Environmental Research Letters

LETTER • OPEN ACCESS

Asymmetric dynamical ocean responses in warming icehouse andcooling greenhouse climatesTo cite this article: Karin F Kvale et al 2018 Environ. Res. Lett. 13 125011

 

View the article online for updates and enhancements.

This content was downloaded from IP address 134.245.215.187 on 07/01/2019 at 10:58

Environ. Res. Lett. 13 (2018) 125011 https://doi.org/10.1088/1748-9326/aaedc3

LETTER

Asymmetric dynamical ocean responses in warming icehouse andcooling greenhouse climates

Karin FKvale1,5 , Katherine ETurner1,4, David PKeller1 andKatrin JMeissner2,3

1 GEOMARHelmholtz Centre forOceanResearch,West shore campus, DuesternbrookerWay 20,D-24105Kiel, Germany2 Climate Change ResearchCentre, Level 4Mathews Building, UNSW, Sydney, NSW,Australia3 ARCCentre of Excellence forClimate Extremes, Australia4 Present address: Department of Earth, Ocean, and Ecological Sciences, University of Liverpool, Nicholson Building, 4 Brownlow St.,

LiverpoolMerseyside, L69 3GP,United Kingdom.5 Author towhomany correspondence should be addressed.

E-mail: kkvale@geomar.de

Keywords: ocean dynamics, greenhouse, cooling climate, warming climate, icehouse, ocean circulation, ocean heat release

AbstractWarmperiods in Earth’s history tend to coolmore slowly than cool periods warm.Herewe exploreinitial differences in how the global ocean takes up and gives up heat and carbon in forced rapidwarming and cooling climate scenarios.We force an intermediate-complexity earth systemmodelusing two atmospheric CO2 scenarios. A ramp-up (1%per year increase in atmospheric CO2 for150 years) starts from an average global CO2 concentration of 285 ppm to represent warming of anicehouse climate. A ramp-down (1%per year decrease in atmospheric CO2 for 150 years) starts froman average global CO2 concentration of 1257 ppm to represent cooling of a greenhouse climate.Atmospheric CO2 is then held constant in each simulation and themodel is integrated an additional350 years. The ramp-down simulation shows aweaker response of surface air temperature to changesin radiative forcing relative to the ramp-up scenario. This weaker response is due to a relatively largeand fast release of heat from the ocean to the atmosphere. This asymmetry in heat exchange in coolingandwarming scenarios existsmainly because of differences in the response of the ocean circulation toforcing. In the ramp-up, increasing stratification andweakening ofmeridional overturningcirculation slows ocean heat and carbon uptake. In the ramp-down, cooling acceleratesmeridionaloverturning and deepens verticalmixing, accelerating the release of heat and carbon stored at depth.Though idealized, our experiments offer insight into differences in ocean dynamics in icehouse andgreenhouse climate transitions.

1. Introduction

The characteristics of ocean water masses regulate theexchange of heat and carbon between the ocean and theatmosphere. These characteristics include water massinitial heat and carbon inventories (Winton et al 2013,Ödalen et al 2018) and capacity for additional storage(Xie and Vallis 2012, Ödalen et al 2018), which is setby circulation (Banks and Gregory 2006, Xie andVallis 2012, Winton et al 2010, 2013, Ödalen et al 2018)and the relative strength of physical and biological oceancarbon pumps (Ödalen et al 2018). Ocean/atmosphereheat and carbon exchange is not necessarily coupledspatially (e.g. Frölicher et al 2015) or temporally (e.g.

Garuba et al 2018). The Southern Ocean is presently aprimary region of anthropogenic heat and carbonuptake, whereas anthropogenic carbon storage is spreadacross lower latitudes, and anthropogenic heat storage ismore restricted to the upper ocean, and is greatest in theSouthern and Atlantic Oceans (Frölicher et al 2015).Similar spatial patterns have been found in naturalvariability of carbon and heat storage by Thomas et al(2018), who described the variability as a function ofconvective states in the Weddell Sea. Regional changesin ocean circulation brought about by climate warmingmight affect the regional ocean heat uptake, resulting inchanges to the global atmospheric temperaturewarmingrate (Garuba et al2018).

OPEN ACCESS

RECEIVED

4 July 2018

REVISED

10October 2018

ACCEPTED FOR PUBLICATION

2November 2018

PUBLISHED

18December 2018

Original content from thisworkmay be used underthe terms of the CreativeCommonsAttribution 3.0licence.

Any further distribution ofthis workmustmaintainattribution to theauthor(s) and the title ofthework, journal citationandDOI.

© 2018TheAuthor(s). Published by IOPPublishing Ltd

Ocean heat and carbon uptake has relevance tometrics commonly used to assess climate responsesto external forcing. These concepts generally seek tocharacterize the response of global mean temperatureto a change in atmospheric CO2 (IntergovernmentalPanel on Climate Change 2014). Two examples ofmetrics are: equilibrium climate sensitivity, or ECS,which scales equilibrium surface temperatureresponse against a forced change in atmospheric CO2

concentration, and transient climate response, orTCR, which scales transient warming against a forcedchange in atmospheric CO2 concentration. Suchmetrics can be used to inform projections of future cli-mate change (e.g. Meinshausen et al 2008). However,recent work has demonstrated they exhibit initialstate-dependent qualities, particularly over longertimescales. For example, data suggest ECS (definedhere as the global equilibrium surface temperatureresponse for a change in radiative forcing) may havebeen 30%–40% lower than intermediate glacial statesduring fully glacial states, and there is limited evidenceof reduced sensitivity to land ice albedo feedback inother colder-than-modern climates (summarized byvon der Heydt et al 2016). TCR has similarly beenshown to be sensitive to initial climate state (e.g.Weaver et al 2007, He et al 2017), in which oceancirculation state (He et al 2017) (and specifically, thestrength and depth of the Atlantic Meridional Over-turning Circulation; Kostov et al 2014), exerts controlon the modelled ocean heat uptake efficacy (Wintonet al 2010,He et al 2017).

Understandably, themajority of the attention paidto climate metrics and the decoupling of ocean/atmosphere heat and carbon uptake is in warmingmodern and glacial climates. While modern or close-to-modern warming studies have obvious relevance,studies of other climate states also have value forunderstanding heat and carbon dynamics. Here weextend the general discussion of ocean/atmosphereexchange to a comparison of the heat and marine car-bon dynamical responses to rapidly warming an equi-librated icehouse (pre-industrial atmospheric CO2

concentration, strong deep ocean circulation, lowocean suboxia) and rapidly cooling an equilibratedgreenhouse (higher-than-modern atmospheric CO2

concentration, weaker deep ocean circulation, higherocean suboxia) climate, in order to improve our under-standing of the ocean dynamics regulating heat and car-bon storage and exchange. A greenhouse climate isinteresting because presently our icehouse earth iswarming, and is likely committed to future reductionsof Atlantic Meridional Overturning Circulation, polarice, and oceanic oxygen (Intergovernmental Panel onClimate Change 2014). These changes indicate a trans-ition toward a greenhouse climate state, which is possi-bly the most stable (and therefore, the most difficult totransition out of) of the climate states (Kidder andWorsley 2010). We force our model using values foratmospheric CO2 concentration within a range

consistent with the climate literature (modern icehouse,285 ppm; Intergovernmental Panel on ClimateChange 2014, and hypothetical greenhouse within thelikely range of the Early EoceneClimateOptimum, 1257ppm; Anagnostou et al 2016). We employ symmetricincreases and decreases in prescribed atmospheric CO2

concentrations for a direct comparison of two hypothe-tical transitions between climate states, in order to inves-tigate the relative symmetryof the transitions.

2.Model description and experiments

For our experiments we use the University of VictoriaEarth SystemClimateModel (UVic ESCM) version 2.9(Weaver et al 2001, Meissner et al 2003, Eby et al 2009,Schmittner et al 2005b, 2008). It contains an oceanmodel, land and vegetation components, dynamic-thermodynamic sea ice model, and sediments. Theatmosphere is represented by a two-dimensionalenergy-moisture balance model.Winds are prescribedfrom monthly NCAR/NCEP reanalysis data and areadjusted geostrophically to surface pressure changes(Weaver et al 2001). Continental ice sheets are fixed.The UVic ESCM has a horizontal resolution of 3.6°longitude× 1.8° latitude, with 19 vertical levels in theocean.We use an updated NPZD-type marine biogeo-chemicalmodel (Keller et al 2012).

We first integrate the model 20 000 years in twoconfigurations to achieve equilibrium climate states.The first configuration, for the ramp-up experiment(hereafter, WARMING), is equilibrated with anatmospheric CO2 concentration of 285 ppm torepresent an icehouse climate. The second, for theramp-down experiment (hereafter, COOLING), isequilibrated with an atmospheric CO2 concentrationof 1257 ppm to represent a greenhouse climate. Solarand orbital forcing is prescribed at modern levels ineach configuration.

We then force each respective configuration with a1% per year increase (for WARMING) or 1% per yeardecrease (for COOLING) in atmospheric CO2 con-centration for 150 years to reach the other experi-ment’s initial CO2 concentration (figure 1). Eachmodel is then integrated an additional 350 years, keep-ing atmospheric CO2 concentrations fixed at year 150levels. No changes in non-CO2 greenhouse gas for-cings are included in the simulations.

3. Comparison and discussion ofWARMINGandCOOLING results

Experiment WARMING has an initial average surfaceair temperature (SAT) of 14.2 °C and an average oceantemperature of 4.1 °C. Experiment COOLING ismuch warmer, and has an initial average SAT of21.5 °C and an average ocean temperature of 10.2 °C.Initial COOLING sea ice is restricted to patches in theArctic Ocean and the Ross and Weddell Seas (not

2

Environ. Res. Lett. 13 (2018) 125011

shown). Initial ocean convection in COOLING com-pared to WARMING is stronger in the Ross andWeddell Seas (with annual mean ventilation depths—defined as the depth of potential instability—exceed-ing 400 m versus 100 m), and in the North Atlanticregion it is more restricted to the Nordic Seas (notshown). The initial greenhouse ocean in COOLING isbetter ventilated than the initial icehouse ocean inWARMING, as evidenced by a higher average radio-carbon value (−77.2 versus −129.9 per mil). A moreglobally ventilated greenhouse ocean is produced bythe strongerwind forcing in this configuration.

3.1. Global climate responses to CO2 forcingFigure 2 shows the global mean responses in WARM-ING and COOLING to the CO2 forcings. SATincreases with increasing atmospheric CO2 concentra-tions in WARMING, and continues to increase afterCO2 is held constant, because of inertia in the climatesystem (Eby et al 2009). Likewise, SAT decreases withdecreasing atmospheric CO2 concentrations inCOOLING. Gaps between end points in WARMINGand COOLING indicate neither simulated SAT has yetreached equilibriumwith the final CO2 concentration.Warming during WARMING is about 1 °C greaterthan cooling duringCOOLINGwithin the 500 years ofintegration, and is reflected in the higher WARMINGproportion of surface warming to radiative forcing(shown here as ΔSAT/ΔR, where R stands forradiative forcing). Previous work has demonstrated anonlinear transient climate response to cumulativeemissions relationship in the negative phase of CO2

removal scenarios (Zickfeld et al 2016). This non-linearity occurs because of the inertia of the ocean, sothe ocean continues to take up heat and carbon afteratmospheric CO2 concentrations have started todecline (Zickfeld et al 2016). In our simulations, the

cumulative diagnosed emissions for WARMING is4305 Pg C over the 500 year integration, and−5472 PgC over the 500 year integration in COOLING. Oursimulations start from equilibrium and are thereforeaffected by a different mechanism (noted but notdescribed by Zickfeld et al (2016)), which we describein detail in the following.

Differences in the proportion of surface warmingto radiative forcing inWARMING and COOLING areproduced by asymmetric ocean heat responses,brought about by differing ocean dynamics describedin the next section. During COOLING, the magnitudeof the oceanic release of heat and carbon (shown asnegative fluxes in figure 2) is larger than themagnitudeof uptake by theWARMINGocean. In COOLING, therate of release also accelerates throughout the period oftransient CO2, whereas during WARMING the ocea-nic uptake of heat and carbon decelerates. The releaseof ocean heat to the atmosphere in COOLING has awarming effect on SAT, while atmospheric CO2 con-centrations are forced to decline in our simulations,regardless of oceanic carbon outgassing. Similarly, theuptake of heat by the ocean inWARMING has a cool-ing effect on SAT. In both scenarios the exchange ofheat works to damp the response. However, this is notthe case for carbon exchange, which is decoupledbecause of prescribed atmospheric CO2 concentra-tions and therefore has no impact on the radiative for-cing. The large-scale ocean dynamics are describedbelow.

3.2. Response of ocean circulation toCO2 forcingDecelerating global heat and carbon fluxes inWARM-INGcanbe partly explained by a reduction in northernhemisphere overturning over the course of the integra-tion (upper right panel of figure 2), which reduces air-sea gas and heat exchange. Increasing stratification

Figure 1.Atmospheric CO2 forcing used forWARMING (dark blue) andCOOLING (light blue) experiments.

3

Environ. Res. Lett. 13 (2018) 125011

and a reduction of overturning circulation withincreasing atmospheric CO2 concentrations is a com-mon result in model studies using modern boundary

conditions, e.g. Sarmiento et al (1998), Bopp et al(2005), Schmittner et al (2005a), Weaver et al (2007),Schmittner et al (2008), Steinacher et al (2010).

Figure 2.Global responses in theWARMING (dark blue) andCOOLING (light blue) experiments: global average SATplotted againstatmospheric CO2 concentrations (top left), global average proportion of surface warming to radiative forcingΔSAT/ΔR (topmiddle), maximumoverturning in theAtlantic (top right), global average ocean heat uptake plotted against atmospheric CO2

concentrations (bottom left), global average ocean carbon uptake plotted against atmospheric CO2 concentrations (bottommiddle),andmaximumAABWoverturning (bottom right). Negativeflux values indicate flux out of the ocean and into the atmosphere andnegative overturning values indicate anti-clockwise rotation. Increasing time is indicated by arrows on theCO2-dependent plots.

Figure 3. Initial and year 150meridional overturning circulation forWARMING (left columns) andCOOLING (right columns).Contours denote intervals of 2 Sv.Negative values are shown as dotted lines and represent anti-clockwise rotation.

4

Environ. Res. Lett. 13 (2018) 125011

Accelerating global heat and carbon fluxes in COOL-ING can likewise be partly explained by a strengthen-ing of overturning in both northern and southernhemispheres (right panels offigure 2).

Basin-averaged plots of meridional overturning infigure 3 show the effect of weakening/strengtheningconvection inWARMINGand COOLING. InWARM-ING, North Atlantic Deep Water (NADW) formationweakens and shoals and Antarctic Bottom Water(AABW) formation nearly collapses, similar to otherpreviously published increasing atmospheric CO2 con-centration experiments using the UVic ESCM (e.g.Weaver et al 2007, Schmittner et al 2008). In COOL-ING, initial overturning plots display a maximumNorth Atlantic meridional overturning strength of18 Sv that extends to 3000m depth (globally averaged),and a maximum AABW overturning strength of 12 Sv.As atmospheric CO2 levels decrease, the NADW cellstrengthens from18 to 38 Sv and extends downwards toreach the vertical extent of theNorthAtlantic. AABW inthe Indo–Pacific also strengthens, reaching 26 Sv at theend of the transient CO2 forcing, and extends north-ward to most of the basin. In both hemispheres,strengthening of the overturning during COOLING isachieved by a surface cooling and salinification in thepolar deep water formation regions over the course ofthe integration (not shown). The strongly asymmetricresponse of the ocean circulation in WARMING andCOOLING is due to the different initial states and thedifferent CO2 forcing. In WARMING, the cool oceanwarms from the surface, which increases stratification

and reduces overturning. In COOLING, the relativelywarmer ocean is cooled from the surface, which tem-porarily accelerates overturning.

3.3. Response of ocean heat and carbon storage tochanges in circulationThermal and carbon anomalies in WARMING andCOOLING oceans (figures 4 and 5, where DIC standsfor dissolved inorganic carbon) are primarily a pro-duct of both the changing atmospheric CO2 concen-trations as well as the response of the oceancirculation. This is suggested by the fact that at a basinscale, the DIC responses well match the concurrentresponses in radiocarbon (of which the atmosphericconcentration is held at 0, figure 6). In WARMING,the rising heat content is limited to the upper half ofthe global ocean. This has a damping effect on air-seafluxes, reducing uptake rates. Carbon anomalies arestrongest in the upper half of the Atlantic. This isbecause AABW continues to form, albeit more slowly,therefore overall raising the DIC content. The Pacificdevelops a thermal anomaly profile similar to theAtlantic due to the surface North Pacific taking up therising atmospheric heat while the deep water circula-tion slows. The strongest DIC anomalies occur in thesubtropical gyres. Strong sub-surface warming occursin the Indian Ocean as heat is advected by the currentsmovingwest out of the central-western Pacific. Surfacetemperature anomalies are strongest in the highlatitudes (due to ice-albedo feedback) and westernbasins (regions dominated by fast poleward surface

Figure 4.Changes inWARMING (left columns) andCOOLING (right columns) zonal average ocean temperature for theAtlantic,Pacific and Indian basins after 150 and 500 years of simulation.

5

Environ. Res. Lett. 13 (2018) 125011

currents), while surface carbon anomalies are stron-gest in the central basin gyres. Strongest accumulationof carbon in the gyres agrees with the fifth phase of the

Coupled Model Intercomparison Project (CMIP5)multi-model comparison of anthropogenic carbonuptake of Frölicher et al (2015). The gyre regions do

Figure 5.Changes inWARMING (left columns) andCOOLING (right columns) zonal average oceanDIC for the Atlantic, Pacific andIndian basins after 150 and 500 years of simulation.

Figure 6.Changes inWARMING (left columns) andCOOLING (right columns) zonal average ocean background radiocarbon for theAtlantic, Pacific and Indian basins after 150 and 500 years of simulation.

6

Environ. Res. Lett. 13 (2018) 125011

not take up the carbon directly; rather the carbon isadvected laterally from the high latitudes (Frölicheret al 2015). Strong Southern Ocean temperatureanomalies also agree with the intermodel comparisonof Frölicher et al (2015) demonstrating the SouthernOcean as an important anthropogenic heat uptake andstorage location.

In COOLING, thermal anomalies roughly mirrorthose found inWARMING, but carbon anomalies dif-fer substantially. Deepening of Atlantic meridionaloverturning reduces the separation of intermediateand deep water masses, warming and ventilating thedeep ocean. This effect is transient, and by year 500the positive thermal anomaly is no longer apparent inthe zonal average. The strongest cooling occurs in theupper half of the Atlantic, due to decreasing SAT andoceanic heat release, as well as the upward mixing ofcooler deep water. However, the stronger ventilationof deepwater produces a carbon anomaly that is stron-gest below 3000 meters. Over the period of equilibra-tion at 1257 ppm atmospheric CO2, a large amount ofcarbon accumulated in the deep ocean.When Atlanticmeridional overturning accelerates, this carbon isquickly brought up into the upper ocean (mitigatingthe effect of dropping atmospheric CO2 concentra-tions on upper ocean DIC concentrations) and even-tually to the surface, where it is released to theatmosphere (mostly between 40° and 60°N). Inthe Pacific, the strongest thermal anomaly forms in thenorthern subsurface, as a shallow clockwise over-turning circulation re-establishes (NPDW). Thisclockwise overturning introduces the dropping atmo-spheric temperature anomaly into the shallow NorthPacific. At year 150, the largest carbon anomalies arenear the surface and in the deep South Pacific, partlydue to these being the first regions to exchange carbonwith the atmosphere (though lateral advection andadjustment of ocean carbon pumps also have a role inresponse; Huiskamp et al 2016, Ödalen et al 2018).Rapidly accelerating anti-clockwise overturning in thedeep Pacific has little effect on temperature until nearthe end of the simulation, but it has a large effect onDIC, as carbon-rich deepwater is flushed upward. Theresult is lowering DIC concentrations in the deepwater formation regions of the Southern Ocean andthe abyssal Pacific, and rising (or stable) concentra-tions in the intermediate and shallowNorth Pacific. Byyear 500, the largest carbon anomalies are found in theabyssal central and North Pacific. A similar pattern inthermal and carbon anomalies is present in the IndianOcean, with accelerating overturning introducing thelargest thermal anomalies (and initially, largest carbonanomalies) to the upper ocean, and eventually, the lar-gest carbon anomalies to the deep ocean. As inWARMING, the largest surface temperature anoma-lies are found in the high latitudes and western

boundary regions, and the largest carbon anomaliesare found in the central gyres.

4. Conclusions

Our simulations demonstrate asymmetry of heat andcarbon ocean uptake and release dynamics in rapidtransitions out of icehouse and greenhouse climates.Strong anomalies of temperature do not alwayscorrespond with strong anomalies of carbon, andmayat times occur in different regions of the ocean;especially apparent in the case of greenhouse climatecooling. In the icehouse warming scenario, the oceanstratifies and meridional overturning reduces. Thisslows the rate of ocean heat and carbon uptake. In thegreenhouse cooling scenario, vertical mixing and deepconvection increase. This flushes deep ocean carboninto the upper layers, where it is released to theatmosphere, but has no effect on radiative forcing inthese simulations. A simultaneous release of near-surface heat to the atmosphere maintains high atmo-spheric temperatures, lowering the proportion ofsurface warming to radiative forcing. In bothWARM-ING and COOLING, atmosphere/ocean exchange inthe Southern Ocean is found to be a primary conduitfor deep ocean adjustment to forcing.

The different ocean dynamics in greenhouse andicehouse climates is one potential mechanism thatmight explain the relatively greater stability of green-house over icehouse climates hypothesized by Kidderand Worsley (2010). Our icehouse-to-greenhousesimulation response bears some resemblance to whatmight be expected from a large and rapid release offossil carbon to the atmosphere by humans. However,our greenhouse-to-icehouse simulation ismore theor-etical because the rate of CO2 draw-down is unreason-ably large given current understanding of naturalcarbon sinks and potential climate engineering meth-ods, i.e. carbon dioxide removal. Also, our model set-up prescribes an atmospheric CO2 reduction despite alarge release of carbon from the ocean (which in thereal world would raise atmospheric CO2 and temper-ature, slow the circulation, and sustain the greenhouseclimate unless this additional carbon was removedfrom the system by somemeans). It would be interest-ing to model a greenhouse-to-icehouse transitionincluding coupled atmospheric CO2, however this isbeyond the scope of the present study, which seeks tocompare two symmetric scenarios.

It is important to mention that the UVic ESCMcontains neither cloud feedbacks, nor dynamic icesheets, and neither aerosol nor dust emissions are con-sidered, all of which influence the real world climateresponse on timescales relevant to our study (recentlydiscussed by Members, PALAEOSENS Project 2012,von der Heydt et al 2016, and Caballero and

7

Environ. Res. Lett. 13 (2018) 125011

Huber 2013). These missing feedbacks could alter sur-face wind stress, precipitation, fresh water runoff fromthe land to the ocean, and surface albedo, all of whichcould alter ocean circulation patterns. Omitting adynamical atmosphere has potentially significant con-sequences for climate sensitivity (summarized by Ull-man and Schmittner 2017). Unfortunately, cloud andaerosol feedbacks are the most uncertain feedbacks ofthe modern climate state, and there are no proxies toreconstruct clouds in past greenhouse climates(Huber 2012). Lastly, the potential for wind stress to bea dominant driver in greenhouse overturning (de Boeret al 2008), and the strong role it plays in our simulatedgreenhouse climate, suggests it would be worthwhileto repeat our simulations with a model including adynamic atmosphere. Despite these deficiencies, ourresults do indicate a strong role for ocean dynamics insetting the transient climate response to a CO2

perturbation.

Acknowledgments

This work was supported by GEOMAR HelmholtzCentre for Ocean Research Kiel and Sonder-forschungsbereich 754. KET and KFK are grateful forcomputing resources provided by GEOMAR and KielUniversity. DPK acknowledges funding received fromthe German Research Foundation’s Priority Program1689 ‘Climate Engineering’ (project CDR-MIA; KE2149/2-1). KJM acknowledges support from theAustralia ResearchCouncil (DP180100048).

ORCID iDs

Karin FKvale https://orcid.org/0000-0001-8043-5431

References

Anagnostou E, John EH, EdgarKM, Foster GL, Ridgwell A,Inglis GN, Pancost RD, LuntD J and Pearson PN2016Changing atmospheric CO2 concentrationwas the primarydriver of early Cenozoic climateNature 533 380

BanksHTandGregory JM2006Mechanisms of oceanheat uptake ina coupled climatemodel and the implications for tracer basedpredictions of oceanheat uptakeGeophys. Res. Lett.33L07608

Bopp L, AumontO, Cadule P, Alvain S andGehlenM2005Response of diatoms distribution to global warming andpotential implications: a globalmodel studyGeophys. Res.Lett. 32 L19606

Caballero R andHuberM2013 State-dependent climate sensitivityin past warm climates and its implications for future climateprojections Proc. Natl Acad. Sci. 110 14162–7

de Boer AM, Toggweiler J R and SigmanDM2008Atlanticdominance of themeridional overturning circulation J. Phys.Oceanogr. 38 435–50

EbyM, Zickfeld K,Montenegro A, ArcherD,Meissner K J andWeaver A J 2009 Lifetime of anthropogenic climate change:millennial time scales of potential CO2 and surfacetemperature perturbations J. Clim. 22 2501–11

Frölicher T L, Sarmiento J L, PaynterD J, Dunne J P, Krasting J P andWintonM2015Dominance of the southern ocean inanthropogenic carbon and heat uptake inCMIP5modelsJ. Clim. 28 862–86

GarubaOA, Lu J, Liu F and SinghHA2018The active role of theocean in the temporal evolution of climate sensitivityGeophys. Res. Lett. 45 306–15

He J,WintonM,Vecchi G, Jia L andRugensteinM2017Transientclimate sensitivity depends on base climate ocean circulationJ. Clim. 30 1493–504

HuberM2012A sensitivity to historyNat. Geosci. 6 15HuiskampWN,Meissner K J and d’OrgevilleM2016Competition

between ocean carbon pumps in simulations with varyingSouthernHemispherewesterly wind forcingClim.Dyn. 463463–80

Intergovernmental Panel onClimate Change: Climate Change 20132014The Physical Science Basis:WorkingGroup IContributionto the Fifth Assessment Report of the Intergovernmental PanelonClimate Change (Cambridge: CambridgeUniversity Press)(https://doi.org/10.1017/CBO9781107415324)

KellerDP,Oschlies A and EbyM2012Anewmarine ecosystemmodel for theUniversity of Victoria Earth SystemClimateModelGeosci.Model Dev. 5 1195–220

KidderD L andWorsley TR 2010 Phanerozoic Large IgneousProvinces (LIPs), HEATT (Haline Euxinic Acidic ThermalTransgression) episodes, andmass extinctions Palaeogeogr.Palaeoclimatol. Palaeoecol. 295 162–91

Kostov Y, ArmourKC andMarshall J 2014 Impact of the Atlanticmeridional overturning circulation on ocean heat storage andtransient climate changeGeophys. Res. Lett. 41 2108–16

MeinshausenM,Raper S CB andWigley TML2008 EmulatingIPCCAR4 atmosphere-ocean and carbon cyclemodels forprojecting global-mean, hemispheric and land/oceantemperatures:MAGICC6.0Atmos. Chem. Phys. Discuss. 86153–272

Meissner K,Weaver A,MatthewsH andCoxP 2003The role of landsurface dynamics in glacial inception: a studywith theUVicEarth SystemModelClim. Dyn. 21 515–37

Members, PALAEOSENSProject 2012Making sense ofpalaeoclimate sensitivityNature 491 683

ÖdalenM,Nycander J, Oliver K IC, Brodeau L andRidgwell A 2018The influence of the ocean circulation state on ocean carbonstorage andCO2 drawdown potential in an Earth systemmodelBiogeosciences 15 1367–93

Sarmiento J L,Hughes TMC, Stouffer R J andManabe S 1998Simulated response of the ocean carbon cycle toanthropogenic climate warmingNature 393 245

Schmittner A, LatifM and Schneider B 2005aModel projections oftheNorth Atlantic thermohaline circulation for the 21stcentury assessed by observationsGeophys. Res. Lett. 32

Schmittner A,Oschlies A,GiraudX, EbyMand SimmonsH2005bA globalmodel of themarine ecosystem for long-termsimulations: sensitivity to oceanmixing, buoyancy forcing,particle sinking, and dissolved organicmatter cyclingGlob.Biogeochem. Cycles 19GB3004

Schmittner A,Oschlies A,MatthewsHDandGalbraith ED2008Future changes in climate, ocean circulation, ecosystems, andbiogeochemical cycling simulated for a business as usual CO2

emission scenario until year 4000ADGlob. Biogeochem.Cycles 22GB1013

SteinacherM et al 2010 Projected 21st century decrease inmarineproductivity: amulti-model analysisBiogeosciences 7979–1005

Thomas J,WaughD andGnanadesikanA 2018Relationshipbetween ocean carbon and heatmultidecadal variabilityJ. Clim. 31 1467–82

UllmanD J and Schmittner A 2017A cloud feedback emulator(CFE, version 1.0) for an intermediate complexitymodelGeosci.Model Dev. 10 945–58

von derHeydt A S et al 2016 Lessons on climate sensitivity frompastclimate changesCurr. Clim. Change Rep. 2 148–58

8

Environ. Res. Lett. 13 (2018) 125011

Weaver A J et al 2001TheUVic Earth SystemClimateModel:Modeldescription, climatology, and applications to past, presentand future climatesAtmos.-Ocean 39 361–428

Weaver A J, EbyM,KienastM and SaenkoOA2007Response of theAtlanticmeridional overturning circulation to increasingatmospheric CO2: sensitivity tomean climate stateGeophys.Res. Lett. 34 L05708

WintonM,Griffies SM, Samuels B L, Sarmiento J L andFrölicher T L 2013Connecting changing ocean circulationwith changing climate J. Clim. 26 2268–78

WintonM, Takahashi K andHeld IM2010 Importance of oceanheat uptake efficacy to transient climate change J. Clim. 232333–44

Xie P andVallis GK 2012The passive and active nature of oceanheat uptake in idealized climate change experimentsClim.Dyn. 38 667–84

Zickfeld K,MacDougall AH andMatthewsHD2016On theproportionality between global temperature change andcumulative CO2 emissions during periods of net negativeCO2 emissionsEnviron. Res. Lett. 11 055006

9

Environ. Res. Lett. 13 (2018) 125011