288
in the lower boundary conditions. However, since the decadal mode is an
inherently coupled phenomenon, the feedback of the atmosphere onto the
ocean is a crucial part of the dynamics of the decadal mode.
Our results are in conflict with several previous studies, since we found
the tropics played a minor role for the generation of the decadal mode.
The studies of Trenberth and Hurrell (1994) and Graham (1994) argue
basically that low-frequency changes in tropical Pacific SST introduce the
signal into the North Pacific through a changed atmospheric circulation.
Jacobs et al. (1994) argues also that decadal variability in the North
Pacific is forced by the tropics, but through the ocean by the propagation
of planetary waves in the aftermath of strong ENSO extremes, such as the
1982/1983 warm event. We did not find much evidence for an active role
of the tropics in the generation of the decadal mode in the North Pacific.
Our view of an independent mid-latitudinal mode is supported by the
findings of Robertson (1996). He investigated a multi-century integration
with another CGCM and found a similar mode in the North Pacific to
that described here. However, the integration Robertson (1996) analyzed
shows virtually no ENSO-type variability in the tropical Pacific so that
some of the proposed tropical forcing mechanisms cannot operate in that
integration.
Another point of controversy which might originate from our study is
the question of whether the atmosphere is sensitive to mid-latitudinal SST
anomalies. Our results suggest indeed that mid-latitudinal SST anomalies
force a significant atmospheric response, as implied by the early study of
Palmer and Sun (1985). Further, we were able to reproduce the dominant atmospheric response pattern found in the coupled integration and
observations by forcing our atmosphere model in a stand-alone mode by
the characteristic North Pacific SST pattern of the decadal mode. We
speculate that changes in the surface baroclinicity and resultant changes
in the transient activity are crucial in establishing the time-mean response.
If correct, this would imply that atmosphere models need to simulate the
eddy activity rather well when they are used in a coupled model to study
decadal variability. This would require quite high resolution for climate
integrations, at least of the order of the T-42 resolution we used in our
coupled integration.
Finally, we would like to discuss critically the predictability of the
decadal modes. Our results suggest that their predictability is probably
less than that of ENSO, for example, if one measures the predictability
in the lower boundary conditions. However, since the decadal mode is an
inherently coupled phenomenon, the feedback of the atmosphere onto the
ocean is a crucial part of the dynamics of the decadal mode.
Our results are in conflict with several previous studies, since we found
the tropics played a minor role for the generation of the decadal mode.
The studies of Trenberth and Hurrell (1994) and Graham (1994) argue
basically that low-frequency changes in tropical Pacific SST introduce the
signal into the North Pacific through a changed atmospheric circulation.
Jacobs et al. (1994) argues also that decadal variability in the North
Pacific is forced by the tropics, but through the ocean by the propagation
of planetary waves in the aftermath of strong ENSO extremes, such as the
1982/1983 warm event. We did not find much evidence for an active role
of the tropics in the generation of the decadal mode in the North Pacific.
Our view of an independent mid-latitudinal mode is supported by the
findings of Robertson (1996). He investigated a multi-century integration
with another CGCM and found a similar mode in the North Pacific to
that described here. However, the integration Robertson (1996) analyzed
shows virtually no ENSO-type variability in the tropical Pacific so that
some of the proposed tropical forcing mechanisms cannot operate in that
integration.
Another point of controversy which might originate from our study is
the question of whether the atmosphere is sensitive to mid-latitudinal SST
anomalies. Our results suggest indeed that mid-latitudinal SST anomalies
force a significant atmospheric response, as implied by the early study of
Palmer and Sun (1985). Further, we were able to reproduce the dominant atmospheric response pattern found in the coupled integration and
observations by forcing our atmosphere model in a stand-alone mode by
the characteristic North Pacific SST pattern of the decadal mode. We
speculate that changes in the surface baroclinicity and resultant changes
in the transient activity are crucial in establishing the time-mean response.
If correct, this would imply that atmosphere models need to simulate the
eddy activity rather well when they are used in a coupled model to study
decadal variability. This would require quite high resolution for climate
integrations, at least of the order of the T-42 resolution we used in our
coupled integration.
Finally, we would like to discuss critically the predictability of the
decadal modes. Our results suggest that their predictability is probably
less than that of ENSO, for example, if one measures the predictability