Quantifying the inﬂuence of the terrestrial biosphere on …...T. Davies-Barnard et al.: Terrestrial biosphere carbon emissions 1383 2.1 Climate model description The GCM used in

Clim Past 13 1381ndash1401 2017httpsdoiorg105194cp-13-1381-2017copy Author(s) 2017 This work is distributed underthe Creative Commons Attribution 30 License

Quantifying the influence of the terrestrial biosphere onglacialndashinterglacial climate dynamicsTaraka Davies-Barnard12 Andy Ridgwell13 Joy Singarayer4 and Paul Valdes1

1BRIDGE Cabot Institute and School of Geographical Sciences University of Bristol Bristol BS8 1SS UK2College of Engineering Mathematics and Physical Sciences University of Exeter Exeter EX4 4QE UK3Department of Earth Sciences University of California Riverside CA 92521 USA4Department of Meteorology and Centre for Past Climate Change University of Reading PO Box 243Whiteknights Campus Reading RG6 6BB UK

Correspondence to Taraka Davies-Barnard (tdavies-barnardbristolacuk)

Received 27 January 2017 ndash Discussion started 10 February 2017Revised 18 August 2017 ndash Accepted 15 September 2017 ndash Published 26 October 2017

Abstract The terrestrial biosphere is thought to be a keycomponent in the climatic variability seen in the palaeo-record It has a direct impact on surface temperature throughchanges in surface albedo and evapotranspiration (so-calledbiogeophysical effects) and in addition has an important in-direct effect through changes in vegetation and soil carbonstorage (biogeochemical effects) and hence modulates theconcentrations of greenhouse gases in the atmosphere Thebiogeochemical and biogeophysical effects generally haveopposite signs meaning that the terrestrial biosphere couldpotentially have played only a very minor role in the dy-namics of the glacialndashinterglacial cycles of the late Qua-ternary Here we use a fully coupled dynamic atmospherendashoceanndashvegetation general circulation model (GCM) to gen-erate a set of 62 equilibrium simulations spanning the last120 kyr The analysis of these simulations elucidates the rel-ative importance of the biogeophysical versus biogeochem-ical terrestrial biosphere interactions with climate We findthat the biogeophysical effects of vegetation account for upto an additionalminus091 C global mean cooling with regionalcooling as large as minus5 C but with considerable variabilityacross the glacialndashinterglacial cycle By comparison whileopposite in sign our model estimates of the biogeochemi-cal impacts are substantially smaller in magnitude Offlinesimulations show a maximum of +033 C warming due toan increase of 25 ppm above our (pre-industrial) baselineatmospheric CO2 mixing ratio In contrast to shorter (cen-tury) timescale projections of future terrestrial biosphere re-sponse where direct and indirect responses may at times can-

cel out we find that the biogeophysical effects consistentlyand strongly dominate the biogeochemical effect over theinter-glacial cycle On average across the period the terres-trial biosphere has a minus026 C effect on temperature withminus058 C at the Last Glacial Maximum Depending on as-sumptions made about the destination of terrestrial carbonunder ice sheets and where sea level has changed the av-erage terrestrial biosphere contribution over the last 120 kyrcould be as much asminus50 C andminus083 C at the Last GlacialMaximum

1 Introduction

Terrestrial vegetation interacts with the climate in complexways both responding to and impacting climate conditionsand hence creating an important feedback in the Earth system(eg Claussen 2009 Davies-Barnard et al 2014b Harrisonand Prentice 2003 Jahn et al 2005 Matthews et al 2003Pongratz et al 2010) The influence of the terrestrial bio-sphere on climate occurs in two distinct ways Firstly thereare a number of biogeophysical mechanisms such as changesin albedo or evapotranspiration that provide a direct physicalinfluence on surface climate via changes in net solar radiationtransfer infrared loss roughness length latent heat loss andless directly via changes in moisture exchange and hencetransport Climate feedbacks driven by these changes in ter-restrial vegetation have been hypothesised to be partially re-sponsible for some of the major past climate states (eg Brad-shaw et al 2015 Claussen et al 2006 Crucifix and Loutre

Published by Copernicus Publications on behalf of the European Geosciences Union

1382 T Davies-Barnard et al Terrestrial biosphere carbon emissions

2002 de Noblet et al 1996 Zhou et al 2012) with manystudies particularly focussing on the biogeophysical effects atthe Last Glacial Maximum (LGM) (eg Hopcroft and Valdes2014 Jahn et al 2005 Kageyama et al 2012 Orsquoishi andAbe-Ouchi 2013) The second way in which the terrestrialbiosphere can influence climate is via variations in the carbonstored in vegetation and soil This is a crucial component forunderstanding changes in the carbon cycle through the lastglacialndashinterglacial cycle (Montenegro et al 2006) and nu-merous attempts have been made to estimate the total carbonstorage using a range of methods such as inferences frommarine and terrestrial carbon isotopes (eg Shackleton et al1977 Bird et al 1994) databases of pollen (eg Adams andFaure 1998 Crowley 1995) and simple and complex mod-elling (eg Prentice et al 1993 Kaplan et al 2002 Koumlh-ler and Fischer 2004 Brovkin et al 2012 Orsquoishi and Abe-Ouchi 2013) The resulting range of carbon storage changeestimates is from a few hundred to about 1000 PgC (Ciaiset al 2012) One could add to this changes in the weath-ering of soil minerals and hence CO2 uptake from the at-mosphere as well as nutrient particularly phosphate supplyto the ocean and hence changes in the ocean productivityFor simplicity we will not address these further here (exceptto include a basic silicate weathering feedback in our modelanalysis of the impacts of terrestrial carbon storage change)

Simulations of future vegetation changes show that thebiogeochemical aspect can globally be around the same mag-nitude as the biogeophysical effects (eg Davies-Barnardet al 2014b) meaning that there is uncertainty even in thesign of the net feedback with climate change Both biogeo-physical and biogeochemical effects likely also play an im-portant role in past climate change and potentially the samefundamental uncertainty in the sign of the climate feedbackmight arise However model simulations have generally fo-cussed on either the biogeophysical impacts of vegetationchanges (eg Bradshaw et al 2015 Claussen et al 2006Jahn et al 2005 Orsquoishi and Abe-Ouchi 2013 Shellito andSloan 2006) or biogeochemical impacts (eg Kaplan et al2002 Ciais et al 2012) and the question of the overall feed-back on climate rarely addressed although Claussen (2009)argues that the net effect at the LGM is dominated by thebiogeophysical effects

One of the few examples where both have been combinedand the net effects of vegetation on past climate have beenestimated over long time periods is Brovkin et al (2012)They used an Earth system model of intermediate complexity(EMIC) to suggest that the net effect of vegetation is to de-crease global temperatures during the last glacialndashinterglacialcycle But the model used is relatively coarse in resolution(10 in latitude and 51 in longitude) and reduced in physicalprocess complexity This may be important because of the lo-cal and spatially heterogeneous nature of biogeophysical ef-fects and depending on the location of the forest the biogeo-physical and biogeochemical effects of forest change can bevery different (Bonan 2008) For this reason fully coupled

general circulation models (GCMs) are commonly used inquantifying future climate changes to vegetated land surface(Brovkin et al 2013a b Davies-Barnard et al 2015 Davinand de Noblet-Ducoudreacute 2010) The importance of consid-ering both biogeophysical and carbon cycle impacts togetherat a finer scale when assessing the climate impacts of vegeta-tion is illustrated by work quantifying the climate impacts offorest changes Studies have found that deforestation wouldcause local high-latitude cooling (Betts 2000) global warm-ing (Davin and de Noblet-Ducoudreacute 2010) or even slightglobal cooling (Davies-Barnard et al 2014b a) These out-comes are not predictable from looking at the biogeophysicsor terrestrial biogeochemistry alone at coarse resolutions

Here we present the first model analysis using a fully cou-pled dynamic atmospherendashoceanndashvegetation GCM over thelast 120 kyr that quantifies the net effect of vegetation on cli-mate (A prior study ndash Singarayer and Valdes 2010 ndash did nothave dynamic vegetation and hence could not directly evalu-ate the biogeophysical effects) We separate the biogeophys-ical and biogeochemical effects of vegetation to understandthe overall climate effect of vegetation over the last glacialcycle We show that over the whole period the biogeophys-ical is the dominant effect and that the biogeochemical im-pacts may have a lower possible range than typically esti-mated We also highlight how the temporal scale affects thenet impact of terrestrial biosphere changes

2 Methods

We use the GCM HadCM3 to run a series of simulationswith and without dynamic vegetation to provide the biogeo-physical changes and the land carbon changes To look atthe climate impact of those vegetation carbon changes wethen use the GCM terrestrial carbon changes as an input tothe EMIC cGENIE to calculate the resulting change in atmo-spheric CO2 and global temperature

For future climate changes studies the response of atmo-spheric CO2 concentrations (and hence climate) to changesin terrestrial carbon storage can be calculated using the tran-sient response to cumulative emissions (TRCE) approach(Gillett et al 2013) which demonstrated proportionalitybetween carbon emissions and temperature rise (Goodwinet al 2015) We include these estimates for completenessHowever this approach is only valid for relatively rapidchanges On the longer timescales of glacialndashinterglacialchange we need to take into account the full changes inocean carbon chemistry and including the interactions ofocean and atmosphere with the solid Earth (eg weathering)To do this we employ the ldquocGENIErdquo EMIC

Clim Past 13 1381ndash1401 2017 wwwclim-pastnet1313812017

T Davies-Barnard et al Terrestrial biosphere carbon emissions 1383

21 Climate model description

The GCM used in the simulations in this study is the UK MetOffice Hadley Centrersquos HadCM3B-M21a and HadCM3B-M21aD (Valdes et al 2017) Though not from the latestgeneration of climate models HadCM3 remains an exten-sively used model for many research applications around theworld due to its computational efficiency which means thatlong integrations and many ensemble members can be run

HadCM3 is a three-dimensional fully coupled fullydynamic ocean non-flux-adjusted global climate model(Collins et al 2001) The atmosphere component HadAM3has a Cartesian grid with a horizontal resolution of 25times375 19 vertical levels and a time step of 30 min (Popeet al 2000) The ocean and sea-ice component has the samehorizontal resolution as the atmosphere with 20 verticalocean levels

The land surface scheme used for the atmosphere com-ponent of HadCM3 is the Met Office Surface ExchangeScheme MOSES21 (Gregory et al 1994 Cox et al 1999)MOSES can also use an additional vegetation and terrestrialcarbon model TRIFFID (Top-down Representation of Inter-active Foliage and Flora Including Dynamics) (Cox 2001Cox et al 1998) TRIFFID predicts the vegetation basedon plant functional types using a competitive hierarchicalmodel TRIFFID has two modes equilibrium mode whichquickly brings the vegetation cover into equilibrium by run-ning 50 years of TRIFFID for each 5 years of the climatemodel run and dynamic which runs TRIFFID every 10days TRIFFID and MOSES have nine land surface typesfive of which are vegetation broadleaf trees needleleaf treesshrubs C3 grasses and C4 grasses These are known as plantfunctional types (PFTs) and have different leaf area indexlimits and other phenological differences in the model Soilmoisture in the model is represented on four layers of thick-nesses (measured from the top) of 01 025 065 and 25 m(Essery et al 2001)

The soil carbon is a single pool increased by litterfall anddecreased by respiration (Cox 2001 Cox et al 1998) Thesoil respiration is controlled by moisture and temperature andreturns carbon dioxide to the atmosphere unless as is thecase here the atmospheric carbon dioxide is fixed The lit-terfall is an area-weight sum of the litterfall of the five PFTsin each grid cell (Cox 2001 Cox et al 1998) There is nopermafrost component in the model and soil in frozen re-gions is treated the same as in any other

Assessment of the pre-industrial (PI) vegetation cover ofHadCM3 by Valdes et al (2017) shows good agreement withreconstructions of 1800 vegetation

GCM simulations and experimental methodology

The simulations used here are revised versions of those de-scribed in Singarayer and Valdes (2010) who used HadCM3version HadCM3-M1 which has an older surface scheme

(MOSES1) than the MOSES21 used here and no dynamicvegetation Two sets of 62 simulations were performed cov-ering the time period 120ndash0 ka

ndash The first set of 62 simulations used TRIFFID to predictvegetation changes Each individual simulation was ini-tialised from the previous MOSES1 simulations (whichwere run for 600 years) and were then run for a fur-ther 300 years with ldquoequilibriumrdquo TRIFFID and a final300 years with fully dynamic vegetation This set willbe referred to as the Dynamic set

ndash A second set of simulations uses static vegetation basedon the pre-industrial simulation of the dynamic set (ex-trapolated to new land areas using a simple nearest-neighbour algorithm) They are otherwise identical tothe ldquoDynamicrdquo set (see details below) These will be re-ferred to as the ldquoStaticrdquo set

The differences between Dynamic and Static allow us toevaluate the biogeophysical and biogeochemical responsesof terrestrial carbon cycle change

Both sets of simulations are forced with the same changesin orbit greenhouse gases (CO2 CH4 and N2O) and icesheets as in Singarayer and Valdes (2010) except that weuse a revised ice sheet extent and elevation as discussed inSingarayer et al (2011)

We have also added a parameterisation of water transportfrom ocean to ice sheet in order to ensure that ocean salinityis conserved during each simulation In the normal config-uration of HadCM3 salinity is conserved by the numericalscheme but water that accumulates as snow on ice sheets isnot interactively considered A predefined (spatially varying)flux of water is prescribed into the model which minimisesthe salinity drift for the pre-industrial simulation but this isnot normally changed for other time periods In our new pa-rameterisation we continue to add the predefined flux butalso add an additional flux which is spatially uniform buttemporally variable to ensure that the volume integral oceansalinity is relaxed back to its initial value with a relaxationtimescale of 10 years This prevents any spurious long-termdrifts in ocean salinity

Note that this model does not have a closed carbon cycleThere is no representation of carbon in the ocean and ter-restrial carbon changes do not feedback to the atmosphere(since the greenhouse gas forcings are prescribed) Howeverthe carbon that would have returned to the atmosphere canbe inferred from the change in the carbon stores in the soiland vegetation allowing the biogeochemical impact of vege-tation to be understood as well as the biogeophysical From22 ka to the pre-industrial era simulations are run for every1000 years From 80 to 22 ka simulations are run for every2000 years For 120 to 80 ka simulations are run for every4000 years (See grey points in Fig 2 for a representationof the temporal distribution of the 62 simulations) Reported

wwwclim-pastnet1313812017 Clim Past 13 1381ndash1401 2017

Table 1 Terrestrial carbon changes in the Dynamic simulations from PI to LGM These numbers are the global totals of the maps in Fig A4cand e with the specified parts includedexcluded For the vegetation carbon these are the values associated with the vegetation changes shownin Fig A3 For storage values at the LGM see Table A1

Name Carbon storage scenarios Soil C Vegetation C Total Cchange (PgC) change (PgC) change (PgC)

GCI_ELE Carbon under ice sheets released to atmosphere No carbon on ex-panded land area

minus145 minus295 minus440

GCE_ELE Carbon under ice sheets stored under the ice No carbon on expandedland area

+77 minus222 minus146

GCE_ELI Carbon under ice sheets stored under the ice Modelled carbon storageon new land included

+211 minus173 +37

GCI_ELI Carbon under ice sheets released to atmosphere Modelled carbonstorage on new land included

final climatologies are based on the last 30 years of each sim-ulation

22 EMIC description

The cGENIE Earth system model is used to calculate the im-pacts on atmospheric CO2 over the glacial cycle and hencemake a time-varying estimate of the contribution of biogeo-chemical changes to glacialndashinterglacial climate change Themodel is based around a fast energy-balance-based atmo-sphere model coupled to a 3-D ocean circulation componentand dynamicndashthermodynamic sea ice (Edwards and Marsh2005) plus representations of oceanndashatmosphere (Ridgwelland Hargreaves 2007) oceanndashsediment (Ridgwell et al2007) and atmospherendashland (terrestrial weathering) (Col-bourn et al 2013) carbon cycling As employed here thenon-seasonally forced ocean has eight levels and the config-uration and selection of model parameterisations and param-eter values is identical to that described in Lord et al (2016)These choices are made to minimise experiment runtime andprovide maximum traceability (to a previously used and in-depth analysed configuration) respectively

23 cGENIE carbon cycle simulations

The evolution of terrestrial carbon storage simulated byHadCM3 from 120 ka to the pre-industrial era was used toderive a forcing for cGENIE In this we created a contin-uous time series of the carbon flux from the terrestrial bio-sphere by calculating the difference in carbon storage cal-culated at the end of each HadCM3 time slice and then as-suming a linear interpolation between these points For theldquoFullrdquo simulations cGENIE was then run for 120 kyr usingthis forcing and starting from a fully spun-up state of globalcarbon cycling including an initial balance between the rateof silicate rock weathering and volcanic CO2 outgassing (seeLord et al 2016 for details) For the ldquoCarbonaterdquo simu-lations the model was run with just carbonate compensa-tion only as per Ridgwell and Hargreaves (2007) For the

ldquoClosedrdquo simulations there was no weathering or sedimentresponse and hence is just oceanndashatmosphere repartitioningFor the ldquoAirOceanrdquo simulations the carbon remains in the at-mosphere Both the resulting history of atmospheric CO2 andannual mean global surface air temperature were extractedand calculated as anomalies relative to the late Holocene(pre-industrial)

Using the ldquoFullrdquo setup cGENIE simulations were run us-ing four different carbon estimations from the GCM sim-ulations (see Table 1) For ldquoCarbonaterdquo ldquoAirOceanrdquo andldquoClosedrdquo a simulation was run with the GCI_ELE carbonscenario (see Table 1) Therefore seven transient cGENIEsimulations were run in total

It should be noted that we do not attempt to change theboundary conditions required by the cGENIE model dynam-ically through the glacialndashinterglacial cycles namely orbitalparameters planetary albedo and sea level (and ocean salin-ity) These are instead kept fixed at modern (following Lordet al 2016) Hence changes in the sensitivity of atmosphericCO2 to unit CO2 input (or removal) will not be accountedfor We expect such an effect to exist due to for examplethe dependence of the Revelle factor (the sensitivity of dis-solved CO2(aq) to changes in total dissolved inorganic carbonZeebe et al 1999) on both (ocean surface) temperature andatmospheric pCO2 changes in ocean circulation and the ef-ficiency of the biological pump and changes in the carbonatebuffering of ocean chemistry Some of these factors could intheory be imposed (eg changes in ocean surface tempera-tures) but others would require the glacialndashinterglacial dy-namics in both ocean circulation and marine carbon cyclingto be sufficiently accurately represented in the model Thelatter is far beyond what the current state of understandingof glacialndashinterglacial global carbon cycling allows for (Ko-hfeld and Ridgwell 2009) Hence our assumption of fixedlate Holocene boundary conditions will impart a small biasin our estimates of the atmospheric CO2 response but notone that would affect our overall conclusions

In addition in making estimates of the mean global airsurface temperature change corresponding to the projected

minus150 minus100 minus50 0 50 100 150

minus4

0minus

Preminusindustrial

Longitude

minus4

0minus

21 k a

Longitude

minus4

0minus

30 k a

Longitude

minus4

0minus

68 k a

LongitudeL

minus4

0minus

100 k a

Longitude

1 Broadleaf tree 2 Needleleaf tree 3 C3 grass4 C4 grass5 Shrub6 Urban7 Inland water8 Bare soil9 Ice

(a) (b)

(c) (d)

Figure 1 Dominant plant functional type (PFT) for some time periods of interest in the 120 kyr covered by the simulations (a) Pre-industrial(b) 21 ka (c) 30 ka (d) 68 ka and (e) 100 ka Note that the dominant PFT is calculated as the land cover with the highest proportion of covercompared to the other land surface types and does not necessarily indicate the highest or a significant amount of net primary productivity(NPP)

change in atmospheric pCO2 in cGENIE it is important toalso note that the climate sensitivity is effectively prescribed(Edwards and Marsh 2005) In the Lord et al (2016) config-ured used here only sea-ice cover via its associated albedocan provide feedback on climate In the absence of a dynam-ical atmosphere glacialndashinterglacial changes in climate sen-sitivity due to changes in atmospheric circulation and cloudsare not possible Nor do we account for the possible influ-ences of changes in total land surface area (from sea-levelchange) or vegetation cover and distribution However theassumption of an effectively fixed climate sensitivity acrossthe glacialndashinterglacial cycle is unlikely to impart significantbias or unduly affect our overall conclusions

3 Results

31 Results vegetation dynamics

The changes in climate over time affects the vegetation coverin the Dynamic simulations (shown in Fig 1) In generalcooling leads to an equatorward shift in vegetation as thehigh latitudes become covered in ice or otherwise inhos-pitable for significant quantities of vegetation There is alsoexposure of continental shelves providing potential for veg-etation increases At the Last Glacial Maximum (LGM) at21 ka we can see needleleaf trees and shrubs giving way tovery low-productivity grasses in the high latitudes Howeverbecause of the small number of PFTs (five) in this model theshifts may be underestimated as each PFT represents a widerange of vegetation types The shrubs and trees do not have asignificant presence in northern Europe after 100 ka until theclimate ameliorates into the Holocene It is the vegetation

changes shown in Fig 1 and their associated soil changesthat drive the climate feedbacks and other changes describedhereafter

Hoogakker et al (2016) have shown that HadCM3broadly reproduces the known changes in vegetation acrossthe glacialndashinterglacial cycle Hoogakker et al (2016) useHadCM3B-M1 (without dynamic vegetation) and then usethe climate to drive BIOME4 The climate is very similar be-tween HadCM3B-M1 and HadCM3B-M21a used here InHoogakker et al (2016) they ran an offline vegetation modelBIOME4 driven by the climate anomalies from HadCM3Our results from TRIFFID are consistent with the relativechanges although since TRIFFID uses the actual climatefrom the models the vegetation can have biases (eg Aus-tralia has a tendency to be too wet in HadCM3 in the presentday and hence the coupled model has too much vegetation inthis region) However during glacial times there is a decreasein biomass consistent with Hoogakker et al (2016)

Comparison with the BIOME6000 mega-biome maps forLGM (Pickett et al 2004 Prentice and Jolly 2000 Bigelowet al 2003 Harrison et al 2001) dataset shows generalagreement The model has considerable expansion of grassesin Eurasia where BIOME6000 has grassland and dry shrub-land Broadly speaking in both North America shows lit-tle change from the mid-Holocene to LGM One key weak-ness of the model is in western Europe where BIOME6000shows grassland and dry shrubland whereas the model hasshrubs and needleleaf trees South-east Asia shows continuedwarm-temperate temperate and tropical forest where ourmodel simulated broadleaf trees which encompasses all ofthese biomes The BIOME6000 reconstructions show arounda dozen tundra points on and near the Bering land bridge andour model simulates this as C3 grasses which is the closestPFT to tundra Over central Asia our model has extensiveareas where the dominant land surface type is bare soil indi-cating desert or sparse dry vegetation BIOME6000 shows amixture of desert and dry grassshrubland which is generallyin keeping with the low-productivity low-density vegetationindicated in our model simulation

The forest extent in the tropics at the LGM is similar toPI (see Appendix Fig A2 for shifts in vegetation at 21 ka)This is supported by pollen and other data (Maslin et al2012 Anhuf et al 2006 Wang et al 2017) and modelling(Cowling et al 2001) which find that although there is di-minished tropical forest there is still substantial tree coverat the LGM and little sign of widespread grasslands Con-versely the BIOME6000 data find that the tropical rainforestarea was reduced during the LGM (Pickett et al 2004 Pren-tice and Jolly 2000 Bigelow et al 2003 Harrison et al2001) and grasslands expanded as do some modelling stud-ies (Martin Calvo and Prentice 2015 Prentice et al 2011Hoogakker et al 2016) It is interesting to note that in thepresent-day Amazon BIOME6000 shows three points oftropical forest two of savanna two of warm-temperature for-est two of temperate forest and three of dry grassshrubland

at the LGM In our simulations the dominant PFT of thesame area is broadleaf trees For comparison Prentice et al(2011) using a different dynamic global vegetation modelhave tropical forest over the same domain Therefore thereis little indication that where TRIFFID may be inconsistentwith BIOME6000 another model is necessarily significantlybetter

Because of the PFT (rather than biome) approach of TRIF-FID and the limited number of PFTs it is difficult to besure whether trees in the tropics are a tropical rainforest atthe LGM as there are a number of biomes with significantamounts of trees Although there is little change in PFT inthe tropics at the LGM on the margins there are reductions invegetation carbon suggesting a change in vegetation withinthe large margins of the PFTs used in this model

32 Results biogeophysical feedbacks

The biogeophysical impacts of vegetation are calculated bysubtracting the Dynamic simulations from the correspondingStatic simulations We find that vegetation is acting as a posi-tive feedback to the climate enhancing the cooling (Fig 2a)Broadly the Static and Dynamic simulations both agree withan approximation of global temperature over the whole pe-riod (the EPICA dataset halved) (Fig 2a) The Static set gen-erally does better in 70 to 10 ka whereas the Dynamic set iscloser to the EPICA data in the period 110 to 70 ka The bio-geophysical differences between the Static and Dynamic setsalter global annual mean surface temperature by as much asminus091 C (see Fig 2b) Regionally this temperature coolingis up to 5 C (Fig 3)

The albedo changes are in the same location as the vegeta-tion carbon changes as well as the main temperature changes(see Figs 3 and 4) These temperature differences are mainlydriven by reductions in tree cover and its replacement withbare soil or grasses which is a result of the vegetation dy-namics in the model (see Figs 1 and 4) Trees have a loweralbedo and when they are replaced by higher albedo grassesthere is a cooling effect The change in tree fraction betweenthe Static and Dynamic sets is a good predictor of the temper-ature changes (r2

= 079 using a linear model of the globaltemperature and tree anomalies) This is exacerbated by thepresence of snow cover as the snow-covered visible and near-infrared albedo of grasses shrubs and bare soil is higher thanthat of trees (Essery et al 2001) Therefore when trees arereplaced by grasses where there is snow cover for part of theyear there is a larger change in albedo than where there isno snow cover Thus the albedo changes can be seen mainlywhere a change between trees and grasses occurs in an areawith snow cover (see Figs 1 and 5) The exact contributionof the snow as opposed to the no-snow albedo is difficult todisentangle but the influence of this effect is well established(Betts 2000)

The land surface albedo changes caused by the vegetationhave an even stronger correlation with these biogeophysical

minus120 000 minus100 000 minus80 000 minus60 000 minus40 000 minus20 000 0

minus6minus4

minus20

Years BP

rial (

EPICA05 Dynamic Static(a)

minus05

Years BP

Biogeochemical Biogeophysical Net(b)

Figure 2 (a) Global mean annual temperature (degrees Celsius at 15 m) for Static vegetation simulation (orange) Dynamic vegetation(blue) and the EPICA core-inferred temperature data (black) halved (to give an approximation of global temperature over the whole period)Time periods of particular interest are highlighted as filled points 21 30 56 68 and 100 ka (b) Temperature anomaly over time of DynamicndashStatic simulations for biogeochemical temperature effects of the vegetation change calculated with GENIE averaged to the same temporalresolution as the HadCM3 simulations (red) biogeophysical temperature effects of vegetation change (blue) and the net (biogeophysicaland biogeochemical) effect of vegetation on temperature (purple) Grey points show the time points of the HadCM3 simulations

temperature changes (r2= 086) However we can see that

although the forcing is land-based (the dynamic vegetation)significant changes occur in the ocean (see Figs 3 and 4)that drive the resulting temperature changes Ocean-only sur-face albedo anomaly as a determinant of global temperatureanomaly has an r2 of 095 ndash lower only than the r2 of theglobal (land and ocean) surface albedo of 096 By compari-son the r2 of the latent heat anomaly as a predictor of tem-perature anomaly is lower for land ocean and globally thansurface albedo (070 093 and 091 respectively) The otherparts of the energy balance in particular the latent heat sen-sible heat and the net shortwave radiation do not have such aclear relationship with the temperature change (see AppendixFig A4 and compare to Fig 3)

Although the biogeophysical changes cause cooling thereare some minima of biogeophysical temperature change seenat 30 56 and 100 ka (Fig 2 filled symbols) These minima

have an oceanic source and are caused by vegetation interact-ing with thermohaline circulation changes In our new simu-lations we account for the net transport of water from oceanto the ice sheets by a parameterisation that instantaneouslybalances any net accumulation of water on ice This parame-terisation results in fresher ocean conditions during times ofprecession-driven N Hemisphere summer insolation highs(less water is being used to build the ice sheets) The instan-taneous nature of the parameterisation is physically unreal-istic but reductions in accumulation and an increase in ab-lation during precession highs has been seen in fully cou-pled climatendashice sheet EMIC simulations (eg Ganopolskiet al 2010) During weaker accumulation periods the pa-rameterisation results in a freshening of ocean surface watersand a reduction in Atlantic meridional overturning circula-tion (AMOC) strength from 16 to 10ndash12 Sv

minus40

Temperature 21 ka

Longitude

minus10

minus5

minus40

Temperature 30 ka

Longitude

minus10

minus5

minus40

Precipitation 21 ka

Longitude

minus2

minus1

minus40

Precipitation 30 ka

LongitudeLa

minus2

minus1

(a) (b)

(c) (d)

Figure 3 Anomaly of (a) and (b) temperature at 15 m (degrees Celsius) and (c) and (d) precipitation (mmdayminus1) between the Dynamicvegetation simulation and the equivalent Static vegetation simulation For (a c) 21 ka and (b d) 30 ka The pattern of increased temperatureat 30 ka is similar to the pattern at 56 and 100 ka

Superimposed upon this general behaviour the addition ofinteractive vegetation generally does not change the AMOCstrength However at times of weak AMOC small changesin runoff and temperature are sufficient to cause somechanges in the response For instance in the static vegetationsimulations there is a relatively weak AMOC in the simula-tions for 60 58 and 56 ka In the interactive vegetation sim-ulations the weakened AMOC only occurs at 60 ka Thusat 60 ka the changes in climate are fairly typical of preced-ing times but at 58 and 56 ka there is a substantial differ-ence between the static and dynamic vegetation simulationsThe cause for this difference is associated with a combina-tion of reduced runoff into the North Atlantic (principallyfrom changes in land surface in North America) and coldertemperatures both of which act to stabilise the AMOC in allthree periods but it is sufficient to prevent the AMOC weak-ening in the 58 and 56 ka simulations This threshold-likebehaviour of the AMOC is almost certainly highly model-dependent and hence the result is not robust

The regional patterns of cooling also temporarily affectthe precipitation regime (see Fig 3) This appears to be re-lated to the AMOC weakening There are some suggestionsof similar relationships between the increases in precipitationand the terrestrial changes to previous studies (Gedney andValdes 2000 Singarayer et al 2009) Similar to the temper-ature changes it is unclear how model-specific these changesare

33 Results biogeochemistry

We now calculate the total change in terrestrial carbon storesin the HadCM3 simulations We consider scenarios of ter-restrial carbon change with combinations of including or ex-cluding uncertain aspects of the carbon cycle specifically de-pending on the fate of soil carbon under ice and the changesrelated to the expansion of land Zeng (2003) suggested thatthe soil and vegetation carbon formed during the warm lastinterglacial could simply get covered by ice and is storedthere rather than being released into the rest of the systemas is typically assumed in past estimates Equally if the car-bon were exported from under the ice sheets it is not certainthat it would return to atmosphere rather than be sequesteredin riverine and oceanic systems

Similarly the amount of carbon stored on newly emergedland is also uncertain as it depends on both the area of emer-gent land and the surface properties It is uncertain whetherthe vegetation and soil inundated in the deglaciation wouldhave been returned to the atmosphere or remained in situTherefore we calculate the changes in soil and vegetationcarbon from these various sources In Table 1 we focus onthe changes between pre-industrial and LGM which cor-responds to the largest overall change through the glacialndashinterglacial cycle

In the model 222 PgC of soil carbon and 73 PgC of veg-etation carbon is associated with areas covered with ice atthe LGM (see Fig A1) Similarly 134 PgC of soil carbonand 49 PgC of vegetation carbon is associated with new landThe resulting range of total carbon storage is large from a

Years BP

Global Land only Ocean only

minus20

minus15

minus10

minus05

Years BP

nt heat

ndash2)

minus01

2minus0

minus00

Years BP

Figure 4 Mean annual anomaly DynamicndashStatic simulations for (a) surface albedo (b) latent heat (Wmminus2) and (c) tree cover (proportionof land area)

loss of 440 PgC at the LGM (no carbon stored under newice sheets with all being released to the oceanndashatmosphereand no build-up of carbon on new land surface) to a possiblesmall increase in carbon (if carbon is stored under the newice sheets and there is no carbon storage on new land)

In reality glacial systems are known to export carbon ina highly labile form (Lawson et al 2014) erode soil andbedrock creating major landscape changes and release largeamounts of methane when they retreat (Wadham et al 2012)Although the conversion of this terrestrial carbon to atmo-spheric carbon may be through riverine or oceanic systemsit seems likely the majority would return to the atmospherewithin the time periods we consider We therefore use thislargest scenario as a conservative option for our main analy-sis

The other major change to soil carbon in the model isnewly exposed land which is revealed when the water in the

ice sheets causes lower sea levels (see Figs 1 and 5) For thenew land we use a nearest-neighbour interpolation of basicsoil properties (eg water-holding capacity) and the model isrun for sufficient length of time for the soil and vegetationcarbon to reach equilibrium

This estimate of carbon on expanded lands also has uncer-tainties We have reasonable confidence in the sea-level es-timates and consequent change in land area but it is muchmore uncertain whether carbon could accumulate on thatland Uncertainty in the positions and timings of ice sheetsparticularly before the LGM exacerbate this For instancein our simulations the East Siberia ice sheet is absent (seeFig 1) whereas some other ice-sheet reconstructions includeit (eg Niessen et al 2013) The area of the ice sheet aloneaccounts for an average of 56 PgC soil carbon in these sim-ulations when it is absent However soil carbon takes a longtime to accumulate especially with low net primary pro-

(a) Albedo 21 ka

(b) Albedo 30 ka

ampl ampl 04 04

0 m 0 N C N

-02 -02

-04 -04

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

(c) Vegetation C 21 ka (d) Vegetation C 30 ka

ampl ampl 20 20

0 m 0 N

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

(e) Soil C 21 ka (f) Soil C 30 ka

ampl ampl 20 20

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

Figure 5 Mean annual anomaly of (a b) surface albedo (unitless) (c d) vegetation carbon (kgCmminus2) and (e f) soil carbon (kgCmminus2)between the Dynamic vegetation simulation and the equivalent Static vegetation simulation For 21 ka (a c e) and 30 ka (b d f) The patternof reduced surface albedo at 30 ka is similar to the pattern at 56 and 100 ka Since in the Static simulations the carbon remains at PI levelspanels (c)ndash(f) also represent the anomaly to PI The ice sheets are excluded from these plots For the carbon under ice sheets see Fig A4

ductivity (NPP) and vegetation carbon storage averages just05 PgC over all the expanded lands

If exposed land carbon was included and glacial land soilcarbon excluded the terrestrial carbon is a gain from PIto LGM of +37 PgC (see Table 1) However as discussedabove we would argue that excluding glacial land soil car-bon change is probably unreasonable Most previous studieshave also assumed that all carbon under ice is removed If weinclude the loss of carbon then the range in total amount ofterrestrial carbon lost in this model between the PI and theLGM at 21 ka is minus440 to minus257 PgC

The change in terrestrial carbon found in our simulationscontributes to atmospheric carbon dioxide change Using thecGENIE model to approximate the carbon uptake by theocean we therefore calculate the atmospheric carbon dioxidechange (see Methods section and Fig 6)

Selecting the largest change in carbon storage (minus440 PgCat the LGM including glacial soil carbon changes and ex-

minus10

Time (years BP)

GCE_ELE GCE_ELI GCI_ELE GCI_ELI

Figure 6 Contribution of terrestrial biosphere carbon emissions toatmospheric CO2 Run with the cGENIE ldquoFullrdquo configuration andnormalised to pre-industrial CO2 levels The four scenarios are asdetailed in Table 1

cluding expanded lands) we find that the results suggest apeak contribution compared to pre-industrial CO2 of 25 ppmCO2 (Fig 6) In all scenarios except GCE_ELI the terres-

trial carbon contribution to atmospheric CO2 acts as a neg-ative feedback to the climate dampening the effect of otherclimate forcings including the net contribution of the terres-trial biosphere (Fig 2b)

Within cGENIE the change in atmospheric CO2 pro-duces a warming at the LGM of 029 C (equivalent to aclimate sensitivity of around 2 Wmminus2Cminus1 see Fig 2b)This is much smaller than the biogeophysical contributionof minus084 C It is also much less variable For most of theglacial period from 100 to 20 ka the implied biogeochemi-cal warming is around 026 C (Fig 2b) This results in thedominance of the biogeophysical impacts over biogeochem-ical feedbacks

However with different Earth system processes includedthe biogeochemical effects vary substantially (see Fig 7)In the simulations discussed above silicate and carbonateweathering are both included and this results in the lowesttemperature change from the same carbon emissions Thetemperature contribution at the LGM increases relative to thebaseline ldquoFullrdquo simulation (029 C) as the silicate weather-ing is excluded (Carbonate 030 C) all weathering is ex-cluded (Closed 047 C) a decadal- to millennial-scale car-bon uptake is used (TRCE 086 C) and if all carbon re-mains in the atmosphere (AirOcean 192 C) Note that theTRCE as shown above includes the terrestrial biosphere as asink so it will slightly overestimate how much carbon willbe removed from the atmosphere when the source is the nat-ural vegetation Comparing these values to the biogeophys-ical terrestrial effect in Fig 2b we can see that the shorterthe timescale the more likely biogeochemical terrestrial pro-cesses will dominate as it weakens over time On longertimescales the biogeophysics dominate because the scale ofthe effect does not diminish over time relative to the control

4 Discussion

Considering the biogeophysical and biogeochemical effectsof the terrestrial biosphere together we find that the biogeo-physics dominate the signal Over the entire 120 kyr periodconsidered the average temperature contribution by the ter-restrial biosphere is minus026 C At the LGM the contributionof the terrestrial biosphere to temperature is minus058 C How-ever if we use the GCE_ELI scenario where the carbon un-der the ice sheets and on land exposed by sea-level changes isassumed not to return to the atmosphere the equivalent val-ues are minus050 and minus083 C For both the conservative andmore uncertain assumptions it is clear that terrestrial carbonis a much smaller contributor to temperature than biogeo-physics over this period

The biogeophysical results found here broadly concur withcomparable model studies of past vegetation biogeophysicsClaussen et al (2006) found the biogeophysical contributionof vegetation to LGM cooling of around 1 C in the North-ern Hemisphere whereas Jahn et al (2005) found around

120 000 100 000 80 000 60 000 40 000 20 000 0

Time (years BP)

us) Full Carbonate Closed AirOcean TRCE

Figure 7 Temperature changes resulting from the same terrestrialcarbon emissions scenario (GCI_ELE) with different model setupfor cGENIE and for the TRCE of HadCM3 The cGENIE simula-tions were ldquoFullrdquo with silicate weathering feedback and just car-bonate compensation as Lord et al (2016) ldquoCarbonaterdquo with justcarbonate compensation only as Ridgwell and Hargreaves (2007)ldquoClosedrdquo with no weathering or sediment response and hence justoceanndashatmosphere repartitioning and ldquoAirOceanrdquo where the car-bon remains in the atmosphere ldquoTRCErdquo is the simple calculationof the TRCE of HadCM3 (taken from Matthews et al 2009) forthe same carbon inputs into the atmosphere as used for the cGENIEsimulations Note that we include the TRCE for completeness butit is not a cGENIE simulation (see Methods section)

minus06 C and up to 2 C locally Our result of minus084 C isin the middle of the other LGM studies

The dominance of the biogeophysical effects found hereis contrary to the results found for short timescale prob-lems which find that biogeochemistry tends to be compa-rable in magnitude to biogeophysical effects (eg Davies-Barnard et al 2014b Pongratz et al 2010) This is becausethe centennial simulations have a stronger biogeochemicaleffect since the transient response to cumulative emissionsis stronger than the equilibrium response In climate simu-lations up to around a century long more carbon tends toremain in the atmosphere This creates a strong warmingeffect that is approximately linearly related to the amountof greenhouse gas emissions (Matthews et al 2009 Gillettet al 2013) The transient response to cumulative emissions(TRCE) accounts for the uptake of atmospheric carbon by theocean and terrestrial biosphere but only on short timescalesThe uptake of atmospheric carbon by the ocean requires hun-dreds or thousands of years and it is slower when the in-crease in carbon into the system is small and staggered (Lordet al 2016) However the simulations we use are on a mil-lennial timescale allowing much of the carbon to be takenup by the ocean (Lord et al 2016) From a climate sensi-tivity point of view this means that on shorter timescalesthe effects of dynamic vegetation can cancel each other outThis provides some rationale for the fact that dynamic veg-etation has been generally not included in the majority ofstate-of-the-art Earth system models used in CMIP5 as itdoes not significantly affect the climate sensitivity At longertimescales it is more important to include dynamic vegeta-

tion as without the positive feedback of the biogeophysicaleffects the climate sensitivity would be underestimated

For the biogeochemical effects of the terrestrial biosphereprevious estimates of carbon stocks on exposed continentalshelves based on models are between 112 and 323 PgC atthe LGM (Montenegro et al 2006) The comparable numberin this simulation is 183 PgC which is on the lower end ofthe wide range of other models However it has good agree-ment with the vegetation reconstruction (not model) valuesby Montenegro et al (2006) of 182 to 220 PgC

The LGM terrestrial carbon change here is minus440 to+37 PgC including a zero contribution of terrestrial car-bon This is smaller than the values of minus900 to minus400 PgCrange reviewed by Kohfeld and Ridgwell (2009) More re-cent modelling studies are also somewhat larger than ourestimate range such as minus500 PgC (Brovkin et al 2012)minus597 PgC (Orsquoishi and Abe-Ouchi 2013) and minus550 tominus694 PgC (Prentice et al 2011) However recent inventoryand isotope approaches are closer to our range of valueseg minus378plusmn 88 PgC (Menviel et al 2017) and minus330 PgC(Ciais et al 2012)

For present day Hugelius et al (2014) shows around 75ndash100 kgCm2 of soil carbon in far north Siberia and 20ndash40 fur-ther south Far northern Canada is much more heterogeneouswith values from 20 to 150 kgCm2 The modelled PI valuesof northern high-latitude soil carbon are on the low side andmuch more homogeneous around 15ndash20 kgCm2 but similarto Hugelius et al (2014) who show far north America to beless consistent with some higher areas of 35ndash40 kg Cm2 inthe far north (See Appendix Fig A4 of the loss of soil car-bon under ice sheets at the LGM) What this suggests is thatwhile of the correct order of magnitude the model has a verymodest amount of soil carbon that could be considered per-mafrost Therefore we think it is reasonable to include thislow estimate of soil carbon in the uncertainties

The soil carbon change under ice sheets between PI andLGM is modelled as sim220 PgC Extrapolating from a com-parison with Hugelius et al (2014) this might be a third toolittle If the true value were sim330 PgC this would make thetotal C change PI to LGM 550 PgC more in line with someprevious model estimates It would affect the global meanannual biogeochemical contribution by sim 01 K This wouldmean the net effect of vegetation was closer to zero but thebiogeophysical effect would still dominate

However the exact size of the terrestrial carbon emissionsis uncertain Other carbon stores not accounted for here arepotentially important for example methane during sea-levelrises or changes to the wetlands in the tropics Modellingstudies that look at wetlands at the LGM suggest that al-though the wetland area is larger the methane emissionsare lower compared to modern day (Kaplan 2002) How-ever palaeohydrological data indicate a drying in the Africantropics (Gasse 2000) Our model does not have a process-based permafrost or wetlands component and therefore thechanges in methane are not accounted for This is a particular

limitation when considering the carbon stored in deep per-mafrost soils in northern peatlands Saito et al (2013) showthat based on the temperature changes there is a substantialexpansion of permafrost area during glacial times but cannotestimate any changes in carbon storage Zimov et al (20062009) have argued that permafrost storage could be a majorsource of carbon through the deglaciation and Ciais et al(2012) argue that there was a large extra pool of inert carbonat the LGM Similarly Koumlhler et al (2014) have argued thatlarge amounts of carbon were locked into permafrost whichwere then released rapidly at the BoslashllingndashAlleroslashd

Research has also suggested that waterlogging and flood-ing as sea-level rises during the Holocene could cause rapidanaerobic decomposition of vegetation causing methaneemissions (Ridgwell et al 2012) This could account foremissions of as much as 25 PgC for 10 m sea-level rise (ibid)Since our simulations do not account for methane or this ef-fect of inundation it is likely it there is a slight underesti-mation of equivalent CO2 effect of the carbon emissions (asmethane is a stronger greenhouse gas than carbon dioxide)

The impacts are mainly determined by the vegetationshifts the dynamic global vegetation model simulates Eachgrid box has the potential for five PFTs but generally theLotkandashVolterra equations used in TRIFFID mean that thegrid box is dominated by one PFT The small number ofPFTs means that the range within each PFT is relativelylarge Therefore the model probably underestimates the ef-fects of small perturbations in climate as the large definitionof the PFTs allows the PFT to remain the same Converselyit makes an abrupt change more likely as the climate tips agrid box from being predominantly one PFT to being pre-dominantly another Overall the model could be slightly un-derestimating the amount of change in vegetation Howeverbecause of the ratio of the biogeophysical to biogeochemi-cal changes if the vegetation change is underestimated thesign of the net effect of the terrestrial biosphere is unlikelyto change Similarly because over the long time periods in-volved much of the released carbon is taken up by the oceanthe changes in carbon densities of the vegetation would needto be wrong by a lot to change the overall signal

Our approach here assumes that there is no non-linear in-teraction between the biogeochemical and biogeophysical ef-fects Since the biogeochemistry acts as a negative feedbackand reduces over time and the biogeophysics acts as a posi-tive feedback and stays the same over time there is no strongreason to believe that at equilibrium there would be any sig-nificant synergy However on shorter timescales and on a re-gional rather than global scale it is quite possible that therecould be some synergies

5 Conclusions

Using a fully coupled atmospherendashoceanndashvegetation modelwith static and dynamic vegetation we find that over the

last 120 kyr the net effect of vegetation feedbacks on globalannual mean 15 m air temperature is a cooling which canbe as much as minus066 C (Fig 2b) For the vast majority ofthe last glacialndashinterglacial cycle cooling associated withbiogeophysical feedbacks dominates over the biogeochem-ical warming associated with reduced terrestrial carbon stor-age The biogeophysical cooling effect is mainly due tothe role that vegetation plays in changing surface albedoand particularly related to snow cover and the taigandashtundratransition (Gallimore and Kutzbach 1996 de Noblet et al1996) and is a well-understood climate feedback The bio-geochemical contribution to atmospheric carbon dioxide issmall (sim 20 ppmv) and hence the temperature contributionis small (on average 026 C with a maximum of 033 C)There are significant uncertainties in this calculation whichwould further diminish the net temperature impact of the ter-restrial biosphere by cancelling out the biogeophysical im-pact In this analysis the only time periods where the effectsare comparable are at times when additional mechanisms op-erate such as changes in ocean circulation but these mecha-nisms may be model-specific

The key uncertainties in this study originate in the biogeo-chemistry especially the soil carbon build-up in newly ex-posed land the fate of soil carbon in glacial systems and theamount of carbon in permafrost (not calculated in this study)Further research is needed to fully understand the functioningof these systems and how they can be best incorporated intoclimate models In addition the technique we use for infer-ring the biogeochemical effects of terrestrial carbon changeshas limitations and is potentially model-dependent How-ever the smaller estimate of terrestrial carbon emissions maymake the low LGM atmospheric carbon dioxide somewhateasier to reconcile (Montenegro et al 2006)

Our work confirms previous results using EMICs (Brovkinet al 2012) that found the net terrestrial biosphere effect tobe primarily biogeophysical and that the terrestrial carboncontribution to atmospheric carbon is comparatively smallOur findings also represent a clear illustration of the netclimatic effect of vegetation being highly dependent on thetimescale with the biogeophysical response dominating inthe longer term in contrast to century-scale future changes

Code and data availability The model code is currentlyavailable to view at httpcmsncasacukcode_browsersUM45UMbrowserindexhtml The GCM simulation data are avail-able at httpwwwpaleobrisacukummodelscriptspapersDavies-Barnard_et_al_2017html

Appendix A

Table A1 Terrestrial carbon storage at the LGM

Carbon storage scenarios Soil C (PgC) Vegetation C (PgC) Total C (PgC)

GCI_ELE Carbon under ice sheets released to atmo-sphere No carbon on expanded land area

893 430 1323

GCE_ELE Carbon under ice sheets remains stored un-der the ice No carbon on expanded land area

1114 502 1617

GCE_ELI Carbon under ice sheets remains stored un-der the ice Modelled carbon storage on new land in-cluded

1249 552 1800

GCI_ELI Carbon under ice sheets released to atmo-sphere Modelled carbon storage on new land included

1027 479 1506

Time (years BP)

GCE_ELEGCE_ELIGCI_ELEGCI_ELI

Total terrestrial CVegetation CSoil C

Figure A1 Absolute changes in carbon stores (vegetation and soil) over time The four scenarios are as detailed in Table 1

minus40

minus20

Net down surface SW flux 21 ka

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

minus150 minus100 minus50 0 50 100 150minus4

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

Figure A2 Maps of mean annual anomalies between Dynamic and Static simulations for net downward shortwave flux sensible heat andlatent heat for the 21 and 30 kyr simulations

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

Figure A3 Maps of mean annual anomalies of vegetation cover between LGM and PI in the Dynamic simulations for the five PFTs pft1 isbroadleaf trees pft2 is needleleaf trees pft3 is C3 grasses pft4 is C4 grasses pft5 is shrubs

(a) Albedo 21 ka (b) Albedo 30 ka

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

(c) (d) Vegetation C 21 ka Vegetation C 30 ka

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

(e) (f) Soil C 21 ka Soil C 30 ka

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

Figure A4 Mean annual anomaly of (a b) surface albedo (unitless) (c d) vegetation carbon (kgCmminus2) and (e f) soil carbon (kgCmminus2)between the Dynamic vegetation simulation and the equivalent Static vegetation simulation For 21 ka (a c e) and 30 ka (b d f) The patternof reduced surface albedo at 30 ka is similar to the pattern at 56 and 100 ka Since in the Static simulations the carbon remains at PI levelspanels (c)ndash(f) also represent the anomaly to PI This figure includes the carbon under ice sheets

Author contributions PJV and JSS ran the climate model simu-lations TDB did the analysis and wrote the manuscript AR ran thecGENIE model simulations All the authors provided comments andcontributed to the manuscript

Competing interests The authors declare that they have no con-flict of interest

Acknowledgements Taraka Davies-Barnard was funded bythe European Research Councilrsquos grant ERC-2013-CoG-617313(PaleoGENIE) This work was carried out using the computationalfacilities of the Advanced Computing Research Centre Universityof Bristol ndash httpwwwbrisacukacrc Taraka Davies-Barnardacknowledges funding from the European Unionrsquos Horizon2020 research and innovation programme under grant agreementno 641816 (CRESCENDO)

Edited by Andreacute PaulReviewed by two anonymous referees

References

Adams J and Faure H A new estimate of changing carbon storageon land since the last glacial maximum based on global landecosystem reconstruction Global Planet Change 16ndash17 3ndash24httpsdoiorg101016S0921-8181(98)00003-4 1998

Anhuf D Ledru M-P Behling Jr H F D C CordeiroR der Hammen T V Karmann I Marengo J OliveiraP D Pessenda L Siffedine A Albuquerque A andDias P D S Paleo-environmental change in Amazo-nian and African rainforest during the LGM Palaeogeog-raphy Palaeoclimatology Palaeoecology 239 510ndash527httpsdoiorg101016jpalaeo200601017 2006

Betts R A Offset of the potential carbon sink from boreal foresta-tion by decreases in surface albedo Nature 408 187ndash190httpsdoiorg10103835041545 2000

Bigelow N H Brubaker L B Edwards M E Harrison S PPrentice I C Anderson P M Andreev A A Bartlein P JChristensen T R Cramer W Kaplan J O Lozhkin A VMatveyeva N V Murray D F McGuire A D RazzhivinV Y Ritchie J C Smith B Walker D A Gajewski K WolfV Holmqvist B H Igarashi Y Kremenetskii K Paus APisaric M F J and Volkova V S Climate change and Arcticecosystems 1 Vegetation changes north of 55 N between thelast glacial maximum mid-Holocene and present J GeophysRes-Atmos 108 8170 httpsdoiorg1010292002JD0025582003

Bird M I Lloyd J and Farquhar G D Terrestrialcarbon storage at the LGM Nature 371 566ndash566httpsdoiorg101038371566a0 1994

Bonan G B Forests and Climate Change Forcings Feedbacksand the Climate Benefits of Forests Science 320 1444ndash1449httpsdoiorg101126science1155121 2008

Bradshaw C D Lunt D J Flecker R and Davies-Barnard T Disentangling the roles of late Miocenepalaeogeography and vegetation Implications for cli-

mate sensitivity Palaeogeogr Palaeocl 417 17ndash34httpsdoiorg101016jpalaeo201410003 2015

Brovkin V Ganopolski A Archer D and Munhoven G GlacialCO2 cycle as a succession of key physical and biogeochemicalprocesses Clim Past 8 251ndash264 httpsdoiorg105194cp-8-251-2012 2012

Brovkin V Boysen L Arora V K Boisier J P CaduleP Chini L Claussen M Friedlingstein P Gayler Vvan den Hurk B J J M Hurtt G C Jones C D KatoE de Noblet-Ducoudreacute N Pacifico F Pongratz J andWeiss M Effect of Anthropogenic Land-Use and Land-CoverChanges on Climate and Land Carbon Storage in CMIP5 Projec-tions for the Twenty-First Century J Climate 26 6859ndash6881httpsdoiorg101175JCLI-D-12-006231 2013a

Brovkin V Boysen L Raddatz T Gayler V Loew A andClaussen M Evaluation of vegetation cover and land-surfacealbedo in MPI-ESM CMIP5 simulations J Adv Model EarthSyst 5 48ndash57 httpsdoiorg1010292012MS000169 2013b

Ciais P Tagliabue A Cuntz M Bopp L Scholze M Hoff-mann G Lourantou A Harrison S P Prentice I C KelleyD I Koven C and Piao S L Large inert carbon pool in theterrestrial biosphere during the Last Glacial Maximum NatureGeosci 5 74ndash79 httpsdoiorg101038ngeo1324 2012

Claussen M Late Quaternary vegetation-climate feedbacks ClimPast 5 203ndash216 httpsdoiorg105194cp-5-203-2009 2009

Claussen M Fohlmeister J Ganopolski A and Brovkin V Veg-etation dynamics amplifies precessional forcing Geophys ResLett 33 L09709 httpsdoiorg1010292006GL026111 2006

Colbourn G Ridgwell A and Lenton T M The Rock Geochem-ical Model (RokGeM) v09 Geosci Model Dev 6 1543ndash1573httpsdoiorg105194gmd-6-1543-2013 2013

Collins M Tett S F B and Cooper C The internal climatevariability of HadCM3 a version of the Hadley Centre cou-pled model without flux adjustments Clim Dynam 17 61ndash81httpsdoiorg101007s003820000094 2001

Cowling S A Maslin M A and Sykes M T Paleovegeta-tion Simulations of Lowland Amazonia and Implications forNeotropical Allopatry and Speciation Quaternary Res 55 140ndash149 httpsdoiorg101006qres20002197 2001

Cox P M Description of the TRIFFID dynamic global vegeta-tion model available at httpjulesjchmrorgsitesjulesjchmrorgfilesHCTN_24pdf (last access 20 October 2017) 2001

Cox P Huntingford C and Harding R A canopy conductanceand photosynthesis model for use in a GCM land surface schemeJ Hydrol 212ndash213 79ndash94 httpsdoiorg101016S0022-1694(98)00203-0 1998

Cox P M Betts R A Bunton C B Essery R L H Rown-tree P R and Smith J The impact of new land surface physicson the GCM simulation of climate and climate sensitivity ClimDynam 15 183ndash203 httpsdoiorg101007s0038200502761999

Crowley T J Ice Age terrestrial carbon changesrevisited Global Biogeochem Cy 9 377ndash389httpsdoiorg10102995GB01107 1995

Crucifix M and Loutre F M Transient simulations overthe last interglacial period (1260ndash115 kyr BP) feed-back and forcing analysis Clim Dynam 19 417ndash433httpsdoiorg101007s00382-002-0234-z 2002

Davies-Barnard T Valdes P J Singarayer J S and Jones C DClimatic impacts of land-use change due to crop yield increasesand a universal carbon tax from a scenario model J Climate 271413ndash1424 httpsdoiorg101175JCLI-D-13-001541 2014a

Davies-Barnard T Valdes P J Singarayer J S Pacifico F Mand Jones C D Full effects of land use change in the repre-sentative concentration pathways Environ Res Lett 9 114014httpsdoiorg1010881748-9326911114014 2014b

Davies-Barnard T Valdes P J Singarayer J S Wiltshire A Jand Jones C D Quantifying the relative importance of landcover change from climate and land use in the representative con-centration pathways Global Biogeochem Cy 2014GB004949httpsdoiorg1010022014GB004949 2015

Davin E L and de Noblet-Ducoudreacute N Climatic Im-pact of Global-Scale Deforestation Radiative ver-sus Nonradiative Processes J Climate 23 97ndash112httpsdoiorg1011752009JCLI31021 2010

de Noblet N I Prentice I C Joussaume S Texier D Botta Aand Haxeltine A Possible role of atmosphere-biosphere inter-actions in triggering the Last Glaciation Geophys Res Lett 233191ndash3194 httpsdoiorg10102996GL03004 1996

Edwards N R and Marsh R Uncertainties due to transport-parameter sensitivity in an efficient 3-D ocean-climate modelClim Dynam 24 415ndash433 httpsdoiorg101007s00382-004-0508-8 2005

Essery R Best M and Cox P MOSES 22 technical documenta-tion available at httpbiodavatmoscolostateedukrausPapersBiosphere20ModelsHCTN_30pdf 2001

Gallimore R G and Kutzbach J E Role of orbitally inducedchanges in tundra area in the onset of glaciation Nature 381503ndash505 httpsdoiorg101038381503a0 1996

Ganopolski A Calov R and Claussen M Simulation ofthe last glacial cycle with a coupled climate ice-sheetmodel of intermediate complexity Clim Past 6 229ndash244httpsdoiorg105194cp-6-229-2010 2010

Gasse F Hydrological changes in the African tropics since theLast Glacial Maximum Quaternary Sci Rev 19 189ndash211httpsdoiorg101016S0277-3791(99)00061-X 2000

Gedney N and Valdes P J The effect of Amazo-nian deforestation on the northern hemisphere circula-tion and climate Geophys Res Lett 27 3053ndash3056httpsdoiorg1010292000GL011794 2000

Gillett N P Arora V K Matthews D and Allen M R Con-straining the ratio of global warming to cumulative CO2 emis-sions using CMIP5 simulations J Climate 130314153438000httpsdoiorg101175JCLI-D-12-004761 2013

Goodwin P Williams R G and Ridgwell A Sensitivity ofclimate to cumulative carbon emissions due to compensationof ocean heat and carbon uptake Nature Geosci 8 29ndash34httpsdoiorg101038ngeo2304 2015

Gregory D Smith R N B and Cox P M CANOPY SUR-FACE AND SOIL HYDROLOGY httpprecismetofficecomUM_Docs025pdf 1994

Harrison S P and Prentice C I Climate and CO2 controlson global vegetation distribution at the last glacial maximumanalysis based on palaeovegetation data biome modelling andpalaeoclimate simulations Global Change Biol 9 983ndash1004httpsdoiorg101046j1365-2486200300640x 2003

Harrison S P Yu G Takahara H and Prentice I CPalaeovegetation (Communications arising) Diversityof temperate plants in east Asia Nature 413 129ndash130httpsdoiorg10103835093166 2001

Hoogakker B A A Smith R S Singarayer J S Marchant RPrentice I C Allen J R M Anderson R S Bhagwat S ABehling H Borisova O Bush M Correa-Metrio A de Ver-nal A Finch J M Freacutechette B Lozano-Garcia S GoslingW D Granoszewski W Grimm E C Gruumlger E Hansel-man J Harrison S P Hill T R Huntley B Jimeacutenez-MorenoG Kershaw P Ledru M-P Magri D McKenzie M MuumlllerU Nakagawa T Novenko E Penny D Sadori L Scott LStevenson J Valdes P J Vandergoes M Velichko A Whit-lock C and Tzedakis C Terrestrial biosphere changes over thelast 120 kyr Clim Past 12 51ndash73 httpsdoiorg105194cp-12-51-2016 2016

Hopcroft P O and Valdes P J Last glacial maximum constraintson the Earth System model HadGEM2-ES Clim Dynam 451657ndash1672 httpsdoiorg101007s00382-014-2421-0 2014

Hugelius G Strauss J Zubrzycki S Harden J W Schuur EA G Ping C-L Schirrmeister L Grosse G Michaelson GJ Koven C D OrsquoDonnell J A Elberling B Mishra UCamill P Yu Z Palmtag J and Kuhry P Estimated stocksof circumpolar permafrost carbon with quantified uncertaintyranges and identified data gaps Biogeosciences 11 6573ndash6593httpsdoiorg105194bg-11-6573-2014 2014

Jahn A Claussen M Ganopolski A and Brovkin VQuantifying the effect of vegetation dynamics on the cli-mate of the Last Glacial Maximum Clim Past 1 1ndash7httpsdoiorg105194cp-1-1-2005 2005

Kageyama M Braconnot P Bopp L Caubel A Foujols M-AGuilyardi E Khodri M Lloyd J Lombard F Mariotti VMarti O Roy T and Woillez M-N Mid-Holocene and LastGlacial Maximum climate simulations with the IPSL model-partI comparing IPSL_CM5A to IPSL_CM4 Clim Dynam 402447ndash2468 httpsdoiorg101007s00382-012-1488-8 2012

Kaplan J O Wetlands at the Last Glacial Maximum Distribu-tion and methane emissions Geophys Res Lett 29 3-1ndash3-4httpsdoiorg1010292001GL013366 2002

Kaplan J O Prentice I C Knorr W and Valdes P JModeling the dynamics of terrestrial carbon storage sincethe Last Glacial Maximum Geophys Res Lett 29 2074httpsdoiorg1010292002GL015230 2002

Kohfeld K E and Ridgwell A Glacial-Interglacial Variability inAtmospheric CO2 in Surface Ocean Lower Atmosphere Pro-cesses edited by Queacutereacute C L and Saltzman E S pp 251ndash286 American Geophysical Union httponlinelibrarywileycomdoi1010292008GM000845summary 2009

Koumlhler P and Fischer H Simulating changes in the ter-restrial biosphere during the last glacialinterglacialtransition Global Planet Change 43 33ndash55httpsdoiorg101016jgloplacha200402005 2004

Koumlhler P Knorr G and Bard E Permafrost thawingas a possible source of abrupt carbon release at theonset of the BollingAllerod Nature Comm 5 5520httpsdoiorg101038ncomms6520 2014

Lawson E C Wadham J L Tranter M Stibal M Lis G PButler C E H Laybourn-Parry J Nienow P Chandler Dand Dewsbury P Greenland Ice Sheet exports labile organic

carbon to the Arctic oceans Biogeosciences 11 4015ndash4028httpsdoiorg105194bg-11-4015-2014 2014

Lord N S Ridgwell A Thorne M C and Lunt D J An im-pulse response function for the long tail of excess atmosphericCO2 in an Earth system model Global Biogeochem Cy 30 2ndash17 httpsdoiorg1010022014GB005074 2016

Martin Calvo M and Prentice I C Effects of fireand CO2 on biogeography and primary production inglacial and modern climates New Phytol 208 987ndash994httpsdoiorg101111nph13485 2015

Maslin M A Ettwein V J Boot C S Bendle J and PancostR D Amazon Fan biomarker evidence against the Pleistocenerainforest refuge hypothesis J Quaternary Sci 27 451ndash460httpsdoiorg101002jqs1567 2012

Matthews H D Weaver A J Eby M and Meissner K J Radia-tive forcing of climate by historical land cover change GeophysRes Lett 30 1055 httpsdoiorg1010292002GL0160982003

Matthews H D Gillett N P Stott P A and Zick-feld K The proportionality of global warming tocumulative carbon emissions Nature 459 829ndash832httpsdoiorg101038nature08047 2009

Menviel L Yu J Joos F Mouchet A Meissner K Jand England M H Poorly ventilated deep ocean at theLast Glacial Maximum inferred from carbon isotopes Adata-model comparison study Paleoceanography 32 2ndash17httpsdoiorg1010022016PA003024 2017

Montenegro A Eby M Kaplan J O Meissner K J andWeaver A J Carbon storage on exposed continental shelvesduring the glacial-interglacial transition Geophys Res Lett 33l08703 httpsdoiorg1010292005GL025480 2006

Niessen F Hong J K Hegewald A Matthiessen J SteinR Kim H Kim S Jensen L Jokat W Nam S-I and Kang S-H Repeated Pleistocene glaciation of theEast Siberian continental margin Nature Geosci 6 842ndash846httpsdoiorg101038ngeo1904 2013

Orsquoishi R and Abe-Ouchi A Influence of dynamic vegetation onclimate change and terrestrial carbon storage in the Last GlacialMaximum Clim Past 9 1571ndash1587 httpsdoiorg105194cp-9-1571-2013 2013

Pickett E J Harrison S P Hope G Harle K DodsonJ R Peter Kershaw A Colin Prentice I Backhouse J Col-houn E A DrsquoCosta D Flenley J Grindrod J HaberleS Hassell C Kenyon C Macphail M Martin H Mar-tin A H McKenzie M Newsome J C Penny D Pow-ell J Ian Raine J Southern W Stevenson J Sutra J-PThomas I van der Kaars S and Ward J Pollen-based recon-structions of biome distributions for Australia Southeast Asiaand the Pacific (SEAPAC region) at 0 6000 and 18000 14C yrBP J Biogeogr 31 1381ndash1444 httpsdoiorg101111j1365-2699200401001x 2004

Pongratz J Reick C H Raddatz T and Claussen M Biogeo-physical versus biogeochemical climate response to historical an-thropogenic land cover change Geophys Res Lett 37 L08702httpsdoiorg1010292010GL043010 2010

Pope V D Gallani M L Rowntree P R and Stratton R AThe impact of new physical parametrizations in the HadleyCentre climate model HadAM3 Clim Dynam 16 123ndash146httpsdoiorg101007s003820050009 2000

Prentice I C and Jolly D Mid-Holocene and glacial-maximumvegetation geography of the northern continents and AfricaJ Biogeogr 27 507ndash519 httpsdoiorg101046j1365-2699200000425x 2000

Prentice I C Sykes M Lautenschlager M Harrison S Denis-senko O and Bartlein P Modelling Global Vegetation Pat-terns and Terrestrial Carbon Storage at the Last Glacial Max-imum Global Ecol Biogeogr 3 67ndash76 httpwwwjstororgstable2997548 1993

Prentice I C Harrison S P and Bartlein P J Global veg-etation and terrestrial carbon cycle changes after the last iceage New Phytol 189 988ndash998 httpsdoiorg101111j1469-8137201003620x 2011

Ridgwell A and Hargreaves J C Regulation of at-mospheric CO2 by deep-sea sediments in an Earthsystem model Global Biogeochem Cy 21 gB2008httpsdoiorg1010292006GB002764 2007

Ridgwell A Hargreaves J C Edwards N R Annan J DLenton T M Marsh R Yool A and Watson A Marine geo-chemical data assimilation in an efficient Earth System Modelof global biogeochemical cycling Biogeosciences 4 87ndash104httpsdoiorg105194bg-4-87-2007 2007

Ridgwell A Maslin M and Kaplan J O Flooding of the conti-nental shelves as a contributor to deglacial CH4 rise J Quater-nary Sci 27 800ndash806 httpsdoiorg101002jqs2568 2012

Saito K Sueyoshi T Marchenko S Romanovsky V Otto-Bliesner B Walsh J Bigelow N Hendricks A andYoshikawa K LGM permafrost distribution how well canthe latest PMIP multi-model ensembles perform reconstruc-tion Clim Past 9 1697ndash1714 httpsdoiorg105194cp-9-1697-2013 2013

Shackleton N J Lamb H H Worssam B C Hodgson J MLord A R Shotton F W Schove D J and Cooper L H NThe Oxygen Isotope Stratigraphic Record of the Late Pleis-tocene [and Discussion] Philos T Roy Soc B 280 169ndash182httpsdoiorg101098rstb19770104 1977

Shellito C J and Sloan L C Reconstructing a lost Eocene Par-adise Part II On the utility of dynamic global vegetation mod-els in pre-Quaternary climate studies Global Planet Change 5018ndash32 httpsdoiorg101016jgloplacha200508002 2006

Singarayer J S and Valdes P J High-latitude climate sensitivityto ice-sheet forcing over the last 120 kyr Quaternary Sci Rev29 43ndash55 httpsdoiorg101016jquascirev2009100112010

Singarayer J S Ridgwell A and Irvine P Assessing the ben-efits of crop albedo bio-geoengineering Environ Res Lett 4045110 httpstacksioporg1748-93264i=4a=045110 2009

Singarayer J S Valdes P J Friedlingstein P Nelson S andBeerling D J Late Holocene methane rise caused by orbitallycontrolled increase in tropical sources Nature 470 723ndash757httpsdoiorg101038nature09739 2011

Valdes P J Armstrong E Badger M P S Bradshaw CD Bragg F Crucifix M Davies-Barnard T Day J JFarnsworth A Gordon C Hopcroft P O Kennedy A TLord N S Lunt D J Marzocchi A Parry L M Pope VRoberts W H G Stone E J Tourte G J L and WilliamsJ H T The BRIDGE HadCM3 family of climate modelsHadCM3Bristol v10 Geosci Model Dev 10 3715ndash3743httpsdoiorg105194gmd-10-3715-2017 2017

Wadham J L Arndt S Tulaczyk S Stibal M Tranter MTelling J Lis G P Lawson E Ridgwell A Dubnick ASharp M J Anesio A M and Butler C E H Potentialmethane reservoirs beneath Antarctica Nature 488 633ndash637httpsdoiorg101038nature11374 2012

Wang X Edwards R L Auler A S Cheng H Kong X WangY Cruz F W Dorale J A and Chiang H-W Hydroclimatechanges across the Amazon lowlands over the past 45000 yearsNature 541 204ndash207 httpsdoiorg101038nature207872017

Zeebe R E Wolf-Gladrow D and Jansen H On the time re-quired to establish chemical and isotopic equilibrium in the car-bon dioxide system in seawater Marine Chem 65 135ndash153httpsdoiorg101016S0304-4203(98)00092-9 1999

Zeng N Glacial-interglacial atmospheric CO2 change - Theglacial burial hypothesis Adv Atmos Sci 20 677ndash693httpsdoiorg101007BF02915395 2003

Zhou J Poulsen C J Rosenbloom N Shields C and BrieglebB Vegetation-climate interactions in the warm mid-CretaceousClim Past 8 565ndash576 httpsdoiorg105194cp-8-565-20122012

Zimov N S Zimov S A Zimova A E Zimova G MChuprynin V I and Chapin F S Carbon storage in per-mafrost and soils of the mammoth tundra-steppe biome Rolein the global carbon budget Geophys Res Lett 36 L02502httpsdoiorg1010292008GL036332 2009

Zimov S A Schuur E A G and Chapin F S Permafrostand the Global Carbon Budget Science 312 1612ndash1613httpsdoiorg101126science1128908 2006

Abstract

Introduction

Methods

Climate model description

EMIC description

cGENIE carbon cycle simulations

Results

Results vegetation dynamics

Results biogeophysical feedbacks

Results biogeochemistry

Discussion

Conclusions

Code and data availability

Appendix A

Author contributions

Competing interests

Acknowledgements

References

2002 de Noblet et al 1996 Zhou et al 2012) with manystudies particularly focussing on the biogeophysical effects atthe Last Glacial Maximum (LGM) (eg Hopcroft and Valdes2014 Jahn et al 2005 Kageyama et al 2012 Orsquoishi andAbe-Ouchi 2013) The second way in which the terrestrialbiosphere can influence climate is via variations in the carbonstored in vegetation and soil This is a crucial component forunderstanding changes in the carbon cycle through the lastglacialndashinterglacial cycle (Montenegro et al 2006) and nu-merous attempts have been made to estimate the total carbonstorage using a range of methods such as inferences frommarine and terrestrial carbon isotopes (eg Shackleton et al1977 Bird et al 1994) databases of pollen (eg Adams andFaure 1998 Crowley 1995) and simple and complex mod-elling (eg Prentice et al 1993 Kaplan et al 2002 Koumlh-ler and Fischer 2004 Brovkin et al 2012 Orsquoishi and Abe-Ouchi 2013) The resulting range of carbon storage changeestimates is from a few hundred to about 1000 PgC (Ciaiset al 2012) One could add to this changes in the weath-ering of soil minerals and hence CO2 uptake from the at-mosphere as well as nutrient particularly phosphate supplyto the ocean and hence changes in the ocean productivityFor simplicity we will not address these further here (exceptto include a basic silicate weathering feedback in our modelanalysis of the impacts of terrestrial carbon storage change)

Simulations of future vegetation changes show that thebiogeochemical aspect can globally be around the same mag-nitude as the biogeophysical effects (eg Davies-Barnardet al 2014b) meaning that there is uncertainty even in thesign of the net feedback with climate change Both biogeo-physical and biogeochemical effects likely also play an im-portant role in past climate change and potentially the samefundamental uncertainty in the sign of the climate feedbackmight arise However model simulations have generally fo-cussed on either the biogeophysical impacts of vegetationchanges (eg Bradshaw et al 2015 Claussen et al 2006Jahn et al 2005 Orsquoishi and Abe-Ouchi 2013 Shellito andSloan 2006) or biogeochemical impacts (eg Kaplan et al2002 Ciais et al 2012) and the question of the overall feed-back on climate rarely addressed although Claussen (2009)argues that the net effect at the LGM is dominated by thebiogeophysical effects

One of the few examples where both have been combinedand the net effects of vegetation on past climate have beenestimated over long time periods is Brovkin et al (2012)They used an Earth system model of intermediate complexity(EMIC) to suggest that the net effect of vegetation is to de-crease global temperatures during the last glacialndashinterglacialcycle But the model used is relatively coarse in resolution(10 in latitude and 51 in longitude) and reduced in physicalprocess complexity This may be important because of the lo-cal and spatially heterogeneous nature of biogeophysical ef-fects and depending on the location of the forest the biogeo-physical and biogeochemical effects of forest change can bevery different (Bonan 2008) For this reason fully coupled

general circulation models (GCMs) are commonly used inquantifying future climate changes to vegetated land surface(Brovkin et al 2013a b Davies-Barnard et al 2015 Davinand de Noblet-Ducoudreacute 2010) The importance of consid-ering both biogeophysical and carbon cycle impacts togetherat a finer scale when assessing the climate impacts of vegeta-tion is illustrated by work quantifying the climate impacts offorest changes Studies have found that deforestation wouldcause local high-latitude cooling (Betts 2000) global warm-ing (Davin and de Noblet-Ducoudreacute 2010) or even slightglobal cooling (Davies-Barnard et al 2014b a) These out-comes are not predictable from looking at the biogeophysicsor terrestrial biogeochemistry alone at coarse resolutions

Here we present the first model analysis using a fully cou-pled dynamic atmospherendashoceanndashvegetation GCM over thelast 120 kyr that quantifies the net effect of vegetation on cli-mate (A prior study ndash Singarayer and Valdes 2010 ndash did nothave dynamic vegetation and hence could not directly evalu-ate the biogeophysical effects) We separate the biogeophys-ical and biogeochemical effects of vegetation to understandthe overall climate effect of vegetation over the last glacialcycle We show that over the whole period the biogeophys-ical is the dominant effect and that the biogeochemical im-pacts may have a lower possible range than typically esti-mated We also highlight how the temporal scale affects thenet impact of terrestrial biosphere changes

2 Methods

We use the GCM HadCM3 to run a series of simulationswith and without dynamic vegetation to provide the biogeo-physical changes and the land carbon changes To look atthe climate impact of those vegetation carbon changes wethen use the GCM terrestrial carbon changes as an input tothe EMIC cGENIE to calculate the resulting change in atmo-spheric CO2 and global temperature

For future climate changes studies the response of atmo-spheric CO2 concentrations (and hence climate) to changesin terrestrial carbon storage can be calculated using the tran-sient response to cumulative emissions (TRCE) approach(Gillett et al 2013) which demonstrated proportionalitybetween carbon emissions and temperature rise (Goodwinet al 2015) We include these estimates for completenessHowever this approach is only valid for relatively rapidchanges On the longer timescales of glacialndashinterglacialchange we need to take into account the full changes inocean carbon chemistry and including the interactions ofocean and atmosphere with the solid Earth (eg weathering)To do this we employ the ldquocGENIErdquo EMIC

+211 minus173 +37

22 EMIC description

minus4

0minus

Preminusindustrial

Longitude

minus4

0minus

21 k a

Longitude

minus4

0minus

30 k a

Longitude

minus4

0minus

68 k a

LongitudeL

minus4

0minus

100 k a

Longitude

(a) (b)

(c) (d)

3 Results

minus6minus4

minus20

Years BP

rial (

minus05

Years BP

minus40

Temperature 21 ka

Longitude

minus10

minus5

minus40

Temperature 30 ka

Longitude

minus10

minus5

minus40

Precipitation 21 ka

Longitude

minus2

minus1

minus40

Precipitation 30 ka

LongitudeLa

minus2

minus1

(a) (b)

(c) (d)

Years BP

minus20

minus15

minus10

minus05

Years BP

nt heat

ndash2)

minus01

2minus0

minus00

Years BP

(a) Albedo 21 ka

(b) Albedo 30 ka

ampl ampl 04 04

0 m 0 N C N

-02 -02

-04 -04

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

0 m 0 N

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

minus10

Time (years BP)

4 Discussion

120 000 100 000 80 000 60 000 40 000 20 000 0

Time (years BP)

5 Conclusions

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

+211 minus173 +37

22 EMIC description

minus4

0minus

Preminusindustrial

Longitude

minus4

0minus

21 k a

Longitude

minus4

0minus

30 k a

Longitude

minus4

0minus

68 k a

LongitudeL

minus4

0minus

100 k a

Longitude

(a) (b)

(c) (d)

3 Results

minus6minus4

minus20

Years BP

rial (

minus05

Years BP

minus40

Temperature 21 ka

Longitude

minus10

minus5

minus40

Temperature 30 ka

Longitude

minus10

minus5

minus40

Precipitation 21 ka

Longitude

minus2

minus1

minus40

Precipitation 30 ka

LongitudeLa

minus2

minus1

(a) (b)

(c) (d)

Years BP

minus20

minus15

minus10

minus05

Years BP

nt heat

ndash2)

minus01

2minus0

minus00

Years BP

(a) Albedo 21 ka

(b) Albedo 30 ka

ampl ampl 04 04

0 m 0 N C N

-02 -02

-04 -04

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

0 m 0 N

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

minus10

Time (years BP)

4 Discussion

120 000 100 000 80 000 60 000 40 000 20 000 0

Time (years BP)

5 Conclusions

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

+211 minus173 +37

22 EMIC description

minus4

0minus

Preminusindustrial

Longitude

minus4

0minus

21 k a

Longitude

minus4

0minus

30 k a

Longitude

minus4

0minus

68 k a

LongitudeL

minus4

0minus

100 k a

Longitude

(a) (b)

(c) (d)

3 Results

minus6minus4

minus20

Years BP

rial (

minus05

Years BP

minus40

Temperature 21 ka

Longitude

minus10

minus5

minus40

Temperature 30 ka

Longitude

minus10

minus5

minus40

Precipitation 21 ka

Longitude

minus2

minus1

minus40

Precipitation 30 ka

LongitudeLa

minus2

minus1

(a) (b)

(c) (d)

Years BP

minus20

minus15

minus10

minus05

Years BP

nt heat

ndash2)

minus01

2minus0

minus00

Years BP

(a) Albedo 21 ka

(b) Albedo 30 ka

ampl ampl 04 04

0 m 0 N C N

-02 -02

-04 -04

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

0 m 0 N

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

minus10

Time (years BP)

4 Discussion

120 000 100 000 80 000 60 000 40 000 20 000 0

Time (years BP)

5 Conclusions

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

minus4

0minus

Preminusindustrial

Longitude

minus4

0minus

21 k a

Longitude

minus4

0minus

30 k a

Longitude

minus4

0minus

68 k a

LongitudeL

minus4

0minus

100 k a

Longitude

(a) (b)

(c) (d)

3 Results

minus6minus4

minus20

Years BP

rial (

minus05

Years BP

minus40

Temperature 21 ka

Longitude

minus10

minus5

minus40

Temperature 30 ka

Longitude

minus10

minus5

minus40

Precipitation 21 ka

Longitude

minus2

minus1

minus40

Precipitation 30 ka

LongitudeLa

minus2

minus1

(a) (b)

(c) (d)

Years BP

minus20

minus15

minus10

minus05

Years BP

nt heat

ndash2)

minus01

2minus0

minus00

Years BP

(a) Albedo 21 ka

(b) Albedo 30 ka

ampl ampl 04 04

0 m 0 N C N

-02 -02

-04 -04

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

0 m 0 N

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

minus10

Time (years BP)

4 Discussion

120 000 100 000 80 000 60 000 40 000 20 000 0

Time (years BP)

5 Conclusions

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

minus6minus4

minus20

Years BP

rial (

minus05

Years BP

minus40

Temperature 21 ka

Longitude

minus10

minus5

minus40

Temperature 30 ka

Longitude

minus10

minus5

minus40

Precipitation 21 ka

Longitude

minus2

minus1

minus40

Precipitation 30 ka

LongitudeLa

minus2

minus1

(a) (b)

(c) (d)

Years BP

minus20

minus15

minus10

minus05

Years BP

nt heat

ndash2)

minus01

2minus0

minus00

Years BP

(a) Albedo 21 ka

(b) Albedo 30 ka

ampl ampl 04 04

0 m 0 N C N

-02 -02

-04 -04

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

0 m 0 N

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

minus10

Time (years BP)

4 Discussion

120 000 100 000 80 000 60 000 40 000 20 000 0

Time (years BP)

5 Conclusions

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

minus6minus4

minus20

Years BP

rial (

minus05

Years BP

minus40

Temperature 21 ka

Longitude

minus10

minus5

minus40

Temperature 30 ka

Longitude

minus10

minus5

minus40

Precipitation 21 ka

Longitude

minus2

minus1

minus40

Precipitation 30 ka

LongitudeLa

minus2

minus1

(a) (b)

(c) (d)

Years BP

minus20

minus15

minus10

minus05

Years BP

nt heat

ndash2)

minus01

2minus0

minus00

Years BP

(a) Albedo 21 ka

(b) Albedo 30 ka

ampl ampl 04 04

0 m 0 N C N

-02 -02

-04 -04

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

0 m 0 N

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

minus10

Time (years BP)

4 Discussion

120 000 100 000 80 000 60 000 40 000 20 000 0

Time (years BP)

5 Conclusions

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

minus40

Temperature 21 ka

Longitude

minus10

minus5

minus40

Temperature 30 ka

Longitude

minus10

minus5

minus40

Precipitation 21 ka

Longitude

minus2

minus1

minus40

Precipitation 30 ka

LongitudeLa

minus2

minus1

(a) (b)

(c) (d)

Years BP

minus20

minus15

minus10

minus05

Years BP

nt heat

ndash2)

minus01

2minus0

minus00

Years BP

(a) Albedo 21 ka

(b) Albedo 30 ka

ampl ampl 04 04

0 m 0 N C N

-02 -02

-04 -04

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

0 m 0 N

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

minus10

Time (years BP)

4 Discussion

120 000 100 000 80 000 60 000 40 000 20 000 0

Time (years BP)

5 Conclusions

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

Years BP

minus20

minus15

minus10

minus05

Years BP

nt heat

ndash2)

minus01

2minus0

minus00

Years BP

(a) Albedo 21 ka

(b) Albedo 30 ka

ampl ampl 04 04

0 m 0 N C N

-02 -02

-04 -04

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

0 m 0 N

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

minus10

Time (years BP)

4 Discussion

120 000 100 000 80 000 60 000 40 000 20 000 0

Time (years BP)

5 Conclusions

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

(a) Albedo 21 ka

(b) Albedo 30 ka

ampl ampl 04 04

0 m 0 N C N

-02 -02

-04 -04

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

0 m 0 N

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

ampl ampl 20 20

-10 -10

-20 -20

1 1 -30 -30

-150 -100 -50 50 100 150 -150 -100 -50 50 100 150

Longitude Longitude

minus10

Time (years BP)

4 Discussion

120 000 100 000 80 000 60 000 40 000 20 000 0

Time (years BP)

5 Conclusions

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

4 Discussion

120 000 100 000 80 000 60 000 40 000 20 000 0

Time (years BP)

5 Conclusions

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

5 Conclusions

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

Appendix A

893 430 1323

1114 502 1617

1249 552 1800

1027 479 1506

Time (years BP)

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Longitude

minus40

minus20

minus40

minus20

Sensible heat 21 ka

Longitude

minus40

minus20

0minus2

Sensible heat 30 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 21 ka

Longitude

minus40

minus20

minus40

minus20

Latent heat 30 ka

Longitude

minus40

minus20

(a) (b)

(c) (d)

(e) (f)

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

minus150 minus50 0 50 100 150

minus40

Longitude

minus10

minus05

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

ii ii 04 04

0 02 02

j 00 00

-02 -02

-04 -04

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

20 20 0 0

j j 0 0

-10 -10

I -20 I -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

-10 -10

7 -20 7 -20

1 -30 1 -30

-150 -100 -so 50 100 150 -150 -100 -so 0 50 100 150

Longitude Longitude

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

Abstract

Introduction

Methods


EMIC description


Results




Discussion

Conclusions


Appendix A


Competing interests

Acknowledgements

References

Documents

Quantifying the inﬂuence of the terrestrial biosphere on …...T. Davies-Barnard et al.: Terrestrial biosphere carbon emissions 1383 2.1 Climate model description The GCM used in