Modelling carbon cycle feedbacks during abrupt climate change

Similar documents
Rapid Climate Change: Heinrich/Bolling- Allerod Events and the Thermohaline Circulation. By: Andy Lesage April 13, 2010 Atmos.

Welcome to ATMS 111 Global Warming.

LETTERS. Influence of the Thermohaline Circulation on Projected Sea Level Rise

Rapid climate change in ice cores

Deep Ocean Circulation & implications for Earth s climate

SUPPLEMENTARY INFORMATION

Continental Hydrology, Rapid Climate Change, and the Intensity of the Atlantic MOC: Insights from Paleoclimatology

Stability of Antarctic Bottom Water Formation to Freshwater Fluxes and Implications for Global Climate

Chapter 15 Millennial Oscillations in Climate

The ocean and abrupt climate change

Climate and Environment

6. What has been the most effective erosive agent in the climate system? a. Water b. Ice c. Wind

We re living in the Ice Age!

IPCC AR5 WG1 - Climate Change 2013: The Physical Science Basis. Nandini Ramesh

The Tswaing Impact Crater, South Africa: derivation of a long terrestrial rainfall record for the southern mid-latitudes

Modes of Global Climate Variability during Marine Isotope Stage 3 (60 26 ka)

/ Past and Present Climate

Ocean Mixing and Climate Change

Lecture 0 A very brief introduction

Ice on Earth: An overview and examples on physical properties

Fast and Slow Response of Sea ice and the Southern Ocean to Ozone Depletion

Lecture 1. Amplitude of the seasonal cycle in temperature

MAR 110 LECTURE #10 The Oceanic Conveyor Belt Oceanic Thermohaline Circulation

1 Carbon - Motivation

Natural and anthropogenic climate change Lessons from ice cores

Ruddiman CHAPTER 13. Earth during the LGM ca. 20 ka BP

Sensitivity of the Younger Dryas climate to changes in freshwater, orbital, and greenhouse gas forcing in CESM1.

Ocean & climate: an introduction and paleoceanographic perspective

Introduction to Quaternary Geology (MA-Modul 3223) Prof. C. Breitkreuz, SS2012, TU Freiberg

Impact of the Ocean s Overturning Circulation on Atmospheric CO 2

Speleothems and Climate Models

8. Climate changes Short-term regional variations

MERIDIONAL OVERTURNING CIRCULATION: SOME BASICS AND ITS MULTI-DECADAL VARIABILITY

ATMS 321: Natural Climate Variability Chapter 11

Thermohaline and wind-driven circulation

isopycnal outcrop w < 0 (downwelling), v < 0 L.I. V. P.

Weather & Ocean Currents

Sensitivity of the Present-Day Climate to Freshwater Forcing Associated with Antarctic Sea Ice Loss

A bit of background on carbonates. CaCO 3 (solid)

Please be ready for today by:

2. Meridional atmospheric structure; heat and water transport. Recall that the most primitive equilibrium climate model can be written

AMOC Impacts on Climate

RAPID TRANSITIONS OF THE THERMOHALINE OCEAN CIRCULATION

An Introduction to Coupled Models of the Atmosphere Ocean System

CORRELATION OF CLIMATIC AND SOLAR VARIATIONS OVER THE PAST 500 YEARS AND PREDICTING GLOBAL CLIMATE CHANGES FROM RECURRING CLIMATE CYCLES

An oceanic mechanism for glacial greenhouse gas fluctuations

Honours Thesis: Impacts of Latitude Shifts in the Southern. Ocean Westerly Winds on Past and Present. Climates

Orbital-Scale Interactions in the Climate System. Speaker:

The Transport Matrix Method (TMM) (for fast, offline simulation of passive tracers in the ocean) Samar Khatiwala

SIO 210 Final Exam December 10, :30 2:30 NTV 330 No books, no notes. Calculators can be used.

Stability and variability of the thermohaline circulation in the past and future: a study with a coupled model of intermediate complexity

Climate Change 2007: The Physical Science Basis

Presentation A simple model of multiple climate regimes

lecture 12 Paleoclimate

Early warning of climate tipping points Tim Lenton

A Broecker Brief Origin of the Atlantic s glacial age lower deep water

ATOC OUR CHANGING ENVIRONMENT Class 19 (Chp 6) Objectives of Today s Class: The Cryosphere [1] Components, time scales; [2] Seasonal snow

Chapter outline. Reference 12/13/2016

SIO 210 Final Exam Dec Name:

Ice core-based climate research in Denmark

Ocean oxygen isotope constraints on mechanisms for millennial-scale climate variability

Simulation of Long-Term Future Climate Changes with the Green McGill Paleoclimate Model: The Next Glacial Inception

An Investigation of Antarctic Circumpolar Current Strength in Response to Changes in Climate. Presented by Matt Laffin

On the Role of Wind-Driven Sea Ice Motion on Ocean Ventilation

1. Introduction 2. Ocean circulation a) Temperature, salinity, density b) Thermohaline circulation c) Wind-driven surface currents d) Circulation and

Jacob Schewe Potsdam Institute for Climate Impact Research. Ocean circulation under climate change: Examples of qualitative changes

Pathways in the ocean

Oceans and Climate. Caroline Katsman. KNMI Global Climate Division

A Seasonally Forced Ocean Atmosphere Model for Paleoclimate Studies

Meltdown Evidence of Climate Change from Polar Science. Eric Wolff

SIO 210 Final examination Answer Key for all questions except Daisyworld. Wednesday, December 10, PM Name:

Effect of the Drake Passage Throughflow on Global Climate

NATS 101 Section 13: Lecture 32. Paleoclimate

The Ice Age sequence in the Quaternary

A GCM Reconstruction of the Last Glacial Inception

Future risk of tipping points

ATOC OUR CHANGING ENVIRONMENT

Atmosphere-ocean interactions and dynamic response of the Southern Ocean to climate variability and trends. Mike Meredith BAS, Cambridge, UK

SIO 210 Final examination Wednesday, December 12, :30-2:30 Eckart 227 Name:

Today s Climate in Perspective: Hendrick Avercamp ( ) ~1608; Rijksmuseum, Amsterdam

The North Atlantic Oscillation: Climatic Significance and Environmental Impact

When Did the Anthropocene Begin? Observations and Climate Model Simulations

LETTERS. Tropical Stabilization of the Thermohaline Circulation in a Greenhouse Warming Simulation

Lecture 3 questions Temperature, Salinity, Density and Circulation

3. Global Warming Research Program

Understanding Paleoclimates ~ Modelling the Glacial/Interglacial Climate with Coupled GCM~

Development of the Global Environment

On the world-wide circulation of the deep water from the North Atlantic Ocean

The Effects of Geoengineering on the Southern Ocean

2/18/2013 Estimating Climate Sensitivity From Past Climates Outline

A box model test of the freshwater forcing hypothesis of abrupt climate change and the physics governing ocean stability

Sensitivity of the thermohaline circulation to tropical and high latitude freshwater forcing during the last glacial-interglacial cycle

Dynamical Paleoclimatology

G 3. AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES Published by AGU and the Geochemical Society

Coastal Antarctic polynyas: A coupled process requiring high model resolution in the ocean and atmosphere

Water mass formation, subduction, and the oceanic heat budget

SUPPLEMENTARY INFORMATION

Directed Reading. Section: Ocean Currents. a(n). FACTORS THAT AFFECT SURFACE CURRENTS

At it s most extreme very low pressure off Indonesia, every high off SA, ~8 o C difference over the Pacific and ½ m water level differential) ENSO is

M. Ballarotta 1, L. Brodeau 1, J. Brandefelt 2, P. Lundberg 1, and K. Döös 1. This supplementary part includes the Figures S1 to S16 and Table S1.

Transcription:

ARTICLE IN PRESS Quaternary Science Reviews 23 (24) 431 448 Modelling carbon cycle feedbacks during abrupt climate change Tracy L. Ewen a, *, Andrew J. Weaver a, Andreas Schmittner b a Department of Earth and Ocean Sciences, University of Victoria, P.O. Box 355, Victoria, BC, Canada V8W 3P6 b Institute of Geosciences, University of Kiel, Ludewig-Meyn-Str. 1, D-2498 Kiel, Germany Received 23 March 23; accepted 12 August 23 Abstract Past climate, both before and after the Last Glacial Maximum, was marked by a series of abrupt transitions from cold to warm states associated with significant changes in atmospheric CO 2 : Mechanisms which led to these transitions most likely include variability in the thermohaline circulation (THC) as inferred from deep sea sediment records. In this study, we investigate the changes in atmospheric CO 2 concentration that arise during abrupt climate change events. This is accomplished through our use of meltwater pulse scenarios applied to an ocean atmosphere sea ice model coupled to an inorganic carbon component. We perform transient simulations with increased freshwater discharge to high latitude regions in both hemispheres from a glacial equilibrium climate to simulate meltwater episodes. We find that changes in ocean circulation and carbon solubility lead to significant increases in atmospheric CO 2 concentrations when we simulate meltwater episodes in both hemispheres. The magnitude of increase in atmospheric CO 2 is between 1 and 4 ppmv; which accounts for some of the changes in CO 2 as recorded in the ice core records. r 23 Elsevier Ltd. All rights reserved. 1. Introduction Present-day atmospheric CO 2 concentrations continue to increase and climate model projections suggest a possibility, albeit remote, of this increase triggering an abrupt climate change event (Manabe and Stouffer, 1994; Stocker and Schmittner, 1997). Climatic variations in the past provide the opportunity to test these climate models and to improve our understanding of abrupt climate change. Although the mechanisms which have led to abrupt climate change in the past are still not fully understood, evidence exists that variability in the thermohaline circulation (THC) was involved in the rapid transitions between cold and warm states recorded in paleoclimate archives from the last glacial period (Stocker, 2; Clarket al., 22b). These so-called Dansgaard Oeschger (D O) oscillations (Dansgaard et al., 1984; Oeschger et al., 1984) have been recorded, for example, in the stable oxygen isotope (do 18 ; a proxy for air temperature) of polar ice. In Greenland ice cores, abrupt warming events of several degrees were recorded, which *Corresponding author. E-mail address: tracy@ocean.seos.uvic.ca (T.L. Ewen). tookplace in only a few decades or less (Grootes et al., 1993). These warming events, initiating the warm phase (interstadial) of the D O oscillation, were followed by periods of slow cooling which were sometimes terminated by an abrupt cooling event. The sequence of events often terminates with meltwater episodes (Bond and Lotti, 1995) during the coldest phase of the D O cycle (stadials), as evident in layers of ice-rafted debris found in marine sediments (Heinrich, 1988). Climate model simulations have shown that freshwater input resulting from meltwater discharges can reduce the overturning strength of the THC through changes in North Atlantic Deep Water (NADW) formation (Manabe and Stouffer, 1995; Clarket al., 22b; Schmittner et al., 22). At the centennial timescale, a reduction in the Atlantic overturning impacts the carbon balance through diminishing the transport of inorganic carbon from the surface to the deep ocean. Changes in sea ice cover, corresponding to changes in the poleward heat transport, also effect CO 2 uptake by changing air sea interaction in the high latitude regions. Associated changes in sea surface temperatures (SSTs), salinities (SSSs) and alkalinity further affect CO 2 solubility and uptake in both hemispheres. 277-3791/$ - see front matter r 23 Elsevier Ltd. All rights reserved. doi:1.116/j.quascirev.23.8.7

432 ARTICLE IN PRESS T.L. Ewen et al. / Quaternary Science Reviews 23 (24) 431 448 Atmospheric greenhouse gas concentrations can be reconstructed from measurements of air bubbles from polar ice cores. CO 2 concentrations have been reconstructed from southern hemisphere cores, e.g., Byrd Station (West Antarctic), Dome Concordia (Dome C) and a more recent high-resolution record from Taylor Dome (Stauffer et al., 1998; Inderm.uhle et al., 2). It is believed that Antarctic ice core records provide more reliable information about past CO 2 concentrations than Greenland ice core records as they have been less affected by acid-carbonate reactions which affect, for example, the GRIP record (Stauffer et al., 1998). In order to investigate the linkbetween abrupt climate change in the northern and southern hemispheres, GRIP CO 2 ice core records have been synchronized to the Byrd cores using methane synchronization techniques (Stauffer et al., 1998; Blunier and Brook, 21). These records reveal that atmospheric CO 2 concentrations have changed by up to 2 ppmv during strong D O oscillations (Stauffer et al., 1998; Inderm.uhle et al., 2). During the warm interstadial phases, CO 2 concentrations were relatively low (B2 ppmv); they increased by 1 2 ppmv during the stadials. The rapid warming events in Greenland led to a decrease of atmospheric CO 2 : Increases in atmospheric CO 2 between 2 and 7 ka BP parallel Antarctic warming events A1 through A4 (Inderm.uhle et al., 2) and seem to be connected with large iceberg discharges in the North Atlantic. Less pronounced D O events do not seem to correspond to significant variations in atmospheric CO 2 ; possibly because the associated freshwater events are too short or occur in a region which does not readily affect the THC. There is also no indication of warming in the southern hemisphere being recorded in ice cores during weaker D O events (Stocker, 2). Abrupt transitions have also occurred between the LGM and the Holocene as recorded in, for example, the GISP2 (Greenland) oxygen isotope records (Grootes et al., 1993). The Oldest Dryas cold period was followed by the B^lling Aller^d warm interval, in which a warming of several degrees tookplace over a few hundred years, bringing an end to the last glacial period. The B^lling Aller^d interstadial event (14,6 kyr BP in the GISP2 oxygen isotope record) was a rapid warming event which corresponds to the Antarctic Cold Reversal (ACR) in the Southern Ocean and a rapid increase in atmospheric CO 2 concentrations as seen in the Dome C CO 2 data (Monnin et al., 21). Although Vostokand Taylor Dome CO 2 data are thought to be the most accurate, the temporal resolution is too low to provide a detailed record of atmospheric CO 2 data over the last glacial termination (Monnin et al., 21). Coincident with the onset of the B^lling Aller^d was an abrupt increase in atmospheric CO 2 of about 8 ppmv in o3 yr (Monnin et al., 21). This interval is also thought to have coincided with a large meltwater pulse (mwp-1a) in the Southern Ocean (Clarket al., 22a; Weaver et al., 23). It is thought that the Southern Ocean played a key role in abrupt transitions of atmospheric CO 2 during the last glacial termination, however, the changes in the North Atlantic THC during mwp-1a probably also played a key role during that time. Through the addition of an inorganic ocean carbon cycle model into the UVic Earth System Climate Model (ESCM Weaver et al., 21), we examine changes to atmospheric CO 2 concentrations during abrupt climate change events both before and after the LGM. To accomplish this, we force the model with meltwater pulses in both the North Atlantic and Southern Oceans to replicate possible meltwater events which may have occurred. We use an inorganic ocean carbon cycle to study the feedbacks of the perturbations on atmospheric CO 2 concentrations through changes to the THC and carbon solubility and compare these to the variations from the paleo record. The outline of the rest of this paper is as follows: in the next section we provide an overview of the UVic ESCM version used here. In Section 3 we examine the stability of the THC with the introduction of linearly varying freshwater input into the North Atlantic. We then investigate the changes in the ocean atmosphere carbon system when we introduce freshwater at a volume equivalent to a potential Heinrich event. With the introduction of freshwater into the North Atlantic, there is a decrease in NADW formation and a reduction of carbon transported to the deep ocean. This leads to an overall reduction in uptake over the global oceans and a increase in atmospheric CO 2 : In Section 3.1 we examine the effect of the magnitude of the freshwater perturbations on atmospheric CO 2 : In Section 4 we examine the sensitivity of atmospheric CO 2 to freshwater events introduced into the Southern Ocean. We find that the introduction of freshwater into a region off the South American continent causes a slight increase in NADW circulation. This leads to an increase in CO 2 uptake in the region of NADW formation. Nevertheless, this increase is smaller than the decrease in uptake associated with changes in solubility in the Southern Ocean, thereby leading to an overall increase in atmospheric CO 2 : In Section 4.1 we examine the role of the representation of sea ice on the uptake of atmospheric CO 2 : The results from experiments which include either dynamic-thermodynamic or thermodynamic-only sea ice models are compared. In Section 5 we examine the sensitivity of atmospheric CO 2 to freshwater perturbations when starting from different equilibria obtained using a variety of prescribed atmospheric CO 2 levels. Finally, our results are discussed and summarized in Section 6.

ARTICLE IN PRESS T.L. Ewen et al. / Quaternary Science Reviews 23 (24) 431 448 433 2. Model description and initialization The ocean component of the ESCM is a threedimensional ocean general circulation model (Pacanowski, 1995) with a resolution of 3:6 zonally and 1:8 meridionally. There are 19 vertical levels in the ocean which vary from 5-m thickness at the surface to 518-m at the ocean bottom. Coupled to the ocean model is a vertically integrated energy-moisture balance model of the atmosphere with no explicit dynamics and in which atmospheric transports of moisture and energy are parametrized using Fickian diffusion. A dynamicthermodynamic sea ice component (Bitz et al., 21) is also employed. The inorganic carbon component is based on the Ocean Carbon Model Intercomparison Project (OCMIP) implementation. Detailed descriptions of each model component as well as the glacial and present-day climatology are available elsewhere (Weaver et al., 21). The solubility component allows atmospheric CO 2 concentrations to vary based on the integrated air sea carbon fluxes, and assumes that the atmosphere is instantly well mixed (see Ewen et al., 24, for more details). This interaction between the oceanic and atmospheric CO 2 allows us to perform transient experiments determining the evolution of atmospheric CO 2 : The evolving atmospheric CO 2 changes the outgoing long-wave radiation in our model which allows assessment of its climatic feedback. In this study, we are interested in the response of the inorganic carbon cycle to freshwater perturbations and have not included biological processes involved in the carbon cycle. Although this neglects potential feedbacks associated with ocean biology in response to these perturbations, the biological response to resulting changes in ocean physics is poorly understood. During meltwater events, biological productivity in high latitude regions may be affected by changes to water column stratification and sea ice extent. Iron supply is thought to limit Southern Ocean productivity and CO 2 uptake (Martin, 199). A reduction in iron through upwelling may result in changes to CO 2 uptake where freshwater input leads to stratification changes. Fran@ois et al. (1997) suggest that surface water stratification may contribute to increased CO 2 uptake in the Southern Ocean through changes in nutrient availability. Changes to sea ice extent may also impact the productivity due to light availability and further changes to stratification in this region. Although we will not address the biological response to meltwater episodes in our sensitivity studies, we show that significant changes to water column properties and deep water formation occur in the high latitude regions. To begin, we start with a Last Glacial Maximum (LGM) equilibrium obtained using 21 kbp (thousand years before present) orbital parameters, an atmospheric CO 2 concentration of 2 ppmv and elevated topography based on the Peltier (1994) continental ice sheet reconstruction. The carbon flux map for this glacial equilibrium is shown in Fig. 1. The main features are similar in pattern to uptake in the preindustrial climate with an atmospheric CO 2 concentration of 28 ppmv (shown in Weaver et al., 21). High fluxes of carbon out of the ocean occur in warm equatorial regions and Annual mean air-sea CO 2 flux 6 N 6 S 15 E 6 W 9 E -3.5-1.8 -.1 1.6 3.3 5. mol m -2 yr -1 Fig. 1. Map of annual mean air sea CO 2 flux (mol m 2 yr 1 ) for the glacial equilibrium climate. The initial condition for this experiment was an equilibrium state with a prescribed atmospheric CO 2 concentration of 2 ppmv and 21 kyr BP orbital forcing. Here, a dynamic-thermodynamic representation of sea ice and the Gent and McWilliams (1999) parametrization for mixing associated with mesoscale eddies were used.

434 ARTICLE IN PRESS T.L. Ewen et al. / Quaternary Science Reviews 23 (24) 431 448 the Southern Ocean has the highest fluxes of carbon into the ocean. Uptake in the northern hemisphere is weaker than in the preindustrial equilibrium, although there is still a significant amount of uptake in the North Atlantic. The overturning strength in the North Atlantic has a maximum overturning of about 5 Sv; much less than present day and less than LGM overturning in Weaver et al. (21). The sea ice edge in the North Atlantic is further south and NADW forms further south which, combined with a weaker overturning, leads to less uptake in the North Atlantic in the LGM run. In this version of the UVic model, we incorporate an improved parametrization of mixing associated with mesoscale eddies (Gent and McWilliams, 1999) which leads to a somewhat weaker overturning (both for the LGM and the present-day) than in Weaver et al. (21). The coefficients for both the isopycnal diffusion and isopycnal thickness diffusion coefficients are 2 1 7 cm 2 s 1 and the maximum slope of the isopycnals is.1. The fact that the LGM overturning might be weaker than inferred from LGM proxy records is inconsequential to our analysis as our LGM equilibrium simply serves as an initial starting point from which we conduct numerous sensitivity studies. That is, in the subsequent analysis, we undertake numerous perturbation experiments, including the tracing out of the hysteresis behaviour of the THC. These experiments were designed to examine the response of the climate system to abrupt transitions in the THC within the context of a variety of mean climatic states (i.e., cold: atmospheric CO 2 ¼ 2 ppmv; warm: atmospheric CO 2 ¼ 28 ppmv). 3. Sensitivity of atmospheric CO 2 to freshwater events in the North Atlantic D O events 8 and 12 correspond to significant variations of atmospheric CO 2 concentrations of up to 2 ppmv as revealed in Byrd and Taylor Dome ice core records from the Antarctic (Stauffer et al., 1998; Inderm.uhle et al., 2). MacAyeal (1993) estimated the freshwater flux from a corresponding Heinrich event discharged into the North Atlantic to be on the order of :16 Sv over 25 5 yr: Alternative reconstructions from ice sheet models (Marshall and Clarke, 1997) resulted in much smaller perturbations which have also been applied to modelling studies (Weaver, 1999). Here, we address the sensitivity of atmospheric CO 2 to freshwater perturbations applied to the North Atlantic. Meltwater experiments are carried out between 55 and 65 N: The volume, duration and rate of freshwater input are varied to cover potential observational based estimates of meltwater pulses (MacAyeal, 1993; Marshall and Clarke, 1997). Our initial condition, Table 1 Summary of the North Atlantic perturbation experiments Experiment Initial condition Forcing NA hys1 CO 2 ¼ 2 ppmv Linear: 7:6 Sv for 1 yr NA hys2 CO 2 ¼ 2 ppmv Linear: 7:8 Sv for 1 yr NA1 CO 2 ¼ 2 ppmv Linear: :42 Sv over 7 yr NA2 CO 2 ¼ 2 ppmv Constant: :35 Sv for 5 yr NA.15 CO 2 ¼ 2 ppmv Constant: :15 Sv for 5 yr NA.15 CO 2 ¼ 2 ppmv Constant: :15 Sv for 5 yr NA.2 CO 2 ¼ 2 ppmv Constant: :2 Sv for 5 yr NA.3 CO 2 ¼ 2 ppmv Constant: :3 Sv for 5 yr NA.35 CO 2 ¼ 2 ppmv Constant: :35 Sv for 5 yr NA16 CO 2 ¼ 16 ppmv Constant:.35 for 5 yr NA2 CO 2 ¼ 2 ppmv Constant:.35 for 5 yr NA22 CO 2 ¼ 22 ppmv Constant:.35 for 5 yr NA26 CO 2 ¼ 26 ppmv Constant:.35 for 5 yr The initial condition for each experiment was an equilibrium obtained under 21KBP orbital parameters and with the atmospheric CO 2 concentration fixed at the value given in column 2. The freshwater forcing is introduced at either a constant or linearly varying rate (column 3). described in the previous section, has NADW formation of B5 Sv; corresponding to a weakglacial state. The study by Marchal et al. (1998), using a zonally-averaged biogeochemical model, showed that when there is no active deep water formation this leads to Southern Ocean warming which affects the solubility of carbon and results in an increase in atmospheric CO 2 : To our knowledge, no study to date has used a coupled ocean atmosphere sea ice model which includes an ocean GCM and an inorganic carbon component to assess the role of circulation on uptake as well as solubility changes. Each North Atlantic freshwater perturbation experiment begins with an extraction of freshwater for the first 5 yr that results in an increase in deep water formation. From this state with active deep water formation, four separate experiments are then carried out. NA hys1 and NA hys2 (Table 1) have a linearly varying rate of freshwater applied in order to assess the hysteresis behaviour on the North Atlantic THC. Surface freshwater forcing is applied at a linear rate of :6 Sv per 1 yr for NA hys1, and :8 Sv per 1 yr for NA hys2 (Fig. 2a). The net freshwater forcing in NA1 and NA2 (Table 1) is larger than that of both NA hys1 and NA hys2 since NA1 and NA2 were designed to understand the response of the climate system to a hypothetical meltwater pulse event. In both NA1 and NA2, the same rate of negative freshwater forcing is applied as in NA hys1 over the first 5 yr: In the case of NA1, the negative freshwater forcing is then turned off ( Sv) and a positive freshwater perturbation is added which increases at the same rate as in NA hys1 (:6 Sv per 1 yr) for 7 yr (Fig. 2b). In the case of NA2,

ARTICLE IN PRESS T.L. Ewen et al. / Quaternary Science Reviews 23 (24) 431 448 435.4 3.3 25.2 Freshwater input (Sv).1.1.2 Overturning (Sv) 2 15 1.3 5 (a) Freshwater Input (Sv) (b).4 5 1 15 2 25 3 35.5.4.3.2.1.1.2.3.4.5 2 4 6 8 1 12 14 16 18 2 Fig. 2. (a) The linearly varying freshwater forcing applied to experiments NA hys1 (red) and NA hys2 (blue). (b) The linearly varying freshwater forcing applied to experiments NA1 (red) and NA2 (blue). Details of the experiments are given in Table 1. Positive freshwater forcing corresponds to freshwater input and negative forcing corresponds to freshwater extraction. freshwater is added at a constant rate of :35 Sv for 5 yr (Fig. 2b). Fig. 3 shows the hysteresis behaviour of the THC for NA hys1 and NA hys2 from year 5 onward, with the corresponding timeseries of the overturning strength shown in Fig. 4b. During the negative freshwater forcing in the first 5 yr; the overturning increases to a maximum of 28 Sv in NA hys2 and 26 Sv in NA hys1. Between year 5 and year 15, the overturning decreases until there is almost no deep water formation (o1 Sv). After the freshwater perturbation peaks at year 15, the overturning remains very weak(b1 Sv) and a threshold level of freshwater extraction of about :1 Sv must be reached before deep water formation becomes active again (Fig. 3)..4.3.2.1.1.2.3.4 Freshwater forcing (Sv) Fig. 3. Hysteresis curves showing the change in overturning strength (Sv) corresponding to a linearly varying change in freshwater forcing (Sv) introduced into the North Atlantic. The curves correspond to experiments NA hys1 (red) and NA hys2 (blue), where freshwater varies linearly at a rate of :6 Sv and.7 per 1 yr; respectively. The corresponding overturning strength for the NA1 and NA2 experiments is shown in Fig. 4d. The forcing and resulting overturning for the first 5 yr correspond to NA hys1. After year 5, there is an abrupt transition between freshwater extraction and freshwater input. There is a sharp decrease in overturning when freshwater is introduced and the overturning remains weak (o1 Sv) until the freshwater perturbation is stopped after year 1 for NA2 and year 12 for NA1. After this point, each experiment is run until year 19 with the freshwater forcing removed. Once the forcing is removed, the overturning slowly begins to increase, although it remains less than o3 Sv and never reaches the initial overturning strength of about B5Sv: In Figs. 4a and c the atmospheric CO 2 concentrations are shown for the NA hys1/na hys2, and NA1/NA2 experiments, respectively. Starting from the state with weak(b5 Sv) NADW formation, the overturning increases as freshwater is extracted from the North Atlantic and atmospheric CO 2 oscillates about the initial value of 2 ppmv in all experiments. When freshwater is added into the North Atlantic and the overturning strength begins to decrease, CO 2 correspondingly begins to increase after year 1 for NA hys1 and NA hys2, and year 5 for NA1 and NA2. For NA hys1 and NA hys2, the minimum in overturning strength and maximum concentration of atmospheric CO 2 occur after the maximum freshwater input (at year 15). The subsequent linearly decreasing freshwater flux causes the overturning to initially increase slightly (Fig. 4b), until the critical threshold forcing ( :1 Sv Fig. 3) is reached, whereupon the overturning quickly increases to 26 Sv in NA hys1 and 28 Sv in NA hys2, as before. A minimum in atmospheric CO 2 of about 197 ppmv is

436 ARTICLE IN PRESS T.L. Ewen et al. / Quaternary Science Reviews 23 (24) 431 448 28 3 26 25 Atmospheric CO 2 (ppmv) 24 22 2 Overturning (Sv) 2 15 1 198 5 (a) 196 5 1 15 2 25 3 35 (b) 5 1 15 2 25 3 35 216 3 Atmospheric CO 2 (ppmv) 214 212 21 28 26 24 22 2 Overturning (Sv) 25 2 15 1 5 (c) 198 2 4 6 8 1 12 14 16 18 2 (d) 2 4 6 8 1 12 14 16 18 2 Fig. 4. Change in atmospheric CO 2 concentration (ppmv) (a,c) and overturning strength (Sv) (b,d) for the (a,b) NA hys1 (red) and NA hys2 (blue) experiments and the (c,d) NA1 (red) and NA2 (blue) experiments. The linearly varying forcing for these experiments is shown in Fig. 2(a) for NA hys1 (red) and NA hys2 (blue) and Fig. 2(b) for NA1 and NA2. The hysteresis behaviour of the NA hys1 (red) and NA hys2 (blue) experiments is shown Fig. 3. Details of the experiments are given in Table 1. reached at about year 23 before it increases along the same path as from 5 to 1 yr: For the NA1 and NA2 experiments, the atmospheric CO 2 begins to increase as the freshwater input causes the overturning to decrease rapidly. As one mechanism (NADW formation) which transports CO 2 from the surface to deep ocean is turned off, CO 2 builds up in surface waters and consequently in the atmosphere. Although solubility effects will counteract these changes, there is still a significant increase in CO 2 from an equilibrium state of 2 ppmv to between 212 and 215 ppmv: Despite a similar evolution of the overturning in both experiments, there is a considerable difference in the response of atmospheric CO 2 : NA2, which has a constant freshwater flux applied, shows a much steeper increase than NA1, where the freshwater flux increases linearly. This suggests that the rate of change of the freshwater forcing strongly determines the atmospheric CO 2 response through dissolution as the transient change in overturning strength between the experiments is similar (Fig. 4d). Eventually, both experiments attain similar equilibria after the forcing is removed with atmospheric CO 2 concentration differing by less than 1 ppmv: A pronounced asymmetry is observed in all NA experiments as the increase of deep water formation during the first 5 years leads to only minor changes in atmospheric CO 2 ; while the reduction of the overturning after year 1 leads to an increase of 4 8 ppmv: When freshwater is extracted again after 2 yr in experiments NA hys1 and NA hys2, atmospheric CO 2 begins to decrease by about 3 ppmv: The first 5 years and between years 2 and 25 differ in that the initial condition at year is in equilibrium with an overturning

ARTICLE IN PRESS T.L. Ewen et al. / Quaternary Science Reviews 23 (24) 431 448 437 circulation of about 5 Sv: The second freshwater extraction beginning at year 2 continues from a state in which there is no deep water formation (o1 Sv). We would expect that both NA1 and NA2 would result in lower atmospheric CO 2 concentrations if there was another freshwater extraction applied after year 12 and year 1, respectively. The physical mechanisms which determine the carbon flux across the air sea interface involve both changes in circulation, in this case affecting the NADW, and solubility changes arising from changes in surface temperature, salinity, alkalinity and DIC. The rate of overturning affects the poleward heat transport, which in turn affects SSTs in the North Atlantic, and hence the uptake of CO 2 into the ocean. As the overturning strength weakens, less carbon is transported to the deep ocean and less CO 2 is taken up across the air sea interface. On the other hand, if the overturning weakens to the point where NADW stops, the southern hemisphere warms and the North Atlantic cools, which leads to large solubility changes. In the North Atlantic the reduction in uptake due to weak NADW formation is partially counteracted by increased solubility in cooler waters. In the Southern Ocean it is mostly changes in solubility that affect CO 2 uptake, although changes in downwelling in this region affect uptake slightly as well. On the other hand, when the overturning increases we would expect atmospheric CO 2 to decrease as more of it will be transported to depths which are not yet saturated in carbon and solubility changes can be expected to change the uptake in the North Atlantic as the water is much more saline and SSTs increase. Fig. 5 shows difference maps of surface DIC and the corresponding atmosphere ocean CO 2 flux after 5 and 1 yr for experiment NA2. After the first 5 yr; zonal mean DIC shows increased concentrations in the North Atlantic corresponding to an increase in CO 2 uptake associated with an increase in NADW formation (Fig. 5a). There is a southward transport of this DIC at depths above 2148 m as shown by the depth-integrated DIC shown in Table 2. The increased DIC signal reaches the surface waters around 5 S resulting in decreased uptake (Fig. 5b) as increased surface DIC increases the CO 2 surface partial pressure. DIC concentrations are also increased in the Southern Ocean at latitudes where Antarctic Intermediate Water (AAIW) forms. In the Southern Ocean, areas of both increased and decreased uptake occur but the deepest ocean, which is filled by Antarctic Bottom Water, is more depleted in DIC, with about 15 GtC less than the equilibrium run. Reduced ventilation of AABW is caused by the increase in sea ice cover in the Southern Ocean (Fig. 5c), as described in Schmittner (23), and probably accounts for the depletion of DIC below 2148 m (Table 2) as less CO 2 is transported to depths. This offsets the high uptake in the North Atlantic leading to the small net atmospheric CO 2 change. After 5 yr; the North Atlantic warms by about 5:5 C: The large (compared to the equilibrium uptake shown in Fig. 1) increase in CO 2 uptake of about 5 mol m 2 yr 1 (Fig. 5b) reflects the dominance of the increased overturning in this region over solubility effects. There is also a corresponding retreat in the sea ice edge (Fig. 5c) which allows more CO 2 to pass from the atmosphere to the ocean. The Southern Ocean, on the other hand, experiences a cooling, predominantly off the western and eastern coasts of South America (Fig. 5d). There is an overall increase in heat flux off the eastern coast extending to the southern tip of Africa and into the Indian Ocean (Fig. 5e) with a corresponding decrease in uptake there (Fig. 5b). A band from about 6 45 S has increased uptake of CO 2 ; although further south, there is a significant expansion of the sea ice edge (by almost 1 to about 5 S) and a resulting reduction in carbon uptake. With the addition of freshwater after year 5, the poleward heat transport begins to decrease until there is a shutdown of NADW formation and a surface temperature decrease of about 4: C (Fig. 5i). The northern hemisphere sea ice edge also moves further south (Fig. 5h) by almost 15 ; which leads to a large reduction in CO 2 uptake (Fig. 5g). The decrease in SSTs and increase in heat flux (Fig. 5j) extends from the equator into the North Atlantic where there is also an overall decrease in uptake. The reduction in overturning, and to a lesser extent the heat flux, together with the change in sea ice concentration, account for most of the changes in CO 2 flux after year 1. Compared to the initial condition, Table 2 shows that there is slightly less DIC below 2148 m (about 3 GtC compared to 15 GtC at year 5), and between 1257 and 2148 there is about 2 GtC more (compared to 3 GtC more at year 5). The surface waters, however, show a significant decrease in DIC of 32 GtC (compared to a gain of 11 GtC at year 5), resulting from an alkalinity dilution and a small decrease in the overturning strength. In the Southern Ocean there is a 2 C increase in SST (Fig. 5i), predominantly in the Atlantic between about 58 42 S off the south African coast. These warmer surface waters are associated with a retreat in the southern hemisphere sea ice edge (although there is an increase in the North Atlantic sea ice cover Fig. 5h), leading to an increase of almost 2 mol m 2 yr 1 in CO 2 uptake in this region (Fig. 5g). The net uptake, however, is negative, predominantly due to a large reduction in the North Atlantic of more than 5 mol m 2 yr 1 : This leads to an overall increase in atmospheric CO 2 of about 15 ppmv:

438 ARTICLE IN PRESS T.L. Ewen et al. / Quaternary Science Reviews 23 (24) 431 448 Zonal mean DIC for Atlantic (yr5 - yr).5 Zonal mean DIC for Atlantic (yr1 - yr).5 Depth in Metres 15 3 9.9 14.3 Level Depth in Metres 15 3 9.9 14.3 Level 45 17.7 45 17.7 (a) 6 2.7 9 S 9 N -23-14 -5 4 13 22 umol kg -1 Difference in air-sea CO 2 flux (yr5 - yr) (f) 6 2.7 9 S 9 N -22.6-173.2-125.7-78.2-3.7 16.8 umol kg -1 Difference in air-sea CO 2 flux (yr1 - yr) 6 N 6 N 6 S 6 S 15 E 6 W 9 E 9 E 12 W 3 E (b) -1..6 2.2 3.8 5.4 7. mol m -2 yr -1 (g) -5. -3.45-1.9 -.35 1.2 2.75 mol m -2 yr -1 Difference in ice concentration (yr5 - yr) Difference in ice concentration (yr1 - yr) 6 N 6 N 6 S 6 S 15 E 6 W 9 E 15 E 6 W 9 E (c) -.7263 -.4836 -.241.16.2442.4868 Ice Concentration (h) -.313 -.187 -.62.64.19.316 Ice Concentration Difference in SST (yr5 - yr) Difference in SST (yr1 - yr) 6 N 6 N 6 S 6 S 15 E 6 W 9 E 15 E 6 W 9 E (d) -2.5 -.9.7 2.3 3.9 5.5 Temperature ( C) (i) -4. -2.8-1.6 -.4.8 2. Temperature ( C) Difference in heat flux (yr5 - yr) Difference in heat flux (yr1 - yr) 6 N 6 N 6 S 6 S 15 E 6 W 9 E 15 E 6 W 9 E (e) -196. -138.8-81.6-24.4 32.8 9. Heat flux in W/m 2 ( j) -76.2-44.2-12.3 19.6 51.6 83.5 Heat flux in W/m 2 Fig. 5. Changes obtained from experiment NA2 relative to the LGM initial condition (year ): (a,f) zonal mean DIC in the North Atlantic; (b,g) air sea CO 2 flux; (c,h) sea ice concentration; (d,i) SST; (e,j) heat flux. Panels (a) (e) show differences between year and year 5 (i.e., after 5 years of freshwater extraction), whereas panels (f) (j) show differences between year and year 1 (i.e., once a positive freshwater forcing has been introduced).

ARTICLE IN PRESS T.L. Ewen et al. / Quaternary Science Reviews 23 (24) 431 448 439 Table 2 Integrated DIC (GtC) over different depths at equilibrium (year ) and after 5 (year 5) and 1 (year 1) years of freshwater forcing in the North Atlantic experiment NA2 Depth (m) year year 5 year 1 1257 949 942 9377 1257 2148 651 6513 6512 2148 5137 145 1435 1447 3.1. Effects of freshwater pulse magnitude on atmospheric CO 2 To address the effects of varying the magnitude of the freshwater perturbation on atmospheric CO 2 concentrations, we apply continuous freshwater forcing at varying levels between.15 and :35 Sv: All experiments were carried out as in NA2, where a freshwater perturbation is introduced at a constant rate after year 5. The five separate experiments are referred to as NA.15, NA.15, NA.2, NA.3 and NA.35, corresponding to a freshwater perturbation of.15,.15,.2,.3 and :35 Sv; respectively (Table 1). As might be expected, the magnitude of the freshwater perturbation has a large effect on the final atmospheric CO 2 concentration (Fig. 6). NA.15 yields a maximum change of only 2 ppmv after the first 15 yr of freshwater input, which subsequently begins to decrease until it reaches 21 ppmv: When the volume of freshwater input is increased to :15 Sv over 5 yr; the maximum in atmospheric CO 2 is reached after 175 yr before decreasing slightly to a final value of about 25 ppmv: At a volume input of :2 Sv over 5 yr; a maximum is reached after 175 yr and then the level of atmospheric CO 2 remains relatively constant at about 27 ppmv to year 5. NA.15, NA.15 and NA.2 all have a minimum overturning strength of > 1Sv after the first 1 yr of freshwater input. Both NA.35 and NA.3, however, have minimum overturning strengths o1 Sv after the first 1 yr: This reduction in overturning, to a point where NADW is no longer forming, leads to a threshold wherein atmospheric CO 2 concentrations continue to increase. In Fig. 6, both experiments NA.3 and NA.35 show a sharp increase in atmospheric CO 2 over the first 12 yr and then CO 2 continues to increase, but at a lower rate, over the remaining 5-yr period of freshwater input. After 5 yr; the atmospheric CO 2 is 212 ppmv for NA.3 and 215 ppmv for NA.35. These experiments suggest that in order to have any significant increase in atmospheric CO 2 there must be a significant reduction in NADW formation (to less than 1 Sv in this case) or a strong dilution of surface waters. These experiments have also been carried out with a linear increase of freshwater input (corresponding to similar total volumes of water above) yielding a similar overturning threshold Atmospheric CO 2 (ppmv) 216 214 212 21 28 26 24 22 2 5 1 15 2 25 3 35 4 45 5 Fig. 6. Atmospheric CO 2 concentrations for freshwater perturbations in the North Atlantic. The experiments correspond to NA.15 (green), NA.15 (pink), NA.2 (blue), NA.3 (red) and NA.35 (black) with initial conditions and forcing given in Table 1. for continued positive atmospheric CO 2 accumulation to year 5. 4. Sensitivity of atmospheric CO 2 to freshwater events in the Southern Ocean Recent sea-level fingerprinting (Clarket al., 22a) has shown that the Laurentide Ice Sheet could not have been the only source of mwp-1a and that a large part probably came from the Antarctic Ice Sheet. In a modelling study, Weaver et al. (23) show that if mwp- 1A originates from the Antarctic ice sheet, rather than from the Laurentide or Fennoscandian ice sheets, an increase in NADW takes place, increasing poleward heat transport and providing an increase in temperatures to sufficiently explain the onset of the B^lling Aller^d. Using a model version similar to Weaver et al. (23) with initial conditions as described in the previous section and the addition of an inorganic solubility model, we introduced a meltwater pulse of the same magnitude and location in the Southern Ocean in order to study changes in atmospheric CO 2 concentrations corresponding to a meltwater pulse of southern hemisphere origin. Monnin et al. (21) show that there were significant changes in atmospheric CO 2 concentrations during the last glacial termination as seen in Dome C CO 2 records. Corresponding to the onset of the B^lling Aller^d interval was a jump of about 8 ppmv in the record. In a study using a zonally averaged coupled model, Marchal et al. (1998) found that the influence of Southern Ocean freshwater forcings did not significantly alter atmospheric CO 2 in their model. In our study, we find large changes in atmospheric CO 2 concentrations

44 ARTICLE IN PRESS T.L. Ewen et al. / Quaternary Science Reviews 23 (24) 431 448 corresponding to freshwater forcings, which may explain some of the variation seen in the ice core record. Freshwater inputs in the Southern Ocean were applied just north of the Antarctic circumpolar current (ACC). We chose the region at the southern tip of South America where AAIW forms in our model (Saenko et al., 22). This region spans a small part of Drake passage extending both into the Atlantic and Pacific Oceans (region B in Weaver et al., 23). An initial freshwater perturbation was applied to the North Atlantic for 2 yr so that there was no active NADW formation as in Weaver et al. (23). The end of this experiment was subsequently used as the initial condition for experiments SO1b and SO2. Fig. 7 shows the transient evolution of the atmospheric CO 2 concentration (Fig. 7a) and overturning strength (Fig. 7b) for three separate experiments. In SO1a and SO1b (Table 3), a linearly increasing freshwater pulse was applied at a rate of :3 Sv per 5 years (Fig. 7c). In SO2 this was increased to 1 Sv per 5 years: SO1a begins from the equilibrium LGM climate, described in Section 2, and SO1b and SO2 use the initial conditions described above in which there is no NADW formation. In the SO1a and SO1b cases, the differences in initial conditions account exclusively for the differences in the evolution of atmospheric CO 2 concentration between year and year 5 (Fig. 7a). In the SO1b run, there is a maximum concentration of 213 ppmv after 5 yr; whereas in SO1a a maximum concentration of 28 ppmv occurs after only about 25 yr: At year 5 the freshwater forcing is removed (Fig. 7c) and a quasi-equilibrium is reached in experiments SO1a and SO1b after 15 yr; at which point the atmospheric CO 2 concentrations converges to 23 ppmv: The Atlantic overturning in SO1b increases from a state with no deep water formation (o1 Sv) to B3 Sv; whereas the overturning in SO1a increases slightly in the first 1 yr; and then decreases and remains constant (at B4 Sv) by 5 yr (Fig. 7b). In both SO1a and SO1b, the freshwater forcing and initial atmospheric CO 2 concentrations are the same. The uptake in the North Atlantic is not significantly different between the two experiments, although small changes in the overturning may contribute slightly to the minor difference in atmospheric CO 2 (Fig. 7a). The largest differences in uptake, however, are due to changes related to the southern sea ice edge between the two runs. SO1b has larger ice concentrations primarily in the southern hemisphere. There is also a slight warming in the Northern Hemisphere, both through increased SSTs and surface air temperatures (SATs), which occurs mainly in the Pacific Ocean. There is a cooling in the southern hemisphere, predominantly in the region where the freshwater flux is applied, and an increase in the southern sea ice edge Atmospheric CO 2 (ppmv) (a) Overturning (Sv) (b) Freshwater forcing (Sv) (c) 235 23 225 22 215 21 25 2 5 1 15 5.5 5 4.5 4 3.5 3 2.5 2 1.5 1.5 5 1 15 1.2 1.8.6.4.2 -.2 -.4 5 1 15 Fig. 7. (a) Atmospheric CO 2 concentration (ppmv) and (b) overturning strength (Sv) corresponding to freshwater perturbations applied to the Antarctic Intermediate Water formation region in Southern Ocean (SO). (c) The timeseries of the freshwater forcing for each experiment with positive forcing indicating freshwater addition. The curves correspond to SO1a (red), SO1b (pink) and SO2 (blue) with forcing and initial conditions outlined in Table 3.

Table 3 Summary of the Southern Ocean perturbation experiments ARTICLE IN PRESS T.L. Ewen et al. / Quaternary Science Reviews 23 (24) 431 448 441 Experiment Initial condition Forcing SO1a equil, CO 2 ¼ 2 ppmv Linear: :3 Sv over 5 yr SO1b trans, CO 2 ¼ 2 ppmv Linear: :3 Sv over 5 yr SO2 trans, CO 2 ¼ 2 ppmv Linear: 1 Sv for 5 yr SO16 equil, CO 2 ¼ 16 ppmv Constant: 1 Sv for 5 yr SO2 equil, CO 2 ¼ 2 ppmv Constant: 1 Sv for 5 yr SO22 equil, CO 2 ¼ 22 ppmv Constant: 1 Sv for 5 yr SO26 equil, CO 2 ¼ 26 ppmv Constant: 1 Sv for 5 yr SOp nodyn pi-equil, CO 2 ¼ 28 ppmv; thermo ice Constant: 1 Sv for 5 yr SOp dyn pi-equil, CO 2 ¼ 28 ppmv Constant: 1 Sv for 5 yr SOn nodyn pi-equil, CO 2 ¼ 28 ppmv; thermo ice Constant: 1 Sv for 5 yr SOn dyn pi-equil, CO 2 ¼ 28 ppmv Constant: 1 Sv for 5 yr The initial conditions correspond to either an equilibrium LGM climate (equil), an equilibrium preindustrial climate (pi-equil) or a transient LGM run (trans). Orbital parameters for the equil and trans are for 21 kyr BP and for pi-equil they are for 185. The trans experiments were integrated from the equil initial condition with freshwater forcing introduced into the North Atlantic over 2 years so that the NADW formation is ceased. Atmospheric CO 2 concentrations are initially prescribed for the equilibrium runs and then allowed to evolve over time. All experiments were carried out with the inclusion of a dynamic-thermodynamic sea ice model except for a few cases where only a thermodynamic sea ice model was used (thermo ice).freshwater forcing is introduced at either a constant or linear varying rate. around Antarctica, except south of Africa, where there is reduced carbon uptake of about 5 mol m 2 yr 1 : SO2 corresponds to a more dramatic increase in atmospheric CO 2 of 33 ppmv (Fig. 7a). There is initially no NADW formation in SO2, although with the introduction of freshwater into the Southern Ocean, it gradually increases until it peaks just over 4 Sv (Fig. 7b). While a positive freshwater forcing applied to the North Atlantic causes a reduction in NADW formation, a positive forcing introduced into the Southern Ocean causes the opposite effect and the rate of NADW formation increases. An increase in NADW can happen whenever the potential density of NADW (r NADW ) is greater than the potential density of AAIW (r AAIW ), as described in Weaver et al. (23). This can happen when either a positive freshwater perturbation is applied to a region in the Southern Ocean or a negative perturbation is applied to a region in the North Atlantic. In SO2, the initial state has reduced NADW formation so that r AAIW > r NADW ; but when we apply a positive freshwater perturbation to the region in the Southern Ocean, the potential density difference becomes, r NADW > r AAIW and NADW formation increases. The increase in overturning strength, however, is not very large due to the weaklgm equilibrium which has an overturning strength of around 5 Sv: After the maximum is reached, the overturning quickly reduces again once the freshwater perturbation in the Southern Ocean is removed (year 5). After a further 1 yr of integration, the atmospheric CO 2 levels reach quasiequilibrium. Although the overturning in SO2 and SO1b is similar, the rate of freshwater input and corresponding changes in atmospheric CO 2 are quite different. The steep increase in atmospheric CO 2 in SO2 corresponds to a small increase in overturning strength in the North Atlantic. Fig. 8 shows difference plots for the SO2 experiment between year 5 (corresponding to the end of the Southern Ocean freshwater forcing) and year (corresponding to a state in which there is no NADW formation Fig. 7b). In Section 3 we found that a large increase in overturning (of B2 Sv) did not lead to a significant change in overall uptake (although there was a large regional increase in the North Atlantic). Here, we apply the freshwater perturbation to the Southern Ocean and dramatic differences in uptake occur in this region. The large increase in uptake (Fig. 8f) in the North Atlantic is largely due to a reduction in sea ice cover (Fig. 8e), which in turn is a result of an increase in the northward heat transport at year 5 relative to the initial state with no NADW formation (year ). There is also an overall increase in SSTs across the northern hemisphere with a maximum increase of about 5 C occurring in the eastern North Atlantic Ocean and Mediterranean Sea (Fig. 8d). Cooling in the southern hemisphere occurs in two regions: off the South American continent; in the southern Indian Ocean where surface air temperatures are also lower by B4 C). Decreased SSTs off the coast of South America have a corresponding heat flux signature (Fig. 8c) and result from the increase in heat transport away from this region into the North Atlantic. There is a net downward heat flux due to cooler SSTs and changes in ice concentration along the Southern Ocean leading to an overall reduction in carbon uptake by the ocean with an average of about 4 mol m 2 yr 1 (Fig. 8f). Both SAT and SST are reduced off the southern coast of South America where the maximum decrease in uptake in this region reaches almost 2 mol m 2 yr 1 :

ARTICLE IN PRESS 442 T.L. Ewen et al. / Quaternary Science Reviews 23 (24) 431 448 Zonal mean DIC (yr5- yr) Difference in surface DIC (yr5- yr).5 6 N 15 9.9 Depth in Metres 3 14.3 Level 45 17.7 6 S 15 E 6 W 9 E 6 2.7 9 S 9 N (a) -436.9-32.8-168.7-34.6 99.5 233.7 umol kg -1 (b) -144.4-8. -15.6 48.7 113.1 177.5 umol kg -1 Difference in heat flux (yr5 - yr) Difference in SST (yr5- yr) 6 N 6 N 6 S 6 S 15 E 6 W 9 E 15 E 6 W 9 E (c) -87.1-44.1-1.2 41.7 84.7 127.6 Heat flux in W/m 2 (d) -3.21-1.49.23 1.95 3.67 5.39 Temperature ( C) Difference in ice concentration (yr5- yr) Difference in CO 2 flux (yr5- yr) 6 N 6 N 6 S 6 S 15 E 6 W 9 E 15 E 6 W 9 E (e) -.3729 -.19 -.71.1757.3586.5414 Ice Concentration (f ) -25. -19.2-13.4-7.6-1.8 4. mol m -2 yr -1 Fig. 8. Changes obtained from experiment SO2 at year 5 relative to the initial condition (year ): (a) surface DIC; (b) zonal mean North Atlantic DIC; (c) heat flux; (d) SST; (e) sea ice concentration; (f) air sea CO 2 flux. Year corresponds to a transient initial condition (Table 3) once formation of NADW is ceased, and year 5 corresponds to the end of the application of the positive freshwater perturbation to the Southern Ocean. The strength of the overturning at year and year 5 and the prescribed freshwater forcing are shown in Fig. 7b and c. Overturning in this region becomes much shallower and slightly weaker and the maximum depth of the overturning cell is reduced from about 37 to 21 m: The zonal mean DIC in the Atlantic (Fig. 8b) shows a large surface decrease in the surface region of the overturning cell (between 74 63 S) and the entire deep basin has an overall negative change in the DIC concentration. The entire water column is depleted in DIC after 5 yr by almost 74 GtC (Table 4). The surface layer to 1257 m is reduced by 15 GtC whereas the deeper layers are reduced by about 3 GtC each. Uptake off the southern coast of South America is also affected by changes in alkalinity due to dilution, which leads to a much more significant local effect on the uptake than either changes in the overturning or heat flux. When globally averaged, there is a large positive outgassing of carbon to the atmosphere which results in an increase in atmospheric CO 2 concentrations of 33 ppmv:

ARTICLE IN PRESS T.L. Ewen et al. / Quaternary Science Reviews 23 (24) 431 448 443 Table 4 Integrated DIC (GtC) after 5 yr of freshwater input in the Southern Ocean (year 5) Depth (m) year year 5 1257 9362 9348 1257 2148 658 6478 2148 5137 1449 1419 Integrated DIC for the equilibrium run is shown in Table 2. year shows values after 2 yr of freshwater input into the North Atlantic corresponding to transient (trans) runs in Table 3. 4.1. The role of sea ice on atmospheric CO 2 changes In this section, we compare oceanic carbon uptake obtained using a dynamic-thermodynamic sea ice model to that obtained using a thermodynamic only model when freshwater perturbations (both positive and negative) are applied in the same region in the Southern Ocean. The thermodynamic sea ice model is based on Semtner (1976) and Hibler (1979) and a elastic-viscousplastic rheology (Hunke and Dukowicz, 1997) is used to represent dynamics. Similar experiments (not shown) have also been carried out in the northern hemisphere although little variation in atmospheric CO 2 was found due to the differences in sea ice physics. For these experiments we use, as an initial condition, a preindustrial equilibrium climate with an equilibrium CO 2 concentration of 28 ppmv and orbital parameters for the calendar year 185. Two experiments were carried out using each representation of sea ice. Both positive and negative freshwater forcing were applied in each case in the same region as in experiment SO2 with a rate of 1 Sv for 5 yr (again the same as in SO2). The runs with dynamical sea ice will be referred to as SOp dyn (positive forcing) and SOn dyn (negative forcing). Similarly, the runs with only thermodynamic sea ice are referred to as SOp nodyn and SOn nodyn (Table 3). Unlike SO2, a preindustrial climate was used for the initial state and freshwater was not added to initially reduce NADW formation. The initial overturning strength for SO-dyn is 14:2 Sv and for SOnodyn 11:5Sv: Fig. 9(b) and (c) shows the difference in atmospheric CO 2 and in overturning between SOp dyn and SOp nodyn. Both experiments driven by a positive freshwater perturbation yield slightly increasing overturning strengths in the North Atlantic over the first 1 yr; and a subsequent decrease (Fig. 9c). The overall increase in atmospheric CO 2 is large for both runs: starting at 28 ppmv SOp dyn increases by B48 ppmv and reaches 328 ppmv; SOp nodyn increases by B4 ppmv: The most significant differences in CO 2 uptake occur where there are changes in the sea ice edge (Fig. 9a). With the inclusion of sea ice dynamics there is a lower concentration of sea ice along the Antarctic continent and in the North Atlantic, as this ice is transported equatorward by the winds. With the movement of the ice edge in both hemispheres, there is less overall uptake (up to 2:5 molm 2 yr 1 ) in SOp dyn compared to SOp nodyn (Fig. 9a). This leads to an atmospheric CO 2 increase of about 8 ppmv in SOp dyn relative to SOp nodyn. When the Southern Ocean is perturbed with a similar negative freshwater flux (extraction), there is a significant increase in uptake in the Southern Ocean (Fig. 9d), which is also reflected in the decrease in atmospheric CO 2 (Fig. 9e). The overturning reduces from about 14:2 Sv in SOn dyn and 11:5 Sv in SOn nodyn to o1 Sv in each run (Fig. 9f), and there is an increase in AAIW formation. In this case the negative freshwater flux causes the potential density difference to become, r AAIW > r NADW and formation of NADW ceases. Since AAIW formation increases, the uptake of CO 2 in this region also increases and there is an overall reduction in atmospheric CO 2 from 28 to 256 ppmv in both SOn experiments. Differences in uptake between SOn dyn and SOn nodyn occur almost entirely in the Southern Ocean and are due to differences in the sea ice edge (Fig. 9d). There is a thin band of sea ice relocation equatorward in SOn dyn off the western tip of South America and extending to the western side of the Ross Sea. There is a negative difference in uptake in this region of B3 mol m 2 yr 1 : Off the south African coast and along the Antarctic continent to the south there are also regions with a negative difference in uptake. This difference in uptake leads to an overall decrease in DIC in the surface layer in the Indian Ocean south of 45 S: There is also a negative difference in uptake in the North Atlantic (Fig. 9d) which leads to a large decrease seen in the surface DIC concentration. The uptake in the Pacific is almost the same in both runs. Overall, the atmospheric CO 2 concentration varies by less than 1 ppmv between these runs although there are significant regional differences in the air sea CO 2 flux as well as surface DIC concentrations. 5. Sensitivity to initial CO 2 concentration In this section we address the sensitivity to initial atmospheric CO 2 concentration. We carry out equilibrium runs where the atmospheric CO 2 concentration is held constant at 16, 2, 22, and 26 ppmv for 38 yr: Two sets of experiments are then carried out using these equilibrium runs: one set in which North Atlantic meltwater events are simulated (NA16, NA2, NA22 and NA26); a second set where meltwater events are introduced in the Southern Ocean (SO16, SO2, SO22 and SO26).