136
prise to some people, given the debate about the role of the thermohaline
circulation on atmospheric decadal variability. In summer, the percentage
of SST-explained variance of decadal fluctuations is somewhat larger in the
extratropics. Of course, one has to bear in mind that these results have
been obtained by integrating with SSTs over a specific 45 year period.
5.2 The role of the oceans on decadal predictability
Although the results in Figs 23 and 24 suggest that internal atmospheric
dynamics can explain a considerable amount of the observed decadal variability (especially in winter), there is modelling evidence that ocean dynamics might enhance the amplitude of decadal fluctuations. For example,
Manabe and Stouffer (1996), have studied the geographical distribution
of the standard deviation of 25-year mean surface air temperature from
1000 year integrations of three different models: a coupled GeM, an atmospheric GeM coupled to a ocean mixed layer model, and an atmospheric
GeM with fixed SSTs. Results show that over continents, the standard
deviations in all three runs are broadly comparable. Over much of the
oceanic regions, the coupled and mixed layer models produce comparable
standard deviations which in turn are larger than the fixed SST run. Over
specific regions such as the Denmark Strait and in some regions over the
circumpolar ocean of the Southern Hemisphere, the standard deviation of
the coupled model is larger than the mixed layer model.
There are in fact reasons to suspect that the influence of the oceans on
the atmosphere may be somewhat larger than suggested in Manabe and
Stouffer (1996) analyses. In particular, if the feedback from the transient
eddies onto the mean flow is an essential component in accounting for the
impact of mid-latitude SSTs on the atmospheric flow (see below), then
this impact might be under-represented in a model in which the transient
eddy covariances were weak. It is known that low-resolution GeMs, and
the R15 GFDL GeM in particular (Held and Phillipps, 1993) does suffer
from excessively weak eddy momentum fluxes. Secondly, if eddy-meanflow interaction is important, then if observed SST anomalies are added
to a model in which the storm track position has systematic error, the
potential impact of midlatitude SST anomalies may be underestimated. In
fact significant interdecadal midlatitude air-sea interaction has been found
in a coupled GeM integration by Latif and Barnett (1994). In this case,
the atmosphere model was integrated at T42 resolution, and, moreover,
prise to some people, given the debate about the role of the thermohaline
circulation on atmospheric decadal variability. In summer, the percentage
of SST-explained variance of decadal fluctuations is somewhat larger in the
extratropics. Of course, one has to bear in mind that these results have
been obtained by integrating with SSTs over a specific 45 year period.
5.2 The role of the oceans on decadal predictability
Although the results in Figs 23 and 24 suggest that internal atmospheric
dynamics can explain a considerable amount of the observed decadal variability (especially in winter), there is modelling evidence that ocean dynamics might enhance the amplitude of decadal fluctuations. For example,
Manabe and Stouffer (1996), have studied the geographical distribution
of the standard deviation of 25-year mean surface air temperature from
1000 year integrations of three different models: a coupled GeM, an atmospheric GeM coupled to a ocean mixed layer model, and an atmospheric
GeM with fixed SSTs. Results show that over continents, the standard
deviations in all three runs are broadly comparable. Over much of the
oceanic regions, the coupled and mixed layer models produce comparable
standard deviations which in turn are larger than the fixed SST run. Over
specific regions such as the Denmark Strait and in some regions over the
circumpolar ocean of the Southern Hemisphere, the standard deviation of
the coupled model is larger than the mixed layer model.
There are in fact reasons to suspect that the influence of the oceans on
the atmosphere may be somewhat larger than suggested in Manabe and
Stouffer (1996) analyses. In particular, if the feedback from the transient
eddies onto the mean flow is an essential component in accounting for the
impact of mid-latitude SSTs on the atmospheric flow (see below), then
this impact might be under-represented in a model in which the transient
eddy covariances were weak. It is known that low-resolution GeMs, and
the R15 GFDL GeM in particular (Held and Phillipps, 1993) does suffer
from excessively weak eddy momentum fluxes. Secondly, if eddy-meanflow interaction is important, then if observed SST anomalies are added
to a model in which the storm track position has systematic error, the
potential impact of midlatitude SST anomalies may be underestimated. In
fact significant interdecadal midlatitude air-sea interaction has been found
in a coupled GeM integration by Latif and Barnett (1994). In this case,
the atmosphere model was integrated at T42 resolution, and, moreover,