The Southern Ocean exhibits more than one
mode of interannual variability. Specifically, a quasistanding wave, zonal wavenumber 3 pattern and a
propagating, zonal wavenumber 2 pattern, have
been identified (see Cai et al., 1999, for a brief
review). The propagating pattern has been named
the Antarctic Circumpolar Wave (ACW), and has
signatures in SST, sea ice extent and atmospheric
and oceanic circulation (White and Petersen, 1996;
Jacobs and Mitchell, 1996). Variability with some
of the observed characteristics has been seen in a
number of coupled model studies (Christoph et al.,
1998; Motoi et al., 1998; Cai et al., 1999). However, the dominant length scale of the variability is
not always as observed (e.g. Cai et al. find ACWlike variability with wavenumber 3), and the mechanisms that drive this variability remain a subject
of debate.
2.3.5.6 Decadal variability
On interannual time scales, the dominant mode of
Pacific SST variability is ENSO. On interdecadal
time scales the centre of action moves to the North
Pacific (see, e.g., Yukimoto, 1999). This variability,
which appears to be a coupled mode involving
wind stress and temperature advection anomalies
in the subtropical gyre, is reproduced in several
models (Latif and Barnett, 1996; Robertson, 1996;
Yukimoto et al., 1996; Knutson and Manabe,
1998; Saravanan, 1998). Decadal modulation of
the interannual ENSO signal is also seen in both
observations and models (Zhang et al., 1997; Lau
and Weng, 1999).
The North Atlantic and Arctic Oscillations
(NAO and AO) dominate the decadal variability
of the Atlantic sector. A number of models produce NAO-like (e.g. Delworth, 1996; Saravanan,
1998) and AO-like (e.g. Broccoli et al., 1998; Fyfe
et al., 1999; Shindell et al., 1999) variability.
It is known that the ocean responds to atmospheric variability on decadal time scales (Dickson
et al., Chapter 7.3; Dickson et al., 1996). An
important question is whether the ocean has a significant role in driving the atmospheric variability
through SST anomalies, producing a fully coupled
mode and opening up the possibility of some level
of climate predictability on decadal time scales.
Diagnosis of CSM model heat fluxes suggests that
over much of the ocean there is a strong local
negative feedback of the surface heat flux on SST
(e.g. Saravanan, 1998). Ocean-only model runs
have been used to show that in the GFDL model
decadal THC variability can be explained as a
largely passive response to surface heat flux variations (Delworth and Greatbatch, 1999). However,
this does not preclude predictable oceanic forcing
of the atmosphere if the ocean dynamics produces
SST anomalies in regions where the atmosphere
is particularly sensitive to SST. The results of
Rodwell et al. (1999) suggest that mid-latitude
SST anomalies are playing some role in driving
NAO variations on decadal time scales. This is a
topic of much current research.
2.3.5.7 Simulation of past climates
Palaeoclimatic evidence suggests that changes in
ocean circulation have played an important role in
past climate change and variability (Broecker,
1997). Coupled models are therefore likely to have
an important part to play in the study of past
climates. Further, coupled models are widely used
to make projections of possible future climate, but
most model evaluation is done by testing the models’ ability to simulate present climate (as in Sections
2.3.5.1–2.3.5.6 above). How do we know that the
models have not simply been ‘tuned’ to give a
good simulation of present climate? A valuable,
orthogonal test of the models is to ask whether,
without ‘retuning’ they can simulate past climate
states.
One test that has been applied to models is to
ask whether, given the history of forcing over the
last century or so from solar variability, volcanic
and anthropogenic aerosols and greenhouse gases,
they can reproduce the observed evolution of
the global mean surface air temperature over
that time. Since the climate (and the models) contains considerable natural variability, one would
not expect the model response to be determined
entirely by the forcing, and ensembles of several
integrations are usually used in these studies. Several models have shown that encouraging agreement can be obtained with the observed record
(Hasselmann et al., 1995; Mitchell et al., 1995;
Haywood et al., 1997; Boer et al., 1999). By
running the models with different combinations
of forcings, insight can be gained into the attribution of observed climate variations to natural or
anthropogenic causes (e.g. Hegerl et al., 1997;
Barnett et al., 1999; Tett et al., 1999).
Moving further back into time, much work
has been done on the simulation of the climatic
SECTION 2 OBSERVATIONS AND MODELS
92
mode of interannual variability. Specifically, a quasistanding wave, zonal wavenumber 3 pattern and a
propagating, zonal wavenumber 2 pattern, have
been identified (see Cai et al., 1999, for a brief
review). The propagating pattern has been named
the Antarctic Circumpolar Wave (ACW), and has
signatures in SST, sea ice extent and atmospheric
and oceanic circulation (White and Petersen, 1996;
Jacobs and Mitchell, 1996). Variability with some
of the observed characteristics has been seen in a
number of coupled model studies (Christoph et al.,
1998; Motoi et al., 1998; Cai et al., 1999). However, the dominant length scale of the variability is
not always as observed (e.g. Cai et al. find ACWlike variability with wavenumber 3), and the mechanisms that drive this variability remain a subject
of debate.
2.3.5.6 Decadal variability
On interannual time scales, the dominant mode of
Pacific SST variability is ENSO. On interdecadal
time scales the centre of action moves to the North
Pacific (see, e.g., Yukimoto, 1999). This variability,
which appears to be a coupled mode involving
wind stress and temperature advection anomalies
in the subtropical gyre, is reproduced in several
models (Latif and Barnett, 1996; Robertson, 1996;
Yukimoto et al., 1996; Knutson and Manabe,
1998; Saravanan, 1998). Decadal modulation of
the interannual ENSO signal is also seen in both
observations and models (Zhang et al., 1997; Lau
and Weng, 1999).
The North Atlantic and Arctic Oscillations
(NAO and AO) dominate the decadal variability
of the Atlantic sector. A number of models produce NAO-like (e.g. Delworth, 1996; Saravanan,
1998) and AO-like (e.g. Broccoli et al., 1998; Fyfe
et al., 1999; Shindell et al., 1999) variability.
It is known that the ocean responds to atmospheric variability on decadal time scales (Dickson
et al., Chapter 7.3; Dickson et al., 1996). An
important question is whether the ocean has a significant role in driving the atmospheric variability
through SST anomalies, producing a fully coupled
mode and opening up the possibility of some level
of climate predictability on decadal time scales.
Diagnosis of CSM model heat fluxes suggests that
over much of the ocean there is a strong local
negative feedback of the surface heat flux on SST
(e.g. Saravanan, 1998). Ocean-only model runs
have been used to show that in the GFDL model
decadal THC variability can be explained as a
largely passive response to surface heat flux variations (Delworth and Greatbatch, 1999). However,
this does not preclude predictable oceanic forcing
of the atmosphere if the ocean dynamics produces
SST anomalies in regions where the atmosphere
is particularly sensitive to SST. The results of
Rodwell et al. (1999) suggest that mid-latitude
SST anomalies are playing some role in driving
NAO variations on decadal time scales. This is a
topic of much current research.
2.3.5.7 Simulation of past climates
Palaeoclimatic evidence suggests that changes in
ocean circulation have played an important role in
past climate change and variability (Broecker,
1997). Coupled models are therefore likely to have
an important part to play in the study of past
climates. Further, coupled models are widely used
to make projections of possible future climate, but
most model evaluation is done by testing the models’ ability to simulate present climate (as in Sections
2.3.5.1–2.3.5.6 above). How do we know that the
models have not simply been ‘tuned’ to give a
good simulation of present climate? A valuable,
orthogonal test of the models is to ask whether,
without ‘retuning’ they can simulate past climate
states.
One test that has been applied to models is to
ask whether, given the history of forcing over the
last century or so from solar variability, volcanic
and anthropogenic aerosols and greenhouse gases,
they can reproduce the observed evolution of
the global mean surface air temperature over
that time. Since the climate (and the models) contains considerable natural variability, one would
not expect the model response to be determined
entirely by the forcing, and ensembles of several
integrations are usually used in these studies. Several models have shown that encouraging agreement can be obtained with the observed record
(Hasselmann et al., 1995; Mitchell et al., 1995;
Haywood et al., 1997; Boer et al., 1999). By
running the models with different combinations
of forcings, insight can be gained into the attribution of observed climate variations to natural or
anthropogenic causes (e.g. Hegerl et al., 1997;
Barnett et al., 1999; Tett et al., 1999).
Moving further back into time, much work
has been done on the simulation of the climatic
SECTION 2 OBSERVATIONS AND MODELS
92