GLOBAL BIOGEOCHEMICAL CYCLES, VOL. 14, NO.

3, PAGES 795-825, SEPTEMBER 2000

Testingthe performanceof a Dynamic Global EcosystemModel:

Water balance, carbon balance, and vegetation structure
ChristopherJ. Kucharik,JonathanA. Foley,ChristineDelire, VeronicaA. Fisher,

Michael
T. Coe,John
D. Lenters,
Christine
Young-Molling,
l andNavinRamankutty
Climate,PeopleandEnvironmentProgram(CPEP), Institutefor EnvironmentalStudies,Universityof
Wisconsin,Madison

John M. Norman

Departmentof Soil Science,Universityof Wisconsin,Madison

Stith T. Gower

Departmentof ForestEcologyandManagement,Universityof Wisconsin,Madison

Abstract. While a new classof DynamicGlobalEcosystemModels(DGEMs) hasemergedin the

pastfew yearsasan importanttool for describing
globalbiogeochemical
cyclesandatmospherebiosphereinteractions,
thesemodelsarestill largelyuntested.Here we analyzethe behaviorof a
new DGEM andcomparethe resultsto global-scaleobservations
of waterbalance,carbon
balance,andvegetationstructure.In thisstudy,we useversion2 of the IntegratedBiosphere
Simulator(IBIS), whichincludesseveralmajorimprovements
andadditionsto the prototype
modeldevelopedby Foley et al. [1996]. IBIS is designedto be a comprehensive
modelof the
terrestrialbiosphere;
the modelrepresents
a wide rangeof processes,
includinglandsurface
physics,canopyphysiology,plantphenology,
vegetation
dynamicsandcompetition,
andcarbon
andnutrientcycling. The modelgenerates
globalsimulations
of the surfacewaterbalance(e.g.,
runoff),theterrestrialcarbonbalance(e.g.,netprimaryproduction,net ecosystem
exchange,soil
carbon,aboveground
andbelowground
litter,andsoilCO2fluxes),andvegetationstructure(e.g.,
biomass,leaf areaindex,andvegetationcompositior0.In orderto testthe performance
of the
model,we haveassembled
a wide rangeof continer/tal-andglobal-scaledata,including
measurements
of river discharge,
net primaryproduction,
vegetationstructure,rootbiomass,soil
carbon,litter carbon,andsoil CO2flux. Usingthesefield dataandmodelresultsfor the
contemporary
biosphere(1965-1994),ourevaluationshowsthatsimulatedpatternsof runoff,
NPP, biomass,leaf areaindex, soil carbon,andtotal soil CO2 flux agreereasonablywell with
measurements
thathavebeencompiledfromnumerousecosystems.Theseresultsalsocompare
favorablyto otherglobalmodelresults.

1. Introduction

Prenticeet al., 1992;NeilsonandMarks,1994;Woodward

et al.,
1995; Haxeltineand Prentice, 1996], and land-atmosphere

Recently, ecologicalresearchhas focusedon the functioning

and dynamic nature of ecosystems,along with their role in the
global carbon,nutrient,and water cycles. Part of this focushas
resulted in the development of global terrestrial ecosystem
models, including models of terrestrial biogeochemistry[e.g.,

exchange
processes
[e.g.,Dickinson
et al., 1986;Sellers
et al.,

Melillo et al., 1993,' Potter et al., 1993; Parton et al., 1993;

Runningand Gower, 1991], global vegetationbiogeography[e.g.,

1996; Bonan,1995]. However,thesemodels,which were

individually
designed
with specificgoalsin mind,consider
a
limitedrangeof ecosystem
processes.Only recentlyhaveglobal
ecosystemmodels been designedto integrate biophysical
processes,
biogeochemical
cycles,andvegetation
dynamics
intoa
single,consistent
framework[Foley, 1995; Foley et al., 1996;
Friend et al., 1997].

Large-scale
integrated
biosphere
modelsareurgentlyneededto
improveour understanding
of how changesin land-use,climate
variability, and increasesin atmosphericCO2 concentrations
mightaffectthe structureandoverallfunctioningof bothnatural
and managedecosystems
acrossthe globe. Suchchangesin
globalecosystems
havethe potentialto dramaticallymodifythe
availabilityof essentialnaturalresourcessuchas water, food,

INowatSpace
Science
andEngineering
Center
(SSEC),University
of
Wisconsin,Madison.

Copyright2000 by the AmericanGeophysicalUnion.

Papernumber1999GB001138.
0886-6236/00/1999GB001138512.00

795

796

KUCHARIK ET AL.: TESTING A DYNAMIC GLOBAL ECOSYSTEM MODEL

timber, and fiber during the next century. Furthermore, any

significantchangesin the terrestrialbiospheremay contributeto
additional feedbackson the climate systemthrough changesin
vegetationpatternsor changesin the productionand consumption
of greenhousegases.
It is thereforeof the utmost importancethat we analyze the
complexinteractionsamongall facetsof the biosphere,including
atmospheric, vegetative, hydrological, and biogeochemical
processes.However, the dynamic interactionsthat occur within
the terrestrial biosphere take place across a continuum of
timescales,ranging from secondsto hundredsof years. This
systemcomplexity has made the developmentand validation of
integratedmodelsof the terrestrialbiospherea difficult endeavor
but one that is essentialto globalchangeresearch.
An integratedmodel of the Earth's biosphere,the Integrated
BiosphereSimulator (or IBIS) [Foley et al., 1996] has been
developed as a first step toward gaining an improved
understandingof global biosphericprocessesand studyingtheir
potential responseto human activity. IBIS is constructedto
explicitly link land surfaceand hydrologicalprocesses,terrestrial
biogeochemicalcycles,and vegetationdynamicswithin a single,
physicallyconsistentframework. Furthermore,IBIS is one of a
new generation of global biosphere models, termed Dynamic
Global Vegetation Models (or DGVMs) [Steffenet al., 1992,'
Walker, 1994,' W. Cramer et al., Dynamic responsesof global
terrestrialvegetationto changesin CO2 and climate, submittedto
Global ChangeBiology, 1999], that considerstransientchangesin
vegetationcompositionand structurein responseto environmental
change.Previousglobalecosystem
modelshavetypicallyfocused
on the equilibrium state of vegetation and could not allow
vegetationpatternsto changeover time.
Here we presenta new versionof the IBIS model (version2),
whichhas improvedrepresentations
of land surfacephysics,plant
physiology, canopy phenology, plant functional type (PFT)
differences,and carbon allocation. In addition, IBIS-2 includesa
new belowgroundbiogeochemistrysubmodel[C.J. Kucharik et
al., Measurementsand modelingof carbonand nitrogendynamics
in managed and natural ecosystemsof southern Wisconsin,
submitted to Ecosystems, 1999] (hereinafter referred to as
Kuchariket al., submittedmanuscript,1999), which is coupledto
detritusproduction(litterfall andfine rootturnover).
In this study, we presentmodel simulationsof the terrestrial
carbonbudget(net primaryproduction,vegetationbiomass,soil
carbon, litter carbon, soil CO2 flux, and microbial biomass),
globalvegetationpatterns,andpatternsof runoff. To evaluatethe
model output acrossthe world s biomes,we use a new global

compilation
of netprimaryproduction
(NPP)measurements
[S.T.
Gower,unpublished
data, 1999],a globalgriddedsoil carbon
database
[InternationalGeosphere-Biosphere
Programme- Data
andInformation
System
(IGBP-DIS),1999]andsoilCO2fluxdata
[RaichandSchlesinger,
1992]. Measurements
of runoff[Cogley,
1991],microbial
biomass
[Landsberg
andGower,1997],fineroot
biomass
[Jackson
et al., 1996]andlittercarbon[Matthews,
1997]
are also used to test the model results.

2. IBIS-2: An Integrated Model of Biospheric

Processes

The IBIS model includes, in a single integratedframework,

representationsof land surface processes(energy, water, and

momentum exchange between soil, vegetation, and the

atmosphere),
canopyphysiology(canopyphotosynthesis
and
conductance),
vegetationphenology(budburstand senescence),
vegetationdynamics(allocation,turnover,and competition
betweenplanttypes),andterrestrial
carbonbalance(netprimary
production,tissueturnover,soil carbon,and organicmatter
decomposition)
[Foley et al., 1996]. These processesare

organized
in a hierarchical
framework
andoperateat different
timesteps,rangingfrom60 min to 1 year(Figure1). Suchan
approachallows for explicit couplingamong ecological,
biophysical,
andphysiological
processes
occurring
on different
timescales.

A description
of the IBIS-1 modelis givenby Foley et al.
[1996],soit is notdescribed
in detailhere.Instead,
we will focus
on describingthe major improvements,
modifications,and
additions to IBIS that have been made within the current version
of the model.

2.1. Land Surface Physics

The IBIS land surface module, which is based on the land

surfacetransfermodel (LSX) packageof Thompsonand Pollard

[1995a,b], simulatesthe energy,water, carbon,and momentum
balanceof the soil-vegetation-atmosphere
system. The model

represents
two vegetationcanopies(i.e., treesversusshrubsand
grasses),eight soil layers, and three layers of snow (when
required)(seeFigure2). Followingthe logicof mostlandsurface
packages,
IBIS explicitlyrepresents
thetemperature
of the soil(or
snow) surface and the vegetationcanopies,as well as the
temperatureand humidity within the canopyair spaces. The
radiation balance of the vegetationand the ground and the
diffusive and turbulent fluxes of sensible heat and water vapor

drive changesin temperature

and humidity. In orderto resolve
the diurnalcycle,the IBIS land surfacemoduleusesa relatively
shorttime step(60 min in thisstudy).
IBIS simulates the exchange of both solar and infrared
radiationbetweenthe atmosphere,the vegetationcanopies,and
the surface. Solar radiation transfer is simulatedfollowing the

two-streamapproximation,
with separatecalculations
for direct
and diffuse radiation in both visible (0.4-0.7 m) and near-

infrared(0.7-4.0 m) wavelengths.Comparedto LSX and IBIS1, the solar radiative transfer scheme of IBIS-2 has been
simplified;sunlit and shadedfractionsof the canopiesare no
longertreatedseparately.Themodelnowusesthesolutionof the
two-stream
approximation
followingtheapproach
of Sellerset al.
[1986] and Bonan[1995]. Infraredradiationis simulatedas if
each vegetation layer is a semitransparentplane; canopy
emissivitydependson foliagedensity.
Turbulent fluxes and wind speedsthroughthe canopiesare
modeledusinga simplelogarithmic
profileaboveandbetweenthe
canopylayersanda simplediffusivemodelof air motionwithin
each layer. In contrastto LSX and IBIS-I, which use a
logarithmicwindprofilebelowthe lowestcanopy,IBIS-2 usesan
empiricallinear functionof wind speedto estimateturbulent
transfer between the soil (or snow) surface and the lower
vegetationcanopy. This relationwasdeducedempiricallyfrom
extensivestudiesin wind tunnelsandbeneathvegetationcanopies
in thefield by Saueret al. [1995].
The totalamountof evapotranspiration
fromthe landsurfaceis
treatedas the sumof threewater vaporfluxes:evaporationfrom

KUCHARIK ET AL.: TESTING A DYNAMIC GLOBAL ECOSYSTEMMODEL

ATMOSPHERE

797

IBIS Structure

(prescribed atmospheric datasets)

VEGETATION

DYNAMICS

MODULE

gross
primary
I total
I net
primary
product
on

grossphotosynthesis,
foliagerespiration,&
optimalC:N ratiosfor
each plant functionaltype
LAND SURFACE
MODULE

respiration production

J/allocation
Igrowthofleaves,Jmo
JJ
ndturnoverJ
stems
&rootsdisturbance

vegetationstructure
and biomass

fitter
0 nitrogen
supply
fall

photosynthesis
I stomatal
&leaf
respiration
]conductance

BELOWGROUND
CARBON & NITROGEN
CYCLING MODULE

canopy nitrogen
allocation

i?}::!ii:i]
&soI organic
matter
[respirat
on[;:://

temperature,

61
chn

li?i:!i?'i']
.nitr,.gen..
nlnitrification
denitrificati

photosynthesis

i,,ii:,:]mlneraHzauo
[

;iiiiii',i
bdburst
senescence
dormancy
[iiiiiii?/i?i
t ,,, minutes

to hours

I::::i
i':i:.':i::':
i:i':i

t ,,, years

t ,,, days to weeks

Figure1. Schematic
of theDynamic
GlobalVegetation
Model(DGVM)IBIS [Foleyet al., 1996]. The
characteristic
timescales
of theprocesses
areindicated
atthebottom
of thefigure.

the soil surface,evaporation

of waterintercepted
by vegetation
canopies,
andcanopytranspiration.
Ratesof transpiration
depend
on canopy conductance(see below) and are calculated
independently
for eachplanttypewithinthecanopy.To account
for evaporation
from intercepted
rain, the modeldescribes
the
interception
and cascadeof precipitation
(bothrain and snow)
throughthecanopies.
IBIS uses a multilayer formulationof soil to simulatethe
diurnal and seasonalvariationsof heat and moisturein the top 12

and carbon dioxide between vegetation canopies and the

atmosphere. IBIS uses a mechanistictreatmentof canopy

photosynthesis
[Farquharet al., 1980;Farquharand Sbarkey,
1982]anda semimechanistic
modelof stomatal
conductance
[Ball
et al., 1986]. Theseformulations
are usedby otherlandsurface
packages
[e.g.,Sellerset al., 1992;Bonan,1995]andhavebeen
testedextensivelyagainstgas exchangemeasurements
[e.g.,
Delire and Foley, 1999].

The photosynthesis
rate of C3 plants,whichincludeall trees
plants,is expressed
followingtheFarquhar
m of the soil. The eightsoil layersin IBIS havetop-to-bottom andmanyherbaceous
thicknessesof 0.10, 0.15, 0.25, 0.50, 1.0, 2.0, 4.0, and 4.0 m. At
equations
[Farquharet al., 1980;Collatzet al., 1991]. The gross
anytimestep,eachlayeris described
in termsof soiltemperature, photosynthesis
rateperunitleafarea,
Ag(molCO2m--2-l),may
volumetricwater contentand ice content[Pollard and Thompson, be expressed
astheminimumof two potentialrates,
1995;Foley et al., 1996]. The IBIS soil physicsmoduleuses
Richards equation
to calculatethe timerateof changeof liquid
An= min(Je,Jc),
soil moisture,andthe verticalflux of water is modeledaccording
to Darcys Law[Campbell
andNorman,1998]. Thewaterbudget whereJE,thelightlimitedrateof photosynthesis,
is givenas
of soilis controlled
by therateof infiltration,evaporation
of water
from the soil surface,the transpirationstreamoriginatingfrom

plants,andredistribution
of waterin the profile. All of these
processes
areinfluenced
by thesoiltextureandamount
of organic
matter within the soil. Moreover, distinct textural differences

(boundaries)
betweenlayersin thesoilareparticularly
influential
in one-dimensional water flow.

2.2. Canopy Physiology

J- t3QP
ci+2F.
"*
)'

(2)

Qpisthefluxdensity
of photosynthetically
activeradiation
(PAR)

absorbed
by theleaf(molquanta
m--2-1),03
is theintrinsic
quantum
efficiency
forCO2uptake
inC3plants
(molCO2Ein--1),
Ci is the concentration
of CO2 in the intercellularair spacesof the

and I-', is thecompensation

pointfor gross
The physiological
processes
involvedin photosynthesis
and leaf(molmol-1),
stomatalregulationdirectlycontrolthe exchangeof water vapor

photosynthesis
(molmol'l).

798

KUCHARIK

ET AL.: TESTING A DYNAMIC

GLOBAL ECOSYSTEM

MODEL

physiologicallybased formulations
of canopy photosynthesis
and conductance

physically-basedmodel of snow
temperature, extension and depth

puddle

15 c_mm____

5O cm

Tsoil
5 OsOiceS..
.

100

cm

Tsoil6 06 0ice6

2:00 cm

physically-basedmodel of soft

'temperature,
soilmoisture
and
soil ice with depth

Tsi17
070ice7

400cm

Tsoi18
08 0ice8

400 cm

Figure 2. IBIS statedescription.The basicstatedescriptionshownhere is carriedthroughthe entire integrated

biospheremodel.

lc-

is the Rubiscolimited rate of photosynthesis,

and dc = kCi is the
CO2 limited rate of photosynthesisat low-CO2 concentrations.
The compensation
pointis assumed
to be zerofor C4 plants.

v(c,- E)

c,++[q])' (3)

Leafrespiration,
Rleaf
(molCO2m--2-1),isdetermined
by

Ko

Rleaf-- ''En,

(5)

deis the Rubisco (enzyme) limited rate of photosynthesis,

where V,. is the maximumcarboxylasecapacityof Rubisco(tool
where T is the leaf respirationcost [Collatz et al., 1991]. The
CO2 m--2 s--l), and and Ko are the Michaelis-Mentenmaintenancerespirationrates of stem and fine root biomassare
coefficients
(molmol'l) forCO2and02, respectively.
In IBIS-I given by
the maximumRubiscocarboxylationcapacity(V,, was predicted
by optimizingthe net assimilationof carbonby the leaf [Haxeltine
and Prentice,1996]. However,IBIS-2, for the sakeof simplicity,
and
prescribesconstantvaluesof the maximumRubiscocapacityfor
the following plant functional types: broadleaf deciduous,
(7)
broadleaf evergreen, needleleaf deciduous and needleleaf
Rroot
-- rootWroot,
if(Tsoil),
evergreentrees,shrubs,and C3 andC4 grasses.
Following Collatz et al. [ 1992], the photosynthesis
rate of C4
in woodyandfine
plants,which includemany tropicaland warm-seasongrasses,is where CstemandCrootarecarboncontained
root
biomass,
respectively,
/
is
a
maintenance
respiration
determinedfrom threepotentialcapacitiesto fix carbon,

Rstem
-- stem&apwoodWstem,if(Tstem)
(6)

coefficient
defined
at 15fC
(0.0125
forstem
sapwood
biomass,

As = min(J,,J.,Jc),

[Sprugel
et al., 1995;Amthor,
(4) 1.250y-i forfinerootbiomass)

wheredi= a4Opis thelightlimitedrateof photosynthesis,

dE= V,,

1984;Ryanet al., 1995],)sapwood

is thesapwood
fractionof the
totalstembiomass
(estimated
froman assumed
sapvelocityand

KUCHARIK

ET AL.: TESTING

A DYNAMIC

the maximumrate of transpiration),and ~(T) is the Arrenhius

temperature
function[Lloydand Taylor, 1994]

f(T)- e

1
(8)

GLOBAL

ECOSYSTEM

MODEL

799

2.3. Phenology

Many plants show an annual cycle of leaf display (e.g.,

deciduoustreesand someshrubsand grasses)and physiological
activity (e.g., evergreentrees). Typically, changesin leaf display
andphysiologicalactivityare triggeredby climaticevents,where

In (8), T is the temperature

(13C)of the appropriate
tissues(stem there are one or more unfavorable seasons. IBIS-2 uses a set of
phenology parameterizationsthat are based on algorithms
temperature,soil temperaturein the rooting zone), Eo is a
developedby Botta et al. [2000], which are calibrated against
temperature
sensitivity
factor,andTois a reference
temperature.
The net carbonbalancefor eachplanttype is calculatedby satellite-basedmeasuresof leaf activity.
plantsis assumedto occurwhen
addingall of the carbonfluxes(includinggrossphotosynthesis Budburstof winter-deciduous
accumulated
growing
degree
days
exceeda certainthreshold.We
andmaintenance
respiration).For eachplanttypetheannualNPP
assume(simplifyingthe approachof Botta et al., 2000) that the
is calculated as
thresholdis 150 growingdegreedays(on a-513Cbase)for shrubs
and grasses,and the thresholdis 100 growingdegreedays(on a

NPP
- (1- r/)f (Ag
--Rleaf
--gstem
- groot
)dt, (9)013Cbase) for winter deciduoustrees.

Leaf fall for trees is

initiated when one of the following conditionsis met; either the
where Agis grosscanopyphotosynthesis
and r/ (0.30) is the averagetemperature(usinga 10-dayrunningaverage)falls below
fractionof carbonlostdueto growthrespiration
[Amthor,1984]. 013Cor is 513Cwarmerthan the coldestmonthlytemperature. For
Stomatalconductance
is simulatedfollowingthe formulationof
grassesand shrubs,leaf fall is initiatedwhenthe 10-dayrunning
Ball et al. [ 1986]andCollatzet al. [ 1991,1992],
averagetemperaturereaches013C. This parameterizationis a
deliberatesimplificationof severalphenologyparameterizations
mAn
that havebeensuggested
in the literature. However,it is possible
to make this calculationmore processorientedby incorporating
the inverserelationshipbetweengrowingdegree-dayrequirements
wheregs.h2o
is the leaf-levelstomatalconductance
of watervapor for budburstandchillingperiodlength[e.g.,Murray et al., 1989].
(molH20m--2s--!)sistheCO2concentration
(molmol-l) atthe
Droughtdeciduousplants(includingtropicaldeciduoustrees,
leaf surface,hs is the relative humidity at the leaf surface shrubs,andgrasses)are assumedto respondto changesin the net
(fraction), and m and b are the slope (nondimensional)and canopycarbonbudget. The modelcalculatesthe mean(usinga
intercept
(molH20m-2 s'l) of theconductance-photosynthesis
10-day running average)net photosynthesis
rate of the plant
relationship. Following Collatz et al. [1991, 1992], the
canopies;
wheneverthe 10-day-mean
photosynthesis
ratebecomes
photosynthesis
andstomatalconductance
submodels
arelinkedby
negative (indicating that leaf respirationnow exceedsgross
consideringthe diffusion of H20 and CO2 between the free
photosynthesis),
the modeldropsleaves.This algorithmcaptures
atmosphere,
theleafboundarylayer,andthestomatal
cavity.
the basicdynamicsof waterstresson the plantcanopiesandhow
In IBIS-1 we scaledphotosynthesis
andtranspiration
from the
plantsmayoptimizethecanopycarbonbudgetduringdroughts
by

gs,h2o
= ss
ms
+b,

(10)

leaf levelto thecanopylevelby calculating

thoseratesseparately reducingtheirleaf area[Linderet al., 1987].
for sunlitandshadedleavesandthenaveraging
themweightedby
the fractionof sunlitandshadedleavesoccupying
the canopy.In
IBIS-2, we usea new and simplerapproachfor canopyscaling. 2.4. Vegetation Dynamics
Our new approach
assumes
thatthe net photosynthesis
withinthe
canopyis proportionalto the absorbedPAR (APAR) within it.
IBIS represents
the vegetationcover as a collectionof PFTs,
where each PFT is characterizedin terms of biomass(carbon in
This assumption
is supported
by severalfield studies[Hiroseand
Werger, 1987; Field, 1983, 1991; Evans, 1989; Kull and Jarvis,
1995] and has a theoretical basis [Field, 1983; Sands, 1995;
HaxeltineandPrentice,1996].
We use our coupled photosynthesis-stomatal
conductance

modelto calculatenet photosynthesis

of a leaf at the top of the
canopy. The net photosynthesis
within the canopyis calculated
by scalingit proportionalto the APAR within it. The vertical
profile of APAR throughthe canopy,calculatedusingthe twostreamapproximation,can be simplifiedto a simpleexponential
functionof leaf areaindex(LAI). The analyticalintegralof this
function,multipliedby photosynthesis
of the top leaf givesus the
canopy-integratedphotosynthesis.We then diagnosea canopyaveragevalue of CO2 concentrationat the leaf surface(C.0 by
usingthe big leaf approach[Arethoret a!., 1994;Arethor, 1994;
Lloyd et al., 1995a,b] and applyinga diffusionequationfor the
leaf level CO2 concentration(not shown) to the canopy level.
Canopy-averagestomatal conductanceis similarly calculated,
using the big leaf approach,by applying (10) to the canopy
level.

leaves,stems,and fine roots) and LAI. The definition of plant

functional types is used to differentiate several important
characteristics, including basic physiolognomy (trees versus
shrubsversusgrasses),leaf habit (evergreenversusdeciduous),
leaf form(broadleafversusneedleleaf),andphysiology(C3 versus

C4 pathway)(Table 1). The geographic

distribution
of eachplant
functionaltype is determinedusing a simpleset of climatic
constraintsthat determinecold tolerancelimits, growing degreeday requirements,
and minimumchilling requirements[Prentice
et al., 1992; Haxeltine and Prentice, 1996; Foley et al., 1996].

In eachgrid cell, a numberof plantfunctionaltypescanexist

simultaneously.For example,a grid cell in the easternUnited
States might include temperate broadleaf deciduous trees,
temperateneedleleafconiferevergreentrees,evergreenshrubs,
deciduousshrubs,and cool grasses.The modelexplicitlyallows

trees,shrubs,
andgrasses
toexperience
different
amounts
of light
and water availability. In addition,plant competitionfor these
two essentialresourcesmay be explicitly representedthrough
shadingand differencesin water uptake[Foley et al., 1996]; the

800

KUCHARIK ET AL.: TESTING A DYNAMIC GLOBAL ECOSYSTEM MODEL

,
V

,
V

,
V

A A 'T',
A

',
A

KUCHARIK

ET AL.: TESTING

A DYNAMIC

nutrients.

This approachprovidesa highly simplified representationof

vegetation dynamics that is appropriate for global ecosystem
modeling. Competition among plant functional types is
characterizedby the ability of plantsto capturecommonresources
(light and water). For example,treesin the uppercanopyare able
to capture light first and therefore shade shrubs and grasses.
However, shrubsand grassesare able to accesssoil moisturefirst
as it infiltrates through the soil. Thus the model can
mechanisticallysimulatethe competitionbetweendifferent plant
forms. In addition, the competitionbetween plant types of the
samebasicform (e.g., trees)is driven by differencesin the annual
carbon balance resulting from differences in phenology
(evergreen versus deciduous), leaf form (needleleaf versus
broadleaf),andphotosynthetic
pathway(C3 versusC4).
IBIS considerseach plant functionaltype in terms of a simple
state description. Each PFT is representedin terms of three
biomasspools:carbonin leaves,carbon in transporttissues(i.e.,
predominantlystems),and carbonin fine roots. Carbon in leaves
may also be expressedin terms of LAI by multiplying by the
specificleaf area (Table 1). Transientchangesin biomasspoolj
of plantfunctionaltype i aredescribedas

Ci,j

oqt--ai/NPPi
- ,
'
'i,j

ECOSYSTEM

MODEL

801

pools. IBIS-2 includes a new soil biogeochemistrymodule

[Kucharik et al., submitted manuscript, 1999] (see Figure 3),
which is similar to other previously documented soil
biogeochemistrymodels [Parton et al., 1987, 1993; Verberneet

competition schememay be further adapted to include mineral

Ci,j_

GLOBAL

(11)

al., 1990].
Following the framework of Verberne et al. [1990], IBIS
allows for the growthof microbialbiomassto be a direct function
of the amount of available substrate (litter return and root
turnover, among other more recalcitrant carbon sourcesin the
soil). Furthermore, IBIS allows for microbial growth (and
turnover)and the rate of nitrogenmineralizationto be dependent
on soil textures (particularly clay content). For example, soils
higher in clay content have the capability to cause increased
physicalprotectionof both the microbialbiomassand soil organic
matter, which can potentially yield lowered rates of nitrogen
mineralizationand higher soil carbon densities[Verberne et al.,
1990]. The microbial biomasspool in IBIS is regardedas the
entire active (highly labile) carbonpool and is kept separatefrom
all other belowgroundsoil carbonpools. By explicitly modeling
the microbial dynamics,simulatedmicrobial biomassvalues can
be comparedto actual field measurements[e.g., Kucharik et al.,
submittedmanuscript,1999].
IBIS allowsmicrobialactivity to be dependenton an Arrhenius
function of physically modeled hourly soil temperatures[Lloyd
and Taylor, 1994] and water-filled pore space(WFPS) [Linn and
Doran, 1984]. These functions are not part of the model
describedby Verberne et al. [1990] and are a modification to

equationsused in the CENTURY model. Additionally, rooting

whereai.jis the fractionof annualNPP allocatedto eachbiomass

profiles describedby Jackson et al. [1996] are used in IBIS to
compartment
and 'i.jis the meanresidencetime (from lossof
biomassthrough mortality, disturbance,and tissue turnover) of
eachbiomasscompartment.
Process-basedglobal terrestrial ecosystemmodels generally
use constantratios of carbon allocation to leaves, roots, and wood
and thereforeare unable to capturethe dynamic natureof carbon
allocation that is driven by changes in nutrient and water
availability during the various stages of plant growth.
Unfortunately, an incomplete understandingof the abiotic and
biotic

influence

on carbon allocation

has made it difficult

to

implementdynamiccarbonallocationinto processmodels[Gower
et al., 1999]. In IBIS-2, we have useda new compositedata set
compiled by Gower et al. [1999] that reports the ratio of
belowgroundNPP to total NPP for major terrestrialbiomes(Table
1). Total NPP values that were calculated using estimatesof
belowground
NPP (e.g., estimatedas a ratioof aboveground
NPP)
were excluded from the data set [Gower et al., 1999].
Furthermore,few estimatesof belowgroundNPP exist for tropical
biomesthussomedegreeof uncertaintyexistswith thesevalues.
The general patterns show that the fraction of annual NPP
allocatedto rootsis high in grasses(55%) and shrublands(35%)
and about 20% for all other forest PFTs, with the exceptionof
boreal conifers(40%). The fractionof NPP allocatedto leavesis
30% for all tree PFTs and 45% for shrubsand grassesin IBIS.
Carbon allocation to wood is calculated from the residual NPP not
allocated to leaves and fine roots.

2.5. SoilBiogeochemistry
In the originalversionof IBIS [Foley et al., 1996] therewas no
explicit belowgroundbiogeochemistry
model to completeflow of
carbon betweenthe vegetation,detritus, and soil organic matter

designatewhere fine rootsand soil carbonis most likely to reside

in the soil profile to 1 m. Theseprofiles allow soil moistureand
temperature values to be weighted by depth accordingly to
coincidewith the approximatelocation(depth)of the majority of
carbon and microbial biomass. Micorrhizae are not explicitly
modeled within IBIS-2.

2.5.1. Flow of carbon and nitrogen. Following Parton et al.

[1987] and Verberneet al. [1990], the main partitioningof carbon
in 1BIS is betweensurfaceand belowgroundcarbonlitter pools
derivedfrom litterfall (leaf turnover,woodydetritus,and fine root
turnover) and the more recalcitrant soil organic matter pools
(Figure 3). Annual litterfall and fine root turnover from the
previous year are divided into equal daily increments (total
detritus productiondivided by total days in year) which are

incorporated
into the litter pools. This approachis an obvious
simplification
to actualfine rootandleaf phenology;
namelyroot
mortality is not a linear function, and future models should

includea more sophisticated

approach.The detritusis divided
between decomposable(DPM, decomposableplant matter),
structural(SPM, structuralplant matter), and lignified (RPM,
resistant
plantmatter)fractions,eachhavinga specifiedC:N ratio
[Kuchariket al., submitted
manuscript,1999].
Leaves, woody debris, and fine root biomass detritus are
treated separately here to divide their respective amounts
accordinglybetweenthe three litter pool compartmentsbased
uponthe C:N ratioof eachresiduetype (leaves,wood,or roots),
and the C:N ratio of eachlitter pool compartment
[Whitmoreet
al., 1997]. IBIS-2 followsthe equationspresented
by Whitmore
et al. [1997] whichcalculatesthe fractionof eachresiduetypein
RPM first,thendetermines
theamountof detritusin theSPM pool
usingthe fractionallocatedto the RPM pool, with the residual

802

KUCHARIK

ET AL.: TESTING A DYNAMIC

GLOBAL

ECOSYSTEM

MODEL

Figure 3. Schematic
of the IBIS soilbiogeochemistry
submodel.The followingsymbols
are usedto identify
carbonpools:DPM, decomposable
plantmatter;SPM, structural
plantmatter;RPM, resistant
(lignin)plantmatter;
POM, protected
slowpoolof organicmatter(OM); NOM, nonprotected
slowpoolof OM. Linesandarrowsdrawn
between
carbonpoolsrepresent
flowsof carbon,whilethickarrowsshowmicrobialCO2respiration
occurring
with
the oxidation of carbon.

detritus assumedto reside in the DPM pool. Following this

approach,eachplantmatterpool (RPM, SPM, andDPM) contains
someamountof leaf,fine root,andwoodydetritus.
IBIS has an abovegroundstandinglitter pool of accumulated
decomposableand structuralleaf and woody biomassat the soil
surface. Similarly, there is a belowground litter pool of
decomposing
fine rootbiomass.Generally,theselitter poolshave
the shortestresidencetimesof any carbonpool in the model,of
the orderof weeksto months[Parton et al., 1987]. Litter in the
DPM and SPM carbon pools moves directly to the microbial
biomassfor decomposition,while the lignified plant biomass
(RPM) is transferredto the soil organicmatter [l/erberneet al.,
1990]. The belowground soil carbon is divided into four
compartments
(Figure 3): (1) microbial biomasscomprisesthe
activecarbonpool with a residencetime varyingfrom a few hours
to months, (2) protected and (3) nonprotectedorganic matter
comprise slow carbon pools [ Parton et al., 1987], with
residencetimesof 10-30 years,and (4) stablizedorganicmatter,
representing
the mostrecalcitrantcarbon,typicallywith residence
times of more than 1000 years [Verberne et al., 1990]. The
carbonto nitrogenratiosfor the variouscarbonpoolsreportedby
l/erberne et al. [1990] are used in our model and are assumedto
remainconstant(Table 2). As carbonin litter and soil organic
matter decomposes,it is partitionedbetween microbial biomass
andrespirationaccordingto the microbialefficienciesassigned
to
eachtransformation
(Table2).

The variablesoutputby the model includetotal soil carbonand

nitrogen, aboveground and belowground litter carbon, total
microbial biomass, soil surface CO2 flux, nitrogen
mineralization/immobilization,and organiccarbonand nitrogen
leaching. Amounts of atmospheric nitrogen deposition and
nitrogenfixation (both functionsof annualprecipitation)are also
determinedin IBIS usingempiricalequationsfoundin
CENTURY [Parton et al., 1987].
2.5.2. Decomposition processes. Soil organic matter
decompositionis simulatedusinga daily time step;hourly values
of modeled soil temperatureare converted into daily average
values. The base decay rates of litter, root, and soil carbonand
microbial biomass turnover (defined at 1513Cand 60% WFPS;
Table 2) are modified by functionsof soil temperatureand soil
moisture. Modifying factors, which are used to simulate the
temperatureand moisture effects on microbial activity, are

calculatedusing two different sets of soil temperatureand

moisture
values.Onesetof modifyingfactorsinfluences
therate
of litter decompositionat the soil surfaceand are basedon the
dailyaverage0-10 cm soiltemperature
(topsoillayerin IBIS) and
0-10 cm soil moisture.

The second set of factors affects the

of

turnover rate of microbial biomass and the transformation

belowground
soil carbonbetweenindividualpools (fine root
detritus,active (microbial),slow, and passivepools). These
factorsare basedon the 0-100 cm soil temperatureand soil
moisturevalues,whicharedepthweightedby the fractionof total

KUCHARIK

ET AL.: TESTING A DYNAMIC

GLOBAL ECOSYSTEM

MODEL

803

Table 2. BaseRate Constants

of CarbonDecomposition,
MicrobialEfficiencyFactorsandC:N RatiosUsedWithin
theIBIS Soil Biogeochemistry
Model

Transformation/Pool

Decomposition
Rate,d'

Microbial

C:N

Efficiency

Pool

DPM

leaf litter
fine roots
wood

0.15
0.10
0.00

0.40
0.40
0.40

6
6
6

0.01
0.005
0.001

0.30
0.30
0.30

150
150
150

0.01
0.005
0.001

1.0
1.0
1.0

100
100
100

0.005
0.045

1.0
1.0

10
10

SPM

leaf litter
fine roots
wood
RPM

leaf litter
fine roots
wood

Pbio ==1>
Po,,
Nbio ==I}Non,

Po,,:= Pbio

1.0x 10'4

0.25

10

No,,

0.001

0.20

15

1.0x 10'6
1.0x 10'6
8.0x 10-7

1.0
1.0
0.20

15
10
15

Nbio

Pon,:= So,,
No,,:= So,,
Son,

The C:N ratio of litterfall and fine root turnover in IBIS is 40 for leaves, 200 for wood and 60 for fine roots,

respectively.
Here 1 - efficiencyis the fractionof carbonoxidationthat is releasedas CO2. Symbolsare definedto
denotethe following C pools: DPM, decomposable
plant material;SPM, structuralplant material;RPM, resistant
(lignin)plantmaterial;Pbo,protected
microbialbiomass;
No, nonprotected
microbialbiomass;Pon,,protectedslowC
pool;No,,,nonprotected
slowC pool;Son,,
stabilizedpassiveC pool.

root biomassin each soil layer followingJacksonet al. [1996].

Carbonaddedto the soil throughthe decomposition
processis not
partitioned
with respectto soil layers(e.g., decomposition
of fine
root biomassis not kept track of for each individual soil layer);
instead,thetotalamountof decomposed
carbonis aggregated
for
the total soil profile.
The soil carbon decompositionfactors used in IBIS were
determinedfrom a simplecalibrationprocedure.First, the model
decayconstants
weremodifiedso thatthe simulatedglobaltotals
of soil carbon(to 1 m total depth)were closeto reportedestimates
of 1200-1600Gigatons(Gt) [e.g., Post et al., 1982; Zinke et al.,
1986; Schlesinger,1991; Eswaran et al., 1993; Sombroeket al.,
1993;Batjes,1996]. Then,the annualamountof leachedorganic
carbonwascalibratedto fall within the rangemeasuredat various
natural and managed ecosystemsin Wisconsin (leaching was

grid,drivenby climatological
dataandsoilboundary
conditions.
The model simulationwent throughan initial 400-year spin-up

procedure
wheresoil carbon,vegetationstructure,and biomass
wereallowedto cometo an equilibriumstateassumed
to existin
1860. Duringthisinitialspin-upperiod,the soilbiogeochemistry
and vegetation dynamics were both subjectedto separate
numericalaccelerationprocedures. The vegetationdynamics
were assumed to occur at 4 times the normal rate (plant

competitionfor light and water, NPP, carbonallocation,and

mortality)duringyears20-100of spin-upand2 timesthenormal
rateduringyears101-150.
The initial 400-year model spin-upassumeda preindustrial
concentration
of atmospheric
CO2 (286.6 ppmv)and was driven
with 1860-1899climatologyof mean monthlytemperatureand

precipitation (M.

Heimann, Max-Planck-lnstitut fuer

typically
between
0.001-0.005
kg C m-2yr-l; seeBryeet al. Biogeochemie,
personalcommunication,
1999),and 1931-1960
[1999]). Thesesimulatedvaluesare also representative
of both
temperateuplandforestsand swampsand lowlandforests,which

climatologyof Leeroans
and Cramer[ 1990]for monthlyaverage
daily temperature
range(Tmax-Tmin),the numberof dayswith
havereported
carbon
leaching
rates
of 0.001-0.014
kgC m-2yr-1 precipitationeach month and potential sunshinehours (W.
[Landsbergand Gower, 1997].
Cramer,personalcommunication,1999). Absoluteminimum
temperature
data,usedto determine
theextremegeographic
limits
of severalplanttypeswithinIBIS, wereobtainedfromP. Bartlein
3. Model Implementation
(personalcommunication,1998). Monthly-meanestimatesof
relative humidity and wind speed were compiled from the
We presentthe resultsof a 135-yearmodel simulation(18601994) performedon a 113latitudeby 113longitudeglobal terrestrial National Center for EnvironmentalPrediction(NCEP)/National

804

KUCHARIK ET AL.' TESTING A DYNAMIC GLOBAL ECOSYSTEM MODEL

Center for Atmospheric Research (NCAR) reanalyzed

meteorologicaldataset[Kalnay et al., 1996]; we averagedthe
NCEP/NCAR

data

from

1958

to

1997

to

determine

the

climatological values. The simulationbegins with zero snow

cover and uniform soil temperature(5BC) and soil moisture(50%
of pore space). All soil carbonand litter pools also have zero
carbonstoresto begin, with the soil carbonhaving reachedan
equilibriumvalue by the end of the model spin-up(year 1860)
usinga numericalacceleration
procedurethat is discussed
below.
After spin-up,the model was run for 135 yearsusingtransient
climateand atmosphericCO2 data from 1860-1994(M. Heimann,
Max-Planck-Institutfuer Biogeochemie,
personalcommunication,
1999) and Leeroansand Cramer [1990] climatologyfor other
meteorologicaldrivingvariables.
3.1. Weather

Generator

IBIS uses a 60-min time step to calculate land-atmosphere

exchange processes.To meaningfully convert monthly-mean
climate data to this time step, we use a stochastic weather
generatorbased on Richardson[ 1981], Richardsonand Wright
[1984], and Geng et al. [1985]. The Richardson-styleweather
generatorstochasticallydeterminesthe probability of weather
events (on the basis of long-term climatic means). The
interrelationships
betweenthe randomvariationsin atmospheric
parameters(e.g., perturbationsin temperature,precipitation,and
cloudiness)and their persistenceover time are determinedfrom a
serial auto-correlativeapproach. This ensuresthat a reasonable
sequenceof atmosphericparameters,mimickingthe day-to-day
behaviorof synopticweatherevents,is providedto the model.
Spatialcorrelationbetweengrid cells, however,is not accounted
for on hourly to daily timescales.Furthermore,the Richardsonstyle weather generator explicitly considers the statistical
distributionof atmospheric
parameters,so extremeweatherevents
can be generatedwithin the model. This is criticalto simulating
the surface water balance, where extreme rainfall events often

producea largefractionof therunoff.

3.2. Soil

Data Sets

We adopteda new globaldataset [IGBP-DIS, 1999]of soil

textureto representsoil conditions
acrossthe globe. Usingthe
SoilData softwarethat was includedwith the data set [IGBP-DIS,

1999],thefractionof sand,silt,andclayweredetermined
for each
IBIS soillayer. For soildepthswheretherewasno defaultdata,
the valueof the layerimmediatelyabovewasused. This allowed
eachIBIS gridcell to havea representative
fractionof sand,silt,
andclay as a functionof depthin the soil. This is a significant
improvement
overtheoriginalsoildatasetusedby IBIS [Zobler,
1986],whichdidnotcontainsoiltextureasa function
of depth.
To determinethe hydraulicandphysicalproperties
of the soil
layers, we classifiedsoil textureclassesinto 11 categories
followingRawIset al. [1992] andCampbellandNorman[1998]
(seeTable 3). IBIS matcheseachsoil layerof eachgrid cell to the
mostcomparableof the 11 soil texturecategories.Soil properties

assigned
to eachsoilcategory
include
porosity
(m3 m-3),field
capacity
(m3m'3),wiltingpoint(m3m'3),saturated
matric(airentry) potential(m H20), and the saturatedhydraulicconductivity

(m s-i). Additionally,
following
Campbell
andNorman
[1998],
thedensity
ofsoilmaterial
(kgm'3)iscomputed
by

KUCHARIK

ET AL.: TESTING A DYNAMIC

2650(1 --forganc
4-1300forgan,c,

(12)

andthespecific
heatofsoilmaterial
(Jkg-K't) isgiven
by
870(1 'forgan,c)
4-1920forganic,

(13)

whereforgamc
is the fractionof organicmattercontentin the soil.
In this versionof IBIS, for simplicity,it is assumedthat all soil
layers contain 1% organic carboncontent in these calculations.
However, it is possibleto modify the physical propertiesof the
soil in terms of the soil organic content, thus allowing soil
hydraulicpropertiesto evolvewith changesin soil organicmatter.
This could have important applicationsin comparing different
land use practices,where differencesin soil organic matter can

causedifferencesin key soil properties(e.g., density,water

holdingcapacity,andheatcapacity).
3.3. Numerical Accelerationof Soil Carbon Dynamics

The amountof soil carbonandnitrogencan be initializedusing

observedquantitiesof soil carbon or zero storesof carbon and
nitrogen. One notableproblemwith initializing the model with
actual data or current measurements

of total litter and soil carbon

densitiesis that thesetotal quantitiesare partitioned into several

compartments in IBIS (like numerous other biogeochemistry
models) (Figure 3). Soil carbonmeasurementsgenerally do not
distinguishbetween active, slow, and passive carbon pools or
distinct litter quality categories. Therefore an initialization
procedurewould be difficult to implementwithout causingsome
degree of uncertaintyin the subsequentbiogeochemicalcycling.
Consequently,for model resultsreportedin this study, the soil
storageof carbonandnitrogenwasinitializedat zero.
During a typical IBIS model run, a numerical acceleration
procedureis usedto simulateapproximately5000 yearsof actual
(geologic)time during a 150-yearmodel integration. To do so,
the soil biogeochemistrymodule in IBIS, which operateson a
daily time step, is called 40 times each day with constantsoil
temperature and moisture (instead of performing just one
execution)duringthe first 100 yearsin a model run. During the
next 50 years of the numerical acceleration(model years 100150), a linear decreasein the degreeof accelerationoccursuntil
the biogeochemistrymodel is only called one time each day
(normal operation)at year 150 (comparedto 40 times at year
100). This progressive decrease applied to the numerical
accelerationover time allows for the biogeochemicalcycling to
adjustgraduallyto an equilibriumstateinsteadof implementing
an abrupt discontinuationof the numerical accelerationat year
150.

GLOBAL ECOSYSTEM

MODEL

805

or CO2 concentrations
did not occur. In the followingsectionswe
presentand evaluatethe resultsof the model in terms of a wide
range of phenomena, including hydrological processes, the
terrestrialcarbonbalance,andglobalvegetationpatterns.
4.1. Terrestrial

Water

Balance

The hydrologiccycle is of critical importanceto climate and

globalecosystem
processes.Delire and Foley [ 1999] havetested
IBIS-2 againsta variety of detailed biophysicalmeasurements,
including evapotranspirationand soil moisture, made at sites
acrossthe globe. However, these site-specificmodel exercises
cannot fully address the ability of the model to capture
hydrologicalprocessesat continentaland global scales. In order
to test the model at these scales, we need to use data collected at

fundamentallylargerscales. Currently,we usespatiallyextensive

terrestrial hydrological data consistingof a network of river
gaugesto evaluatehydrologicalprocesses
in IBIS.
We compare the global patternsof simulatedrunoff to river
gaugeobservations
compiledby Cogley[1991]. Cogleydigitized
the United Nations Educational, Scientifi'c, and Cultural
Organization(UNESCO) [1978] global runoff maps,which were
compiledfrom over 1000 long-termriver gaugemeasurements
takenfi'omaroundthe world to providea preliminaryestimateof
global runoff patterns. IBIS simulatesthe global patternsof
runoff quite well comparedto observations(Plate 1). The
pronouncedrunoff peak in the tropics is particularly well
captured,as are the weaker naxima in midlatitudesand the dry
belts of the subtropics.The only systematicbias in the model
resultsappearsin the high northemlatitudes,where precipitation
andriver dischargedataare still questionable.
In our work with IBIS-2, we have tested the water balance

budgetof the modelagainstthreesourcesof data:(1) site-specific

biophysicalmeasurementsfrom meteorologicaltowers [Delire
and Foley, 1999], (2) global compilations of river gauge
networks,and (3) continental-scale
compilationsof long-termsoil
moisture measurements. However, this is still only a cursory
evaluationof the landsurfaceandhydrological
processes
involved
with terrestrialecosystemprocesses.Much more attentionshould
be paid to this aspectof global ecosystemmodels,but improved
datasetsof globalhydrologicalprocesses
are needed.
4.2. Net Primary Productivity (NPP)

For individual 113x 113grid cells, model estimatesof annual

average
NPPrangefrom0.0 to 1.5kg C m'2 yr' (Plate2a).
Averaged
foreachof the1-5basicvegetation
typesdefined
by
IBIS,NPPranges
from0.001kgC m'2yr' indeserts
to0.90kgC
m'2 y-r in tropicalevergreen
forests
(Table4). Theseresults

4. Results

obviouslyshowthat simulatedNPP is highestin warm, moist

tropicalclimates
andlowestin subtropical
deserts
or polarregions
We presentthe long-term characteristicsof the simulation, experiencing
coldanddryconditions
throughout
theyear.
basedon a 30-yearaverageof modelresultsbetweenyears1965
Theglobal
totalannual
NPPisapproximately
54.3GtC yr't
and1994.Theseresultsareintended
to represent
theaveragestate andis 47.3% of the annualgrossprimaryproduction(GPP) (114.7
of present-day
ecosystems.
Because
themodeldoesnotexplicitly Gt C yr'), whichfallswithintherangeof measured
NPP/GPP
includeperiodicdisturbance
mechanisms
(e.g., fire, wind, and ratiosreportedby Waringet al. [ 1998]. Table4 alsoshowsthat

herbivory),a stateof equilibrium(wherevaluesof NPP and soil

GPPbiomeaverages
range
from1.89kgC m'2yr
- fortropical

carbonalongwithvegetation
structure
donotchangesignificantly evergreen
forests
to0.020kgC m'2yr-within
desert
regions.
For
overtime)wouldbe attainedif atmospheric
CO2andclimatewere majorecosystem
types(tropical,boreal,temperate,andgrasses)
assumedto remainconstant. In naturalecosystems,
however, the ratio of NPP/GPP varies from 10% in desertsto 30% in open
where disturbanceis part of the system,a true steadystate shrublandsand savannasand up to 55% in tundraand temperate
conditionmay neverbe reachedevenif otherchangesin climate broadleafevergreenforests.

806

KUCHARIK

ET AL.' TESTING A DYNAMIC

GLOBAL ECOSYSTEM

MODEL

mmb/ear
O- 50
50-100
100- 200
200 - 300
300- 500
500- 750
750- 1000
1000-2000
> 2000

mmear
0 - 50
50- 100
100 - 200
200 - 300
300 - 500
500- 750
750 - 1000
1000 - 2000
> 2000

1200

----Cogley

Simulated

yrs 176-200

ooo

800 .......
600
400

2oo
in,-.,,

latitude

Plate 1. Globaldistributionof (a) simulatedrunoff(averageof years 1965-1994),(b) observedrunoff [Cogley,

1991], and(c) comparison
of zonalmeansof simulatedandobservedrunoffversuslatitude

KUCHARIK ET AL.' TESTING A DYNAMIC GLOBAL ECOSYSTEM MODEL

kg/m/yr

0 - O.1

0.4- 0.5 ................

J0.8 - 0.9

0.1 - 0.2

0.5 - 0.6 ::
.......
] 0.9- 1

0.2 - 0.3

0.6 - 0.7

I - 1.1

0.3- 0.4

0.7- 0.8

> 1.1

Plate2. Distribution
of (a)simulated
annual
netprimary
productivity
(NPP;kgC m-2yr'f) and(b)observations
compiledby S.T. Gower(unpublished
data, 1999).

807

808

KUCHARIK ET AL.: TESTING A DYNAMIC GLOBAL ECOSYSTEM MODEL

To evaluateIBIS simulationsof annualNPP, we comparethe

model resultsto a datasetof field measurementscompiled at the
University of Wisconsin [S.T. Gower, unpublisheddata, 1999]
(Plate 2b). NPP data were compiled from three major sources:
Esser et al. [1997], Cannell [1982], and individual NPP data
complied by S.T. Gower [unpublisheddata, 1999]. These data
were processedto remove redundancyamong the data sets and
quality screenedby removingany value that differed more than 3

shortestresidencetimes (Table 4). Temperateforestsand boreal

forestsgenerally have the longestresidencetime of carbon in
biomass, around 10-25 years and slightly higher in a few
locations;however,this is a difficult valueto measuredirectly.
We also compare simulationsof fine root biomassto data
compiledby Jacksonet al. [1996] (Table 5). Simulatedvaluesof
the annual averagelive fine root biomassin IBIS range from a

s.d. from the mean biome value.

temperateneedleleafevergreenforests(Table 4, Table 5); tropical

and boreal forestsgenerallyhave simulatedvaluesbetween0.08

NPP was calculated as the sum

of wood (stem + branch) + foliage production. Many studiesuse

leaf litterfall as a surrogatefor foliage production. Only a small
fraction (12%) of the studies have belowground NPP data; to
standardizethe data we use the aboveground:belowground
NPP
ratiosfrom Gower ct al. [1999] to approximatetotal NPP. In this
comparison,we usea total of 1882 measurements
madeat various

lowof 0.004kgC m'2 in deserts

to a highof 0.200kgC m-2in
and0.20kgC m-2. Therangeof livefinerootbiomass
reported
byJackson
et al. [1996]is 0.07-0.47kg C m'2. Whenbiome-

averageIBIS resultsare comparedto equivalentdatacompiledby

.lacksonet al. [1996], temperate and tropical forestsgenerally
agree to within 10-25%. The largestdiscrepancybetweenmodel
sites.
simulationsand measurements
(outsideof desertregions)existed
biome, which yieldeda difference
In order to describebiome-averagevariationsin NPP, the 1882 with the IBIS grassland/steppe
of about71% (Table 5).
field measurementsare categorized into one of the 15 IBIS
The biome classificationsdiffer betweenJacksonet al. [1996]
vegetation types (Table 4). However, not all of the 15 IBIS
vegetation types had associated NPP field measurements and lB IS makinga direct one-to-onecomparisondifficult. Only
(savanna, polar desert, and open shrublandswere omitted), one boreal forest category (including both deciduous and
leaving12 biomesfor comparison(Figure4). There is reasonable coniferous)was presentedby Jacksonet al., and grasslandswere
agreementbetweenthe simulatedand observedNPP acrossmost partitionedinto temperateand tropical categories. Furthermore,
biomes. Simulated and measuredNPP differed on average by
no distinctionis made in the Jacksonet al. data set betweenopen
and dense shrublandtypes. For the purpose of this study, the
39% for the 12 biomes in which comparisonswere made; the
Jacksonet al. temperategrasslandbiome is comparedto the IBIS
smallestdifferencewas 4.0% for temperatedeciduousforests(403
biome, and the Jacksonet al. tropicalgrasslandis
measurements),while the greatestdifference(within a nondesert grassland/steppe
area) was 89% for boreal mixed forests (44 measurements). compared to the IBIS savanna biome. The shrub data are
Much of the large discrepancy between measurementsand compared to the dense shrubland biome category in IBIS.
simulationsin boreal mixed forestscouldbe explainedby the fact Jacksonet al. did not present data for mixed forests. For those
that the IBIS

mixed forest biome classification

includes mixed

biomes

that do correlate

between

IBIS

and the Jackson et al.

categories(tropical and temperateecosystems),generally good

agreementexists. However, it is obviousthat somerather large
differencesexist in tundra and desertecosystems(40-75%), and
IBIS greatlyunderestimates
(by 71%) grasslandfine rootbiomass
simulations and measurements.
The IBIS model simulates the
reportedby Jacksonet al. This large differencemay be partly
averageNPP over a homogeneous
113x 113grid cell, while field due to IBIS having an oversimplified root turnover
fine rootsare assumedto only turnoverone time
measurements
of NPP are generallymadeon individualallometric parameterization;
studyplots,sometimesas small as 10 m on a side. Furthermore, eachyear. We note howeverthat the simulatedvaluesin Table 5
belowgroundNPP measurementsare exceedinglydifficult to are annualaverages,while the datacompiledby Jacksonet al. are
from specific point measurementsmade during different time
make and are sometimesestimatedon the basis of assuminga
value.
fixedratiobetweenbelowground
to aboveground
production
(e.g., periodsandmay not representa true annual-average
50:50contribution
to totalNPP).
forestsfrom all regionsof the globe.
While we believe that this data set represents the most
completeand accurateset of globalNPP measurements
currently
available, caution should still be taken when comparingmodel

4.4. Leaf Area Index

4.3. VegetationBiomass

Plate4a showsthe distributionof LAI (definedas single-sided,

projectedleaf area)for trees. The LAI of treesrangefrom 1.0-4.0
The distribution of vegetation biomass (including both in boreal and tundraregionsto 4.0-8.0 in temperateconiferous
abovegroundand belowgroundcomponents),divided into trees and deciduousforeststo a high of 8.0-11.0 in tropicalrainforests.
and shrubs/grasses,is highlighted in Plates 3a and 3b. In
For forestedregionsin IBIS the biome-averageLA! rangesfrom a
individual l g x 113grid cells, the simulatedtotal carbon density high of 6.8 in tropicalevergreenforeststo a low of about2.7 for
ranges
from0.01kgC m'2in subtropical
deserts
to 14.1kgC m'2 boreal deciduousforests (Table 4). Values of tree LAI in
in the core of tropical rainforests. The biome-averagebiomass savannas,grasslands,shrublands,deserts,and tundraregionsare
density
ranges
from0.03kgC m'2in desert
ecosystems
to9.1kg all lessthan 1.0. These LAI valuesfor main forestedregionsare
C m'2in tropical
evergreen
forests
(Table4). Theglobalstorage realisticand comparereasonablywell to measurements
madein
several forested areas.
Measurements
made in boreal forests
of carbonin vegetationbiomassis 557.4 Gt C.
We also show the mean residencetime of carbon in living
during the Boreal Ecosystem-AtmosphereStudy (BOREAS)
biomass (Plate 3c), estimated as the ratio of equilibrium
[Gower et al., 1997; Kucharik et al., 1997, 1999; Gower et al.,
vegetationbiomass to NPP. Tropical rainforests,which have 1999] showthat LAI was approximately2.0 in jack pine forests,
relatively large storesof carbonin vegetation,generallyhave the 3.5 in aspen, and 6.0 in black spruce stands. LA! has been

KUCHARIK ET AL.' TESTING A DYNAMIC

GLOBAL ECOSYSTEM MODEL

809

1.6

1.4

'1.2

O.8
,", 0.6

0.4

0.2

Figure4. Comparison
of simulated
versus
observed
NPPvalues
(kgC m'2yr'l) [S.T.Gower,unpublished
data,
1999] reportedon a biome-by-biome
basis(averageand+/- 1 s.d.are shown).The averageandstandarddeviations
of modeloutputreflectvariationsoccurringwithineachspecificbiomeoverthe 30-yearperiod(years1965-1994)
andare not intendedto showyear-to-yeardifferences.

measuredto rangefrom 1.3 to 8.4 in temperateconiferousand 4.0) preventsignificantquantitiesof light from reachingthe forest
floor,thusinhibitingthedevelopment
of grasses
andshrubs.
[Fassnacht and Gower, 1997; Landsberg and Gower, 1997;
A plot of severalNPP versusLAI measurements
for different
Gower et al., 1999]. LAI can reachvaluesgreaterthan 10 in foreststandsare shownin Figure 5 along with all biome-average
someevergreenconiferousforestsof the Pacific Northwest in the values simulatedby IBIS. The compiledmeasurementdata are
United States[Landsbergand Gower, 1997]. LAI of tropical from the following sources:Fassnachtand Gower [1997], who
forests is also variable. LAI has been measured at 5.1-7.5 for
report abovegroundNPP versusLA! for evergreenneedleleaf,
deciduous forests of the Midwestern and eastern United States

lowland tropical rainforestin Venezuela[Saldarriaga and

broadleaf deciduous, and mixed forest stands in northern

Luxernoore, 1991], 6.6 for a lowland rainforest in Puerto Rico

Wisconsin;Runyonet al. [1994], who report abovegroundNPP

[Jordan, 1971], 5.5 for an evergreenmontane rainforest in

Jamaica,7.5 for a lowlandevergreen
rainforestin Malaysia[Kira,
1978], and 12.3 for a lowlandevergreenrainforestin Thailand
[Kira et al., 1964]. Robertset al. [1996] showedthat the LAI
valuesat two sitesin the Amazonforest(Manausand Ji-Paran)
were6.0 and4.7, respectively.
The simulatedLAI of grassesand shrubsrangesfrom 1.0-3.0

in tundraand boreal regionsto 3.0-6.0 in the grasslandand

savanna regions of northern Africa, Australia, and the Great

Plainsregionof theUnitedStates(Plate4b). Somelocallyhigher

LAIs are apparentin Africa and Australia, where some values

approach8.0 to 10.0. The biome-average

grassandshrubLAI in
nonforest
biomesarehighestin savannas
(3.3) andobviously
very
low in desertareas (0.10) (Table 4). In all forest biomesthe
averageLAI of grassesand shrubsis generallylessthan 1.0. A
significant,negativecorrelationexistsbetweentree LAI and shrub
and grassLAI in forestbiomesin IBIS. The generalpatternis
thathigherLAI foreststhatachievecanopyclosure(LAI > about

versus LAI

for the Otter

transect

in the Pacific

Northwest

(Oregon);andSaldarriagaandLuxrnoore[1991], who showNPP

(aboveground
andcoarseroot) versusLAI tbr a tropicalrainforest
age sequencein Amazonia. While it is apparent that IBIS
simulationsand measurement
datacomparequite well for several
ecosystems,
somerather interestingpatternsassociatedwith the
measureddataare worthnoting. In particular,in both the Runyon
et al. [1994] and Saldarriaga and Luxrnoore [1991]
measurements,
data coincidewith the generalIBIS relationship
until reachinga LAI between4.0-5.0 (generally when canopy
closureoccursin a canopy with randomlydistributedfoliage
elements). However, after canopyclosure,NPP per unit LAI
declinessignificantly(Figure 5), which may be related to a
decline in NPP occurring with increasing stand age. The
interestingrelationshipsbetween NPP and LAI in both model
resultsand measurementsdeservemore investigationin future
studies.

IBIS assumesthat foliage in any grid cell (canopy) has a

g10

KUCHARIK ET AL.' TESTINGA DYNAMIC GLOBALECOSYSTEMMODEL

KUCHARIK

ET AL.: TESTING

A DYNAMIC

GLOBAL

ECOSYSTEM

MODEL

kg/m2
0-0.1
0.1 - 0.25
0.25 - 0.5
0.5- 1
1-2
2-3
3-4

6-7

?-9
9-11
11 -13

kg/m2
0-0.1
0.1 - 0.25
0.25 - 0.5
...... 10.5- 1
1-2
2-3

]3-4
..

15-6
6-7
7-9
....9-11
11 -13

years
0-1
1-3
3-5
5-7

]7-10
10-15
15-20
20- 25
>25

Plate3. Distribution
of simulated
biomass
(kgC m-2)of(a)trees
and(b)grasses
andshrubs.
(c)Simulated
mean
carbonresidence
time(years)in biomasscomputed
asthetotalvegetation
biomassdividedby annualNPP.

81 1

812

KUCHARIK

ET AL.: TESTING

A DYNAMIC

GLOBAL

ECOSYSTEM

MODEL

Table 5. SimulatedAnnualAverageLiving FineRootBiomassComparedto Data Compiledby dacksonet al., [1996]

on a Biome-by-BiomeBasis
IBIS Biome

IBIS,

dacksonet al., [1996],

kgC m'2

kgC m'2

Obs

Tropicalevergreenforest
Tropicaldeciduous
forest
Temperatebroadleafevergreenforest
Temperateneedleleafevergreenforest
Temperatedeciduous
forest

0.178
0.110
0.164
0.199
0.163

(0.035)
(0.031)
(0.047)
(0.039)
(0.032)

0.161
0.137
......
0.244
0.214

12
6

Borealevergreen
forest
Borealdeciduous
forest

0.127
0.080

(0.034)
(0.024)

0.112
0.112

Evergreen/deciduous
mixedforest

0.143

(0.040)

......

Savanna

0.133 (0.080)

0.249

Grassland/steppe

0.135

(0.075)

0.464

21c

Dense
shrubland

0.075 (0.032)

0.136

6a

Openshrubland
Tundra

0.034
0.096

(0.022)
(0.031)

......
0.166

Desert
Polardesert/rock/ice

0.004
0.007

(0.007)
(0.005)

0.063
......

10
14

5a
5a

5
4

The column"Obs" denotesthe numberof field measurements

usedto performaveragingfor thosebiomes.The

averageand+/- 1 s.d.areshownfor IBIS output.Missingdata(dashes)

denotenomeasurements
madein locations
fallingwithinthosebiomeclassifications.
The averageandstandard
deviations
for modeloutputreflectvariations
occurring
withineachspecificbiomeoverthe30-yearperiod(years1965-1994)andarenotintended
to showyearto-yeardifferences.
aNo distinction
is madebydackson
et al. [ 1996]betweenborealdeciduous
andconifers.

bIBISsavanna
biome
iscompared
totropical
grassland
biome
bydackson
etal. [1996].
cIBIS grassland/steppe
biomeiscompared
totemperate
grassland
biomebydackson
etal. [ 1996].

aIBISdense
shrubland
category
iscompared
tosclerophyllous
shrubs
andtreesbydackson
etal. [1996].

randomdistribution,thus the clumping of foliage that generally

occurs in boreal forestsand influencesradiation penetrationis
ignored[Kuchariket al., 1997; Chen et al., 1997]. This is a

A more meaningful way to evaluate simulationsof global

vegetationstructurewould be to directly comparevegetation

significantlimitationto manyecosystem
models.For example,in
someblack spruceforestsof centralCanada,forestsstandswith
LA! greaterthan4.0 only achieve60-70% canopyclosuredue to
clumping [Kucharik et al., 1999]. Future ecosystemmodels
shouldexplicitlyconsiderclumpingwithinthecanopy.

example,DeFries et al. [1999] have produceda satellite-based

data set which delineatesthe fraction of vegetation cover in
evergreenbroadleaf trees, evergreen conifer trees, deciduous
broadleaftrees,deciduousconifer trees, herbaceous
plants,and
bare ground. This data set is directly comparable to our
simulation of vegetation structure, without any semantic

4.5. Vegetation Structure

differences in the description of vegetation cover. However,

satellite-basedmaps representcurrent vegetationnot potential

It is often convenient to categorize global patterns of

vegetationstructureinto a simplelist of vegetationtypes(Plate
4c); this allows for an easily interpreted,visual inspectionof
global vegetationgeography. However, while these global
vegetationmaps are qualitatively useful, it is difficult to
meaningfullycomparesimulatedand observedmapsof global
vegetation
patterns.Manyof thedifferences
arelargelytheresult
of differencesin the definitionof vegetationtypes. For example,
there are significant semantic problems in defining many

structural

attributes

with

satellite-based

measurements.

For

vegetation. Future satellite-basedaltimetry measurements

of
vegetationheightandbiomasscouldalsoprovidea powerfuldata
setfor validatingglobalvegetationmodels.
To more quantitativelydescribe the global patternsof
vegetation structure, we calculate the fractional cover of
evergreentrees,deciduoustrees,grassesand shrubs,and bare
ground(Plate5). Thisquantityis definedby the fully projected
foliagecover,approximated
by 1 --exp(-0.5eLAJ, where0.5 is
the
canopy
extinction
coefficient.
This equationis alsodirectly
vegetationtypes,includingsavannas
versusgrasslands,
dense
relatedto the PAR interceptedby the canopy. The fractional
shrublandsversus open shrublands,and mixed forests. A
treesis highest(>90%) in thetropicalareasof
comparison(not shown)of ground-based
and satellite-based coverfor evergreen
vegetation
maps[e.g.,Ramankutty
and Foley, 1999;DeFrieset South America, Africa, and SoutheastAsia (Plate 5a). These
havecorrespondingly
high LAI values(>
al., 1998;Haxeltineand Prentice,1996; Matthews,1983; Olson, areas,not surprisingly,
1983]suggests
thatthereare majordifferences
betweenexisting 4.0; Plate 4a) typical of closedforestcanopies. Evergreentrees
mapsof globalvegetationpatterns,therebycomplicating
direct alsocomprisethe majorityof the vegetationcover(60-90%) in
forestedareasof northernEurope throughwest central Russia,
model-to-data
comparisons.

KUCHARIK ET AL.' TESTING A DYNAMIC

GLOBAL ECOSYSTEM MODEL

813

1.6
IBIS-2

1.4

BF & G [1997]

Runyon et al. [1994]

1.2

Tropical Rainforest

AS & L [1991]

0.8A

0.6-

+
l
&
l

0.4

0.2

10

12

14

LAI
Figure 5. Biomeaverages
of NPP (total)andLAI for IBIS plottedagainstmeasurements
of NPP andLAI acquired
fromthefollowingsources:Fassnachtand Gower,[1997] (evergreenneedleleaf,
broadleafdeciduous,
andmixed
forestsin northemWisconsin- only aboveground
NPP is accounted
for, F & G [1997] in figurelegend);Runyonet
al. [1994] for the PacificNorthwestUnited States(Oregon)(only aboveground
NPP was measured);various
measurements
of total NPP (aboveground,
fine, and coarseroot) and LAI in tropicalrainforestlocations;and a
tropical rainforestage sequencein Amazonia [Saldarriaga and Luxmoore, 1991], where data includeboth
aboveground
andcourserootNPP but excludefine rootNPP (S & L [ 1991] in figurelegend).

east central Canada, and the northwestern United States and

Australia,and the Great Plainsregionof the United States. These

BritishColumbia(Plate5a).

regions,similarto thosehigh fractionalcoverareasfor forests,

havehigherLAIs (> 5; Plate4b). The northemborealandtundra
southernEurope generallyhave lower vegetativecover comprised regionsof CanadaandAsia aremuchmoreopencanopies
dueto
of evergreentrees (10-30%). Deciduoustrees form a majority of colderanddrier climates,andonly have40-70% vegetativecover,
the vegetativecover in the forestedareasof the United Stateseast whichcorresponds
to LAI typically< 3.0 (Plate4b).
of the MississippiRiver, and in central and southernEurope,
whichare all areasthatexperience
temperateclimates(Plate5b). 4.6. Soil Organic Matter and Litter
Forested

areas of the eastern United

States and central

and

Some minimal cover of deciduoustrees (10-30%) (e.g., aspen,

The globaltotal amountof soil carbon(not countinglitter) to 1

m depth is simulatedto be 1408 Gt C (Table 4). This calibrated
(30-40%) of deciduous
treesis alsoapparentin the peripheryof value falls within the range (1200-1600 Gt C) reported from
compilationsof soil carbondatareportedin the literature[Postet
tropicalforestsin bothSouthAmericaandcentralAfrica.
Plate 5c showsthe simulatedfractional vegetationcover for al., 1982; Zinke et al., 1986; Schlesinger, 1991; Eswaran et al.,
both grassesand shrubs. Greaterthan 80% fractionalcover 1993;Sombroeket al., 1993;Batjes, 1996]. However,it hasbeen
generallyexistsin the grassland
andsavannaregionsof southern suggestedthat these numbersmight be an underestimatedue to
South America, southernand north central Africa, India, interior additionalorganiccarbonstoredin the upperpermafrostlayersin

birch,balsampoplar)is alsopresentin theborealforestregionsof

central Manitoba, Saskatchewan,and Alberta. Some dominance

814

KUCHARIK

ET AL,' TESTING A DYNAMIC

GLOBAL ECOSYSTEM

MODEL

m?-/m2
0-0.5
..... 0.5- 1

/' 1-2

\.

2-3
3-4

5-6
8-!0

m2/m2
0- 0.5
0.5-1

1-2

2-3

3-4
4-5
5-6
6-8
8-10

,-

n r

Savanna

. Grassland/Steppe
Dense Shrubland

.......OpenShrubland
.... i, Tundra
Desert

i' { Polar
Desert/Rock/Ice
TropicalEvergreen
TropicalDeciduous
TemperateBroadleafEvergreen
TemperateNeedleleafEvergreen
TemperateDeciduous
BorealEvergreen
Boreal Deciduous

Evergreen/Deciduous
MixedForest

Plate 4. Distributionof simulated(a) leaf areaindex(LAI) of trees,(b) simulatedLAI of grassesand shrubs,and

(c) globalvegetation
types.We classifythe simulated
combination
of 12 plantfunctionaltypesinto 15 vegetation
biomes.The vegetation
classification
is determined
by thedominantplantfunctionaltype(in termsof LAI) anda
few basic climatic

rules.

KUCHARIK

ET AL.' TESTING

A DYNAMIC

GLOBAL

ECOSYSTEM

MODEL

815

g 16

KUCHARIK

ET AL.: TESTING

A DYNAMIC

GLOBAL ECOSYSTEM

MODEL

[] IBIS

[]Observed

o25

'020

TT
o10-

5-

Figure6. Biomeaverages
and+/- 1 s.d.of simulated
totalsoilcarbon
density
(kgC m-2)to a depthof 1 m
comparedwith datacompiledby IGBP-DIS [ 1999]. The averageand standarddeviationsreflectvariationsoccurring
within eachspecificbiomeover the 30-year period(years 1965-1994) in modeloutputand for measurement
data,
respectively. The averagesand standarddeviationsshownfor model outputare not intendedto showyear-to-year
differences.

sometundraand boreal ecosystems

that has not been accounted over an entire,homogeneous,113grid cell. It is also possiblethat
for [Michaelson
et al., 1996].
the IBIS vegetationtype simulatedfor a particulargrid cell is not
The spatialvariationin soil organiccarbonis shownin Plate characteristic of the actual vegetation. This obviously can

6a. Forindividual113
x 113
gridcells,totalsoilcarbonranges
from

contribute to the observed differences between model and data.

0.1 kg C m'2to 53.1kg C m'2. Tundra

andboreal
deciduous

The global patternof total standingabovegroundlitter carbon

is shownin Plate 6c. Generally, as is the case with soil carbon
24.4kg C m'2,respectively,
whiledesert
ecosystems
havean storage,high latitudesgenerally have a larger accumulationof
average
ofonly0.4kgC m'2(Table
4). Theaverage
soilcarbon litter carboneven thoughannualvaluesof NPP are low compared
density
in thetropical
evergreen
forests
is approximately
9.5kg C to tropicalareas. Becausethe decompositionof carbonis highly
m'2(Table
4).
dependenton soil temperatureand soil moisture,high latitudes
have a slower rate of carbondecay,causinga larger buildup of
Recently, a global compilationof more than a thousandsoil
profiles with organic carbon content measurements(and other carboneven thoughNPP is low in theseregions[Landsbergand
relevantdata)wasusedto constructa mapof soil carbondensity Gower, 1997].
to a 1 m depth(Plate6b) [Scholeset al., 1995,'IGBP-DIS, 1999].
The total amount of carbon stored in the abovegroundlitter
This global pedon database is based on the WISE database pool simulatedby IBIS is 65.0 Gt C (includingwoody and
developedby the International Soil Referenceand Information nonwoodytissues),while the amountof dead fine root carbon
storedis 22 Gt C. Matthews[1997] analyzedboth measurements
Centre (ISRIC) in Wageningen,Netherlands [Scholeset al.,
to estimategloballitter pools
1995]. Here we comparesimulatedsoil carbondensitiesto these andmodelingof detritusproduction
the aboveground
fine litter(nonwoody)poolto be
data (Figure 6). Biome averagesand standarddeviationsfor the andestimated
68 Gt C, with a possible75 Gt C containedin
IGBP-DIS datawereconstructed
by assigning
eachgrid cell in the approximately

biomeshavethehighestbiome-average
carbondensities,
30.1 and

Mark Harmon (personal

dataset(using113resolution)to thecorresponding
IBIS vegetation coarse woody debris (CWD).
classification
for that grid cell. Generally,goodagreementexists communication,1998) estimatesglobal CWD carbonstorageto
for mostvegetationclasseswith the exceptionof desertandpolar be 55-60 Gt, with an additional 20-25 Gt in fine surface litter.
desertregions(Figure 6). As is the casewith NPP measurements, IBIS outputcomparesfavorablywith the total carbonstorage
soilprofilesareplot-scalemeasurements
characterized
by extreme estimatesof Harmon (75-85 Gt) but is roughlyhalf of the value
variability,while IBIS simulatesthe averagesoil carbondensity reportedby Matthews[1997].

KUCHARIK ET AL.: TESTING A DYNAMIC

For individual grid cells, the simulated abovegroundlitter

GLOBAL ECOSYSTEM MODEL

g 1'7

the decayrate of carbon,are significantand are simulatedby the

carbon
ranges
from0.01to2.9kgC m'2(Plate6c). Ona biome- model. The amountof microbialbiomasscan fluctuateby a factor
by-biomebasis,desertbiomeshave the lowest annualaverageat

of 2 duringthe year in many global ecosystems[Kuchariketa!.,

0.01kg C m'2,andborealdeciduous
forests
having
thehighest submittedmanuscript,1999].
densityof aboveground
litter,near1.3 kg C m-2 (Table4).
Belowground litter carbon storage of decomposingfine roots

ranged
froma highof 1.3kgC m'2inboreal
regions
to a lowof
0.01kgC m-2indesert
regions.
Tundra
regions,
onaverage,
have

4.8. Soil Respiration and Net EcosystemExchange

The efflux of carbon from the soil in IBIS is the sum of two

fine root respiration(autotrophicrespirationrequired

the highestequilibriumstorageof deadfine root carbon,roughly processes:
for
maintenance)
and carbon that is respiredby the microbial
0.7kgC m'2.
biomass(heterotrophic
respiration)
duringits oxidationof the soil
Plate 6d shows the mean residence time of carbon in litter and
soil organicmatter,estimatedas the ratio of the equilibriumtotal organicmatter.The exchangeof CO2 betweenthe soil and the
of litter and soil carbonto NPP. The spatialpatternof residence atmosphereis the secondlargestterm of some forest carbon
[Raich and
time shows that climates with colder temperatures and less budgets,next to gross canopy photosynthesis
Schlesinger,
1992].
Because
climate
affects
the
soil
temperature
precipitation (mainly higher elevation areas and northern
latitudes)have meancarbonresidencetimes in the soil and litter and moistureconditionsand the functioningof microbes,climate
between 50-100 years. A few specific biome averagesare 81 is a controlling factor in the productionof soil CO2 flux
yearsin borealdeciduousand 57 yearsin openshrublands
(Table [Landsbergand Gower, 1997]. Furthermore,strongpositive
4).
Temperate ecosystems found in central Europe, the
Mediterranean region, and parts of central North America and
Australiahave medianresidencetimes around15-25 years(Table
4). Tropical areas, which characteristicallyhave the highest
biomass,LAI, and NPP might depositthe mostdetritusto the soil
through litterfall, but the warm, moist climate of these regions
causes rapid decompositionof carbon in the model. These
regionsgenerally have carbon residencetimes in soil and litter
less than 20 years on average,with specificbiome averagesof
11.4 and 16.5 yearsfor tropicalevergreenand deciduousforests,
respectively(Table 4).

correlations have been shown to exist between NPP and soil

carbonefflux [Raichand Schlesinger,1992].

Soil CO2 fluxes generally increasefrom colder to warmer

latitudesandarehighestin biomesthathavethe highestsimulated

NPP values(Plate7b). Quantitiesof the annualsoil carbonflux

rangefromnearzeroin desertandotherunproductive
regionsof

theglobeto a highof 1.9kgC m'2 in tropical

rainforests.
The
distributionof soil carbon flux showsa strongcorrelationto the

spatialdistributionof simulatedNPP (Plate 2a). The highest

biomeaverageoccurredin tropicalevergreen
forests,whichhave

anaverage
annual
fluxof about
1.15kgC m'2(Table4). The
globalannualaveragesoil CO2respiration
is 70.2 Gt, whichis

about30% higherthan globalNPP (54.3 Gt). Thesevaluesare

consistent
with otherestimatesof the globaltotal soil carbonflux
A key componentof the IBIS biogeochemistry
model is the [RaicbandSchlesinger,1992].
For the purposeof evaluatingmodelsimulations
of soil CO2
living microbialbiomasspool. The decomposition
of all litterfall
and fine root turnover and more recalcitrant soil carbon takes
flux, we used the data set of soil carbon flux measurements
place throughthe microbialassimilationof thesematerials. In compiled by Raich and Schlesinger[1992]. Measurement
general,microbeshave a muchstrongeraffinity for labile carbon locations were correlated with the specific IBIS simulated
classsothatbiomeaverages
andstandard
deviations
of
(recentlyaddedlitterfall)thanrecalcitrantsoil carbon[Landsberg vegetation
and Gower, 1997]. As a result,microbialbiomassgrowth and the measurementdata could be compared with IBIS results.
overall density are somewhat correlated to patterns of NPP Unfortunately,not all of the IBIS biomes had representative
to compareto (Table 6). Some obviouspatterns
(litterfall) [Zak et al., 1994] and total soil carbonstorage. IBIS is measurements
constructedin sucha mannerthat the growthof microbialbiomass emerge;namely,IBIS slightlyunderestimates
measured
soil CO2
is directlyrelatedto theamountof availablesubstrate
anddirectly flux in tropical evergreenand temperatebroadleafevergreen
measuredfluxesin othertemperateand
influencedby the soil environmentaffecting the activity of the biomesand overestimates
microbial biomass.
boreal ecosystems(Table 6). Furthermore, the standard
The globaldistributionof simulatedmicrobialbiomassranges deviationsof measurementsmade in somebiomes are larger than
from0.01kgC m'2in unproductive
deseregions
of Africaarid the variationsimulatedby IBIS.
thewestern
UnitedStates
to 0.49kg C ni'2in tropical
evergreen Soil-surfaceCO2 flux and the partitioning of the flux into
forests(Plate7a). Generally,thesequantitiesare similarto values separate
contributions
of rootsandmicrobesis quitedifficultto
reportedby Landsbergand Gower [1997] for differentbiomes. measure because of substantial errors associated with various
a highdegreeof spatialvariability,and
Moreover, measurementsof microbial biomass made in natural methodsof measurement,
on soil environmentand nutrientcharacteristics
prairiesystemsin Wisconsinduringlate-spring
andearlysummer dependence
problemis thoughtto
[ Wagaiet al., 1998;Kuchariket al., submittedmanuscript,1999], [Normanet al., 1992,1997]. Oneparticular
whichaveraged
around0.25-C..50
kg C m-2, alsocompareexist with the data set of Raich and Schlesinger[1992]; namely
reasonablywell with IBIS mocleloutputfor the MidwestUnited most of the measurementswere probably made using a static
States. On a biome-by-biome basis, temperate evergreen alkali absorption
technique[RaicbandNadelhoffer,1989],which
soil CO2 flux, especiallyat higherrates[Ewel et
broadleafforestshavethe highestannualaveragedensity,around underestimates
0.24kgC m-2,withtropical
forests
between
0.12-0.20
kgC m-2 al., 1987;Haynesand Gower,1995;Normanet al., 1997].Thus
thatIBIS doesan adequate
job of simulating
(Table 4). We note that the annual variations in microbial we canonly suggest
biomass,which are correlatedto the timing of litt/rfall and soil the exchangeof CO2betweenthe soil-surface
andthe atmosphere
environmentalconditions(temperatureand moisture)that affect basedon the limited datawe havefor comparison.
4.7. Microbial

Biomass

818

KUCHARIK

ET AL.' TESTING

A DYNAMIC

o.

GLOBAL

ECOSYSTEM

MODEL

KUCHARIK

ET AL.: TESTING

A DYNAMIC

GLOBAL

ECOSYSTEM

MODEL

kg/m2
O- 0.05

0.15-

0.2

0.1 - 0.15 l-- 0.25

0.3 - 0.35
0.35 - 0.4
0.4- 0.43

7a

kg/m2/yr
0.3 - 0.4

0.6 - 0.7 -

o.1 - 0.2

o - o.1

0.4 - 0.5

0.7 - 0.8

0.2 - 0.3

0.5 - 0.6 t..--] 0.8 - 0.9 -'-

0.9 - 1
1 - 1.2

1.2 - 1.4

Plate7. Distribution
of annual
average
of (a) simulated
microbial
biomass
(kgC m-2)and(b) totalsoilcarbon
respiration
(kgC m'2yr-;finerootandmicrobial).

819

820

KUCHARIK ET AL.: TESTING A DYNAMIC GLOBAL ECOSYSTEM MODEL

Table 6. Comparison
of IBIS SimulatedAnnualAverageandStandardDeviationof SoilCO2Flux With
Point-Measurement
Data Compiledby RaichandSchlesinger[1992] (R&S).
IBIS Biome

IBIS,

R & S,

kgC m-2yr-
Tropicalevergreen
forest
Tropicaldeciduous
forest
Temperate
broadleaf
evergreen
forest
Temperate
coniferevergreen
forest
Temperate
deciduous
forest
Borealevergreen
forest

1.15
0.69
0.96
0.71
0.66
0.37

Obs

kgC m'2yr'
(0.202)
(0.144)
(0.256)
(0.126)
(0.139)
(0.103)

1.28
0.47
1.36
0.37
0.53
0.16

(0.20)
(0.09)
(0.64)
(0.19)
(0.27)
(0.08)

10
1
4
9
61
18

11

Borealdeciduousforest

0.29

(0.074)

......

Evergreen/deciduous
mixedforest

0.50

(0.181)

0.72

(0.36)

Savanna

0.61

(0.244)

0.80

(0.20)

Grassland/steppe

0.53

(0.252)

0.48

(0.15)

Denseshrubland

0.27

(0.111)

0.16

(0.03)

Openshrubland

0.15

(0.096)

0.22

(0.05)

Tundra
Desert
Polardesert/rock/ice

0.24
0.02
0.02

(0.098)
(0.029)
(0.017)

0.15
0.711
......

(0.15)
(0.11)

13
3

Standard deviation +/- 1 is shown. The column denoted "Obs" is the number of field measurements that were

categorized
intoeachIBIS biome. The averageandstandard
deviationreportedfor Raichand Schlesinger
[1992]
are classifiedinto IBIS biomecategoriesbasedon the actualmeasurement
location(latitude,longitude). Missing
data (dashes)denoteno measurements
made in locationsfalling within thoseIBIS biome classifications.
The
averageandstandarddeviationsfor modeloutputreflectvariationsoccurringwithineachspecificbiomeoverthe
30-yearperiod(years1965-1994)andarenotintended
to showyear-to-year
differences.

However,thereare somelimitationsto IBIS that canalsohelp during the winter season in temperate climates, the root

explaindeparture
fromfieldmeasurements.
First,soilprofilesof contribution
to total carbonrespiredincreases
significantlyas
02 and CO2, which are influencedby gaseousdiffusionand the microbialactivitydeclines[Winstonet al. 1997].
verticalmovementof soil air, are not accountedfor in IBIS [Fang
The annualnetecosystem
exchange
(NEE) of carbonis defined
respiration;
this quantityis usedin
and Moncrieff, 1999]. The gaseous environment (e.g., as NPP minusheterotrophic
to help determineif
concentrations
of both O2 and CO2) surroundingmicrobesmight modelingstudiesand field experiments
be viewedasbeingas importantas soil moistureandtemperature terrestrial ecosystemsare contributingto or mitigating the
in influencingmicrobialactivity. Furthermore,other factorsnot observedrise in atmosphericCO2. Many modeling studies
parameterized
in IBIS suchas soil pH, the type of microbial diagnosethe future stateof the globalterrestrialcarboncycle
fauna, and presenceof heavy metals can also effect CO2 under future atmosphericCO2 concentrationsand climate
production [Fang and Moncrieff, 1999]. Moreover, the scenariosat regional and global scales.Here we presentthe

oversimplificationof fine root dynamicsin IBIS could also annualaverageNEE over the 30-yearperiodfrom 1965-1994,
which reflects a period of increasing atmospheric CO2
concentration,
and transientclimatechanges.Model simulations
the IBIS fine root respirationis controlledby soil temperature,
savannas,
openshrublands,
anddesertsall
anyerrorsin describing
theannualcycleandtimingof fine root showthat grasslands,
arenet sources
of
turnovercouldpotentiallyyield significanterrorsin soil CO2 havenegativeNEE, meaningtheseecosystems
carbonto the atmosphere,
with a rangein fluxesbetween13-40 g
production.
contributeto errors in annual estimatesof soil CO2 flux. Because

IBISbiomes
aresequestering
carbon
Landsbergand Gower[1997]reportthatthe averagefraction C m-2yr-. Theremaining
of rootrespiration
to totalCO2flux is about0.45 but is highly during this time (Table 4), with the highestbiome average
forests,wherecarbonwasbeing
variable(evenwithinthe sameforesttype),rangingfrom 0.22- occurringin tropicaldeciduous

ata rateof 53g C m-2yr-. Wenotethattheannual

0.62. Thesevaluesarehighlydependent
on whattimeof yearthe sequestered
average
global
total NEE during 1965-1994 (2.3 Gt C) is
measurementswere taken. Figure 7 showsthe annual average
equalto reportsfromthe IPCC whichestimatea
contribution of fine roots to total soil CO2 flux in IBIS. These approximately
2 Gt yr- terrestrial
carbon
sink(Schimel,1995;
valuesrangefrom 0.16 in borealdeciduous
foreststo 0.51 in present-day
grassland/steppe
biomes,with the averagenear 0.30. We note Steffenet al., 1998).
that the ratio of root to total carbonrespiredis highly variable
duringthe year when comparedto the annualaverage. In
5. Summary and Conclusions
grassland
ecosystems
duringpeakgrowthperiods,
therootto total
In this study, we presenta new versionof the IBIS global
CO2 flux ratio in IBIS can approach0.70, for example. Field
measurements
of the partitioningare generallyonly performed biosphere model. IBIS-2 has improved canopy physiology
duringa few periodsor daysduringthe year andmighthavebeen formulations,additional plant functional types, more realistic
of
measuredduring the peak of the growing season. Furthermore, carbonallocationcoefficients,and improvedparameterizations

KUCHARIK

07'

0.6

ET AL.' TESTING A DYNAMIC

GLOBAL

ECOSYSTEM

MODEL

821

0.5
0

''

r,

0.4

-03

mE 0.2
m

c::

0.1

Figure 7. Biomeaveragesand+/- 1 s.d.of the ratio of simulatedannualaveragefine root respirationto total soil
carbonrespiration.
The averageandstandarddeviationsreflectvariationsoccurringwithin eachspecificbiomeover
the30-yearperiod(years1965-1994)andarenotintended
to showyear-to-year
differences.

vegetationdynamicsandplantphenology.Moreover,the addition
of a comprehensive
belowground
biogeochemistry
submodelnow
allows for the descriptionof belowgroundbiogeochemical
processes
and simulatesthe flow of carbonbetweenvegetation,
detritus,andthesoilorganicmatter.
While global-scale ecosystem models are becoming
increasinglysophisticated,
they still suffer from a severelack of
rigorous testing and evaluation. Thankfully, this situation is
startingto change. For example,many modelsare now being
compared to detailed biophysical measurementsobtained at
individualfield sites. However, for global- and regional-scale
ecosystemmodels,it is necessaryto make these comparisons
acrossa wide rangeof climateandecosystem
types[e.g.,Delire
and Foley, 1999]. Unfortunately, many ecosystem types
(includingbothnaturaland managedsystems)have not yet been
observedwith sophisticated
eddy-fluxmeasurement
systems.
In two parallel studies designed to test the biophysical,
hydrological,andbiogeochemical
processes
simulatedin IBIS-2,
the modelhasbeentestedagainstseveraluniquesetsof detailed
measurements[Delire and Foley, 1999; Kucharik et al., submitted

manuscript,1999]. In the Delire and Foley [1999] study,field

measurementsof soil moisture and temperature and surface
energyandcarbonfluxeswere comparedwith model simulations
for five ecosystems:
a soybeancrop(HAPEX-Mobilhy), meadow
(Cabauw,Netherlands),
grassland
(Valdai, Russia),prairie(FIFE,
Konza prairie, Kansas) and tropical rainforest (ABRACOS,
Brazil). In the Kucharik et al. [submittedmanuscript,1999]

study, field measurementsof microbial biomass, soil carbon

density,soil CO2 flux, and nitrogenmineralizationand leachingin
agroecosystemsunder different management practices in the
upper Midwest were comparedto output generatedby IBIS-2.
From thesetwo independentstudies,we concludethat IBIS-2 is
able to accurately simulate biophysical, hydrologic, and
biogeochemical
processes
occurringat contrastingsites.
Even a much more extensive set of detailed site-specific
biophysicalmeasurementsfrom acrossthe globe would not be
completelysufficientto evaluateglobal ecosystemmodels. It is
also necessary to consider more fundamental ecosystem
properties,which includeabovegroundand belowgroundbiomass,
leaf area index, net primary production,litter carbon,and soil
carbon. These quantities generally define the productivity,
longevity, and physical statureof global ecosystems. A much
more completedescriptionof terrestrialecosystemprocessescan
be obtainedby combiningdetailedbiophysicalmeasurementsof
canopy gas-exchange processes with spatially extensive
measurements
of basic ecosystempropertiesand biogeochemical
stocks. Fortunately, more comprehensivecompilationsof these
data are being developedby a number of researchgroups and
internationalorganizations.

In this study,we use severalglobal-scalecompilationsof basic

ecological and biogeochemical measurements that are now
becomingavailable. Even thoughmostof thesedatacompilations
are not fully comprehensivein nature because they do not
describeeachsquarekilometerof the Earths surface,they are still

822

KUCHARIK ET AL.: TESTING A DYNAMIC GLOBAL ECOSYSTEM MODEL

extremelyvaluableto our modelevaluationprocess.Usingthese

data,our modelevaluationshowsthat simulatedpatternsof NPP,
biomass,andleafareaindexarecomparable
to otherglobalmodel
resultsand agree reasonablywell with measurements
that have
been compiled from numerousecosystems. We also find our
simulations of total soil carbon, accumulated litter at the soil

surface,microbialbiomass,and total soil CO2 flux compare

favorablywith variousdataavailableboth in the literatureand in
globaldatacompilations.

Generallyspeaking,
it appears
thatIBIS-2 is ableto represent
a
widerangeof ecosystem
processes
occurring
at bothlocal[Delire
and Foley, 1999,'Kuchariket al., submittedmanuscript,
1999]
andglobalscales.Whileour evaluationof IBIS-2 represents
the
mostexhaustivetest of globalecosystem
modelsperformedto

Institutefor Biogeochemistry)
hasgivenvaluableinputandadvice
throughout
thedevelopment
of IBIS. Thisworkwassupported
by
NASA throughan Interdisciplinary
ScienceInvestigation
(IDS)
grant (NAG5-3513) and by the NSF through grant ATM9711484. AdditionalsupportwasprovidedthroughtheClimate,
People,and EnvironmentProgram(CPEP) of the Universityof
Wisconsin.

References
Amthor, J.S., The role of maintenance
respirationin plant growth,Plant
Cell Environ., 7, 561-569, 1984.

Amthor,J.S., ScalingCO2photosynthesis
relationships
from the leaf to the
canopy,Photosynth.
Res'.,39, 321-350, 1994.
date, it is importantto note the limitations of this work. For Arethor, J.S., M.L. Goulden,J.W. Munger, and S.C. Wofsy, Testing a
mechanisticmodel of forest-canopy
massand energyexchangeusing
example,our ability to test global ecosystemmodelsagainst
eddy correlation:Carbondioxide and ozone uptakeby a mixed oakmeasurements
is still hamperedby the scarcityof measurements
maplestand,Aust.J. Plant. Physiol.,21,623-651, 1994.
from key biomesand the extremespatialvariabilityof the Ball, J.T., I.E. Woodrow,and J.A. Berry, A model predictingstomatal
conductanceand its contributionto the controlof photosynthesis
under
measurements
(which are difficult to compareto homogeneous
different environmentalconditions,in Progress in Photosynthesis'
modelgridcells). Furthermore,
we arestill limitedby thehighly
Research,vol. 4, editedby J. Biggins,pp. 221-224, MartinusNijhoff,

biasednatureof temporalsampling
thatis typicalof biophysical
Zoetermeer, Netherlands, 1986.
and ecological measurements. The timing of ecological Batjes, N.H., Total carbonand nitrogenin the soils of the world, Eur. J.
Soi! Sci., 47, 151-163, 1996.
measurements
is critical for modelcomparisons
whendealing
CO2exchangesimulatedby a land surface
with highlydynamicquantities
suchas standing
litter,livingfine Bonan,G.B., Land-atmosphere
root biomass, microbial biomass, and the contribution of fine
rootsto the total soil CO2 efflux.

We believethatfuturedevelopments
in global-scale
terrestrial
ecosystem
modelingmustfocuson two majorobjectives.First,
globalecosystem
modelsmustbe muchmorerigorouslytested
againsta suiteof site-specific(e.g., flux tower measurements
of
carbon,water,andenergyexchange)and spatiallyextensivedata
(e.g., compilations of field-based and satellite-based
measurements
that characterizethe basic structureof vegetation
andothercarbonstocks).Second,globalecosystem
modelsneed
to seriouslyaddressimportantaspectsof the Earths biosphere
suchastheeffectsof naturaldisturbance
(e.g.,fire, herbivory,and
windthrow)andhumanlanduseactivities[e.g.,Ramankutty
and

processmodelcoupledto an atmospheric
generalcirculationmodel,J.
Geophys.Res.,100(D2), 2817-2831, 1995.
Botta, A., N. Viovy, P. Ciais, and P. Friedlingstein,and P. Monfray, A
global prognosticschemeof leaf onset using satellite data, Global
ChangeBio/. in press,2000.
Brye, K.R., J.M. Norman, L.G. Bundy,and S.T. Gower, An equilibrium
tensionlysimeterfor measuringdrainagethroughsoil, Soil Sci. Soc.
Am. J., 63, 536-542, 1999.

Campbell, G.S., and J.M. Norman, An Introduction to Environmental

Biophysics,
286 pp., Springer-Verlag,
New York, 1998.
Cannell, M.G.R., World Forest Biomassand Primary Production Data,
Academic,SanDiego, Calif., 1982.

Cannell,M.G.R.,andJ.E.Jackson,
Attributes
of Treesas CropP/ants,592
pp., Inst.OfTerr. Ecol,AbbotsRipton,Huntington,U.K., 1985.

Foley, 1998].

Chen, J.M., P.M. Rich, S.T. Gower, J.M. Norman, and S. Plummer, Leaf
area indexof borealforests:Theory,techniques,and measurements,
J.
Geophys.Res., 102(D24), 29429-29443,1997.

Note

Cogley, J.G., GGHYDRO --Global hydrologicaldata release2.0, Trent

Clim. Note 91-1, Dep. of Geogr., Trent Univ., Peterborough,Ont.,
Canada, 1991.

The IBIS-2 code is written in Fortran 77, and has been

implementedon a variety of Unix-based computingplatforms.

The codeandmanyof the associated
datasetsare freely available.
For more information,pleasecontactthe authorsor refer to our
web site (http://cpep.aos.wisc.edu).

Acknowledgements.We wishto acknowledge

individualswho
havehelpedmakethiswork possible.Climatedatawereobtained
from Wolfgang Cramer (PotsdamInstitute for Climate Impacts
Research)and Patrick Bartlein (University of Oregon), and a
preliminaryversionof the IGBP-DIS soilsdata setwas provided
by Robert Scholes(Division of Forest Scienceand Technology,
Pretoria, South Africa). Gower compiled the NPP data while
supportedby a NASA grant(NAG5-7298). Guiling Wang (MIT),
David Price (Canadian Forest Service), and Debbie Shannon
(University of Cape Town) testedpreliminaryversionsof IBIS-2
and were instrumentalin the model development. The land
surfacephysicsroutinesof IBIS were basedon the LSX land
surfacemodel code, which was generouslyprovidedby David
Pollard (Penn StateUniversity)and StarleyThompson(National
Center for AtmosphericResearch). Colin Prentice(Max Planck

Colbert, S.T., E.J. Jokela,and D.G. Neary, Effectsof annualfertilization

and sustainedweed controlon dry matterpartitioning,leaf area, and
growth efficiency of juvenile loblolly and slashpine, For. Sci., 36,
995-1014, 1990.

Collatz, G.J., J.T. Ball, C. Grivet, and J.A. Berry, Physiologicaland

environmentalregulationof stomatalconductance,photosynthesis
and
transpiration:A modelthat includesa laminarboundarylayer,Agric.
Pbr. Meteoro!., 53, 107-136, 1991.
Collatz, G.J., M. Ribas-Carbo,and J.A. Berry, Coupled photsynthesisstomatalconductancemodel for leaves of C4 plants,Aust. J. Plant
Physiol.,19, 519-538, 1992.
DeFries,R.S., M.C. Hansen,J.R.G. Townshend,and R. Sohlberg,Global
land cover classificationsat 8 km spatial resolution: The use of
trainingdataderivedfrom Landsatimageryin decisiontree classifiers,
Int. J. RemoteSens.,19, 3141-3168, 1998.
DeFries, R.S., J.R.G. Townshend,and M.C. Hansen, Continuousfields of
vegetationcharacteristics
at the global scale at l-km resolution,J.
Geophys.Res'.,104(D14), 16,911- 16,923,1999.
Delire, C., and J.A. Foley, Evaluating the performance of a land

surface/ecosystemmodel with biophysical measurementsfrom

contrastingenvironments,
J. Geophys.Res., 104(D14), 16,895-16,909,
1999.

Dickinson,R.E., A. Henderson-Sellers,
P.J. Kennedy,and M.F. Wilson,
Biosphere-Atmosphere
TransferScheme(BATS) for the NCAR CCM,
NCAR/TN-275-STR,Nat. Cent. for Atmos. Res., Boulder,Colo., 1986.

KUCHARIK

ET AL.: TESTING

A DYNAMIC

Esser,G., H.F.H. Lieth, J.M.O. Scurlock,and R.J. Olson, Worldwide

estimatesand bibliographyof net primaryproductivityderivedfrom

re-1982publications,
ORNL/TM-1348.5,
104pp.,OakRidgeNat.
Lab., Oak Ridge,Tenn., 1997.

Eswaran,H., E. Van den Berg,and P. Reich,Organiccarbonin soilsof

the world, Soil Sci. Soc. Am. d. , .57, 192-194, 1993.

Evans,J.R., Partitioningof nitrogenbetweenand within leavesgrown

underdifferentirradiances,
Aust.d. Plant.Physio!.,16, 533-548, 1989.
Ewel, K.C., W.P. Cropper,andH.L. Gholz,Soil CO2evolutionin Florida
slashpineplantations,II, Importanceof rootrespiration,
Can.d. For.
Res., 17, 330-333, 1987.

Fang, C., and J.B. Moncrieff, A model for soil CO2 productionand
transport,1, Model development,Agr. For. Meteorol.,9.5, 225-236,
1999.

GLOBAL

ECOSYSTEM

MODEL

g23

Kira, T., Communityarchitecture

and organicmatterdynamicsin tropical
lowland rainforestsof SoutheastAsia with specialreferenceto Pasoh
ForestMalaysia,in Tropical Treesas Living Systems,editedby P.B.
Tomlinsonand M.H. Zimmermann, CambridgeUniv. Press,New
York, 1978.

Kira, T., H. Ogawa, K. Yoda, and K. Ogino, Primaryproductionby a

tropicalrain forestin southernThailand,Bot. Mag., 77, 428-429, 1964.
Kucharik, C.J., J.M. Norman, L.M.

Murdock, and S.T. Gower,

Characterizingcanopynonrandomness
with a multibandvegetation
imager(MVI), J. Geopbys.Res'.,102(D24), 29455-29473,1997.
Kucharik, C.J., J.M. Norman and S.T. Gower, Characterization of

radiation regimes in nonrandom forest canopies: Theory,

measurements,
anda simplifiedmodelingapproach,Tree Pbysio!.,19,
695-706, 1999.

Farquhar, G.D., and T.D. Sharkey, Stomatal conductance and

Kull, O., and P.G. Jarvis,The role of nitrogenin a simpleschemeto scale
up photosynthesis
from leaf to canopy,Plant Cell Environ., 18, 1174photosynthesis,
Ann.Rev.PlantPhysiol.,33, 317-345,1982.
1182, 1995.
Farquhar,G.D., S. von Caemmerer,and J.A. Berry, A biogeochemical
modelof photosynthetic
CO2 assimilationin leavesof C3 species, Landsberg,J.J.,and S.T. Gower, Applicationsof PhysiologicalEcologyto
Planta, 149, 78-90, 1980.
ForestManagement,354 pp., Academic,SanDiego,Calif., 1997.
Fassnacht,K.S., and S.T. Gower, Interrelationships
amongthe edaphic Larcher,W., PhysiologicalPlant Ecology,303 pp., Springer-Verlag,New
York, 1983.
andstandcharacteristics,
leaf areaindex,andaboveground
netprimary
productionof uplandforestecosystems
in north centralWisconsin, Leemans,R., and W.P. Cramer,The IIASA databasefor mean monthly
Can. d. For. Res'.,27, 1058-1067, 1997.
values of temperature, precipitation and cloudinesson a global
Field, C.B., Allocatingleaf nitrogenfor the maximizationof carbongain:
terrestrialgrid, IIASA WP-90-41, Int. Inst. For Appl. Syst. Anal.,
Laxenburg,Austria,1990.
Leaf age as a controlon the allocationprogram,Oecologia,56, 341347, 1983.
Lieberman,D., M. Lieberman,R. Peralta,and G.S. Hartshorn,Mortality
patternsand standturnoverratesin a wet tropicalforestin CostaRica,
Field, C.B., Ecologicalscaling of carbon gain to stressand resource
J. Eco!., 73, 915-924, 1985.
availability,in Response
of Plantsto MultipleStresses,
editedby H.A.
Linder,S.L., M.L. Benson,B.J. Myers,andR.J. Raison,Canopydynamics
Mooney,W.E. WinnerandE.J. Pell, pp. 35-65, Academic,SanDiego,
Calif., 1991.
and growthof Pinus radiata, l, Effectsof irrigationand fertilization
duringdrought,Can. J. For. Res., 17, 1157-1165, 1987.
Fitter,A.H., and R.K.M. Hay, EnvironmentalPhysiologyof Plants, 355
Linn, D.M., and J.W. Doran, Effect of water-filledpore spaceon carbon
pp.,Academic,SanDiego,Calif., 1981.
dioxideand nitrousoxide productionin tilled and non tilled soils,Soil
Foley, J.A., Numericalmodelsof the terrestrialbiosphere,d. Biogeogr.,
22, 837-842, 1995.

Foley,J.A., I.C. Prentice,N. Ramunkutty,

S. Levis,D. Pollard,S. Sitch,
and A. Haxeltine, An integratedbiospheremodel of land surface
processes,
terrestrialcarbonbalanceandvegetation
dynamics,Global
Biogeochem.
Cycles,10(4), 603-628, 1996.
Friend,A.D., A.K. Stevens,R.G. Knox, and M.G.R. Cannell, A process-

based,terrestrialbiospheremodelof ecosystemdynamics(Hybrid
v3.0), Ecol. Mode#.,95(2-3), 249-287, 1997.
Geng,S., F.W.T. PenningDe Vries, and I. Supit,A simplemethodfor
generatingdaily rainfall data,Agric. For. Meteoro!.,36, 363-376,
1985.

Gower,S.T., J.G.Vogel,J.M. Norman,C.J. Kucharik,S.J.Steele,andT.K

Stow,Carbondistribution
andaboveground
net primaryproduction
in
aspen,jack pine, and black sprucestandsin Saskatchewanand
Manitoba,Canada,d. Geophys.
Res.,102(D24)29029-29041,1997.
Gower, S.T., C.J. Kucharik, and J.M. Norman, Direct and indirect
estimationof leaf area index, faparand net primaryproductionof
terrestrialecosystems,
RemoteSens.Environ.,70, 29-51, 1999.
Harcombe,P.A., Tree life tables,BioScience,37, 557-574, 1987.
Haxeltine, A., and l.C. Prentice, BIOME3: An equilibrium terrestrial
biospheremodel basedon ecophysiologicalconstraints,resource
availability,and competitionamongplant functionaltypes,Global
Biogeochem.
Cycles,10(4), 693-709, 1996.

Haynes,B.E., and S.T. Gower, Belowgroundcarbonallocationin

Sci. Soc. Am. J., 48, 1267-1272, 1984.

Lloyd, J., and J.A. Taylor, On the temperature dependenceof soil

respiration,FunctionalEco!., 8, 315-323, 1994.
Lloyd, J., J. Grace,A.C. Miranda, P. Meir, S.C. Wong, H.S. Miranda, l.R.
Wright, J.H.C. Gash, and J. Mcintyre, A simple calibratedmodel of
Amazon rainforestproductivitybasedon leaf biochemicalproperties,
Plant, Cell Environ., 18, 1129-1145, 1995a.
Lloyd, J., S.C. Wong, J.M. Styles,D. Batten,R. Priddle,C. Turnbull, and
C.A. McConchie,Measuringand modellingwhole-treegasexchange,
Aust.J. Plant Physio!.,22, 987-1000, 1995b.
Matthews, E., Global vegetationand land use: New high resolutiondata
basesfor climatestudies,J. Clim. Appl. Meteorol., 22, 474-487, 1983.
Matthews, E., Global litter production, pools, and turnover times:
Estimatesfrom measurement
dataand regressionmodels,J. Geophys.
Res., 102(D 15), 18771- 18800, 1997.

Melillo, J.M., A.D. McGuire, D.W. Kicklighter, B. Moore Ill, C.J.

Vorosmarty,and A.L. Schloss,Global climate changeand terrestrial
net primaryproduction,Nature, 363, 234-240, 1993.
Michaelson, G.J., C.L. Ping, and J.M. Kimble, Carbon storageand
distributionin tundra soils of Arctic Alaska, USA, Arct. Alp. Res.,
28(4), 414-424, 1996.
Murray, M.B., M.G.R. Cannell, and I. Smith, Date of budburstof fifteen
tree speciesin Britain folowingclimaticwarming,J. Appl., Eco!., 26,
693-700, 1989.
Neilson,R. P., and D. Marks, A globalperspectiveof regionalvegetation
andhydrologicsensitivities
from climaticchange,J. VegetationSci., .5,

unfertilizedand fertilizedred pine plantationsin northernWisconsin,

TreePhysio!.,15(5), 317-325, 1995.
Hendrick,R.L., andK.S. Pregitzer,Patterns
of fine rootmortalityin two
715-730, 1994.
sugarmapleforests,Nature,361, 59-61, 1993.
Norman, J.M., R. Garcia, and S.B. Verma, Soil surfaceCO2 fluxes and the
Hirose,T., andM.J.A. Werger,Maximizingdaily canopyphotosynthesis
carbonbudgetof a grassland,J. Geophys.Res., 97(D17), 18845with respectto the leaf nitrogenallocationpatternin the canopy,
18853, 1992.
Oecologia,72, 520-526, 1987.
Norman, J.M., C.J. Kucharik, S.T. Gower, D.D. Baldocchi, P.M. Crill, M.
InternationalGeosphere-Biosphere
Programme- Data and Information
Rayment,K. Savageand R.G. Striegl, A comparisonof six methods
System(IGBP-DIS),GlobalSoil DataTask:SpatialDatabase
of Soil
for measuringsoil surfacecarbondioxide fluxes,J. Geophys.Res.,
Properties,
Toulouse,France,1999.
102(D24), 28771-28777, 1997.
Jackson,
R.B., J. Canadell,andJ.R. Ehleringer,A globalanalysisof root Olson, J.S., J.A. Watts, and L.J. Allison, Carbon in Live Vegetationof
distributions
for terrestrialbiomes,Oecologia,108, 389-411, 1996.
Major WorldEcosystems,
Oak Ridge Nat. Lab., Oak Ridge, Tenn.,
Jordan,C.F., Productivityof tropicalforestand its relationto a world
1983.
patternof energystorage,
d. Eco!.,59, 127-142,1971.
Parton, W.J., D.S. Schimel, C.V. Cole, and D.S. Ojima, Analysis of
Kalnay, E., et al., The NCEP/NCAR 40-Year ReanalysisProject,Bull.
factors controlling soil organic matter levels in Great Plains
Am. Meteorl. Soc. 77, 437-471, 1996.

grasslands,Soil Sci. Soc.Am. J., .51, 1173-1179, 1987.

824

KUCHARIK

ET AL.: TESTING A DYNAMIC

Parton,W.J., et al., Observationsand modelingof biomassand soil

organicmatterdynamicsfor the grassland
biomeworldwide,Global
Biogeocbem.
Cycles',7(4), 785-809, 1993.
Pollard,D., and S. L. Thompson,Use of a land-surface-transfer
scheme
(LSX) in a globalclimatemodel: The response
to doublingstomatal
resistance,
GlobalPlanet.Change,1O, 129-161,1995.
Post,W.M., W.R. Emanuel,P.J. Zinke, and A.G. Stangenberger,
Soil
carbonpoolsandworldlife zones,Nature,258, 156-159,1982.
Potter, C.S., J.T. Randerson,C.B. Field, P.A. Matson, P.M. Vitousek,
H.A. Mooney,andS.A. Klooster,Terrestrialecosystem
production:
A
processmodel based on global satellite and surface data, Global
Biogeocbem.Cycles,7(4), 811-841, 1993.
Prentice,I.C., and H. Helmisaari,Silvics of north Europeantrees:
Compilation, comparisonsand implicationsfor forest succession
modelling,For. Ecol. Manage.,42, 79-93, 1991.
Prentice,I.C., W. Cramer,S.P. Harrison,R. Leeroans,R.A. Monserud,and

A.M. Solomon,A globalbiomemodelbasedon plantphysiologyand

dominance,soil propertiesand climate,J. Biogeogr.,19, 117-134,
1992.

Raich, J.W., and K.J. Nadelhoffer, Below groundcarbonallocationin

forestecosystems:
Globaltrends,Ecology,70, 1346-1354,1989.
Raich,J.W., andW.A. Schlesinger,
The globalcarbondioxideflux in soil
respiration
andits relationship
to vegetation
andclimate,Tellus,44B,
81-99, 1992.

Ramankutty,N., and J.A. Foley, Characterizing

patternsof globalland
use:An analysisof globalcroplands
data,GlobalBiogeochem.
Cycles,
12(4), 667-685, 1998.

Ramankutty,N. and J.A. Foley, Estimatinghistoricalchangesin global

landcover:Croplands
from 1700to 1992,GlobalBiogeocbem.
Cycles,
13, 997-1027, 1999.

RawIs,W.J., L.R. Ahuja, and D.L. Brakensiek,Estimatingsoil hydraulic

properties from soils data, in Indirect Methodsfor Estimating
Hydraulic Propertiesof UnsaturatedSoils',edited by M. th. Van
Genuchthen,
F.J. Leij, and L.J. Lund, U.C. RiversidePress,Riverside,
Calif., 1992.

Richardson,
C.W., andD.A. Wright,WGEN:A modelforgenerating
daily
weathervariables,ARS-8,83pp.,U.S. Dep.of Agric.,Agric.Res.Serb'.,
Washington,D.C., 1984.
Richardson, C.W., Stochastic simulation of daily precipitation,
temperature, and solar radiation, Water Resour. Res., 17, 182-190,
1981.

Roberts,J.M., O.M.R. Cabral, J.P. da Costa,A.-L.C. McWilliam, and T.D.

de A. S, An overview of leaf area index and physiological

measurements
during ABRACOS, in AmazonianDeforestationand
Climate, editedby J.H.C. Gashet al., pp. 287-306, JohnWiley, New
York, 1996.

GLOBAL ECOSYSTEM

MODEL

Schlesinger,
W.H.,Biogeochemistry,
AnAnalysis'
of GlobalChange,
443
pp., Academic,SanDiego, Calif., 1991.

Scholes,R.J., D. Skole,and J.S. Ingram,A globaldatabase

of soil
properties:
Proposal
for implementation,
IGBP-DISworking
pap. 10,
GCTE Focus3 Associate
Office,Oxford,UK, 1995.

Sellers,
P.J.,Y. Mintz,Y.C. Sud,andA. Dalcher,
A simplebiosphere
model(SiB) for usewithingeneralcirculation
models,J. Atmos.Sci.,
43, 505-531, 1986.

Sellers,
P.J.,J.A. Berry,G. J.Collatz,
C. B.Field,andF. G. Hall,Canopy
reflectance,
photosynthesis,
andtranspiration,
III, A reanalysis
using
improved
leafmodelsanda newcanopyintegration
scheme,Remote
Sens.Environ., 42, 187-216, 1992.

Sellers,P.J.,et al., Comparison

of radiativeandphysiological
effectsof
doubled
atmospheric
CO2onclimate,Science,
271, 1402-1406,1996.
Singh,S.P.,B.S.Adhikari,andD.B. Zobel,Biomass,
productivity,
leaf
longevity
andforeststructure
in thecentralHimalaya,
Ecol.Monogr.,
64, 401-421, 1994.

Sombroek,
W.G.,F.O.Nachtergaele,
andA. Hebel,Amounts,
dynamics
andsequestering
of carbonin tropicalandsubtropical
soils,Ambio,22,
417-429, 1993.

Sprugel,D.G., M.G. Ryan,J.R. Brooks,K.A. Vogt,andT.A. Martin,

Respiration
fromtheorganlevelto thestand,in Resource
Physiology
of Conikrs,
pp.255-300,Academic,
SanDiego,Calif.,1995.
Steffen,W.L., B.H. Walker,J.S.Ingram,andG.W. Koch,(Eds.),Global
changesand terrestrialecosystems:
The operational
plan, Global
ChangeRep. 21, 95 pp., Int. Geosphere-Biosphere
Programme,
Stockholm,1992.

Steffen,W.L., et al., The terrestrial

carboncycle: Implications
for the
KyotoProtocol,Science,280, 1393-1394,1998.

Thompson,
S.L.,andD. Pollard,A globalclimatemodel(GENESIS)with
a land-surface-transfer
scheme
(LSX), 1, Present-day
climate,
J. C!im.,
8, 732-761, 1995a.

Thompson,
S.L.,andD. Pollard,A globalclimatemodel(GENESIS)with
a land-surface-transfer
scheme
(LSX), 2, CO2sensitivity,
J. C!im.,8,
1104-1121, 1995b.

United Nations Educational,Scientific,and Cultural Organization

(UNESCO),WorldWaterBalanceand WaterResources
of theEarth,
editedby V.I. Korzun,663 pp.,Paris,1978.
Verberne,E.L., J. Hassink,P. De Willigen,J.J.R.Groot,and J.A. Van
Veen, Modelingorganicmatterdynamicsin differentsoils,Neth. J.
Agr. Sci., 38, 221-238, 1990.

Vinnikov,K.Y., andI.B. Yeserkepova,

Soilmoisture:
Empirical
dataand
modelresults,J. C!im., 4, 66-79, 1991.

Wagai,R., K.R. Brye,S.T. Gower,J.M.Norman,andL.G. Bundy,Land

useandenvironmental
factorsinfluencingsoil surfaceCO2flux and
microbialbiomassin naturaland manageecosystems
in southern

Wisconsin,
Soil Biol. Biochem.,30, 1501- 1509, 1998.
Robock,A., K.Y. Vinnikov,C.A. Schlosser,
N.A. Speranskaya,
and Y.
Xue, Useof midlatitudesoil moistureandmeteorological
observations Walker, B., Landscapeto regional-scaleresponsesof terrestrial
to validate soil moisture simulationswith biosphereand bucket
ecosystems
to globalchange,Ambio,23, 67-73, 1994.
models,J. Clim., 8, 15-35, 1995.

Waring,R.H.,J.J.Landsberg,
andM. Williams,Netprimaryproduction
of

Running,S.W., and S.T. Gower, Forest-BGC,a generalmodelof forest

forests:A constantfractionof grossprimaryproduction?,
Tree
Physio!.,18, 129-134, 1998.
ecosystem
processes
for regionalapplications,
II, Dynamicallocation
andnitrogenbudgets,TreePbysio!.,9, 147-160,1991.
Whitmore, A.P., H. Klein-Gunnewiek,G.J. Crocker, J. Klir, M.
Runyon,J., R.H. Waring, S.N. Goward,and J.M. Welles, Environmental
Korschens,
andP.R.Poulton,Simulating
trendsin soilorganiccarbon
limits on net primaryproductionand light use efficiencyacrossthe
in long-term experimentsusing the Verberne/MOTOR model,
Geoderma,81, 137-151, 1997.
Oregontransect,Ecol.Appl., 4, 226-237, 1994.
Ryan, M.G., S.T. Gower, R.M. Hubbard,R.H. Waring, H.L. Gholz, W.P.
Winston,G.C., E.T. Sundquist,
B.B. Stephens,
andS.E.Tumbore,Winter
CropperJr., andS.W. Running,Woodytissuemaintenance
respiration
CO2 fluxesin a borealforest,J. Geophys.Res, 102(D24),2879528804, 1997.
of four conifersin contrastingclimates,Oecologia,101, 133-140,
1995.

Saldarriaga,J.G., and R.J. Luxmoore,Solar energyefficiencyduring

succession
of a tropical rain forestin Amazonia,J. Trop. Eco!., 91,
233-242, 1991.

Sands,P.J., Modelling canopyproduction,I, Optimal distributionof

photosynthetic
resources,
Aust.J. Plant. Pbysio!.,22, 593-601, 1995.
Sauer, T.J., J.M. Norman, C.B. Tanner and T.B. Wilson, Measurement of

heatandvaportransfer-coefficients
at the soil surfacebeneatha maize
canopyusingsourceplates,Agric. For. Meteorol., 75(1-3), 161-189,
1995.

Schimel, D.S., Terrestrial ecosystemsand the carbon-cycle,Global

ChangeBio!., 1, 77-91, 1995.

Woodward,
F.I., T.M. Smith,andW.R. Emanuel,
A globallandprimary
productivity
andphytogeography
model,GlobalBiogeochem.
Cycles,
9(4), 471-490, 1995.
Zak, D.R., D. Tilman, R.P. Paramenter,C.W. Rice, F.M. Fisher, J. Vose,

D. Milchunas, and C.W. Martin, Plant production and soil

microorganisms
in late-successional
ecosystems:A continental-scale
study,Ecology,75, 2333-2347, 1994.
Zinke, P.J., A.G. Stangenberger,W.M. Post, W.R. Emanuel,and J.S.
Olson,Worldwideorganiccarbonand nitrogendata,ORNL/TM-8857,
Oak RidgeNat. Lab.,Oak Ridge,Tenn.,1986.
Zobler, L., A world soil file for global climate modeling, NASA/TM87802, (NASA), Washington,D.C., 1986.

KUCHARIK

ET AL.: TESTING A DYNAMIC

M.T. Coe, C. Delire, V. A. Fisher,J. A. Foley, C. J. Kucharik and J.

D. Lenters,Climate,People,and EnvironmentProgram(CPEP), Institute
for EnvironmentalStudies,Universityof Wisconsin,1225 West Dayton
Street,Madison,WI 53706.(kucharikfacstaff.wisc.edu)
S. T. Gower, Department of Forest Ecology and Management,
Universityof Wisconsin,1630LindenDrive, Madison,Wl 53706.

GLOBAL ECOSYSTEM

MODEL

825

J. M. Norman, Departmentof Soil Science,Universityof Wisconsin,

1525 ObservatoryDrive, Madison,W1 53706.
C. Young-Molling, Space Scienceand EngineeringCenter (SSEC),
Universityof Wisconsin,1225 West Dayton Street,Madison,WI 53706.
(ReceivedJanuary25, 1999; revisedOctober14, 1999;
acceptedOctober14, 1999.)