zhaomin wang two climatic states and feedbacks on ... climatic states.pdfstable modes for the...

Zhaomin Wang

Two climatic states and feedbacks on thermohaline circulation in anEarth system model of intermediate complexity

Received: 15 November 2004 / Accepted: 13 April 2005� Springer-Verlag 2005

Abstract The McGill Paleoclimate Model-2 (MPM-2) isemployed to study climate–thermohaline circulation(THC) interactions in a pre -industrial climate, with aspecial focus on the feedbacks on the THC from otherclimate system components. The MPM-2, a new versionof the MPM, has an extended model domain from 90Sto 90N, active winds and no oceanic heat and freshwaterflux adjustments. In the MPM-2, there are mainly twostable modes for the Atlantic meridional overturningcirculation (MOC) under the ‘present-day’ forcing(present-day solar forcing and the pre-industrial atmo-spheric CO2 level of 280 ppm). The ‘on’ mode has anactive North Atlantic deep water formation, while the‘off’ mode has no such deep water formation. By com-paring the ‘off’ mode climate state with its ‘on’ modeanalogue, we find that there exist many large differencesbetween the two climate states, which originate fromlarge changes in the oceanic meridional heat transports.By suppressing or isolating each process associated witha continental ice sheet over North America, sea ice, theatmospheric hydrological cycle and vegetation, feed-backs from these components on the Atlantic MOC areinvestigated. Sensitivity studies investigating the role ofvarying continental ice growth and sea ice meridionaltransport in the resumption of the Atlantic MOC arealso carried out. The results show that a fast ice sheetgrowth and an enhanced southward sea ice transportsignificantly favor the resumption of the Atlantic MOCin the MPM-2. In contrast to this, the feedback from theatmospheric hydrological cycle is a weak positive one.The vegetation-albedo feedback could enhance conti-nental ice sheet growth and thus could also favor theresumption of the Atlantic MOC. However, before the

shut-down of the Atlantic MOC, feedbacks from thesecomponents on the Atlantic MOC are very weak.

1 Introduction

Today the oceanic heat transport in the Atlantic ismainly northward (Ganachaud and Wunsch 2000;Trenberth and Caron 2001) because of the relativelystable Atlantic meridional overturning circulation(MOC). However, during a glacial period it is believedthat the Atlantic MOC experienced notable changes,which were responsible for large climate fluctuations on aglobal scale (Keigwin et al. 1994; Sarnthein et al. 1994;Stocker 1998; Alley 2003). In view of these climate fluc-tuations, there has been considerable theoretical workdone investigating the stability and variation of theAtlantic MOC for both present-day and glacial periods.For example, in coupled atmosphere–ocean general cir-culation models, Schiller et al. (1997) and Vellinga et al.(2002) simulated the weakening or shut-down of thepresent-day Atlantic MOC by freshwater discharge orsalinity perturbation, but the Atlantic MOC that isweakened or shut down is not stable in their models, incontrast to the results in Manabe and Stouffer (1988) andGanopolski et al. (2001). In their application to a glacialclimate, Earth system models of intermediate complexity(EMIC) (Claussen et al. 2002), also known as climatemodels of reduced complexity, show various propertiesof stability for the Atlantic MOC (Ganopolski andRahmstorf 2001; Crucifix and Berger 2002; Prange et al.2002; Schmittner et al. 2002; Knorr and Lohman 2003).

In this paper, we carry out a model study of climate–thermohaline circulation (THC) interactions using theMcGill Paleoclimate Model-2 (MPM-2), with a specialfocus on feedbacks on the THC from other climatesystem components. The MPM-2 is a multi-componentclimate model of reduced complexity, which is a majorextension of the previous versions of the MPM (Wang

Z. WangEarth System Modelling Group,Department of Atmospheric and Oceanic Sciences,McGill University, Montreal, QC, H3A2K6 CanadaE-mail: [email protected].: +1-514-3987448Fax: +1-514-3986115

Climate Dynamics (2005) 25: 299–314DOI 10.1007/s00382-005-0033-4

and Mysak 2000, 2002; Wang et al. 2005). In the MPM-2, the model domain is extended from (75S, 75N) to(90S, 90N) to include the Antarctic continent and theArctic Ocean; the surface wind is calculated followingPetoukhov et al. (2000); and there are no oceanic heatand freshwater flux adjustments.

In the first part of this work, we briefly analyze thesimulation of two climatic states, corresponding to anactive Atlantic MOC or ‘on’ mode with North Atlanticdeep water (NADW) formation, and an ‘off’ modeAtlantic MOC without NADW formation, under thepresent-day forcing. Since the MPM-2 includes mostcomponents of the Earth system, we are able to inves-tigate the responses of those components, including theatmosphere, sea ice, continental ice and dynamic vege-tation, to the Atlantic MOC mode change.

In the second major part of this paper, we investigatethe feedbacks from the atmosphere, sea ice, continentalice and vegetation on the THC under the present-dayforcing. This investigation is necessary for us to under-stand not only the present-day THC stability, but also theglacial THC stability. We believe that the description ofthe major components in the Earth system is more com-plete in the MPM-2 than in most earlier coupled climatemodels, albeit each component is of reduced complexity.

In this paper, we use the hysteresis diagram of theAtlantic MOC to study the stability of the AtlanticMOC and identify feedbacks on the THC from othercomponents in the MPM-2. The hysteresis diagramwas first obtained by Stocker and Wright (1991) in azonally averaged ocean model and then by Mi-kolajewicz and Maier-Reimer (1994) and by Rahm-storf (1995) in ocean general circulation models.Recently, various studies have used the present-dayand/or glacial hysteresis diagram to explore the THCstability (for example, see Ganopolski and Rahmstorf2001; Prange et al. 2002; Schmittner et al. 2002; Knorrand Lohman 2003; Romanova et al. 2004). In thispaper, in order to identify feedbacks from othercomponents on the THC, various sensitivity experi-ments are carried out by suppressing or isolating sev-eral processes and the hysteresis diagrams fromdifferent experiments are compared.

The remainder of this paper is structured as follows.In Sect. 2, the MPM-2 is described. In Sect. 3, the hys-teresis diagram obtained from the control run is pre-sented. In Sect. 4, the two climatic states are compared.In Sect. 5, feedbacks from the atmosphere, continentalice, sea ice and vegetation on the THC are investigated.Concluding remarks are presented in Sect. 6.

2 Model description and experimental design

2.1 Model description

The current version of the MPM (MPM-2) is anextension of Wang and Mysak (2000, 2002) and Wang

et al. (2005), with several major improvements. TheMPM-2 is now a global climate model, which consistsof the atmosphere, ocean, sea ice, land surface, conti-nental ice and vegetation components. The variables inthe MPM-2 are sectorially or zonally averaged acrosseach continent or ocean basin, with variables in orover North America and Eurasia downscaled to a 5 by5 degree resolution in order to couple a two-dimen-sional ice sheet model (Marshall and Clarke 1997) tothe MPM (Wang and Mysak 2002). Following thestudy of Wang and Mysak (2002), a new solar energydisposition parameterization (Wang et al. 2004) hasbeen employed in the MPM and a global dynamicvegetation [VECODE: VEgetation COntinuousDEscription (Brovkin et al. 2002)] has been coupled tothe MPM (Wang et al. 2005). In the following, a briefdescription of the three most recent major improve-ments is given.

1. Incorporation of the Arctic Ocean and the Antarcticcontinent

By including the Arctic Ocean and Antarctic conti-nent, the domain of the MPM is now extended from(75S, 75N) to (90S, 90N). Since the MPM-2 is a coarseresolution model, the detailed features of the geogra-phy, atmosphere, ocean, sea ice and continental icecirculation in the polar regions cannot be resolved.Thus, it was decided to include only highly simplifiedpolar regions. Both the Arctic and Antarctic regions arezonally averaged and have a 5� resolution in themeridional direction. The Arctic Ocean has only amixed layer and no freshwater storage is considered.The meridional advection of sea ice is prescribed as inHarvey (1988). Heat is diffused from the subpolarAtlantic to the Arctic Ocean. The net freshwater budget(associated with precipitation, evaporation, runoff andsea ice salt rejection/freshwater release) in the entireArctic Ocean is put into the northernmost box of theNorth Atlantic (72.5N) to mimic the freshwater ex-change between the subpolar North Atlantic and theArctic Ocean, which we assume is instantaneous. Ant-arctica is snow or ice covered with prescribed elevationat the present-day level, and the snow or ice does notflow.

2. Parameterization of active winds using Petoukhovet al. (2000)

The surface wind is calculated following Petoukhovet al. (2000). In this method, the structure of the atmo-spheric circulation in the meridional-vertical plane isparameterized in terms of surface air temperature(SAT). The vertical profiles of temperature and specifichumidity are also parameterized as functions of SATand surface specific humidity. The zonally averaged sealevel pressure is calculated from the parameterizedzonally averaged meridional surface wind through anageostrophic relation. The azonal component of sea

300 Wang: Two climatic states and feedbacks on thermohaline circulation in an Earth system model of intermediate complexity

level pressure is derived using the relationship betweenazonal sea level temperature and pressure in a field ofquasi-stationary planetary-scale waves. The geostrophicwinds at any height are calculated using the thermalwind approximation. The large-scale surface winds arederived using the Taylor model and the synoptic-scalesurface wind speed is derived using a macroturbulentdiffusion assumption.

3. Elimination of flux adjustments

With the new parameterization of solar energy dis-position (Wang et al. 2004) and the new surface albedocalculation (Wang et al. 2005), the energy cycle in theMPM has been significantly improved. By slightly tun-ing the cloud amount and cloud optical depth (cloudamount and cloud optical depth are decreased or in-creased by less than 30%), no explicit heat and fresh-water flux adjustments are needed in this version. [Notethat cloud amount and optical depth are prescribed ashaving present-day observed values, which have uncer-tainties (see Zhang et al. 1995; Wang et al. 2004)].

2.2 Experimental design

We follow a commonly used method for varying thefreshwater forcing on the THC in order to derive ahysteresis loop. Starting from an equilibrium state underthe present-day forcing (present-day solar forcing at thetop of the atmosphere and an atmospheric CO2 con-centration of 280 ppm), the net freshwater flux (precip-itation minus evaporation plus runoff) is evenlyincreased over 20–50N of the North Atlantic to weakenand eventually shut down the Atlantic MOC. The totaladded freshwater flux over this region changed at a rateof 0.05 Sv (1 Sv=106 m3/s) per 1,000 years in order tomaintain a quasi-equilibrium oceanic state. After theshut-down of the Atlantic MOC, the freshwater flux isdecreased at the same rate to restore the Atlantic MOC.After the resumption of the Atlantic MOC, the fresh-water flux is increased again to its starting value, i.e.,with no freshwater perturbation, to complete the hys-teresis loop. This design is adopted from Stocker andWright (1991) and is used in many studies. There is nocompensation in the other regions for the freshwaterperturbation over 20–50N of the North Atlantic sinceprevious studies show that this effect is negligible(Rahmstorf and Ganopolski 1999).

In this study, the ice sheet component, which has amuch longer equilibration time-scale than the ocean, isactive. Since one purpose of this study is to investigatethe feedback from the ice sheet component on achanging Atlantic MOC, we choose 0.05 Sv per1,000 years as the freshwater flux change rate, ratherthan a much smaller value. For the purpose of modelintercomparisons, the ice sheets should be inactive orat equilibrium states. The run with suppressed icesheets can be used for this purpose (see Sect. 5.2,Fig. 9).

3 Hysteresis diagram

The maximum stream function below the Ekman layerin the Atlantic basin is used to represent the intensityof the Atlantic MOC. Hereafter, when we mention theintensity of the Atlantic MOC, this is what we mean.The Atlantic MOC intensity versus the freshwaterperturbation is plotted in Fig. 1. Similar to many othermodel results, the hysteresis behavior of the AtlanticMOC is clearly seen in Fig. 1. This hysteresis behavioris mainly due to the salt advection mechanism, whichsustains NADW formation for an active Atlantic MOCby transporting salt northward. When the AtlanticMOC is inactive, this northward salt advection isstopped, which favors a state without NADW forma-tion.

In Fig. 1, point A marks the present state of theAtlantic MOC, with an intensity of around 20 Sv andit is defined as the ‘on’ mode. Point B marks thetransition state from the ‘on’ mode to the ‘off’ modeof the Atlantic MOC. This transition happens whenthe integrated freshwater input rate over 20–50Nreaches a threshold value, here around 0.1 Sv. Whenthe added freshwater forcing is decreased to zero, the‘off’ mode, as marked by point C, is stable. By con-tinually decreasing the freshwater perturbation to anegative value, corresponding to point D, the AtlanticMOC is restored and then jumps to a significantlyintensified state. This jump is caused by substantialsalt advection to the deep water formation region. Avery large negative freshwater perturbation in thesubtropical region produces very large salinity valuesin this region since the northward advection of salt isstopped. Once the advection is turned on, a largeamount of salt is transported toward the deep waterformation region and the deep convection is thus sig-nificantly enhanced. After the release of the subtropi-cal salt, the Atlantic MOC returns to a smaller value.The restored Atlantic MOC is gradually weakened

AB

C

D

E

Fig. 1 The hysteresis diagram of the intensity of the Atlantic MOC

Wang: Two climatic states and feedbacks on thermohaline circulation in an Earth system model of intermediate complexity 301

when the freshwater perturbation is slowly returned tozero and another equilibrium state is reached, asmarked by point E. (State E is just slightly differentfrom state A with a northward shift of the deepconvection site. The climatic state corresponding to Eis very similar to that corresponding to A. The mul-tiple ‘on’ states were also obtained by Rahmstorf(1995). But note that this feature is related to the shut-down of the Labrador Sea convection in Rahmstorf(1995),which is not resolved by the two-dimensionalstructure of the ocean model in the MPM-2.) Thus,multi-equilibrium states are obtained for the present-day forcing. In the following, the two climate states,corresponding to points A and C, are first comparedin Sect. 4. Feedbacks resulting from the climate changeassociated with the Atlantic MOC mode change arethen investigated in Sect. 5.

4 Two climate states

4.1 Atmosphere

The zonally averaged annual mean SAT, precipitation,sensible heat (internal plus potential energy) transportand moisture transport are shown in Fig. 2. The solidline is for the ‘on’ mode, while the dashed line is for the‘off’ mode. In order to clearly show the differences be-tween the two modes, Fig. 3 is plotted.

From Figs. 2a and 3a, a large temperature drop (rise)is seen to occur in the middle and high latitudes of thenorthern hemisphere (NH) [southern hemisphere (SH)],while the temperature change is relatively small in thetropical region. This clearly shows the bipolar see-saweffect associated with the Atlantic MOC mode change(Crowley 1992; Stocker 1998; Broecker 1998). Thecooling in the NH and warming in the SH are furtheramplified by the ice-albedo feedback and vegetation-al-bedo feedback in the high latitudes. The simulatedmaximum annual cooling of 7.9�C occurs at 62.5N overthe North Atlantic. This value is very close to that (7�C)in Ganopolski et al. (2001). The largest cooling of12.3�C occurs at 62.5N over the North Atlantic inwinter, which is slightly smaller than that (15�C) inGanopolski et al. (2001).

The precipitation change (Figs. 2b, 3b) shows a morecomplicated pattern, but still indicates an overall in-crease in the NH and a decrease in the SH after theswitch from the ‘on’ to ‘off’ mode. The cooling (warm-ing) and the advance (retreat) of sea ice in the northern(southern) hemisphere produce decreased (increased)precipitation in the northern (southern) hemisphere. Themoisture transport change (Figs. 2d, 3d) regulates theprecipitation pattern change locally. The sensible heattransport change (Figs. 2c, 3c) is positive in bothhemispheres, which means an increased polewardtransport in the NH and a decreased poleward transportin the SH after the switch from the ‘on’ to the ‘off’ mode.The moisture transport change (Figs. 2d, 3d) is generally

–90 –60 –30 0 30 60 90–50

–40

–30

–20

–10

0

10

20

30

°C

surface air temperature

–90 –60 –30 0 30 60 900

1

2

3

4

5

6

mm

/day

precipitation

–90 –60 –30 0 30 60 90 –4

–2

0

2

4

latitude

PW

sensible heat transport

–90 –60 –30 0 30 60 90 –0.75

–0.5

–0.25

0

0.25

0.5

0.75

latitude

Sv

moisture transport

a

c d

bFig. 2 The annual meanlatitudinal profiles of SAT (a),precipitation (b), atmosphericsensible heat transport (c), andatmospheric moisture transport(d) for the Atlantic MOC ‘on’mode (solid line) and ‘off’ mode(dashed line)


small in the middle and high latitudes. In the lowerlatitudes, due to the Hadley circulation change, theequatorward moisture transport is increased in the NHand decreased in the SH. The poleward transport is in-creased in the NH and decreased in the SH in the middlelatitudes. The surface wind is generally increased afterthe mode switch in the NH, with an especially largeincrease in winter (see Fig. 4). In the SH, the surfacewind change is less regular than that in the NH, with adecrease in the subtropics and high latitudes and anincrease in the middle latitudes.

4.2 Ocean

As seen in Fig. 5, the heat transport is northwardeverywhere in the North Atlantic for the ‘on’ mode witha peak value of 0.87 PW at 30N. This value is somewhatsmaller than the peak value estimated by Ganachaudand Wunsch (2000) (1.3±0.15 PW), but lies in the rangeestimated by Macdonald (1998) (1.07±0.26 PW), and islarger than that in Stammer et al. (2003) (around0.65 PW). For the ‘off’ mode, the northward transport isvery small or close to zero in the NH and the southwardtransport appears in the SH. The heat transport changeat 30N in the North Atlantic is 0.8 PW. This value isvery close to that in Ganopolski et al. (2001), but it islarger than those in Manabe and Stouffer (1988) andVellinga et al. (2002). In Manabe and Stouffer (1988),although the maximum northward heat transport for the‘off’ mode is very close to the value in this study, the

maximum northward heat transport for the ‘on’ mode isjust around 0.45 PW. In Vellinga et al. (2002), themaximum northward heat transport is close to the valuein this study, but the value for the weakest AtlanticMOC they obtained is around 0.5 PW since their oceanhad not reached an equilibrium state. Note that a wind-driven gyre component, which is not resolved in theMPM-2, also makes a contribution to the northwardheat transport in the northern subtropics when theAtlantic MOC is in the weakest state in Vellinga et al.(2002). The Atlantic oceanic heat transport changeassociated with the Atlantic MOC mode change is theorigin of the bipolar see-saw. In the Indo-Pacific, heattransports are poleward in both hemispheres withsmaller values in the NH than in the SH for the ‘on’mode. For the ‘off’ mode, the northward transportbecomes larger in the NH and southward transportbecomes smaller in the SH. The zonally integrated glo-bal ocean poleward transport decreases in the NH andincreases in the SH since the heat transport change in theNorth Atlantic overrides that in the Indo-Pacific, whichis also an important feature of the bipolar see-saw.

Figure 6 shows oceanic heat flux (positive means thatthe ocean loses heat to the atmosphere) and freshwaterflux (positive means that the ocean gains freshwater) forthe global ocean and the Atlantic ocean. Generally, thechanges are small except in the North Atlantic. Theshut-down of the Atlantic MOC leads to a significantadvance of sea ice, which consequently insulates theocean from the atmosphere. Figure 6b thus shows thatfor the ‘off’ mode heat flux is almost zero in the middle

–90 –60 –30 0 30 60 90–4

–2

0

2

4

°C

surface air temperature

–90 –60 –30 0 30 60 90 –0.4

–0.2

0

0.2

0.4

mm

/day

precipitation

–90 –60 –30 0 30 60 90 –0.25

0

0.25

0.5

0.75

latitude

PW

sensible heat transport

–90 –60 –30 0 30 60 90 –0.15

–0.1

–0.05

0

0.05

0.1

0.15

latitude

Sv

moisture transport

a b

c d

Fig. 3 The differences (‘off’mode minus ‘on’ mode) for theSAT (a), precipitation (b),atmospheric sensible heattransport (c), and atmosphericmoisture transport (d)


and high latitudes in the North Atlantic. Since the heattransport from lower to middle and high northern lati-tudes is decreased to a very small value (see Fig. 5a), theheat loss from the ocean to the atmosphere must bereduced to maintain a heat balance, which is realized by

a significant advance of sea ice. For the freshwater fluxchange, relatively large changes occur between 40N and60N and at 72.5N (Fig. 6c, d). The increase of fresh-water flux between 40N and 60N is induced mainly byan increased freshwater release due to the melting of sea

90 75 60 45 30 15 0 15 30 45 60 75 900

2

4

6

8

10

12

surf

ace

win

d (m

/s)

DJF

90 75 60 45 30 15 0 15 30 45 60 75 900

2

4

6

8

10

12

latitude

surf

ace

win

d (m

/s)

JJA

a

b

Fig. 4 The latitudinal profilesof surface wind for DJF (a) andJJA (b), for the Atlantic MOC‘on’ mode (solid line) and ‘off’mode (dashed line)

–2

0

2

PW

Atlantic

–2

0

2

PW

Indo–Pacific

–75 –60 –45 –30 –15 0 15 30 45 60 75 90

–2

0

2

latitude

PW

Global

a

b

c

Fig. 5 The annual meanoceanic heat transport in theAtlantic ocean (a), Indo-Pacificocean (b), and global ocean (c)for the Atlantic MOC ‘on’mode (solid line) and ‘off’ mode(dashed line)


ice. The melted sea ice in this region is transported fromhigher latitudes where sea ice forms and salt is rejected.Both the salt rejection at 72.5N and ice sheet growthover North America (see Cryosphere below) consider-ably reduce the freshwater flux there. The regionalintegrated freshwater budgets for the Atlantic basin areshown in Table 1. Clearly, a large change occurs in thebands (45N, 60N) and (60N, 75N), with a very smallchange in other regions and over the whole Atlanticbasin.

Globally, the decrease of the zonally integrated oceanpoleward heat transport is very close to the increase of

the atmospheric heat (sensible plus latent heat) trans-port, and the total poleward heat transport of theatmosphere–ocean system thus remains almost un-changed (figure not shown).

4.3 Cryosphere

Table 2 shows annual mean NH and SH sea ice areasand continental ice volumes in North America andEurasia for the ‘on’ and ‘off’ modes. The ‘on’ mode NHsea ice area is somewhat smaller than the observed valuein Ropelewski (1989) (12.4·106 km2) because a simpli-fied land–sea configuration, which results in less oceanarea, is used in the northern high latitudes and polarregion. The SH sea ice area is close to the value(13.8·106 km2) given in Ropelewski (1989). The NH seaice area increases from 9.4 for the ‘on’ mode to14.4·106 km2 for the ‘off’ mode, while the SH sea icearea decreases from 13.3·106 km2 to 7.4·106 km2. Thesesea ice area changes clearly indicate the bipolar see-saweffect associated with the global ocean circulationchange, and also enhance the effect through the ice-al-bedo feedback and insulation effect. For the ‘on’ mode,continental ice volumes are zero for both North Americaand Eurasia, consistent with the interglacial state of thepresent-day cryosphere. For the ‘off’ mode, continentalice forms in North America, due to the northern cooling,with a small ice volume and ice covered area(3.3·106 km2). (The ice volume change over Greenlandis very small in the MPM-2.)

Table 1 The regional integrated freshwater budgets (in Sv) for theAtlantic basin for the ‘on’ and ‘off’ modes

40S–75N 40S–5S 5S–10N 10N–45N 45N–60N 60N–75N

On �0.3435 �0.3644 0.1356 �0.4584 0.1858 0.1578Off �0.3507 �0.3596 0.1348 �0.4672 0.2710 0.0704

Table 2 The global annual mean NH and SH sea ice areas andcontinental ice volumes over North America and Eurasia for the‘on’ and ‘off’ modes

NH sea ice (106

km2)SH seaice (106 km2)

North Americanice sheet(106 km3)

Eurasianice sheet(106 km3)

On 9.4 13.3 0.0 0.0Off 14.4 7.4 0.72 0.0

–60 –30 0 30 60 90–100

–50

0

50

100

W/m

2

oceanic heat flux (global)

–60 –30 0 30 60 90 –100

–50

0

50

100oceanic heat flux (Atlantic)

–60 –30 0 30 60 90

–4

–2

0

2

4

6

latitude

mm

/day

W/m

2m

m/d

ay

oceanic freshwater flux (global)

–60 –30 0 30 60 90

–4

–2

0

2

4

6

latitude

oceanic freshwater flux (Atlantic)

a b

c d

Fig. 6 The annual meanoceanic heat flux over the globalocean (a) and Atlantic (b) andfreshwater flux over the globalocean (c) and Atlantic (d) forthe ‘on’ mode and ‘off’ mode ofthe Atlantic MOC


4.4 Biosphere

Relatively large forest area changes occur in the north-ern part of boreal forest regions as well as in the sub-tropical region of the SH (Fig. 7). For the ‘off’ mode, theboreal forests retreat southward due to the cooling in-duced by the shutdown of the Atlantic MOC. The in-crease of forest area in the SH subtropics is induced bythe increase of precipitation in this region (see Fig. 3b).The southward shift of the boreal forest enhances thecooling in the NH by the vegetation-albedo feedback.The retreat of the boreal forest occurs mainly in NorthAmerica (figure not shown). In one sensitivity experi-ment with fixed vegetation at the starting state (corre-sponding to point A in Fig. 1), the permanent ice inNorth America does not appear after the shutdown ofthe Atlantic MOC (see Discussion in Sect. 5.5).

5 Feedbacks from other components

In this section, the overall feedback from the hydro-logical cycle on the THC stability is first studied. Then,processes associated with a continental ice sheet, sea ice,atmospheric moisture transport and vegetation are iso-lated or suppressed and the feedback on the THC sta-bility from each process is studied. Several sensitivityexperiments are also carried out in order for us to have abetter understanding of the role of continental ice andsea ice in the THC stability.

5.1 The role of the hydrological cycle

In order to investigate the overall effects on the THCof an active hydrological cycle, which consists of

atmospheric moisture transport, sea ice brine rejection/freshwater release and transport and continental icegrowth, the hysteresis diagram obtained from a runwith a fixed oceanic freshwater flux at point A in Fig. 1is compared with that from the control run (Fig. 8). Inthe run with a fixed oceanic freshwater flux, the activerole of all above hydrological processes are excluded.The right branch for the run with fixed oceanic fresh-water flux shifts slightly to the right side of that of thecontrol run. This small shift indicates a weak positivefeedback on the THC from an active hydrological cyclebefore the shut-down of the THC. This weak effect isalso simulated by Ganopolski et al. (2001) and sup-ports the conclusion drawn in Hughes and Weaver(1996). In contrast, a large destabilizing feedback onthe THC from the atmospheric moisture transport isobtained by Nakamura et al. (1994) in a box modelstudy.

The left branch for the fixed freshwater flux run shiftsto the left side of that of the control run. This shift islarger than the one to the right and indicates that anactive hydrological cycle has a negative feedback on theTHC and thus favors the resumption of the AtlanticMOC. In other words, less freshwater extraction isneeded in the latitude band (20N, 50N) of the NorthAtlantic to restore the Atlantic MOC in the run with anactive hydrological cycle.

In the following, in order to understand the dif-ferent roles played by different hydrological processes,each process is either suppressed or isolated in thesensitivity experiments. Also, because of the uncer-tainties in the simulation of ice mass balance and seaice processes, further sensitivity experiments arecarried out to investigate the effects on the THC sta-bility of different ice sheet growths and sea iceadvections.

–40 –30 –15 0 15 30 45 60 750

1

2

3

4

5

6

latitude

106 k

m2

forest area

Fig. 7 The latitudinal profile of total forest area (solid line ‘on’mode, dashed line ‘off’ mode)

–0.3 –0.25 –0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15

0

10

20

30

40

50

Freshwater perturbation (Sv)

Atla

ntic

MO

C in

tens

ity (

Sv)

Fig. 8 The hysteresis diagram of the Atlantic MOC for the controlrun (solid line) and the run with fixed oceanic freshwater flux(dashed line)


5.2 The role of continental ice growth

In this part, we show that the above negative feedbackon the THC stability partially results from the growth ofa continental ice sheet in North America. The shut-downof the Atlantic MOC leads to the growth of an ice sheetin North America (see Table 2). The growth of a con-tinental ice sheet accumulates freshwater over land andhence reduces the freshwater input into the subpolarNorth Atlantic. In order to see the role of the ice sheetgrowth, we designed an experiment in which the ice sheetgrowth is suppressed. This is realized by fixing the per-manent snow depth at 2 m. Any excessive snow is con-verted into liquid water and run off into the ocean andthe necessary latent heat to melt the excessive snow istaken from the atmosphere component to conserve theenergy.

In Fig. 9, the hysteresis diagram obtained from thisrun is compared with that from the control run. Thisfigure shows that the left branch of the hysteresis dia-gram for the run with suppressed ice sheet growth shiftsto the left side of that for the control run. This changeclearly indicates that an ice sheet growth favors theAtlantic MOC resumption and hence has a negativefeedback on an ‘off’ mode Atlantic MOC.

Around the transition point from the ‘off’ to ‘on’mode, within 1,000 years, the North American ice vol-ume change is 0.373·106 km3 in the control run. Wenote that the ice sheet growth during the glacial incep-tion is underestimated in early versions of the MPM(Wang and Mysak 2002; Cochelin 2004). It is also truethat all climate models have large uncertainties for theice mass balance, resulting from insufficient model res-olution and the accuracy of the energy and water cyclesimulations since many important mechanisms are stillunknown. For example, one recent modeling studysuggests that ice-dammed lakes could enhance ice sheet

growth (Krinner et al. 2004). Therefore, we designedanother experiment in order to examine the effects offaster ice sheet growth. In this experiment, the fractionof freezing of rain and refreezing of melt water (Wangand Mysak 2002) is unrealistically set to be 1.0 to en-hance the ice sheet growth. In this case, around thetransition point from the ‘off’ to ‘on’ mode, within1,000 years, the North American ice volume change is2.026·106 km3. The hysteresis diagram of this run iscompared with that of the control run in Fig. 10. It isfound that the left branch for the run with fast ice sheetgrowth shifts to the right side significantly. This largeshift clearly indicates that fast ice sheet growth signifi-cantly favors the resumption of the Atlantic MOC.

The latitudinal profiles of the Atlantic oceanicfreshwater fluxes for the control run and the run withfaster ice sheet growth are shown in Fig. 11 in order forus to see more detail with respect to the freshwaterforcing induced by fast ice sheet growth. More rapid icesheet growth mainly decreases the freshwater input ratein the subpolar North Atlantic, while the freshwater fluxremains unchanged in other regions. The freshwater fluxreduction of about 0.05 Sv changes the freshwaterbudget directly in the NADW formation region (around65N). This change is more effective than that in otherregions and hence narrows the hysteresis diagram byabout 0.12 Sv.

We thus conclude that the ice sheet growth favors theresumption of the THC. The restoring effect of ice sheetgrowth on the Atlantic MOC is much larger for a case ofmore rapid ice sheet growth. This result has very strongimplications for the THC–ice sheet interactions during aglacial period. If the cooling induced by the shut-downof the Atlantic MOC enhances the ice sheet mass in-crease, the freshwater input into the North Atlantic mustbe reduced and hence the resumption of the AtlanticMOC is favored. This kind of ice sheet–THC interactionhas been simulated during a glacial period by Schmittner

–0.3 –0.25 –0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15

0

10

20

30

40

50


Atla

ntic

MO

C in

tens

ity (

Sv)

Fig. 9 The hysteresis diagram of the Atlantic MOC for the controlrun (solid line) and the run with suppressed ice sheet growth (dashedline)

–0.3 –0.25 –0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15

0

10

20

30

40

50


Atla

ntic

MO

C in

tens

ity (

Sv)

Fig. 10 The hysteresis diagram of the Atlantic MOC for thecontrol run (solid line) and the run with a faster ice sheet growthrate (dashed line)


et al. (2002). However, it is important to keep in mindthe large uncertainties regarding the net ice accumula-tion in the current modeling studies on ice sheet–THCinteractions.

5.3 The role of sea ice brine rejection/freshwater release

When SST reaches the freezing point, sea ice forms. Dueto the relatively low heat conductivity of sea ice, itspresence causes the oceanic heat flux to decrease dra-matically. Consequently, the water temperature underthe ice is prevented from further decrease and theatmosphere above the ice is cooled down significantly. Ina model without sea ice, the weakening of the AtlanticMOC decreases the SST of the subpolar North Atlanticmonotonically. The drop of SST increases the sea waterdensity there. Therefore, there is a negative SST–THCfeedback (Rahmstorf and Willebrand 1995). However,this stabilization mechanism should be modified if therole of sea ice is taken into account in the model (forexample, see Lohman and Gerdes 1998). Since SST canonly drop until the freezing point is reached and then seaice forms, the sea ice thermal effect in some sense sta-bilizes an ‘off’ mode Atlantic MOC. In this study, wefocus on the hydrologic effects of sea ice on the THCstability.

When sea ice forms, salt is rejected; when sea icemelts, freshwater is released. If sea ice forms and melts inthe same region in a certain period, the integratedfreshwater forcing is canceled out over this period.However, sea ice is not motionless and it is generallytransported away from a sea ice formation region to asea ice melting region on a large scale. In addition, wenote that over sea ice covered regions, the area-inte-

grated freshwater forcing is zero except for the case inwhich total sea ice volume changes with climate changes.In the MPM-2 the total sea ice volume change rate issmall in terms of freshwater forcing (on the order of0.001 Sv or less). Despite the latter, the latitudinalstructure of the oceanic freshwater flux is altered by thesea ice brine rejection/freshwater release and transportprocesses, with a decreased freshwater flux in the sub-polar region and an increased one in lower latitudes (seeFig. 6d and the related discussion).

In order to investigate the effect of sea ice salt rejec-tion/freshwater release and transport on the THC sta-bility, we design another experiment in which the annualcycle of sea ice brine rejection/freshwater release at everygrid is fixed at its starting state (corresponding to pointA) and the ice sheet growth is suppressed. We comparethe hysteresis diagram from this run with that from therun with suppressed ice sheet growth only (see Fig. 12).Since the atmospheric hydrological cycles are found tobe identical in these two runs for both ‘on’ and ‘off’modes and there is no ice sheet growth, the only differ-ence is in the sea ice brine rejection/freshwater releaseand transport processes. The right branches of the dia-grams for these two runs are very close to each other,which suggests that the feedback on the THC stabilityfrom sea ice salt rejection/freshwater release is very weakas the Atlantic MOC is gradually weakened. A closeinspection finds that both the sea ice concentration andthickness changes are very small before the shut-down ofthe Atlantic MOC in these runs.

However, the shut-down of the Atlantic MOC causesa large southward advance of sea ice and changes the seaice formation and melting region, and the salt rejectionand freshwater release sites are consequently changed.This causes a larger difference between the left branchesof the hysteresis diagrams. Figure 12 thus indicates that

–60 –30 0 30 60 90

–4

–2

0

2

4

6

latitude

mm

/day


Fig. 11 The latitudinal profiles of oceanic freshwater fluxes overthe Atlantic for the control run (solid line) and the run with fasterice sheet growth as in Fig. 10 (dashed line). Both are for an ‘off’mode and correspond to a zero freshwater perturbation (seeFig. 10)

–0.3 –0.25 –0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15

0

10

20

30

40

50


Atla

ntic

MO

C in

tens

ity (

Sv)

Fig. 12 The hysteresis diagram of the Atlantic MOC for the runwith suppressed ice sheet growth (solid line) and the run with bothsuppressed ice sheet growth and fixed NH sea ice brine rejection(dashed line)


the sea ice brine rejection/freshwater release and trans-port processes favor the resumption of the AtlanticMOC. At the transition state from the ‘off’ to the ‘on’mode, total sea ice volume decreases and hence there is anet freshwater input into the subpolar North Atlantic.While this does not favor the resumption of the AtlanticMOC, its effect is very small simply because the inducedfreshwater flux change associated with the sea ice vol-ume change rate is very small. Since the overall effect ofsea ice processes favors the resumption of the AtlanticMOC, we may only conclude that the shifts of the saltrejection and freshwater release sites favor thisresumption. In fact, the subpolar North Atlantic is afreshwater release region for the ‘on’ mode, while for the‘off’ mode it becomes a region with salt rejection. Thismainly explains why the freshwater flux increases be-tween 45N and 60N and decreases to the north of 60N inFig. 6d.

The SH sea ice salt rejection/freshwater release pro-cess has little effect on the Atlantic MOC resumption inthe MPM-2 (figure not shown), as demonstrated byanother experiment in which only salt rejection/fresh-water release in the SH is fixed. However, Keeling andStephens (2001) suggested that the SH sea ice has a largeeffect on the global THC during a glacial period, basedon their hydraulic model. A more sophisticated climatemodel with a 3-D ocean model suggests that the SH seaice may play a role in the glacial North Atlantic THC(Shin et al. 2003). We note that our results are obtainedbased on present-day forcing. It would be very inter-esting for us to investigate the role of the SH sea iceduring a glacial period, in which climate forcing and seaice conditions are very different from the present day.

Generally, the effect of sea ice brine rejection/fresh-water release on the THC stability is not large as illus-trated in Fig. 12. Many modeling studies (Zhang et al.1995; Lenderink and Haarsma’s 1996; Lohmann andGerdes 1998) also show that this sea ice effect is minor.However, we note that in these modeling studies, thereare no sea ice transport processes and in the MPM-2, seaice meridional advection is prescribed and remains un-changed. A shut-down of the Atlantic MOC leads to avery cold climate and hence a stronger wind field in theNorth Atlantic. The sea ice meridional advection wouldmost likely be stronger. Based on this reasoning, we didanother two experiments in which we increased theprescribed sea ice advection velocity by two and fourtimes when the Atlantic MOC is in the ‘off’ mode. Inthese experiments, ice sheet growth is also suppressed inorder to isolate the sea ice transport effect. The hyster-esis diagrams obtained from these two runs are com-pared with that from the run with unchanged sea iceadvection velocity and suppressed ice sheet growth(Fig. 13). Since the atmospheric hydrological cycle dif-fers very little among these three experiments, only thedifference in the sea ice advection velocity makes thehysteresis diagrams differ from each other. Inthe experiments with increased sea ice advection veloc-ity, the Atlantic MOC recovers earlier than in the run

with unchanged sea ice advection velocity. This dem-onstrates that an intensified southward transport of seaice in the North Atlantic favors the resumption of theAtlantic MOC.

The latitudinal profiles of the Atlantic freshwaterfluxes are shown in Fig. 14, for the run with unchangedsea ice advection velocity and the run with four times thesea ice advection velocity for the ‘off’ mode. In the lattercase, the freshwater flux is decreased to the north of55N, while it is increased in the region from 40N to 55N.Enhanced sea ice advection produces more sea ice for-

–0.3 –0.25 –0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15

0

10

20

30

40

50


Atla

ntic

MO

C in

tens

ity (

Sv)

Fig. 13 The hysteresis diagram of the Atlantic MOC for the runwith prescribed present-day sea ice advection velocity (solid line),the run with increased sea ice advection velocity by two times(dashed line) and the run with increased sea ice advection velocityby four times (dotted line). The ice sheet growth is suppressed inthese three experiments

–60 –30 0 30 60 90

–4

–2

0

2

4

6

8

latitude

mm

/day


Fig. 14 The latitudinal profiles of oceanic freshwater fluxes overthe Atlantic for the control run (solid line) and the run with fourtimes sea ice advection velocity as in Fig. 13 (dashed line). Both arefor an ‘off’ mode and correspond to a zero freshwater perturbation(see Fig. 13)


mation to the north of 55N and more sea ice meltingbetween 40N and 55N in the MPM-2. Unlike thefreshwater flux change induced by rapid ice sheet growth(see Fig. 11), the net freshwater flux change is almostzero in the North Atlantic since the total sea ice volumechange rate is very small. It is interesting to note that thesalt rejection in the very northern region favors theresumption of the Atlantic MOC, although the basin-scale integrated freshwater forcing perturbation is al-most zero. On the other hand, the situation with the saltrejection/freshwater release and southward sea icetransport is equivalent to a northward salt transport. Inan ‘off’ mode, the northward salt advection by theAtlantic MOC is stopped, which can stabilize this mode.However, this ‘off’ mode can also be destabilized by theequivalent northward salt transport associated with seaice salt rejection/freshwater release and southward seaice transport.

Because of the coarse model resolution and simplicityof the sea ice dynamics in the MPM, there may be largeuncertainties in the simulation of THC–sea ice interac-tions in this EMIC. Future investigations are necessaryusing more sophisticated coupled atmosphere–ocean–sea ice models. However, the enhanced salt rejectionduring stadials or the very weak phase of the AtlanticMOC are observed during the last glacial period. Thisenhanced salt rejection process is believed to be impor-tant in the restoration of the Atlantic MOC during thelast glacial (Dokken and Jansen 1999; van Kreveld et al.2000). Also, in their transient THC perturbation exper-iment using HadCM3 (Vellinga et al. 2002), sea ice brinerejection is found to be responsible for the sustaining ofGreenland and Norwegian Sea deep convection.Therefore, the role of sea ice brine rejection/freshwaterrelease and possibly enhanced sea ice advection in theresumption of the THC must be taken into account.

5.4 The role of atmospheric hydrological cycle

By suppressing the ice sheet growth and fixing the saltrejection/freshwater release process in both hemispheres,the effect of the atmospheric hydrological cycle can beisolated. The hysteresis diagram from this run is com-pared with that from the run with a fixed oceanicfreshwater flux in Fig. 15. It is found that the polewardatmospheric moisture transport in the middle latitudesof the NH for the ‘off’ mode is larger than that for the‘on’ mode (see Figs. 2d, 3d). Nakamura et al. (1994)argued that the increase of the moisture eddy transportdue to the weakening of the THC has a strong destabi-lizing effect on the THC. However, Fig. 15 suggests thatthis destabilizing effect is very weak before the shut-down of the Atlantic MOC, presumably due to a verysmall change in the atmospheric moisture transport.Different parameterization schemes and model resolu-tions could produce different changes. However, anobvious reason for this weak effect is that the climatechange is very small and hence the atmospheric hydro-

logical cycle change is also small before the shut-down ofthe Atlantic MOC in the MPM-2. An ‘off’ modeAtlantic MOC causes a much larger climate change andhence a larger poleward atmospheric moisture transport.Consequently, the shift of the left branch is relativelylarge. The leftward shift of this branch for the run withan active atmospheric hydrological cycle indicates thatthis cycle does not favor the resumption of the AtlanticMOC. The small change in the atmospheric hydrologicalcycle associated with the THC mode change is also ob-tained in other climate models of reduced complexity(Lohmann and Gerdes 1998; Ganopolski et al. 2001)and in a coupled atmosphere–ocean general circulationmodel (Manabe and Stouffer 1988). The weak effect ofthe atmospheric hydrological cycle on the THC in theMPM-2 simulation is consistent with many previousstudies (for example, see Hughes and Weaver 1996;Lohmann and Gerdes 1998; Ganopolski et al. 2001).

5.5 The role of vegetation

By fixing the vegetation at the starting point (point A),the interactive role of vegetation is excluded. Figure 16shows that the vegetation change has virtually no effecton the right branch of the diagram but a notable effecton the left branch. This is due to a very small climatechange before the shut-down of the Atlantic MOC and alarge change after. Table 2 shows that the shut-down ofthe Atlantic MOC leads to ice sheet growth over NorthAmerica. This growth is made possible by the vegeta-tion-albedo feedback in the MPM-2. In the run with afixed vegetation cover, permanent ice is prevented fromappearing in North America. Due to the vegetation-al-bedo feedback, permanent snow also appears overNorth America in a global deforestation experimentusing a more sophisticated climate model in Renssen

–0.3 –0.25 –0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15

0

10

20

30

40

50


Atla

ntic

MO

C in

tens

ity (

Sv)

Fig. 15 The hysteresis diagram of the Atlantic MOC for the runwith fixed oceanic freshwater flux (solid line) and the run withsuppressed ice sheet growth and fixed SH and NH sea ice brinerejection (dashed line)


et al. (2003). By comparing the hysteresis diagram fromthe run with fixed vegetation with that from the run withsuppressed ice sheet growth, it is found that the twodiagrams are identical (figure not shown). It is obviousthat vegetation change can change the surface albedoand hence change the energy budget at the land surface,but this thermal effect on the hysteresis diagram is veryweak. Rather, it was found that with fixed vegetation,the ice sheet growth does not occur and hence there is nofreshwater accumulation over land. This hydrologiceffect is relatively large in the hysteresis diagram.

The above thermal effect on the THC is weak becausesea ice advances southward significantly. The thermalinsulation effect of sea ice decouples the ocean from theatmosphere for the heat exchange. The presence of seaice is directly determined by the SST, which is effectivelydetermined by the ocean heat transport. Moreover, inthe high latitude, since SST is close to the freezing point,water density is sensitive to the salinity change.

5.6 Summary

In this section, it is found that atmospheric hydrologicfeedbacks on the THC stability are generally weak in theMPM-2. This general conclusion is consistent with someearlier work (Hughes and Weaver 1996; Lohmann et al.1996; Weber 1998; Lohmann and Gerdes 1998; Gano-polski et al. 2001), but it contrasts with the conclusiondrawn from box model studies (Nakamura et al. 1994;Lohmann 1995). Box models have a much lower reso-lution than other climate models and hence some feed-backs are possibly exaggerated (Prange et al. 1997). Thesea ice brine rejection/freshwater release and sea icetransport favor the resumption of the THC. Enhancedsea ice advection significantly favors the resumption ofthe Atlantic MOC. Because the sea ice transport is not

included, the sea ice brine rejection/freshwater releaseeffect on the THC is minor in Zhang et al. (1995),Lenderink and Haarsma (1996), and Lohmann andGerdes (1998). The strong restoring effect of sea ice saltrejection on the THC is observed in Dokken and Jansen(1999) and van Kreveld et al. (2000) for a glacial climate.In HadCM3, sea ice brine rejection is found to beresponsible for the sustaining of Greenland and Nor-wegian Sea deep convection in a THC perturbationexperiment (Vellinga et al. 2002). Another interestingfeature in this study is the role of ice sheet growth in theresumption of the Atlantic MOC. The effect of thesimulated ice sheet growth on the THC is relativelylarge. More rapid ice sheet growth has a more significantinfluence on the resumption of the Atlantic MOC.Therefore both enhanced sea ice transport and ice sheetgrowth are two important processes in the NH to restorethe THC. This has strong implications for glacial THCstability, because both sea ice and continental iceactivities are much stronger during a glacial period.

6 Concluding remarks

A new version of the MPM, namely, the MPM-2, isdescribed and employed in this paper. Following thestudy of Wang and Mysak (2002), a new solar energydisposition scheme has been developed by Wang et al.(2004) and is used to calculate the solar insolation ab-sorbed by the atmosphere and by the surface in theMPM. A global dynamic vegetation model, called VE-CODE, is coupled to the MPM (Wang et al. 2005). Themodel domain is extended from (75S, 75N) to (90S,90N). The surface wind is calculated following Petouk-hov et al. (2000). This new version MPM has no heatand freshwater flux adjustments.

The hysteresis diagram of the THC is first obtainedby slowly adding freshwater into the latitude band (20N,50N) of the North Atlantic or slowly extracting fresh-water from this region. The width of the hysteresis dia-gram is 0.27 Sv in the control run (0.32 Sv for the runwith suppressed ice sheet growth), and there exist twostable modes of the THC, the ‘on’ mode and the ‘off’mode in the MPM-2.

The two climate states, corresponding to the ‘on’ andthe ‘off’ modes of the THC under present-day forcing,are then compared. The shutting down of the THC leadsto a cooling in the NH and a warming in the SH, indi-cating a bipolar see-saw effect of the THC. Precipitationis decreased in the NH and increased in the SH. Thepoleward atmospheric sensible heat transport is in-creased in the NH and decreased in the SH. A relativelylarge change of poleward atmospheric moisture trans-port is confined to (40S, 40N), with an increase (a de-crease) in the NH (SH) subtropics and a decrease (anincrease) in the NH (SH) tropical region.

The oceanic heat transport in the Atlantic has alarge change associated with the THC mode change.For the ‘on’ mode, the Atlantic heat transport is

–0.3 –0.25 –0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15

0

10

20

30

40

50


Atla

ntic

MO

C in

tens

ity (

Sv)

Fig. 16 The hysteresis diagram of the Atlantic MOC for thecontrol run (solid line) and the run with a fixed vegetation (dashedline)


northward, with a peak value of 0.87 PW at 30N, whilefor the ‘off’ mode, the Atlantic heat transport almostvanishes in the NH and becomes directed southward inthe SH. Although the oceanic heat transport in theIndo-Pacific has an opposite change to that in theAtlantic, the zonally integrated oceanic heat transporthas a decreased poleward transport in the NH and anincreased poleward transport in the SH. This kind ofoceanic heat transport change is the origin of thebipolar see-saw effect (Crowley 1992; Stocker 1998;Broecker 1998), as seen in the SAT change shown inFigs. 2a and 3a. The large decrease of the oceanic heattransport in the North Atlantic causes a large drop ofthe SST in the middle and high latitudes of the NorthAtlantic and hence a significant southward advance ofsea ice. The heat loss from the ocean to the atmosphereis consequently decreased in the middle and high lati-tudes due to the thermal insulation effect of sea ice.Due to the increased sea ice melting, the oceanicfreshwater flux increases in the middle latitudes; mainlydue to the sea ice formation, the oceanic freshwaterflux decreases in the subpolar region.

The shutting down of the THC causes an increase ofthe annual mean NH sea ice area from 9.4·106 km2 to14.4·106 km2 and a decrease of the annual mean SH seaice area from 13.3·106 km2 to 7.4·106 km2, followingand amplifying the NH cooling and the SH warming.Permanent ice appears over North America due to thecooling caused by the THC shut down, ice-albedofeedback and vegetation-albedo feedback. However, thetotal ice volume is small (0.72·106 km3 at point C inFig. 1). The tree line retreats southward due to thecooling induced by the shutdown of the Atlantic MOC.The forest area increases in the SH subtropics due to theincrease of precipitation.

In order to identify feedbacks on the THC fromother components, various sensitivity experiments arecarried out. The overall hydrologic feedback favorsthe shut-down of the THC and also favors theresumption of the THC, but these effects are small inthe control run. The positive hydrologic feedback be-fore the shut-down results from the atmospherichydrologic cycle, while the negative feedback after theTHC shut-down results from the continental icegrowth and sea ice processes. The weak feedback ofthe atmospheric hydrological cycle on the THC isconsistent with many previous studies (Hughes andWeaver 1996; Lohmann and Gerdes 1998; Ganopolskiet al. 2001), but contrasts with a box model study(Nakamura et al. 1994).

Since there exist large uncertainties in the simulationof continental ice sheet growth and sea ice processes,further sensitivity experiments are carried out to inves-tigate the roles of rapid ice sheet growth and an en-hanced sea ice transport in the resumption of the THC.Rapid ice sheet growth (with a growth rate close to thatduring the last glacial inception, i.e., 2.026·106 km3 over1,000 years) significantly favors the THC resumption.Such an effect mainly results from the growth of conti-

nental ice, which causes a reduction in river runoff andhence reduces the oceanic freshwater flux in the subpolarregion of the North Atlantic. Enhanced sea ice transportalso significantly favors the THC resumption. With suchenhanced sea ice transport, salt rejection in the sea iceformation region is intensified and the freshwater releasein the sea ice melting region is also intensified. Obviouslysea ice forms in the higher latitudes and melts in thelower latitudes. Therefore, enhanced sea ice transport isequivalent to enhanced southward (northward) fresh-water (salt) transport, which certainly favors theresumption of the THC.

These findings have strong implications for the THCstability during a glacial period. The THC was believedto be shut down during Heinrich events in some studies(Seidov and Maslin 1999; Ruhlemann et al. 2004). Butthe THC recovered after each Heinrich event. Thereare probably many mechanisms that are responsible forthe resumption of the THC (for example, see Prange etal. 2004). But a possible regrowth of continental icesheets after each Heinrich event is a potential mecha-nism to restore the THC. This, of course, needs to beverified by further observational and modeling work.The role of sea ice salt rejection in the resumption hasbeen demonstrated in some recent paleoceanographicstudies (Dokken and Jansen 1999; van Kreveld et al.2000). During a glacial period, continental ice sheetscovered a much larger area and atmospheric CO2

concentration was much lower. Therefore, a glacialclimate was much colder. Much more sea ice formationmust happen in a much colder climate. Along withstronger wind, sea ice mass transport must be muchstronger, especially during a cold stadial phase or thephase of an ‘off’ mode THC. It is suggested in thispaper that enhanced sea ice formation and large sea icetransport could favor the THC resumption during aglacial period.

Note that these feedbacks are identified in one singleversion of the MPM-2. Model parameters have uncer-tainties. If different values of some key parameters (suchas oceanic diffusivities) are used, some other parametersneed to be re-tuned. Hence, multiple versions of theMPM-2 will be obtained. However, we believe that theidentified feedbacks are qualitatively robust. It is robustthat the shut-down of the Atlantic MOC leads tonorthern cooling and hence possible ice sheet growth,extended sea ice cover and enhanced southward sea icetransport, as well as a reduction of forest area in theNH.

The model structure of the MPM-2 is simple, rela-tive to comprehensive climate models. In particular, azonally averaged ocean model with a flat bottom isemployed. Thus the effects of complicated topographyin the deep water formation region on the AtlanticMOC stability cannot be addressed in this study. It isdifficult for a zonally averaged ocean model to resolvedetailed topographical effects. It is also a challenge forcomprehensive 3D ocean circulation models toparameterize the dense overflow in regions of steep


topography and so far there is no agreement on theeffect of the dense overflow on the Atlantic MOCstability (Lohmann 1998; Kosters et al. 2005). Futurework needs to done with respect to this issue. Aphysically based parameterization of the effects oftopography is especially needed for a zonally averagedocean model.

We also note that in earlier and also recent studies, ithas been suggested that physical processes in the SHhave strong influences on the Atlantic MOC (Toggweilerand Samuels 1993; Keeling and Stephens 2001; Shinet al. 2003; Weaver et al. 2003; Knorr and Lohmann2003; Peeters et al. 2004; Rohling et al. 2004). Particu-larly, recent studies suggested that the Antarctic ice sheetis actively associated with the re-organization of theAtlantic MOC (Weaver et al. 2003; Rohling et al. 2004).Freshwater discharge from the Antarctic continent canintensify or restore the Atlantic MOC. This issue is notaddressed in this study because in the MPM-2, (1) thesouthern ocean dynamics is too simple; (2) the resolutionis too low for both the Antarctic continent and thesouthern ocean (zonally averaged); and (3) the Antarcticice sheet is prescribed and hence it is difficult to addressthe ice discharge and re-growth.

Another limitation in a zonally averaged ocean modelis that the heat and salt transports by wind-driven gyrescannot be explicitly resolved. In some 3D coupledatmosphere–ocean–sea ice model studies (Schiller et al.1997; Vellinga et al. 2002; Timmermann and Goosse2004), it is suggested that the wind-driven gyres havelarge effects on an ‘off’ mode or significantly weakenedTHC. Also in Schiller et al. (1997), it was suggested thatthe intensification of the cyclone over the deep waterformation region after the shut-down of the AtlanticMOC enhances the upwelling and hence increases thesurface density, which favors the resumption of theAtlantic MOC.

Another unresolved issue for all kinds of oceanmodels is the role of mixing in the THC stability. Tur-bulent mixing is highly parameterized and constantdiffusivity parameters are generally prescribed in oceanmodels. Model studies using both a zonally averagedocean model and a 3D ocean model (Manabe andStouffer 1999; Schmittner and Weaver 2001; Oka et al.2001; Prange et al. 2003) suggest that both the THCintensity and stability are sensitive to the specified dif-fusivity parameters.

However, we suggest that even in the light of theabove caveats, the hydrologic processes in the NH,especially continental ice growth and sea ice salt rejec-tion/freshwater release and transportation, must be in-cluded in order for us to have a more completeunderstanding of the THC stability.

Acknowledgements Vladimir Petoukhov is gratefully acknowledgedfor helping us to calculate active winds in the MPM-2. The com-ments of L. A. Mysak and A. Ganopolski on a first draft of thispaper are much appreciated. The comments of two anonymousreviewers and the editorial assistance of Rebekah Kipp are alsogratefully acknowledged. This work was supported by a Discovery

Grant from the Canadian Natural Sciences and Engineering Re-search Council, and a Canadian Foundation for Climate andAtmospheric Sciences Project Grant awarded to L. A. Mysak.

References

Alley RB (2003) Paleoclimatic insights into future climate chal-lenges. Philos Trans R Soc Lond A 361:1831–1849

Broecker WS (1998) Paleocean circulation during the last deglaci-ation: a bipolar seesaw? Paleoceanography 13:119–121

Brovkin V, Bendtsen J, Claussen M, Ganopolski A, Kubatzki C,Petoukhov V, Andreev A (2002) Carbon cycle, vegetation andclimate dynamics in the holocene: experiments with theCLIMBER-2 model. Global Biogeochem Cycles 16(4):1139.DOI 10.1029/2001GB001662

Claussen M, Mysak LA, Weaver AJ, Crucifix M, Fichefet T,Loutre MF, Weber SL, Alcamo J, Alexeev VA, Berger A, CalovR, Ganopolski A, Goosse H, Lohmann G, Lunkeit F, MokhovII, Petoukhov V, Stone P, Wang Z (2002) Earth system modelsof intermediate complexity: closing the gap in the spectrum ofclimate system models. Clim Dyn 18:579–586

Cochelin ASB (2004) Simulation of glacial inceptions with the‘‘green’’ McGill Paleoclimate Model. MSc Thesis Available asCCGCR Report: 2004–2, McGill University, Montreal

Crowley TJ (1992) North Atlantic deep water cools the southernhemisphere. Paleoceanography 7:489–497

Crucifix M, Berger A (2002) Simulation of ocean–ice sheets inter-actions during the deglaciation. Paleoceanography 17(4). DOI10.1029/2001PA000702

Dokken TM, Jansen E (1999) Rapid changes in the mechanism ofocean convection during the last glacial period. Nature401:458–461

Ganachaud A, Wunsch C (2000) Improved estimates of globalocean circulation, heat transport and mixing from hydro-graphic data. Nature 408:453–457

Ganopolski A, Rahmstorf S (2001) Rapid changes of glacial cli-mate simulated in a coupled climate model. Nature 409:153–158

Ganopolski A, Petoukhov V, Rahmstorf S, Brovkin V, ClaussenM, Eliseev A, Kubatzki C (2001) CLIMBAR-2: a climate sys-tem model of intermediate complexity. Part II: model sensitiv-ity. Clim Dyn 17:735–751

Harvey LDD (1988) Development of a sea ice model for use inzonally averaged energy balance climate models. J Clim 1:1221–1238

Hughes TMC, Weaver AJ (1996) Sea surface temperature-evapo-ration feedback and the ocean’s thermohaline circulation.J Phys Oceanogr 26:644–654

Keeling RF, Stephens BB (2001) Antarctic sea ice and the controlof Pleistocene climate instability. Paleoceanography 16(1):112–131

Keigwin LD, Curry WB, Lehman SJ, Johnsen S (1994) The role ofthe deep ocean in North Atlantic climate change between 70and 130 kyr ago. Nature 371:323–326

Knorr G, Lohmann G (2003) Southern Ocean origin for theresumption of Atlantic thermohaline circulation during degla-ciation. Nature 424:532–536

Kosters F, Kase RH, Schmittner A, Herrmann P (2005) The effectof Denmark Strait overflow on the Atlantic meridional over-turning circulation. Geophys Res Lett 32(4). DOI 10.1029/2004GL022112

van Kreveld SA, Sarnthein M, Erlenkeuser H, Grootes P, Jung S,Nadeau MJ, Pflaumann U, Voelker A (2000) Potential linksbetween surging ice sheets, circulation changes and theDansgaard-Oeschger cycles in the Irminger Sea, 60-18 kyr.Paleoceanography 15:425–442

Krinner G, Mangerud J, Jakobsson M, Crucifix M, Ritz C,Svendsen JI (2004) Enhanced ice sheet growth in Eurasia owingto adjacent ice-dammed lakes. Nature 427:429–432

Lenderink G, Haarsma RJ (1996) Modeling convective transitionsin the presence of sea ice. J Phys Oceanogr 26:1448–1467


Lohmann G (1995) Stability of the thermohaline circulation inanalytical and numerical models. PhD Thesis, University ofBreman, Germany, p 128

Lohmann G (1998) The influence of a near-bottom transportparameterization on the sensitivity of the thermohaline circu-lation. J Phys Oceanogr 28:2095–2103

Lohmann G, Gerdes R (1998) Sea ice effects on the sensitivity ofthe thermohaline circulation. J Clim 11:2789–2803

Lohmann G, Gerdes R, Chen D (1996) Sensitivity of the thermo-haline circulation in coupled oceanic GCM-atmospheric EBMexperiments. Clim Dyn 12:403–416

MacdonaldAM (1998) The global ocean circulation: a hydrographicestimate and regional analysis. Prog Oceanogr 41:281–382

Manabe S, Stouffer RJ (1988) Two stable equilibria of a coupledocean-atmosphere model. J Clim 1:841–866

Manabe S, Stouffer RJ (1999) The role of thermohaline circulationin climate. Tellus 51A–B:91–109

Marshall SJ, Clarke GKC (1997) A continuum mixture model ofice stream thermomechanics in the Laurentide Ice Sheet, 1.Theory. J Geophys Res 102:20599–20613

Mikolajewicz U, Maier-Reimer E (1994) Mixed boundary condi-tions in ocean general circulation models and their influence onthe stability of the model’s conveyor belt. J Geophys Res99(C11):22633–22644

Nakamura M, Stone PH, Marotzke J (1994) Destabilization of thethermohaline circulation by atmospheric eddy transports.J Clim 7:1870–1882

Oka A, Hasumi H, Suginohara N (2001) Stabilization of thermo-haline circulation by wind-driven and vertical diffusive salttransport. Clim Dyn 18:71–83

Peeters FJC, Acheson R, Brummer GJA, de Ruijter WPM,Schneider RR, Ganssen GM, Ufkes E, Kroon D (2004) Vig-orous exchange between the Indian and Atlantic oceans at theend of the past five glacial periods. Nature 430:661–665

Petoukhov V, Ganopolski A, Brovkin V, Claussen M, Eliseev A,Kubatzki C, Rahmstorf S (2000) Climber-2: a climate systemmodel of intermediate complexity. Part I: model descriptionand performance for present climate. Clim Dyn 16:1–17

Prange M, Lohmann G, Gerdes R (1997) Sensitivity of the ther-mohaline circulation for different climates—investigations witha simple atmosphere-ocean model. Paleoclimates 2:71–99

Prange M, Romanova V, Lohmann G (2002) The glacial thermo-haline circulation: stable or unstable? Geophys Res Lett 29(21).DOI 10.1029/2002GL015337

Prange M, Lohmann G, Paul A (2003) Influence of vertical mixingon the thermohaline hysteresis: analyses of an OGCM. J PhysOceanogr 33:1707–1721

Prange M, Lohmann G, Romanova V, Butzin M (2004) Modellingtempo-spatial signatures of Heinrich events: influence of theclimatic background state. Quaternary Sci Rev 23(5–6):521–527

Rahmstorf S (1995) Bifurcation of the Atlantic thermohaline cir-culation in response to changes in the hydrological cycle. Nat-ure 378:145–149

Rahmstorf S, Ganopolski A (1999) Long-term global warmingscenarios computed with an efficient coupled climate model.Clim Change 43:353–367

Rahmstorf S, Willebrand J (1995) The role of temperature feed-back in stabilizing the thermohaline circulation. J Phys Ocea-nogr 25:787–805

Renssen H, Goosse H, Fichefet T (2003) On the non-linear re-sponse of the ocean thermohaline circulation to global defor-estation. Geophys Res Lett 30(2). DOI 10.1029/2002GL016155

Rohling EJ, Marsh R, Wells NC, Siddall M, Edwards NR (2004)Similar meltwater contributions to glacial sea level changesfrom Antarctic and northern ice sheets. Nature 430:1016–1021

Romanova V, Prange M, Lohmann G (2004) Stability of the glacialthermohaline circulation and its dependence on the backgroundhydrological cycle. Clim Dyn 22:527–538

Ropelewski CF (1989) Monitoring large-scale cryosphere/atmo-sphere interactions. Adv Space Res 9:213–218

Ruhlemann C, Mulitza S, Lohmann G, Paul A, Prange M, WeferG (2004) Intermediate depth warming in the tropical Atlanticrelated to weakened thermohaline circulation: combining pa-leoclimate data and modeling results for the last deglaciation.Paleoceanography 19(1). DOI 10.1029/2003PA000948

Sarnthein M et al (1994) Changes in east Atlantic deepwater cir-culation over the last 30,000 years: eight time slice reconstruc-tions. Paleoceanography 9:209–267

Schiller A, Mikolajewicz U, Voss R (1997) The stability of theNorth Atlantic thermohaline circulation in a coupledocean-atmosphere general circulation model. Clim Dyn13:325–347

Schmittner A, Weaver AJ (2001) Dependence of multiple climatestates on ocean mixing parameters. Geophys Res Lett 28:1027–1030

Schmittner A, Yoshimori M, Weaver AJ (2002) Instability of gla-cial climate in a model of the ocean-atmosphere-cryospheresystem. Science 295:1489–1493

Seidov D, Maslin M (1999) North Atlantic deep water circulationcollapse during the Heinrich events. Geology 27:23–26

Shin SI, Liu Z, Otto-Bliesner BL, Kutzbach JE, Vavrus SJ (2003)Southern Ocean sea-ice control of the glacial North Atlanticthermohaline circulation. Geophys Res Lett 30(2):1096. DOI10.1029/2002GL015513

Stammer D, Wunsch C, Giering R, Eckert C, Heimbach P, Mar-otzke J, Adcroft A, Hill CN, Marshall J (2003) Volume, heat,and freshwater transports of the global ocean circulation 1993–2000, estimated from a general circulation model constrained byWorld Ocean Circulation Experiment (WOCE) data. J GeophysRes 108(C1). DOI 10.1029/2001JC001115

Stocker TF (1998) The seesaw effect. Science 282:61–62Stocker TF, Wright DG (1991) Rapid transitions of the ocean’s

deep circulation induced by changes in surface water fluxes.Nature 351:729–732

Timmermann A, Goosse H (2004) Is the wind stress forcingessential for the meridional overturning circulation? GeophysRes Lett 31:L04303. DOI 10.1029/2003GL018777

Toggweiler JR, Samuels B (1993) Is the magnitude of the deepoutflow from the Atlantic Ocean actually governed by SouthernHemisphere winds? In: Heimann M (ed) The global carboncycle. Springer, Berlin Heidelberg New York, pp 303–331

Trenberth KE, Caron JM (2001) Estimates of meridional atmo-sphere and ocean heat transports. J Clim 14:3433–3443

Vellinga M, Wood RA, Gregory JM (2002) Processes governingthe recovery of a perturbed thermohaline circulation in Had-CM3. J Clim 15:764–780

Wang Z, Mysak LA (2000) A simple coupled atmosphere-ocean-sea ice-land surface model for climate and paleoclimate studies.J Clim 13:1150–1172

Wang Z, Mysak LA (2002) Simulation of the last glacial inceptionand rapid ice sheet growth in the McGill Paleoclimate Model.Geophys Res Lett 29(23). DOI 10.1029/2002GL015120

Wang Z, Hu R, Mysak LA, Blanchet JP, Feng J (2004) Aparameterization of solar energy disposition in the climatesystem. Atmos Ocean 42(2):113–125

Wang Y, Mysak LA, Wang Z, Brovkin V (2005) The greening ofthe McGill Paleoclimate Model. Part I: Improved land surfacescheme with vegetation dynamics. Clim Dyn 24:469–480

Weaver AJ, Saenko OA, Clark PU, Mitrovica JX (2003) Meltwaterpulse 1A from Antarctica as a trigger of the bolling-allerodwarm interval. Science 299:1709–1713

Weber SL (1998) Parameter sensitivity of a coupled atmosphere-ocean model. Clim Dyn 14:201–212

Zhang S, Lin CA, Greatbatch RJ (1995) A decadal oscillation dueto the coupling between an ocean model and a thermodynamicsea ice model. J Mar Res 53:79–106

Zhang YC, Rossow WB, Lacis AA (1995) Calculation of surfaceand top of atmosphere radiative fluxes from physical quantitiesbased on ISCCP data sets 1. Method and sensitivity to inputdata uncertainties. J Geophys Res 100:1149–1165


zhaomin wang two climatic states and feedbacks on ... climatic states.pdfstable modes for the...

Documents