133
explained by internal atmospheric chaotic dynamics, with the fraction of
variance explained by 'external' forcing (eg from the ocean). From a practical point of view, if chaotic variations are significant, then the predictability
of the atmosphere will be limited on these timescales.
Simple model estimates (eg James and James, 1989) suggest that internal chaotic processes could be significant even on decadal timescales.
In order to be able to have a more quantitative estimate, one needs to
resort to more comprehensive models. To start to address these questions
I will show results from some decadal timescale integrations made using
the UKMO unified model (D. Rowell, personal communication). These
integrations were made as part of a coordinated study using a number of
GeMs worldwide. Results are based on an ensemble of 6 integrations in
which the model was run for 45 years with observed prescribed SSTs from
1948-1993 using the UKMO GISST (Folland and Rowell, 1995) data set.
The ensemble members differ only in terms of their initial conditions.
Figs 23 and 24 show the percentage of variance of simulated surface
pressure that can be attributed to the time-varying SSTs for seasonal and
decadal averages. This diagnostic was estimated by taking the ratio of the
temporal variance of the ensemble-mean fields, to the total variance of all
surface pressure fields within the ensemble. The temporal variance of the
ensemble mean is assumed to be attributable to the underlying SST variability, hence in regions where the ratio is large, we can assume that the
influence of time-varying SST dominates over internal atmospheric variability. Figs 23 and 24 show results for DJF and JJA respectively. For
each figure the top panel shows results for interannual fluctuations (ie seasonal means), the bottom panel shows results for interdecadal fluctuations
(ie based on 10-year running means of a given season).
Results for seasonal timescale fluctuations are consistent with many previous studies. In particular, over much of the tropics, the total ensemble
variance is dominated by the effects of SST variability, whilst in the (more
chaotic) extratropics the percentage is generally smaller. In the extratropics, the maximum percentage occurs over the north-east Pacific and is
probably associated with the PNA response to EI Nino SST anomalies. In
summer, the percentage of variance in the north Pacific is smaller, consistent with a reduction in tropical-extratropical teleconnectivity associated
with weak potential vorticity gradients.
Interestingly, the percentage of variance of the decadal fluctuations explained by SST variability is, in many areas, smaller than that associated
explained by internal atmospheric chaotic dynamics, with the fraction of
variance explained by 'external' forcing (eg from the ocean). From a practical point of view, if chaotic variations are significant, then the predictability
of the atmosphere will be limited on these timescales.
Simple model estimates (eg James and James, 1989) suggest that internal chaotic processes could be significant even on decadal timescales.
In order to be able to have a more quantitative estimate, one needs to
resort to more comprehensive models. To start to address these questions
I will show results from some decadal timescale integrations made using
the UKMO unified model (D. Rowell, personal communication). These
integrations were made as part of a coordinated study using a number of
GeMs worldwide. Results are based on an ensemble of 6 integrations in
which the model was run for 45 years with observed prescribed SSTs from
1948-1993 using the UKMO GISST (Folland and Rowell, 1995) data set.
The ensemble members differ only in terms of their initial conditions.
Figs 23 and 24 show the percentage of variance of simulated surface
pressure that can be attributed to the time-varying SSTs for seasonal and
decadal averages. This diagnostic was estimated by taking the ratio of the
temporal variance of the ensemble-mean fields, to the total variance of all
surface pressure fields within the ensemble. The temporal variance of the
ensemble mean is assumed to be attributable to the underlying SST variability, hence in regions where the ratio is large, we can assume that the
influence of time-varying SST dominates over internal atmospheric variability. Figs 23 and 24 show results for DJF and JJA respectively. For
each figure the top panel shows results for interannual fluctuations (ie seasonal means), the bottom panel shows results for interdecadal fluctuations
(ie based on 10-year running means of a given season).
Results for seasonal timescale fluctuations are consistent with many previous studies. In particular, over much of the tropics, the total ensemble
variance is dominated by the effects of SST variability, whilst in the (more
chaotic) extratropics the percentage is generally smaller. In the extratropics, the maximum percentage occurs over the north-east Pacific and is
probably associated with the PNA response to EI Nino SST anomalies. In
summer, the percentage of variance in the north Pacific is smaller, consistent with a reduction in tropical-extratropical teleconnectivity associated
with weak potential vorticity gradients.
Interestingly, the percentage of variance of the decadal fluctuations explained by SST variability is, in many areas, smaller than that associated