Abrupt Climate Change Revisited - Abrupt Climate Change: Mechanisms, Patterns, and Impacts - page 4

AMOC as modulated by the Antarctic climate. Furthermore,
Flower et al. hypothesize that LIS melting is linked to the
Antarctic climate, so that it was the AMOC reduction (via the
bipolar seesaw and Antarctic warming) that drove increased
LIS meltwater input to the GOM and not vice versa. This
causality, however, may have been limited as LIS meltwater
input to the GOM continued throughout D-O 8 and Bølling/
Allerød events despite a lack of evidence for AMOC reduc-
tion during these intervals [Rahmstorf, 2002].
Since the D-O cycles were

archives, the climatic expression of the D-O cycles is evident
from the southwest Pacific Ocean to northeastern Siberia.

3.2. Deglaciation Period

One of the most studied periods of the last glacial cycle is
the time between 10 and 25 ka, commonly known as the last
deglaciation. This period contains four very well studied
abrupt climate events: the Younger Dryas (also known as
H0 event, 11.6 - 12.9 ka); Bølling-Allerød (B-A) (14.6 ka);
the so-called “ Mystery Interval ” (14.5 - 17.5 ka); and the last
glacial maximum (LGM) (19 - 26.5 ka). The rationales behind
focusing on the last deglaciation interval include the avail-
ability of paleoclimate records with high temporal resolution
and a robust age control constrained by 14 C - accelerator mass
spectrometry and U-Th dating. It is not surprising that 5 out
of 14 chapters in this volume focus on climate events from
this interval. In addition, we outline several new hypotheses
(relative to the deglaciation events) that have emerged since
the 2009 conference.
Simulating the impact of freshwater discharge into the
North Atlantic during the Heinrich and other meltwater
events has gained a significant momentum after the work of
Stouffer et al. [2006] and Liu et al. [2009]. Numerous mod-
eling studies, where freshwater discharge was artificially
added to the North Atlantic, have shown large climate im-
pacts associated with abrupt AMOC changes. These results
highlight the need to better understand the mechanisms of the
AMOC variability under the past, present, and future climate
conditions. The mechanisms controlling the recovery of the
AMOC are, however, difficult to investigate as they require
long simulations with models that can capture very different
internal variabilities. Cheng et al. [this volume] describe the
AMOC recovery stages from a model simulation of the last
deglaciation built on the work of Liu et al. [2009]. These
authors perform a remarkably long simulation using an
atmosphere-ocean general circulation model [Liu et al.,
2009], which allows for a clear identification of transient
states of the recovery process including the AMOC overshoot
phenomenon [Renold et al., 2010]. This 5000 year simulation
time covers the last deglaciation period using insolation and
fresh water estimated from paleoproxies as forcings. In par-
ticular, the simulation offers an explanation for the B-Awarm
period by a recovery of the AMOC in less than 400 years. An
in-depth analysis of the AMOC recovery suggests that the
two convection sites in the North Atlantic simulated in the
model do not recover at the same time: the first stage of
the recovery occurred in the Labrador Sea and was then
followed by convection recovery in the Greenland-Iceland-
Norwegian (GIN) Seas. The study suggests that reinitiation
of convection in the Labrador Sea is related to the reduction

first discovered in Greenland ice
cores and then later confirmed in deep-sea sediments of the
Atlantic, a large number of modeling efforts were directed
toward understanding the origin and signi

cance of these
paleoclimatic events. Rial and Saha [this volume] simulate
the D-O cycles using a conceptual model, based on simple
stochastic differential equations, termed the sea ice oscillator
(SIO), which borrowed elements from the Saltzman et al.
[1981] concept. Simulation results show that sea ice extent
and mean ocean temperature can be driven by changes in
orbital insolation, which play a dominant role in controlling
atmospheric temperature variability. The SIO model has two
internal parameters controlling the abruptness of temperature
change: the free frequency of the oscillator and the intensity of
positive feedbacks. The model best reproduces the D-O cy-
cles if the free oscillation period is set around 1.5 kyr, as
originally proposed by Bond et al. [1997]. The performance
of the SIO model is tested against two of the best resolved
paleoclimate records: the Greenland ice core

18 O[North
Greenland Ice Core Project members, 2004] and planktonic
and benthic foraminiferal

δ

18 O records of sea surface and
deep-sea temperature from the Portuguese margin (core
MD95-2042 [Shackleton et al., 2000]). Shackleton et al.
[2000] have shown that the

δ

18 O of planktonic foraminifer
exhibits changes similar to those in the Greenland ice core,
whereas the δ

δ

18 O record of benthic foraminifer (water depth
below 2200 m) varies in a manner similar to the Antarctic air
temperature. At any rate, Rial and Saha conclude that the time
integral of the surface ocean proxy history is proportional to
that of the deep ocean temperature, and the latter is shifted by
π /2 with respect to sea ice extent. Surprisingly, a similar

relationship was found between methane-synchronized tem-
perature proxies from Greenland and East Antarctic ice core
records [Blunier and Brook, 2001]. Rial and Saha

s model
output is reproduced using ECBilt-Clio, another well re-
garded model [Goosse et al., 2002].
To summarize up-to-date results on abrupt climate events
of the last glacial cycle, Figures 1 and 2 show records
representative of the global climate during this period. These
records were selected based on the geographic location and
temporal resolution adequate to assess millennial- or finer-
scale climate events. Regardless of the nature of climate

'