vegetation model for
climate studies
IBIS (version 1) is designed around a hierarchical modelling framework (Fig. 2), in which information flows
between various subsystems at appropriate frequencies.
IBIS represents a number of land surface and ecosystem
phenomena, including:
d land surface and physiological processes, including photosynthesis, respiration, and stomatal behaviour, as well as
the surface energy, water, and momentum balance
d phenological behaviour of leaf display and plant activity
in response to changing climatic conditions
d transient changes in carbon balance and vegetation structure
resulting from changes in primary productivity, competition, carbon allocation, carbon turnover, and mortality
Land surface processes and phenology
IBIS includes a land surface module which simulates the
energy, water, carbon, and momentum balance of the soil–
vegetation–atmosphere interface. This module borrows
most of its basic structure from LSX (Pollard & Thompson
1995; Thompson & Pollard 1995a,b), which is the original
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
565
land-surface parameterization scheme of the GENESIS
climate model. Here the land-surface module operates
on a relatively short timestep (30 min), which is also the
timestep used by the AGCM to simulate atmospheric
dynamics. The module represents two vegetation layers
(i.e. ‘woody plants’ and ‘herbaceous plants’) and six soil
layers to simulate soil temperature, soil water, and soil
ice content over a total depth of 4.25 m. Physical processes
in the vertical soil column include heat diffusion, liquid
water transport, uptake of liquid water by roots for
transpiration, and the freezing and thawing of soil ice.
Physiologically based formulations of C3 and C4 photosynthesis (Farquhar et al. 1980; Collatz et al. 1991, 1992),
stomatal conductance (Leuning 1995) and respiration
(Amthor 1984) are used to simulate canopy gas exchange
processes. This approach provides a mechanistic link
between the exchange of water and CO2 between vegetation canopies and the atmosphere (Collatz et al. 1991;
Sellers et al. 1992; Amthor 1994; Bonan 1995). Because of
the nonlinear response of physiological parameters to
varying environmental conditions, leaf-level processes
are scaled to the canopy by assuming that the vertical
profile of Rubisco capacity is nearly optimized with
respect to net canopy photosynthesis (Field 1983; Sellers
et al. 1992; Haxeltine & Prentice 1997).
Currently, IBIS uses very simple temperature and productivity criteria to control leaf display in winter-deciduous and drought-deciduous trees. While these algorithms
reflect some basic control of climate on phenology, they
could be improved to consider more processes, including
the relationships between growing degree day requirements for budburst and chilling period length (e.g. Murray et al. 1989).
Vegetation structure
While previous vegetation models have focused on equilibrium vegetation patterns (e.g. Holdridge 1947; Prentice
et al. 1992; Neilson & Marks 1994; Haxeltine & Prentice
1996), recent efforts have produced more highly generalized approaches to simulating vegetation dynamics in a
wide range of environmental conditions. For global-scale
applications, IBIS uses a simplified representation of
vegetation dynamics where plant growth and competition
are characterized by the ability of plants to capture light
and water from common resource pools. For example,
tall woody plants (trees and shrubs) intercept light first,
and therefore shade herbaceous plants. However, herbaceous plants are able to preferentially capture soil moisture as it infiltrates through the soil. In this way, the model
can simulate some of the basic features of competition
between trees and herbaceous plants. Furthermore, competition among woody plants and among herbaceous
plants is driven by differences in the annual carbon
566
J . A . F O L E Y et al.
Fig. 2 Overview of the IBIS dynamic global vegetation model. IBIS represents a number of ecosystem processes occurring at different
timescales. Land surface physics and canopy physiological processes are simulated on a short timestep (30 min), while vegetation
phenology and vegetation dynamics are simulated on longer timesteps (1 day and 1 year, respectively).
balance resulting from differences in allocation, phenology, leaf form, and photosynthetic pathway. Unlike most
other ecosystem models, IBIS explicitly simulates light
and water availability as part of the physics of the land
surface package during each 30 minute timestep.
In IBIS the vegetation cover is represented as a combination of nine plant functional types (7 trees and 2 herbaceous plants) that are adapted from Prentice et al. (1992).
Plant functional types are defined to resolve a few
important features: basic physiognomy (e.g. trees and
grasses), leaf habit (e.g. evergreen and deciduous), photosynthetic pathway (C3 and C4), and leaf form (broadleaf and needle-leaf) (Woodward & Cramer 1996). In
order to determine which plant functional types may
potentially exist within each gridcell, the model uses a
set of simple climatic constraints (i.e. winter temperature
limits, growing degree days, and minimum chilling
requirements). In off-line simulations, IBIS uses longterm climatic means to determine the limits of each plant
functional type. However, when coupled to an AGCM,
IBIS uses a 5-year running mean of the monthly average
climatic parameters.
While IBIS-1 correctly simulates the basic global-scale
patterns of ecosystem processes and vegetation cover
(Foley et al. 1996), the current version of the model has
several limitations, including a lack of explicit disturbance
mechanisms (e.g. fire and windthrow). This omission has
important consequences on the ability of the model to
describe savannas and mixed forests, whose composition
is strongly affected by disturbance frequency. We also
expect that the representation of vegetation dynamics
could be improved by including parameterizations of
other subgrid and landscape-scale processes (Walker
1994).
GENESIS — a global climate model
The global climate model used here is version 2 of the
GENESIS (Global ENvironmental and Ecological Simulation of Interactive Systems) AGCM, which was developed
at the National Center for Atmospheric Research (NCAR)
for the main purposes of performing greenhouse and
palaeoclimatic experiments. An earlier version of the
model (version 1.02) has been described in Thompson &
Pollard (1995a,b) and Pollard & Thompson (1994, 1995).
The primary improvements to the model physics in
version 2 are documented in Thompson & Pollard (1997),
along with a discussion of the model’s present-day climate
and polar ice-sheet mass balance. In addition, a few
palaeoclimatic applications of version 2 are described in
Pollard & Thompson (1997a,b). For brevity, only the most
salient features of the model are outlined below.
The atmospheric component of GENESIS is a 3D
spectral general circulation model, which is nominally
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
C O U P L I N G D Y N A M I C M O D E L S O F C L I M AT E A N D V E G E TAT I O N
operated at T31 spectral horizontal resolution (µ 3.75°
latitude by 3.75° longitude), with 18 vertical levels, and
an overall timestep of 30 min. In this exploratory study,
we use a coarser configuration of R15 spectral resolution
(µ 4.5° latitude by 7.5° longitude) to save on computing
expenses. We expect the GENESIS model’s simulation of
modern-day climate to degrade as the resolution becomes
coarser. Yet, for the purposes of illustrating the general
behaviour of a coupled climate–vegetation model, this
resolution should be sufficient.
Solar radiation is modelled using a delta-Eddington
two-stream approximation for clear and cloudy fractions
of each atmospheric layer, with cloud overlap assumed
to be random between layers. The radiation code of the
model treats the separate effects of greenhouse trace
gases (CO2, CH4, N2O, CFC11, and CFC12) explicitly,
assuming globally uniform mixing ratios (Wang et al.
1991). O3 is prescribed from 3D monthly average climatological fields (Wang et al. 1995). The radiative effects
of tropospheric aerosols are also included, assuming a
uniform aerosol column over all nonice land points, with
a prescribed exponential decrease with height.
GENESIS uses prognostic 3D cloud water amounts to
separately predict stratus, convective, and anvil cirrus
clouds (Senior & Mitchell 1993; Smith 1990). A semiLagrangian algorithm is employed to transport water,
while convective plumes and background diffusion mix
water vertically. The model includes the processes of
evaporation, conversion to precipitation, cloud droplet
aggregation, and turbulent deposition of lowest-layer
cloud particles onto the surface. The model physics also
accounts for both ice and liquid cloud particles, but
neglects the latent heat of freezing.
While the atmospheric portion of GENESIS operates
at coarse (R15) to moderate (T31) resolutions, land and
ocean surfaces are represented at a much finer spatial
resolution of 2° latitude by 2° longitude. For each land
gridcell, the topography and fraction of surface water
are derived from the high-resolution U.S. Navy FNOC
global elevation dataset (Cuming & Hawkins 1981; Kineman 1985). While GENESIS has been previously coupled
to a variety of ocean models, in this study we prescribe
monthly mean ocean temperatures from climatological
data (Shea et al. 1992).
567
simplified to represent only 7 vegetation types, instead
of the original 18 (Table 1). For each vegetation type, a
set of leaf area index and biomass parameters are assigned
following Table 1. To avoid artifacts of the initial vegetation map (e.g. unrealistically sharp boundaries between
vegetation types), the resultant distributions of LAI and
biomass are smoothed through a 3-cell by 3-cell spatial
averaging filter.
During the first 10 years of the coupled model simulation, the model maintains the initial characteristics of the
vegetation cover, so that the atmospheric circulation will
be allowed to equilibrate without interactive vegetation.
Starting in Year 11, the coupled model allows the vegetation cover to change, directly responding to the simulated
climate. Simultaneously, a 5-year running-mean of
monthly temperatures and growing degree days is used
to broadly bound the distribution of plant types. Clearly
this way of applying climatic limits on the geographical
distribution of plant types is not process-based, but rather
is a very rough approximation for how plant distributions
respond to shifting climatic patterns. We plan to replace
this with a more mechanistic formulation in a future
version of the model, when we have explicit representations of these processes.
In this exercise we have not employed a ‘flux correction’
procedure, which is very common in atmosphere–ocean
modelling (e.g. Meehl 1990). In most coupled atmosphere–ocean models, key variables at the atmosphere–
ocean interface (i.e. sea surface temperature and salinity)
are restored towards observed climatological values, so
as to keep the coupled model from ‘drifting’ into a poor
representation of the current climate. The main drawback
of this procedure is that it lacks a physical basis. Furthermore, it is unclear how one would restore a coupled
atmosphere–biosphere model towards observations,
because the system interface cannot easily be characterized by only a few well-observed parameters. As a result,
we have decided not to use a flux correction technique
to correct for biases in the AGCM and DGVM simulations.
Thus we retain the physical consistency which this coupling scheme achieves. As coupled climate–vegetation
models become more commonly used, this issue may
need to be re-examined.
Simulated climate
Results of the coupled simulation
In this simulation, the coupled GENESIS–IBIS model is
run 30 years and is initialized with globally uniform air
temperatures (15 °C), soil temperatures (15 °C), and soil
moisture (50% of available pore space). The model’s
vegetation cover is initialized from a modified version of
Haxeltine & Prentice’s (1996) potential vegetation map.
In particular, the Haxeltine & Prentice (1996) map is
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
According to Thompson & Pollard (1997), the present-day
climate in GENESIS version 2 is significantly improved in
many regions compared to the earlier results for version
1.02. Most of the improvements apparently stem from
the greater horizontal and vertical resolutions, the new
prognostic cloud scheme, and adjustments to the convective plume scheme. In particular, precipitation, highlatitude surface temperatures, and snow and sea-ice
568
J . A . F O L E Y et al.
Table 1 Initial vegetation types. The types of vegetation defined in the potential modern vegetation map used as initialization for this
simulation. These vegetation types directly correspond to the types used by Haxeltine & Prentice (1996). Each vegetation type is
assigned an upper canopy and a lower canopy initial LAI. In every continental gridcell, this LAI is partitioned among the IBIS plant
functional types (pfts) permitted to exist according to simple climatic requirements (Foley et al. 1996). The resultant LAI maps for
individual pfts are smoothed through a 3 by 3 low pass filter and converted into leaf carbon using specific leaf area values for each
pft. Initial root carbon is assumed to equal initial leaf carbon. Initial wood carbon is such that in forests it adds up to 10 kg-C m–2
Woody
biomass
(kg-C m–2)
Simplified H & P
H & P types
Tree
LAI
1 EVERGREEN FOREST
2, 4, 6, 7, 8
6.0
10.0
2 DECIDUOUS FOREST
1, 5, 9
6.0
10.0
3 MIXED FOREST
3
6.0
10.0
4 SAVANNA GRASSLAND
10, 11, 12, 13
0.5
0.8
5 XERIC VEGETATION
14, 15
0.8
1.3
6 TUNDRA
7 DESERT
17
16, 18
0.0
0.0
0.0
0.0
distributions are all more realistic. However, Thompson
& Pollard (1997) operated GENESIS at T31 horizontal
resolution, while we use the more economical R15 resolution. As a result, we expect to have a somewhat poorer
simulation of the modern climate. To compare the climate
of the coupled model to observations, we average the
results over the last five years (Years 26–30) of the
simulation.
The general distribution of surface air temperatures
(Fig. 3) appears to be well captured by the model;
however, a significant warm bias is noticeable in the high
northern latitudes. Comparing the simulated surface air
temperatures with observations of Legates & Willmott
(1990, 1992) (Fig. 4), we see some significant regionalscale biases in the simulation. For example, there is a
large cold bias in the northern hemisphere over the
Himalayan plateau and China, which we attribute to
erroneous wintertime dynamical effects of the Himalayan
plateau. Large areas of the northern continental interiors
are too warm, notably central Canada, Eastern Europe,
and central Asia, which may have serious effects on the
local vegetation. Despite these regional climatic biases,
the zonal-average temperature fields (not shown) compare very favourably to observations and the results of
other coarse-resolution GCMs.
The global-average precipitation rate in the simulation
is 1095 mm y–1, which is within the range of observed
climatologies (e.g. 1153 mm y–1, Legates & Willmott 1992).
Over land, annual-average precipitation is 865 mm y–1
(compared to 806 mm y–1 and 864 mm y–1 according
to Baumgartner & Reichel (1975) and UNESCO (1978),
IBIS tree pfts
(%)
Herb.
LAI
Herb.
biomass
(kg-C m–2)
IBIS herbaceous
pfts(grasses)
(%)
75% evergreen
25% deciduous
25% evergreen
75% deciduous
50% evergreen
50% deciduous
50% evergreen
50% deciduous
50% evergreen
50% deciduous
none
none
1.0
0.10
1.0
0.10
1.0
0.10
4.0
0.40
0.2
0.02
2.0
0.0
0.20
0.00
50% warm
50% cool
50% warm
50% cool
50% warm
50% cool
75% warm
25% cool
50% warm
50% cool
100% cool
none
Fig. 3 Initial vegetation cover of the coupled model. We initialize
the coupled climate–vegetation model (see Table 1) with the
observed distribution of potential vegetation types (adapted
from Haxeltine & Prentice 1996).
respectively), and annual evapotranspiration is simulated
to be µ 75% of the precipitation (compared to 65% and
61%, according to Baumgartner & Reichel and UNESCO,
respectively). While the zonal mean patterns of precipita© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
C O U P L I N G D Y N A M I C M O D E L S O F C L I M AT E A N D V E G E TAT I O N
569
Fig. 4 Simulated patterns of surface air temperature. The general distribution of surface air temperatures is well captured by the
model; however, a significant warm bias is noticeable in the high northern latitudes. Surface fields from the atmosphere are passed to
the surface grid (2° latitude by 2° longitude), where they are used in the surface physics and ecosystem dynamics calculations.
tion (not shown) agree well with observations, regional
scale differences between observed and simulated (Fig. 5)
precipitation may be large (Fig. 6). There are a few areas
whose annual rainfall is much less than observed, most
notably the Amazon Basin and southern South America,
central Africa and Indonesia, which may have serious
effects on the vegetation cover. The model also tends to
slightly underestimate rainfall over northern continental
interiors between µ 40° and 60° N. The model simulates
wetter than observed conditions in northern Africa and
China. Work is in progress to improve these shortcomings;
however, the general patterns of simulated precipitation
are reasonably realistic for medium- to low-resolution
AGCMs, and may be adequate for the initial exploration
of vegetation–climate feedback mechanisms.
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
Simulated net primary productivity and vegetation
cover
While this is a relatively long (30 year) simulation by
most AGCM standards, it is important to note that the
vegetation component of the coupled model has not fully
reached equilibrium. By the end of the simulation, global
net primary productivity (NPP) has roughly stabilized
(Fig. 8a), but leaf area index (LAI) (Fig. 8b) and biomass
(Fig. 8c) have not, particularly in regions which have
experienced a large change in vegetation cover during
the simulation. We expect, based on forest LAI and
biomass data, as well as our experience with IBIS offline,
that LAI will equilibrate after another decade or two,
while biomass will equilibrate after a few more decades
570
J . A . F O L E Y et al.
Fig. 5 Simulated minus observed patterns of surface air temperature. Here we see regional-scale biases in the temperature simulation
plotted only over land. GENESIS produces a large cold bias over the Himalayan plateau and China, while large areas of the northern
continental interiors are too warm, notably central Canada, Eastern Europe, and central Asia. Observed temperatures are taken from
the Legates & Willmott (1990, 1992) climate dataset.
in the tropics and will continue to change for a century
or more in slower-growing forests of cold climates.
Net primary productivity is often used as an index to
compare ecosystem activity, and its relationship to climate. The global patterns of NPP (Fig. 9a) simulated by
the coupled model are strongly driven by the modelled
climate: areas of high NPP are in warm and moist
climates, while the least productive areas include subtropical deserts, mountain ranges, and polar regions. In this
simulation, global total NPP is 83 Gt-C y–1, which is
fairly high compared to other model-based estimates (e.g.
Foley 1994). This value would have declined in a longer
simulation due to the accumulation of respiring woody
biomass.
The simulated vegetation cover may be represented by
the geographical distributions of leaf area index (LAI)
and biomass. The global distribution of total LAI (Fig. 9b)
shows the relative density of vegetation cover, which is
strongly determined by the modelled climate: areas of
highest LAI include the rainforests of South America,
Africa, and Asia, and forests of North America and East
Asia. Total vegetation biomass (Fig. 9c) follows this same
basic pattern, with high biomass in tropical and temperate
forests, and low biomass in deserts and high-latitude
areas.
While some very basic geographical patterns of global
vegetation cover appear to be qualitatively captured by
the coupled model, there are still many areas that are
poorly simulated. For example, the forest and grassland
cover is underestimated in a large section of South
America as a result of the drier than observed conditions
simulated by GENESIS. A similar bias in the climate
model (i.e. warmer than observed temperatures, lower
than observed precipitation) results in very low produc© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
C O U P L I N G D Y N A M I C M O D E L S O F C L I M AT E A N D V E G E TAT I O N
571
Fig. 6 Simulated Patterns of Precipitation.
tivity and forest cover south-west of Canada’s Hudson
Bay. Conversely, anomalously wet conditions permit
grasslands to penetrate parts of the southern Sahara and
Arabian deserts.
We may also compare the geographical distributions
of LAI for individual plant types, to demonstrate that
many basic patterns of vegetation geography are simulated fairly well by the coupled model (Figs 10 and 11).
The model retains a clear separation of forests, grasslands,
and tundra, with grasslands and tundra dominating
the semiarid and high-latitude regions of the world.
Grasslands are scattered throughout central Asia, central
North America, central Australia, and the semiarid
regions of Africa and South America. Because of the lack
of natural disturbance mechanisms, particularly fire, the
model does not simulate the large areas of savanna
vegetation in Africa and South America.
In the forested portions of the world, the model clearly
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
indicates a separation of evergreen-dominated and
deciduous-dominated forests (Fig. 10a and b). For
example, in the high-latitudes, the model successfully
sustains a boreal deciduous (i.e. Larix dominated) forest
in eastern Siberia, and evergreen-dominated forests
throughout the rest of the circumpolar boreal zone. The
mid-latitude forests of Europe, eastern North America,
and east Asia are a mixture of evergreen and deciduous
trees, but deciduous trees clearly dominate the core of
the temperate forest regions (eastern North America,
western Europe, and portions of east Asia). In the tropics,
the model includes regions of drought deciduous forests
in the areas of seasonal rainfall, and tropical evergreen
forests in the core of annual rainfall in the Amazon basin,
central Africa, and Indonesia. In south-central Africa,
however, tropical deciduous trees are shown to give way
to lush grasslands, because of cooler than observed
temperatures in the region.
572
J . A . F O L E Y et al.
Fig. 7 Simulated minus observed patterns of precipitation. While the zonal patterns of precipitation agree well with observations (not
shown), regional-scale differences between observed and simulated precipitation (plotted only over land) may be large. The model
simulates drier than observed conditions in much of South America, equatorial Africa and Indonesia, with wetter than observed
conditions in northern Africa and China. The model also tends to slightly underestimate rainfall over northern continental interiors
between µ 40° to 60° N. Observed precipitation rates are taken from the Legates & Willmott (1990, 1992) climate dataset.
Fig. 8 Simulated trends of (a) Net Primary Production (NPP), (b) Average Leaf Area Index (LAI), and (c) average biomass. The trend
in the global total NPP (a) appears to roughly stabilize by the end of the simulation, while the trends in the LAI (b) and especially
the biomass (c) do not.
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
C O U P L I N G D Y N A M I C M O D E L S O F C L I M AT E A N D V E G E TAT I O N
573
Fig. 9 Simulated patterns of (a) Net Primary Production (NPP),
(b) Total Leaf Area Index (LAI), and (c) total biomass. The
resulting patterns of NPP (a) in the coupled model (averaged
between years 26 and 30) are strongly driven by the simulated
climate: areas of high NPP are in warm and moist climates,
while the least productive areas include subtropical deserts,
mountain ranges, and polar regions. The same applies to total
LAI (b) and total biomass (c).
Vegetation feedbacks on climate
To assess the potential feedbacks the interactive vegetation cover may have on the climate, we compare two
periods of simulation: one with fixed vegetation (Years
6–10), and one with interactive vegetation (Years 26–30).
The interactive changes in vegetation cover result in
significant (at the 95% confidence level) temperature
changes in several regions (Fig. 12). For example, the
changing vegetation cover causes cooling over the
southern Sahara and Arabian deserts, mainly during the
drier seasons (boreal autumn, winter, and spring), which
can be attributed to the northward expansion of grasslands. In the early stages of the simulation, the model
places a stronger than observed monsoon over northern
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
Africa, permitting grasslands to penetrate northward. The
ensuing changes in vegetation result in greater evapotranspiration, which in part explains lower temperatures
during the drier times of year. Furthermore, greater
amounts of atmospheric moisture may cause additional
cloudiness, less solar radiation at the surface, and cooler
temperatures. Conversely, a reduction in vegetation cover
causes significant warming over south-central Africa in
austral summer, over much of south-eastern South
America in austral summer, and in central Asia during
boreal summer. Differences in relative humidity (not
shown) also show a clear signal of the effects of interactive
vegetation, with moister conditions over much of the
Sahara and Arabian deserts, and drier conditions in
central Asia.
574
J . A . F O L E Y et al.
Fig. 10 Simulated Leaf Area Index for (a) Evergreen Trees, (b)
Deciduous Trees, and (c) Herbaceous Plants. The resulting
patterns of LAI at the end of the simulation (averaged between
Years 26 and 30) result from the interactions between the climate
and vegetation cover. These results show that the general
placement of forests and grasslands is roughly captured by the
coupled model. The model also simulates a roughly correct
separation of evergreen and deciduous forests in the tropical,
temperate, and boreal zones. However, the general patterns of
global vegetation cover are only approximately correct: there
are still significant regional biases in the simulation. In particular,
forest cover is not simulated correctly in large portions of central
Canada and southern South America, and grasslands extend too
far into northern Africa.
Changes in precipitation resulting from the interactive
vegetation cover are not as clear at the 95% confidence
level (Fig. 13). Most notably, we observe an increase in
precipitation over part of the Sahara late in the wet
season (boreal autumn) and the southern Arabian desert
during the wettest season (boreal summer). The effects
of interactive vegetation cover also cause significant
increases in snow cover at the southern edge of the boreal
forest and significant decreases in regions dominated
by tundra (not shown). It is important to note that
precipitation displays greater interannual variability than
temperature and relative humidity, making it difficult for
a t-test to capture the significance of the changes resulting
from interactive vegetation.
Overall, the differences in climate resulting from interactive changes in vegetation cover are relatively small
compared to the intrinsic biases of the AGCM. The largest
changes in climate are associated with changes in the
position of desert/grassland and forest/tundra ecotones,
which have already been implicated as centres of action
in previous climate–vegetation modelling studies (e.g.
Claussen 1994; Foley et al. 1994).
Conclusions
Scenarios of future climate change must be re-evaluated
to consider the potential for vegetation feedback
mechanisms. In particular, the effects of several hypo© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
C O U P L I N G D Y N A M I C M O D E L S O F C L I M AT E A N D V E G E TAT I O N
575
Fig. 11. Simulated Leaf Area Index for each plant functional type.
thesized vegetation feedback mechanisms should be
investigated, including: (a) changes in albedo resulting
from shifting boreal forest and tundra boundaries, (b)
changes in the extent of deserts and the resulting
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
changes in albedo and evapotranspiration, (c) increases
in mid-continental aridity, with the consequent changes
in vegetation cover and soil moisture, and (d) changes
in vegetation cover and evapotranspiration directly
576
J . A . F O L E Y et al.
Fig. 12 Differences in simulated temperature between interactive vegetation (Years 26–30) and Fixed Vegetation (Years 6–10) Conditions.
To assess how the interactive vegetation cover may affect the overall climate of the model, we compare the surface air temperature
patterns between interactive vegetation (Years 26–30) and fixed vegetation (Years 6–10) portions of the run. We employ a student ttest to assess the statistical significance (at the 95% level) of the differences in temperature. The significant changes in temperature are
plotted only over land.
resulting from the physiological effects of increased
CO2 concentrations. Using coupled climate–vegetation
models could help elucidate the role of vegetation
feedback mechanisms on future climate.
Previous studies, including Henderson-Sellers (1993)
and Claussen (1994), have demonstrated the feasibility
of linking simple equilibrium vegetation models with
atmospheric general circulation models. While these
exploratory efforts have highlighted the importance of
representing vegetation cover within AGCMs, they
have suffered from two limitations: (a) the equilibrium
nature of current vegetation models, and (b) possible
inconsistencies between vegetation models and land
surface parameterizations. Ideally, AGCMs should be
directly coupled to fully dynamic and integrated models
of the biosphere, where land surface physics, ecological
processes, and vegetation dynamics would be simulated
within a single, physically consistent framework (Foley
1995). Here we have taken the next logical step of
coupling an AGCM with a fully dynamic and integrated
global vegetation model.
In any coupled modelling exercise, it is important to
examine the variability of the system across a wide
variety of timescales. Unfortunately, this initial simulation is not long enough for us to judge whether the
climate–vegetation system will remain stable around
the simulated mean conditions or it will oscillate with
a discernible long-term trend. Longer simulations would
also be required to see if the coupled model exhibits
a ‘climate drift’, equivalent to what has been seen in
many coupled atmosphere–ocean models.
Modelling the dynamical interactions between the
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
C O U P L I N G D Y N A M I C M O D E L S O F C L I M AT E A N D V E G E TAT I O N
577
Fig. 13 Differences in simulated precipitation between interactive vegetation (Years 26–30) and fixed vegetation (Years 6–10) Conditions.
Here we compare the difference in precipitation patterns between interactive and fixed vegetation portions of the run. We employ a
student t-test to assess the statistical significance (at the 95% level) of the differences in precipitation. The significant changes in
precipitation are plotted only over land.
physical climate system and the terrestrial biosphere is
still in its very early stages. While incremental improvements to both climate and global vegetation models are
needed, important methodological issues still need to be
resolved. Among these are issues of stability, modes
of variability, potential for ‘drift’, sensitivity to initial
conditions, as well as the major limitation of coupled
models which stems from the computing resources
required for long-term simulations. Ideally, a coupled
climate-vegetation model would be run over several
decades or centuries, which may be more feasible with
reduced-form climate models rather than AGCMs
(Claussen, pers. comm.). While many of these issues still
remain largely unexplored in coupled climate–vegetation
models, most of them have been examined by atmosphere–ocean modellers over the last two decades. Atmosphere–biosphere modellers may gain considerable
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
experience from comparing methods with atmosphere–
ocean modellers.
Acknowledgements
We would like to thank Navin Ramankutty and Christine
Molling of the Climate, People, and Environment Program
(CPEP) for their help in this effort. In addition, we would like
to thank Martin Claussen, Jim Collatz, and an anonymous
reviewer for their helpful comments on the manuscript. This
work was supported by the Climate Dynamics Program of the
U.S. National Science Foundation through grant ATM-94–08836.
The development of the GENESIS model at the National Center
for Atmospheric Research (NCAR) is supported in part by the
U.S. Environmental Protection Agency Interagency Agreement
DW49935658–01–0. The simulations presented in this study were
performed at the NCAR Climate Simulation Laboratory (CSL)
under computing grant 93300185. The National Center for
Atmospheric Research is sponsored by the U.S. National Science
Foundation.
578
J . A . F O L E Y et al.
References
Amthor JS (1984) The role of maintenance respiration in plant
growth. Plant, Cell and Environment, 7, 561–569.
Amthor JS (1994) Scaling CO2-photosynthesis relationships from
the leaf to the canopy. Photosynthesis Research, 39, 321–350.
Baumgartner A, Reichel E (1975) The World Water Balance:
Mean Annual Global Continental and Maritime Precipitation,
Evaporation, and Runoff. Elsevier, New York, 179pp.
Betts RA, Cox PM, Lee SE, Woodward FI (1997) Contrasting
physiological and structural vegetation feedbacks in climate
change simulations. Nature, 387, 796–799.
Bonan GB (1995) Land-atmosphere CO2 exchange simulated by
a land surface process model coupled to an atmospheric
general circulation model. Journal of Geophysical Research, 100,
2817–2831.
Charney J, Stone PH, Quirk WJ (1975) Drought in the Sahara: A
biogeophysical mechanism. Science, 187, 434–435.
Claussen M (1994) Coupling global biome models with climate
models. Climate Research, 4, 203–221.
Claussen M (1997) Modeling bio-geophysical feedback in the
African and Indian monsoon region. Climate Dynamics, 13,
247–257.
Claussen M (1998) On multiple solutions of the atmospherevegetation system in present-day climate. Global Change
Biology, this issue.
Claussen M, Gayler V (1997) The greening of Sahara during the
mid-Holocene: results of an interactive atmosphere–biome
model. Global Ecology and Biogeography Letters, in press.
Cohmap (1988) Climatic changes of the last 18000 years:
observations and model simulations. Science, 241, 1043–1052.
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.
Costa MH, Foley JA (1997) The water balance of the Amazon
basin dependence on vegetation cover and canopy
conductance. Journal of Geophysical Research – Atmospheres,
102, 23,973–23,989.
Cuming MJ, Hawkins BA (1981) TERDAT: The FNOC system
for terrain data extraction and processing. Technical Report
M11 Project, M254 (2nd edn). Prepared for Fleet Numerical
Oceanography Center, Monterey: Meteorology International
Inc.
Dickinson RE, Henderson-Sellers A, Kennedy PJ, Wilson MF
(1986) Biosphere-Atmosphere Transfer Scheme (BATS) for
the NCAR CCM. NCAR Technical Note, NCAR/TN-275STR, Boulder, CO, 69pp.
Farquhar GD, von Caemmerer S, Berry JA (1980) A biochemical
model of photosynthetic CO2 assimilation in leaves of C3
species. Planta, 149, 78–90.
Field C (1983) Allocating leaf nitrogen for the maximization of
carbon gain: leaf age as a control on the allocation program.
Oecologia, 56, 341–347.
Field CB, Jackson RB, Mooney HA (1995) Stomatal responses to
increased CO2 implications from the plant to the global scale.
Plant, Cell And Environment, 18, 1214–1225.
Foley JA (1994) Net primary productivity in the terrestrial
biosphere: the application of a global model. Journal of
Geophysical Research, 99, 20773–20783.
Foley JA (1995) Numerical models of the terrestrial biosphere.
Journal of Biogeography, 22, 837–842.
Foley JA, Kutzbach JE.Coe MT, Levis S (1994) Feedbacks between
climate and boreal forests during the Holocene epoch. Nature,
371, 52–54.
Foley JA, Prentice CI.Ramankutty N.Levis S.Pollard D.Sitch S,
Haxeltine A (1996) An integrated biosphere model of land
surface processes, terrestrial carbon balance, and vegetation
dynamics. Global Biogeochemical Cycles, 10, 603–628.
Friend AD, Cox PM (1995) Modelling the effects of atmospheric
CO2 on vegetation–atmosphere interactions. Agricultural
Forest Meteorology, 73, 285–295.
Gallimore RG, Kutzbach JE (1996) Role of orbitally induced
changes in tundra area in the onset of glaciation. Nature, 381,
503–505.
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.
Haxeltine A, Prentice IC (1997) A general model for the light
use efficiency of primary production. Functional Ecology, 10,
551–561.
Henderson-Sellers A (1993) Continental vegetation as a dynamic
component of a global climate model: a preliminary
assessment. Climatic Change, 23, 337–377.
Henderson-Sellers A, Mcguffie K, Gross C (1995) Sensitivity
of global climate model simulations to increased stomatal
resistance and CO2 increases. Journal of Climate, 8, 1738–1756.
Holdridge LR (1947) Determination of world plant formations
from simple climatic data. Science, 105, 367–368.
Kineman J (1985) FNOC/NCAR global elevation, terrain, and
surface characteristics. Digital Dataset, 28 MB, Boulder:
NOAA National Geophysical Data Center.
Kutzbach JE, Bonan G, Foley JA, Harrison SP (1996) Vegetation
and soil feedbacks on the response of the African monsoon
to orbital forcing in the early to middle Holocene. Nature,
384, 623–626.
Legates DR, Willmott CJ (1990) Mean seasonal and spatial
variability in gauge-corrected, global precipitation. Int. J.
Climatol., 111–127.
Legates DR, Willmott CJ (1992) A comparison of GCM simulated
and observed mean January and July precipitation.
Paleogeography, Paleoclimatology, Paleoecology (Global and
Planetary Change Section), 97, 345–363.
Leuning R (1995) A critical appraisal of a combined stomatalphotosynthesis model for C3 plants. Plant, Cell and
Environment, 18, 339–355.
Meehl GA (1990) Development of global coupled oceanatmosphere general circulation models. Climate Dynamics, 5,
19–33.
Murray MB, Cannell MGR, Smith I (1989) Date of budburst of
fifteen tree species in Britain following climatic warming.
Journal of Applied Ecology, 26, 693–700.
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
C O U P L I N G D Y N A M I C M O D E L S O F C L I M AT E A N D V E G E TAT I O N
Neilson RP, Marks D (1994) A global perspective of regional
vegetation and hydrologic sensitivities from climatic change.
Journal of Vegetation Science, 5, 715–730.
de Noblet NI, Prentice IC, Joussaume S.Texier D.Botta A,
Haxeltine A (1996) Possible role of atmosphere–biosphere
interactions in triggering the last glaciation. Geophysical
Research Letters, 23, 3191–3194.
Pollard D, Thompson SL (1994) Sea-ice dynamics and CO2
sensitivity in a global climate model. Atmosphere-Ocean, 32,
449–467.
Pollard D, Thompson SL (1995) Use of a land-surface-transfer
scheme (LSX) in a global climate model: the response to
doubling stomatal resistance. Global and Planetary Change, 10,
129–161.
Pollard D, Thompson SL (1997a) Climate and ice-sheet mass
balance at the last glacial maximum from the GENESIS
version 2 global climate model. Quaternary Science Reviews,
in press.
Pollard D, Thompson SL (1997b) Driving a high-resolution
dynamic ice-sheet model with GCM climate: Ice-sheet
initiation at 116000 B.P. Annals of Glaciology, in press.
Prentice C, Cramer W, Harrison SP, Leemans R, Monserud RA,
Solomon AM (1992) A global biome model based on plant
physiology and dominance, soil properties and climate.
Journal of Biogeography, 19, 117–134.
Sellers PJ, Berry JA, Collatz GJ, Field CB, Hall FG (1992)
Canopy reflectance, photosynthesis, and transpiration. III. A
reanalysis using improved leaf models and a new canopy
integration scheme. Remote Sensing of the Environment, 42,
187–216.
Sellers PJ, Bounoua L, Collatz GJ, Randall DA, Dazlich DA, Los
SO, Berry JA, Fung I, Tucker CJ, Field CB, Jensen TG
(1996) Comparison of radiative and physiological effects of
atmospheric CO2 on climate. Science, 271, 1402–1406.
Sellers PJ, Dickinson RE, Randall DA, Betts AK, Hall FG, Berry
JA, Collatz GJ, Denning AS, Mooney HA, Nobre CA, Sato
N, Field CB, Henderson-Sellers A (1997) Modeling the
exchanges of energy, water, and carbon between continents
and the atmosphere. Science, 275, 502–509.
Sellers PJ, Mintz Y, Sud YC, Dalcher A (1986) A simple biosphere
© 1998 Blackwell Science Ltd., Global Change Biology, 4, 561–579
579
model (SiB) for use within general circulation models. Journal
of Atmospheric Science, 43, 505–531.
Senior CA, Mitchell JFB (1993) Carbon dioxide and climate:
the impact of cloud parameterization. Journal of Climate, 6,
393–418.
Shea DJ, Trenberth KE, Reynolds RW (1992) A Global Monthly
Sea Surface Temperature Climatology. Journal of Climate, 5,
987–1001.
Smith RNB (1990) A scheme for predicting layer clouds and
their water content in a general circulation model. Quarterly
Journal of the Royal Meteorological Society, 116, 435–460.
Steffen WL, Walker BH, Ingram JS, Koch GW (eds) (1992) Global
changes and terrestrial ecosystems: The operational plan.
Global Change Rep., 21, 95pp.. International Geosphere–
Biosphere Programme, Stockholm.
Tempo (1996) Potential role of vegetation feedback in the climate
sensitivity of high-latitude regions: A case study at 6000
years B.P. Global Biogeochemical Cycles, 10, 727–736.
Thompson SL, Pollard D (1995a) A global climate model
(GENESIS) with a land-surface transfer scheme (LSX). Part
I: present climate simulation. Journal of Climate, 8, 732–761.
Thompson SL, Pollard D (1995b) A global climate model
(GENESIS) with a land-surface transfer scheme (LSX). Part
II: CO2 sensitivity. Journal of Climate, 8, 1104–1121.
Thompson SL, Pollard D (1997) Greenland Antarctic mass
balances for present and doubled atmospheric CO2 from the
GENESIS version 2 global climate model. Journal of Climate,
10, 871–900.
Unesco (1978) World Water Balance and Water Resources of the
Earth. Paris/Leningrad, 663pp.
Walker B (1994) Landscape to regional-scale responses of
terrestrial ecosystems to global change. Ambio, 23, 67–73.
Wang W-C, Liang X-Z.Dudek MP.Pollard D, Thompson SL (1995)
Atmospheric ozone as a climate gas. Atmospheric Research,
37, 247–256.
Wang W-C, Shi G-Y, Kiehl JT (1991) Incorporation of thermal
radiative effect of CH4, N2O, CF2Cl2, and CFCl3 into the
NCAR Community Climate Model. Journal of Geophysical
Research, 96, 9097–9103.
Woodward FI, Cramer W (1996) Plant functional types and
climatic changes: Introduction. Journal of Vegetation Science,
7, 306–308.