North American Carbon Program (NACP) regional interim synthesis: Terrestrial biospheric model intercomparison

Nb

DJCJa

b

c

d

e

f

g

h

i

j

k

l

m

n

o

p

q

r

s

t

u

v

a

ARRA

KTICNR

0d

Ecological Modelling 232 (2012) 144ndash 157

Contents lists available at SciVerse ScienceDirect

Ecological Modelling

jo u r n al hom ep age wwwelsev ier com locate eco lmodel

orth American Carbon Program (NACP) regional interim synthesis Terrestrialiospheric model intercomparison

N Huntzingeralowast WM Postb Y Weib AM Michalakc TO Westd AR Jacobsonef IT BakergM Chenh KJ Davis i DJ Hayesb FM Hoffmanb AK Jain j S Liuk AD McGuire l RP Neilsonmhris Pottern B Poultero David Pricep BM Raczka i HQ Tianq P Thorntonb E Tomelleri r N Viovyo

Xiaos W Yuant N Zengu M Zhaov R Cookb

School of Earth Science and Environmental Sustainability Northern Arizona University PO Box 5694 Flagstaff AZ 86011-5694 United StatesEarth Science Division Oak Ridge National Laboratory Oak Ridge TN United StatesDepartment of Global Ecology Carnegie Institute for Science Stanford CA United StatesJoint Global Change Research Institute College Park MD United StatesNOAA Earth System Research Lab Global Monitoring Division Boulder CO United StatesCooperative Institute for Research in Environmental Sciences University of Colorado Boulder CO United StatesDepartment of Atmospheric Sciences Colorado State University Fort Collins CO United StatesDepartment of Geography and Program in Planning University of Toronto Toronto Ontario CanadaDepartment of Meteorology The Pennsylvania State University University Park PA United StatesAtmospheric Sciences University of Illinois Urbana Champaign Urbana IL United StatesUnited States Geologic Survey National Center for EROS Sioux Falls SD United StatesUS Geological Survey Alaska Cooperative Fish and Wildlife Research Unit University of Alaska Fairbanks Fairbanks AK United StatesDepartment of Botany and Plant Pathology University of Utah Salt Lake City UT United StatesNASA Ames Research Center Moffett Field CA United StatesLaboratoire des Sciences du Climat et de lrsquoEnvironnement LSCE Gif sur Yvette FranceNorthern Forestry Centre Natural Resources Canada Edmonton Alberta CanadaEcosystem Dynamics and Global Ecology Laboratory Auburn University Auburn AL United StatesMax Planck Institute for Biogeochemistry Jena GermanyEarth Systems Research Center Institute for the Study of Earth Oceans and Space University of New Hampshire Durham NH United StatesCollege of Global Change and Earth System Science Beijing Normal University Beijing ChinaDepartment of Atmospheric and Oceanic Science University of Maryland College Park MD United StatesNumerical Terradynamics Simulation Group University of Montana Missoula MT United States

r t i c l e i n f o

rticle historyeceived 5 October 2011eceived in revised form 7 February 2012ccepted 8 February 2012

eywordserrestrial biospheric modelsntercomparisonarbon fluxesorth American Carbon Programegional

a b s t r a c t

Understanding of carbon exchange between terrestrial ecosystems and the atmosphere can be improvedthrough direct observations and experiments as well as through modeling activities Terrestrial biospheremodels (TBMs) have become an integral tool for extrapolating local observations and understandingto much larger terrestrial regions Although models vary in their specific goals and approaches theircentral role within carbon cycle science is to provide a better understanding of the mechanisms currentlycontrolling carbon exchange Recently the North American Carbon Program (NACP) organized severalinterim-synthesis activities to evaluate and inter-compare models and observations at local to continentalscales for the years 2000ndash2005 Here we compare the results from the TBMs collected as part of theregional and continental interim-synthesis (RCIS) activities The primary objective of this work is tosynthesize and compare the 19 participating TBMs to assess current understanding of the terrestrialcarbon cycle in North America Thus the RCIS focuses on model simulations available from analysesthat have been completed by ongoing NACP projects and other recently published studies The TBM flux

estimates are compared and evaluated over different spatial (1 times 1 and spatially aggregated to differentregions) and temporal (monthly and annually) scales The range in model estimates of net ecosystemproductivity (NEP) for North America is much narrower than estimates of productivity or respirationwith estimates of NEP varying between minus07 and 22 PgC yrminus1 while gross primary productivity andheterotrophic respiration vary between 122 and 329 PgC yrminus1 and 56 and 132 PgC yrminus1 respectivelyThe range in estimates from the models appears to be driven by a combination of factors includingthe representation of photosynthesis the source and of environmental driver data and the temporal

lowast Corresponding author Tel +1 928 523 1669 fax +1 928 523 7423E-mail address deborahhuntzingernauedu (DN Huntzinger)

304-3800$ ndash see front matter copy 2012 Elsevier BV All rights reservedoi101016jecolmodel201202004

DN Huntzinger et al Ecological Modelling 232 (2012) 144ndash 157 145

variability of those data as well as whether nutrient limitation is considered in soil carbon decompositionThe disagreement in current estimates of carbon flux across North America including whether NorthAmerica is a net biospheric carbon source or sink highlights the need for further analysis through theuse of model runs following a common simulation protocol in order to isolate the influences of modelformulation structure and assumptions on flux estimates

1

(bH(c2sren22ccfapcsatePip

toaflriccrcuo2

ititraiuriuses

scribing consistent driver data and a detailed simulation protocol)in order to better understand what is driving the differences amongmodel estimates this approach provides an unrealistic assessment

Introduction

North America has been identified as both a significant sourceeg fossil fuel emissions) and biospheric sink of atmospheric car-on dioxide (CO2) (Gurney et al 2002 CCSP 2007 Prentice 2001)owever as summarized in the State of the Carbon Cycle Report

SOCCR CCSP 2007) estimates of the North American biospherearbon sink vary widely ranging from less than 01 PgC yrminus1 to over0 PgC yrminus1 While some of the mechanisms responsible for thisink are understood (eg forest regrowth) the current and futureole of other mechanisms such as extreme weather events (Jentscht al 2007) changes in land-use CO2 and nitrogen fertilizationatural disturbances (eg Kurz et al 2007 Bond-Lamberty et al007) and other carbon-climate feedbacks (Friedlingstein et al006 Pan et al 1998) in controlling the North American carbonycle are highly uncertain (CCSP 2007) Thus a basic goal of carbonycle studies has been to address key scientific questions rangingrom carbon flux diagnosis (What are net carbon sources and sinksnd how do they change with time) to attribution (What are therocesses controlling flux variability) and prediction (How mighthanges in climate and other factors alter future fluxes) Under-tanding the sources and sinks of carbon and their distributioncross North America is critical for the successful management ofhe carbon cycle (CCSP 2007) and for useful predictions of its futurevolution and requires a strong understanding of carbon dynamicsroviding useful information about the carbon cycle and project-ng future CO2 concentrations is also urgently needed for informingolicies addressing fossil fuel emissions

Understanding of carbon exchange between terrestrial ecosys-ems and the atmosphere can be improved through directbservations and experiments as well as through modelingctivities Terrestrial biosphere models (TBMs) sometimes calledorward models have become an integral tool for extrapolatingocal observations and understanding to much larger terrestrialegions (Waring and Running 2007 Davis 2008) as well as for test-ng hypotheses about how ecosystems will respond to changes inlimate and nutrient availability Although TBMs vary in their spe-ific goals and approaches their central role within carbon cycleesearch is to provide a better understanding of the mechanismsurrently controlling carbon exchange This understanding is thensed as the basis of prediction and ultimately to inform the devel-pment of any potential carbon management plans (Schimel et al000)

The ultimate objective is to model all the processes that resultn the net carbon exchange between the terrestrial system andhe atmosphere called the net ecosystem exchange (NEE) Thisncludes many processes most importantly gross primary produc-ion (GPP) autotrophic and heterotrophic respiration (Ra and Rhespectively which together add up to ecosystem respiration Re)nd losses due to fire and other disturbance processes (herbivorynsects disease physical disturbance from storms etc) Thereforenderstanding how TBM estimates of ecosystem photosynthesisespiration and net carbon exchange vary spatially and temporallys of great importance not only for improving TBMs but also fornderstanding their contribution to uncertainty in global climate

imulations By extension it is also important to know why differ-nt TBMs product different estimates even when forced with theame driving conditions The former can be examined by bringing

copy 2012 Elsevier BV All rights reserved

together existing model results and comparing them within a con-sistent framework while the later requires a substantial formalintercomparison effort

Individual TBMs are often based on different simplifyingassumptions use different environmental driving data and ini-tial conditions and formulate the processes controlling carbonexchange in different ways Thus there is diversity in both the com-plexity of the model structure and formulation as well as modelestimates of regional net carbon exchange Each TBM therefore isa complex combination of scientific hypotheses and choices andtheir estimates depend on these inherent assumptions (Beer et al2010) Available observations of carbon flux components as wellas our current understanding of the processes controlling carbonexchange over regional scales however are not sufficient to rankmodels in terms of which is ldquobestrdquo at representing current fluxesor predicting carbon exchange under future climate conditions(Melillo et al 1995) Therefore in order to move towards morerobust estimates of carbon cycle dynamics we must first compareestimates from a variety of model types as well as evaluate esti-mates against those measurements that are available (Cramer et al1999 Melillo et al 1995 Beer et al 2010)

Recently the North American Carbon Program (NACP) (Denninget al 2005 Wofsy and Harriss 2002) organized several interim-synthesis activities to evaluate and inter-compare models andobservations at local to continental scales for the time period of2000 through 2005 These interim synthesis activities include threecompanion studies each conducted on different spatial scales (1)site-level analyses that examine process-based model estimatesand observations at over 30 AmeriFlux1 and Fluxnet-Canada2 towersites across North America (2) a regional mid-continent inten-sive study centered in the agricultural regions of the United Statesand focused on comparing inventory-based estimates of net car-bon exchange with those from atmospheric inversions and (3)a regional and continental synthesis evaluating model estimatesagainst each other and available inventory-based estimates acrossNorth America A number of other interim syntheses are underwayincluding ones focusing on non-CO2 greenhouse gases the impactof disturbance on carbon exchange and coastal carbon dynamics

Here we compare the model estimates from the regional andcontinental interim-synthesis (RCIS) activities The primary objec-tive of this work is to synthesize and compare TBMs to assesscurrent understanding of the terrestrial carbon cycle in NorthAmerica Thus the RCIS focuses on ldquooff-the-shelfrdquo model simula-tions ie existing model results currently available from analysesthat have been completed by ongoing NACP projects and otherrecently published studies Although there is a challenge in inter-preting existing results compared to prescribing new simulationsdesigned for the controlled comparison of different modeling sys-tem there is also great value in using independent estimates toassess the overall spread or variability in model results While it isnecessary to limit variability between models (by for example pre-

1 httppublicornlgovameriflux2 httpwwwfluxnet-canadaca

1 cal Mo

obpspTrecteposaNadaitmispbdimi

2

hvsdctvbotttatotep

psasmptf

iia

46 DN Huntzinger et al Ecologi

f the true uncertainty in our ability to model land-atmosphere car-on exchange Models differ structurally in how they represent therocesses controlling carbon exchange between the land and atmo-phere in their input or driver data (land cover climate) and in thearameter values used within their varying process descriptionshese varying approaches to modeling terrestrial carbon exchangeesult in a large degree of variability in the land-atmosphere fluxstimates Thus this work provides a valuable assessment of theurrent status of terrestrial carbon modeling in NA by bringingogether model estimates that incorporate a wide range of mod-ling choices and input data This work also serves as a startingoint for analyses that compare these model results to differentbservational data products Specifically Raczka and Davis (per-onal communication) evaluated flux estimates of RCIS modelsgainst observations from 30 flux towers across a wide range ofA ecosystems In addition Hayes et al (2012) has assembled andnalyzed available agricultural and forest biomass inventory-basedata for NA and compared them alongside estimates from TBMnd inverse approaches available from the RCIS In addition ongo-ng work is comparing TBM estimates of net ecosystem exchangeo flux estimates derived from atmospheric inversions Flux esti-

ates from atmospheric inverse models are more comprehensiven the sense that all ecosystem sources and sinks fossil fuel emis-ions and any other processes emitting or absorbing CO2 are inrinciple captured in the atmospheric signal (GCP 2010) Com-ined the comparison of TBM estimates to different observationalata products and modeling approaches can provide further insight

nto our ability to model land-atmosphere carbon dynamics Thisanuscript provides the foundation for these types of compar-

sons

Overview of participating models

TBMs represent processes controlling carbon cycle dynamicsowever the level of detail with which processes are representedaries across models Whereas some models are empirically ortatistically-based with relatively simple relationships betweenriver variables and flux others are more complex simulating theoupled carbon nutrient and water cycles in terrestrial ecosys-ems Models also differ in their representation of soil propertiesegetation type and environmental forcings as well as how car-on pools are initialized Here we compare carbon flux estimatesver North America (NA) for the 19 TBMs that participated inhe RCIS Key features of the models participating in this study inerms of how they represent photosynthesis autotrophic respira-ion decomposition and other processes affecting carbon fluxesre summarized in Tables 1ndash3 (see Supplemental Material for addi-ional model descriptions) The TBM flux estimates are evaluatedver different land cover regions of NA and with respect to pho-osynthetic formulation soil carbon dynamics and whether theyxplicitly account for the impact of fire disturbances on carbonools and stocks

TBMs can be divided into two general classes diagnostic andrognostic models In order to specify the internal (time-varying)tate of the system diagnostic models rely on forcing data (eg leafrea) provided directly or indirectly from satellite or other externalources In contrast the internal states of the system in prognosticodels are computed as part of the system equations Therefore in

rinciple prognostic models can be used to predict future condi-ions using external climate forcing alone in addition to being usedor diagnostic analyses (eg reproducing past or measured fluxes)

The distinction between diagnostic and prognostic models ismportant Diagnostic models frequently use observed leaf areandex (LAI) as a specified driving variable along with empiricallgorithms of varying complexity to estimate fluxes over regional

delling 232 (2012) 144ndash 157

or global domains and changes in carbon pool over time (Table 2models BEPS CASA NASA-CASA CASA GFEDv2 EC-MOD EC-LUEISAM MODIS MOD17+) Conversely prognostic models determinethe amount of leaf area as the result of carbon allocation andwater balance dynamics within the model As a result they canproject or estimate carbon cycle dynamics into the future underchanging environmental conditions (Can-IBIS CLM-CASAprime CLM-CN DLEM LPJ-wsl MC1 ORCHIDEE SiB31 TEM6 VEGAS2) Inaddition some prognostic models also contain dynamic algorithmsto estimate vegetation distribution over time (Can-IBIS LPJ-wslMC1 ORCHIDEE and VEGAS2) Although prognostic models canbe used for future predictions they are much less constrained byobservations than diagnostic models As a result one would expecttheir results to be more variable (and perhaps less reliable) evenwhen used in a diagnostic mode

The model results submitted to the interim synthesis activityalso vary in terms of the processes included the choice of driv-ing data and the types of algorithms employed to represent theseprocesses (Tables 1ndash3 Supplementary Information) For exampleeight of the nineteen models represent photosynthesis using anenzyme kinetic formulation (Farquhar et al 1980) normally ata sub-daily time step while nine of the models use a light-useefficiency calculation at daily to monthly time steps The modelsalso differ in how they model soil carbon decomposition Five ofthe models use a zero-order calculation where decomposition isa function of temperature and moisture only Two of the modelsomitted soil carbon decomposition altogether and the remainderof the models represent decomposition through first-order kinet-ics where decomposition depends on the magnitude of soil carbonstocks in addition to environmental drivers and interactive pro-cesses such as N dynamics In addition models differ in the types ofdisturbance considered (eg wind or storm fire disease) and howthese disturbances are included within the model (eg explicitlydescribed or implicitly accounted for through vegetation indices)Most of the models in this study do not directly account for theimpacts of fire disease or storm events on carbon fluxes or poolsIn addition those that do include the impact of fire disturbances(eg Can-IBIS TEM6 MC1 LPJ-wsl) do so in varying ways (refer toTable 3 and Supplementary Information)

This diversity in model structure and process representationmakes evaluation and comparison of model performance challeng-ing However information on model differences helps to inform theanalysis and was used here to define subsets or groups of modelsbased on specific defining characteristics and aid in the interpre-tation of observed differences

3 Methods for comparison

Prior to analysis all model output was processed as necessaryto a spatial resolution of one-degree by one-degree temporallyaggregated to monthly fluxes and placed on a grid with a spatialextent of 10ndash84 North and 50ndash120 West Fluxes are comparedfor the six years covering the period of 2000 through 2005

31 Regional analysis of TBM output

Several of the model estimates lack full spatial coverage of NorthAmerica (Fig 1) therefore in order to better compare net fluxacross models 1 times 1 flux estimates were spatially aggregated toregions defined by the TransCom intercomparison study (Gurneyet al 2002) and the Global Land Cover classification for 2000

(GLC2000 Latifovic et al 2004 NRCan and USGS 2003) The aggre-gation of fluxes to large contiguous regions with similar land coveror biome types and climatic conditions allows for the examinationof regional differences between the models This approach is similar

DN Huntzinger et al Ecological Modelling 232 (2012) 144ndash 157 147

Table 1Terrestrial biospheric models participating in the NACP regional interim synthesis

Model Spatial range Native spatialresolution

Native temporalresolution

Fluxes submitted Temporal range Selected references

Can-IBIS Canada and US ndash 30 min GPP NEE NEP NPP Ra Rh 2000ndash2005 Wang et al (2011)Kucharik et al (2000)and Foley et al (1996)

CLM-CASAprime Global 28 20 min GPP NPP Rh NEE NEP 2000ndash2004 Randerson et al (2009)CLM-CN Global 28 20 min GPP NPP Rh NEE NEP 2000ndash2004 Thornton et al (2009)

and Randerson et al(2009)

DLEM N America 32 km Daily GPP NEE NPP Ra Rh 2000ndash2005 Tian et al (2010)ISAM N America 1 Weekly NEE Rh NPP 2000ndash2005 Jain and Yang (2005)

and Yang et al (2009)LPJ-wsl N America 05 Daily GPP NPP Rh NEE CFire NEEF 2000ndash2005 Bondeau et al (2007)

and Sitch et al (2003)MC1 Global Continental US 05 Monthly NPP Rh NEE CFire NEEF 2000ndash2005 Bachelet et al (2000)

Daly et al (2000) andLenihan et al (2008)

ORCHIDEE Global 05 30 min GPP NPP Rh NEE CO2Flux 2000ndash2005 Krinner et al (2005)and Viovy et al (2000)

SiB3 Global 1 Hourly NEE GPP Reco 2000ndash2005 Baker et al (2008)TEM6 N A gt 45N 05 Monthly GPP NPP Rh NEE CFire NECB 2000ndash2005 McGuire et al (2010)

and Hayes et al (2011)VEGAS2 N America 1 Daily GPP NPP Ra Rh NEE CFire 2000ndash2005 Zeng (2003) and Zeng

et al (2004 2005)BEPS N America 1 Hourly GPP NEE NEP NPP Rh 2000ndash2004 Chen et al (1999) and

Ju et al (2006)CASA Global 1 Monthly NEE 2002ndash2003 Randerson et al (1997)NASA CASA Continental US 8 km Monhly NPP Rh NEE NEP 2001ndash2004 Potter et al (2007)CASA GFEDv2 Global 1 Monthly GPP NPP Rh CFire NEE 2000ndash2005 van der Werf et al

(2004 2006)EC-LUE N America 1 Weekly GPP 2004ndash2005 Yuan et al (2007)EC-MOD N America 1 8-Day GPP NEE 2000ndash2006 Xiao et al (2008 2010

2011)MODIS N America ndash 8-Day GPP annual NPP 2000ndash2005 Heinsch et al (2003)

and Running et al(2004)

MOD17+ Global 05 Daily GPP NEE Reco 2000ndash2004 Reichstein et al (2005)

G produe chang(

ttb(cmefd

echroctmi

3

et(aAt

ross primary productivity (GPP) net ecosystem exchange (NEE) net ecosystemrotrophic respiration (Rh) carbon emissions from fires (CFire) net ecosystem exCO2Flux) ecosystem respiration (Reco) net ecosystem carbon balance (NECB)

o that used by Kicklighter et al (1999) where net primary produc-ivity (NPP) estimates were averaged across global biomes definedy the potential natural vegetation map developed by Melillo et al1993) The choice of land cover classification for defining spatiallyontiguous regions is somewhat subjective As with the Potsdamodel intercomparison study (eg Cramer et al 1999 Kicklighter

t al 1999) landcover classification is used here solely as a maskor flux aggregation to smaller regions in order to examine regionalifferences among models

The models used (or prognostically generated) different veg-tation maps with varying classification schemes Therefore thehoice of land cover scheme applied in this analysis does not reflectow well a model predicts flux for a particular biome type butather how predicted fluxes compare over large spatially contigu-us regions with similar land cover or climatic conditions To avoidomparing models with limited spatial coverage in a region onlyhose models with at least 80 representation (ie those that esti-

ate fluxes for at least 80 of the cells) in a given land region werencluded in the comparison within that region

2 Subsetting models based on model formulation

In addition to comparing aggregated carbon fluxes fluxstimates were also compared by grouping models by their pho-osynthetic formulation and treatment of soil carbon dynamics

Table 2) Both the spread in model estimates and the across-modelverage for these different subsets were evaluated and compareds mentioned above the models in this study can be divided into

wo predominant photosynthetic formulation classes light-use

ctivity (NEP) net primary productivity (NPP) autotrophic respiration (Ra) het-e including fire emissions (NEEF) net carbon flux including fire and disturbance

efficiency (LUE) and enzyme kinetic (EK) Light-use efficiencymodels estimate productivity by quantifying the fraction of pho-tosynthetically active radiation (fPAR) absorbed by the vegetationand then adjust the conversion of solar energy to photosynthesisor biomass production through climatological and physiologicalrestrictions (eg temperature moisture) Thus carbon fixation isa strong function of solar radiation and leaf area index (LAI) or aproxy such as normalized vegetative difference index (NDVI) Incontrast models with enzyme kinetic formulations are more phys-iologically based simulating photosynthesis using equations thatrepresent biochemicalbiophysical reactions driven by absorbedPAR atmospheric CO2 concentration leaf temperature and leafwater status (Farquhar et al 1980) Thus EK models quantify pho-tosynthesis by emphasizing the light and enzyme limiting rates thataffect photosynthesis In addition to LUE and EK formulations somemodels employ more statistical or regression-based approachesmodeling productivity as an empirical function of different envi-ronmental drivers Photosynthetic formulation controls to someextent estimates of carbon uptake or productivity predicted bythe models Photosynthesis can also be influenced by other factorsincluding driving meteorology atmospheric CO2 concentrationnutrient availability and moisture and temperature limitations

In addition to photosynthesis models were grouped based ontheir treatment of soil carbon dynamics and decomposition TheCO2 released (ie heterotrophic respiration Rh) from the decompo-

sition of above and below-ground dead organic matter is controlledby three factors including substrate quality and quantity moistureavailability and temperature (Waring and Running 2007) Thusthe degree to which these limitations are accounted for in the model

148 DN Huntzinger et al Ecological Modelling 232 (2012) 144ndash 157

Table 2Comparison of environmental drivers vegetation and soil distribution phenology compartments and photosynthetic and soil carbon decomposition formulations amongmodels

Modela Vegetationdistribution

Soil distribution Weatherclimatedata

Phenology PFTs Vegpools

Soilpools

Photo-syntheticformulationb

Soil carbondecomposition

Can-IBIS Dynamic CSL (Canada)STATSGO (Alaska)VEMAP (cont US)

Canadian ForestServices (CFS)

Prognostic 12 3 7 EK 1st Order

CLM-CASArsquo MODIS IGBP-DIS (GSDTG2000)

NCEP reanalysis Prognostic 15 3 5 EK 1st Order

CLM-CN MODIS IGBP-DIS (GSDTG2000)

NCEP reanalysis Prognostic 15 4 7 EK 1st Order with N

DLEM Multiple sources(Tian et al 2010)

Zobler (1986)FAO(19952003)

NARR and PRISM Prognostic 21 + 10 7 3 EK 1st Order with N

ISAM Loveland andBelward (1997)and Haxeltine andPrentice (1996)

Zobler (1986)FAO(19952003)

Mitchell et al(2005)

ndash 13 5 8 LUE 1st Order with N

LPJ-wsl Dynamic Zobler (1986)FAO(19952003)

CRU TS 30 Prognostic 9 3 2 EK 1st Order

MC1 Dynamic STATSGO PRISM Prognostic 6 7 6 Statistical 1st Order with NORCHIDEE Dynamic Zobler (1986)FAO

(19952003)CRU05 and NCEPreanalysis

Prognostic 12 8 8 EK 1st Order with N

SiB3 IGBP IGBP-DIS (GSDTG2000)

NARR MODIS LAI 14 1 0 EK Zero Order

TEM6 Loveland et al(2000) and Hurttet al (2006)

IGBP-DIS (GSDTG2000)

CRU05 and NCEPreanalysis

Prognostic 23 1 3 EK 1st Order with N

VEGAS2 Dynamic Related tovegetation

CRU05 and NCEPreanalysis

Prognostic 4 3 6 LUE 1st Order

BEPS GLC2000 STATSGO (SSS2011)

NCEP reanalysis VGETATION LAI 6 4 9 EK 1st Order with N

CASA DeFries andTownshend (1994)

Zobler (1986)FAO(19952003)

Leemans andCramer (1991) andHansen et al(1999)

GIMMS NDVIderived LAI

11 3 5 LUE 1st Order

NASA CASA MODIS STATSGO (SSS2011)

NCEP reanalysis MODIS EVI 11 3 5 LUE 1st Order with N

CASA GFEDv2 MODIS Batjes (1996) IISAS GISSTEMPand GPCPv2

GIMMS NDVIderived LAI

3 3 5 LUE 1st Order

EC-LUE ndash ndash GMAODAO MODIS NDVI ndash ndash ndash LUE ndashEC-MOD MODIS ndash ndash MODIS EVI LAI 7 0 0 statistical Zero OrderMODIS MODIS ndash DAO MODIS LAI ndash 0 ndash LUE ndashMOD17+ SYNMAP Jung et al

(2006)ndash ERA-Interim

reanalysisMODIS LAI 10 0 0 LUE Zero Order

S the mupple

wd

ectekop(icbsalofc

dlo

haded boxes refer to model components that are not considered or needed withina Model acronyms are defined and additional model information is provided in Sb Enzyme kinetic (EK) and light-use efficiency (LUE)

ill likely impact their estimations of Rh and overall net carbonynamics

Some models lack soil carbon poolslayers altogether and het-rotrophic respiration is thus not explicitly calculated Othersalculate soil respiration as an empirical function of moisture andemperature conditions (eg zero-order) In most models how-ver soil organic matter decomposition is based on first-orderinetics where the rate of decomposition is a function of the sizef the soil carbon pool (eg amount of carbon) a simple decom-osition constant as well as temperature and moisture limitationsReichstein and Beer 2008) The influence of nitrogen (N) dynam-cs and cycling on soil carbon decomposition may or may not beonsidered by the model (Table 2) In this analysis two soil car-on dynamics classifications are used models with (1) dynamicoil carbon pools with first-order soil carbon decomposition ratesnd (2) dynamic soil carbon pools that include nitrogen cycling andimitations with first-order soil carbon decomposition rates A fewf the models consider zero-order soil decomposition and there-ore lack soil carbon pools altogether and were not included in theomparison of heterotrophic respiration

Models were also classified by other factors that affect theirynamics including whether they consider fire disturbances and

and-use change and whether transient CO2 or the combinationf transient CO2 and N deposition forcings are included within the

odelmentary Information

model (Tables 1 and 3) Although many of these classifications arenot mutually exclusive (eg many prognostic models use an EKphotosynthetic formulation) their use of in model evaluation helpsto identify potential sources of variability that drive differencesin GPP and Rh which translate into differences in net ecosystemproductivity (NEP)

4 Results and discussion

41 Magnitude and distribution of carbon sources and sinks

The carbon flux that all the models submitted to the RCIShave in common is net ecosystem production (NEP) where NEPis the difference between GPP and the sum of autotrophic and het-erotrophic respiration (Chapin et al 2006) NEP does not includedirect disturbance-induced carbon fluxes which many models inthis study do not consider If a model does consider disturbances(Table 3) however this can alter carbon pools and as a resultimpact both NPP and Rh In some models such as Can-IBIS the

effects of disturbances on NEP are only accounted for at yearrsquos endAs a result if NEP is compared over the summer months (June JulyAugust) the flux estimates from these months will not account forlosses due to disturbance Instead disturbances will cause additions

DN Huntzinger et al Ecological Modelling 232 (2012) 144ndash 157 149

Table 3Components and processes (including disturbance events) influencing the estimation of net ecosystem productivity by each model

Modela NEPb Land-uselandcover change

Firec Insect stormdamaged

Transient forcingse DIC DOCPC lossesf

Can-IBIS GPP minus (Ra + Rh) ndash Prognostic ndash CO2 Ndep DOCCLM-CASArsquo GPP minus (Ra + Rh) Prescribed

land-usendash ndash CO2 ndash

CLM-CN GPP minus (Ra + Rh) Prescribedland-use

Prognostic ndash CO2 Ndep ndash

DLEM GPP minus (Ra + Rh) Prescribedland-use

ndash ndash CO2 Ndep CH4 loss

ISAM NPP minus Rh Prescribedland-use

ndash ndash CO2 Ndep ndash

LPJ-wsl GPP minus (Ra + Rh) ndash Prognostic ndash CO2 ndashMC1 NPP minus Rh Prescribed

land-useprognostic forestharvest

Prognostic ndash CO2 Ndep ndash

ORCHIDEE GPP minus (Ra + Rh) minus cropharvest

No land-useland-cover change 40of croplandbiomass isharvested

ndash ndash CO2 ndash

SiB31 GPP minus (Ra + Rh) ndash ndash ndash CO2 ndashTEM6 GPP minus (Ra + Rh) Prescribed

land-use andforest harvest

Prescribed ndash CO2 Ndep DOC

VEGAS2 GPP minus (Ra + Rh) ndash ndash Constantbackgroundmortality ratefrom cold anddrought stress

CO2 ndash

BEPS GPP minus (Ra + Rh) ndash ndash ndash CO2 ndashCASA NPP minus Rh ndash ndash ndash ndash ndashNASA CASA NPP minus Rh ndash ndash ndash CO2 Ndep ndashCASA GFEDv2 NPP minus Rh ndash Prescribed ndash ndash ndashEC-LUE GPP only ndash ndash ndash ndash ndashEC-MOD -NEE ndash ndash ndash ndash ndashMOD17+ GPP minus Re ndash ndash ndash ndash ndash

Shaded boxes refer to processes that are not included or considered in the modela Model acronyms are defined and additional model information is provided in Supplementary Informationb Net ecosystem productivity (NEP) gross primary productivity (GPP) heterotrophic respiration (Rh) autotrophic respiration (Ra)c Models without prognostic or prescribede Transient atmospheric carbon dioxide concentration (CO2) transient nitrogen deposition (Ndep)

arbon

tw

NaoraipargvsUwibcma

Nc

f Dissolved inorganic carbon (DIC) dissolved organic carbon (DOC) particulate c

o litter pools and removals of live vegetation at year end whichill affect the NEP in the following (and subsequent) years

The spatial distribution of average summer (June July August)EP predicted by the models is shown in Fig 1 Table 3 provides

list of processes or factors that influence each modelrsquos estimatef productivity Although as mentioned above the direct and indi-ect effects of fires influence some model estimates of carbon fluxnd pools direct CO2 emissions from forest fires are not includedn model NEP estimates Throughout the following discussion aositive (+) sign on NEP indicates net uptake of carbon from thetmosphere by the land while a negative (minus) sign signifies a netelease of carbon from the land back to the atmosphere During therowing season the magnitude and spatial distribution of fluxesary substantially among the models (Fig 1) Some models showtrong carbon sources in the Midwest and Southeast portions of theS (eg MC1 LPJ-wsl) Central Plains West and Southwest (LPJ-sl MOD17+ DLEM) while others estimate large sinks particularly

n the Southeast (eg BEPS EC-MOD NASA-CASA Can-IBIS) In theoreal regions of North America however there appears to be moreonsistency among the models In these northern regions mostodels show an overall sink of carbon during the summer months

lthough the strength of that sink varies across models (Fig 1)The overall similarities and differences among modeled mean

EP estimates were quantified for each one-degree cell by cal-ulating the across-model standard deviation in estimated flux

(PC)

During the summer months of June July and August the largestdifferences between NEP estimates are located in the Midwesternand Southeast regions of the continental US (Fig 2) Much of theacross-model spread in summertime NEP in the southeast is drivenby differences in predicted GPP (Fig 2) Overall as expected thegreatest difference in model estimates occurs in areas of larger fluxmagnitude

When fluxes are spatially aggregated to all of North America theTBMs predict annual NEP ranging from minus07 to +17 PgC yrminus1 forprognostic models and minus03 to +22 PgC yrminus1 for diagnostic mod-els with an overall model average of +065 PgC yrminus1 for the NorthAmerican continent (Table 4) This model average is consistentwith previous estimates of the strength of the North Americansink of 035ndash075 PgC yrminus1(Goodale et al 2002 Houghton et al1999 CCSP 2007 Pacala et al 2001 Xiao et al 2011) Much ofthe spread in NEP estimates comes from the range in model esti-mates of photosynthesis or GPP because the majority of modelsscale autotrophic respiration (Ra) based on their estimates of pho-tosynthesis TBM estimates of GPP and heterotrophic respiration forNorth America vary considerably between 122 and 329 PgC yrminus1

and 56 and 132 PgC yrminus1 respectively (Table 4) Overall prognostic

models exhibit greater across-model spread or variability in theirnet GPP estimates relative to diagnostic models Prognostic mod-els also estimate a larger net GPP or uptake across North Americacompared to diagnostic models

150 DN Huntzinger et al Ecological Modelling 232 (2012) 144ndash 157

F by moa ognosw

tLflisneum

Fp

ig 1 Long-term mean summer (June July August) net ecosystem productivity

tmosphere while a negative sign signifies net carbon release to the atmosphere Prith a purple background

One potential reason for the narrower spread in GPP amonghe diagnostic models is that several of the diagnostic models (EC-UE EC-MOD MOD17+) presented in this study are calibrated toux tower data and use similar satellite observations for provid-

ng LAI and fPAR As a result their flux estimates tend to be moreimilar among themselves relative to the differences among prog-

ostic models However only three of the eight diagnostic modelsxplicitly calibrate their models using flux tower data so this isn-likely to be the only cause of similarly among the diagnosticodels

ig 2 Across-model standard deviation in long-term mean (2000ndash2005) summer (Junrimary productivity

del (2000ndash2005) A positive sign indicates net terrestrial carbon uptake from thetic models are shown above with a green background diagnostic models are below

It is surprising that diagnostic models have a greater range andstandard deviation in NEP than prognostic models given that diag-nostic models have smaller ranges in the component fluxes GPPand Rh (Table 4) This indicates that the production and respirationcomponents are less correlated within diagnostic models

Fluxes were also spatially aggregated to Boreal and Temperate

North America regions defined by the TransCom inverse modelintercomparison (Gurney et al 2003) The TransCom regions werechosen for comparison because they cover a majority of NorthAmerica (minus Greenland the Northern Queen Elizabeth Islands

e July August) model estimates of (A) net ecosystem productivity and (B) gross

DN Huntzinger et al Ecological Modelling 232 (2012) 144ndash 157 151

Fig 3 Model estimates of the long-term mean (2000ndash2005) seasonal cycle of (A) net ecosystem productivity and (B) gross primary productivity for boreal and temperateNorth America

Fig 4 Model estimates of annual gross primary productivity (GPP) for 2000 through 2005 for Boreal and Temperate North America Prognostic models are shown in shadesof green diagnostic models are shown in purple

152 DN Huntzinger et al Ecological Modelling 232 (2012) 144ndash 157

Table 4Long-term mean (2000ndash2005) net ecosystem productivity gross primary productivity and heterotrophic respiration estimated by the models in PgC yrminus1 for North AmericaNot all models submitted all three fluxes (NEP GPP and Rh) To avoid comparing models with limited spatial coverage in a region only those models with at least 80representation (ie those that estimate fluxes for at least 80 of the cells) in a given land region were included in the comparison within that region

Prognostic models Diagnostic models

Number of models (min max) Mean Std dev Number of models (min max) Mean Std dev

Net ecosystem productivity (n = 17)North America 9 (minus07 17) 04 04 6 (minus03 22) 09 07Boreal NA 10 (minus02 07) 01 02 4 (minus04 06) 01 03Temperate NA 10 (minus05 11) 02 03 6 (minus01 16) 07 06

Gross primary productivity (n = 15)North America 8 (122 329) 200 66 6 (122 187) 148 19Boreal NA 9 (22 116) 57 27 5 (26 44) 36 06Temperate NA 8 (77 213) 123 40 6 (82 126) 100 10

Heterotrophic respiration (n = 13)North America 8 (56 132) 82 23 2 (74 86) 82 ndashBoreal NA 9 (13 46) 26 11 2 (21 29) 24 ndashTemperate NA 9 (34 75) 48 13 3 (24 56) 45 ndash

Cainsse(

cea

FB

entral America and parts of southern Mexico) Estimates of NEPnd GPP by prognostic versus diagnostic models differ considerablyn both the depth and timing of the seasonal cycle with prog-ostic models estimating greater overall productivity during theummer months compared to diagnostic models (Fig 3) These sea-onal cycle differences translate into large variability in net annualstimates of NEP for 2000ndash2005 ranging from minus04 to 07 PgC yrminus1

Boreal NA) and minus05 to 16 PgC yrminus1 (Temperate NA) (Table 4)

The differences among TBMs are even more apparent when

omparing GPP over similar land regions Overall prognostic mod-ls exhibit a significantly greater across model variability in netnnual uptake than diagnostic models (Table 4 and Fig 4) In order

ig 5 Model estimates of the long-term mean (2000ndash2005) net ecosystem productivity (Niome or vegetative cover classification based on the Global Land Cover 2000 classificatio

to examine regional differences among the models that may becontributing to variation in their estimates of North American netannual flux model estimates of NEP GPP and Rh were comparedacross biomes (Fig 5) To better compare model estimates aggre-gated fluxes were normalized by the total land area covered by agiven model for a given land cover region and therefore the resultsare presented as gC mminus2 yrminus1 Recall that to be included in the com-parison for a given biome a model must have at least 80 spatial

coverage within that region

Model estimates vary considerably in their net annual estimatesof flux with the greatest discrepancies occurring in more pro-ductive regions (eg mixed and deciduous forest cultivated and

EP) gross primary productivity (GPP) and heterotrophic respiration (Rh) by biomen scheme

cal Mo

maathalic(frnbsaeFaRalrar

am

4a

ttrdcosla

4

stvteaftimpetra(

elec

DN Huntzinger et al Ecologi

anaged lands Fig 5) Model estimates of the long-term meannnual NEP in mixed and deciduous forested regions varies frombout minus25 gC mminus2 yrminus1 to +250 gC mminus2 yrminus1 One explanation forhis difference is that models and their estimates of GPP and Rhave varying sensitivities to limitations such as water availabilitynd temperature In low productivity systems (eg shrublands)imitations are likely strong regardless of a given modelrsquos sensitiv-ty to these limitations In more productive systems (eg forests andultivated lands) however a modelrsquos sensitivity to limiting factorseg water availability) will have a much larger effect and slight dif-erences in the sensitivity of GPP and Rh to these limitations couldesult in more divergent NEP estimates In addition from exami-ation of model estimates of long-term mean seasonal cycle at theiome level it appears that across-model differences in growingeason net uptake may be driving some of the average annual NEPnd GPP variability among models Conversely a similar range instimated NEP is seen in areas of cultivated and managed landsor most models NEP is calculated as the difference between GPPnd ecosystem respiration (Rh + Ra) Model estimates of GPP andh vary considerably across biomes However in more productivereas (eg deciduous shrublands evergreen and needleleaf) thearger productivity results in more decomposable substrate As aesult Rh tends to be highly correlated with GPP which yields rel-tively similar estimates of NEP across models compared to otheregions (variability Rh and GPP somewhat cancel each other out)

The potential factors driving the differences seen across modelsre examined further below by subsetting models based on sharedodel attributes

2 Attribution of intermodel differences to model formulationnd driver data

Attribution of intermodel differences in net flux and the long-erm mean seasonal cycle of NEP can best be examined throughhe component fluxes of GPP (photosynthetic uptake) and respi-atory release of carbon (Rh) Thus in order to identify potentialrivers of differences between models we compare estimates ofomponent fluxes (eg GPP and Rh) by subsetting models basedn differences in their photosynthetic and soil carbon decompo-ition formulations as well as their treatment of fire disturbanceand cover change and external forcings such as time-varying CO2nd N deposition

21 Differences in gross primary productivityIt is generally assumed that the physiology of photosynthe-

is and the kinetics of Rubisco are relatively well understood athe leaf-level (Collatz et al 1992 Dai et al 2004 Farquhar andon Caemmerer 1982) However there is a great deal of uncer-ainty as to how to scale leaf-level processes up to the canopy orcosystem level (Chen et al 1999 Baldocchi and Amthor 2001) Inddition there are uncertainties concerning the exact influence ofactors such as nitrogen content nitrogen allocation and radiativeransfer on productivity These processes must be parameterizedn models and can lead to a potentially large spread in GPP esti-

ates across a collection of models The complications in modelingroductivity leads to significant disagreement among the modelstimates of GPP with peak growing season differences of greaterhan 2 PgC monthminus1 in both Temperate and Boreal NA TransComegions (Fig 3) and over 1000 gC mminus2 yrminus1 in regions of mixednd deciduous broadleaf forests and cultivated and managed landsFig 5)

Overall models with photosynthetic formulations based on

nzyme kinetics predict a greater mean annual GPP with aarger range in estimates than light-use efficiency-based mod-ls (Fig 6) Whether photosynthetic formulation is the drivingause of variability in modeled GPP is not clear For example

delling 232 (2012) 144ndash 157 153

Medvigy et al (2010) found that high-frequency meteorologicaldata profoundly impacts simulated terrestrial carbon dynamicsUsing the Ecosystem Demography model version 2 (ED2) forcedwith observed meteorology as well as reanalysis weather thisstudy found that over an 8-year period differences in climaticdriver data alone resulted in a 10 difference in net GPP and 25difference in NEP This work suggests that precipitation and radia-tion data with higher temporal variability yield lower overall GPPand cumulative above ground biomass due to non-linearities in thephotosynthetic functions Conversely climate drivers with lowervariability eg from reanalysis weather products may lead tohigher GPP (Medvigy et al 2010) Model estimates of GPP and NEPare also highly sensitive to biases in solar radiation (eg Ricciutoet al in prep Poulter et al 2011 Zhao et al 2011) Finally manyof the EK models examined in this study also model phenologyprognostically which could also explain much of the spread in GPP(Figs 4 and 6) among the prognostic models Therefore much ofthe spread in GPP estimates in this study is likely to be driven by acombination of differences including the source of driver data thetemporal variability of meteorological data prognostic representa-tion of phenology andor how changes in sunlight and precipitationaffect productivity through the modelsrsquo choice of photosyntheticformulation

Disturbances can have a significant and immediate influence onecosystems by redistributing stocks among live and dead organicmatter pools and in the case of fire the atmosphere Disturbancescan also greatly alter the natural community (eg succession)which can influence biogeochemical cycling long after the directimpacts of a disturbance event have passed To examine the poten-tial impacts of a modelrsquos treatment of disturbance on GPP modelswere grouped based on how they account for fire disturbancesSome models explicitly account for the effect of fire either prog-nostically or diagnostically (refer to Supplemental Information)However a majority of the models in this study do not directlyaccount for fire disturbances or do so implicitly through the useof satellite-based vegetative indices such as LAI or fPAR which arethemselves impacted by fire disturbance

Overall models that explicitly account for fire disturbances andtheir associated impact on carbon pools predict a greater meanannual GPP with a larger range in flux estimates than models with-out disturbance included (Fig 6) The impacts of fire on a givenecosystem depend on a number of factors including the ecosystemtype (eg ponderosa pine forest versus grasslands) fire intensityand type (ie stand replacing) and overall scale For example alarge stand-replacing fire would likely result in suppressed pro-ductivity (and GPP) for several years following the fire Converselygiven the right conditions a fire event could make more nitrogenavailable for growth (and thereby increase production of leaf tis-sue) andor for photosynthesis (through higher leaf tissue N in theform of Rubisco) This however is balanced by any losses in leafarea during the fire Many of the models that directly account forfire also employ an enzyme kinetic approach in their formulation ofphotosynthesis Although how a model accounts for disturbances(including fire) impacts their estimates of carbon pools and stocksit is not likely the dominant driver for the differences in GPP seenamong the participating models in this study

There are limited datasets with which to compare modeled GPPAlthough MODIS-derived estimates of GPP (Heinsch et al 2006Running et al 2004 Zhao et al 2005) have been favorably com-pared to flux tower measurements tower-by-tower comparisonsstill show significant residuals MODIS GPP is fundamentally a mod-eled product not a direct observation The MODIS product and

other LUE-based models are similar in their estimates of net uptakeand generally predict lower productivity than models in which pho-tosynthesis is more physiologically based (Figs 4 and 6 and Table 3)For example when totalled over the growing season and annually

154 DN Huntzinger et al Ecological Modelling 232 (2012) 144ndash 157

F hic rep er fireb

mtmtftieabic

anb(ic

4

mpv1tbd

ig 6 Model estimates of (A) gross primary productivity (GPP) and (B) heterotrophotosynthetic formulation (enzyme kinetic versus light-use efficiency) and whethy the models See Tables 2 and 3 for more information

any of the prognostic models in this study estimate 12ndash2 timeshe GPP predicted by the diagnostic or light-use efficiency based

odels Razcka and Davis (personal communications) comparedhe TBM estimates in this study to flux tower measurements Theyound that the mean GPP and ecosystem respiration (Ra + Rh) fromhe models is about 30ndash40 greater in most biomes (not includ-ng deciduous broadleaf forests) compared to those derived fromddy-covariance (EC) measurements at flux tower sites As a resultlthough similarities exist between the lower end of the model-ased GPP estimates and those derived from EC measurements

t is difficult to say whether these lower GPP estimates are moreorrect

In addition to the influence of environmental drivers discussedbove whether a model accounts for time-varying CO2 andoritrogen deposition could contribute to the differences in net car-on uptake simulated by the models (Fig 6) Friedlingstein et al2006) for example showed greater carbon uptake by ecosystemsn uncoupled TBMs as a result of increased atmospheric CO2 con-entration

22 Variability in heterotrophic respirationHeterotrophic respiration is also difficult to model at a funda-

ental scale due to its dependence on poorly understood complexrocesses as well as the need to track diverse carbon pools ofarying decomposability (Jastrow 1996 Oades 1988 Parton et al

987) While the overall magnitude in Rh is smaller than that of GPPhe variation among models is still large with estimates differingy 50ndash600 gC mminus2 yrminus1 (Fig 5) Models that estimate soil carbonecomposition based on zero-order kinetics (ie decomposition

spiration (Rh) for Temperate North America grouped by decomposition kinetics disturbance land-coverland-use changes and transient forcings were considered

rate independent of concentration) do not explicitly calculate Rhand they are therefore not included in Fig 6 Estimates of Rhfrom models with both first-order soil carbon decomposition rateswhich also include nitrogen cycling tend to exhibit a shallowerseasonal cycle and less overall soil C release than models withoutN cycling Nitrogen limitations on microbial decomposition couldresult in slower decomposition rates (Thornton et al 2007 Yanget al 2009) However this in turn would reduce the rate of Navailability for plant growth The models that consider nitrogendeposition (in addition to CO2) do not have lower GPP and may havea slightly larger GPP than the models that do not include N depo-sition (Fig 6A) This added N from atmospheric deposition may atleast for North America be enough to compensate for the reductionin N from decomposition thus supplying the N required for GPP

Overall the differences in modeled GPP and Rh do not translateinto large differences in the long-term mean seasonal cycle of NEP(Fig 3) in part because within many models respiration is highlycorrelated to GPP This is also observed in other studies (eg Poulteret al 2011) where modeled Rh tends to respond proportionally tochanges in GPP or productivity resulting in a smaller net range inabsolute NEP among the models (Table 4)

5 Conclusions

This study brings together estimates of land-atmosphere carbon

exchange from nineteen prognostic and diagnostic TBMs in orderto assess the current understanding of the terrestrial carbon cyclein North America The models differ substantially in their estimatesof net ecosystem productivity as well as gross primary productivity

cal Mo

atA

mitMabsRappi

oatadtmttairtwda

midmfpiistacdfDpaqmmmtosoatesmce

DN Huntzinger et al Ecologi

nd respiration Prognostic models exhibit greater overall range inheir estimates and predict larger net uptake of carbon over Northmerica relative to diagnostic models

Photosynthetic formulation the source and variability of cli-atic driver data and how phenology is described all appear to

nfluence the across-model difference in estimated fluxes andhe magnitude of overall carbon uptake predicted by the models

uch of the variability in modeled Rh is likely driven by vari-bility in GPP because the majority of models scale respirationased on their estimates of photosynthesis While this type ofcaling may be appropriate for forested regions where GPP andh are closely linked this assumption is probably not appropri-te for more managed lands (eg agricultural lands and forestlantations in the US Southeast) where harvest lateral trans-ort and other management activities can impact where carbon

s respiredFor many biome types (eg evergreen and needleleaf decidu-

us and herbaceous shrublands) there is a large range in both GPPnd Rh but a relatively small range in model-estimated NEP Thisrend in simulation results is consistent with the work of Raczkand Davis (2011 personal communication) which compares modelerived estimates of GPP and respiration to those inferred from fluxower observations Thus models that overestimate (or underesti-

ate) GPP and Rh can still predict plausible values for NEP but forhe wrong reasons For example models that are calibrated to fluxower observations may be ldquotunedrdquo to NEP particularly when GPPnd Rh observations are scarce The flux tower records can help tonterpret the cause of model difference and suggest that the lowerange of GPP in this collection of TBM models may be closer toower-based observations What we cannot tell from comparisonsith observations however whether the model estimates repro-uce observations for the right reasons (ie whether processesccurately are represented in the model)

Overall flux estimates are a function not only of model algorith-ic formulation but also how models were calibrated (or tuned)

nitial conditions (eg soil properties vegetation and land-use)river data (eg weather CO2 concentration) and their treat-ent of disturbances (eg fire wind disease) The entire modeling

ramework contributes to the results and therefore all of the com-onents require evaluation The study reveals the large variation

n TBM estimates of long-term mean net ecosystem productiv-ty as well as discrepancies in the magnitude and timing of theeasonal cycle The results also provide a sobering picture ofhe current lack of consensus among model estimates of land-tmosphere carbon exchange across North America Attributing theross-model variability to differences in modeling approaches andriving data is difficult however given the focus on existing resultsrom models run using a wide range of assumptions and inputseveloping improving and evaluating TBMs such that they canrovide useable forecasts (and past diagnoses) at near-term inter-nnual decadal and century timescales requires developments inuantitative model evaluation and rigorous benchmark develop-ent While we were able to attribute some of this variation toodel structure and aspects of model driver data a more formalodel-data comparison is required to more definitively quantify

he impact of model formulation and supporting and driver datan the accuracy of the simulation outputs Such efforts requireubstantial technical support for model participation the devel-pment of consistent and optimal environmental driver datasets

unified intercomparison protocol as well as coordination ofhe intercomparison effort across research groups These types offforts are underway including several projects working to under-

tand how model formulation and model choices impact overallodel performance through the use of detailed simulation proto-

ol and controlled input environmental driver data (eg Schwalmt al 2010) and the Multi-Scale Synthesis and Terrestrial Model

delling 232 (2012) 144ndash 157 155

Intercomparison Project (MsTMIP) which directly builds of theNACP regional interim synthesis present here

Acknowledgements

The interim-synthesis activity represents a grass-roots effortby the carbon cycle community conducted largely on a volun-teer basis We would particularly like to thank all of the modelingteams that participated in the synthesis activities sharing resultsfrom their ongoing work and providing feedback during the work-shops We also thank MAST-DC at Oak Ridge National Laboratoryfor data management support MAST-DC (Project NNH06AE47I)is a Carbon Cycle Interagency Working Group Project funded byNASArsquos Terrestrial Ecology Program Funding was also provided bythe National Aeronautics and Space Administration (NASA) underGrant No NNX06AE84G ldquoConstraining North American Fluxes of Car-bon Dioxide and Inferring their Spatiotemporal Covariances throughAssimilation of Remote Sensing and Atmospheric Data in a Geosta-tistical Frameworkrdquo issued through the ROSES A6 North AmericanCarbon Program

Appendix A Supplementary data

Supplementary data associated with this article can be found inthe online version at doi101016jecolmodel201202004

References

Bachelet D Lenihan JM Daly C Neilson RP 2000 Interactions between firegrazing and climate change at Wind Cave National Park SD Ecological Mod-elling 134 (2ndash3) 229ndash244

Baker IT Prihodko L Denning AS Goulden M Miller S da Rocha HA 2008Seasonal drought stress in the Amazon reconciling models and observationsJournal of Geophysical Research-Biogeosciences 113

Baldocchi DD Amthor JS 2001 Canopy Photosynthesis History Measurementsand Models Terrestrial Global Productivity Academic Press San Diego ISBN978-0-12-505290-0 doi101016B978-012505290-050003-X pp 9ndash31

Batjes N 1996 Total carbon and nitrogen in the soils of the world European Journalof Soil Science 47 doi101111j1365-23891996tb01386x

Beer C Reichstein M Tomelleri E Ciais P Jung M Carvalhais N et al 2010Terrestrial gross cabon dioxide uptake global distribution and covariation withclimate Science 329 (5993) 834ndash838 doi101126science1184984

Bond-Lamberty B Peckham SD Ahl DE Gower ST 2007 Fire as the dominantdriver of central Canadian boreal forest carbon balance Nature 450 89ndash92

Bondeau A SmithPC Zaehle S Schaphoff S Lucht W Cramer W Gerten D2007 Modelling the role of agriculture for the 20th century global terrestrialcarbon balance Global Change Biology 13 (3) 679ndash706

Carbon Cycle Science Program (CCSP) 2007 In King AW Dilling L ZimmermanGP Fairman DM Houghton RA Marland G et al (Eds) The First State of theCarbon Cycle Report (SOCCR) The North American Carbon Budget and Implica-tions for the Global Carbon Cycle A Report by the US Climate Change ScienceProgram and the Subcommittee on Global Change Research National Oceanicand Atmospheric Administration National Climatic Data Center Asheville NCUSA p 242

Chapin FS Woodwell GM Randerson JT Rastetter EB Lovett GM BaldocchiDD et al 2006 Reconciling carbon-cycle concepts terminology and methodsEcosystems 9 1041ndash1050 doi101007s10021-005-0105-7

Chen JM Liu J Cihlar J Goulden ML 1999 Daily canopy photosynthesis modelthrough temporal and spatial scaling for remote sensing applications EcologicalModelling 124 (2ndash3) 99ndash119

Collatz GJ Ribas-Carbo M Berry JA 1992 Coupled photosynthesis-stomatal con-ductance model for leaves of c4 plants Australian Journal of Plant Physiology19 (5) 519ndash538

Cramer W Kicklighter DW Bondeau A Moore B Churkina C Nemry B et al1999 Comparing global models of terrestrial net primary productivity (NPP)overview and key results Global Change Biology 5 1ndash15

Dai YJ Dickinson RE Wang YP 2004 A two-big-leaf model for canopy tem-perature photosynthesis and stomatal conductance Journal of Climate 17 (12)2281ndash2299

Daly C Bachelet D Lenihan JM Neilson RP Parton W Ojima D 2000 Dynamicsimulation of treendashgrass interactions for global change studies Ecological Appli-cations 10 (2) 449ndash469

Davis KJ 2008 Integrating field measurements with flux tower and remote sens-ing data In Hoover CM (Ed) Field Measurements For Landscape-Scale ForestCarbon Monitoring XVIII ISBN 978-1-4020-8505-5 p 242

DeFries R S Townshend JRG 1994 1 Degree Global Land Cover DataSet Derived from AVHRR Available on-line [httpglcfumiacsumdedu

1 cal Mo

D

F

F

F

F

F

G

G

G

G

G

H

H

H

H

H

H

H

J

J

J

J

J

K

K

K

K

56 DN Huntzinger et al Ecologi

datalandcoverindexshtml] from the Global Land Cover Facility University ofMaryland Institute for Advanced Computer Studies College Park MarylandUSA

enning AS et al 2005 Science implementation strategy for the North Ameri-can Carbon Program In Report of the NACP Implementation Strategy Groupof the US Carbon Cycle Interagency Working Group US Carbon Cycle ScienceProgram Washington DC p 68

AO 19952003 The Digitized Soil Map of the World and Derived Soil Properties(Version 35) FAO Land and Water Digital Media Series 1 FAO Rome

arquhar GD von Caemmerer S 1982 Modeling of photosynthetic response toenvironmental conditions In Lange OL Nobel PS Osmond CB Zeigler H(Eds) Physiological Plant Ecology II Water Relations and Carbon AssimilationSpringer-Verlag New York

arquhar GD Caemmerer SV Berry JA 1980 A biochemical-model of photo-synthetic CO2 assimilation in leaves of C-3 species Planta 149 (1) 78ndash90

oley JA Prentice IC Ramankutty N Levis S Pollard D Sitch S HaxeltineA 1996 An integrated biosphere model of land surface processes terrestrialcarbon balance and vegetation dynamics Global Biogeochemical Cycles 10 (4)603ndash628

riedlingstein P et al 2006 Climate-carbon cycle feedback analysis results fromthe (CMIP)-M-4 model intercomparison Journal of Climate 19 (14) 3337ndash3353

lobal Change Project (GCP) 2010 REgional Carbon Cycle Assessment and Processes(RECCAP) Soft Protocol Version 4 Global Carbon Project

lobal Soil Data Task Group 2000 Global Gridded Surfaces of Selected Soil Char-acteristics (IGBP-DIS) Global Gridded Surfaces of Selected Soil Characteristics(International Geosphere-Biosphere Programme-Data and Information System)Data Set Oak Ridge National Laboratory Distributed Active Archive CenterOak Ridge Tennessee USA doi103334ORNLDAAC569 Available on-linehttpwwwdaacornlgov

oodale CL et al 2002 Forest carbon sinks in the Northern Hemisphere EcologicalApplications 12 (3) 891ndash899

urney KR Law RM Denning AS Rayner PJ Baker D Bousquet P et al 2002Towards robust regional estimates of CO2 sources and sinks using atmospherictransport models Nature 415 (6872) 626ndash630

urney KR Law RM Denning AS Rayner PJ Baker D Bousquet P et al 2003Transcom 3 CO2 inversion intercomparison 1 Annual mean control results andsensitivity to transport and prior flux information Tellus 55B 555ndash579

ansen J Ruedy R Glascoe J Sato M 1999 GISS analysis of surfacetemperature change Journal of Geophysical Research 104 30997ndash31022doi1010291999JD900835

axeltine A Prentice IC 1996 BIOME3 An equilibrium terrestrial biospheremodel based on ecophysiological constraints resource availability and com-petition among plant functional types Global Biogeochemical Cycles 10 (4)693ndash709

ayes DJ McGuire AD Kicklighter DW Gurney KR Burnside TJ MelilloJM 2011 Is the northern high latitude land-based CO2 sink weakening GlobalBiogeochemical Cycles 25 (3) GB3018 doi1010292010gb003813

ayes DJ Turner DP Stinson G McGuire AD Wei Y West TO et al2012 Reconciling estimates of the contemporary North American carbon bal-ance among inventory-based approaches terrestrial biosphere models andatmospheric inversions Global Change Biology 18 (3) doi101111j1365-2486201102627x

einsch FA et al 2003 Userrsquos Guide GPP and NPP (MOD17A2A3) Products NASAMODIS Land Algorithm

einsch FA et al 2006 Evaluation of remote sensing based terrestrial produc-tivity from MODIS using regional tower eddy flux network observations IEEETransactions on Geoscience and Remote Sensing 44 (7) 1908ndash1925

oughton RA Hackler JL Lawrence KT 1999 The US carbon budget contrib-utions from land-use change Science 285 (5427) 574ndash578

ain AK Yang JK 2005 Modeling the effects of two different land cover changedata sets on the carbon stocks of plants and soils in concert with CO2 and climatechange Global Biogeochemical Cycles 19 (2)

astrow JD 1996 Soil aggregate formation and the accrual of particulate andmineral-associated organic matter Soil Biology amp Biochemistry 28 (4ndash5)665ndash676

entsch A Kreyling J Beierkuhnlein C 2007 A new generation of climate-changeexperiments events not trends Frontiers in Ecology and the Environment 5 (7)365ndash374

u WM Chen JM Black TA Barr AG Liu J Chen BZ 2006 Modelling multi-year coupled carbon and water fluxes in a boreal aspen forest Agricultural andForest Meteorology 140 (1ndash4) 136ndash151

ung M Henkel K Herold M Churkina G 2006 Exploiting synergies of globalland cover products for carbon cycle modeling Remote Sensing of Environment101 534ndash553

icklighter DW Bondeau A Schloss AL Kaduk J McGuire AD et al 1999Comparing global models of terrestrial net primary productivity (NPP) globalpattern and differentiation by major biomes Global Change Biology 5 16ndash24

rinner G Viovy N de Noblet-Ducoudre N Ogee J Polcher J FriedlingsteinP et al 2005 A dynamic global vegetation model for studies of the coupledatmospherendashbiosphere system Global Biogeochemical Cycles 19 (1)

ucharik CJ Foley JA Delire C Fisher VA Coe MT Lenters JD et al 2000

Testing the performance of a Dynamic Global Ecosystem Model water balancecarbon balance and vegetation structure Global Biogeochemical Cycles 14 (3)795ndash825

urz WA Stinson G Rampley G 2007 Could increased boreal forest ecosystemproductivity offset carbon losses from increased disturbances Philosophical

delling 232 (2012) 144ndash 157

Transactions of the Royal Society of London Series B Biological Sciencesdoi101098rstb20072198

Latifovic R Zhu ZL Cihlar J Giri C Olthof I 2004 Land cover mapping of northand central America ndash Global Land Cover 2000 Remote Sensing of Environment89 (1) 116ndash127

Leemans R Cramer W 1991 The IIASA Database for Mean Monthly Values of Tem-perature Precipitation and Cloudiness of a Global Terrestrial Grid InternationalInstitute for Applied Systems Analysis (IIASA) RR-91-18

Lenihan JM Bachelet D Neilson RP Drapek R 2008 Simulated response ofconterminous United States ecosystems to climate change at different levels offire suppression CO2 emission rate and growth response to CO2 Global andPlanetary Change 64 (1ndash2) 16ndash25

Loveland TR Belward AS 1997 The International Geosphere Biosphere Pro-gramme Data and Information System global land cover data set (DISCover)Acta Astronautica 41 (4ndash10) 681ndash689

Loveland TR Reed BC Brown JF Ohlen DO Zhu J Yang L Merchant JW2000 Development of a Global Land Cover Characteristics Database and IGBPDISCover from 1-km AVHRR Data International Journal of Remote Sensing 21(67) 1303ndash1330

McGuire AD Hayes DJ Kicklighter DW Manizza M Zhuang Q Chen M et al2010 An analysis of the carbon balance of the Arctic Basin from 1997 to 2006Tellus 62B 455ndash474 doi101111j1600-0889201000497x

Medvigy DS Wofsy C Munger JW Moorcroft PR 2010 Responses of terrestrialecosystems and carbon budgets to current and future environmental variabilityProceedings of the National Academy of Science of the United States of America201 (18) 8275ndash8280 doi101073pnas0912032107

Melillo JM McGuire AD Kicklighter DW Moore Vorosmarty CJ Schloss AL1993 Global climate-change and terrestrial net primary production Nature 363(6426) 234ndash240

Melillo JM Borchers J Chaney J Fisher H Fox S Haxeltine A et al 1995 Vege-tation ecosystem modeling and analysis project ndash comparing biogeography andbiogeochemistry models in a continental-scale study of terrestrial ecosystemresponses to climate-change and CO2 doubling Global Biogeochemical Cycles 9(4) 407ndash437

NRCan and USGS 2003 Land Cover Database of North America 2000Oades JM 1988 The retention of organic-matter in soils Biogeochemistry 5 (1)

35ndash70Pacala SW et al 2001 Consistent land- and atmosphere-based US carbon sink

estimates Science 292 (5525) 2316ndash2320Pan YD Melillo JM McGuire AD Kicklighter DW Pitelka LF Hibbard K

Pierce et al 1998 Modeled responses of terrestrial ecosystems to elevatedatmospheric CO2 a comparison of simulations by the biogeochemistry modelsof the vegetationecosystem modeling and analysis project (VEMAP) Oecologia114 389ndash404

Parton WJ Schimel DS Cole CV Ojima DS 1987 Analysis of factors control-ling soil organic-matter levels in great-plains grasslands Soil Science Society ofAmerica Journal 51 (5) 1173ndash1179

Potter C Klooster S Huete A Genovese V 2007 Terrestrial carbon sinks forthe United States predicted from MODIS satellite data and ecosystem modelingEarth Interactions 11

Poulter BD Frank C Hodson EL Zimmerman NE 2011 Impacts of land coverand climate data selection on understanding terrestiral carbon dynamics and theCO2 ariborne fraction Biogeosciences Discuss 8 1617ndash1642 doi105194bgd-9-1617-2011

Prentice IC 2001 The Carbon Cycle and Atmospheric Carbon Dioxide Rep 3rdAssessment Intergovernmental Panel on Climate Change

Randerson JT Thompson MV Conway TJ Fung IY Field CB 1997 The con-tribution of terrestrial sources and sinks to trends in the seasonal cycle ofatmospheric carbon dioxide Global Biogeochemical Cycles 11 (4) 535ndash560

Randerson JT Hoffman FM Thornton PE Mahowald NM Lindsay KLee YH et al 2009 Systematic assessment of terrestrial biogeochemistryin coupled climate-carbon models Global Change Biology 15 2462ndash2484doi101111j1365-2486200901912x

Reichstein M Falge E Baldocchi D Papale D Aubinet M Bebigier P et al2005 On the separation of net ecosystem exchange into assimilation and ecosys-tem respiration review and improved algorithm Global Change Biology 111424ndash1439

Reichstein M Beer C 2008 Soil respiration across scales the importance ofa modelndashdata integration framework for data interpretation Journal of PlantNutrition and Soil Science 171 2344ndash2354

Running SW Nemani RR Heinsch FA Zhao MS Reeves M Hashimoto H2004 A continuous satellite-derived measure of global terrestrial primary pro-duction Bioscience 54 (6) 547ndash560

Schimel DS Enting IG Heimann M WIgley Raynaud D Alves D SiegenthalerU 2000 CO2 and the carbon cycle In Wigley TML (Ed) The Carbon CycleCambridge University Press

Schwalm CR Williams CA Schaefer K Anderson R Arain MA Baker I et al2010 A model-data intercomparison of CO2 exchange across North Americaresults from the North American Carbon Program site synthesis Journal of Geo-physical Research-Biogeosciences 115 doi1010292009JG001229

Sitch S et al 2003 Evaluation of ecosystem dynamics plant geography and terres-

trial carbon cycling in the LPJ dynamic global vegetation model Global ChangeBiology 9 (2) 161ndash185

Soil Survey Staff Natural Resources Conservation Service United States Departmentof Agriculture US General Soil Map (STATSGO2) for (State) Available online athttpsoildatamartnrcsusdagov accessed

cal Mo

T

T

T

v

v

V

W

W

W

DN Huntzinger et al Ecologi

hornton PE Lamarque JF Rosenbloom NA Mahowald NM 2007 Influenceof carbonndashnitrogen cycle coupling on land model response to CO2 fertil-ization and climate variability Global Biogeochemical Cycles 21 GB4018doi1010292006GB002868

hornton PE Doney SC Lindsay K Moore JK Mahowald N RandersonJT et al 2009 Carbonndashnitrogen interactions regulate climate-carbon cyclefeedbacks results from an atmospherendashocean general circulation model Bio-geosciences 6 (10) 2099ndash2120

ian HQ Chen G Liu M Zhang C Sun G Lu C et al 2010 Model estimatesof ecosystem net primary productivity evapotranspiration and water use effi-ciency in the Southern United States during 1895ndash2007 Forest Ecology andManagement 259 1311ndash1327

an der Werf GR Randerson JT Collatz GJ Giglio L Kasibhatla PS ArellanoAF et al 2004 Continental-scale partitioning of fire emissions during the 1997to 2001 El NinoLa Nina period Science 303 (5654) 73ndash76

an der Werf GR Randerson JT Giglio L Collatz GJ Kasibhatla PS ArellanoAF 2006 Interannual variability in global biomass burning emissions from1997 to 2004 Atmospheric Chemistry and Physics 6 3423ndash3441

iovy N Francois C Bondeau A Krinner G Polcher J Kergoat L et al 2000Assimilation of Remote Sensing Measurements into the ORCHIDEESTOMATEDGVM Biosphere Model

ang Z Grant RF Arain MA Chen BN Coops N Hember R Kurz WA PriceDT Stinson G Trofymow JA Yeluripati J Chen Z 2011 Evaluating weathereffects on interannual variation in net ecosystem productivity of a coastal tem-perate forest landscape a model intercomparison Ecological Modelling 2223236ndash3249

aring RH Running SW 2007 Forest Ecosystems Analysis at Multiple Scales

third ed Elsevier Academic Press Burlington MA

ofsy SC Harriss RC 2002 The North American Carbon Program (NACP)Report of the NACP Committee of the US Interagency Carbon Cycle Sci-ence Program Rep US Global Change Research Program Washington DC56 pp

delling 232 (2012) 144ndash 157 157

Xiao JF et al 2008 Estimation of net ecosystem carbon exchange for the contermi-nous United States by combining MODIS and AmeriFlux data Agricultural andForest Meteorology 148 (11) 1827ndash1847

Xiao JF et al 2010 A continuous measure of gross primary productivity for theconterminous US derived from MODIS and AmeriFlux data Remote Sensing ofEnvironment 114 576ndash591 doi101016jrse200910013

Xiao JF et al 2011 Assessing net ecosystem carbon exchange of US ter-restrial ecosystems by integrating eddy covariance flux measurements andsatellite observations Agricultural and Forest Meteorology 151 60ndash69doi101016jagrformet201009002

Yang XJ Wittig V Jain AK Post W 2009 Integration of nitrogen cycle dynam-ics into the Integrated Science Assessment Model for the study of terrestrialecosystem responses to global change Global Biogeochemical Cycles 23

Yuan WP et al 2007 Deriving a light use efficiency model from eddy covari-ance flux data for predicting daily gross primary production across biomesAgricultural and Forest Meteorology 143 (3ndash4) 189ndash207

Zeng N 2003 Glacialndashinterglacial atmospheric CO2 change ndash the glacial burialhypothesis Advances in Atmospheric Sciences 20 (5) 677ndash693

Zeng N Qian HF Munoz E Iacono R 2004 How strong is carbon cycle-climatefeedback under global warming Geophysical Research Letters 31 (20)

Zeng N Mariotti A Wetzel P 2005 Terrestrial mechanisms of interannual CO2

variability Global Biogeochemical Cycles 19 (1)Zhao MS Heinsch FA Nemani RR Running SW 2005 Improvements of the

MODIS terrestrial gross and net primary production global data set RemoteSensing of Environment 95 (2) 164ndash176

Zhao Y Ciais P Peylin P Viovy N Longdoz B Bonnefond JM et al2011 How errors on meteorological variables impact simulated ecosystem

fluxes a case study for six French sites Biogeosciences Discuss 8 2522ndash5467doi105194bgd-8-2467-2011

Zobler L 1986 A World Soil File for Global Climate Modeling Technical Mem-orandum 87802 NASA Goddard Institute for Space Studies (GISS) New YorkNY