C4MIP/CMIP5CoupledCarbonClimateModel

IntercomparisonsProjectNatalieMahowald,Cornell

Presenta=onderivesfrom

• Friedlingsteinetal.,submiCedtoJOC,“CMIP5cliamteprojec=onsanduncertain=esduetocarboncyclefeedbacks”

• Jonesetal.,submiCedtoJOC,“21stcenturycompa=bleCO2emissionsandairbornefrac=onsimuatedbyCMIP5EarthSystemmodelsunder4representa=veconcentra=onpathways.”

• Aroraetal.submiCedtoJOC,“Carbon‐concentra=onandcarbon‐climatefeedbacksinCMIP5EarthSystemmodels”

• HoffmanandRanderson,andhasaworking=tleof"TheCausesandImplica=onsofPersistentAtmosphericCO2BiasesinEarthSystemModels”,inprep

• Mahowaldetal.,inprep• OthersonPCMDIwebsite:16total

Black:concforced

Red:emisforced

Friedlingsteinetal.,submiCed

Land/Atmospherefluxes

Friedlingsteinetal.,submiCed

Ocean/atmospherefluxes

40

716

Figure 9. Model estimate of 2100 w arming re lative to present day (a vera ge, standard de viat ion and 717

full ra nge) f or the concentrat ion dri ven runs from the CMIP5 models (f ul l database avai lable) and 718

from the 7 CM IP5 ESMs analysed here, and for t he emiss ion dri ven runs from the 7 CMIP5 E SMs 719

analysed here and from the CMIP 3/C4MIP emulation using the MAGICC6 m odel . 720

Friedlingsteinetal.,submiCed

37

FIG. 2. Changes in total land carbon store (top) vegetation carbon (bo ttom left) and soil carbon (defined as cSoil + cLit ter; bottom right) for the CMIP5 models. An observationally derived est imate o f net changes (Ar ora et al., 2011) is shown by the vertical pink bar in the top panel.

Jonesetal.,sub

ModelvariabilityaslargeasRCPvariability

39

FIG. 4. Changes in annual oceanic carbon uptake (top) and cumulat ive uptake s ince 1850 (bottom) from the CMIP5 models. An observationally derived est imate o f net changes (Arora et al., 2011; Le Quere, personal communicat ion) is shown by the vertical p ink bar in the bottom panel.

Insomecases,uncertain=esinmodelsaslargeasrcp

Jonesetal.,

40

FIG. 5. Compatible foss il fuel emissions from CMIP5 models for the historical period (black) and the 4 RCP scenario s for the 21 st century (colours). Timeseries o f annual emissions (top panel): the thick solid lines denote the mult i-model mean and the thick dashed l ines the histo rical and RCP scenarios. Ind ividual model es timates are show n in the thin lines. The bo ttom panel shows cumulat ive emissions for historical (1850-2005) and 21 st century (2006 -2100). The left hand bars in each pair show the cumulative emissions from the histo rical reconstruct ion o r from the RCP scenario as generated by IAM models, and the right hand bars the CMIP5 mu lti-model-mean. Black/gr ey circles sho w individual model values.

Le\bar:RCPRightbar:modelmean,dotsarespread

Jonesetal.,

Jonesetal.,submCed

• TheuncertaintyinlandcarbonuptakeduetodifferencesamongmodelsiscomparablewiththespreadofdifferencesamongRCPscenariosandisdueinparttodifferingrepresenta=onofanthropogeniclandusechange.TheCMIP5modelses=matecumula=ve(2006‐2100)fossil‐fuelemissionsof331±117,861±160,1147±124,1783±187PgCrespec=velyfortheRCP2.6,RCP4.5,RCP6.0andRCP8.5scenarios.

Aroraetal.,submiCed

Aroraetal.,

Aroraetal

(a) Future (2048_2052) vs. Contemporary (2006_2010) Atmospheric Carbon Diox i d e

Contempor ar y (2006_2010) Carbon Dio xide Mole Fr action (ppm)

Futu

re (2

048_

2052

) Ca r

bon

Di o

xid e

Mol

e F r

actio

n ( p

pm)

360 365 370 375 380 385 390 395 400 405 410 415 420

460

480

500

520

540

560

5 80

6 00

620

6 40

460

4 80

5 00

5 20

540

5 60

580

6 00

6 20

6 40Obser ved Contempor ary Mole Fr action

Chinese Model (BCC_CSM1.1)Chinese Model (BCC_CSM1.1_M)

Chinese Model(BNU_ESM)

Canadian Model(CanESM2)

Russian Model(INM_CM4)

U.S. Model(CESM1_BGC)

Chinese Model(FGO ALS_S2.0)

U.S. Model (GFDL_ESM2M)

Japanese Model(MIR OC_ESM)

Ger man Model(MPI_ESM_LR)

Japanese Model(MRI_ESM1)

U.K. Model(HadGEM2_ES)

Histor ical + RCP8.5

R2 = 0.804

Hoffmanetal.

Whatiscausingthebias?

• Oceanuptakeistooweak

• In5ofthe8models,thelandsinkovercompensates,andistoostrong

Hoffmanetal.

CarbonCycleBiasConclusions• Tuningthecarboncycletocontemporaryatm.CO2levelswould

reducefutureuncertaintyinagivenRCPscenariobecausecarboncyclebiasesarepersistentondecadal=mescales

• Forthenext50yearsorso,structuralbiasesregula=ngcarbon‐concentra=onfeedbacksandlandusechangeemissionsmaybemoreimportanterrorsourcesthanclimate‐carbonfeedbacks

• Carbon–concentra=onfeedbacksarelinkedwithmanystructuralmodelcomponentsthatdonotchangerapidly:– RatesofSouthernOceanoverturning– Sensi=vityofphotosynthesistoelevatedcarbondioxide– Alloca=onofNPPtowoodandliCer

• Morefundamentally,onland,carboncycleresponse=meslimitedby:– Residence=mesofliCerandsoilcarbon– Life=meoftrees– Timescalesforrecruitmentandestablishmentofspecies

Hoffmanetal.

(b) Zero ! B ias Model Relative to Model CO2 E stimates fo r 2008

Year

CO2

(pp

m)

1850 1875 1900 1925 1950 1975 2000 2025 2050 2075 2100

250

350

450

550

650

750

850

950

1050

250

350

450

550

650

750

850

950

1050

Obser vationsRCP 8.5Multi_Model MeanZero_Bias Model

Hoffmanetal.,inprep.

• Temperatureat2050

• RCP4.5:medium‐lowscenario

• It’swarming

Meanchange

Change/Stddev

Modelconsistency

Mahowaldetal.,inprep

• Someareasgetmoreprecip,someless:somecoherenceoverland

Mahowaldetal.,inprep

• P‐E:someareasmoisture,somenot:notsomuchcoherenceoverland

Mahowaldetal.,inprep

:• ModelsuggestshigherNetprimaryproduc=on(growth),consistently,almosteverywhere

RCP4.5:medium‐lowscenarioIt’swarming

Mahowaldetal.,inprep

Meanchange

Change/Stddev

Modelconsistency

• Abovegroundvegeta=onhigher,almosteverywhere

• ModelspreCyop=mis=c!True?Mostmodelshavetoostrongofcarbonfer=liza=on,probably

• Ifnottrue:morecarbonwillstayinatmosphere.

Mahowaldetal.,inprep

Modelconsistency

Change/Stddev

Meanchange

Aerosols?VerylimitedcomparisoninliteratureofCMIP5(onlyShindellet

al.,ACCMIP;BCinACCMIP)AOD

1900 2000 2100 2200 23000.00

0.05

0.10

0.15

0.20

0.25

CanESM2CCSM4CESM1CSIRO-Mk3-6-0GFDL-ESM2GGISS-E2-HGISS-E2-RHadGEM2-CCHadGEM2-ESIPSL-CM5A-LRIPSL-CM5A-MRMIROC4hMIROC5MIROC-ESMMIROC-ESM-CHEMMRI-CGCM3NorESM1-MNorESM1-ME

historical: solidrcp26: dottedrcp45: dashedrcp60: dash-dotrcp85: dash-dot-dot