Validation of terrestrial biogeochemistry in CMIP6 Earth

Geosci Model Dev 14 5863ndash5889 2021httpsdoiorg105194gmd-14-5863-2021copy Author(s) 2021 This work is distributed underthe Creative Commons Attribution 40 License

Validation of terrestrial biogeochemistry in CMIP6 Earth systemmodels a reviewLynsay Spafford12 and Andrew H MacDougall11Climate and Environment Saint Francis Xavier University Antigonish Canada2Environmental Sciences Memorial University St Johnrsquos Canada

Correspondence Lynsay Spafford (lspafforstfxca)

Received 12 May 2021 ndash Discussion started 25 May 2021Revised 1 September 2021 ndash Accepted 2 September 2021 ndash Published 27 September 2021

Abstract The vital role of terrestrial biogeochemical cy-cles in influencing global climate change is explored bymodelling groups internationally through land surface mod-els (LSMs) coupled to atmospheric and oceanic componentswithin Earth system models (ESMs) The sixth phase ofthe Coupled Model Intercomparison Project (CMIP6) pro-vided an opportunity to compare ESM output by provid-ing common forcings and experimental protocols Despitethese common experimental protocols a variety of terrestrialbiogeochemical cycle validation approaches were adoptedby CMIP6 participants leading to ambiguous model perfor-mance assessment and uncertainty attribution across ESMsIn this review we summarize current methods of terrestrialbiogeochemical cycle validation utilized by CMIP6 partici-pants and concurrent community model comparison studiesWe focus on variables including the dimensions of evalua-tions observation-based reference datasets and metrics ofmodel performance To ensure objective and thorough val-idations for the seventh phase of CMIP (CMIP7) we rec-ommend the use of a standard validation protocol employinga broad suite of certainty-weighted observation-based refer-ence datasets targeted model performance metrics and com-parisons across a range of spatiotemporal scales

1 Introduction

The terrestrial biosphere is presently responsible for seques-tering about a quarter of anthropogenic carbon emissionssubstantially reducing the severity of ongoing climate change(Friedlingstein et al 2020) The future capacity of the ter-restrial biosphere to sequester CO2 emissions is uncertain

due to non-linear feedbacks such as CO2 fertilization grow-ing season extension in cold-limited regions enhanced het-erotrophic respiration and potentially other feedbacks aswell as environmental and physiological constraints such asmoisture availability nutrient limitations and stomatal clo-sure (Fleischer et al 2019 Green et al 2019 Xu et al2016 Wieder et al 2015) Earth system models (ESMs) area means to simulate past present and future terrestrial bio-geochemical cycles examine the influence of changes in cli-mate and atmospheric CO2 concentration on CO2 uptake ex-plore feedbacks and limitations and estimate anthropogeniccarbon emissions compatible with avoiding a given thresh-old in global temperature change ESMs simulate global ex-changes of matter and energy through the coupling of landatmospheric and oceanic components Through concertedefforts successive generations of ESMs have improved interms of spatiotemporal resolution complexity and processrepresentation (Anderson et al 2016) Despite this progressterrestrial biogeochemical cycles remain a major source ofuncertainty in future climate projections (Arora et al 2020Lovenduski and Bonan 2017) This uncertainty stems fromlimited process understanding lacking observational con-straints inherent cycle variability temporal discrepancy be-tween forcings and responses (Sellar et al 2019 Ciais et al2013) and uncertain stock quantifications (Ito et al 2020Wieder et al 2015) which together compound uncertaintywithin models Among models this uncertainty is ampli-fied by artefacts in the form of inconsistent model struc-ture boundary conditions forcing datasets experimentalprotocols and benchmarking observational datasets which ismagnified by the increasing number diversity and complex-ity of ESMs (Eyring et al 2020) Subsequently a study on

Published by Copernicus Publications on behalf of the European Geosciences Union

5864 L Spafford and A H MacDougall Validation of terrestrial biogeochemistry in CMIP6

uncertainty in projected terrestrial carbon uptake based upon12 Coupled Model Intercomparison Project phase 5 (CMIP5)ESMs indicated that uncertainty stemming from model struc-ture may be 4 times greater than uncertainty from differentemission scenarios and internal variability (Lovenduski andBonan 2017) Some progress has been made in addressingthe large uncertainty associated with the terrestrial biogeo-chemistry in ESMs as comparison of the carbonndashclimate andcarbonndashconcentration feedback among ESMs participatingin the sixth phase of CMIP (CMIP6) by Arora et al (2020)shows a reduced model spread amongst models that includeda nitrogen cycle which provided a realistic constraint onphotosynthesis in the context of elevated atmospheric CO2concentration However the spread in estimated feedback pa-rameters across ESMs overall has not been significantly re-duced from CMIP6 relative to CMIP5 (Arora et al 20202013)

To answer scientific questions regarding climate changethe CMIP was initiated in 1995 by the World Climate Re-search Programmersquos (WCRP) Working Group of CoupledModelling (WCRP 2020) The CMIP designates standardexperimental protocols model output formats and modelforcings to diagnose climate change variability predictabil-ity and uncertainty following various scenarios within amulti-model framework CMIP6 began in 2013 with 3 yearsof planning and community consultation to address knowl-edge gaps prior to the conduction of simulations and anal-yses in 2016 and onwards Model validation in the contextof CMIP consists of demonstrating sufficient agreement be-tween model output data and historical observation-basedreference data following model development and is a cru-cial process in model advancement Such comparison facili-tates model improvement by identifying model limitations inperformance or sources of modelndashdata uncertainty (Loven-duski and Bonan 2017) and informs the weighting of dif-ferent ESMs in influencing climate projections and policy(Eyring et al 2019) CMIP6 specified detailed experimentalprotocols for modelling group participants to facilitate ob-jective comparisons of the output of different models withcommon forcings (Eyring et al 2016a)

Here we focus on validations of the stocks and biologi-cal fluxes of fully coupled ESMs and associated land sur-face model (LSM) releases from 2017 onwards with explicitterrestrial biogeochemical cycle representation contributedby participating CMIP6 modelling groups (hereafter partici-pants Table 1 Arora et al 2020) Validations are analyzed interms of variables included spatiotemporal scales referencedatasets and metrics of performance Section 2 compares themethods of historical terrestrial biogeochemical cycle valida-tion used by participants Section 3 summarizes the methodsused in community analyses of CMIP5 era models and pro-vides a critique of these methods A future outlook is pre-sented in Sect 4

2 Participant methods of validating terrestrialbiogeochemical cycles

To participate in CMIP6 participants had to submit four Di-agnosis Validation and Characterization of Klima (DECK)experimental simulations which included a control simula-tion with prescribed idealized pre-industrial (1850) forcingfor at least 500 years to demonstrate stability in global cli-mate and biogeochemical exchanges Additionally partic-ipants had to conduct historical simulations from 1850 to2014 using designated CMIP6 forcings (available at httpsesgf-nodellnlgovsearchinput4MIPs last access 8 Febru-ary 2021) and initialization from the pre-industrial forc-ing control run (Eyring et al 2016a) Each modellinggroup demonstrated stability in the global carbon cycle withglobal net carbon exchange below the suggested limit of01 Pg C yrminus1 by Jones et al (2016) while no suitable pre-industrial simulation global nitrogen or phosphorus flux wasspecified for CMIP6 though these were generally below20 Pg yrminus1 (Ziehn et al 2020) Each modelling group val-idated terrestrial biogeochemical cycle components for thehistorical simulation in a unique fashion which is summa-rized below and detailed in Appendix A

21 Variables included in validations

The number of terrestrial biogeochemical cycle variablesevaluated against observation-based estimates by partici-pants varied considerably (from 0 to 21) with a total of38 unique variables evaluated by all participants combinedThe variable validated most often was gross primary produc-tion (GPP) which was validated by all but one participantThe next nine most validated variables in descending orderwere soil carbon the global land carbon sink leaf area in-dex (LAI) vegetation carbon ecosystem respiration globallandndashatmosphere CO2 flux surface CO2 concentrations to-tal biomass and burned area (Fig 1) For a list of variabledefinitions see Table 2

The majority of variables were validated by just one or twoparticipants (Fig 2) Danabasoglu et al (2020) and Lawrenceet al (2019) validated a relatively extensive suite of vari-ables with the International Land Model Benchmarking (IL-AMB) package version 21 (ILAMBv21 Collier et al 2018Fig 3) including an explicit uncertainty analysis of the influ-ences of interannual variability forcing datasets and modelstructure in the form of prescribed versus prognostic vegeta-tion phenology While no nitrogen cycle variable was val-idated by more than one group soil N2O flux and totalN2O emissions were evaluated by Hajima et al (2020) andLawrence et al (2019) respectively

A variety of spatiotemporal scales of these variables wereconsidered in validations both within and among partici-pants Spatial scales consisted of site-level model grid celldegree of latitude region and global scales with the lat-ter being the most common across participants Temporal

Geosci Model Dev 14 5863ndash5889 2021 httpsdoiorg105194gmd-14-5863-2021

L Spafford and A H MacDougall Validation of terrestrial biogeochemistry in CMIP6 5865

Table 1 Modelling group contributions to C4MIP of CMIP6 from Arora et al (2020)

Modelling ESM Land surface model Explicit Dynamic Prognostic Prognostic leaf Reference(s)group biogeochemistry component N cycle vegetation LAI phenology

CSIRO ACCESS-ESM15 CABLE24 Yes No Yes No Ziehn et al (2020)

BCC BCC-CSM2-MR BCC-AVIM2 No No Yes Yes (for Wu et al (2019)deciduous) Li et al (2019)

CCCma CanESM5 CLASS-CTEM No No Yes Yes Swart et al (2019)

CESM CESM2 CLM5 Yes No Yes Yes Danabasoglu etal (2020)Lawrence et al (2019)

CNRM CNRM-ESM2-1 ISBA-CTRIP No No Yes Yes (from leaf Seacutefeacuterian et al (2019)carbon balance) Delire et al (2020)

GFDL GFDL-ESM4 LM41 No Yes ndash ndash Dunne et al (2020)

IPSL IPSL-CM6A-LR ORCHIDEE version 20 No No Yes Yes Boucher et al (2020)Vuichard et al (2019)

JAMSTEC MIROC-ES2L VISIT-e Yes No Yes Yes Hajima et al (2020)

MPI MPI-ESM12-LR JSBACH32 Yes Yes Yes Yes Mauritsen et al (2019)Goll et al (2017)

NCC NorESM2-LM CLM5 Yes No Yes Yes Seland et al (2020)

UK UKESM1-0-LL JULES-ES-10 Yes Yes Yes Yes Sellar et al (2019)

Table 2 Terms associated with terrestrial biogeochemical cycles and their definitions as used by participants

Term CMIP6 definition

Gross primary production (GPP) The quantity of CO2 removed from the atmosphere byvegetation

Net primary productivity (NPP) The quantity of CO2 removed from the atmosphere byvegetation minus the quantity of CO2 from autotrophicrespiration

Autotrophic respiration (AR) The quantity of CO2 from cellular respiration in plants

Ecosystem respiration (ER) The quantity of CO2 from autotrophic respiration andheterotrophic respiration

Heterotrophic respiration (HR) The quantity of CO2 from cellular respiration by het-erotrophs

Net ecosystem production (NEP) The quantity of CO2 removed from the atmosphere byvegetation minus the quantity of CO2 from autotrophicand heterotrophic respiration

Net biome production (NBP) The net rate of organic carbon accumulation mi-nus autotrophic and heterotrophic respiration and non-respiratory losses from disturbance

Net ecosystem carbon balance (NECB) The net rate of organic carbon accumulation in anecosystem (independent of scale)

httpsdoiorg105194gmd-14-5863-2021 Geosci Model Dev 14 5863ndash5889 2021

Figure 1 Validation (green) or omission (grey) of the 10 mostfrequently validated variables by participants (treating ESMs andLSMs separately) gross primary productivity (GPP) soil carbon(SC) global land carbon sink (GLCS) leaf area index (LAI) veg-etation carbon (VC) ecosystem respiration (ER) landndashatmosphereCO2 flux (LACF) surface CO2 concentrations (Surf[CO2]) totalbiomass (TB) and burned area (BA)

Figure 2 Frequency of a given variable being validated across par-ticipants (treating ESMs and LSMs separately) Most variables werevalidated only once across participants (leftmost x axis) while GPPwas validated by 11 participants (rightmost bar)

scales included daily seasonal annual decadal select pe-riods and long-term trends accumulations or averages overthe whole historical simulation period from 1850 to 2014For more detail on the spatiotemporal scales of validationused by each participant readers should refer to Appendix A

Dynamic variables such as LAI were subject to a detailed as-sessment including annual maximum and minimum magni-tude (Seacutefeacuterian et al 2019) and month (Li et al 2019) sea-sonality (Ziehn et al 2020) and seasonal average as wellas global averages GPP was also evaluated across a vari-ety of scales including in terms of the daily seasonal andannual magnitude on a plant functional type (PFT) spatialand global basis against site-level observations (Vuichard etal 2019) as well as globally in terms of functional relation-ships with temperature and precipitation (Swart et al 2019)and the relative contribution of drivers of variation (Vuichardet al 2019) Biomass and carbon stock variables were eval-uated in terms of spatial distributions or global averages overthe chosen time periods often on a decadal scale (Li et al2019) Global vegetation and soil carbon turnover times werealso evaluated for selected time periods (Delire et al 2020Lawrence et al 2019)

22 Reference datasets

For variables which were validated by more than one mod-elling group such as GPP a variety of observation-basedreference datasets were utilized For example across par-ticipants several different GPP reference datasets were used(Table 3) though most participants utilized model tree en-semble (MTE) machine-learning upscaled ground eddy-covariance meteorological and satellite observation-basedestimates of GPP from Jung et al (2011) Interestingly onegroup the Centre National de Recherches Meacuteteacuteorologiques(CNRM Delire et al 2020) used a more recent Fluxnet-based GPP dataset (FluxComv1 Jung et al 2016 Tramon-tana et al 2016) and further used the mean of 12 prod-ucts therein CNRM and the Institut Pierre Simon Laplace(IPSL Vuichard et al 2019) were the only groups to in-clude a comparison to site-level GPP observations A vari-ety of reference datasets were also utilized for the secondmost frequently validated variable soil carbon (Table 4)spanning a 12-year publication range (Batjes 2016 GlobalSoil Data Task Group 2002) Several participants used morethan one reference dataset for evaluation of soil carbon de-pending upon regional or global focus such as the NorthernCircumpolar Soil Carbon Database provided by Hugelius etal (2013) for midndashhigh latitudes while global soil carbonestimates were obtained from Batjes (2016) Carvalhais etal (2014) Todd-Brown et al (2013) and FAO (2012) Whilebiomass and carbon stocks were predominantly compared topresent-day observations Delire et al (2020) used recordsfrom the Global Database of Litterfall Mass and Litter PoolCarbon and Nutrients database which extends from 1827 to1997 (Holland et al 2015)

23 Statistical metrics of model performance

A variety of statistical metrics were used to quantify modelperformance in simulating historical variables in comparison

Figure 3 Validation results for terrestrial variables within the CLM5 by Lawrence et al (2019) using ILAMB analysis (Collier et al2018) including three different climate forcing data products (individual columns) and two forms of model structure (column groups)CLM5SP denotes MODIS-prescribed (Zhao et al 2005) vegetation phenology while CLM5GBC denotes prognostic phenology Climateforcing data products include WATCHWFDEI from Mitchell and Jones (2005) CRUNCEPv7 the default forcing dataset used by the GlobalCarbon Project (Le Queacutereacute et al 2018) and GSWP3v1 the default forcing dataset used in the Land Surface Snow and Soil Moisture ModelIntercomparison Project (van den Hurk et al 2016) This figure was made available under a Creative Commons Attribution License (CCBY)

to observations though chosen metrics were more consistentthan selected variables The comparison of simulated andobservation-based averages calculated over space and timewas the most common metric used by all but two partici-pants (Table 5) The next most commonly used metric wasroot-mean-squared error (RMSE) followed by bias (simu-latedminus observed) on a spatial or global basis Evaluationsof global accumulations seasonal phase seasonal maximumandor minimum and global totals were also used The Tay-lor diagram which geometrically combines spatiotemporalcorrelation standard deviation and root mean square (rms)

difference (Taylor 2001) was used to summarize model per-formance by three participants (Li et al 2019 Collier et al2018 Goll et al 2017) The correlation coefficient (r) wasalso used by three participants (Swart et al 2019 Maurit-sen et al 2019 Goll et al 2017) RMSE normalized bythe standard deviation of observations (NRMSE) was onlyused by Swart et al (2019) while the coefficient of deter-mination (r2) was only used by Mauritsen et al (2019) Atargeted metric in the form of dissected mean squared devia-tion (Kobayashi and Salam 2000) the sum of squared biassquared difference between standard deviations and lack of

Table 3 The source for gross primary production (GPP) data referenced by each modelling group for ESM or LSM simulations Adjacentcontributions from the same modelling group are banded in a common fashion for readability LSM-focused validations by each modellinggroup are presented with the associated ESM in brackets

Model validation GPP reference data

ACCESS-ESM15 Jung et al (2011) Ziehn et al (2011) Beer et al (2010)BCC-CSM2-MR ndashBCC-AVIM20 (BCC-CSM2-MR) Jung et al (2011)CanESM5 Jung et al (2009)CESM2 Jung et al (2011)CLM5 (CESM2) Jung et al (2011)CNRM-ESM2-1 ndashISBA-CTRIP (CNRM-ESM2-1) Jung et al (2016) Tramontana et al (2016) Joetzjer et al (2015)IPSL-CM6A-LR ndashORCHIDEE (IPSL-CM6A-LR) Jung et al (2011)GFDL-ESM41 ndashMIROC-ES2L Jung et al (2011)MPI-ESM12-LR ndashJSBACH310 (MPI-ESM12-LR) ndashNORESM2 Jung et al (2011)UKESM1-0-LL Jung et al (2011)

Table 4 The source for soil carbon data referenced by each modelling group for ESM or LSM simulations Adjacent contributions from thesame modelling group are banded in a common fashion for readability LSM-focused validations by each modelling group are presented withthe associated ESM in brackets

Model validation Soil carbon reference data

ACCESS-ESM15 ndashBCC-CSM2-MR ndashBCC-AVIM20 (BCC-CSM2-MR) ndashCanESM5 ndashCESM2 Hugelius et al (2013) Todd-Brown et al (2013)CLM5 (CESM2) FAO (2012)CNRM-ESM2-1 ndashISBA-CTRIP (CNRM-ESM2-1) FAO (2012)IPSL-CM6A-LR ndashORCHIDEE (IPSL-CM6A-LR) ndashGFDL-ESM41 ndashMIROC-ES2L Batjes (2016) Hugelius et al (2013) Todd-Brown et al (2013)MPI-ESM12-LR Goll et al (2015)JSBACH310 (MPI-ESM12-LR) ndashNORESM2 FAO (2012)UKESM1-0-LL Batjes (2016) Carvalhais et al (2014) Global Soil Data Task Group (2002)

correlation weighted by standard deviation was used to dis-tinguish model sources of error by Vuichard et al (2019)In addition to quantitative metrics the qualitative aspects ofsimulations were compared to observational reference datasuch as in demonstrating source or sink behaviour over time(Danabasoglu et al 2020) or in visual comparison of spatialdistribution maps

3 Community methods of validating terrestrialbiogeochemical cycles

A variety of software and projects have been dedicated tothe communal evaluation of ESM (Gleckler et al 2016) andLSM performance (Kumar et al 2012 Gulden et al 2008)with CMIP6-era collaborative efforts including the EarthSystem Model Evaluation Tool version 2 (ESMValToolv20Eyring et al 2016b) and ILAMBv21 (Danabasoglu et al2020 Lawrence et al 2019 Collier et al 2018) Both ES-MValToolv20 and ILAMBv21 are openly available tools

Table 5 Model performance metrics used by each modelling group for ESM or LSM simulations Adjacent contributions from the samemodelling group are banded in a common fashion for readability LSM-focused validations by each modelling group are presented with theassociated ESM in brackets

Model validation Presented model performance assessment metrics

ACCESS-ESM15 Space-time averages seasonal amplitude timing and magnitude of annual maximums and min-imums

BCC-CSM2-MR ndash

BCC-AVIM20 (BCC-CSM2-MR) Average annual cycle phase global mean bias RMSE Taylor score

CanESM5 Space-time averages geographic distribution of time averages and bias latitudinal averagescorrelation coefficient (r) RMSE NRMSE ( RMSE

standard deviation of observations ) change in NRMSE

CESM2 Space-time averages seasonal cycles spatial distributions time series interannual variabilityglobal accumulations functional relationships relative bias RMSE ILAMB relative scale

CLM5 (CESM2) Space-time averages seasonal cycles annual monthly maximum spatial distributions globaltotals turnover time time series interannual variability functional relationships bias relativebias RMSE ILAMB relative scale

CNRM-ESM2-1 Average annual maximums and minimums spatial distribution bias RMSE model correlationbetween spatial pattern of error

ISBA-CTRIP (CNRM-ESM2-1) Geographic distribution of time averages and bias latitudinal averages global accumulationsbias spatial correlation turnover time average annual maximums seasonal cycle amplitudeand phase

IPSL-CM6A-LR Global annual averages and accumulations over time

ORCHIDEE (IPSL-CM6A-LR) Daily seasonal annual averages spatial distribution regional averages global averagesRMSE NRMSE dissected mean-squared deviation (squared bias squared difference betweenstandard deviations lack of correlation weighted by standard deviations from Kobayashi andSalam 2000) relative drivers of variation

GFDL-ESM41 Spatial distribution of seasonal amplitude interannual variability RMSE correlation coefficient(r) coefficient of determination (r2)

MIROC-ES2L Space-time averages latitudinal averages spatial distribution gradient seasonality densityglobal accumulations

MPI-ESM12-LR Spatial variability latitudinal average density global accumulations

JSBACH310 (MPI-ESM12-LR) Space-time averages spatial variability frequency distribution response ratio correlation coef-ficient (r) RMSE Taylor score

NORESM2 Global averages and totals

UKESM1-0-LL Space-time averages spatial distribution latitudinal averages global accumulations and totals

for the evaluation of a variety of model output against re-processed observations (httpswwwesmvaltoolorg httpspypiorgprojectILAMB last access 1 May 2021 Eyringet al 2020 2016b Collier et al 2018) The observation-based reference datasets for each are displayed in Table 6For the ESMValToolv20 dataset re-processing for compat-ible comparison in space and masking of missing obser-vations is detailed in Righi et al (2020) The analysis ofthe land carbon cycle in ESMValToolv20 (Eyring et al2020) is based upon the approach of Anav et al (2013)for considering long-term trends interannual variability andseasonal cycles A variety of tailored model performance

metrics are available with ESMValToolv20 (Eyring et al2020) The relative space-time root-mean-square deviation(RMSD) indicates model success relative to the multi-modelmedian in simulating the seasonal cycle of key variables thatwas originally detailed in Flato et al (2013) and allows si-multaneous comparison to more than one observational ref-erence for each simulated variable (where available) ES-MValTool20rsquos AutoAssess function provides a highly re-solved model performance evaluation for 300 individual vari-ables originally developed by the UK Met Office Furtherland cover can be comprehensively evaluated with ESM-ValToolv20 in terms of area mean fraction and bias on a

regional and global basis accommodating different modelrepresentations of land cover ILAMBv21 was used to vali-date terrestrial biogeochemical cycle components in CESM2(Danabasoglu et al 2020) and CLM5 (Fig 3 Lawrence etal 2019) ILAMBv21 was also used to demonstrate the ab-solute and relative performance of Dynamic Global Vegeta-tion Models (DGVMs) within several iterations of the GlobalCarbon Project (Friedlingstein et al 2020 2019 Le Queacutereacuteet al 2018) In addition to variables presented in Table 6functional relationships between these variables and temper-ature and precipitation are provided for validation purposesin ILAMBv21 ILAMBv21 employs a weighting systemto assign scores to observation-based datasets which en-compasses certainty measures spatiotemporal-scale appro-priateness and process implications In computing statisticalmodel performance scores ILAMBv21 acknowledges howreference observations represent discontinuous constants intime and space For example if a reference dataset containsaverage information across a span of years the annual cycleof such a dataset is assumed to be undefined and is there-fore not used as a reference The calculation of averages overtime in ILAMBv21 addresses spatiotemporally discontinu-ous data by performing calculations over specific intervalsfor which data are considered valid For each variable evalua-tion ILAMBv21 generates a series of graphical diagnosticsincluding spatial contour maps time series plots and Taylordiagrams (Taylor 2001) as well as statistical model perfor-mance scores including period mean bias RMSE spatialdistribution interannual coefficient of variation seasonal cy-cle and long-term trend These scores are then scaled basedupon the weighting of reference observation-based datasetsand for multi-model comparisons they are presented acrossmetrics and datasets to provide a single score

4 Critique of validation approaches

While standard protocols were used by participants for his-torical simulations in CMIP6 no standard protocol in termsof variables evaluated reference data performance metricsor acceptable performance threshold was adopted for terres-trial biogeochemical cycle validation The validation of par-ticular variables by different participants occasionally em-ployed the same datasets though in many cases inconsis-tent reference datasets were used for the same variable andthe spatial and temporal dimension of validations was oftendistinct This contrasts with other works employing multi-ple models such as the Global Carbon Project (Friedlingsteinet al 2020 2019 Le Queacutereacute et al 2018) which providesexplicit validation criteria such as simulating recent histori-cal net landndashatmosphere carbon flux within a particular rangeand within the 90 confidence interval of specified observa-tions The stringency of such criteria must be carefully cho-sen to acknowledge the role of observational uncertainty anduncertainty stemming from potential model tuning to forcing

datasets The use of different validation approaches impedesthe comparison of performance across models however italso provides a diverse collection of example methods

41 Variable choice

A comprehensive validation of a process-based model shouldinclude all simulated interacting variables for which a re-liable empirical reference is available Improvement in thesimulation of one variable through altered parameters struc-ture or algorithms may translate into degradation for othervariables which would be otherwise obscured in a restrictedvariable analysis (Deser et al 2020 Ziehn et al 2020Lawrence et al 2019) Given the scope of CMIP6 pub-lications in demonstrating model improvements relative toprevious versions and the results of CMIP6 experiments itis understandable that most participants validated a few se-lect variables and that more extensive validations may be inpreparation Essential climate variables (ECVs) prioritizedfor land evaluation in the ESMValToolv20 included GPPLAI and NBP (Eyring et al 2020 2016b) as these variablesintersect with other ESM components in matter and energyexchanges (Reichler and Kim 2008) In contrast LAI andNBP were not as frequently validated as GPP by CMIP6 par-ticipants (Fig 1) though the third most validated variablethe global land carbon sink is equivalent to NBP minus landuse emissions The most common variable chosen for val-idation by participants was GPP which is advantageous asit represents a crucial carbon cycle flux GPP designates thequantity of CO2 removed from the atmosphere and assimi-lated into structural and non-structural carbohydrates duringphotosynthesis by vegetation part of which is later respiredback to the atmosphere This quantity is limited by nutri-ent availability light soil moisture stomatal response to at-mospheric CO2 concentration and other environmental fac-tors (Davies-Barnard et al 2020) and is the largest carbonflux between the land biosphere and atmosphere (Xiao et al2019) Over- or under-estimations of GPP can lead to biasesin carbon stocks which are exacerbated through time (Car-valhais et al 2014)

An emergent ecosystem property that integrates a varietyof influential model processes is carbon turnover time calcu-lated as the ratio of a long-term average total carbon stockcompared to GPP or NPP (Eyring et al 2020 Yan et al2017 Carvalhais et al 2014) Carbon turnover times canbe the source of pervasive uncertainty within ESMs andtheir misrepresentation can lead to long-term drifts in car-bon stocks fluxes and feedbacks (Koven et al 2017) Theevaluation capacity of turnover times was seldom utilized byCMIP6 participants despite soil carbon being a relativelycommonly validated variable Many CMIP5 models werefound to under-estimate turnover times both globally and ona latitudinal basis (Eyring et al 2020 Fan et al 2020) whiletwo participants here Delire et al (2020) and Lawrence et

Table 6 Select observation-based reference dataset sources for ESMValToolv20 (Eyring et al 2020) and ILAMBv21 (Collier et al 2018)including net biome production (NBP) leaf area index (LAI) land cover (LC) gross primary production (GPP) net ecosystem exchange(NEE) soil carbon (SC) vegetation carbon (VC) ecosystem carbon turnover (ECT) vegetation biomass (VB) and burned area (BA) Notethat vegetation carbon is dependent upon vegetation biomass

Variables ESMValToolv20 ILAMBv21

NBP Le Queacutereacute et al (2018)Maki et al (2010)

Le Queacutereacute et al (2016)Hoffman et al (2014)

LAI Zhu et al (2013)Baret et al (2007)

De Kauwe et al (2011)Myneni et al (1997)

LC Defourny et al (2016) ndash

GPP NEE Jung et al (2019 2011) Lasslop et al (2010)Jung et al (2010)

SC Wieder (2014) Todd-Brown et al (2013)Hugelius et al (2013)

VC Gibbs 2006 ndash

ECT Carvalhais et al (2014) ndash

VB ndash Saatchi et al (2011)Kellndorfer et al (2013)Blackard et al (2008)

BA ndash Giglio et al (2010)

al (2019) reported over-estimated carbon turnover times de-spite demonstrating improvement from previous models

Another approach to validation that combines high-levelvariables and re-parameterization efforts is the assessment offunctional relationships or emergent constraints such as therelationship between GPP or turnover times and temperaturemoisture growing season length and nutrient stoichiometry(Danabasoglu et al 2020 Swart et al 2019 Anav et al2015 McGroddy et al 2004) Physically interpretable emer-gent constraints can aid in identifying model components thatare particularly influential for climate projections (Eyring etal 2019) such as the temperature control on carbon turnoverin the top metre of soil in cold climates (Koven et al 2017)GPP responses to soil moisture availability (Green et al2019) or regional carbonndashclimate feedbacks (Yoshikawa etal 2008) With the goal of realistically simulating Earthsystem processes to develop informed predictions of futureclimate large-scope variables that inherit uncertainty froman amalgamation of processes are often prioritized for val-idation Several participants focused on comparing simu-lated long-term trends or accumulations in global land car-bon fluxes to observation-based estimates from the GlobalCarbon Project (Friedlingstein et al 2019 Le Queacutereacute et al2018 2016) While this summation approach can signal alarge bias (Eyring et al 2020 2016b Reichler and Kim2008) and reduce the effect of sub-scale noise it does notidentify sources of model error or may even obscure modelerror For example if simulated landndashatmosphere carbon flux

from the pre-industrial era to the 2010s is found to concurwith observation-based estimates this could be due in partto compounding underlying biases which neutralize one an-other over time (Fisher et al 2019 Yoshikawa et al 2008)or alternatively suitable global averages may be susceptibleto antagonistic regional biases such as between the trop-ics and northern high latitudes Plant-functional-type-levelevaluations such as that of the maximum rate of RuBisCOcarboxylation and canopy height by Lawrence et al (2019)demonstrate the performance of underlying variables in in-fluencing large-scale carbon fluxes and stocks Several par-ticipants included latitudinal-scale evaluations (Delire et al2020 Hajima et al 2020 Mauritsen et al 2019) whichare both informative and readily comparable to observationsA comprehensive validation should therefore encompass arange of scales and a variety of variables to demonstratemodel performance not only for producing suitable averagesor accumulations but also for representing processes

Satellite-based remote sensing of terrestrial biogeochemicalcomponents has been conducted for almost 50 years sincethe launch of the Landsat satellite in 1972 (Xiao et al 2019Mack 1990) while field-based experimental and observa-tional data has been available since at least the early 19thcentury (Holland et al 2015) Just in terms of satellite-basedobservational data products there are currently thousands ofexamples available (Waliser et al 2020) Despite this seem-

ing wealth of observational data and observation-based dataproducts the implementation of a variety of observation-based references for validation of terrestrial biogeochemi-cal cycles within ESMs and LSMs is challenging for severalreasons These include the specifications required for directmodel output comparison inconsistent spatial and temporaldomains missing observations logistical biases and largeuncertainty in global-scale data products (Delire et al 2020Collier et al 2018 Lovenduski and Bonan 2017) The in-complete coverage of observational datasets in space-time di-mensions has led to significant bias in comparisons of modeldata and observation data previously (de Mora et al 2013)though this has not been generally discussed in validation ex-ercises by CMIP6 participants Observational discontinuityhas been addressed previously in a LSM validation by Orth etal (2017) which excluded daily observation reference aver-ages when more than 1 h of data from a 24 h period was miss-ing and through exclusion criteria in Collier et al (2018)For example the compilation of satellite observations to de-velop a LAI data product with one observation-based esti-mate every 15 d by Zhu et al (2013) for monthly averageor seasonal extrema comparison would require careful con-sideration for comparison to model averages computed frommore resolved output In an analysis of how sparse histori-cal measurements compare to continuous model output deMora et al (2013) demonstrate that where data are lackingin time or space the discrete comparison of model output torecords from site-level measurements may provide a strate-gic assessment of model performance over time especiallyin producing interannual variability Site-level comparisonsof GPP and or CO2 concentrations were performed by Delireet al (2020) Dunne et al (2020) and Vuichard et al (2019)while Collier et al (2018) caution against the use of spatiallysparse data but indicate that inclusion of site level evaluationsis a key future focus for the ILAMB project

Another approach to overcome spatial discontinuity maybe to compare broad gradients or trends in a given variablewith reference datasets such as regional and functional-typetrends in forest carbon stocks rather than a global summa-tion or average (Thurner et al 2014) to investigate whetheror not the model captures enduring spatial patterns In addi-tion some observational methods may invoke inherent biassuch as satellite-based observation estimates of LAI in mid tohigh latitudes seasonally under-estimating LAI due to snowcover leading to ambiguous model performance assessment(Ziehn et al 2020 Liu et al 2018) Observational uncer-tainty can be addressed by applying a weighting to referencedatasets as in ILAMBv21 as well as by using more than oneobservational reference when available (Eyring et al 2020Sellar et al 2019 Collier et al 2018) Careful considera-tion of spatiotemporal discontinuity in observations and in-herent bias is warranted in future validations which can beachieved through filtered exclusions site-level comparisonspattern comparison certainty weighting of datasets and theuse of more than one reference dataset

The globally gridded 1982ndash2008 GPP data product fre-quently used for GPP validation by CMIP6 participants wasdeveloped from machine learning upscaling of site-leveleddy-covariance Fluxnet observations with model tree en-sembles based on remote sensing vegetation indices meteo-rological data and land use (Jung et al 2011) Observation-based estimates of GPP can be obtained through satellite-derived vegetation indices such as the normalized differencevegetation index (NDVI Phillips et al 2008) and solar-induced chlorophyll fluorescence (Zhang et al 2020) in ad-dition to ground-based monitoring of turbulent CO2 fluxeswith the eddy covariance technique (Jung et al 2009) Lo-gistical challenges with eddy covariance-based techniques ofestimating GPP can result in potentially extensive data gapsand systematic omission of diel cycle observations (Roddaet al 2021 Erkkilauml et al 2018 Jung et al 2011 Lass-lop et al 2010 2008 Desai et al 2008) For example ina study of eddy-covariance monitoring of CO2 flux Jonssonet al (2008) report only 34 data coverage of a growingseason period of which 54 was discarded as it did notdemonstrate energy balance closure To address these chal-lenges Jung et al (2011) employ Bowen ratio correctionsof energy imbalance (Twine et al 2000) quality controlcriteria to exclude sites with more than 20 missing ob-servations and monthly averages to alleviate noise WhereNEE observations are missing in space over time driver re-lationships can be utilized for multi-decadal extrapolationthough only 38 and 60 of Fluxnet sites with less than15 years of observations capture mean conditions and inter-annual variability of drivers sufficiently well for this extrapo-lation as of 2015 and most have been operating for less than5 years (Chu et al 2017) While the site-level observationsfrom Jung et al (2011) originate from 212 sites presentinga globally extensive network regions with an important con-tribution to overall carbon stocks and fluxes are underrepre-sented (Jung et al 2020) and even the recent global FluxnetGPP data product by Jung et al (2016) has just 14 tropi-cal and 5 Arctic sites GPP observations from Fluxnet prod-ucts currently do not account for fire and waterbody emis-sions which prompts regional and interannual bias (Jung etal 2020) Despite these caveats such global-scale data prod-ucts provide a critical resource to the CMIP community inconducting model validation (Collier et al 2018) and therelatively common use of Jung et al (2011) for validationsby CMIP6 participants coincidentally reduces the influenceof observational contradiction (Xie et al 2020 Anav et al2015) Site-level GPP evaluation with observations from thetropics by Delire et al (2020) and Vuichard et al (2019)demonstrates a strategic approach to addressing the repre-sentation bias in GPP validations Site-level evaluations of-ten benefit from a wealth of available information includingspatially consistent meteorological forcing and avoid the in-fluence of spatial extrapolation error While Jung et al (2011)do not provide uncertainty measures several forms of uncer-tainty are explicitly presented for the Fluxnet2015 dataset by

Pastorello et al (2020) Therefore the utility of Fluxnet GPPdata products could be improved with standardized use byparticipants in conjunction with other independent data prod-ucts select site-level evaluations explicit uncertainty quan-tifications and improved ecological representation in under-lying site-level data

43 Statistical metrics and validation approaches

Several participants relied primarily on residual-based met-rics such as bias (simulated-observed) for validation ofterrestrial biogeochemical cycle model components On aspatial basis bias can identify significant regional over- orunder-estimations of a given variable However the attribu-tion of model error from global maps of bias can be am-biguous as the displayed bias is the combined result of dif-ferent forms of uncertainty including model structural rep-resentations unforced variability and spatial disagreement(Deser et al 2020 Lovenduski and Bonan 2017 Koch etal 2016) Such residual-based metrics may not indicate howwell the model would perform in simulating future condi-tions beyond the current contextual envelope of observations(Gulden et al 2008) and neglect the contribution of uncer-tainty from observations These limitations are considerablein the context of ESMs and LSMs as tools for predictingterrestrial biogeochemical function A more contextualizedbias assessment is the Wilcoxon test as applied by Swartet al (2019) to filter insignificant bias In a LSM evalua-tion Orth et al (2017) provides an observationally robustbias assessment by subtracting mean seasonal cycles fromeach grid cell and correlating the resulting anomalies be-tween observation-based datasets and model output In ad-dition RMSE normalized by the mean or standard deviationof the observed quantity (NRMSE) contextualizes the differ-ence between simulated and observed variable quantities interms of the magnitude or inherent variability of the variableof interest (Swart et al 2019 Fan et al 2018) which is ad-vantageous for variables such as GPP with large interannualvariability

Beyond these a variety of targeted model skill metricshave been published for process-based modelling that pro-vide detailed assessments of different forms of model un-certainty (Collier et al 2018 Orth et al 2017 Eyring etal 2016b Koch et al 2016 Law et al 2015 Kumar etal 2012 Taylor 2001 Kobayashi and Salam 2000) Meansquared deviation the sum of squared bias squared differ-ence between standard deviations and lack of correlationweighted by standard deviations presented by Kobayashiand Salam (2000) was used by Vuichard et al (2019) Thismetric is readily applicable to the objective validation andimprovement of mechanistic models as its dissection allowsfor the accurate attribution of different sources of model er-rors Additionally a Taylor diagram (Fig 4 Taylor 2001)conveys several dimensions of model error allows for theconcise simultaneous display of variables and models was

utilized in the evaluation of BCC-AVIM2 (Li et al 2019)NORESM2 (Seland et al 2020) and several LSMs andESMs by Anav et al (2015) and is incorporated into IL-AMBv21 (Collier et al 2018) The Taylor diagram wasdesigned for simultaneous performance comparison of sev-eral simulated variables and serves as a concise and infor-mative validation tool Caution is warranted however in theevaluation of fully coupled model output due to the inabil-ity of fully coupled models to reproduce the timing of inter-nal climate variability phenomena such as El NintildeondashSouthernOscillation (ENSO Flato et al 2013) While the magni-tude of observed and simulated internal climate variabilitymay be statistically consistent bias RMSE and NRMSE as-sessments of fully coupled model output should encompassdecadal or longer periods to address the influence of temporalmismatches in simulated internal climate variability relativeto observational records Alternatively as offline simulationscan be directly forced with historical observation data theoutput of offline simulations can be validated on a finer tem-poral scale

For example Taylor diagrams of global and regional NPPby Anav et al (2015) demonstrate a consistent low correla-tion and high standard deviation for model estimates in thetropics that is substantially reduced in the extratropics andglobally warranting focus on tropical NPP The validationprocess of terrestrial biogeochemical cycles and dissection ofmodel uncertainty may also be enhanced through offline sim-ulations or models with intermediate complexity as these al-low for a greater replication of simulations with different ini-tializations forcing datasets and model configurations due totheir computational affordability (Bonan et al 2019 Umairet al 2018 Orth et al 2017) Offline simulations also reducethe potential for incidental compounding error from couplingcomponents though this leads to an under-estimation in un-certainty for equivalent fully coupled simulations Replicatesimulations with different initial conditions such as thoseperformed by Danabasoglu et al (2020) allow for the at-tribution of uncertainty from unforced variability which ac-counted for half of the inter-model spread in key variablespreviously (Deser et al 2020 Eyring et al 2019) In addi-tion replicate simulations with different forcing datasets canindicate the role of forcing uncertainty (Wei et al 2018)which Lawrence et al (2019) found to be significant Fur-ther sensitivity analyses or perturbed parameter analyses in-volving replicated simulations with one or more variablesfixed as performed by Hajima et al (2020) and Lawrence etal (2019) illuminate structural uncertainty The use of well-established statistical and model performance metrics in ad-dition to strategic simulations facilitates a detailed analysisof model uncertainty

44 Moving forward

A model can only be expected to perform well in simulat-ing past present and future conditions if it is provided with

Figure 4 Taylor diagram from Taylor (2001) The standard deviation of model fields is displayed as the radial distance from the origin andcan be visually compared to the observed (reference) point which is indicated by a circle on the abscissa The correlation between the modeland observed fields decreases with azimuthal angle (dotted lines) and the root-mean-square difference between the model and observedfields is proportional to the distance from the reference point (quantified by dashed contours)

high-quality observational constraints Lovenduski and Bo-nan (2017) suggest that obtaining accurate observations andimproving process understanding should take precedenceover reducing model spread as constraining models to un-certain observations does not improve their predictive capac-ity and even models that agree well with observations canprompt divergent projections Several of the challenges in-herent in implementing observations in model validation anddevelopment are now a key focus of the Observations forModel Intercomparison Project (obs4MIPs Waliser et al2020) which strives to deliver long-term high-quality ob-servations from international efforts An obs4MIPs meetingheld in preparation for CMIP6 with more than 50 satellitedata and global climate modelling experts identified under-utilized observation products and recommended new effortsto address knowledge gaps including an expanded inven-tory of datasets higher-frequency datasets and model outputmore reliable uncertainty measures more datasets tailored tooffline simulations and more explicit metadata for modellers(Waliser et al 2020) Further recent satellite missions suchas the Sentinel2A twin satellite launched in 2015 have un-precedented spectral spatial and temporal resolution com-binations which can be used alone or in combination withother satellite-based observations to provide higher-fidelityreferences for validation (Vafaei et al 2018) Field experi-mental data provide unique insight as to the functional re-sponses of vegetation to elevated CO2 concentration (Gollet al 2017) temperature change (Richardson et al 2018)moisture availability (Williams et al 2019 Hovenden andNewton 2018) and nutrient limitations (Fleischer et al2019) outside the current context of observations The inte-gration of experimental findings in evaluations is challeng-

ing given the environmentally rapid application of treatmentsand limited ecological representation (Nowak et al 2004)though sophisticated relationship-based techniques such asused by Goll et al (2017) alleviate some of these issues In-creased collaboration between field and model researchersin designing experiments could improve the applicability offuture experiments In addition enhanced field and remotesensing collaboration would allow for higher-fidelity cali-brated global data products (Orth et al 2017 Verger et al2016) Thus future CMIPs will benefit from forthcoming col-laborations and reference data products tailored for valida-tion

A standard protocol for the validation of terrestrial biogeo-chemical variables would facilitate a thorough and objectiveassessment of model performance within and among partici-pants Further the collective merits and limitations of the cur-rent variety of approaches utilized by participants could beconsolidated and addressed in a comprehensive protocol Inthe interest of model improvement and weighting for predic-tions validation with an exhaustive assessment of variablesacross a range of spatiotemporal scales against all availablepeer-recommended observation-based references is optimalDataset-specific expertise is also warranted to correctly im-plement reference datasets in these evaluations (Waliser etal 2020 Liu et al 2018) The procurement and applica-tion of reference datasets within validations is demanding forparticipants considering their presiding obligation to con-tinuously refine model components and participate in CMIPwith computationally expensive ESM simulations Addition-ally the universal inclusion of often overlooked processessuch as moisture limitation nitrogen and phosphorus cy-cles dynamic vegetation prognostic leaf phenology and

natural disturbance regimes should be a priority focus forparticipants in developing diagnostic models as these pro-cesses are highly influential on terrestrial biogeochemistryand physics (Eyring et al 2020 Fleischer et al 2019 Piaoet al 2019 Wieder et al 2015 Achard et al 2014 Richard-son et al 2013 Heimann and Reichstein 2008 Tucker etal 1986) and the omission of these processes contributes towidespread bias (Green et al 2019 Anav et al 2015) Whileoutside the focus of this review equal attention should be ap-plied to the physical components of terrestrial biogeochem-ical cycles including explicit representation of permafrostand riverine carbon transport dynamics In fact a study in-cluding four CMIP5 ESMs found that soil moisture variabil-ity prompted variability in terrestrial NBP on the order of gi-gatonnes with non-linear responses to both moisture scarcityand excess (Green et al 2019) Further many of the meritsand limitations of the validation approaches discussed hereinapply to the validation of these physical components as well

The communal use of software packages such as ESM-ValToolv20 and ILAMBv21 (Eyring et al 2020 Collier etal 2018) could liberate time and computational resources formodellers In addition this would standardize validation pro-tocols address long-overlooked model uncertainty distinc-tions (Deser et al 2020) and avoid terminology confusion(Lovett et al 2006) While these packages include exten-sive suites of peer-verified observational reference datasetsand performance metrics these packages do not yet includeevaluation of nitrogen and phosphorus cycles which maybe due to the combined scarcity of observations upscal-ing approaches and model representations (Lawrence et al2019 Zhu et al 2018 Wieder et al 2015 Zaehle and Dal-monech 2011) The strategic situation of nitrogen phospho-rus and soil moisture monitoring which coincides with cur-rent Fluxnet sites (Jung et al 2020) could provide high-fidelity insight into nutrient and environmental limitations onGPP coherent turnover time assessments and broadly ap-plicable functional relationships to facilitate upscaling Theco-situation of multiple observational monitoring objectivesat Fluxnet sites would enhance the utility of each site-leveldataset and alleviate errors due to spatiotemporal inconsis-tencies between datasets in both performing evaluations anddeveloping large-scale data products Following increasedcollaboration between empirical and modelling communitiesto strategically expand observations and their inclusion ina comprehensive evaluation software the CMIP-designateduse of such software would standardize conserve and aug-ment validation efforts

5 Conclusion

The current generation of ESMs that participated in thesixth phase of the Coupled Model Intercomparison Projectadopted a broad assortment of approaches to validate histori-cally simulated terrestrial biogeochemical cycles Validationswhich encompassed a large suite of variables over a range ofspatiotemporal scales in conjunction with informative modelperformance metrics demonstrated relatively comprehensiveassessments of model performance Across CMIP6 partic-ipants the variety of variables reference datasets evalua-tion dimensions and statistical metrics utilized make generalassessments of model performance in simulating terrestrialbiogeochemistry challenging To address this inconsistencyand alleviate the immense responsibilities of participants werecommend the designation of a standard validation proto-col for CMIP participants which is consolidated in an open-source software (such as the Earth System Model Evalua-tion Tool version 2 (ESMValToolv20) or the InternationalLand Model Benchmarking version 21 (ILAMBv21)) Thisprotocol should utilize a comprehensive suite of certainty-weighted observational reference datasets targeted modelperformance metrics and comparisons across a range ofspatiotemporal dimensions The insights from a universallyadopted validation protocol would precisely attribute modeluncertainty and aid in directing future observational effortsto improve crucial process understanding within terrestrialbiogeochemical cycles

Appendix A Technical summary of validation activitiesby participants

A1 CSIRO

The Australian Community Climate and Earth SystemSimulator (ACCESS-ESM15) was developed by the Aus-tralian modelling group Commonwealth Scientific and In-dustrial Research Organization (CSIRO) for participationin CMIP6 (Ziehn et al 2020) The land surface modelused in ACCESS-ESM15 is the Community AtmosphereBiosphere Land Exchange (CABLE) model (Kowalczyket al 2013 2006) version 24 Ziehn et al (2020) com-pared ACCESS-ESM15 simulated land carbon cycle vari-ables against observation-based estimates for the 1986ndash2005period The spatial distribution of simulated average an-nual GPP was compared to upscaled Fluxnet observationsfrom Jung et al (2011) while average annual global GPPwas compared to observation-based estimates from Beer etal (2010) and Ziehn et al (2011) Simulated LAI magni-tude and seasonality was compared to global and regionalestimates based on Moderate Resolution Imaging Spectrora-diometer (MODIS) and Advanced Very High-Resolution Ra-diometer (AVHRR) data from Zhu et al (2013) Simulatedsurface CO2 concentrations in terms of mean seasonal cycleamplitude and timing were compared to four NOAAEarthSystem Research Laboratory station flask samples providedin the GLOBAL VIEW data product (GLOBAL VIEW-CO22013)

A2 BCC

The Beijing Climate Centre (BCC) participated in CMIP6with the BCC Climate System Model version 2 with mediumresolution (BCC-CSM2-MR Wu et al 2019) Land bio-geochemistry in BCC-CSM2-MR was simulated through theBCC Atmosphere and Vegetation Interactive Model version20 (BCC-AVIM2 Li et al 2019) While Wu et al (2019)did not provide validation results for terrestrial biogeochem-istry from BCC-CSM2-MR a detailed validation with offlinesimulations of BCC-AVIM2 was provided by Li et al (2019)using the Princeton global forcing dataset (Sheffield et al2006) Li et al (2019) compared the annual peak month sea-sonal average and global average of LAI to satellite obser-vations from 1982 to 2010 by the AVHRR (Myneni et al1997) Surface carbon fluxes including GPP and ER werecompared to upscaled Fluxnet observations from Jung etal (2011) Aboveground biomass was compared to Avitabileet al (2016) while global total biomass carbon from 1990to 2010 was compared to Saatchi et al (2011) The perfor-mance of BCC-AVIM2 in estimating each of these variableswas assessed through bias RMSE and Taylor diagram met-rics (Taylor 2001)

A3 CCCma

The Canadian Centre for Climate Modelling and Analy-sis (CCCma) participated in CMIP6 with the CCCma fifth-generation Earth System model (CanESM5 Swart et al2019) The land biogeochemistry component of CanESM5is the Canadian Terrestrial Ecosystem Model (CTEMArora and Boer 2010 2005) Swart et al (2019) com-pared CanESM5 simulated GPP from 1982 to 2009 withobservation-based estimates from Jung et al (2009) in termsof geographical distribution zonal averages and functionalrelationships with air temperature and precipitation Sev-eral metrics were used to illustrate CanESM5rsquos performancein simulating GPP including the correlation coefficient (r)between simulated and observed spatial patterns in GPPbias (simulatedminus observed) and root-mean-squared error(RMSE) normalized (NRMSE) by observed spatial standarddeviation Global average decadal landndashatmosphere CO2 fluxand net cumulative atmospherendashland CO2 flux from 1850 to2014 were compared to observation-based estimates from theGlobal Carbon Project (GCP Le Queacutereacute et al 2018) the lat-ter by subtracting cumulative land use emissions from cumu-lative land carbon uptake

A4 Climate and Global Dynamics Laboratory (NCAR)

The Community Earth System Model version 2 (CESM2)was developed by the Climate and Global Dynamics Labo-ratory at the American National Centre for Atmospheric Re-search (NCAR) for participation in CMIP6 (Danabasoglu etal 2020) The land component of CESM2 is the Commu-nity Land Model Version 5 (CLM5 Lawrence et al 2019)Danabasoglu et al (2020) and Lawrence et al (2019) com-prehensively assessed terrestrial biogeochemical cycle vari-able outputs from simulations of CESM2 and CLM5 re-spectively with the International Land Model Benchmark-ing package (ILAMBv21 Collier et al 2018) includingan explicit analysis of interannual variability with a three-member ensemble from different pre-industrial control ini-tialization years (CESM2) the influence of forcing throughthe use of three forcing datasets (CLM5) and the influence ofprescribed versus prognostic vegetation phenology (CLM5)ILAMBv21 utilizes a suite of data products weighted by cer-tainty These included vegetation biomass (tropical Saatchiet al 2011 global Kellndorfer et al 2013 Blackard etal 2008) burned area (Giglio et al 2010) CO2 concentra-tions GPP (Fluxnet Lasslop et al 2010 Global biospherendashatmosphere flux Jung et al 2010) LAI (AVHRR Myneni etal 1997 MODIS de Kauwe et al 2011) global net ecosys-tem carbon balance (GCP Le Queacutereacute et al 2014 Hoffman etal 2014) net ecosystem exchange (Fluxnet Lasslop et al2010 GBAF Jung et al 2010) NBP ER NEP (equivalentto GPP-ER) soil carbon (Harmonized World Soil Database(HWSD) Todd-Brown et al 2013 Northern CircumpolarSoil Carbon Database (NCSCDV22) Hugelius et al 2013)

and 10 functional relationships Lawrence et al (2019) alsocompared the relationship between apparent soil carbonturnover times versus air temperature to observation-basedestimates developed from HWSD NCSDV22 and MODISLawrence et al (2019) additionally compared maximummonthly LAI and average Vcmax25 (maximum RuBisCO car-boxylation rate at 25 C and high irradiance per unit leaf areain micromol mminus2 sminus1) at the PFT-level for the year 2010 to Zhaoet al (2005) and Kattge et al (2009) respectively as well ascanopy height for the year 2005 for tree PFTs to Simard etal (2011) Nitrogen cycle variables evaluated by Lawrenceet al (2019) with observational references included nitro-gen deposition (Fowler et al 2013) symbiotic fixed nitrogen(Vitousek et al 2013) soy fixed nitrogen (Herridge et al2008) crop nitrogen fertilization (Fowler et al 2013) den-itrification (Fowler et al 2013) hydrologic nitrogen losses(Fowler et al 2013) fire losses (Lamarque et al 2010)and N2O flux (Fowler et al 2013) Different climate forc-ing datasets and anthropogenic forcings were utilized to ex-amine the effect of climate CO2 emissions land use changeand nitrogen additions on carbon cycle variables and threeCLM versions to partition total uncertainty into forcing andmodel contributions using fixed-effect analysis of variancewith additional PFT-level analysis and prognostic versus pre-scribed vegetation and carbon cycling for CLM5 In addi-tion to the ILAMB validation Danabasoglu et al (2019)and Lawrence et al (2019) compared simulated global netbiome production (NBP) and cumulative land carbon sinkto observation-based estimates from 1850 to 2014 from theGCP for 1959ndash2014 (Le Queacutereacute et al 2016) and from Hoff-man et al (2014) for 1850ndash2010 Observation-based GPPER and NEP (equivalent to GPP-ER) comparison data wereobtained from Jung et al (2011 2010) Vegetation carbonwas evaluated relative to observations for the tropics fromSaatchi et al (2011) and the GEOCARBON and Global-Carbon datasets (Collier et al 2018 Avitabile et al 2016Santoro et al 2015) ILAMBv21 results from these inves-tigations comprised a collection of statistical metrics for an-nual mean bias relative bias RMSE seasonal cycle phasespatial distribution and interannual variability in addition tofunctional relationships Bonan et al (2019) provides a de-tailed analysis of the role of climate forcing uncertainty ininfluencing CLM5 output

A5 CNRM and CERFACS

The Centre National de Recherches Meacuteteacuteorologiques(CNRM) and Centre Europeacuteen de Recherche et de Forma-tion Avanceacutee en Calcul Scientifique (CERFACS) contributedthe CNRM-ESM2-1 to CMIP6 (Seacutefeacuterian et al 2019) Theland component in CNRM-ESM2-1 is the Interaction Soil-Biosphere-Atmosphere with Total Runoff Integrating Path-ways with carbon cycling (ISBA-CTRIP Delire et al 2020)Seacutefeacuterian et al (2019) compared CNRM-ESM2-1 simulatedannual minimum and maximum LAI to AVHRR observa-

tions from 1998 to 2011 (Zhu et al 2013) The simulatedland carbon sink from 1982 to 2010 was compared to a multi-model estimate by Huntzinger et al (2013) These valida-tions included spatial bias global mean bias RMSE andspatial error correlation between CNRM ESM versions todistinguish model sources of error Delire et al (2020) val-idated offline ISBA-CTRIP simulated GPP NPP autotrophicrespiration and ER from 1980 to 2010 with estimates withthe mean of 12 products from the FluxComv1 dataset (Junget al 2017 2016 Tramontana et al 2016) and a satel-lite product from the Numerical Terradynamic SimulationGroup MODIS17A3 (NASA LP DAAC 2017 Zhao et al2005) with reference autotrophic respiration calculated asthe mean of FLUXCOM GPP products minus MODIS17A3NPP Simulated crop NPP for the 2000s was comparedto the Harvested Area and Yield dataset (Monfreda et al2008) Carbon use efficiency (CUE) calculated as the ra-tio of NPP to GPP was evaluated with observation andmodel-based estimates for tropical evergreen forest fromMalhi et al (2009) and tropical deciduous temperate andboreal forests from He et al (2018) Zhang et al (2014)and theoretical derivations by Amthor (2000) Simulatedheterotrophic respiration was evaluated with a data prod-uct from Hashimoto et al (2015) which combines globaland Amazonian in situ observations from the Soil Respi-ration database (Bond-Lamberty et al 2018) and Malhi etal (2009) respectively and global gridded climate data Thesimulated burned area and fire CO2 emissions were com-pared to Mouillot and Field (2005) and the Global Fire Emis-sions Database version 41 (Randerson et al 2017 van derWerf et al 2017) Simulated dissolved organic carbon yieldleached from soil was compared to model results of May-orga et al (2010) and observations by Dai et al (2012)Simulated global aboveground biomass carbon was validatedwith observation-based estimates from 1993ndash2012 from Liuet al (2015) regional datasets for midndashhigh northern lati-tudes from Thurner et al (2014) and tropical datasets fromSaatchi et al (2011) and Baccini et al (2012) Simulatedaboveground litter carbon was compared to site measure-ments from 1827 to 1997 from the Global Database of Litter-fall Mass and Litter Pool Carbon and Nutrients (Holland etal 2015) Simulated belowground organic carbon was vali-dated with the HWSDv12 (FAO 2012) Vegetation turnovertime calculated as biomass divided by NPP and soil turnovertime calculated as the combination of litter and soil carbondivided by NPP for 1984ndash2014 were also computed for val-idation Delire et al (2020) also used local scale Fluxnetdata from Joetzjer et al (2015) to assess ISBA-CTRIP per-formance Each variable was validated through comparisonof the distribution of simulated and observation-based esti-mates of annual averages as well as zonal averages and thespatial distribution of the bias (simulated minus observed)Average simulated carbon fluxes from 2006ndash2015 and thetrend from 1960ndash2015 were also compared to observation-

based estimates from the GCP (Le Queacutereacute et al 2018) andCiais et al (2019)

A6 IPSL

The Institut Pierre Simon Laplace (IPSL) participatedin CMIP6 with IPSL-CM6A-LR the land component ofwhich was the ORCHIDEE land surface model version 20(Boucher et al 2020 Hourdin et al 2020) Boucher etal (2020) evaluated IPSL-CM6A-LR simulated average an-nual carbon fluxes from 1990 to 1999 and 2009 to 2018resulting from land cover change fossil fuel emissions theterrestrial sink and total net land fluxes (the terrestrial sinkminus land cover change) with observation-based estimatesfrom the 2019 GCP (Friedlingstein et al 2019) Vuichardet al (2019) validated ORCHIDEE simulated GPP in termsof the mean annual seasonal and daily simulated GPP on aPFT spatial and global basis against observations from 78Fluxnet sites (Vuichard and Papale 2015) and the global-scale MTE-GPP product based upon upscaled Fluxnet ob-servations for 1982ndash2008 (Jung et al 2011) RMSE anddissected mean-squared deviation (MS the sum of squaredbias squared difference between standard deviations andlack of correlation weighted by standard deviations based onKobayashi and Salam 2000) metrics were used to attributedifferent sources of uncertainty The relative contribution ofdrivers of variation in present-day GPP were also assessedincluding seasonal variability in NOx and NHx depositionand the leaf carbon to nitrogen ratio The sensitivity of OR-CHIDEE output to model structure in terms of MSE was alsoanalyzed on a global and PFT-level basis including fixed anddynamic fully coupled carbonndashnitrogen cycles

A7 GFDL

The American National Oceanic and Atmospheric Admin-istration Geophysical Fluid Dynamics Laboratory (GFDL)participated in CMIP6 with GFDL-ESM41 (Dunne et al2020) in which land biogeochemistry is simulated with theGFDL Land Model version 41 (LM41 Shevliakova et al2021) Dunne et al (2020) validated GFDL-ESM 41rsquos sim-ulated spatial distribution of seasonal amplitude in CO2 con-centrations and interannual variability of CO2 concentrationscompared to NOAA Global Monitoring Division sites with atleast 15-year records (Global Monitoring Laboratory 2005)using RMSE and the coefficient of determination (r2) aswell as the correlation coefficient (r) for individual sites

A8 JAMSTEC University of Tokyo and NationalInstitute for Environmental Studies

The Japanese Agency for Marine-Earth Science and Tech-nology (JAMSTEC) University of Tokyo and National In-stitute for Environmental Studies participated in CMIP6 withthe Model for Interdisciplinary Research on Climate EarthSystem version 2 for Long-term simulations (MIROC-ES2L

Hajima et al 2020) The land biogeochemical componentin MIROC-ES2L is the Vegetation Integrative Simulator forTrace gases model (VISIT-e Ito and Inatomi 2012) Hajimaet al (2020) evaluated MIROC-ES2L simulated terrestrialcarbon gain with and without land use as well as land useemissions from 1850 to 2014 in comparison to multi-modelestimates from the GCP (Le Queacutereacute et al 2018) Observation-based data products used for other comparisons included (1)the spatial pattern gradient across biomes magnitude sea-sonality and length of growing season of global gridded GPPfrom 1986 to 2005 from Fluxnet (Jung et al 2011) (2) themagnitude and density of forest carbon stock (Kindermannet al 2008) and (3) global and regional soil organic car-bon from the harmonized soil property values for broad-scalemodelling (WISE30Sec Batjes 2016) the northern high lat-itudes from the Northern Circumpolar Soil Carbon Databaseversion 2 (NCSCDv2 Hugelius et al 2013) and an estimatefrom Todd-Brown et al (2013) developed from the HWSDversion 13 (FAO 2012) Hajima et al (2020) also comparedsimulated and observation-based estimates of annual biolog-ical nitrogen fixation (BNF) from 1850 to 2014 (Gruber andGalloway 2008) present-day BNF (Galloway et al 2008Herridge et al 2008) annual unperturbed state terrestrialN2 flux (Gruber and Galloway 2008) and change in annualsoil nitrous oxide emissions from 1850 to 2014 relative to amodel comparison study by Tian et al (2018)

A9 MPI

The Max Planck Institute for Meteorology (MPI) Earth Sys-tem Model version 12 Low Resolution (MPI-ESM12-LR)was developed for participation in CMIP6 (Mauritsen et al2019) by the MPI the land component of which is JS-BACH32 (Goll et al 2017) Mauritsen et al (2019) com-pared the spatial variability and zonally averaged densityof MPI-ESM12-LR simulated soil and litter carbon stocksto estimates by Goll et al (2015) developed from the Har-monized World Soil Database The simulated evolution inglobal total land carbon from 1850ndash2013 was compared toestimates provided by Ciais et al (2013) Additionally sim-ulated land use change carbon emissions from 1860 to 2013were compared to estimates provided by Ciais et al (2013)In a model description paper of JSBACH version 310 whichwas set to be used in CMIP6 Goll et al (2017) com-pare JSBACH31 simulated present-day NPP to Ito (2011)while simulated present-day biomass carbon was comparedto Saugier and Roy (2001) and Ciais et al (2013) Thesimulated response of NPP and GPP to increases in atmo-spheric CO2 were compared to experimentally observed es-timates from four free-air CO2 enrichment (FACE) exper-iments (Norby et al 2005) and an intramolecular isotopedistribution examination of plant metabolic shifts (Ehlers etal 2015) Simulated present-day biomass nitrogen was com-pared to Schlesinger (1997) while simulated present-day to-tal nitrogen was compared to Galloway et al (2013) Simu-

lated values of pre-industrial (1850) and present-day leach-ing and BNF were compared to Galloway et al (2013 2004)Vitousek et al (2013) and short-term experimental resultsfrom a meta-analysis by Liang et al (2016) while simu-lated present-day denitrification was compared to Gallowayet al (2013) Goll et al (2017) also verified the simulatedspatial variability in reactive nitrogen-loss pathways usinga compilation of nitrogen-15 isotopic data (Houlton et al2015) with the statistical metrics r RMSE and Taylor score(Taylor 2001)

A10 NCC

The Norwegian Earth System Model (NORESM2) was de-veloped for participation in CMIP6 (Seland et al 2020) bythe Norwegian Climate Consortium (NCC) and is basedon CESM2 As in CESM2 the land model in NORESM2is CLM5 (Lawrence et al 2019) The performance ofNORESM2 was validated through a three-member ensem-ble of historical simulations from 1850 to 2014 with slightlyvarying initial conditions Simulated carbon cycle variablesthat were compared to observation variables included GPPsoil carbon and vegetation carbon from Jung et al (2011)FAO (2012) and Avitabile et al (2016) and Santoro etal (2015) respectively Seland et al (2020) NORESM2 re-sults in terms of carbon stocks and fluxes broadly agree withthose of Lawrence et al (2019) while conducting land-onlysimulations of CLM5

A11 NERC and Met Office

The United Kingdom Community Earth System Model(UKESM1-0-LL) was developed for participation in CMIP6by the United Kingdom Natural Environmental ResearchCouncil (NERC) and National Meteorological Service (MetOffice Sellar et al 2019) The land component in UKESM1-0-LL is an updated version of the Joint UK Land Envi-ronment Simulator (JULES Clark et al 2011) with anadditional PFT-updated competition scheme (Harper et al2018) Sellar et al (2019) evaluated UKESM1-0-LL simu-lated global GPP magnitude and evolution in time throughcomparisons to recent decadal GPP from the Fluxnet modeltree ensemble data product (Jung et al 2011) The arealland cover of aggregated plant functional types (PFTs) wasvalidated with satellite observation-based datasets from theEuropean Space Agency Climate Change Initiative LandCover data (Poulter et al 2015) and the InternationalGeosphere-Biosphere Programme (IGBP) Land Use andCover Change project (Loveland et al 2000) using themodel year 2005 The coverage of PFTs were validated usingthese observation-based datasets as references both spatiallyand as a fraction of biomes based upon regions defined byOlson et al (2006) The simulated vegetation carbon distri-bution was validated on a latitudinal basis with observation-based estimates from GEOCAROBON (Avitabile et al

2016) and Saatchi et al (2011) while the spatial distribu-tion of soil carbon was validated with observation-based es-timates WISE30sec (Batjes 2016) IGBP-DIS (Global SoilData Task Group 2002) and Carvalhais et al (2014) Themagnitude of simulated global total soil carbon was com-pared to whole soil profile observation-based estimates fromCarvalhais et al (2014) and upper 2 m observation-based es-timates from Batjes (2016) Cumulative carbon uptake andland use emissions from 1850 to 2014 was compared toobservation-based estimates from the GCP (Le Queacutereacute et al2018)

Data availability The data used to generate Figs 1 and 2 areopenly available from CMIP6 model participant papers (see Ta-ble 1)

Author contributions LS and AHMD both initiated the researchand significantly contributed to the writing of the paper LS con-ducted the analysis and wrote the original draft AHMD providedsupervisory support

Competing interests The contact author has declared that neitherthey nor their co-author have any competing interests

Disclaimer Publisherrsquos note Copernicus Publications remainsneutral with regard to jurisdictional claims in published maps andinstitutional affiliations

Acknowledgements We wish to thank Susan Ziegler Hugo Bel-trami Entcho Demirov Joel Finnis and Evan Edinger for their in-sightful comments on an early draft of this manuscript This re-search was supported by the National Science and EngineeringResearch Council of Canada (NSERC) Discovery Grant ProgramLynsay Spafford is grateful for support from a NSERC CanadaGraduate Scholarships ndash Doctoral Program Scholarship

Financial support This research has been supported by the NaturalSciences and Engineering Research Council of Canada (DiscoveryGrant and Canada Graduate Scholarships ndash Doctoral Program)

Review statement This paper was edited by David Lawrence andreviewed by two anonymous referees

References

Achard F Beuchle R Mayaux P Stibig H J Bodart C BrinkA and Simonetti D Determination of tropical deforestationrates and related carbon losses from 1990 to 2010 Glob ChangeBiol 20 2540ndash2554 httpsdoiorg101111gcb12605 2014

Amthor J S The McCreendashde WitndashPenning de VriesndashThornley res-piration paradigms 30 years later Ann Bot-London 86 1ndash20httpsdoiorg101006anbo20001175 2000

Anav A Friedlingstein P Kidston M Bopp L Ciais P CoxP Jones C Jung M Myneni R and Zhu Z Evaluatingthe land and ocean components of the global carbon cycle inthe CMIP5 earth system models J Climate 26 6801ndash6843httpsdoiorg101175JCLI-D-12-004171 2013

Anav A Friedlingstein P Beer C Ciais P Harper A JonesC and Zhao M Spatiotemporal patterns of terrestrial grossprimary production A review Rev Geophys 53 785ndash818httpsdoiorg1010022015RG000483 2015

Anderson T R Hawkins E and Jones P D CO2 the greenhouseeffect and global warming from the pioneering work of Arrhe-

nius and Callendar to todayrsquos Earth System Models Endeavour40 178ndash187 httpsdoiorg101016jendeavour2016070022016

Arora V K and Boer G J A parameterization of leaf phenol-ogy for the terrestrial ecosystem component of climate modelsGlob Change Biol 11 39ndash59 httpsdoiorg101111j1365-2486200400890x 2005

Arora V K and Boer G J Uncertainties in the 20th cen-tury carbon budget associated with land use change GlobChange Biol 16 3327ndash3348 httpsdoiorg101111j1365-2486201002202x 2010

Arora V K Boer G J Friedlingstein P Eby M Jones CD Christian J R Bonan G Bopp L Brovkin V Cad-ule P Hajima T Ilyina T Lindsay K Tjiputra J F andWu T Carbonndashconcentration and carbonndashclimate feedbacksin CMIP5 Earth system models J Climate 26 5289ndash5314httpsdoiorg101175JCLI-D-12-004941 2013

Arora V K Katavouta A Williams R G Jones C D BrovkinV Friedlingstein P Schwinger J Bopp L Boucher O Cad-ule P Chamberlain M A Christian J R Delire C FisherR A Hajima T Ilyina T Joetzjer E Kawamiya M KovenC D Krasting J P Law R M Lawrence D M LentonA Lindsay K Pongratz J Raddatz T Seacutefeacuterian R TachiiriK Tjiputra J F Wiltshire A Wu T and Ziehn T Carbonndashconcentration and carbonndashclimate feedbacks in CMIP6 modelsand their comparison to CMIP5 models Biogeosciences 174173ndash4222 httpsdoiorg105194bg-17-4173-2020 2020

Avitabile V Herold M Heuvelink G B M Lewis S LPhillips O L Asner G P Armston J Ashton P S Banin LBayol N Berry N J Boeckx P de Jong B H J DeVries BGirardin C A J Kearsley E Lindsell J A Lopez-GonzalezG Lucas R Malhi Y Morel A Mitchard E T A Nagy LQie L Quinones M J Ryan C M Ferry S J F Sunder-land T Laurin G V Gatti R C Valentini R Verbeeck HWijaya A and Willcock S An integrated pan-tropical biomassmap using multiple reference datasets Glob Change Biol 221406ndash1420 httpsdoiorg101111gcb13139 2016

Baccini A Goetz S J Walker W S Laporte N T SunM Sulla-Menashe D Hackler J Beck P S A DubayahR Friedl M A Samanta S and Houghton R A Esti-mated carbon dioxide emissions from tropical deforestation im-proved by carbon-density maps Nat Clim Change 2 182ndash185httpsdoiorg101038nclimate1354 2012

Baret F Hagolle O Geiger B Bicheron P Miras B HucM Berthelot B Nintildeo F Weiss M Samain O RoujeanJ L and Leroy M LAI fAPAR and fCover CYCLOPESglobal products derived from VEGETATION Part 1 Princi-ples of the algorithm Remote Sens Environ 110 275ndash286httpsdoiorg101016jrse200702018 2007

Batjes N H Harmonized soil property values forbroad-scale modelling (WISE30sec) with estimatesof global soil carbon stocks Geoderma 269 61ndash68httpsdoiorg101016jgeoderma201601034 2016

Beer C Reichstein M Tomelleri E Ciais P Jung M Carval-hais N Roumldenbeck C Arain M A Baldocchi D Bonan GB Bondeau A Cescatti A Lasslop G Lindroth A LomasM Luyssaert S Margolis H Oleson K W Roupsard OVeenendaal E Viovy N Williams C Woodward F I andPapale D Terrestrial gross carbon dioxide uptake global dis-

tribution and covariation with climate Science 329 834ndash838httpsdoiorg101126science1184984 2010

Blackard J A Finco M V Helmer E H Holden G R Hop-pus M L Jacobs D M Lister A J Moisen G G NelsonM D Riemann R Ruefenacht B Salajanu D WeyermannD L Winterberger K C Brandeis T J Czaplewski R LMcRoberts R E Patterson P L and Tymcio R P MappingUS forest biomass using nationwide forest inventory data andmoderate resolution information Remote Sens Environ 1121658ndash1677 httpsdoiorg101016jrse200708021 2008

Bonan G B Lombardozzi D L Wieder W R Oleson K WLawrence D M Hoffman F M and Collier N Model struc-ture and climate data uncertainty in historical simulations of theterrestrial carbon cycle (1850ndash2014) Global Biogeochem Cy33 1310ndash1326 httpsdoiorg1010292019GB006175 2019

Bond-Lamberty B Bailey V L Chen M Gough C M and Var-gas R Globally rising soil heterotrophic respiration over recentdecades Nature 560 80ndash83 httpsdoiorg101038s41586-018-0358-x 2018

Boucher O Servonnat J Albright A L Aumont O Balkan-ski Y Bastrikov V Bekki S Bonnet R Bony S Bopp LBraconnot P Brockmann P Cadule P Caubel A Cheruy FCodron F Cozic A Cugnet D DrsquoAndrea F Davini P deLavergne C Denvil S Deshayes J Devilliers M DucharneA Dufresne J-L Dupont E Eacutetheacute C Fairhead L Fall-etti L Flavoni S Foujols M-A Gardoll S Gastineau GGhattas J Grandpeix J-Y Guenet B Guez L GuilyardiEacute Guimberteau M Hauglustaine D Hourdin F IdelkadiA Joussaume S Kageyama M Khodri M Krinner GLebas N Levavasseur G Leacutevy C Li L Lott F LurtonT Luyssaert S Madec G Madeleine J-B Maignan FMarchand M Marti O Mellul L Meurdesoif Y MignotJ Musat I Ottleacute C Peylin P Planton Y Polcher JRio C Rochetin N Rousset C Sepulchre P Sima ASwingedouw D Thieacuteblemont R Khadre Traore A Van-coppenolle M Vial J Vialard J Viovy N and VuichardN Presentation and evaluation of the IPSL-CM6A-LR cli-mate model J Adv Model Earth Sy 12 e2019MS002010httpsdoiorg1010292019MS002010 2020

Carvalhais N Forkel M Khomik M Bellarby J Jung MMigliavacca M Mu M Saatchi S Santoro M Thurner MWeber U Ahrens B Beer C Cescatti A Randerson J Tand Reichstein M Global covariation of carbon turnover timeswith climate in terrestrial ecosystems Nature 514 213ndash217httpsdoiorg101038nature13731 2014

Chu H Baldocchi D D John R Wolf S and Reichstein MFluxes all of the time A primer on the temporal representa-tiveness of Fluxnet J Geophys Res-Biogeo 122 289ndash307httpsdoiorg1010022016JG003576 2017

Ciais P Sabine C Bala G Bopp L Brovkin V Canadell JChhabra A DeFries R Galloway J Heimann M Jones CLe Queacutereacute C Myneni R B Piao S and Thornton P Carbonand Other Biogeochemical Cycles in Climate Change 2013The Physical Science Basis Contribution of Working Group Ito the Fifth Assessment Report of the Intergovernmental Panelon Climate Change edited by Stocker T F Qin D Plat-tner G-K Tignor M Allen S K Boschung J NauelsA Xia Y Bex V and Midgley P M Cambridge Univer-sity Press Cambridge United Kingdom and New York NY

USA available at httpswwwipccchsiteassetsuploads201802WG1AR5_Chapter06_FINALpdf (last access 1 April 2021) 2013

Ciais P Tan J Wang X Roedenbeck C Chevallier F Piao S-L Moriarty R Broquet G Le Queacutereacute C Canadell J G PengS Poulter B Liu Z and Tans P Five decades of northernland carbon uptake revealed by the interhemispheric CO2 gradi-ent Nature 568 221ndash225 httpsdoiorg101038s41586-019-1078-6 2019

Clark D B Mercado L M Sitch S Jones C D Gedney NBest M J Pryor M Rooney G G Essery R L H BlythE Boucher O Harding R J Huntingford C and Cox PM The Joint UK Land Environment Simulator (JULES) modeldescription ndash Part 2 Carbon fluxes and vegetation dynamicsGeosci Model Dev 4 701ndash722 httpsdoiorg105194gmd-4-701-2011 2011

Collier N Hoffman F M Lawrence D M Keppel-Aleks GKoven C D Riley W J Mu M and Randerson J T TheInternational Land Model Benchmarking (ILAMB) system de-sign theory and implementation J Adv Model Earth Sy 102731ndash2754 httpsdoiorg1010292018MS001354 2018

Dai M Yin Z Meng F Liu Q and Cai W J Spa-tial distribution of riverine DOC inputs to the ocean an up-dated global synthesis Curr Opin Env Sust 4 170ndash178httpsdoiorg101016jcosust201203003 2012

Danabasoglu G Lamarque J F Bacmeister J Bailey D ADuVivier A K Edwards J Emmons LK Fasullo J Gar-cia R Gettelman A Hannay C Holland M Large W Lau-ritzen P Lawrence D Lenaerts J Lindsay K Lipscomb WMills M J Neale R Oleson K Otto-Bliesner B PhillipsA Sacks W Tilmes S van Kampenhout L VertensteinM Bertini A Dennis J Deser C Fischer C Fox-KemperB Kay J Kinnison D Kushner P Larson V Long MMickelson S Moore J Nienhouse E Polvani L RaschP and Strand W The community earth system model ver-sion 2 (CESM2) J Adv Model Earth Sy 12 e2019MS001916 httpsdoiorg1010292019MS001916 2020

Davies-Barnard T Meyerholt J Zaehle S Friedlingstein PBrovkin V Fan Y Fisher R A Jones C D Lee H PeanoD Smith B Waringrlind D and Wiltshire A J Nitrogen cy-cling in CMIP6 land surface models progress and limitationsBiogeosciences 17 5129ndash5148 httpsdoiorg105194bg-17-5129-2020 2020

Defourny P Boettcher M Bontemps S Kirches G LamarcheC Peters M Santoro M and Schlerf M Land cover CCIProduct user guide version 2 Technical report European SpaceAgency London United Kingdom 1ndash91 2016

de Kauwe M G Disney M I Quaife T Lewis Pand Williams M An assessment of the MODIS Collec-tion 5 leaf area index product for a region of mixedconiferous forest Remote Sens Environ 115 767ndash780httpsdoiorg101016jrse201011004 2011

Delire C Seacutefeacuterian R Decharme B Alkama R Cal-vet J C Carrer D Gibelin A Joetzjer E Morel XRochner M and Tzanos D The global land carbon cy-cle simulated with ISBA-CTRIP Improvements over thelast decade J Adv Model Earth Sy 12 e2019MS001886httpsdoiorg1010292019MS001886 2020

de Mora L Butenschoumln M and Allen J I How shouldsparse marine in situ measurements be compared to a con-tinuous model an example Geosci Model Dev 6 533ndash548httpsdoiorg105194gmd-6-533-2013 2013

Desai A R Richardson A D Moffat A M KattgeJ Hollinger D Y Barr A and Stauch V J Cross-site evaluation of eddy covariance GPP and RE decom-position techniques Agr Forest Meteorol 148 821ndash838httpsdoiorg101016jagrformet200711012 2008

Deser C Lehner F Rodgers K B Ault T Delworth T LDiNezio P N and Ting M Insights from Earth system modelinitial-condition large ensembles and future prospects Nat ClimChange 10 277ndash286 httpsdoiorg101038s41558-020-0731-2 2020

Dunne J P Horowitz L W Adcroft A J Ginoux P Held IM John J G Krasting J P Malyshev S Naik1 V PaulotF Shevliakova E Stock C A Zadeh N Balaji V Blan-ton C Dunne K A Dupuis C Durachta J Dussin R Gau-thier P P G Griffies S M Guo H Hallberg R W Har-rison M He J Hurlin W McHugh C Menzel R MillyP C D Nikonov S Paynter D J Ploshay J Radhakrish-nan A Rand K Reichl B G Robinson T SchwarzkopfD M Sentman L T Underwood S Vahlenkamp H Win-ton M Wittenberg A T Wyman B Zeng Y and ZhaoM The GFDL Earth System Model version 41 (GFDL-ESM41) Overall coupled model description and simulation char-acteristics J Adv Model Earth Sy 12 e2019MS002015httpsdoiorg1010292019MS002015 2020

Ehlers I Augusti A Betson T R Nilsson M B Marshall JD and Schleucher J Detecting long-term metabolic shifts us-ing isotopomers CO2-driven suppression of photorespiration inC3 plants over the 20th century P Natl Acad Sci USA 11215585ndash15590 httpsdoiorg101073pnas1504493112 2015

Erkkilauml K-M Ojala A Bastviken D Biermann T Heiska-nen J J Lindroth A Peltola O Rantakari M Vesala Tand Mammarella I Methane and carbon dioxide fluxes overa lake comparison between eddy covariance floating cham-bers and boundary layer method Biogeosciences 15 429ndash445httpsdoiorg105194bg-15-429-2018 2018

Eyring V Bony S Meehl G A Senior C A Stevens BStouffer R J and Taylor K E Overview of the CoupledModel Intercomparison Project Phase 6 (CMIP6) experimen-tal design and organization Geosci Model Dev 9 1937ndash1958httpsdoiorg105194gmd-9-1937-2016 2016a

Eyring V Righi M Lauer A Evaldsson M Wenzel S JonesC Anav A Andrews O Cionni I Davin E L Deser CEhbrecht C Friedlingstein P Gleckler P Gottschaldt K-D Hagemann S Juckes M Kindermann S Krasting JKunert D Levine R Loew A Maumlkelauml J Martin G Ma-son E Phillips A S Read S Rio C Roehrig R Sen-ftleben D Sterl A van Ulft L H Walton J Wang S andWilliams K D ESMValTool (v10) ndash a community diagnos-tic and performance metrics tool for routine evaluation of Earthsystem models in CMIP Geosci Model Dev 9 1747ndash1802httpsdoiorg105194gmd-9-1747-2016 2016b

Eyring V Cox P M Flato G M Gleckler P J AbramowitzG Caldwell P Collins W D Gier B K Hall A D Hoff-man F M Hurtt G C Jahn A Jones C D Klein S AKrasting J P Kwiatkowski L Lorenz R Maloney E Meehl

G A Pendergrass A G Pincus R Ruane A C RussellJ L Sanderson B M Santer B D Sherwood S C Simp-son I R Stouffer R J and Williamson M S Taking climatemodel validation to the next level Nat Clim Change 9 102ndash110 httpsdoiorg101038s41558-018-0355-y 2019

Eyring V Bock L Lauer A Righi M Schlund M AndelaB Arnone E Bellprat O Broumltz B Caron L-P CarvalhaisN Cionni I Cortesi N Crezee B Davin E L Davini PDebeire K de Mora L Deser C Docquier D EarnshawP Ehbrecht C Gier B K Gonzalez-Reviriego N Good-man P Hagemann S Hardiman S Hassler B Hunter AKadow C Kindermann S Koirala S Koldunov N Leje-une Q Lembo V Lovato T Lucarini V Massonnet FMuumlller B Pandde A Peacuterez-Zanoacuten N Phillips A PredoiV Russell J Sellar A Serva F Stacke T SwaminathanR Torralba V Vegas-Regidor J von Hardenberg J WeigelK and Zimmermann K Earth System Model Evaluation Tool(ESMValTool) v20 ndash an extended set of large-scale diagnos-tics for quasi-operational and comprehensive evaluation of Earthsystem models in CMIP Geosci Model Dev 13 3383ndash3438httpsdoiorg105194gmd-13-3383-2020 2020

Fan J Chen B Wu L Zhang F Lu X and Xiang Y Evalua-tion and development of temperature-based empirical models forestimating daily global solar radiation in humid regions Energy144 903ndash914 httpsdoiorg101016jenergy2017120912018

Fan N Koirala S Reichstein M Thurner M AvitabileV Santoro M Ahrens B Weber U and CarvalhaisN Apparent ecosystem carbon turnover time uncertaintiesand robust features Earth Syst Sci Data 12 2517ndash2536httpsdoiorg105194essd-12-2517-2020 2020

FAO Harmonized World Soil Database v 12 available at httpwwwfaoorgsoils-portaldata-hubsoil-maps-and-databasesharmonized-world-soil-database-v12en (last access 29 Jan-uary 2021) 2012

Fisher R A Wieder W R Sanderson B M Koven CD Oleson K W Xu C and Lawrence D M Paramet-ric controls on vegetation responses to biogeochemical forc-ing in the CLM5 J Adv Model Earth Sy 11 2879ndash2895httpsdoiorg1010292019MS001609 2019

Flato G Marotzke J Abiodun B Braconnot P Chou S CCollins W Cox P Driouech F Emori S Eyring V ForestC Gleckler P Guilyardi E Jakob C Kattsov V Reason Cand Rummukainen M Evaluation of climate models in Cli-mate Change 2013 The Physical Science Basis Contribution ofWorking Group I to the Fifth Assessment Report of the Intergov-ernmental Panel on Climate Change edited by Stocker T FQin D Plattner G-K Tignor M Allen S K Boschung JNauels A Xia Y Bex V and Midgley P M Cambridge Uni-versity Press Cambridge United Kingdom and New York NYUSA 2013

Fleischer K Rammig A De Kauwe M G Walker AP Domingues T F Fuchslueger L and Lapola D MAmazon forest response to CO2 fertilization dependent onplant phosphorus acquisition Nat Geosci 12 736ndash741httpsdoiorg101038s41561-019-0404-9 2019

Fowler D Coyle M Skiba U Sutton M A Cape J NReis S Sheppard L J Jenkins A Grizzetti B Gal-loway J N Vitousek P Leach A Bouwman A F

Butterbach-Bahl K Dentener F Stevenson D AmannM and Voss M The global nitrogen cycle in the twenty-first century Philos T R Soc B 368 20130164ndash20130164httpsdoiorg101098rstb20130164 2013

Friedlingstein P Jones M W OrsquoSullivan M Andrew R MHauck J Peters G P Peters W Pongratz J Sitch S LeQueacutereacute C Bakker D C E Canadell J G Ciais P Jack-son R B Anthoni P Barbero L Bastos A Bastrikov VBecker M Bopp L Buitenhuis E Chandra N ChevallierF Chini L P Currie K I Feely R A Gehlen M GilfillanD Gkritzalis T Goll D S Gruber N Gutekunst S Har-ris I Haverd V Houghton R A Hurtt G Ilyina T JainA K Joetzjer E Kaplan J O Kato E Klein Goldewijk KKorsbakken J I Landschuumltzer P Lauvset S K Lefegravevre NLenton A Lienert S Lombardozzi D Marland G McGuireP C Melton J R Metzl N Munro D R Nabel J E M SNakaoka S-I Neill C Omar A M Ono T Peregon APierrot D Poulter B Rehder G Resplandy L Robertson ERoumldenbeck C Seacutefeacuterian R Schwinger J Smith N Tans P PTian H Tilbrook B Tubiello F N van der Werf G R Wilt-shire A J and Zaehle S Global Carbon Budget 2019 EarthSyst Sci Data 11 1783ndash1838 httpsdoiorg105194essd-11-1783-2019 2019

Friedlingstein P OrsquoSullivan M Jones M W Andrew R MHauck J Olsen A Peters G P Peters W Pongratz J SitchS Le Queacutereacute C Canadell J G Ciais P Jackson R B AlinS Aragatildeo L E O C Arneth A Arora V Bates N RBecker M Benoit-Cattin A Bittig H C Bopp L BultanS Chandra N Chevallier F Chini L P Evans W FlorentieL Forster P M Gasser T Gehlen M Gilfillan D Gkritza-lis T Gregor L Gruber N Harris I Hartung K Haverd VHoughton R A Ilyina T Jain A K Joetzjer E Kadono KKato E Kitidis V Korsbakken J I Landschuumltzer P LefegravevreN Lenton A Lienert S Liu Z Lombardozzi D MarlandG Metzl N Munro D R Nabel J E M S Nakaoka S-INiwa Y OrsquoBrien K Ono T Palmer P I Pierrot D Poul-ter B Resplandy L Robertson E Roumldenbeck C SchwingerJ Seacutefeacuterian R Skjelvan I Smith A J P Sutton A J Tan-hua T Tans P P Tian H Tilbrook B van der Werf GVuichard N Walker A P Wanninkhof R Watson A JWillis D Wiltshire A J Yuan W Yue X and Zaehle SGlobal Carbon Budget 2020 Earth Syst Sci Data 12 3269ndash3340 httpsdoiorg105194essd-12-3269-2020 2020

Galloway J N Dentener F J Capone D G Boyer E WHowarth R W Seitzinger S P Asner G P Cleveland CC Green P A Holland E A Karl D M Michaels AF Porter J H Townsend A R and Vo C J Nitrogen cy-cles past present and future Biogeochemistry 70 153ndash226httpsdoiorg101007s10533-004-0370-0 2004

Galloway J N Townsend A R Erisman J W Bekunda MCai Z Freney J R Martinelli L A Seitzinger S P andSutton M A Transformation of the nitrogen cycle recenttrends questions and potential solutions Science 320 889ndash892httpsdoiorg101126science1136674 2008

Galloway J N Leach A M Bleeker A and Eris-man J W A chronology of human understanding ofthe nitrogen cycle Philos T R Soc B 368 20130120httpsdoiorg101098rstb20130120 2013

Gibbs H K Olsonrsquos Major World Ecosystem ComplexesRanked by Carbon in Live Vegetation An UpdatedDatabase Using the GLC2000 Land Cover Product NDP-017b Oak Ridge National Laboratory Oak Ridge TNhttpsdoiorg103334CDIACluendp0172006 2006

Giglio L Randerson J T van der Werf G R Kasibhatla PS Collatz G J Morton D C and DeFries R S Assess-ing variability and long-term trends in burned area by mergingmultiple satellite fire products Biogeosciences 7 1171ndash1186httpsdoiorg105194bg-7-1171-2010 2010

Gleckler P J Doutriaux C Durack P J Taylor KE Zhang Y Williams D N and Servonnat J Amore powerful reality test for climate models EOS97 20ndash24 available at httpseosorgscience-updatesa-more-powerful-reality-test-for-climate-models (last access1 April 2021) 2016

Global Monitoring Laboratory Global monitoring Laboratory ndashcarbon cycle greenhouse gases available at httpswwwesrlnoaagovgmdccggtrends (last access 1 April 2021) 2005

Global Soil Data Task Group Global Gridded Surfaces of Se-lected Soil Characteristics (IGBP-DIS) Tech Rep available athttpsdoiorg103334ORNLDAAC569 2002

GLOBAL VIEW-CO2 Cooperative Global Atmospheric Data In-tegration Project updated annually Multi-laboratory compila-tion of synchronized and gap-filled atmospheric carbon diox-ide records for the period 1979ndash2012 NOAA Boulder COhttpsdoiorg103334OBSPACK1002 2013

Goll D S Brovkin V Liski J Raddatz T Thum T and Todd-Brown K E Strong dependence of CO2 emissions from an-thropogenic land cover change on initial land cover and soil car-bon parametrization Global Biogeochem Cy 29 1511ndash1523httpsdoiorg1010022014GB004988 2015

Goll D S Winkler A J Raddatz T Dong N Prentice IC Ciais P and Brovkin V Carbonndashnitrogen interactionsin idealized simulations with JSBACH (version 310) GeosciModel Dev 10 2009ndash2030 httpsdoiorg105194gmd-10-2009-2017 2017

Green J K Seneviratne S I Berg A M Findell K L Hage-mann S Lawrence D M and Gentine P Large influence ofsoil moisture on long-term terrestrial carbon uptake Nature 565476ndash479 httpsdoiorg101038s41586-018-0848-x 2019

Gruber N and Galloway J N An Earth-system perspec-tive of the global nitrogen cycle Nature 451 293ndash296httpsdoiorg101038nature06592 2008

Gulden L E Rosero E Yang Z L Wagener T andNiu G Y Model performance model robustness andmodel fitness scores A new method for identifying goodland-surface models Geophys Res Lett 35 L11404httpsdoiorg1010292008GL033721 2008

Hajima T Watanabe M Yamamoto A Tatebe H Noguchi MA Abe M Ohgaito R Ito A Yamazaki D Okajima H ItoA Takata K Ogochi K Watanabe S and Kawamiya MDevelopment of the MIROC-ES2L Earth system model and theevaluation of biogeochemical processes and feedbacks GeosciModel Dev 13 2197ndash2244 httpsdoiorg105194gmd-13-2197-2020 2020

Harper A B Wiltshire A J Cox P M Friedlingstein PJones C D Mercado L M Sitch S Williams K andDuran-Rojas C Vegetation distribution and terrestrial carbon

cycle in a carbon cycle configuration of JULES46 with newplant functional types Geosci Model Dev 11 2857ndash2873httpsdoiorg105194gmd-11-2857-2018 2018

Hashimoto S Carvalhais N Ito A Migliavacca M NishinaK and Reichstein M Global spatiotemporal distribution of soilrespiration modeled using a global database Biogeosciences 124121ndash4132 httpsdoiorg105194bg-12-4121-2015 2015

He Y Piao S L Li X Y Chen A P and Qin D HGlobal patterns of vegetation carbon use efficiency and their cli-mate drivers deduced from MODIS satellite data and process-based models Agr Forest Meteorol 256ndash257 150ndash 158httpsdoiorg101016jagrformet201803009 2018

Heimann M and Reichstein M Terrestrial ecosystem car-bon dynamics and climate feedbacks Nature 451 289ndash292httpsdoiorg101038nature06591 2008

Herridge D F Peoples M B and Boddey R M Global inputsof biological nitrogen fixation in agricultural systems Plant Soil311 1ndash18 httpsdoiorg101007s11104-008-9668-3 2008

Hoffman F M Randerson J T Arora V K Bao Q Cadule PJi D Jones C D Kawamiya M Khatiwala S Lindsay Kand Wu T Causes and implications of persistent atmosphericcarbon dioxide biases in Earth System Models J Geophys Res-Biogeo 119 141ndash162 httpsdoiorg1010022013JG0023812014

Holland E A Post W M Matthews E Sulzman J M StauferR and Krankina O N A global database of litterfall massand litter pool carbon and nutrients ORNL DAAC availableat httpsdaacornlgovVEGETATIONguidesGlobal_Litter_Carbon_Nutrientshtml (last access 1 April 2021) 2015

Houlton B Z Marklein A R and Bai E Representation of ni-trogen in climate change forecasts Nat Clim Change 5 398ndash401 httpsdoiorg101038nclimate2538 2015

Hourdin F Rio C Grandpeix J-Y Madeleine J-B CheruyF Rochetin N Jam A Musat I Idelkadi A Fairhead LFoujols M-A Mellul L Traore A-K Ghattas J GastineauG Dufresne J-L Boucher O Lefebvre M-P Millour EVignon E Jouaud J Bint Diallo F Bonazzola M and LottF LMDZ6 Improved atmospheric component of the IPSL cou-pled model J Adv Model Earth Sy 12 e2019MS001892httpsdoiorg1010292019MS001892 2020

Hovenden M and Newton P Plant responses toCO2 are a question of time Science 360 263ndash264httpsdoiorg101126scienceaat2481 2018

Hugelius G Bockheim J G Camill P Elberling B GrosseG Harden J W Johnson K Jorgenson T Koven C DKuhry P Michaelson G Mishra U Palmtag J Ping C-LOrsquoDonnell J Schirrmeister L Schuur E A G Sheng YSmith L C Strauss J and Yu Z A new data set for estimatingorganic carbon storage to 3 m depth in soils of the northern cir-cumpolar permafrost region Earth Syst Sci Data 5 393ndash402httpsdoiorg105194essd-5-393-2013 2013

Huntzinger D N Schwalm C Michalak A M Schaefer KKing A W Wei Y Jacobson A Liu S Cook R B PostW M Berthier G Hayes D Huang M Ito A Lei H LuC Mao J Peng C H Peng S Poulter B Riccuito DShi X Tian H Wang W Zeng N Zhao F and Zhu QThe North American Carbon Program Multi-Scale Synthesis andTerrestrial Model Intercomparison Project ndash Part 1 Overview

and experimental design Geosci Model Dev 6 2121ndash2133httpsdoiorg105194gmd-6-2121-2013 2013

Ito A A historical meta-analysis of global terrestrial net primaryproductivity are estimates converging Glob Change Biol 173161ndash3175 httpsdoiorg101111j1365-2486201102450x2011

Ito A and Inatomi M Use of a process-based model for as-sessing the methane budgets of global terrestrial ecosystemsand evaluation of uncertainty Biogeosciences 9 759ndash773httpsdoiorg105194bg-9-759-2012 2012

Ito A Hajima T Lawrence D M Brovkin V Delire CGuenet B Jones C Malyshev S Materia S McDermidS Peano D Pongratz J Robertson E Shevliakova EVuichard N Warlind D Wiltshire A and Ziehn T Soil car-bon sequestration simulated in CMIP6-LUMIP models impli-cations for climatic mitigation Environ Res Lett 15 124061httpsdoiorg1010881748-9326abc912 2020

Joetzjer E Delire C Douville H Ciais P Decharme B Car-rer D Verbeeck H De Weirdt M and Bonal D Improvingthe ISBACC land surface model simulation of water and carbonfluxes and stocks over the Amazon forest Geosci Model Dev8 1709ndash1727 httpsdoiorg105194gmd-8-1709-2015 2015

Jones C D Arora V Friedlingstein P Bopp L Brovkin VDunne J Graven H Hoffman F Ilyina T John J GJung M Kawamiya M Koven C Pongratz J RaddatzT Randerson J T and Zaehle S C4MIP ndash The CoupledClimatendashCarbon Cycle Model Intercomparison Project experi-mental protocol for CMIP6 Geosci Model Dev 9 2853ndash2880httpsdoiorg105194gmd-9-2853-2016 2016

Jonsson A Aringberg J Lindroth A and Jansson M Gas trans-fer rate and CO2 flux between an unproductive lake and the at-mosphere in northern Sweden J Geophys Res-Biogeo 113G04006 httpsdoiorg1010292008JG000688 2008

Jung M Reichstein M and Bondeau A Towards globalempirical upscaling of FLUXNET eddy covariance obser-vations validation of a model tree ensemble approachusing a biosphere model Biogeosciences 6 2001ndash2013httpsdoiorg105194bg-6-2001-2009 2009

Jung M Reichstein M Ciais P Seneviratne S I Sheffield JGoulden M L Bonan G Cescatti A Chen J De Jeu Rand Zhang K Recent decline in the global land evapotranspira-tion trend due to limited moisture supply Nature 467 951ndash954httpsdoiorg101038nature09396 2010

Jung M Reichstein M Margolis H A Cescatti A RichardsonA D Arain M A Arneth A Bernhofer C Bonal D ChenJ Gianelle D Gobron N Kiely G Kutsch W Lasslop GLaw B E Lindroth A Merbold L Montagnani L MoorsE J Papale D Sottocornola M Vaccari F and Williams CGlobal patterns of land-atmosphere fluxes of carbon dioxide la-tent heat and sensible heat derived from eddy covariance satel-lite and meteorological observations J Geophys Res-Biogeo116 G00J07 httpsdoiorg1010292010JG001566 2011

Jung M Reichstein M Schwalm C R Huntingford CSitch S Ahlstroumlm A Arneth A Camps-Valls G CiaisP Friedlingstein P Gans F Ichii K Jain A K KatoE Papale D Poulter B Raduly B Roumldenbeck CTramontana G Viovy N Wang Y-P Weber U Za-ehle S and Zeng N FLUXCOM (RS+METEO) GlobalLand Carbon Fluxes using CRUNCEP climate data FLUX-

COM Data Portal httpswwwdoiorg1017871FLUXCOM_RS_METEO_CRUNCEPv6_1980_2013_v1 2016

Jung M Reichstein M Schwalm C R Huntingford CSitch S Ahlstroumlm A Arneth A Camps-Valls G CiaisP Friedlingstein P Gans F Ichii K Jain A K KatoE Papale D Poulter B Raduly B Roumldenbeck C Tra-montana G Viovy N Wang Y Weber U Zaehle S andZeng N Compensatory water effects link yearly global landCO2 sink changes to temperature Nature 541 516ndash 520httpsdoiorg101038nature20780 2017

Jung M Koirala S Weber U Ichii K Gans F Camps-VallsG Papale D Schwalm C Tramontana G and ReichsteinM The FLUXCOM ensemble of global land-atmosphere en-ergy fluxes Sci Data 6 74 httpsdoiorg101038s41597-019-0076-8 2019

Jung M Schwalm C Migliavacca M Walther S Camps-VallsG Koirala S Anthoni P Besnard S Bodesheim P Carval-hais N Chevallier F Gans F Goll D S Haverd V KoumlhlerP Ichii K Jain A K Liu J Lombardozzi D Nabel J EM S Nelson J A OrsquoSullivan M Pallandt M Papale DPeters W Pongratz J Roumldenbeck C Sitch S TramontanaG Walker A Weber U and Reichstein M Scaling carbonfluxes from eddy covariance sites to globe synthesis and eval-uation of the FLUXCOM approach Biogeosciences 17 1343ndash1365 httpsdoiorg105194bg-17-1343-2020 2020

Kattge J Knorr W Raddatz T and Wirth C Quantifying pho-tosynthetic capacity and its relationship to leaf nitrogen contentfor global-scale terrestrial biosphere models Glob Change Biol15 976ndash991 httpsdoiorg101111j1365-2486200801744x2009

Kellndorfer J Walker W Kirsch K Fiske G Bishop J La-point L Hoppus M and Westfall J NACP abovegroundbiomass and carbon baseline data V2 (NBCD 2000) USA2000 httpsdoiorg103334ORNLDAAC1161 2013

Kindermann G McCallum I Fritz S and ObersteinerM A global forest growing stock biomass and carbonmap based on FAO statistics Silva Fenn 42 387ndash396httpsdoiorg1014214sf244 2008

Kobayashi K and Salam M U Comparing simu-lated and measured values using mean squared de-viation and its components Agron J 92 345ndash352httpsdoiorg102134agronj2000922345x 2000

Koch J Siemann A Stisen S and Sheffield J Spatial val-idation of large-scale land surface models against monthlyland surface temperature patterns using innovative perfor-mance metrics J Geophys Res-Atmos 121 5430ndash5452httpsdoiorg1010022015JD024482 2016

Koven C D Hugelius G Lawrence D M and Wieder WR Higher climatological temperature sensitivity of soil carbonin cold than warm climates Nat Clim Change 7 817ndash822httpsdoiorg101038nclimate3421 2017

Kowalczyk E A Wang Y P Law R M Davies H L McGre-gor J L and Abramowitz G The CSIRO Atmosphere Bio-sphere Land Exchange (CABLE) model for use in climate mod-els and as an offline model CSIRO Marine and Atmospheric Re-search Paper 13 1ndash43 httpwwwcmarcsiroaue-printopenkowalczykea_2006apdf (last access 1 April 2021) 2006

Kowalczyk E A Stevens L Law R M Dix M Wang Y PHarman I N Haynes K Srbinovsky J Pak B and Ziehn

T The land surface model component of ACCESS descriptionand impact on the simulated surface climatology Aust Meteo-rol Oceanogr J 63 65ndash82 httpwwwbomgovaujshessdocs2013kowalczyk_hrespdf (last access 1 April 2021) 2013

Kumar S V Peters-Lidard C D Santanello J Harrison KLiu Y and Shaw M Land surface Verification Toolkit (LVT)ndash a generalized framework for land surface model evaluationGeosci Model Dev 5 869ndash886 httpsdoiorg105194gmd-5-869-2012 2012

Lamarque J-F Bond T C Eyring V Granier C Heil AKlimont Z Lee D Liousse C Mieville A Owen BSchultz M G Shindell D Smith S J Stehfest E VanAardenne J Cooper O R Kainuma M Mahowald NMcConnell J R Naik V Riahi K and van Vuuren DP Historical (1850ndash2000) gridded anthropogenic and biomassburning emissions of reactive gases and aerosols methodol-ogy and application Atmos Chem Phys 10 7017ndash7039httpsdoiorg105194acp-10-7017-2010 2010

Lasslop G Reichstein M Kattge J and Papale D In-fluences of observation errors in eddy flux data on inversemodel parameter estimation Biogeosciences 5 1311ndash1324httpsdoiorg105194bg-5-1311-2008 2008

Lasslop G Reichstein M Papale D Richardson A DArneth A Barr A Stoy P and Wohlfahrt G Separa-tion of net ecosystem exchange into assimilation and res-piration using a light response curve approach Critical is-sues and global evaluation Glob Change Biol 16 187ndash208httpsdoiorg101111j1365-2486200902041x 2010

Law K Stuart A and Zygalakis K Data assimilation A Math-ematical Introduction Texts in Applied Mathematics ChamSwitzerland Springer 141 1ndash242 httpsdoiorg101007978-3-319-20325-6 2015

Lawrence D M Fisher R A Koven C D Oleson K W Swen-son S C Bonan G Collier N Ghimire B van KampenhoutL Kennedy D and Zeng X The Community Land Model ver-sion 5 Description of new features benchmarking and impactof forcing uncertainty J Adv Model Earth Sy 11 4245ndash4287httpsdoiorg1010292018MS001583 2019

Le Queacutereacute C Peters G P Andres R J Andrew R M BodenT A Ciais P Friedlingstein P Houghton R A Marland GMoriarty R Sitch S Tans P Arneth A Arvanitis A BakkerD C E Bopp L Canadell J G Chini L P Doney S CHarper A Harris I House J I Jain A K Jones S D KatoE Keeling R F Klein Goldewijk K Koumlrtzinger A KovenC Lefegravevre N Maignan F Omar A Ono T Park G-HPfeil B Poulter B Raupach M R Regnier P Roumldenbeck CSaito S Schwinger J Segschneider J Stocker B D Taka-hashi T Tilbrook B van Heuven S Viovy N WanninkhofR Wiltshire A and Zaehle S Global carbon budget 2013Earth Syst Sci Data 6 235ndash263 httpsdoiorg105194essd-6-235-2014 2014

Le Queacutereacute C Andrew R M Canadell J G Sitch S Kors-bakken J I Peters G P Manning A C Boden T A TansP P Houghton R A Keeling R F Alin S Andrews O DAnthoni P Barbero L Bopp L Chevallier F Chini L PCiais P Currie K Delire C Doney S C Friedlingstein PGkritzalis T Harris I Hauck J Haverd V Hoppema MKlein Goldewijk K Jain A K Kato E Koumlrtzinger A Land-schuumltzer P Lefegravevre N Lenton A Lienert S Lombardozzi

D Melton J R Metzl N Millero F Monteiro P M SMunro D R Nabel J E M S Nakaoka S OrsquoBrien KOlsen A Omar A M Ono T Pierrot D Poulter B Rouml-denbeck C Salisbury J Schuster U Schwinger J SeacutefeacuterianR Skjelvan I Stocker B D Sutton A J Takahashi T TianH Tilbrook B van der Laan-Luijkx I T van der Werf GR Viovy N Walker A P Wiltshire A J and Zaehle SGlobal Carbon Budget 2016 Earth Syst Sci Data 8 605ndash649httpsdoiorg105194essd-8-605-2016 2016

Le Queacutereacute C Andrew R M Friedlingstein P Sitch S HauckJ Pongratz J Pickers P A Korsbakken J I Peters G PCanadell J G Arneth A Arora V K Barbero L BastosA Bopp L Chevallier F Chini L P Ciais P Doney S CGkritzalis T Goll D S Harris I Haverd V Hoffman F MHoppema M Houghton R A Hurtt G Ilyina T Jain AK Johannessen T Jones C D Kato E Keeling R F Gold-ewijk K K Landschuumltzer P Lefegravevre N Lienert S Liu ZLombardozzi D Metzl N Munro D R Nabel J E M SNakaoka S Neill C Olsen A Ono T Patra P PeregonA Peters W Peylin P Pfeil B Pierrot D Poulter B Re-hder G Resplandy L Robertson E Rocher M RoumldenbeckC Schuster U Schwinger J Seacutefeacuterian R Skjelvan I Stein-hoff T Sutton A Tans P P Tian H Tilbrook B TubielloF N van der Laan-Luijkx I T van der Werf G R Viovy NWalker A P Wiltshire A J Wright R Zaehle S and ZhengB Global Carbon Budget 2018 Earth Syst Sci Data 10 2141ndash2194 httpsdoiorg105194essd-10-2141-2018 2018

Li W Zhang Y Shi X Zhou W Huang A Mu M Qiu Band Ji J Development of land surface model BCC_AVIM20and its preliminary performance in LS3MIPCMIP6 J Meteo-rol Res-Prc 33 851ndash869 httpsdoiorg101007s13351-019-9016-y 2019

Liang J Qi X Souza L and Luo Y Processes reg-ulating progressive nitrogen limitation under elevated car-bon dioxide a meta-analysis Biogeosciences 13 2689ndash2699httpsdoiorg105194bg-13-2689-2016 2016

Liu Y Xiao J Ju W Zhu G Wu X Fan W andZhou Y Satellite-derived LAI products exhibit large discrep-ancies and can lead to substantial uncertainty in simulated car-bon and water fluxes Remote Sens Environ 206 174ndash188httpsdoiorg101016jrse201712024 2018

Liu Y Y Van Dijk A I De Jeu R A Canadell J G Mc-Cabe M F Evans J P and Wang G Recent reversal in lossof global terrestrial biomass Nat Clim Change 5 470ndash474httpsdoiorg101038nclimate2581 2015

Loveland T R Reed B C Brown J F Ohlen D O ZhuZ Yang L W M J and Merchant J W Development of aglobal land cover characteristics database and IGBP DISCoverfrom 1 km AVHRR data Int J Remote Sens 21 1303ndash1330httpsdoiorg101080014311600210191 2000

Lovenduski N S and Bonan G B Reducing uncertainty in pro-jections of terrestrial carbon uptake Environ Res Lett 12044020 httpsdoiorg1010881748-9326aa66b8 2017

Lovett G M Cole J J and Pace M L Is net ecosystem pro-duction equal to ecosystem carbon accumulation Ecosystems9 152ndash155 httpsdoiorg101007s10021-005-0036-3 2006

Mack P E Viewing the Earth The social construction of theLandsat satellite system MIT Press Cambridge Massachusetts

United States available at httpsbooksgooglecabooksid=Pk7WtI2MJPgC (last access 1 May 2021) 1990

Maki T Ikegami M Fujita T Hirahara T Yamada K MoriK Takeuchi A Tsutsumi Y Suda K and Conway T JNew technique to analyse global distributions of CO2 concentra-tions and fluxes from non-processed observational data Tellus B62 797ndash809 httpsdoiorg101111j1600-0889201000488x2010

Malhi Y Aragao L E O Metcalfe D B Paiva R Que-sada C A Almeida S Anderson L Brando P ChamberJ Q da Costa A C L Hutyra L R Oliveira P PatinoS Pyle E Robertson A and Teixeira L Comprehensiveassessment of carbon productivity allocation and storage inthree Amazonian forests Glob Change Biol 15 1255ndash1274httpsdoiorg101111j1365-2486200801780x 2009

Mauritsen T Bader J Becker T Behrens J Bittner MBrokopf R Brovkin V Claussen M Crueger T Esch MFast I Fiedler S Flaumlschner D Gayler V Giorgetta MGoll D S Haak H Hagemann S Hedemann C Hoheneg-ger C Ilyina T Jahns T Jimeneacutez-de-la-Cuesta D JungclausJ Kleinen T Kloster S Kracher D Kinne S Kleberg DLasslop G Kornblueh L Marotzke J Matei D Meraner KMikolajewicz U Modali K Moumlbis B Muumlller W A NabelJ E M S Nam C C W Notz D Nyawira S-S PaulsenH Peters K Pincus R Pohlmann H Pongratz J Popp MRaddatz T J Rast S Redler R Reick C H Rohrschnei-der T Schemann V Schmidt H Schnur R Schulzweida USix K D Stein L Stemmler I Stevens B von Storch J-S Tian F Voigt A Vrese P Wieners K-H WilkenskjeldS Winkler A and Roeckner E Developments in the MPI-M Earth System Model version 12 (MPI-ESM1 2) and its re-sponse to increasing CO2 J Adv Model Earth Sy 11 998ndash1038 httpsdoiorg1010292018MS001400 2019

Mayorga E Seitzinger S P Harrison J A DumontE Beusen A H W Bouwman A F Fekete BM Kroeze C and Van Drecht G Global nutrient ex-port from WaterSheds 2 (NEWS 2) model developmentand implementation Environ Modell Softw 25 837ndash853httpsdoiorg101016jenvsoft201001007 2010

McGroddy M E Daufresne T and Hedin L O Scalingof C N P stoichiometry in forests worldwide Implicationsof terrestrial redfield-type ratios Ecology 85 2390ndash2401httpsdoiorg10189003-0351 2004

Mitchell T D and Jones P D An improved method of con-structing a database of monthly climate observations and as-sociated high-resolution grids Int J Climatol 25 693ndash712httpsdoiorg101002joc1181 2005

Monfreda C Ramankutty N and Foley J A Farmingthe planet Part 2 Geographic distribution of crop ar-eas yields physiological types and net primary productionin the year 2000 Global Biogeochem Cy 22 GB1022httpsdoiorg1010292007GB002947 2008

Mouillot F and Field C B Fire history and the global carbonbudget A 1times 1 fire history reconstruction for the 20th centuryGlob Change Biol 11 398ndash420 httpsdoiorg101111j1365-2486200500920x 2005

Myneni R B Ramakrishna R Nemani R and Running S WEstimation of global leaf area index and absorbed PAR using ra-

diative transfer models IEEE T Geosci Remote 35 1380ndash1393httpsdoiorg10110936649788 1997

NASA LP DAAC MOD17A3 TerraMODIS net primaryproduction yearly L4 global 1 km NASA EOSDIS LandProcesses DAAC USGS Earth Resources Observationand Science (EROS) Center Sioux Falls South Dakotahttpsdoiorg105067ASTERAST_L1T003 2017

Norby R J DeLucia E H Gielen B Calfapietra C Gi-ardina C P King J S and Oren R Forest responseto elevated CO2 is conserved across a broad range ofproductivity P Natl Acad Sci USA 102 18052ndash18056httpsdoiorg101073pnas0509478102 2005

Nowak R S Ellsworth D S and Smith S D Func-tional responses of plants to elevated atmospheric CO2 ndashdo photosynthetic and productivity data from FACE experi-ments support early predictions New Phytol 162 253ndash280httpsdoiorg101111j1469-8137200401033x 2004

Olson D M Dinerstein E Wikramanayake E D BurgessN D Powell G V N Underwood E C Drsquoamico JA Itoua I Strand H E Morrison J C Loucks CJ Allnutt T F Ricketts T H Kura Y Lamoreux JF Wettengel W W Hedao P and Kassem K R Ter-restrial Ecoregions of the World A New Map of Life onEarth BioScience 51 933ndash938 httpsdoiorg1016410006-3568(2001)051[0933TEOTWA]20CO2 2006

Orth R Dutra E Trigo I F and Balsamo G Ad-vancing land surface model development with satellite-basedEarth observations Hydrol Earth Syst Sci 21 2483ndash2495httpsdoiorg105194hess-21-2483-2017 2017

Pastorello G Trotta C Canfora E Chu H Christianson DCheah Y W and Li Y The Fluxnet2015 dataset and the ONE-Flux processing pipeline for eddy covariance data Sci Data 71ndash27 httpsdoiorg101038s41597-020-0534-3 2020

Phillips L B Hansen A J and Flather C H Evaluat-ing the species energy relationship with the newest mea-sures of ecosystem energy NDVI versus MODIS pri-mary production Remote Sens Environ 112 4381ndash4392httpsdoiorg101016jrse200804012 2008

Piao S Liu Q Chen A Janssens I A Fu Y Dai J andZhu X Plant phenology and global climate change Currentprogresses and challenges Glob Change Biol 25 1922ndash1940httpsdoiorg101111gcb14619 2019

Poulter B MacBean N Hartley A Khlystova I Arino OBetts R Bontemps S Boettcher M Brockmann C De-fourny P Hagemann S Herold M Kirches G Lamarche CLederer D Ottleacute C Peters M and Peylin P Plant functionaltype classification for earth system models results from the Eu-ropean Space Agencyrsquos Land Cover Climate Change InitiativeGeosci Model Dev 8 2315ndash2328 httpsdoiorg105194gmd-8-2315-2015 2015

Randerson J T van der Werf G R Giglio L Collatz G Jand Kasibhatla P S Global fire emissions database version41 (GFEDv4) ORNL DAAC Oak Ridge Tennessee USAhttpsdoiorg103334ORNLDAAC1293 2017

Reichler T and Kim J How well do coupled models simu-late todayrsquos climate B Am Meteorol Soc 89 303ndash312httpsdoiorg101175BAMS-89-3-303 2008

Richardson A D Keenan T F Migliavacca M Ryu YSonnentag O and Toomey M Climate change phenol-

ogy and phenological control of vegetation feedbacks tothe climate system Agr Forest Meteorol 169 156ndash173httpsdoiorg101016jagrformet201209012 2013

Richardson A D Hufkens K Milliman T Aubrecht DM Furze M E Seyednasrollah B and Hanson P JEcosystem warming extends vegetation activity but height-ens vulnerability to cold temperatures Nature 560 368ndash371httpsdoiorg101038s41586-018-0399-1 2018

Righi M Andela B Eyring V Lauer A Predoi V SchlundM Vegas-Regidor J Bock L Broumltz B de Mora L Di-blen F Dreyer L Drost N Earnshaw P Hassler BKoldunov N Little B Loosveldt Tomas S and Zimmer-mann K Earth System Model Evaluation Tool (ESMValTool)v20 ndash technical overview Geosci Model Dev 13 1179ndash1199httpsdoiorg105194gmd-13-1179-2020 2020

Rodda S R Thumaty K C Praveen M S S Jha CS and Dadhwal V K Multi-year eddy covariance mea-surements of net ecosystem exchange in tropical dry decid-uous forest of India Agr Forest Meteorol 301 108351httpsdoiorg101016jagrformet2021108351 2021

Saatchi S S Harris N L Brown S Lefsky M Mitchard E TSalas W Zutta B R Buermann W Lewis S L Hagen Sand Morel A Benchmark map of forest carbon stocks in tropi-cal regions across three continents P Natl Acad Sci USA 1089899ndash9904 httpsdoiorg101073pnas1019576108 2011

Santoro M Beaudoin A Beer C Cartus O Fransson J B SHall R J Pathe C Schmullius C Schepaschenko D Shvi-denko A Thurner M and Wegmuumlller U Forest growing stockvolume of the northern hemisphere Spatially explicit estimatesfor 2010 derived from Envisat ASAR Remote Sens Environ168 316ndash334 httpsdoiorg101016jrse201507005 2015

Saugier B Roy J and Mooney H A 23 ndash Estimationsof Global Terrestrial Productivity Converging toward a Sin-gle Number in Physiological Ecology Global TerrestrialProductivity Academic Press San Diego USA 543ndash557httpsdoiorg101016B978-012505290-050024-7 2001

Schlesinger W H Biogeochemistry An analysis of globalchange 2nd edn Academic Press Oxford United Kingdomhttpsdoiorg101017S0016756898231505 1997

Seacutefeacuterian R Nabat P Michou M Saint-Martin D VoldoireA Colin J and Madec G Evaluation of CNRM Earth Sys-tem Model CNRM-ESM2-1 Role of Earth System Processes inPresent-Day and Future Climate J Adv Model Earth Sy 114182ndash4227 httpsdoiorg1010292019MS001791 2019

Seland Oslash Bentsen M Olivieacute D Toniazzo T Gjermundsen AGraff L S Debernard J B Gupta A K He Y-C KirkevaringgA Schwinger J Tjiputra J Aas K S Bethke I Fan YGriesfeller J Grini A Guo C Ilicak M Karset I H HLandgren O Liakka J Moseid K O Nummelin A Spens-berger C Tang H Zhang Z Heinze C Iversen T andSchulz M Overview of the Norwegian Earth System Model(NorESM2) and key climate response of CMIP6 DECK histor-ical and scenario simulations Geosci Model Dev 13 6165ndash6200 httpsdoiorg105194gmd-13-6165-2020 2020

Sellar A A Jones C G Mulcahy J P Tang Y Yool A Wilt-shire A and Zerroukat M UKESM1 Description and evalu-ation of the UK Earth System Model J Adv Model Earth Sy11 4513ndash4558 httpsdoiorg1010292019MS001739 2019

Sheffield J Goteti G and Wood E F Development of a50-year high-resolution global dataset of meteorological forc-ings for land surface modeling J Climate 19 3088ndash3111httpsdoiorg101175JCLI37901 2006

Shevliakova E Malyshev S Martinez-Cano I Milly P C DPacala S W Ginoux P Dunne K A Dunne J P Dupius CFindell K Ghannam K Horowitz L W John J G KnutsonT R Krasting J P Naik V Zadeh N Zeng F and Zeng YThe land component LM4 1 of the GFDL Earth System ModelESM4 1 biophysical and biogeochemical processes and inter-actions with climate J Adv Model Earth Sy 2019MS002040in review 2021

Simard M Pinto N Fisher J B and Baccini A Mapping forestcanopy height globally with spaceborne lidar J Geophys Res-Biogeo 116 G04021 httpsdoiorg1010292011JG0017082011

Swart N C Cole J N S Kharin V V Lazare M ScinoccaJ F Gillett N P Anstey J Arora V Christian J R HannaS Jiao Y Lee W G Majaess F Saenko O A Seiler CSeinen C Shao A Sigmond M Solheim L von Salzen KYang D and Winter B The Canadian Earth System Modelversion 5 (CanESM503) Geosci Model Dev 12 4823ndash4873httpsdoiorg105194gmd-12-4823-2019 2019

Taylor K E Summarizing multiple aspects of model performancein a single diagram J Geophys Res-Atmos 106 7183ndash7192httpsdoiorg1010292000JD900719 2001

Thurner M Beer C Santoro M Carvalhais N Wutzler TSchepaschenko D Shvidenko A Kompter E Ahrens BLevick S R and Schmullius C Carbon stock and density ofnorthern boreal and temperate forests Global Ecol Biogeogr23 297ndash310 httpsdoiorg101111geb12125 2014

Tian H Yang J Lu C Xu R Canadell J G Jackson R B Ar-neth A Chang J Chen G Ciais P Gerber S Ito A HuangY Joos F Lienert S Messina P Olin S Pan S PengC Saikawa E Thompson R L Vuichard N WiniwarterW Zaehle S Zhang B Zhang K and Zhu Q The globalN2O model intercomparison project B Am Meteorol Soc 991231ndash1251 httpsdoiorg101175BAMS-D-17-02121 2018

Todd-Brown K E O Randerson J T Post W M Hoffman FM Tarnocai C Schuur E A G and Allison S D Causesof variation in soil carbon simulations from CMIP5 Earth systemmodels and comparison with observations Biogeosciences 101717ndash1736 httpsdoiorg105194bg-10-1717-2013 2013

Tramontana G Jung M Schwalm C R Ichii K Camps-VallsG Raacuteduly B Reichstein M Arain M A Cescatti A KielyG Merbold L Serrano-Ortiz P Sickert S Wolf S andPapale D Predicting carbon dioxide and energy fluxes acrossglobal FLUXNET sites with regression algorithms Biogeo-sciences 13 4291ndash4313 httpsdoiorg105194bg-13-4291-2016 2016

Tucker C J Fung I Y Keeling C D and Gammon RH Relationship between atmospheric CO2 variations anda satellite-derived vegetation index Nature 319 195ndash199httpsdoiorg101038319195a0 1986

Twine T E Kustas W P Norman J M Cook D R Houser PMeyers T P and Wesely M L Correcting eddy-covarianceflux underestimates over a grassland Agr Forest Meteorol103 279ndash300 httpsdoiorg101016S0168-1923(00)00123-42000

Umair M Kim D Ray R L and Choi M Estimating land sur-face variables and sensitivity analysis for CLM and VIC simu-lations using remote sensing products Sci Total Environ 633470ndash483 httpsdoiorg101016jscitotenv201803138 2018

Vafaei S Soosani J Adeli K Fadaei H Naghavi H PhamT D and Tien Bui D Improving accuracy estimation of For-est Aboveground Biomass based on incorporation of ALOS-2PALSAR-2 and Sentinel-2A imagery and machine learning Acase study of the Hyrcanian forest area (Iran) Remote Sens 10172 httpsdoiorg103390rs10020172 2018

van den Hurk B Kim H Krinner G Seneviratne S I Derk-sen C Oki T Douville H Colin J Ducharne A CheruyF Viovy N Puma M J Wada Y Li W Jia B Alessan-dri A Lawrence D M Weedon G P Ellis R HagemannS Mao J Flanner M G Zampieri M Materia S Law RM and Sheffield J LS3MIP (v10) contribution to CMIP6 theLand Surface Snow and Soil moisture Model IntercomparisonProject ndash aims setup and expected outcome Geosci Model Dev9 2809ndash2832 httpsdoiorg105194gmd-9-2809-2016 2016

van der Werf G R Randerson J T Giglio L van Leeuwen TT Chen Y Rogers B M Mu M van Marle M J E MortonD C Collatz G J Yokelson R J and Kasibhatla P S Globalfire emissions estimates during 1997ndash2016 Earth Syst Sci Data9 697ndash720 httpsdoiorg105194essd-9-697-2017 2017

Verger A Filella I Baret F and Pentildeuelas J Vege-tation baseline phenology from kilometric global LAIsatellite products Remote Sens Environ 178 1ndash14httpsdoiorg101016jrse201602057 2016

Vitousek P M Menge D N Reed S C and Cleveland CC Biological nitrogen fixation rates patterns and ecologicalcontrols in terrestrial ecosystems Philos T R Soc B 36820130119 httpsdoiorg101098rstb20130119 2013

Vuichard N and Papale D Filling the gaps in meteoro-logical continuous data measured at FLUXNET sites withERA-Interim reanalysis Earth Syst Sci Data 7 157ndash171httpsdoiorg105194essd-7-157-2015 2015

Vuichard N Messina P Luyssaert S Guenet B Zaehle SGhattas J Bastrikov V and Peylin P Accounting for car-bon and nitrogen interactions in the global terrestrial ecosystemmodel ORCHIDEE (trunk version rev 4999) multi-scale evalua-tion of gross primary production Geosci Model Dev 12 4751ndash4779 httpsdoiorg105194gmd-12-4751-2019 2019

Waliser D Gleckler P J Ferraro R Taylor K E Ames SBiard J Bosilovich M G Brown O Chepfer H CinquiniL Durack P J Eyring V Mathieu P-P Lee T Pinnock SPotter G L Rixen M Saunders R Schulz J Theacutepaut J-Nand Tuma M Observations for Model Intercomparison Project(Obs4MIPs) status for CMIP6 Geosci Model Dev 13 2945ndash2958 httpsdoiorg105194gmd-13-2945-2020 2020

WCRP CMIP Phase 6 (CMIP6) available at httpswwwwcrp-climateorgwgcm-cmipwgcm-cmip6 (last access 23 Jan-uary 2021) 2020

Wei J Dirmeyer P A Yang Z L and Chen H Effect ofland model ensemble versus coupled model ensemble on thesimulation of precipitation climatology and variability TheorAppl Climatol 134 793ndash800 httpsdoiorg101007s00704-017-2310-7 2018

Wieder W Regridded Harmonized World Soil Databasev12 ORNL DAAC Oak Ridge Tennessee USAhttpsdoiorg103334ORNLDAAC1247 2014

Wieder W R Cleveland C C Smith W K and Todd-Brown K Future productivity and carbon storage limitedby terrestrial nutrient availability Nat Geosci 8 441ndash444httpsdoiorg101038ngeo2413 2015

Williams K E Harper A B Huntingford C Mercado L MMathison C T Falloon P D Cox P M and Kim J Howcan the First ISLSCP Field Experiment contribute to present-day efforts to evaluate water stress in JULESv50 GeosciModel Dev 12 3207ndash3240 httpsdoiorg105194gmd-12-3207-2019 2019

Wu T Lu Y Fang Y Xin X Li L Li W Jie W ZhangJ Liu Y Zhang L Zhang F Zhang Y Wu F Li J ChuM Wang Z Shi X Liu X Wei M Huang A Zhang Yand Liu X The Beijing Climate Center Climate System Model(BCC-CSM) the main progress from CMIP5 to CMIP6 GeosciModel Dev 12 1573ndash1600 httpsdoiorg105194gmd-12-1573-2019 2019

Xiao J Chevallier F Gomez C Guanter L Hicke J A HueteA R and Zhang X Remote sensing of the terrestrial carbon cy-cle A review of advances over 50 years Remote Sens Environ233 111383 httpsdoiorg101016jrse2019111383 2019

Xie X Li A Tan J Lei G Jin H and Zhang Z Uncer-tainty analysis of multiple global GPP datasets in characterizingthe lagged effect of drought on photosynthesis Ecol Indic 113106224 httpsdoiorg101016jecolind2020106224 2020

Xu Z Jiang Y Jia B and Zhou G Elevated-CO2 responseof stomata and its dependence on environmental factors FrontPlant Sci 7 657 httpsdoiorg103389fpls201600657 2016

Yan Y Zhou X Jiang L and Luo Y Effects of carbon turnovertime on terrestrial ecosystem carbon storage Biogeosciences 145441ndash5454 httpsdoiorg105194bg-14-5441-2017 2017

Yoshikawa C Kawamiya M Kato T Yamanaka Y and Mat-suno T Geographical distribution of the feedback between fu-ture climate change and the carbon cycle J Geophys Res-Biogeo 113 G03002 httpsdoiorg1010292007JG0005702008

Zaehle S and Dalmonech D Carbonndashnitrogen interactions onland at global scales current understanding in modelling cli-mate biosphere feedbacks Curr Opin Env Sust 3 311ndash320httpsdoiorg101016jcosust201108008 2011

Zhang Y J Yu G R Yang J Wimberly M C Zhang XZ Tao J Jiang Y B and Zhu J T Climate-driven globalchanges in carbon use efficiency Global Ecol Biogeogr 23144ndash155 httpsdoiorg101111geb12086 2014

Zhang Z Zhang Y Zhang Y Gobron N Frankenberg CWang S and Li Z The potential of satellite FPAR productfor GPP estimation An indirect evaluation using solar-inducedchlorophyll fluorescence Remote Sens Environ 240 111686httpsdoiorg101016jrse2020111686 2020

Zhao M Heinsch F A Nemani R R and Running S W Im-provements of the MODIS terrestrial gross and net primary pro-duction global data set Remote Sens Environ 95 164ndash176httpsdoiorg101016jrse200412011 2005

Zhu Q Castellano M J and Yang G Coupling soil waterprocesses and the nitrogen cycle across spatial scales Poten-tials bottlenecks and solutions Earth-Sci Rev 187 248ndash258httpsdoiorg101016jearscirev201810005 2018

Zhu Z Bi J Pan Y Ganguly S Anav A Xu L Samanta APiao S Nemani R R and Myneni R B Global data sets ofvegetation leaf area index (LAI)3g and fraction of photosynthet-ically active radiation (FPAR)3g derived from global inventorymodeling and mapping studies (GIMMS) normalized differencevegetation index (NDVI3g) for the period 1981 to 2011 RemoteSens 5 927ndash948 httpsdoiorg103390rs5020927 2013

Ziehn T Kattge J Knorr W and Scholze M Improv-ing the predictability of global CO2 assimilation ratesunder climate change Geophys Res Lett 38 L10404httpsdoiorg1010292011GL047182 2011

Ziehn T Chamberlain M A Law R M Lenton A BodmanR W Dix M Stevens L Wang Y P and Srbinovsky JThe Australian Earth System Model ACCESS-ESM15 Journalof Southern Hemisphere Earth Systems Science 70 193ndash214httpsdoiorg101071ES19035 2020

Abstract

Introduction

Participant methods of validating terrestrial biogeochemical cycles

Variables included in validations

Reference datasets

Statistical metrics of model performance

Community methods of validating terrestrial biogeochemical cycles

Critique of validation approaches

Variable choice

Reference datasets

Statistical metrics and validation approaches

Moving forward

Conclusion

Appendix A Technical summary of validation activities by participants

Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma

Appendix A4 Climate and Global Dynamics Laboratory (NCAR)

Appendix A5 CNRM and CERFACS

Appendix A6 IPSL

Appendix A7 GFDL

Appendix A8 JAMSTEC University of Tokyo and National Institute for Environmental Studies

Appendix A9 MPI

Appendix A10 NCC

Appendix A11 NERC and Met Office

Data availability

Author contributions

Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

uncertainty in projected terrestrial carbon uptake based upon12 Coupled Model Intercomparison Project phase 5 (CMIP5)ESMs indicated that uncertainty stemming from model struc-ture may be 4 times greater than uncertainty from differentemission scenarios and internal variability (Lovenduski andBonan 2017) Some progress has been made in addressingthe large uncertainty associated with the terrestrial biogeo-chemistry in ESMs as comparison of the carbonndashclimate andcarbonndashconcentration feedback among ESMs participatingin the sixth phase of CMIP (CMIP6) by Arora et al (2020)shows a reduced model spread amongst models that includeda nitrogen cycle which provided a realistic constraint onphotosynthesis in the context of elevated atmospheric CO2concentration However the spread in estimated feedback pa-rameters across ESMs overall has not been significantly re-duced from CMIP6 relative to CMIP5 (Arora et al 20202013)

To answer scientific questions regarding climate changethe CMIP was initiated in 1995 by the World Climate Re-search Programmersquos (WCRP) Working Group of CoupledModelling (WCRP 2020) The CMIP designates standardexperimental protocols model output formats and modelforcings to diagnose climate change variability predictabil-ity and uncertainty following various scenarios within amulti-model framework CMIP6 began in 2013 with 3 yearsof planning and community consultation to address knowl-edge gaps prior to the conduction of simulations and anal-yses in 2016 and onwards Model validation in the contextof CMIP consists of demonstrating sufficient agreement be-tween model output data and historical observation-basedreference data following model development and is a cru-cial process in model advancement Such comparison facili-tates model improvement by identifying model limitations inperformance or sources of modelndashdata uncertainty (Loven-duski and Bonan 2017) and informs the weighting of dif-ferent ESMs in influencing climate projections and policy(Eyring et al 2019) CMIP6 specified detailed experimentalprotocols for modelling group participants to facilitate ob-jective comparisons of the output of different models withcommon forcings (Eyring et al 2016a)

Here we focus on validations of the stocks and biologi-cal fluxes of fully coupled ESMs and associated land sur-face model (LSM) releases from 2017 onwards with explicitterrestrial biogeochemical cycle representation contributedby participating CMIP6 modelling groups (hereafter partici-pants Table 1 Arora et al 2020) Validations are analyzed interms of variables included spatiotemporal scales referencedatasets and metrics of performance Section 2 compares themethods of historical terrestrial biogeochemical cycle valida-tion used by participants Section 3 summarizes the methodsused in community analyses of CMIP5 era models and pro-vides a critique of these methods A future outlook is pre-sented in Sect 4

2 Participant methods of validating terrestrialbiogeochemical cycles

To participate in CMIP6 participants had to submit four Di-agnosis Validation and Characterization of Klima (DECK)experimental simulations which included a control simula-tion with prescribed idealized pre-industrial (1850) forcingfor at least 500 years to demonstrate stability in global cli-mate and biogeochemical exchanges Additionally partic-ipants had to conduct historical simulations from 1850 to2014 using designated CMIP6 forcings (available at httpsesgf-nodellnlgovsearchinput4MIPs last access 8 Febru-ary 2021) and initialization from the pre-industrial forc-ing control run (Eyring et al 2016a) Each modellinggroup demonstrated stability in the global carbon cycle withglobal net carbon exchange below the suggested limit of01 Pg C yrminus1 by Jones et al (2016) while no suitable pre-industrial simulation global nitrogen or phosphorus flux wasspecified for CMIP6 though these were generally below20 Pg yrminus1 (Ziehn et al 2020) Each modelling group val-idated terrestrial biogeochemical cycle components for thehistorical simulation in a unique fashion which is summa-rized below and detailed in Appendix A

21 Variables included in validations

The number of terrestrial biogeochemical cycle variablesevaluated against observation-based estimates by partici-pants varied considerably (from 0 to 21) with a total of38 unique variables evaluated by all participants combinedThe variable validated most often was gross primary produc-tion (GPP) which was validated by all but one participantThe next nine most validated variables in descending orderwere soil carbon the global land carbon sink leaf area in-dex (LAI) vegetation carbon ecosystem respiration globallandndashatmosphere CO2 flux surface CO2 concentrations to-tal biomass and burned area (Fig 1) For a list of variabledefinitions see Table 2

The majority of variables were validated by just one or twoparticipants (Fig 2) Danabasoglu et al (2020) and Lawrenceet al (2019) validated a relatively extensive suite of vari-ables with the International Land Model Benchmarking (IL-AMB) package version 21 (ILAMBv21 Collier et al 2018Fig 3) including an explicit uncertainty analysis of the influ-ences of interannual variability forcing datasets and modelstructure in the form of prescribed versus prognostic vegeta-tion phenology While no nitrogen cycle variable was val-idated by more than one group soil N2O flux and totalN2O emissions were evaluated by Hajima et al (2020) andLawrence et al (2019) respectively

A variety of spatiotemporal scales of these variables wereconsidered in validations both within and among partici-pants Spatial scales consisted of site-level model grid celldegree of latitude region and global scales with the lat-ter being the most common across participants Temporal

BCC-CSM2-MR ndash

41 Variable choice

VC Gibbs 2006 ndash

44 Moving forward

5 Conclusion

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

BCC-CSM2-MR ndash

41 Variable choice

VC Gibbs 2006 ndash

44 Moving forward

5 Conclusion

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

BCC-CSM2-MR ndash

41 Variable choice

VC Gibbs 2006 ndash

44 Moving forward

5 Conclusion

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

BCC-CSM2-MR ndash

41 Variable choice

VC Gibbs 2006 ndash

44 Moving forward

5 Conclusion

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

BCC-CSM2-MR ndash

41 Variable choice

VC Gibbs 2006 ndash

44 Moving forward

5 Conclusion

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

BCC-CSM2-MR ndash

41 Variable choice

VC Gibbs 2006 ndash

44 Moving forward

5 Conclusion

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

41 Variable choice

VC Gibbs 2006 ndash

44 Moving forward

5 Conclusion

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

VC Gibbs 2006 ndash

44 Moving forward

5 Conclusion

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

44 Moving forward

5 Conclusion

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

44 Moving forward

5 Conclusion

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

5 Conclusion

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

5 Conclusion

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

A1 CSIRO

A2 BCC

A3 CCCma

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

A5 CNRM and CERFACS

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

A6 IPSL

A7 GFDL

A9 MPI

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

A10 NCC

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

Abstract

Introduction


Reference datasets




Variable choice

Reference datasets


Moving forward

Conclusion


Appendix A1 CSIRO

Appendix A2 BCC

Appendix A3 CCCma



Appendix A6 IPSL

Appendix A7 GFDL


Appendix A9 MPI

Appendix A10 NCC


Data availability


Competing interests

Disclaimer

Acknowledgements

Financial support

Review statement

References

Documents

Validation of terrestrial biogeochemistry in CMIP6 Earth