Reduced Atlantic storminess during Last Glacial Maximum:Evidence from a coupled climate model

Camille Li†∗

David S. Battisti

Department of Atmospheric Sciences, University of Washington, Seattle, Washington

†Current affiliation: Bjerknes Centre for Climate Research,Bergen, Norway∗Corresponding author address: Camille Li, Bjerknes Centre for Climate Research, Allegaten 55, 5007 Bergen, Norway.

E-mail: camille@uib.no

Abstract

The Last Glacial Maximum (LGM), 21 thousand years before present, was the time of max-

imum land ice extent during the last ice age. A recent simulation of LGM climate by a state-of-

the-art fully coupled global climate model is shown to exhibit strong, steady atmospheric jets and

weak transient eddy activity in the Atlantic sector compared to today’s climate. In contrast, pre-

vious work based on uncoupled atmospheric model simulations has shown that the LGM jets and

eddy activity in the Atlantic sector are similar to those observed today, with the main difference

being a northeastward extension of their maxima. The coupled model simulation is shown to agree

better with paleoclimate proxy records, and thus, is taken as the more reliable representation of

LGM climate. The existence of this altered atmospheric circulation state during LGM in the model

has implications for our understanding of the stability of glacial climates, for the possibility of

multiple atmospheric circulation regimes, and for the interpretation of paleoclimate proxy records.

1

1. Introduction

The Last Glacial Maximum was a cold period approximately 21 thousand years before present

(21 ka), when massive ice sheets covered much of the NorthernHemisphere continents. Paleocli-

mate proxy records provide valuable information on the climate forcings and climate state at the

Last Glacial Maximum (LGM), but they are limited in spatial and temporal resolution and can be

difficult to interpret, especially if one wishes to deduce circulation features such as flow fields or

energy fluxes. Climate models have proved to be increasinglyuseful as a tool for tackling some of

the questions that cannot be answered using proxy data alone. Of interest in this study is the large-

scale circulation of the atmosphere during glacial times, and how it compares to the circulation

observed in the present day climate. In particular, we direct our attention towards the North At-

lantic sector, a region of deep water formation that experienced large, abrupt climate events during

the last glacial period.

In the 1990s, the Paleoclimate Model Intercomparison Project was undertaken to evaluate past

climates using a collection of climate models. The first phase of the project (PMIP1) comprised,

for the most part, uncoupled atmospheric general circulation models forced with LGM boundary

conditions: a sea surface temperature and sea ice cover reconstruction from the Climate: Long

range Investigation, Mapping, and Prediction (CLIMAP) project (CLIMAP 1981) and the ICE-

4G land ice reconstruction (Peltier 1994). The resulting simluations offered a glimpse into how

atmospheric circulation may have been during LGM. Subsequent studies focused on describing

and understanding specific features such as atmospheric heat transport, transient activity and storm

tracks (Hall et al. 1996; Kageyama et al. 1999; Kageyama and Valdes 2000). Their findings have

been used to endorse the idea that the glacial world was a stormier world thanks to stronger equator-

to-pole temperature gradients and enhanced production of baroclinic eddies. Revisiting the original

studies, however, discloses some often-overlooked details.

Synthesizing results from all the European models in PMIP1,Kageyama et al. (1999) report

2

a northeastward extension of both Northern Hemisphere storm tracks across the board, and an

elongation of the Atlantic storm track in six of seven models. They note “no systematic increase

or decrease in the storminess from the present climate to thelast glacial maximum”, although the

figures in the paper do seem to indicate an increase in peak lowlevel transient eddy activity for

most of the models (see Kageyama et al. 1999, Figures 1–2, 6–7). In the specific case of the U.K.

Universities’ Global Atmospheric Modelling Project (UGAMP) prescribed SST simulation, the

increase in Atlantic eddy activity is limited to low levels of the atmosphere and is associated with

enhanced low level baroclinicity, thus suggesting the presence of stronger but shallower synoptic

waves (Hall et al. 1996). Furthermore, it was found that changes in normal modes largely account

for changes in the position and dominant wavenumber of the storms, but not the amplitude of the

actual storm tracks (Kageyama and Valdes 2000).

Included in the analysis of Kageyama et al. (1999) are simulations from select models that

were run with a slab ocean, a configuration that allowed sea surface temperatures (SSTs) and sea

ice to be computed based on their thermal response to the climate forcings. Results from these

simulations are somewhat equivocal as the oceanic heat flux that must be specified for a slab ocean

was set to present day values, thereby determining in large part the SST and sea ice distributions

as well as the location of the storm tracks. However, among slab ocean simulations, there is some

support for increased storminess at low to middle levels of the atmosphere during LGM (Dong and

Valdes 1998). Going one step further, a study involving an intermediate complexity model with a

simple dynamical ocean model reports similar findings (Justino et al. 2005).

Mounting paleoclimate evidence has since made it clear thatboth the CLIMAP SST recon-

struction and the ICE-4G land ice reconstruction used in PMIP1 are flawed. In view of such de-

velopments, the question of how jets and storminess changedduring LGM is by no means settled.

There are now simulations of the LGM by more complex, fully coupled climate models with up-

dated climate forcings based on the paleoclimate record, some of which fall within the framework

3

of the second phase of PMIP (PMIP2). The majority of the literature documenting these simu-

lations is concerned with issues such as the following: atmosphere-ocean processes in the tropics

and subtropics, where the CLIMAP reconstruction is known tohave problems (Bush and Philander

1998; Broccoli 2000; Kitoh and Murakami 2002; Timmermann etal. 2004); the general features

of the LGM climate (Kitoh and Murakami 2001; Kim et al. 2003; Shin et al. 2003; Otto-Bliesner

et al. 2006); and the role of ocean dynamics in the maintenance of this climate state (Dong and

Valdes 1998; Hewitt et al. 2003). Initial PMIP2 intercomparison studies have found an improve-

ment in many aspects of the new LGM simulations (Kageyama et al. 2006; Braconnot et al. 2007),

although considerable problems remain in terms of reproducing the peak glacial-interglacial tem-

perature change from the Greenland ice cores (Masson-Delmotte et al. 2006) and of simulating

changes in the ocean thermohaline circulation (Otto-Bliesner et al. 2007).

In this study, we examine the atmospheric circulation at Last Glacial Maximum in a PMIP2 sim-

ulation from a state-of-the-art, fully coupled climate model, the Community Climate System Model

(CCSM3) developed at the National Center for Atmospheric Research (NCAR). The simulation

was set up and performed by Otto-Bliesner et al. (2006) in order to contrast pre-industrial, mid-

Holocene and LGM climate. One major objective of the presentstudy is to characterize the mean

state and variability of atmospheric circulation in the Atlantic sector during the LGM by examin-

ing the jet and transient eddies, with an emphasis on identifying differences between the LGM and

present day climates.

Interestingly, when the CCSM3 is forced with PMIP2 boundaryconditions, the simulated LGM

climate shows a strong, steady Atlantic jet and enhanced lowlevel baroclinicity, but diminished

wintertime eddy activity at all levels of the atmosphere compared to the present day. These results

appear to be at odds with the atmosphere-only simulations from PMIP1, and furthermore suggest

the existence of an altered atmospheric circulation regimeduring LGM compared to PD. As an

initial illustration of the differences, Figure 1 shows contours of total horizontal wind speed at

4

200 mb for PD and LGM simulations from a selection of the higher resolution PMIP1 models

(details in Table 1) and from the CCSM3 coupled model. For each uncoupled model (c–h), the

LGM simulation (forced with PMIP1 boundary conditions) produces an Atlantic jet with strength

and orientation (SW-NE tilt) comparable to the present day jet (both simulated and observed),

but with a northeastward extension. This is in stark contrast to the CCSM3 LGM simulation (j),

which features a strong Atlantic jet with a more zonal orientation. Two of the PMIP1 slab ocean

simulations are also shown (a–b); while there are difficulties in interpreting these simulations, they

do serve to emphasize the qualitative difference between the CCSM3 LGM simulation and all the

LGM simulations that use PMIP1 boundary conditions.

The paper is organized as follows. A brief description of themodel and methods is contained

in section 2. Section 3 presents the model results indicating that, during Last Glacial Maximum,

the atmosphere exhibited a strong, steady mean circulationwith decreased eddy activity. Section 4

discusses some possible mechanisms for the suppression of eddy activity. Section 5 summarizes

evidence, based on paleoclimate observations, that the CCSM3 simulation produces a more realis-

tic representation of the LGM climate than the PMIP1 simulations; the focus is on the roles of land

ice forcing and the SST/sea ice distributions. Finally, themain results of this work are summarized

in section 6.

2. Model description and methods

We investigate changes between present day (PD) and Last Glacial Maximum (LGM) climates

as simulated by the Community Climate System Model 3 (CCSM3;Collins et al. 2006a), a global

coupled atmosphere-ocean-sea ice-land surface climate model developed at the National Center

for Atmospheric Research (NCAR). The setup of and results from these model simulations are

documented in detail in Collins et al. (2006a) and Otto-Bliesner et al. (2006). Briefly, the CCSM3

comprises the primitive equation Community Atmosphere Model version 3 (CAM3) at T42 hori-

5

zontal resolution with 26 hybrid coordinate vertical levels (Collins et al. 2006b); a land model with

land cover and plant functional types, prognostic soil and snow temperature and a river routing

scheme (Dickinson et al. 2006); the NCAR implementation of the Parallel Ocean Program (POP)

on a 320×384 dipole grid (nominal horizontal resolution of 1◦) with 40 vertical levels (Smith and

Gent 2002); and a dynamic-thermodynamic sea ice model on thesame grid as the ocean model

(Briegleb et al. 2004).

The forcings for the LGM simulation are in accordance with the protocols established by

PMIP2 (http://pmip2.lsce.ipsl.fr): 21 ka insolation; atmospheric greenhouse gas concentrations

based on ice core measurements (Fluckiger et al. 1999; Dallenbach et al. 2000; Monnin et al. 2001);

atmospheric aerosols at preindustrial values; land ice andcoastlines corresponding to 120 m sea

level depression from the ICE-5G reconstruction (Peltier 2004).

We use monthly mean output from 50 years of the simulations. Daily output for transient

eddy analyses are taken from 25 year branch runs. A bootstrapmethod was used to determine

that 25 year samples are adequate for stable eddy statistics. Using wintertime (DJFM) Northern

Hemisphere 850 mbv′T ′ calculated from high-pass filtered dailyv andT fields as our metric,

we estimated the statistical uncertainty associated with adata set of finite length as follows. We

generated 1000 realizations of the full data set (n0 = 40 years of daily data from the PD simulation),

each realization created by sampling the full data setn0 times with replacement. The result is

considered a distribution of “truth” based on the full data set. We then repeated this procedure for

sample sizesn < n0 to produce subsampled realizations ofv′T ′. Defining an acceptablen to be

one for which the distribution ofv′T ′ falls within the distribution of “truth” 95% of 1000 times, we

found a sample size ofn = 25 years to satisfy this criterion. As an additional check,we repeated

the eddy analysis on the full 40 year data set and found no change to the main results of the study.

Eddy fields were filtered with a sixth order high-pass Butterworth filter to emphasize variability at

periods less than 8 days. Such filters are often used in the study of storm tracks, with the high-pass

6

cutoff varying between 6–10 days (for example Nakamura 1992; Trenberth 1991; Yin 2002). The

exact choice of filter is relatively unimportant since baroclinic waves dominate the eddy statistics

at these synoptic time scales.

3. Last Glacial Maximum climate in the coupled model (CCSM3)

Of interest here is the large scale atmospheric circulationin the model’s simulations of the

Last Glacial Maximum (LGM) and present day (PD) climates. Wewill focus on the Northern

Hemisphere Atlantic sector, where differences in forcingsbetween the LGM and PD, and conse-

quently circulation features, are most dramatic. All differences between the LGM and PD climates

discussed here are significant at the 95% confidence level unless otherwise noted.

a. Circulation and heat transport

Upper level zonal wind and geopotential height provide a useful broad-brush picture of large

scale flow characteristics of the atmosphere. Figure 2 showswintertime maps of these two fields

for the LGM and PD simulations. Under the LGM forcings described in the previous section,

we observe an enhanced stationary wave associated with the Laurentide ice sheet covering most

of North America, and a stronger, more zonal jet in the Atlantic sector downstream of the ice

sheet. The stronger winds during LGM are consistent with thestronger equator-to-pole surface

temperature gradient seen in Figure 3. Changes in the Pacificare more subtle, showing a slight

equatorward shift of the jet (Figure 3) and the development of a split flow over Siberia evident

from the region of weak easterlies outlined by the thick zerocontour in Figure 2b.

The change in atmospheric circulation over the Atlantic sector during LGM is particularly

striking. There is an inverse relationship between maximumjet strength and jet width in January

for the PD simulation (red circles in Figure 4), where jet strength uATL is the maximum zonal wind

7

in the sector, and jet width is the latitude range over which uATL decreases to half its maximum

value. The relationship is not evident in the LGM climate (blue triangles), which furthermore

inhabits a sector of width-strength space separate from thePD climate. The LGM jet is 20%

stronger on average, with half the variability in maximum speeds of the PD jet; it is also 30%

narrower, with almost six times less variability in width. These results are robust to the choice of

month(s) used to define winter, and to the exact longitude range used to define the Atlantic jet.

Figure 5 shows implied annual meridional energy transportscalculated from the model output.

The total (atmosphere plus ocean) heat transport RT by the climate system is determined by in-

tegrating the top-of-atmosphere (TOA) radiation imbalance RTOA over all longitudesθ from the

North Pole to each latitudeφ:

RT(φ) = R 2E

∫

2π

0

∫ π/2

φRTOA(φ′, θ′) cos φ′dφ′dθ′ (1)

where RE is the radius of the Earth. Next, we sum the shortwave, longwave, latent heat and sensible

heat fluxes from the model to get the annual mean surface heat flux Rsfc. The implied atmospheric

heat transport AT is found by integrating the difference between the net TOA fluxRTOA and the

net surface heat fluxRsfc:

AT(φ) = R 2E

∫

2π

0

∫ π/2

φ(RTOA(φ′, θ′) − Rsfc(φ

′, θ′)) cos φ′dφ′dθ′ , (2)

Finally, the ocean heat transport OT is calculated as the residual:

OT = RT − AT . (3)

From Figure 5b, the discrepancy between the model’s PD totalheat transport (black line) and the

satellite-derived radiatively required total heat transport from Trenberth and Caron (2001) (filled

grey area) is less than 0.5 PW, or 10% of the maximum heat transport.

8

We find that the amount of heat transported towards the poles is remarkably similar in the PD

and LGM simulations. Between the two model simulations, theLGM does indeed have slightly

more transport, with the atmosphere helping to increase thepeak Northern Hemisphere (NH) value

to 6.33± 0.01 PW, about 0.3 PW greater than in PD. The robustness of thetotal heat transport curve

found across the CCSM3’s simulations of PD and LGM climates is corroborated by other coupled

climate models (Hewitt et al. 2003; Shin et al. 2003) and fromuncoupled general circulation mod-

els using a prescribed sea surface temperature forcing (Hall et al. 1996), with the magnitude of the

LGM increase varying from 0.2–1.5 PW for peak NH values.

All else being equal, a back-of-the-envelope calculation tells us that this relatively modest

0.3 PW boost in heat transport at 35◦N translates to a 3 W m−2 boost in heating rate north of this

latitude circle. Assuming a mid-range climate sensitivityof 0.5◦C per W m−2, this is equivalent to

a 1.5◦C warming of the mid-to-high latitude regions. Clearly, allelse is not equal, and the actual

surface temperature difference between the two climates poleward of 35◦N is closer to 10◦ C. Note

that technically, the 0.5◦C per W m−2 value is a global climate sensitivity, and should not be usedto

estimate the response of the polar cap to increased heat fluxfrom the lower latitudes. However, the

polar cap as defined in our calculation (35◦N to the North Pole) is a large region over which there is

a net TOA radiation loss to space. Furthermore, the use of this global climate sensitivity to estimate

the contribution of heat transport changes to the simulatedtemperature change is supported by an

experiment in Seager et al. (2002) in which ocean heat transport was turned off, leading to a

1.3 PW reduction in energy moved across 35◦N or a 13 W m−2 decrease in heating rate north of

this latitude circle, which caused a 6◦C cooling of the mid-to-high latitude regions.

Upon closer inspection, there are interesting differencesin the partitioning of these heat fluxes

in the PD and LGM simulations. Overall, the net (atmosphere plus ocean) energy transport in

the NH during LGM is very similar to that in the PD. The total atmospheric transport (AT) in the

LGM is also comparable to that in the PD, but there is a large difference in the balance of processes

9

responsible for this AT. The pronounced ridging forced by the Laurentide ice sheet enhances the dry

stationary wave heat transport contributionv∗ T ∗ to AT in the LGM compared to PD (Figure 5c).

By inference, the transient heat transport by eddies must bediminished in this region. We note that

there is a slight decrease in latent energy transport in the LGM simulation (not shown), an effect

which is to be expected in a drier climate (Held and Soden 2006), but that this change is small

compared to the dry stationary wave change. In a global view,these results are in fact in agreement

with the energy budget analysis of the UGAMP LGM simulation by Hall et al. (1996). Although

they observe an increase in transient heat transport at low levels, there is a compensating decrease

aloft such that the column-integrated transient eddy transport is smaller during LGM.

b. Transient eddy activity

We can diagnose the eddy activity associated with this reduction in energy transport by the tran-

sient eddies in the LGM by calculating high-pass filtered quantities such as low level temperature

flux (v′T ′), upper level momentum flux (u′v′) and upper level eddy kinetic energy ((u′2 + v′2)/2).

In addition to presenting maps of these eddy statistics, we use several metrics to quantify the steadi-

ness of the atmospheric circulation at low levels and aloft during the winter (DJFM) season. The

first metric is the maximum zonal wind at 200 mb in the Atlanticsector (15◦N–65◦N, 90◦W–0◦).

The second metric is the kinetic energy associated with the departure of the 200 mb flow from

climatological monthly means, averaged over the Atlantic sector,

KEIATL =1

2AATL

∫

AATL

(u200 − u200)2 + (v200 − v200)2 dA , (4)

whereu andv are monthly mean fields,A represents area, and overbars indicate climatological

means. KEIATL thus gives an indication of the interannual variability of the upper level flow. The

final metric is the northward eddy heat flux at 850 mb averaged over the Atlantic sector,

10

vTATL =1

AATL

∫

AATL

v′850

T ′850

dA , (5)

where primes indicate daily fields that have been high-pass filtered to retain variability at periods

less than 8 days.

Concentrating on boreal winter in the NH, the eddy fields reveal an LGM climate that is more

quiescent than the PD (Figure 6). In the Atlantic sector, thereduction in eddy activity from PD to

LGM is observed at both low levels (15% decrease in sector-averagedv′T ′) and upper levels (30%

decrease in sector-averaged(u′2 + v′2)/2). Compared to PD, the LGM jets are strong and narrow

(Figure 3), and the eddy fluxes occupy a narrower latitudinalband hugging the axis of the jet core

rather than a broad band perched on the poleward flank of the jet (for example, compare Figure 6b

and c). The differences between the two climates can also be seen in Figure 4, in which January

poleward heat fluxes in the LGM simulation (blue triangles) span a narrow range of smaller values

than in the PD simulation (red circles). Although we will notdiscuss the Pacific sector, it too

exhibits changes in jet structure and eddy fluxes.

The measures of atmospheric flow in Table 2 provide another way to compare the Atlantic

sector in today’s climate and in glacial climates. The weak eddy activity in the LGM simulation

coexists with a stronger, narrower Atlantic jet, as shown inFigure 4. From this plot, we inferred a

less variable upper level flow field during LGM compared to PD.The winter season flow metrics

provide additional and more direct ways to evaluate the steadiness of the LGM jet. From Table 2,

we see a 35% decrease in monthly departures of kinetic energyfrom its climatological mean state

(KEIATL in column three). Together, these results are consistent with the picture of global heat

transport in Figure 5, in which a slight overall increase in meridional energy flux during LGM is

achieved by a greatly enhanced stationary wave, with the implication that the contribution from

transient eddies must be weaker.

As a final remark on this topic, we note that the gross structure of the Atlantic jet and ed-

11

dies described here for the PD simulation is consistent withobservational data from the National

Centers for Environmental Prediction NCEP-NCAR reanalysis (Kalnay et al. 1996) and the Euro-

pean Medium-Range Weather Forecasts (ECMWF) Reanalysis ERA-40 (Uppala et al. 2005). Both

NCEP (not shown) and ERA-40 (Figure 6 a,d,g) show a broad wintertime jet with a SW-NE tilt,

and eddy activity peaking poleward of the jet, just as we haveseen for the PD simulation (Fig-

ure 6 b,e,h). One feature that does not compare well is the absolute jet strength, which the model

overestimates in each winter month (as seen for January in Figure 4) such that the DJFM winter

season average of uATL is 20% too strong (Table 2). However, the variability in the large-scale flow

field in the PD simulation is comparable to that in the two reanalyses prodcuts; furthermore, the

PD variability is substantially different from that in the LGM simulation. For example, uATL spans

a range of approximately 25 m s−1 in both the observations and in 50 years of the PD simulation

(Figure 4), while the range in jet strengths in the LGM simulation is approximately half this value.

Also, the variability as measured by the poleward heat flux vTATL and kinetic energy KEIATL in-

dices (Table 2) appear to show characteristic values for thePD climate (3.42–3.56 K m s−1, 31–34

m2 s−2) that are distinct from those of the LGM climate (2.93 K m s−1, 22 m2 s−2).

c. Baroclinicity of the atmosphere

The result of diminished meridional heat transport in a climate with sharper temperature gra-

dients and stronger jets (Figures 2–4) is somewhat surprising. To express this apparent paradox

in more quantitative terms, we use the Eady growth rate parameter (Eady 1949) formulated by

Lindzen and Farrell (1980) as a means of predicting transient behaviour from the time mean flow.

The parameter is the growth rate of the fastest growing Eady mode, and is given by

σ = 0.31f

N

∂uh

∂z, (6)

wheref is the Coriolis parameter,N is the Brunt-Vaisala buoyancy frequency and∂uh/∂z is the

12

vertical shear of the horizontal wind component. We calculate this growth rate for the 850–700 mb

layer (Figure 7).

For each climate, regions of high Eady growth rate correspond by and large to regions of

enhanced eddy activity; across different climate states, however, regions where Eady growth rates

increase are regions where eddy amplitudes decrease. For example, there is a general increase in

midlatitude growth rates during LGM relative to PD, especially over the Atlantic, which suggests

that the glacial climate should be stormier than today’s climate. That the eddy diagnostics indicate

decreased eddy activity during LGM despite more baroclinicconditions (largerσ) points to the

complicated nature of the relationship between transient activity and the mean flow.

4. Suppression of eddy activity

The LGM climate as simulated by the CCSM3 coupled model exhibits stronger jets but weaker

storms. This quiescent glacial climate runs counter to the intuitive idea that a more vigorously

circulating atmosphere should produce a more tempestuous world. It also runs counter to more

theoretically based indicators such as the stronger meridional temperature gradient and increased

low level baroclinicity. However, a range of factors not captured by the Eady growth rate parameter

also affect the lifecycle of eddies. These factors range from diabatic heating (Hoskins and Valdes

1990) and seeding effects (Zurita-Gotor and Chang 2005) to “governors” that limit the ability of

eddies to tap into the available baroclinicity (James 1987;Lee and Kim 2003; Harnik and Chang

2004). In this section, we discuss whether such factors may play a role in reducing the eddy activity

in the coupled model LGM simulation. A pre-requisite to thisdiscussion is an assessment of the

structure of the eddies.

13

a. Eddy structure

To study the structure of eddies, we use one-point regression analysis (Lim and Wallace 1991;

Chang 1993) in which a regression coefficientb(i) is calculated for a specific pressure, latitude,

longitude locationi by linearly regressing the input variable of interestyt(i) against a reference

time seriesxt:

b(i) =

N∑

t=1

y′t(i)x′t

( N∑

t=1

x′t2)1/2

. (7)

Here,N is the number of time samples and primes denote daily fields which have been high-pass

filtered to retain variability at periods less than 8 days.

We examine regression structures at 200 mb and 850 mb in the CCSM3 simulations of PD and

LGM climates (Figure 8). The reference time series against which variables are linearly regressed

is the 200 mb meridional wind fieldv′200 at 43◦N, 28◦W for the PD climate, and at 38◦N, 28◦W for

the LGM climate. These points were selected based on the location of maximum variance in the

DJFMv′200 field. We normalize theb(i) coefficients such that the regression maps show amplitude,

expressed in the units of the variable of interest, per standard deviation of the reference time series.

Structures calculated using reference time series at different locations, reference time series based

on different fields, and alternate definitions of the winter season are similar to these, and are not

shown here.

The eddy structure in both the PD and LGM climates resemble localized wave trains. The LGM

eddy amplitudes peak just south of 40◦N, the latitude marked by the solid grey line in Figure 8,

while the PD eddy amplitudes peak just north of this latitude. Although meridional wind variations

v′ are much stronger aloft (Figure 8 a,d) than at low levels (Figure 8 b,e) in both simulations (note

the different contour intervals), there is a greater discrepancy in the LGM. Thus, eddies in glacial

times appear to be trapped closer to the tropopause, with weaker low-level disturbance amplitudes.

14

Furthermore, though the LGM has slightly weaker eddy amplitudes, the eddy structures are cor-

related over greater distances and are more persistent, as seen in lag regression analyses in which

the time seriesyt(i) is shifted by a certain number of days relative toxt (not shown). The larger

spatial correlations are in fact consistent with the eastward extension of storm tracks seen in the

PMIP1 simulations Kageyama et al. (1999), although the location and strength of the storm tracks

themselves are not. These results suggest that differencesin eddy structure may be an important

part of why the LGM simulation exhibits such weak eddy activity compared to the PD simulation.

There is a large body of theoretical work on why eddy activitymight be suppressed in conditions

that appear conducive to eddy growth. Next, we examine the relevance of these hypotheses for the

reduction of eddy activity in the LGM.

b. Discussion

One established idea in the literature is the barotropic governor mechanism, in which horizontal

shear is thought to limit baroclinic instability, and hence, storm growth (James 1987). The sharp-

ening of the Atlantic jet (Figure 3) makes this an attractivecandidate, but some telltale signs of the

barotropic governor are absent. If the barotropic governorwere in operation, we would expect the

eddies to bend and elongate into a boomerang or banana shape as they are deformed by the strongly

sheared mean flow; this characteristic shape is not evident in the LGM simulation (Figure 8).

Another explanation for the suppression of eddy activity isinspired by the existence of a mod-

ern analogue to the strong jet/weak storms dichotomy, namely the Pacific midwinter suppression

of storms (Nakamura 1992) which occurs every January when the jet is at its strongest (Figure 9).

According to Yin (2002), this phenomenon has to do with the southward migration of tropical con-

vection in the western Pacific, and the related strengthening of the subtropical jet, during boreal

winter. Yin (2002) diagnosed that these changes increase upper level static stability and lower the

tropopause height in a manner that discourages the growth ofstorms. In the modern climate, the

15

climatological latitude of the Atlantic jet is 2–3◦ north of the climatological latitude of the Pacific

jet in winter, perhaps positioning it too far poleward to be affected by changes in tropical heating.

In the LGM simulation, however, the more zonal orientation of the Atlantic jet pulls it towards the

equator, in closer proximity to the tropical convection sites. Preliminary inspection reveals little

difference in tropical heating or precipitation between the LGM and PD simulations in the Atlantic

sector. Nonetheless, subtle changes in the tropics can havea large impact, so a more thorough

investigation should be carried out before this mechanism for jet stability can be rejected.

In a related vein, Lee and Kim (2003) used idealized numerical experiments to demonstrate

that flow regimes with strong (weak) subtropical jets exhibit weaker (stronger) eddy activity lo-

cated just (far) poleward of the jet core. In a series of experiments, they varied the strength of the

tropical convective heating and of the high latitude cooling in order to strengthen or weaken the

subtropical jet (Lee and Kim 2003; Son and Lee 2005). They showed that the strong subtropical

jet regime supported more isotropic eddies and weaker barotropic conversion than the weak sub-

tropical jet regime. The LGM simulation exhibits several features in common with their strong

subtropical jet case, including relatively weak eddy fluxeslocated at roughly the same latitude as

the tropospheric jet maximum (Figure 6) and a slightly more isotropic eddy structure (Figure 8).

Rather than requiring changes in tropical heating, this idea allows for the possibility of a strong jet

regime being set up some other way, for example, by topographic effects from the Laurentide or

additional high latitude cooling due to extensive sea ice cover in the North Atlantic.

The Lorenz energy cycle (Lorenz 1955) provides further insight into why the eddy activity in

the LGM simulation is suppressed compared to the PD simulation. The terms in the energy budget

equation describe how energy is exchanged between the mean flow and the transient eddies as

the eddy disturbances grow and decay. Different suppression mechanisms affect different parts of

the eddy life cycle, and thus should have different signatures on these budget terms. Preliminary

results (A. Donohoe, personal communication, 2007) indicate that the LGM Atlantic mean state

16

does indeed support faster eddy growth due to its sharper temperature gradients, despite competing

influences from the increased static stability and the strong barotropic shear of the jet. However,

the baroclinic conversion of potential energy into eddy kinetic energy is greatly reduced in the

LGM compared to PD. The explanation for the reduced eddy activity appears to be related to the

fact that there is substantially less seeding of the Atlantic baroclinic zone by upstream disturbances

in the LGM, and that the disturbances that do enter the baroclinic zone are weak. The cause of the

reduced seeding is under continuing investigation.

5. Evaluation of the CCSM3 simulation

The PMIP1 simulations of LGM climate share certain atmospheric flow characteristics that

are very similar to those in the present day climate, while the coupled model simulation of LGM

exhibits fundamental differences when compared to the present day climate. Specifically, in the

PMIP1 models, the amount of storminess in the Atlantic sector during LGM is comparable to that

in the PD climate, with slight increases in peak low level transient eddy activity during the LGM

(Kageyama et al. 1999). Furthermore, in the PMIP1 LGM simulations, the Atlantic jet shows the

same northeastward extension as the Atlantic storm track, with little change in strength (Figure 1).

Conversely, the CCSM3 LGM simulation exhibits a strong, steady jet with zonal orientation, and

weaker transient eddies at all levels of the atmosphere (Figure 6).

We have reproduced the PMIP1 results with CAM3, the atmospheric component of the CCSM3,

to illustrate more clearly the differences between the circulation simulated by the coupled model

and the generic result obtained by the suite of uncoupled PMIP1 models (Figure 10). The at-

mosphere model has now been forced with sea surface temperature (SST), sea ice and land ice

prescribed exactly as specified in the PMIP1 experimental setup. The resulting simulation is the

same that appears in Figure 1 h and will henceforth be referred to as PMIPa, where the “a” stands

for “atmosphere-only”. Compared to the coupled model LGM simulation (Figure 10 c,d), PMIPa

17

produces a stormier Atlantic sector and a weaker jet that exhibits a marked SW-to-NE tilt (Fig-

ure 10 e,f). Indeed, the eddy heat transport and jet orientation in PMIPa bear a closer resemblance

to the coupled PD simulation (Figure 10 a,b) than to the coupled LGM simulation.

The source of the differences between the LGM climate simulated by the uncoupled models

and the LGM climate simulated by the coupled model lies at theEarth’s surface. Recall that

the uncoupled PMIP1 models experience or “see” one set of prescribed land ice and SST/sea

ice boundary conditions. The coupled model experiences a somewhat different set of land ice

boundary conditions, those of the improved ICE-5G (Peltier2004) reconstruction. In addition, the

coupled model has interactive ocean and sea ice components that calculate their own SST/sea ice

fields. The fact that CCSM3’s atmosphere model achieves a PMIP1-like LGM climate when run

in the PMIP1 configuration is further support that it is indeed these land and sea surface forcings,

rather than a disparity in models or atmosphere model physics, that are key for creating the different

LGM climates (Figure 1h).

The question becomes one of evaluating which simulated LGM climate, the uncoupled PMIP1-

like LGM or the coupled CCSM3-like LGM, is a more realistic representation of the actual climate

of the Last Glacial Maximum. Detailed comparisons have beenperformed between PMIP1 model

simulations and a selection of the fully coupled model simulations, including CCSM3, that partic-

ipated in PMIP2. These studies conclude that the fully coupled PMIP2 simulations using ICE-5G

land ice are in better agreement with the paleoclimate observations currently available, particularly

in terms of surface air temperature, SST and tropical climate (Braconnot et al. 2007), as well as

regional temperatures over northern Eurasia and the North Atlantic (Kageyama et al. 2006). These

intercomparisons include a number of partially coupled simulations from PMIP1 in which SSTs

and sea ice were calculated by a slab ocean model. These will not be included in our discussion

as their use of present day oceanic heat transport in the slabocean heavily influences the resulting

SST patterns.

18

The following sections will point out certain aspects of theboundary conditions and model re-

sults where the CCSM3 simulation shows marked improvements, based on the paleoclimate record,

over the PMIP1 simulations. (For a comprehensive treatmentof this topic, the reader is referred

to the intercomparison studies mentioned above and throughout the rest of the section.) Also dis-

cussed are the potential consequences of these features foratmospheric dynamics in general, and

for the Atlantic jet and storm track in particular.

a. Land ice

The ICE-4G (Peltier 1994) and ICE-5G (Peltier 2004) reconstructions of deglaciation history

are based in part on the theory of glacial isostatic adjustment, a process by which the external

surface load of the continental ice sheets is compensated bychanges in the surface of the Earth.

ICE-5G is widely considered to be an improvement over ICE-4G: its land ice reconstruction ac-

counts for new refinements in the reconstruction of global sea level, and it also corrects many of

the regional shortcomings of ICE-4G (see Peltier 2004, and references therein). As a result, at the

time of the LGM, ICE-5G has more land-based ice than did ICE-4G (approximately 12 m sea level

equivalent). In ICE-5G, there is much less land ice on Greenland and the Eurasian continent, while

the Laurentide ice sheet complex over North America is significantly larger in volume compared

to ICE-4G (by 15 m sea level equivalent, or 25% of the Laurentide ice volume), with the bulk

of the extra ice forming the Keewatin Dome west of Hudson Bay (Figure 11). The addition of

this Dome reconciles the original ICE-4G results with more recent observations, namely the large

crustal uplift rates near Yellowknife, Canada (Argus et al.1999) and gravity measurements south

and west of Hudson Bay (Lambert et al. 2001).

A higher Laurentide ice sheet forces a stronger stationary wave pattern in the Northern Hemi-

sphere (Otto-Bliesner et al. 2006), and along with this we expect changes in surface temperatures

and in the interannual variability and seasonal cycle of surface temperature. In general, these fea-

19

tures are better simulated in the PMIP2 runs using ICE-5G than in the PMIP1 runs using ICE-4G,

with the exception of western European winter temperatures, which are still underestimated com-

pared to pollen data (Kageyama et al. 2006). The altered stationary wave and the physical barrier

of this larger ice sheet may also contribute to reduced seeding of the Atlantic jet (section 4b), and

hence, explain both the strengthening and stabilization observed in the LGM simulation.

b. SST and sea ice

The CCSM3 coupled model provides a self-consistent simulation of Last Glacial Maximum

climate. It produces a climatological SST distribution andsea ice coverage that differ from those

determined in CLIMAP and used as boundary conditions in PMIP1 (CLIMAP 1981). At first

glance, one might expect the observation-based CLIMAP dataset to provide a more trustworthy

picture of the glacial ocean than a model. CLIMAP produced the first SST maps of the glacial

ocean by using transfer functions to translate population distributions of fossil plankton species

found in ice age marine sediments (CLIMAP 1981) into sea surface temperatures. The trans-

fer functions (Imbrie and Kipp 1971) were derived from knowledge of how these same plankton

species are distributed in today’s ocean. In the intervening decades, however, the proliferation of

sediment core data and the advent of new statistical and geochemical methodologies for recon-

structing past SSTs have challenged some of the assumptionsand results of the CLIMAP method.

Moreover, modelling studies have shown that the consequences of such SST errors on the atmo-

spheric circulation are potentially dramatic, whether theerrors themselves are in the tropics or high

latitudes (Yin and Battisti 2001; Toracinta et al. 2003; Hostetler et al. 2006).

Though the data still fall short of providing definitive SST and sea ice distributions in the

glacial ocean, in regions where there is growing consensus,the CCSM3 simulation shows a clear

improvement over the CLIMAP reconstruction. Among the moreproblematic areas of CLIMAP

are the lack of cooling in tropical and subtropical regions,and the extensive sea ice in the North

20

Atlantic (Figure 12a). These features are very different inthe CCSM3 simulation (Figure 12b),

and in fact in the coupled PMIP2 simulations in general (Kageyama et al. 2006; Braconnot et al.

2007, and references therein).

An intercomparison of eight coupled model simulations of the LGM shows a 1–2◦ C annual

mean cooling over the global tropics, in agreement with proxy data (Braconnot et al. 2007, and

references therein). But more relevant than changes in the absolute temperature field in the tropics

and subtropics are changes in the distribution of SSTs and SST gradients in the tropics. Eval-

uating whether CLIMAP or the CCSM3 simulation does a better job in this respect is difficult

given the uncertainties in the SST reconstructions and the differences between them. However,

several interesting features are robust enough in the simulation and the records to merit mention.

At Last Glacial Maximum, the Pacific warm pool is contracted meridionally in the CCSM3 simu-

lation (Figure 12), a feature which is consistent with the foraminifera census data (Trend-Staid and

Prell 2002; Kucera et al. 2005); in contrast, CLIMAP has a warm pool that is expanded meridion-

ally compared to the present climate. Several studies pointto stronger zonal gradients across the

tropical Pacific (Lea et al. 2000; Trend-Staid and Prell 2002; Kucera et al. 2005), but the recon-

structed SST patterns themselves are quite dissimilar. Nevertheless, the CCSM3 simulation does

exhibit stronger tropical SST gradients – both zonal and meridional – compared to CLIMAP. In

the northern subtropics, the SST simulated by the CCSM3 is more similar to the reconstruction of

Trend-Staid and Prell (2002) than is the CLIMAP product, including more zonal isotherms in the

regions of the Kuroshio and Gulf Stream currents.

Another region where there are significant differences between the CLIMAP reconstruction

and the CCSM3-simulated SSTs is the North Atlantic (Figure 13). Compared to CLIMAP, the

simulated winter SSTs are 3–5◦C warmer along the west coast of Norway and 1–3◦C warmer

in a broad swath of the Atlantic Ocean just south of the Greenland-Scotland ridge. These areas

remain ice free in winter in CCSM3 and in other PMIP2 simulations (Kageyama et al. 2006),

21

while in CLIMAP, perennial sea ice cover extends as far southas 50◦N. The presence of open

ocean and relatively high temperatures in portions of the eastern North Atlantic and, at least sea-

sonally, in the Nordic Seas is supported by a large body of proxy data, much of which has been

synthesized by the MARGO (Multiproxy Approach for the Reconstruction of the Glacial Ocean

surface, http://margo.pangaea.de/) project. These data include estimates of SST and sea ice ex-

tent based on coccoliths (Hebbeln et al. 1994), biomarker pigments (Rosell-Mele and Koc 1997),

alkenones (Rosell-Mele and Comes 1999), dinoflagellates (de Vernal and Hillaire-Marcel 2000)

and foraminifera (Pflaumann et al. 2003; Sarnthein et al. 2003; Meland et al. 2005). The simulated

North Atlantic conditions in CCSM3 are further corroborated by other coupled atmosphere-ocean

models (Hewitt et al. 2003; Shin et al. 2003), and by inferredNorth Atlantic circulation patterns

from foraminiferal assemblages (Lassen et al. 1999).

CLIMAP has very extensive sea ice in the Atlantic, with the winter ice edge reaching at least

45◦N over the entire basin (Figure 12). Consequently, as the storms track northeastwards into the

Nordic Seas in the PMIP1 simulations, they have access to thestrong temperature gradients at

the sea ice edge. In the CCSM3 coupled simulation of the LGM, sea ice in the western Atlantic

spreads to similarly southern latitudes in winter, but in the eastern Atlantic, the ice edge never

ventures past 60◦N (Figure 12). With the sea ice so far north, the strongest midlatitude temperature

gradient in the eastern Atlantic is now due to the band of rapidly changing SSTs around 35–40◦N

(see Figure 12 or Kageyama et al. 2006). The location and orientation of the jet and storm track in

the coupled simulation are consistent with this SST gradient, which clearly points to its importance

in determining the state of the Atlantic atmospheric circulation. However, the SSTs are most likely

not the whole story. The size of the Laurentide ice sheet upstream seems to exert a large influence

on the location of the North Atlantic storm track and jet, andto the SST distribution in this region

(Li 2007).

22

6. Concluding remarks

We have presented evidence for an altered atmospheric circulation regime in the Atlantic sector

during Last Glacial Maximum from a global coupled climate model. This LGM Atlantic circula-

tion is characterized by a stronger, more zonally oriented Atlantic jet, and yet reduced storminess.

In other words, the coupled model produces a glacial climatethat is quiescent compared to today’s

climate. Analysis of the eddy structures suggests that the barotropic governor is unlikely to play a

role in suppressing the storms. Reduced seeding of the Atlantic jet is currently under investigation

as the most likely cause of the decrease in storminess duringLGM.

The LGM climate simulated by the coupled model is remarkablydifferent from the LGM

climate simulated by atmosphere models forced with PMIP1 boundary conditions. Whereas the

coupled simulation exhibits a strong, zonal jet with reduced eddies, the uncoupled simulations

feature a North Atlantic circulation regime that is similarto the one in today’s climate. We have

found that this discrepancy can be traced to differences in the ice sheet topography and sea surface

conditions, and not to differences in the atmosphere modelsand model physics. The LGM land

ice was updated from the ICE-4G reconstruction in PMIP1 to the improved ICE-5G reconstruction

in the coupled simulation. In addition, the SST and sea ice distributions seen by the atmosphere

were prescribed to CLIMAP in PMIP1 but calculated by sophisticated ocean and sea ice models

in CCSM3. The SST and sea ice distributions simulated by the coupled model are more consistent

with the proxy data than are the prescribed SST and sea ice boundary conditions used in PMIP1.

Thus, to the extent that changes in atmospheric circulationare linked to changes in orographic

forcing and sea surface conditions, the CCSM3 coupled modelsimulation can be regarded as a

better estimate of the LGM climate.

The existence of a different Atlantic atmospheric circulation regime during LGM has impli-

cations for our understanding of global heat transport and the stability of glacial climates. In

particular, it can offer new outlooks on modes of climate variability that appear to be unique to

23

glacial times, such as the abrupt Dansgaard-Oeschger warming events recorded in Greenland ice

cores.

Acknowledgments.

Data access:

http://pmip.lsce.ipsl.fr/

http://data.ecmwf.int/data/

http://www.cdc.noaa.gov/cdc/data.ncep.reanalysis.html

We wish to thank Joe Barsugli, Kerim Nisancioglu, Richard Seager, Mike Wallace, Justin

Wettstein and Jeff Yin for stimulating and constructive discussions, and Cecilia Bitz and Marc

Michelson for assistance with the model simulations. We arealso grateful to Michel Crucifix

and two anonymous reviewers for their thoughtful comments.This research was supported by the

Comer Fellowship Program.

24

References

Argus, D., W. Peltier, and M. Watkins, 1999: Glacial isostatic adjustment observed using very

long baseline interferometry and satellite laser ranging geodesy.J. Geophys. Res., 104 (B12),

29 077–29 094, doi:10.1029/1999JB900237.

Braconnot, P., B. Otto-Bliesner, S. Harrison, S. Joussaume, J.-Y. Peterchmitt, A. Abe-Ouchi,

M. Crucifix, E. Driesschaert, T. Fichefet, C. D. Hewitt, M. Kageyama, A. Kitoh, A. Lan, M.-

F. Loutre, O. Marti, U. Merkel, G. Ramstein, P. Valdes, S. L. Weber, Y. Yu, and Y. Zhao,

2007: Results of PMIP2 coupled simulations of the mid-Holocene and Last Glacial Maxi-

mum – part 1: experiments and large-scale features.Climate of the Past, 3 (2), 261–277, URL

http://www.clim-past.net/3/261/2007/.

Briegleb, B., C. Bitz, E. Hunke, W. Lipscomb, M. Holland, J. Schramm, and R. Moritz, 2004: Sci-

entific description of the sea ice component in the CommunityClimate System Model, version

3. Tech. Note NCAR/TN-463+STR, NCAR, 70 pp.

Broccoli, A. J., 2000: Tropical cooling at the Last Glacial Maximum: An atmosphere-mixed layer

model simulation.Nature, 13, 951–976.

Bush, A. B. G. and S. G. H. Philander, 1998: The role of ocean-atmosphere interactions in tropical

cooling during the Last Glacial Maximum.Science, 279, 1341–1344.

Chang, E. K. M., 1993: Downstream development of baroclinicwaves as inferred from regression

analysis.J. Atmos. Sci., 50, 2038–2053.

CLIMAP, 1981: Seasonal reconstructions of the Earth’s surface at the last glacial maximum. Tech.

Rep. MC-36, Geological Society of America, 18 pp.

25

Collins, W. D., C. M. Bitz, M. L. Blackmon, G. B. Bonan, C. S. Bretherton, J. A. Carton, P. Chang,

S. C. Doney, J. J. Hack, T. B.Henderson, J. T. Kiehl, W. G. Large, D. S. McKenna, B. D. Santer,

and R. D. Smith, 2006a: The Community Climate System Model version 3 (CCSM3).J. Clim.,

19, 2122–2143.

Collins, W. D., P. J. Rasch, B. A. Boville, J. J. Hack, J. R. McCaa, D. L. Williamson, B. P. Briegleb,

C. M. Bitz, S.-J. Lin, and M. Zhang, 2006b: The formulation and atmospheric simulation of the

Community Atmosphere Model version 3 (CAM3).J. Clim., 19, 2144–2161.

Dallenbach, A., T. Blunier, J. Fluckiger, B. Stauffer, J.Chappellaz, and D. Raynaud, 2000:

Changes in the atmospheric CH4 gradient between Greenland and Antarctica during the Last

Glacial Maximum and the transition to the Holocene.Geophys. Res. Lett., 27, 1005–1008.

de Vernal, A. and C. Hillaire-Marcel, 2000: Sea-ice cover, sea-surface salinity and

halo/thermocline structure of the northwest North Atlantic: Modern versus full glacial condi-

tions.Quat. Sci. Rev., 19, 65–85.

Dickinson, R. E., K. W. Oleson, G. B. Bonan, F. Hoffman, P. Thornton, M. Vertenstein, Z.-L. Yang,

and X. Zeng, 2006: The Community Land Model and its climate statistics as a component of

the Community Climate System Model.J. Clim., 19, 2302–2324.

Dong, B. and P. Valdes, 1998: Simulations of the Last GlacialMaximum climates using a general

circulation model: prescribed versus computed sea surfacetemperatures.Clim. Dyn., 14, 571–

591.

Eady, E., 1949: Long waves and cyclone waves.Tellus, 1, 33–52.

Fluckiger, J., A. Dallenbach, B. Stauffer, T. Stocker, D.Raynaud, and J.-M. Barnola, 1999: Varia-

tions of the atmospheric N2O concentration during abrupt climatic changes.Science, 285, 227–

230.

26

Hall, N. M. J., B. Dong, and P. J. Valdes, 1996: Atmospheric equilibrium, instability and energy

transport at the last glacial maximum.Clim. Dyn., 12, 497–511.

Harnik, N. and E. K. Chang, 2004: The effects of variations injet width on the growth of baroclinic

waves: Implications for midwinter Pacific storm track variability. J. Atmos. Sci., 61, 23–40.

Hebbeln, D., T. Dokken, E. Andersen, M. Held, and A. Elverhoi, 1994: Moisture supply for

northern ice-sheet growth during the Last Glacial Maximum.Nature, 370, 357–360.

Held, I. M. and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming.

J. Clim., 19, 5686–5699.

Hewitt, C., R. Stouffer, A. Broccoli, J. Mitchell, and P. J. Valdes, 2003: The effect of ocean

dynamics in a coupled GCM simulation of the Last Glacial Maximum. Clim. Dyn., 20, 203–

218, doi:10.1007/s00382-002-0272-6.

Hoskins, B. J. and P. J. Valdes, 1990: On the existence of storm tracks.J. Atmos. Sci., 47, 1854–

1864.

Hostetler, S., N. Pisias, and A. Mix, 2006: Sensitivity of Last Glacial Maximum climate to uncer-

tainties in tropical and subtropical ocean temperatures.Quat. Sci. Rev., 25, 1168–1185.

Imbrie, J. and N. Kipp, 1971: A new micropaleontological method for quantitative paleoclimatol-

ogy: application to late Pleistocene Carribean core.Late Cenozoic glacial ages, Turekian, K.,

Ed., Yale University Press, New Haven, 71–181.

James, I., 1987: Suppression of baroclinic instability in horizontally sheared flows.J. Atmos. Sci.,

44 (24), 3710–3720.

Justino, F., A. Timmermann, U. Merkel, and E. P. Souza, 2005:Synoptic reorganization of atmo-

spheric flow during the Last Glacial Maximum.J. Clim., 18, 2826–2846.

27

Kageyama, M., A. Lan, A. Abe-Ouchi, P. Braconnot, E.Cortijo, M. Crucifix, A. de Vernal, J. Guiot,

C. D. Hewitt, A. Kitoh, R. O. M. Kucera, O. Marti, B. Otto-Bliesner, W. R. Peltier, A. Rosell-

Mel, G. Vettoretti, S. L. Weber, and Y. Yu, 2006: Last GlacialMaximum temperatures over

the North Atlantic, Europe and western Siberia: a comparison between PMIP models, MARGO

sea-surface temperatures and pollen-based reconstructions. Quaternary Science Reviews, 25,

2082–2102.

Kageyama, M. and P. Valdes, 2000: Synoptic-scale perturbations in AGCM simulations of the

present and Last Glacial Maximum climates.Clim. Dyn., 16, 517–533.

Kageyama, M., P. Valdes, G. R. C. Hewitt, and U. Wyputta, 1999: Northern Hemisphere storm

tracks in present day and Last Glacial Maximum climate simulations: A comparison of the

European PMIP models.J. Clim., 12, 742–760.

Kalnay, E., M. Kanamitsu, R. Kistler, W. Collins, D. Deaven,L. Gandin, M. Iredell, S. Saha,

G. White, J. Woollen, Y. Zhu, A. Leetmaa, B. Reynolds, M. Chelliah, W. Ebisuzaki, W. Higgins,

J. Janowiak, K. Mo, C. Ropelewski, J. Wang, R. Jenne, and D. Joseph, 1996: The NCEP-

NCAR 40-year reanalysis project.Bull. Am. Meteor. Soc., 77 (3), 437–471, doi:10.1175/1520-

0477(1996)077<0437:TNYRP>2.0.CO;2.

Kim, J.-H., R. R. Schneider, S. Mulitza, and P. J. Muller, 2003: Reconstruction of SE trade-wind

intensity based on sea-surface temperature gradients in the southeast Atlantic over the last 25

kyr. Geophys. Res. Lett., 30 (22), 2144, doi:10.1029/2003GL017557.

Kitoh, A. and S. Murakami, 2001: A simulation of the Last Glacial Maximum with a coupled

atmopshere-ocean GCM.Geophys. Res. Lett., 28 (11), 2221–2224.

———, 2002: Tropical Pacific climate at the mid-Holocene and the Last Glacial Maximum sim-

ulated by a coupled ocean-atmosphere general circulation model.Paleoceanogr., 17 (3), 1037,

doi:10.1029/2001PA000724.

28

Kucera, M., M. Weinelt, T. Kiefer, U. Pflaumann, A. Hayes, M. Weinelt, M.-T. Chen, A. C. Mix,

T. T. Barrows, E. Cortijo, J. Duprat, S. Juggins, and C. Waelbroeck, 2005: Reconstruction of sea-

surface temperatures from assemblages of planktonic foraminifera: multi-technique approach

based on geographically constrained calibration sets and its application to glacial Atlantic and

Pacific Oceans.Quat. Sci. Rev., 24, 951–998, doi:10.1016/j.quascirev.2004.07.014.

Lambert, A., N. Courtier, G. Sasegawa, F. Klopping, D. Winester, T. James, and J. Liard, 2001:

New constraints on Laurentide postglacial rebound from absolute gravity measurements.Geo-

phys. Res. Lett., 28 (10), 2109–2112, doi:10.1029/2000GL012611.

Lassen, S., E. Jansen, K. Knudsen, A. K. M. Kristensen, and K.Christensen, 1999: Northeast

Atlantic sea surface circulation during thepast 30–1014c kyr B.P.Paleoceanogr., 14 (5), 616,

doi:0.1029/1999PA900029.

Lea, D. W., D. K. Pak, and H. Spero, 2000: Climate impact of late Quaternary equatorial Pacific sea

surface temperature variations.Science, 289, 1719–1724, doi:10.1126/science.289.5485.1719.

Lee, S. and H.-K. Kim, 2003: The dynamical relationship between subtropical and eddy-driven

jets.J. Atmos. Sci., 60, 1490–1503.

Li, C., 2007: A general circulation modelling perspective on abrupt climate change during glacial

times. Ph.D. thesis, University of Washington.

Lim, G. H. and J. M. Wallace, 1991: Structure and evolution ofbaroclinic waves as inferred from

regression analysis.J. Atmos. Sci., 48, 1718–1732, doi:10.1029/2005BL023492.

Lindzen, R. and B. Farrell, 1980: A simple approximate result for the maximum growth rate of

baroclinic instabilities.J. Atmos. Sci., 37, 1648–1654.

Lorenz, E., 1955: Available potential energy and the maintenance of the general circulation.Tellus,

7, 157–167.

29

Masson-Delmotte, V., M. Kageyama, P. Braconnot, S. Charbit, G. Krinner, C. Ritz, E. Guilyardi,

J. Jouzel, A. Abe-Ouchi, M. Crucifix, R. M. Gladstone, C. D. Hewitt, A. Kitoh, A. N. LeGrande,

O. Marti, U. Merkel, T. Motoi, R. Ohgaito, B. Otto-Bliesner,W. R. Peltier, I. Ross, P. J. Valdes,

G. Vettoretti, S. L. Weber, F. Wolk, and Y. Yu, 2006: Past and future polar amplification of

climate change: climate model intercomparisons and ice-core constraints.Climate Dynamics,

27 (4), 437–440.

Meland, M. Y., E. Jansen, and H. Elderfield, 2005: Constraints on SST estimates for

the northern North Atlantic/Nordic Seas during LGM.Quat. Sci. Rev., 24, 835–852, doi:

10.1016/j.quascirev.2004.05.011.

Monnin, E., A. Indermuhle, A. Dallenbach, J.Fluckiger,B. Stauffer, T. Stocker, D. Raynaud, and

J.-M. Barnola, 2001: Atmospheric CO2 concentrations over the last termination.Science, 291,

112–113.

Nakamura, H., 1992: Midwinter suppression of baroclinic wave activity in the Pacific.J. Atmos.

Sci., 49 (17), 1629–1642.

Otto-Bliesner, B., E. Brady, G. Clauzet, R. Tomas, S. Levis,and Z. Kothavala, 2006: Last Glacial

Maximum and Holocene climate in CCSM3.J. Clim., 19, 2526–2544.

Otto-Bliesner, B. L., C. D. Hewitt, T. M. Marchitto, E. Brady, A. Abe-Ouchi, M. Crucifix, S. Mu-

rakami, and S. L. Weber, 2007: Last Glacial Maximum ocean thermohaline circulation: PMIP2

model intercomparisons and data constraints.Geophys. Res. Lett., 34 (12).

Peltier, W. R., 1994: Ice age paleotopography.Science, 265, 195–201.

———, 2004: Global glacial isostasy and the surface of the ice-age Earth.Annu. Rev. Earth Planet.

Sci., 32, 111–149.

Pflaumann, U., M. Sarnthein, M. Chapman, J. D. H. Schulz, S. van Kreveld, E. Vogelsang, and

30

M. Weinelt, 2003: Glacial North Atlantic: Sea-surface conditions reconstructed by GLAMAP

2000.Paleoceanogr., 18 (3), 1065, doi:10.1029/2002PA000774.

Raynor, N., D. Parker, E. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. Kent, and A. Ka-

plan, 2003: Global analyses of sea surface temperature, seaice, and night marine air temperature

since the late nineteenth century.J. Geophys. Res., 108, doi:10.1029/2002JD002670.

Rosell-Mele, A. and P. Comes, 1999: Evidence for a warm LastGlacial Maximum in the Nordic

seas or an example of shortcomings in U-37(K)’ and U-37(k) toestimate low sea surface tem-

perature?Paleoceanogr., 14, 770–776.

Rosell-Mele, A. and N. Koc, 1997: Paleoclimatic significance of the stratigraphic occurence of

photosynthetic biomarker pigments in the Nordic seas.Geology, 25, 49–52.

Sarnthein, M., U. Pflaumann, and M. Weinelt, 2003: Past extent of sea ice in the northern North

Atlantic inferred from foraminiferal paleotemperature estimates.Paleoceanogr., 18 (27), 1047,

doi:10.1029/2002PA000771.

Seager, R., D. S. Battisti, J. H. Y. N. Gordon, N. Naik, A. Clement, and M. Cane, 2002: Is the Gulf

Stream responsible for Europe’s mild winters?Q. J. R. Meteorol. Soc., 128 (586), 2563–2586.

Shin, S.-I., Z. Liu, B. Otto-Bliesner, E. Brady, J. Kutzbach, and S. Harrison, 2003: A simulation of

the Last Glacial Maximum climate using the NCAR-CCSM.Clim. Dyn., 20, 127–151.

Smith, R. and P. Gent, 2002: Reference manual for the Parallel Ocean Program (POP) ocean

component of the Community Climate System Model (CCSM3.0 and 3.0). Tech. Rep. LA-IR-

02-2484, Los Alamos National Laboratory.

Smith, T. and R. Reynolds, 2003: Extended reconstruction ofglobal sea surface temperatures based

on COADS data (1854-1997).J. Clim., 16, 1495–1510.

31

Son, S.-W. and S. Lee, 2005: The response of westerly jets to thermal driving in a primitive

equation model.J. Atmos. Sci., 62, 3741–3757.

Timmermann, A., F. Justino, F.-F. Jin, and U. K. H. Goosse, 2004: Surface temperature control

in the north and tropical Pacific during the last glacial maximum. Clim. Dyn., 23, 353–370,

doi:10.1007/s00382-004-0434-9.

Toracinta, E. R., R. J. Oglesby, and D. Bromwich, 2003: Atmospheric response to modified

CLIMAP ocean boundary conditions during the Last Glacial Maximum.J. Clim., 17, 504–522.

Trenberth, K. E., 1991: Storm tracks in the Southern Hemisphere.J. Atmos. Sci., 48, 2159–2178.

Trenberth, K. E. and J. M. Caron, 2001: Estimate of meridional atmosphere and ocean heat trans-

ports.J. Clim., 14, 3433–3443.

Trend-Staid, M. and W. L. Prell, 2002: Sea surface temperature at the Last Glacial Maxi-

mum: A reconstruction using the modern analog technique.Paleoceanogr., 17 (4), 1065, doi:

10.1029/2000PA000506.

Uppala, S., P. Kallberg, A. Simmons, U. Andrae, V. da Costa Bechtold, M. Fiorino, J. Gib-

son, J. Haseler, A. Hernandez, G. Kelly, X. Li, K. Onogi, S. Saarinen, N. Sokka, R. Allan,

E. Andersson, K. Arpe, M. Balmaseda, A. Beljaars, L. van de Berg, J. Bidlot, N. Bormann,

S. Caires, F. Chevallier, A. Dethof, M. Dragosavac, M. Fisher, M. Fuentes, S. Hagemann,

E. Holm, B. Hoskins, L. Isaksen, P. Janssen, R. Jenne, A. McNally, J.-F. Mahfouf, J.-J. Mor-

crette, N. Rayner, R. Saunders, P. Simon, A. Sterl, K. Trenberth, A. Untch, D. Vasiljevic,

P. Viterbo, and J. Woollen, 2005: The ERA-40 re-analysis.Quart. J. Roy. Meteor. Soc., 131,

2961–3012.

Yin, J. H., 2002: The peculiar behavior of baroclinic waves during the midwinter suppression of

the Pacific storm track. Ph.D. thesis, University of Washington.

32

Yin, J. H. and D. S. Battisti, 2001: The importance of tropical sea surface temperature patterns in

simulations of Last Glacial Maximum climate.J. Clim., 14, 565–581.

Zurita-Gotor, P. and E. K. M. Chang, 2005: The impact of zonalpropagation and seeding on

the eddy and mean flow equilibrium of a zonally varying two-layer model.J. Atmos. Sci., 62,

2261–2273, doi:10.1175/JAS3473.1.

33

List of Figures

1 Wintertime jet position in simulations of present day and Last Glacial Maximum

climate. Each panel shows the 70th percentile of 200 mb horizontal wind from one

model; warm colours show the present day jet and cool coloursshow the glacial

jet. The contour intervals appear at the top right of the panel and were chosen to

delineate clearly the jet position in each case. The colour scales give the wind speed

(m/s). The results shown are (a–b) two PMIP1 models coupled to slab oceans,

(c–g) five PMIP1 models forced with standard PMIP1 boundary conditions and

(h) the atmosphere component (CAM3) of CCSM3 forced with PMIP1 boundary

conditions. Included for reference are (i) ECMWF ERA-40 reanalysis data for

1958–2001 and (j) the CCSM3 fully coupled simulation. Details of the models

appear in Table 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2 Wintertime atmospheric circulation in CCSM3 simulations. Zonal wind at 250 mb

from the (a) PD and (b) LGM simulations (10 m s−1 contours). The Atlantic jet is

stronger and more zonal while the Pacific jet is largely unchanged. Geopotential

height at 500 mb from the (c) PD and (d) LGM simulations (120 m contours with

an offset of -5400 m). The Laurentide ice sheet over North America forces a strong

stationary wave that intensifies the flow downstream. The thick solid lines in all

the maps denote the zero contour. . . . . . . . . . . . . . . . . . . . . . . .. . . . 41

34

3 Wintertime zonal wind and temperature profiles by ocean sector. Zonal wind (con-

tours) and temperature (colours) for DJFM averaged over (top) the Atlantic sector

(90◦W–0◦) and (bottom) the Pacific sector (150◦E–27◦W). The contour interval is

10 m s−1 and the thick solid lines denote the zero contour. The rightmost panels

show the horizontal zonal wind shear in a 10 degree latitude band, marked on each

panel by horizontal bars, equatorward of the jet core (red isPD, blue is LGM). The

sharp midlatitude temperature front and strong jet in the LGM Atlantic are robust

features no matter how the ocean sector is defined. . . . . . . . . .. . . . . . . . 42

4 Atlantic jet and eddy characteristics in CCSM3 coupled simulations (left) and from

reanalysis data (right). These plots show the relationshipbetween the width and

strength of zonal mean zonal winds and northward eddy heat flux vTATL (K m s−1)

at 850 mb over the Atlantic sector (90◦W–0◦). Left: The top panel shows the

monthly mean jet width versus jet strength for 50 Januarys inthe PD (red) and

LGM (blue) simulations; the bottom panels show the strengthof the eddy heat flux

versus jet strength for 25 Januarys. The horizontal line is the mean vTATL for

each simulation, with the vertical dash marking the 95% confidence limits. The

jet strength is the maximum zonal wind speed in the sector; the jet width is the

latitude range over which the jet speed decreases to half itsmaximum value. These

results are robust to the choice of month(s) used to define winter, and to whether

the jet and eddy strengths are taken as the maximum in the zonal mean of different

longitude ranges straddling the jet/storm track core, or asthe area-weighted mean

of the largest 30–50 values at the jet/storm track core. Right: The corresponding

plots for 44 Januarys of reanalysis data in the period 1958–2001. . . . . . . . . . . 43

35

5 Meridional heat transport from observations and simulations. (a) Recent estimates

(Trenberth and Caron 2001) of total (RT, entire shaded area), atmosphere (AT, dark

grey shading) and ocean (OT, light grey shading) heat transport using data from the

Earth Radiation Budget Experiment (ERBE) and the ERA-40 reanalysis product

from ECMWF. The sum of the dark and light grey areas represents the total heat

transport by the climate system. (b) Comparison of heat transports from observa-

tions and in the PD and LGM simulations by CCSM3. The shaded curves are the

same observational estimates shown in the top panel. Maximum uncertainties in

the model simulations are<0.10 PW, which is less than one third of the value of

the largest differences between the simulations. (c) The dry stationary wave (SW)

and total transient (TR) components of atmospheric heat transport in CCSM3 sim-

ulations. The SW heat transport was estimated by taking the vertical integral of

[v∗ T ∗], where∗ denotes departures from the zonal mean. . . . . . . . . . . . . . . 44

6 Wintertime jet position and eddy diagnostics from the ECMWF ERA-40 reanlaysis

data (1958–2001) and in the CCSM3 simulations. colours show(a–c) eddy tem-

perature fluxv′T ′ (K m s−1) at 850 mb, (d–f) eddy kinetic energy(u′2 + v′2)/2

(m2 s−2) at 200 mb and (g–i) zonal momentum fluxu′v′ (m2 s−1) at 200 mb, all

for DJFM. Black contours show zonal wind at 250 mb (10 m s−1 contours start-

ing at 30 m s−1). All eddy fields are calculated from daily data which have been

high-pass filtered to retain variability at periods less than 8 days. . . . . . . . . . . 45

7 Wintertime baroclinicity in CCSM3 simulations. The Eady growth rate parameter,

σ = 0.31f0N−1(∂uh/∂z), in the 850–700 mb layer for DJFM (0.2 day−1 contours

with values greater than 0.8 day−1 shaded). . . . . . . . . . . . . . . . . . . . . . 46

36

8 1-point regression analysis of wintertime eddy structurein CCSM3 simulations.

DJFM regression maps for the present day (left) and LGM (right) climates: (a,d)

v′200 (2 m s−1) meridional wind at 200 mb; (b,e)v′850 (1 m s−1) meridional wind at

850 mb; (c,f)v′850T′850 (5 K m s−1) meridional heat flux at 850 mb. Black contours

shows positive correlations, grey contours show negative correlations and the zero

contour is omitted. The black crosses mark the locations of the reference time

series, which is thev′200 wind at 43◦N, 28◦W for the PD simulation, and at 38◦N,

28◦W for the LGM simulation. The solid grey line marks 40◦N. All eddy fields are

calculated from daily data which have been high-pass filtered to retain variability

at periods less than 8 days. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . 47

9 Midwinter suppression of the Pacific storm track. This is anupdate of the figure

from Nakamura (1992) showing the midwinter suppression of the Pacific storm

track from NCEP reanalysis data 1948-2002. The top panel shows the seasonal

cycle of eddy temperature fluxv′T ′ at 850 mb (2.5 K m s−1 contours) and sur-

face temperature (colours) in the Pacific sector (160E-180E). The bottom panel

shows the maximumv′T ′ (dark red line with 95% confidence limits shown in

pink); superimposed are the seasonal cycles of maximum zonal wind at 250 mb

(solid blue line) and the equator-to-pole temperature gradient ∆T(E-P), actually

taken between 20 and 70N, at 850 mb (dashed blue line), both arbitrarily scaled to

fit on the plot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

37

10 Wintertime jet position and eddy diagnostics in CCSM3 simulations compared to

the simulation PMIPa reproducing the PMIP1 results. Colours show (a,c,e) eddy

temperature fluxv′T ′ (K m s−1) at 850 mb and (b,d,f) zonal momentum fluxu′v′

(m2 s−2) at 200 mb, all for DJFM. Black contours show zonal wind at 250mb

(10 m s−1 contours starting at 30 m s−1). All eddy fields are calculated from daily

data which have been high-pass filtered to retain variability at periods less than 8

days. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

11 Reconstructions of land ice topography over North America at Last Glacial Max-

imum. The left panel shows profiles of the Laurentide ice sheet complex at 21 ka

across latitude 55◦N from the ICE-4G (Peltier 1994) and ICE-5G (Peltier 2004)

deglaciation history data sets. The right panel shows topography and bathymetry

in the North American sector from the ICE-5G data set, with the thick white con-

tour marking the extent of the ice sheet and the black line (A–A’) marking the

latitude of the profiles in the left panel. . . . . . . . . . . . . . . . .. . . . . . . . 50

12 Sea surface conditions during Last Glacial Maximum. (a) CLIMAP reconstruction

compared to present day observations. (b) LGM as simulated by the coupled model

forced by ICE-5G and 21 ka insolation and greenhouse gases, compared to present

day observations. (c) LGM compared to CLIMAP. Contours in top panels show

annual mean sea surface temperature (4◦C), and colours show differences (◦C).

The white contours mark the 50% sea ice concentration line for August (dashed)

and February (solid). Observed SSTs and sea ice are taken from the period 1943–

2001 from the ERSST (Smith and Reynolds 2003) and HadlSST (Raynor et al.

2003) reconstructions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . 51

38

13 Sea surface conditions in the North Atlantic during Last Glacial Maximum from

reconstructions and as simulated by CCSM3. The reconstructions are from the

CLIMAP project (CLIMAP 1981), the Glacial Atlantic Ocean Mapping (GLAMAP)

project (Sarnthein et al. 2003) and Meland et al. (2005). Colours show sea surface

temperature (◦C), with the 50% sea ice concentration line marked by the thick

white contour, for August (top) and February (bottom). . . . .. . . . . . . . . . . 52

39

8 m/s,7a GENESIS2som 6 m/s,8b UGAMPsom

7 m/s,7c GENESIS2 7 m/s,6d UGAMP

6 m/s,7e CCC GCM2 8 m/s,8f ECHAM3

6 m/s,7g LMD5.3 8 m/s,7h CAM3

7 m/si ECMWF 7 m/s,10j CCSM3

wind speed (m/s)

40 45 50 55 60 65

FIG. 1. Wintertime jet position in simulations of present day and Last Glacial Maximum climate. Each panel

shows the 70th percentile of 200 mb horizontal wind from one model; warm colours show the present day jet

and cool colours show the glacial jet. The contour intervalsappear at the top right of the panel and were chosen

to delineate clearly the jet position in each case. The colour scales give the wind speed (m/s). The results shown

are (a–b) two PMIP1 models coupled to slab oceans, (c–g) five PMIP1 models forced with standard PMIP1

boundary conditions and (h) the atmosphere component (CAM3) of CCSM3 forced with PMIP1 boundary

conditions. Included for reference are (i) ECMWF ERA-40 reanalysis data for 1958–2001 and (j) the CCSM3

fully coupled simulation. Details of the models appear in Table 1.

40

DJFM

aPRESENT DAY

JJA

DJFM

bLGM

JJA

DJFM

c

JJA

DJFM

d

JJA

FIG. 2. Wintertime atmospheric circulation in CCSM3 simulations. Zonal wind at 250 mb from the (a) PD and

(b) LGM simulations (10 m s−1 contours). The Atlantic jet is stronger and more zonal whilethe Pacific jet is

largely unchanged. Geopotential height at 500 mb from the (c) PD and (d) LGM simulations (120 m contours

with an offset of -5400 m). The Laurentide ice sheet over North America forces a strong stationary wave that

intensifies the flow downstream. The thick solid lines in all the maps denote the zero contour.

41

ATL

PRESENT DAYp(

mb)

200

400

600

800

1000ATL

LGM

latitude

p (m

b)

PAC30 45 60

200

400

600

800

1000

latitude

PAC30 45 60 0 2 4

m/s/deg

K190 212 234 256 278 300

FIG. 3. Wintertime zonal wind and temperature profiles by ocean sector. Zonal wind (contours) and temper-

ature (colours) for DJFM averaged over (top) the Atlantic sector (90◦W–0◦) and (bottom) the Pacific sector

(150◦E–27◦W). The contour interval is 10 m s−1 and the thick solid lines denote the zero contour. The right-

most panels show the horizontal zonal wind shear in a 10 degree latitude band, marked on each panel by

horizontal bars, equatorward of the jet core (red is PD, blueis LGM). The sharp midlatitude temperature front

and strong jet in the LGM Atlantic are robust features no matter how the ocean sector is defined.

42

jet w

idth

(la

titud

e)

20

30

40

50

60

jet strength (m/s)

vTA

TL (

Km

s−1 )

35 45 55 65

2

3

4

5

LGMPD

jet strength (m/s)35 45 55 65

NCEPECMWF

FIG. 4. Atlantic jet and eddy characteristics in CCSM3 coupled simulations (left) and from reanalysis data

(right). These plots show the relationship between the width and strength of zonal mean zonal winds and

northward eddy heat flux vTATL (K m s−1) at 850 mb over the Atlantic sector (90◦W–0◦). Left: The top

panel shows the monthly mean jet width versus jet strength for 50 Januarys in the PD (red) and LGM (blue)

simulations; the bottom panels show the strength of the eddyheat flux versus jet strength for 25 Januarys. The

horizontal line is the mean vTATL for each simulation, with the vertical dash marking the 95% confidence

limits. The jet strength is the maximum zonal wind speed in the sector; the jet width is the latitude range over

which the jet speed decreases to half its maximum value. These results are robust to the choice of month(s)

used to define winter, and to whether the jet and eddy strengths are taken as the maximum in the zonal mean

of different longitude ranges straddling the jet/storm track core, or as the area-weighted mean of the largest

30–50 values at the jet/storm track core. Right: The corresponding plots for 44 Januarys of reanalysis data in

the period 1958–2001.

43

−6

−3

0

3

6

PW

aECMWF ATECMWF OT

PW

b

−6

−3

0

3

6 PD RTLGM RTPD OTLGM OT

−90 −60 −30 0 30 60 90−6

−3

0

3

6

latitude

PW

transient eddytransport

dry stationarywave transport

cPD TRLGM TRPD SWLGM SW

FIG. 5. Meridional heat transport from observations and simulations. (a) Recent estimates (Trenberth and

Caron 2001) of total (RT, entire shaded area), atmosphere (AT, dark grey shading) and ocean (OT, light grey

shading) heat transport using data from the Earth RadiationBudget Experiment (ERBE) and the ERA-40 re-

analysis product from ECMWF. The sum of the dark and light grey areas represents the total heat transport

by the climate system. (b) Comparison of heat transports from observations and in the PD and LGM simula-

tions by CCSM3. The shaded curves are the same observationalestimates shown in the top panel. Maximum

uncertainties in the model simulations are<0.10 PW, which is less than one third of the value of the largest

differences between the simulations. (c) The dry stationary wave (SW) and total transient (TR) components of

atmospheric heat transport in CCSM3 simulations. The SW heat transport was estimated by taking the vertical

integral of [v∗ T ∗], where∗ denotes departures from the zonal mean.

44

g ECMWF

ua va at 200 mb

h PD

i LGM

a ECMWF

va Ta at 850 mb

b PD

c LGM

d ECMWF

EKE at 200 mb

e PD

f LGM

0

20

40

0

48

96

0

10

20

FIG. 6. Wintertime jet position and eddy diagnostics from the ECMWF ERA-40 reanlaysis data (1958–2001)

and in the CCSM3 simulations. colours show (a–c) eddy temperature fluxv′T ′ (K m s−1) at 850 mb, (d–f)

eddy kinetic energy(u′2 + v′2)/2 (m2 s−2) at 200 mb and (g–i) zonal momentum fluxu′v′ (m2 s−1) at 200 mb,

all for DJFM. Black contours show zonal wind at 250 mb (10 m s−1 contours starting at 30 m s−1). All eddy45

PRESENT DAY LGM

FIG. 7. Wintertime baroclinicity in CCSM3 simulations. The Eady growth rate parameter,

σ = 0.31f0N−1(∂uh/∂z), in the 850–700 mb layer for DJFM (0.2 day−1 contours with values greater than

0.8 day−1 shaded).

46

15

30

45

60

v200

(2 ms−1)

a

latit

ude

PRESENT DAY

15

30

45

60

v200

(2 ms−1)

d

LGM

15

30

45

60

v850

(1 ms−1)

b

latit

ude

15

30

45

60

v850

(1 ms−1)

e

−90 −60 −30 0 30

15

30

45

60

vT850

(5 Kms−1)

c

latit

ude

longitude−90 −60 −30 0 30

15

30

45

60

vT850

(5 Kms−1)

f

longitude

FIG. 8. 1-point regression analysis of wintertime eddy structure in CCSM3 simulations. DJFM regression

maps for the present day (left) and LGM (right) climates: (a,d) v′200 (2 m s−1) meridional wind at 200 mb; (b,e)

v′850 (1 m s−1) meridional wind at 850 mb; (c,f)v′850T′850 (5 K m s−1) meridional heat flux at 850 mb. Black

contours shows positive correlations, grey contours show negative correlations and the zero contour is omitted.

The black crosses mark the locations of the reference time series, which is thev′200 wind at 43◦N, 28◦W for the

PD simulation, and at 38◦N, 28◦W for the LGM simulation. The solid grey line marks 40◦N. All eddy fields are

calculated from daily data which have been high-pass filtered to retain variability at periods less than 8 days.

47

latit

ude

JAN APR JUL OCT JAN APR

15

45

75

eddy

hea

t flu

x 85

0mb

[Km

/s]

JAN APR JUL OCT JAN APR

5

10

15

20

eddy heat fluxu 250mb∆T(E−P) 850mb

FIG. 9. Midwinter suppression of the Pacific storm track. This isan update of the figure from Nakamura (1992)

showing the midwinter suppression of the Pacific storm trackfrom NCEP reanalysis data 1948-2002. The top

panel shows the seasonal cycle of eddy temperature fluxv′T ′ at 850 mb (2.5 K m s−1 contours) and surface

temperature (colours) in the Pacific sector (160E-180E). The bottom panel shows the maximumv′T ′ (dark red

line with 95% confidence limits shown in pink); superimposedare the seasonal cycles of maximum zonal wind

at 250 mb (solid blue line) and the equator-to-pole temperature gradient∆T(E-P), actually taken between 20

and 70N, at 850 mb (dashed blue line), both arbitrarily scaled to fit on the plot.

48

a PD v′ T′ at 850 mb

c LGM

e PMIPa

0 10 20

b PD u′ v′ at 200 mb

d LGM

f PMIPa

0 20 40

FIG. 10. Wintertime jet position and eddy diagnostics in CCSM3 simulations compared to the simulation

PMIPa reproducing the PMIP1 results. Colours show (a,c,e) eddy temperature fluxv′T ′ (K m s−1) at 850 mb

and (b,d,f) zonal momentum fluxu′v′ (m2 s−2) at 200 mb, all for DJFM. Black contours show zonal wind at

250 mb (10 m s−1 contours starting at 30 m s−1). All eddy fields are calculated from daily data which have

been high-pass filtered to retain variability at periods less than 8 days.

49

ò�

ò�

ò�

ò�

0

�

�

�

90W ��>���>

3 km

��RT

1 km

��RT

A’

A A’

A

ICE-5G

ICE-4G

km

FIG. 11. Reconstructions of land ice topography over North America at Last Glacial Maximum. The left panel

shows profiles of the Laurentide ice sheet complex at 21 ka across latitude 55◦N from the ICE-4G (Peltier 1994)

and ICE-5G (Peltier 2004) deglaciation history data sets. The right panel shows topography and bathymetry in

the North American sector from the ICE-5G data set, with the thick white contour marking the extent of the ice

sheet and the black line (A–A’) marking the latitude of the profiles in the left panel.

50

CLIMAP−OBS

60° S

30° S

0°

30° N

60° N

LGM−OBS

60° S

30° S

0°

30° N

60° N

−15−10 −8 −6 −3 −2 −1−0.5 0 0.5 1 1.5

LGM−CLIMAP

180° W 120° W 60° W 0° 60° E 120° E 180° E

60° S

30° S

0°

30° N

60° N

FIG. 12. Sea surface conditions during Last Glacial Maximum. (a) CLIMAP reconstruction compared to

present day observations. (b) LGM as simulated by the coupled model forced by ICE-5G and 21 ka insolation

and greenhouse gases, compared to present day observations. (c) LGM compared to CLIMAP. Contours in top

panels show annual mean sea surface temperature (4◦C), and colours show differences (◦C). The white contours

mark the 50% sea ice concentration line for August (dashed) and February (solid). Observed SSTs and sea ice

are taken from the period 1943–2001 from the ERSST (Smith andReynolds 2003) and HadlSST (Raynor et al.

2003) reconstructions.

51

Meland et al. 2005CLIMAP GLAMAP CCSM3

ï�ï� 0 � � 3 4 5 6 7 8 9 ��

FIG. 13. Sea surface conditions in the North Atlantic during Last Glacial Maximum from reconstructions and

as simulated by CCSM3. The reconstructions are from the CLIMAP project (CLIMAP 1981), the Glacial

Atlantic Ocean Mapping (GLAMAP) project (Sarnthein et al. 2003) and Meland et al. (2005). Colours show

sea surface temperature (◦C), with the 50% sea ice concentration line marked by the thick white contour, for

August (top) and February (bottom).

52

List of Tables

1 Atmosphere models shown in Figure 1. The model abbreviations are: Global En-

vironmental and Ecological Simulation of Interactive Systems (GENESIS2), the

U.K. Universities’ Global Atmospheric Modelling Project (UGAMP), Canadian

Centre for Climate Modelling and Analysis (CCC GCM2), European Centre Ham-

burg Model (ECHAM3.6), Laboratoire de Meteorologie Dynamique (LMD5.3),

National Center for Atmospheric Research Community Atmosphere Model (CAM3).

The second column indicates whether the model participatedin PMIP1; the next

columns indicate the resolution of the model. . . . . . . . . . . . .. . . . . . . . 54

2 Jet and eddy characteristics in the Atlantic sector (15◦N–65◦N, 90◦W–0◦) for

DJFM winter. uATL is the maximum zonal wind at 200 mb in the sector;σu is the

standard deviation of uATL; KEIATL is the sector-mean monthly departure from

the climatological mean of the kinetic energy of the horizontal wind at 200 mb;

and vTATL is the sector-mean northward eddy heat flux at 850 mb. The 95% con-

fidence intervals of the seasonally averaged quantities were determined using a

Student’s t-test for the means, and a chi-squared test for the standard deviations. . . 55

53

TABLE 1. Atmosphere models shown in Figure 1. The model abbreviations are: Global Environmental and

Ecological Simulation of Interactive Systems (GENESIS2),the U.K. Universities’ Global Atmospheric Mod-

elling Project (UGAMP), Canadian Centre for Climate Modelling and Analysis (CCC GCM2), European Cen-

tre Hamburg Model (ECHAM3.6), Laboratoire de Meteorologie Dynamique (LMD5.3), National Center for

Atmospheric Research Community Atmosphere Model (CAM3). The second column indicates whether the

model participated in PMIP1; the next columns indicate the resolution of the model.

Model PMIP1 lat lon levels

GENESIS2 Y 64 128 18 spectral

UGAMP Y 64 128 19 spectral

CCC GCM2 Y 48 96 10 spectral

ECHAM3.6 Y 64 128 19 spectral

LMD5.3 Y 50 64 11 grid point

CAM3 N 64 128 26 spectral

54

TABLE 2. Jet and eddy characteristics in the Atlantic sector (15◦N–65◦N, 90◦W–0◦) for DJFM winter. uATL is

the maximum zonal wind at 200 mb in the sector;σu is the standard deviation of uATL; KEIATL is the sector-

mean monthly departure from the climatological mean of the kinetic energy of the horizontal wind at 200 mb;

and vTATL is the sector-mean northward eddy heat flux at 850 mb. The 95% confidence intervals of the

seasonally averaged quantities were determined using a Student’s t-test for the means, and a chi-squared test

for the standard deviations.

uATL σu KEIATL vTATL

m s−1 m s−1 m2 s−2 K m s−1

ERA-40 46.0± 1.0 3± 2 31± 2 3.56± 0.07

NCEP 45.9± 0.9 3± 2 31± 2 3.49± 0.08

PD 53.6± 1.0 4± 2 34± 2 3.42± 0.19

LGM 64.7± 0.6 2± 1 22± 1 2.93± 0.13

PMIPa 55.5± 1.3 3± 3 34± 2 3.63± 0.22

55