Simulated Two-Stage Recovery of Atlantic Meridional Overturning Circulation During the Last Deglaciation - Abrupt Climate Change: Mechanisms, Patterns, and Impacts - page 89

however, the time scale should not be ignored when addres-
sing the relative contribution of NADW multiple origins to
AMOC change. For a time scale of interdecades or shorter in
control simulation, the conclusions about the relative contri-
bution do not seem consistent; but for a centuries or longer
time scale in water-hosing experiments, the conclusions are
consistent because of the big perturbation and large response
according to current literature.
In this chapter, we use one new method, which is based on
the zonal mean stream function of the Atlantic, to present the
two-stage feature of the AMOC recovery in multiple regions
of origin of NADW formation primarily. However, the two-
stage feature not only depends on the new method but must
also be confirmed in the traditional way, through examina-
tion of the mixed layer depth in Figure 12c and other differ-
ent variables (Figures 2

the southward order of NADW reinitiation from the GIN seas
to the Labrador Sea by Vellinga and Wood [2002] may be
caused by an absence of a shift in sea ice cover during
periods of suppressed AMOC, so the sequence

of reinitiation of convection in multiple regions in the Vel-
linga and Wood work may not be correct. The conclusions of
Renold et al. [2009] are reasonably correct for their experi-
ments with a “ real ” AMOC suppressing process. Beyond
our results, one point should be mentioned: the contrast
reinitiation sequence of NADW is based on two models,
northward sequence by Renold et al. [2009] and our study
from CCSM3 and southward sequence by Vellinga and
Wood [2002] from HadCM3. Despite the dependence of the
experimental scheme we mentioned before, the sequence of
NADW reinitiation may also be dependent on models, but
this needs further validation.
In our DGL-A model run, the time delay between the onset
of stage 1 and stage 2 generally coincides with the recovery
of the AMOC to its intensity at the initial glacial state.
However, according to Renold et al. [2009], under modern
conditions the time delay is shorter, showing stage 2 onset
before the AMOC has made its initial recovery. This may be
an indication that the relationship between the onset time of
each stage and the timing of initial AMOC recovery can have
an important connection to the climate state.

“ arti cially ”

13). The new method is not
only capable of representing the change of NADW formation
in multiple origins, it can quantify the volume of NADW
formation in multiple origins, which cannot be achieved
through the traditional way.
The results of the DGL-A run are consistent with an
idealized water-hosing experiment under the Last Glacial
Maximum state (not shown here). In this idealized water-
hosing experiment, AMOC intensity fully recovered by the
end of integration (1500 years). In this idealized experiment,
stage 2 of AMOC recovery is mainly dominated by the
adjustment process of NADW formation in the GIN seas
with a time scale of about 500 years. The difference between
these two simulations in the recovery of AMOC is that the
AMOC intensity of DGL-A jumped to interglacial levels
during the recovery process, forced by increased GHG con-
centration and orbital insolation [Liu et al., 2009]. This
difference is not critical to our dynamical analysis of the
AMOC recovery process.
In considering the recovery process of the AMOC during
last deglaciation, the AMOC overshoot is important in the
generation of BA warming in TraCE-21000. So the AMOC
overshoot phenomenon may be critical to the understanding
of abrupt climate change. As shown above, the development
of the AMOC overshoot depends on enhanced NADW for-
mation in the GIN seas during stage 2. Another important
aspect of AMOC recovery in DGL-A is the mean state
transition of the AMOC from a glacial to an interglacial
state. This transition depends significantly on the intensified
NADW formation in the GIN seas during BA onset as well.
So the GIN seas are not only a key region for the develop-
ment of an AMOC overshoot but also the key region for
mean state transition of AMOC within glacial/interglacial
cycles.
With our interpretation of the asynchronous reinitiation of
NADW in the Labrador Sea and GIN seas, we speculate that

- 5 and 8 -

Acknowledgments. This research was supported by the Paleocli-
mate Program of NSF, NCAR, DOE, Peking University, Special
public sector research of CMA, China (GYHY200906016), and
Innovation Plan to Graduate Students in the Universities of Jiangsu
Province, China (CX07B_043z) and NUIST (Y602).

REFERENCES

Arzel, O., M. H. England, and W. P. Sijp (2008), Reduced stability
of the Atlantic meridional overturning circulation due to wind
stress feedback during glacial times, J. Clim., 21, 6260 - 6282,

doi:10.1175/2008JCLI2291.1.
Barker, S., P. Diz, M. J. Vautravers, J. Pike, G. Knorr, I. R. Hall, and
W. S. Broecker (2009), Interhemispheric Atlantic seesaw re-
sponse during the last deglaciation, Nature, 457, 1097

- 1102,

doi:10.1038/nature07770.
Bentsen, M., H. Drange, T. Furevik, and T. Zhou (2004), Simulated
variability of the Atlantic meridional overturning circulation,
Clim. Dyn., 22, 701

720, doi:10.1007/s00382-004-0397-x.
Bitz, C. M., J. C. H. Chiang, W. Cheng, and J. J. Barsugli (2007),
Rates of thermohaline recovery from freshwater pulses in mod-
ern, Last Glacial Maximum, and greenhouse warming climates,
Geophys. Res. Lett., 34, L07708, doi:10.1029/2006GL029237.
Broecker, W. S. (1990), Salinity history of the northern Atlantic
during the last deglaciation, Paleoceanography, 5,459

-

- 467,

doi:10.1029/PA005i004p00459.