172 Mojib Latif, Axei Timmermann, Anselm Grotzner, Christian Eckert, Reinhard Voss
destroys the predictability at re1atively long time scales. Another factor limiting the
predictability of the two modes discussed here is the width of the spectral peaks,
implying a relatively strong damping ofthe modes.
We have applied our coupled ocean-atmosphere circulation model (which was
also used to study the dynamics ofthe quasi-decadal and interdecadal variabilities;
see sections 9.3 and 9.4) to assess systematically the predictability of climate variations in the North Atlantic region at decadal time scales (further details can be
found in Grotzner et al. (1999)). Classical predictability experiments were conducted in ensemble mode. We have chosen four states from the control integration,
from which the coupled model was restarted. While the oceanic initial conditions
remained unchanged, the atmospheric conditions were perturbed. Each predictability ensemble comprises of ten individual members, i.e. for each ofthe four oceanic
states ten experiments were performed with different atmospheric initial conditions. Each predictability experiment has a duration oftwenty years (the duration in
the last forecast ensemble amounts to only ten years). The divergence ofthe trajectories within an ensemble provides an indication of the predictability at decadal
time scales. aur experimental setup yields an upper limit of decadal predictability,
since we assume a perfect knowledge of the oceanic initial conditions, an assumption which will be never fulfilled in a real forecast situation.
We present in Figs. 9.18, 9.19, and 9.20 the results from our forecast ensembles.
We present the results in terms of the leading EOFs of the anomalous meridional
overtuming, SST, and 500 hPa heights. The individual forecast trajectories show
relatively little spread, when the meridional overtuming is considered (Fig. 9.18),
which is visualized by the ensemble means and normalized ensemble variances.
The variance was norrnalized using the variance computed from the control integration. When the normalized variance reaches unity, the variations between the
individual forecast members are as large as typical variations in the control run,
which defines a predictability limit. The North Atlantic meridional overtuming is
predictable for about one to twO decades, depending on the forecast ensemble considered. The anomalous North Atlantic SST, however, appears not to be predictable
beyond time lags of a few years (Fig. 9.19), which is also true for the leading EOF
of the anomalous Northern Hemisphere 500 hPa heights (Fig. 9.20). This latter
result, however, depends somewhat on the forecast ensemble considered: when
predictability experiments are started from states characterized by strong SST
anomalies, the predictability of anomalous SST and 500 hPa heights seems to be
enhanced.
We anticipated these results from the spectral analyses shown in Fig. 9.10. Thus,
our predictability experiments highlight the ocean's role in generating the interdecadal variability. Although the coupling between the ocean and the atmosphere
seems also to play a key role in the generation of the quasi-decadal and interdecadal modes, the coupling appears to be too weak (or equivalently the damping too
strong) to influence the predictability of surf ace quantities at decadal time scales in
the presence of the internal high-frequency noise. The results of our predictability
experiments are consistent with those shown by Griffies and Bryan (1997), who
destroys the predictability at re1atively long time scales. Another factor limiting the
predictability of the two modes discussed here is the width of the spectral peaks,
implying a relatively strong damping ofthe modes.
We have applied our coupled ocean-atmosphere circulation model (which was
also used to study the dynamics ofthe quasi-decadal and interdecadal variabilities;
see sections 9.3 and 9.4) to assess systematically the predictability of climate variations in the North Atlantic region at decadal time scales (further details can be
found in Grotzner et al. (1999)). Classical predictability experiments were conducted in ensemble mode. We have chosen four states from the control integration,
from which the coupled model was restarted. While the oceanic initial conditions
remained unchanged, the atmospheric conditions were perturbed. Each predictability ensemble comprises of ten individual members, i.e. for each ofthe four oceanic
states ten experiments were performed with different atmospheric initial conditions. Each predictability experiment has a duration oftwenty years (the duration in
the last forecast ensemble amounts to only ten years). The divergence ofthe trajectories within an ensemble provides an indication of the predictability at decadal
time scales. aur experimental setup yields an upper limit of decadal predictability,
since we assume a perfect knowledge of the oceanic initial conditions, an assumption which will be never fulfilled in a real forecast situation.
We present in Figs. 9.18, 9.19, and 9.20 the results from our forecast ensembles.
We present the results in terms of the leading EOFs of the anomalous meridional
overtuming, SST, and 500 hPa heights. The individual forecast trajectories show
relatively little spread, when the meridional overtuming is considered (Fig. 9.18),
which is visualized by the ensemble means and normalized ensemble variances.
The variance was norrnalized using the variance computed from the control integration. When the normalized variance reaches unity, the variations between the
individual forecast members are as large as typical variations in the control run,
which defines a predictability limit. The North Atlantic meridional overtuming is
predictable for about one to twO decades, depending on the forecast ensemble considered. The anomalous North Atlantic SST, however, appears not to be predictable
beyond time lags of a few years (Fig. 9.19), which is also true for the leading EOF
of the anomalous Northern Hemisphere 500 hPa heights (Fig. 9.20). This latter
result, however, depends somewhat on the forecast ensemble considered: when
predictability experiments are started from states characterized by strong SST
anomalies, the predictability of anomalous SST and 500 hPa heights seems to be
enhanced.
We anticipated these results from the spectral analyses shown in Fig. 9.10. Thus,
our predictability experiments highlight the ocean's role in generating the interdecadal variability. Although the coupling between the ocean and the atmosphere
seems also to play a key role in the generation of the quasi-decadal and interdecadal modes, the coupling appears to be too weak (or equivalently the damping too
strong) to influence the predictability of surf ace quantities at decadal time scales in
the presence of the internal high-frequency noise. The results of our predictability
experiments are consistent with those shown by Griffies and Bryan (1997), who