The 1500 Year Quasiperiodicity During the Holocene - Abrupt Climate Change: Mechanisms, Patterns, and Impacts - page 160

to the low-frequency (>550 years) solar forcing represented
by the long-term reconstruction of sunspot number [Solanki
et al., 2004].
The model experiments by Schulz et al. [2007] show that
the AMOC has two states corresponding to a strong and a
weak return

earlier Holocene followed by an abrupt change to well-mixed
flow oscillating with the about 1500 year period after about
8.4 ka B.P.
Here we investigate the 1500 year variability during the
Holocene using paleodata at three locations of the AMOC
and suggest a simple conceptual model simulating a possible
cause of this oscillation. The model includes the interaction
of solar and ocean centennial variabilities as well as the
threshold response of the AMOC to forcing.
The standard spectral (Fourier and integrated wavelet)
analysis sums spectral features over the whole time period
and assumes that underlying process is stationary, which is
not the case as indicated straight away by inspection of
the data. To analyze the nonstationary time series of the
paleodata, we apply two methods of analysis to the inter-
polated data. The

flow (Figure 1). These experiments take into
account that the surface AMOC flow in the North Atlantic

consists of the current into the Nordic Seas passing between
the subpolar and subtropical gyres from which it draws
water. When a continuous freshwater input to the Labrador
Sea exceeds a small threshold, the deep convection is sup-
pressed leading to a reduced circulation (the weak state) in
the subpolar gyre. The deepwater formation in the Nordic
Seas, which is always active, provides a negative feedback
returning the system to the strong state. The duration be-
tween the transitions from one state to another is found to be
random with a mean value close to 1500 years but with a
very large standard deviation [Schulz et al. 2007]. Follow-up
modeling [Jongma et al., 2007] indicates that the transitions
are susceptible to a weak periodic solar irradiance forcing via
a noise-assisted stochastic resonance. The reconstruction of
temperature and salinity of the in

first is the wavelet method [Torrence and
Compo, 1998] previously used in studies of a selected set
of North Atlantic paleodata by Debret et al. [2007]. The
wavelet method tracks the nonstationary features well but
is not adaptive to data since it is based on the use of given
basic functions (the Morlet functions in this case). To
complement the wavelet analysis, we apply the empirical
mode decomposition (EMD) method, which has no pre-
scriptions and is data adaptive [Huang and Wu, 2009]. The
EMD represents the data as a sum of a small number of

ow to the Nordic Seas
[Thornalley et al., 2009] provides some support for this
mechanism of the Labrador Sea-Nordic Seas interaction. The
reconstruction shows a strati ed upper ocean flow during the

Figure 1. (a) A map of the ocean circulation in the northeast Atlantic. Abbreviations are SPG, subpolar gyre; STG,
subtropical gyre; and EGC, East Greenland Current. Adopted from Thornalley et al. [2009]. Reprinted by permission from
Macmillan Publishers Ltd: Nature, copyright 2009. (b) Time series of the AMOC poleward of 30°N below 500 m depth
generated for 7.7 mSv freshwater perturbation. (c) Conceptual two-state (hysteresis loop) model. Figures 1b and 1c are
adopted from Schulz et al. [2007].