176 Mojib Latif, AxeI Timmermann, Anseim Grotzner, Christian Eckert, Reinhard Voss
9.6 Discussion
We have shown that the North Atlantic climate variability at decadal time scales
is largely consistent with the picture that the noise immanent to the coupled oceanatmosphere system drives its low-frequency variability. This stochastic concept is
supported by observations, simulations and predictability experiments with a
sophisticated coupled ocean-atmosphere general circulation model. Large-scale
ocean-atmosphere interactions play an important role in the generation of the interdecadal variability, and eigenmodes of the coupled ocean-atmosphere may exist in
the North Atlantic region. It is likely, however, that these eigenmodes are strongly
damped.
In this article the stochastic climate model scenario served as a paradigm for the
description of climate variations. Many observations and model ing results agree
with its predictions. There are, however, alternative paradigms which as weU capture essential features of climate variability. Imagine, for example, the NAO. One
may hypothesize that the dynamics of this climate fluctuation may be represented
by a low-dimensional non-linear system.lts attractor in phase space may consist of
regions where the system stays for a relatively long time. These are the minima and
maxima of the observed time series. Transitions between states occur quasi-regularly (see Lorenz (1963), Palmer (1993), and references therein for further details).
The observational records do not permit a distinction between these two paradigms, since they are too short. In the near future long simulations with coupled
ocean atmosphere GCMs could allow such inferences.
Our studies rai se some important scientific questions that need to be addressed in
more detail in the future. One ofthe most important challenges is to understand the
atmospheric response to mid-Iatitudinal SST anomalies. This response is likely to
be highly non-linear, since it involves changes in the statistics of the transient
eddies (e. g. Palmer and Sun (1985)) which feed back onto the mean flow. We have
reasons to believe that the atmosphere is indeed sensitive to mid-Iatitudinal SST
anomalies, as argued by Latif and Bamett (1994). The spectral peaks in the model's
NAO, the pan-oceanic connections between the North Atlantic and the North
Pacific described by Timmermann et al. (1998), and additional atmospheric
response experiments with prescribed SST anomalies (not shown), aU ofthese indicate that the atmosphere feels the slowly changing North Atlantic SSTs. However,
the internal atmospheric variability is relatively strong, and it remains unclear as to
whether the atmosphere is affected by the slowly varying boundary conditions to a
degree that enables useful predictions at decadal time scales.
The oceanic processes that lead to the interdecadal variations in the North Atlantic SST need also to be analysed in more detail. We need to quantify the relative
contributions of mixed layer processes and ocean dynamics to the SST changes.
This can be achieved, for instance, by conducting an experiment, in which the
atmosphere is coupled to a fixed-depth mixed layer (slab) model. Such an experiment is underway, and its outcome will be discussed in a forthcoming paper. A
series of experiments was performed by Manabe et al. (1996), who compare three
9.6 Discussion
We have shown that the North Atlantic climate variability at decadal time scales
is largely consistent with the picture that the noise immanent to the coupled oceanatmosphere system drives its low-frequency variability. This stochastic concept is
supported by observations, simulations and predictability experiments with a
sophisticated coupled ocean-atmosphere general circulation model. Large-scale
ocean-atmosphere interactions play an important role in the generation of the interdecadal variability, and eigenmodes of the coupled ocean-atmosphere may exist in
the North Atlantic region. It is likely, however, that these eigenmodes are strongly
damped.
In this article the stochastic climate model scenario served as a paradigm for the
description of climate variations. Many observations and model ing results agree
with its predictions. There are, however, alternative paradigms which as weU capture essential features of climate variability. Imagine, for example, the NAO. One
may hypothesize that the dynamics of this climate fluctuation may be represented
by a low-dimensional non-linear system.lts attractor in phase space may consist of
regions where the system stays for a relatively long time. These are the minima and
maxima of the observed time series. Transitions between states occur quasi-regularly (see Lorenz (1963), Palmer (1993), and references therein for further details).
The observational records do not permit a distinction between these two paradigms, since they are too short. In the near future long simulations with coupled
ocean atmosphere GCMs could allow such inferences.
Our studies rai se some important scientific questions that need to be addressed in
more detail in the future. One ofthe most important challenges is to understand the
atmospheric response to mid-Iatitudinal SST anomalies. This response is likely to
be highly non-linear, since it involves changes in the statistics of the transient
eddies (e. g. Palmer and Sun (1985)) which feed back onto the mean flow. We have
reasons to believe that the atmosphere is indeed sensitive to mid-Iatitudinal SST
anomalies, as argued by Latif and Bamett (1994). The spectral peaks in the model's
NAO, the pan-oceanic connections between the North Atlantic and the North
Pacific described by Timmermann et al. (1998), and additional atmospheric
response experiments with prescribed SST anomalies (not shown), aU ofthese indicate that the atmosphere feels the slowly changing North Atlantic SSTs. However,
the internal atmospheric variability is relatively strong, and it remains unclear as to
whether the atmosphere is affected by the slowly varying boundary conditions to a
degree that enables useful predictions at decadal time scales.
The oceanic processes that lead to the interdecadal variations in the North Atlantic SST need also to be analysed in more detail. We need to quantify the relative
contributions of mixed layer processes and ocean dynamics to the SST changes.
This can be achieved, for instance, by conducting an experiment, in which the
atmosphere is coupled to a fixed-depth mixed layer (slab) model. Such an experiment is underway, and its outcome will be discussed in a forthcoming paper. A
series of experiments was performed by Manabe et al. (1996), who compare three