Rates of thermohaline recovery from freshwater pulses in modern, Last Glacial Maximum, and greenhouse warming climates

Abstract

Recovery rates of the thermohaline circulation after a freshwater pulse in the North Atlantic vary consider-ably depending on the background climate, as demon-strated in the Community Climate System Model. The recovery is slowest in a Last Glacial Maximum (LGM) climate, fastest in a modern climate, and intermedi-ate between the two in a greenhouse warming (4XCO 2) climate. Previously proposed mechanisms to explain thermohaline circulation stability involving altered hor-izontal freshwater transport in the North Atlantic are consistent with relative recovery rates in the modern and 4XCO 2 climates, but fail to explain the slow LGM recovery. Instead, sea ice expansion inhibits deep-water formation after freshening in the LGM climate by re-ducing heat loss to the atmosphere and providing addi-tional surface freshwater. In addition, anomalous verti-cal freshwater transport across ∼1km depth after fresh-ening is most effective at weakening the stratification in the modern case but is negligible in the LGM case.

Full text

GEOPHYSICAL RESEARCH LETTERS, VOL. ???, XXXX, DOI:10.1029/,
Rates of thermohaline recovery from freshwater pulses in
Modern, Last Glacial Maximum and Greenhouse Warming
Climates
C. M. Bitz1, J. C. H. Chiang2, W. Cheng3, and J. J. Barsugli4
Recovery rates of the thermohaline circulation after
a freshwater pulse in the North Atlantic vary consider-
ably depending on the background climate, as demon-
strated in the Community Climate System Model. The
recovery is slowest in a Last Glacial Maximum (LGM)
climate, fastest in a modern climate, and intermedi-
ate between the two in a greenhouse warming (4XCO2)
climate. Previously proposed mechanisms to explain
thermohaline circulation stability involving altered hor-
izontal freshwater transport in the North Atlantic are
consistent with relative recovery rates in the modern
and 4XCO2climates, but fail to explain the slow LGM
recovery. Instead, sea ice expansion inhibits deep-water
formation after freshening in the LGM climate by re-
ducing heat loss to the atmosphere and providing addi-
tional surface freshwater. In addition, anomalous verti-
cal freshwater transport across ∼1km depth after fresh-
ening is most effective at weakening the stratification in
the modern case but is negligible in the LGM case.
1. Introduction
Shutdown of the thermohaline circulation (THC) in
a modern climate from enhanced freshwater input of
roughly ∼10 Sv-yr (1 Sv = 106m3s−1) to the North
Atlantic is usually only temporary in climate models:
Once the freshwater input returns to normal, the cir-
culation recovers in a few decades to a century at most
[Stouffer et al., 2006]. These models agree roughly with
the duration of climate change that followed meltwater
pulses approximately 8200 year ago, which resulted in
the largest global cooling in the Holocene [Ellison et al.,
2006]. However, meltwater pulses of a similar magni-
tude during the most recent glacial are hypothesized to
have caused the THC to remain collapsed or greatly re-
duced for much longer durations [see e.g., Clark et al.,
1Dept. of Atmospheric Sciences, University of Washington
2Dept. of Geography and Center for Atmospheric
Sciences, University of California, Berkeley
3JISAO/School of Oceanography, University of
Washington
4CIRES, University of Colorado
Copyright 2006 by the American Geophysical Union.
0094-8276/07/$5.00
2002]. It remains to be seen how anthropogenic climate
change and the anticipated increase in precipitation and
runoff in the northern North Atlantic will influence the
THC response to freshwater pulses, say from Greenland.
Here we investigate the influence of the climate state
on THC recovery following a freshwater addition to the
North Atlantic using the Community Climate System
Model Version 3 (CCSM3). We add the same fresh-
water anomaly to modern, greenhouse warming, and
glacial climates to investigate the mechanisms of THC
recovery in a spectrum of background climates.
2. Model
We impose freshwater perturbations to three nearly
equilibrated background climate cases: a 1,000 yr inte-
gration with “modern” (i.e., 1990s) conditions [Collins
et al., 2006], a 440 yr Last Glacial Maximum LGM (with
ice-sheet topography, ocean bathymetry, orbital config-
uration, and greenhouse gases prescribed for ∼21 ka)
integration [Otto-Bliesner et al., 2006], and a 4XCO2
stabilization integration [Bryan et al., 2006] where CO2
is initially increased at the rate of 1% per year from
present-day to four times present-day levels and then
held fixed for 300 yr. These particular background
climates, or “controls”, were chosen to sample a wide
range of initial THC strengths and structures and sur-
face heat and freshwater fluxes.
We branched freshwater pulse experiments from each
control by instantaneously freshening the upper 970m
of the North Atlantic and Arctic Oceans from 55-90◦N,
90◦W-20◦E by an average of 2psu (higher at the top and
tapering with depth). This is equivalent to adding 16
Sv-yr of freshwater. Our method is similar to Vellinga
et al. [2002] and contrasts with that used in the inter-
comparison study described by Stouffer et al. [2006],
where freshwater is added over 100 yr to the surface.
The latter distributes the freshening over time, while
the former distributes it in depth. Neither is obvi-
ously more realistic, but instantaneous freshening is
more computationally efficient for studying THC re-
covery. We conducted seven freshened runs for a min-
imum of 20 yr each (three modern, three LGM, and
one 4XCO2). One member from each climate is longer
(75 yr for modern and 4XCO2and 135 yr for LGM) to
capture the decade to century time-scale recovery. Ad-
1

X - 2 BITZ ET AL: LGM, MODERN, AND GREENHOUSE WARMING THC RECOVERY
ditional information about the model and experiments
is given in the auxiliary materials.
3. Results
Freshening in the North Atlantic causes the THC to
immediately collapse with a cessation of North Atlantic
Deep Water (NADW) formation in all seven runs as
shown in Fig. 1. In no case is the collapse permanent,
but the recovery rates are strikingly different depending
on the background climate state. The range of recov-
ery rates within each three-member ensemble is clearly
much smaller. After a complete collapse and even re-
versal in year one, every run exhibits an initial rapid
increase to 2–5 Sv by year 4, and then weakens again in
years 5–10 (see Fig 1). After year 10 the modern THC
rapidly recovers at a rate of ∼0.2 Sv yr−1. In contrast,
the LGM THC recovery rate is about 10 times slower,
or ∼0.02 Sv yr−1. The 4XCO2recover rate is ∼0.07 Sv
yr−1— intermediate to the other two. According to Fig
1, the THC recovery takes about ∼40 yr in the mod-
ern case and about ∼70 yr in the 4XCO2case, while
the LGM case appears as though it could take several
centuries. This dependence of the THC recovery rate
on the climate state is the main finding of this study.
In the rest of this section we evaluate several possible
explanations.
One might expect the recovery rate to depend on
the degree to which the THC is controlled by thermal
versus haline forcing in the different climates. Using a
simple box model, Rahmstorf [1996] argued that the di-
rection of freshwater transport by the meridional over-
turning circulation, WTHC, at 33◦S indicates whether
the THC at equilibrium is exclusively thermally driven
(southward WTHC) or both thermally and haline driven
(northward WTHC). WTHC is computed from
WTHC =−
1
SoZ¯v(¯
S−So)dz, (1)
where ¯vis the zonally integrated northward velocity,
¯
Sis the zonal mean salinity, and Sois the reference
salinity, taken here as the global mean salinity. Rahm-
storf [1996] explained that a southward WTHC salini-
fies the Atlantic, so at equilibrium the sum of all other
Atlantic freshwater contributers must brake the THC,
rather than drive it. In this case, a THC shutdown
eliminates a salinity source to the Atlantic, which sus-
tains the shutdown by providing a positive feedback.
If the shutdown is only temporary, then positive feed-
backs slow the recovery. A parallel argument suggests
a northward WTHC yields a negative feedback, which
aids recovery. de Vries and Weber [2005] found this rea-
soning worked well to describe a series of experiments
in an intermediate complexity model where they var-
ied WTHC at 33◦S by applying surface freshwater flux
adjustments.
Table 1. THC recovery rates and Atlantic northward fresh-
water fluxes
Background Recovery Control Perturbed∗
Climate Rate WTHC 33◦S ∆Wgyre 33◦N
LGM 0.02 Sv yr−10.26 Sv -0.47 Sv
Modern 0.2 Sv yr−10.03 Sv -0.33 Sv
4XCO20.07 Sv yr−1-0.02 Sv -0.30 Sv
∗2nd decade after freshening
0 20 40 60 80 100
−5
0
5
10
Sinking Rate − Sv
Year Since Freshening
Figure 1. Annual mean thermohaline circulation in-
dex (solid lines = modern, dashed = LGM, dot-dash =
4XCO2). The light gray lines show the mean index for
the corresponding control runs with error bar indicat-
ing plus/minus one standard deviation. The index is the
sinking flux across 1022 m depth from 60-65◦N, which
emphasizes changes in NADW formation rate, as sug-
gested by Gent [2001]. Positive (negative) values indi-
cate sinking (upwelling). An index of the THC using the
maximum of the Atlantic meridional overturning stream-
function yields the same general conclusions (see auxil-
iary materials).
In our runs, only the 4XCO2control has a southward
WTHC (see Table 1). Because freshening the 4XCO2
climate does not cause a permanent THC collapsed,
other negative feedbacks must be operative. However,
if we assume these other feedbacks are about the same
strength for all three climate states, so that WTHC de-
termines the relative rate of recovery, then the LGM
case should recover fastest, with the modern case next,
and the 4XCO2case last. Because the LGM case is
actually the slowest to recover, we can conclude that
other feedbacks must control the LGM recovery.
Vellinga et al. [2002] argued that the THC recovery
after freshening was controlled by anomalous salt ad-
vection from the North Atlantic subtropical gyre. The
salinity flux by the gyre is typically expressed in terms
of a freshwater flux:
Wgyre =−
1
SoZv0S0dz, (2)
where S0and v0are deviations from the zonal means and
the overbar indicates a zonal integral. After freshening,
Vellinga et al. [2002] found Wgyre at 25◦N decreased

BITZ ET AL: LGM, MODERN, AND GREENHOUSE WARMING THC RECOVERY X - 3
owing to a southward shift of the Atlantic ITCZ. The
nature of the ITCZ shift after freshening depends on the
mean climate state (Cheng et al, submitted). The ITCZ
shift serves as a negative feedback on the THC, such
that a larger decrease in Wgyre in the subtropical North
Atlantic promotes a faster recovery. However, Wgyre
at 33◦N in the Atlantic decreases more after freshening
the LGM climate than the modern or 4XCO2climates
(see Table 1), and thus cannot explain the slow LGM
recovery.
The role of surface buoyancy forcing on NADW for-
mation is another potential factor influencing the THC
recovery rate. The presence of the sea ice strongly al-
ters the surface buoyancy forcing in the subpolar seas,
so we compare the sea ice cover in relation to the sur-
face density in the freshened region for the three control
climate states. Figure 2a shows that in the LGM con-
trol sea ice of 30–80% concentration covers the densist
outcroppings, where NADW forms (also roughly where
the mixed layer reaches its maximum depth, see auxil-
iary material.), suggesting a larger haline influence on
the THC in the LGM climate. This is in agreement
with paleoclimate evidence of brine rejection contribut-
ing to glacial NADW formation [Vidal et al., 1998]. In
contrast, NADW formation occurs south of the sea ice
edge in the modern and 4xCO2 controls (Fig. 2b and
c). Freshening causes the sea ice cover to expand in all
cases, but little sea ice reaches the NADW formation
sites in the modern and 4XCO2climates. This disparity
in sea ice cover among the climates changes the nature
of feedbacks involving the surface fluxes that drive the
THC recovery through deep-water formation.
To quantify the effect of surface buoyancy flux forc-
ing on watermass formation [see Speer and Tziperman,
1992], we first compute components of the surface den-
sity flux, −ραFTand ρβFS, where ρis the seawater
density, αand βare seawater thermal and haline con-
traction coefficients, and FTand FSare surface heat
and freshwater fluxes. The density flux components
are integrated over density outcroppings in the Atlantic
north of 55◦N, south of the Arctic Ocean, and west of
the Barents Sea (henceforth called the “North Atlantic
freshened region”) to arrive at a watermass transfor-
mation function τS,T . Finally we take the derivative,
∂τS,T /∂ρ, to give the surface watermass formation rate
(WFR) as a function of ρ(with units Sv(kg/m3)−1) for
each component. As such, the only other sources of
buoyancy forcing to a given density range are from di-
apycnal mixing and boundary flow. If we imagine plac-
ing a bottom “boundary” on isopycnals at 1022 m depth
a) b) c)
1
2
3
4
5d) LGM−haline
Decade
1.026 1.028 1.030
0
g cm−3
1
2
3
4
5e) LGM−thermal
1.026 1.028 1.030
0
g cm−3
1
2
3
4
5f) Modern−thermal
1.026 1.028 1.030
0
g cm−3
1
2
3
4
5g) 4XCO2−thermal
1.026 1.028 1.030
0
g cm−3
−40
−20
0
20
40
Figure 2. (a-c) Annual mean surface density in the controls in g cm−3with the 15% sea ice concentration contour in
the controls (solid lines) and in the second decade after freshening (dashed lines). (d-g) Watermass formation rate in Sv
(kg/m3)−1for the North Atlantic freshened region in the control (decade 0) and after freshening (decades 1–5). Haline
components for the modern and 4XCO2are negligible and not shown.

X - 4 BITZ ET AL: LGM, MODERN, AND GREENHOUSE WARMING THC RECOVERY
in the sinking region, then the flow across this boundary
(i.e., our index of the THC in Fig. 1) is equal to the
WFR from surface fluxes summed over all isopycnals
that cross this boundary plus the diapycnal mixing into
these isopycnals between the surface and 1 km depth.
Figure 2d-g shows WFR for the North Atlantic
freshened region for all non-negligible thermal and ha-
line components in our runs. The rate that surface
fluxes further densify the heaviest outcroppings (cre-
ating NADW) is indicated by the magnitude of the
rightmost maxima. The thermal component dominates
NADW formation in all three controls and also after
freshening in the modern and 4XCO2climates. Only
the LGM case develops a substantial haline component
after freshening. Freshening shifts the whole pattern
to the left, as outcroppings are made lighter. Then as
the freshwater anomaly erodes, the pattern travel right-
ward, back towards its position in the control. (How-
ever, the physical location of the heaviest outcroppings
and deepest mixed layer sites are not altered by fresh-
ening, see auxiliary material.) In the modern case, sur-
face heat fluxes create dense watermasses at an even
higher rate than in the control within two decades af-
ter freshening, despite the shift in density (Fig. 2f ).
In the 4XCO2case, surface heat fluxes create dense
watermasses after freshening at about 70% of the con-
trol rate (Fig. 2g). In contrast, the LGM case has
greatly reduced dense water formation after freshening
(Fig. 2d and e) owing to expanded sea ice cover over
deep-water outcroppings. Sea ice in the LGM climate
is transported further south and east, where it reduces
heat loss and provides a new source of surface fresh-
water after freshening that destroys outcroppings with
ρ∼1.029gcm−3. There is some compensating increased
brine rejection along with reduced precipitation (dis-
cussed in the auxiliary material) over less dense waters,
with ρ∼1.028gcm−3(Fig. 2d). Such a mechanism
involving sea ice was proposed by Stocker et al. [2001].
The reversal of the effective salt flux over the densest
water outcroppings, counteracting the thermal forcing
after freshening, is the signature of a fundamental dif-
ference between the LGM climate and the modern and
4XCO2climates.
The WFR we computed only depends on surface
buoyancy forcing, yet the THC strength and struc-
ture also depends on the internal ocean stratification.
Among our control simulations, the modern ocean is
the least stable below about 900 m (see Fig. 3a), con-
sistent with its higher sinking rate in Fig. 1 and the
greater sinking depths reached by its NADW compared
to the other controls [Otto-Bliesner et al., 2006; Bryan
et al., 2006]. Comparably greater stratification in the
LGM control is due mainly to salinity, in broad agree-
ment with paleo evidence [Labeyrie, 1992], while in the
4XCO2control it is due mainly to temperature (not
shown). Weaker stratification below the freshened layer
in the modern case allows faster vertical dissipation of
the freshwater anomaly, as can be seen in the potential
density anomaly five decades after freshening in Fig.
3b. The downward anomalous freshwater flux at 1022
m and 1501 m depths is largest in the modern case and
smallest in the LGM case (Fig.3c). Anomalous down-
ward freshwater transport erodes the high stratification
that was imposed by freshening and assists in the THC
recovery. This salinity homogenization after freshening
is most effective in the modern case and least effective
in the LGM case. After five decades, LGM and 4XCO2
anomaly profiles below 1km depth are clearly more sta-
bilizing than the modern anomaly profile (Fig. 3b).
The strong implication is that the originally more sta-
ble density profiles below the surface layer in the LGM
and 4XCO2controls inhibit dissipation of the salinity
anomaly by transport [as noted by Tziperman et al.,
1994] and by convection [as noted by Lenderink and
Haarsma, 1993].
1.025 1.03
0
1
2
3
4
Depth − km
g cm−3
a)
−1 −0.5 0
0
1
2
3
4
10−3 g cm−3
b)
1 2 3 4 5
−0.05
0
0.05
0.1 c)
Sv
Dedade After Freshening
Figure 3. Density profiles in the (a) controls and (b)
anomalies five decade after freshening and anomalous
vertical freshwater flux (c) through 1022m (black) and
1501m (grey). All quantities are averaged over the North
Atlantic freshened region. The reference salinity used to
compute the freshwater flux in (c) is the average salin-
ity above 3.125 km depth in the region. (Solid line =
modern, dashed = LGM, dot-dash = 4XCO2.)
We also examined the horizontal freshwater transport
into and out of the northern North Atlantic (integrated
over the entire water column in the freshened region),
as well as the surface freshwater flux. The total rate
at which the freshwater anomaly dissipates from the
water column depends little on the climate state (see
auxiliary material). Hu et al. [2007] found the THC
recovers more rapidly when the Bering Strait is open
(compared to closed) in a modern climate in CCSM ver-

BITZ ET AL: LGM, MODERN, AND GREENHOUSE WARMING THC RECOVERY X - 5
sion 2 after adding freshwater at the rate of 1 Sv for 100
yr. Because their freshwater was added at the surface,
the anomaly remains more surface bound. In turn, a
sizeable fraction escapes through the 50 m deep Bering
Strait when it is open. In contrast, when we prescribe
less than 1/6 of the total freshwater anomaly that was
used by Hu et al and we distribute it over the upper
970 m, we find only about 1/16 of this anomaly escapes
through Bering Strait in our modern case. Thus Fresh-
water need not escape through Bering Strait in CCSM3
for the THC recovery to be rapid. It is also noteworthy
that the THC recovery in our 4XCO2freshened case is
much less rapid than in the modern case, despite Bering
Strait being open in both.
4. Summary and Conclusions
We have shown that the recovery rate from a tempo-
rary collapse of the thermohaline circulation after fresh-
ening in the north Atlantic depends on the background
climate state in CCSM3. Clearly, the recovery rate from
a given freshwater pulse in models with present day con-
ditions is not a good indication of the recovery rate from
a pulse of the same magnitude and location in a 4XCO2
or glacial climate.
We analyzed recovery mechanism that are influenced
by horizontal freshwater transport by the mean merid-
ional circulation at 33◦S and by the subtropical gyre in
the North Atlantic. If these were the dominant feedback
mechanism in our model, then the LGM case would
have the fastest recovery followed by the modern, with
the 4XCO2case being the slowest.
A mechanism that slows the recovery in the LGM
case is the expansion of sea ice in the NADW forma-
tion regions. NADW formation via surface fluxes in the
LGM freshened case alone is inhibited by sea ice ex-
pansion, which effectively reduces surface heat loss and
supplies meltwater over deep-water outcroppings. In
contrast, surface heat loss drives NADW formation in
the modern and 4XCO2climates within two decades af-
ter freshening. Sea ice in poised to play a greater role in
the LGM climate because NADW production sites are
partially covered with sea ice even before freshening.
After freshening the combination of the anomalous
horizontal freshwater transport and surface freshwater
flux in the North Atlantic freshened region are roughly
the same in all three cases. Therefore the dissipation
rate of the total freshwater anomaly through the sur-
face and lateral boundaries cannot explain the differ-
ing recover rates. Instead, we find that the anomalous
downward freshwater transport across ∼1km depth is
effective at weakening the stratification in the modern
case after freshening, while the same flux is smaller in
the 4XCO2and nearly zero in the LGM case.
Acknowledgments. We gratefully acknowledge support
from the National Science Foundation through grant ATM-
0502204. Computational facilities were provided by the National
Center for Atmospheric Research (NCAR).
References
Bryan, F., G. Danabasoglu, N. Nakashiki, Y. Yoshida, D.-H.
Kim, J. Tsutsui, and S. Doney, Response of the North At-
lantic thermohaline circulation and venticlation to increasing
carbon dioxide in CCSM3, J. Climate,19, 2382–2397, 2006.
Clark, P. U., N. G. Pisias, T. F. Stocker, and A. J. Weaver, The
role of the thermohaline circulation in abrupt climate change,
Nature,415, 863–869, 2002.
Collins, W. D., et al., The Community Climate System Model,
Version 3, J. Climate,19, 2122–2143, 2006.
de Vries, P., and S. L. Weber, The atlantic freshwater budget
as a diagnostic for the existence of a stable shut down of the
meridional overturning circulation, Geophys. Res. Lett.,32,
doi:10.1029/2004GL021450, 2005.
Ellison, C. R. W., M. R. Chapman, and I. R. Hall, Surface and
deep ocean interactions during the cold climate event 8200
years ago, Science,312, 1929–1932, 2006.
Gent, P. R., Will the North Atlantic Ocean thermohaline circu-
lation weaken during the 21st century?, Geophys. Res. Lett.,
28, 1023–1028, 2001.
Hu, A., G. A. Meehl, and W. Han, Role of the Bering Stait in the
thermohaline circulation and abrupt climate change, Geophys.
Res. Lett.,in press, 2007.
Labeyrie, L., Changes in vertical structure of the North Atlantic
Ocean between glacial and modern times, Quat. Sci. Rev,11,
401–413, 1992.
Lenderink, G., and R. J. Haarsma, Variability and multiple equi-
libria of the thermohline circulation associated with deep-
water formation, J. Phys. Oceanogr.,24, 1480–1493, 1993.
Otto-Bliesner, B. L., E. C. Brady, G. Clauzet, R. Tomas, S. Levis,
and Z. Kothavala, Last glacial maximum and holocene climate
in CCSM3, J. Climate,19, 2526–2544, 2006.
Rahmstorf, S., On the freshwater forcing and transport of the
Atlantic thermohaline circulation, Clim. Dyn.,12, 799–811,
1996.
Speer, K., and E. Tziperman, Rates of watermass formation in
the North Atlantic Ocean, J. Phys. Oceanogr.,22, 93–110,
1992.
Stocker, T. F., R. Knutti, and G.-K. Plattner, The future of
the thermohaline circulatio — a perspective, in The Ocean
and Rapid Climate Changes: Past, Present, and Future, pp.
277–293. Geophysical Monograph 126, American Geophysical
Union, 2001.
Stouffer, R. J., et al., Investigating the causes of the response
of the thermohaline circulation to past and future climate
changes, J. Climate,19, 1365–1387, 2006.
Tziperman, E., J. R. Toggweiler, Y. Feliks, and K. Bryan, In-
stability of the thermohaline circulation with respect to mixed
boundary conditions: Is it really a problem for realistic models,
J. Phys. Oceanogr.,24, 217–232, 1994.
Vellinga, M., R. Wood, and J. M. Gregory, Processes govern-
ing the recovery of a perturbed thermohaline circulation in
HadCM3, J. Climate,16, 2002.
Vidal, L., L. Labeyrie, and T. C. E. van Weering, Benthic δ18O
records in the North Atlantic over the last glacial period, Pa-
leoceanography,13, 245–251, 1998.
C. M. Bitz, Dept. of Atmospheric Sciences, MS
351640, University of Washington, Seattle, WA 98195-1640.
(bitz@atmos.washington.edu)

X - 6 BITZ ET AL: LGM, MODERN, AND GREENHOUSE WARMING THC RECOVERY
Auxiliary Materials - online only
The Community Climate System Model (CCSM3)
used in this study has an atmosphere component with
approximately 2.8◦horizontal resolution (T42 spectral
truncation) and 26 vertical levels. The ocean and sea
ice have zonal resolution of 1.125◦and the meridional
resolution of 0.54◦, except in the subtropics and trop-
ics where the meridional resolution is finer. CCSM3
is well documented and may be downloaded freely from
http://www.ccsm.ucar.edu. Present-day forcing is fixed
at 1990 levels for anthropogenic greenhouse gases (CO2
= 355ppm, CH4= 1714ppb, and N2O= 311ppb), ozone,
and aerosols. and has LGM greenhouse gas concen-
trations (CO2= 185ppm, CH4= 350ppb, and N2O=
200ppb) estimated from icecores. Ozone and aerosol
forcing are set to pre-industrial estimates for the LGM
and to 1990 estimates for the modern and 4XCO2sim-
ulations.
Individual ensemble members were branched from
different times in the controls to sample a range of ini-
tial conditions. Results for the first 20 yr are averaged
over the three ensemble members for the modern and
LGM freshened cases.
Figure A shows a timeseries of the maximum of the
meridional overturning streamfunction in the North At-
lantic, which is an often-used measure of the THC
strength and is an alternative to the index used in
the main text. The maximum overturning occurs at
about 35◦N at a depth of 850m in the modern con-
trol in CCSM3. The average value of the maximum
overturning for the three controls is approximately dou-
ble the sinking flux across 1022 m depth from 60-65◦N,
the measure of THC strength used in the main text.
The transient behavior of the THC after freshening ap-
pears qualitatively similar in the two measures. The
recovery rate in the modern climate is fastest, at least
in the first 40 years after freshening, and the recovery
rate in the 4XCO2modern climate is intermediate be-
tween the modern and LGM rates. There are subtle
differences in the details of the recovery as measured by
the two THC indices. Such differences in the various
measures of the THC strength during transient changes
have been described previously and are expected [Gent,
2004]. In fact, these subtle differences are evidence that
more than one measure of the THC should be consid-
ered when analyzing North Atlantic climate. However,
none of the results and conclusions in the main text are
influenced substantially by the choice of THC index.
Figure 2 shows the annual mean mixed layer depths
in the controls and their anomalies after freshening.
This figure shows that the locations with annual mean
mixed layer depths over 200 m roughly coincide with
the locations of the heaviest density outcroppings at
the surface in Fig 2. The mixed layer becomes shal-
lower in most places after freshening, but the location
of the maxima do not shift in any remarkable way. The
maximum mixed layer depths remain underneath the
sea ice before and after freshening in the LGM climate
and south of the sea ice edge before and after freshening
in the modern and 4XCO2climates.
0 20 40 60 80 100
0
5
10
15
20
Sinking Rate − Sv
Year Since Freshening
Figure A. Alternative index of the annual mean thermo-
haline circulation shown in Fig 1. Here the index is the
maximum of the Atlantic meridional overturning stream-
function north of 20◦N and below 418m depth. (Solid line
= modern, dashed = LGM, dot-dash = 4XCO2).
Figure 3a shows horizontal (depth integrated) fresh-
water fluxes into the northern North Atlantic freshened
region across ∼55◦N and at the interface of the Arctic
and Barents Seas and the freshwater flux into the sur-
face of the region. These individual components of the
freshwater budget depend on the climate state. How-
ever, the total freshwater dissipation rate (Fig. 3b) is
roughly the same for all three climates, so the total
freshwater dissipation does not provide an explanation
for the large differences in the THC recovery rates. Be-
cause the horizontal transports are integrated vertically,
the important influence of the vertical freshwater flux
(see Fig 3 and the main text) is not apparent. Nonethe-
less the routes of horizontal freshwater dissipation are
worth mentioning. In the first decade after freshening,
Fig. 3a shows that there is anomalous freshwater ex-
port to the north and south in all cases. The anomalous
southward export at ∼55◦N, which is comparable in the
first decade in all three cases, is due to the export of the
freshened, near-surface waters by the Labrador Current.
The anomalous northward export into the Arctic and
Barents Sea is at least three times larger in the first
decade in the modern and 4XCO2cases, which have
Bering Strait open, than in the LGM case. Instead, the
LGM has a considerable reduction in precipitation on
average over the freshened region, which is unseen in the
modern and 4XCO2cases. The reduction in precipita-
tion in the LGM climate occurs over a large area, while
there is increase in the convergence of sea ice with even
greater magnitude but localized over deep-water out-
croppings in the LGM (see Fig 2d). In all three cases
the total freshwater dissipation rate is about an order
of magnitude larger in the first decade after freshening
that in subsequent decades.

BITZ ET AL: LGM, MODERN, AND GREENHOUSE WARMING THC RECOVERY X - 7
Figure 2. Annual mean mixed layer depths in the controls (upper panels) and the anomalies two decades after freshening
(lower panels) with the 15% sea ice concentration contour in the controls (solid black line) and two decades after freshening
(dashed, lower panels only).

X - 8 BITZ ET AL: LGM, MODERN, AND GREENHOUSE WARMING THC RECOVERY
−0.4
−0.2
0
0.2
0.4
Sv
(a)
1 2 3 4 5
−1
−0.5
0
Sv
(b)
Dedade After Freshening
Figure 3. Anomalous freshwater flux into the northern
North Atlantic (a) across ∼55◦N on average (black lines),
into the Arctic Ocean and Barents Sea (gray lines), and
downward into the top surface in the region inbetween
(red lines) and (b) the net flux (black plus red minus gray
curves). (solid line = modern, dashed = LGM, dot-dash
= 4XCO2). The transport is computed on the native
model grid, which is not aligned with true latitudes and
longitudes, so the transport across the southern bound-
ary actually spans true latitudes from 53.2–58.6◦N, de-
pending on the longitude.
Auxiliary Reference
Gent, P. R, and G. Danabasoglu, Heat uptake and
the thermohaline circulation in the Community Climate
System Model, Version 2, J. Clim.,17, 4058–4069, 2004.