139
example near Newfoundland, the baroclinic gradient near the climatological region of cyclogenesis is weakened, which reduces storm-track activity.
Through wave-mean flow interaction processes (diagnosed using the Evector), the mean flow over and immediately downstream of this region is
weakened throughout most of the depth of the troposphere. By weakening
the mean flow, the heat loss from ocean to atmosphere is reduced and the
SST anomaly can be maintained. However, as simple mixed- layer ocean
models suggest (eg Daly, 1978) if the surface wind forcing is weakened over
the North Atlantic, then SSTs will warm (cfBjerknes, 1964). Hence there
is the possibility of a weak positive feedback between ocean and atmosphere
over the north Atlantic.
The net effect of this feedback is to redden the spectrum of atmospheric
variability by making atmospheric weather regimes (eg associated with the
North Atlantic Oscillation in the case of Atlantic SSTs, the PNA pattern
in the case of Pacific SSTs) somewhat more stable. However, there is no
evidence from these experiments to suggest that the regime transitions are
themselves any more predictable through coupling to an ocean. Indeed,
if such transitions are ultimately associated with fast baroclinic processes,
it is unlikely that coupling to the oceans would alter the predictability
timescale of these transitions.
An initial study of coupled-model ensemble experimentation appears
to support this view about decadal predictability. Fig 25a (from Griffies
and Bryan, 1996) shows 200 year of linearly detrended anomalous yearlyaveraged thermohaline circulation (THe) index from a run of the GFDL
coupled model. Also shown is the time series after a 10-year low pass filter
has been applied (thick solid line). Fig 25b (also from Griffies and Bryan,
1996) shows the THe index for a 9-member ensemble made by adding
atmospheric perturbations at year 500 of the main run. Although the
THe index clearly has substantial variability on multi-decadal timescales,
its predictability appears to be determined by the overlying chaotic atmospheric variability, which is much shorter than the dominant THe timescale.
5.3 A simple chaotic 'coupled' model paradigm for decadal fluctuations and predictability
To explore these ideas further, I want to put forward another simple extension of the Lorenz model. As before, the basic Lorenz model (4.5) describes
the atmosphere, now it is coupled to a simple ocean model, which here is
example near Newfoundland, the baroclinic gradient near the climatological region of cyclogenesis is weakened, which reduces storm-track activity.
Through wave-mean flow interaction processes (diagnosed using the Evector), the mean flow over and immediately downstream of this region is
weakened throughout most of the depth of the troposphere. By weakening
the mean flow, the heat loss from ocean to atmosphere is reduced and the
SST anomaly can be maintained. However, as simple mixed- layer ocean
models suggest (eg Daly, 1978) if the surface wind forcing is weakened over
the North Atlantic, then SSTs will warm (cfBjerknes, 1964). Hence there
is the possibility of a weak positive feedback between ocean and atmosphere
over the north Atlantic.
The net effect of this feedback is to redden the spectrum of atmospheric
variability by making atmospheric weather regimes (eg associated with the
North Atlantic Oscillation in the case of Atlantic SSTs, the PNA pattern
in the case of Pacific SSTs) somewhat more stable. However, there is no
evidence from these experiments to suggest that the regime transitions are
themselves any more predictable through coupling to an ocean. Indeed,
if such transitions are ultimately associated with fast baroclinic processes,
it is unlikely that coupling to the oceans would alter the predictability
timescale of these transitions.
An initial study of coupled-model ensemble experimentation appears
to support this view about decadal predictability. Fig 25a (from Griffies
and Bryan, 1996) shows 200 year of linearly detrended anomalous yearlyaveraged thermohaline circulation (THe) index from a run of the GFDL
coupled model. Also shown is the time series after a 10-year low pass filter
has been applied (thick solid line). Fig 25b (also from Griffies and Bryan,
1996) shows the THe index for a 9-member ensemble made by adding
atmospheric perturbations at year 500 of the main run. Although the
THe index clearly has substantial variability on multi-decadal timescales,
its predictability appears to be determined by the overlying chaotic atmospheric variability, which is much shorter than the dominant THe timescale.
5.3 A simple chaotic 'coupled' model paradigm for decadal fluctuations and predictability
To explore these ideas further, I want to put forward another simple extension of the Lorenz model. As before, the basic Lorenz model (4.5) describes
the atmosphere, now it is coupled to a simple ocean model, which here is