46
pattern be featureless, apart from the inevitable bull's eye surrounding the
reference gridpoint? This question has been addressed by Wallace and
Gutzler (1981) and, in a somewhat different context, by Kushnir and Wallace (1987), who showed that anisotropic patterns with multiple centers of
action far removed from the reference gridpoint are, in fact, observed on
virtually all such maps. The zonal and meridional scales of the features
on different maps tend to be similar, though the intensity and the position of the features relative to the reference grid point varies from map to
map. Hence, the interannual variability of the wintertime circulation is
characterized by a continuum of anisotropic (zonally elongated) structures
qualitatively similar to the ones that have just been examined. But the
pattern in Fig. 9 stands out clearly above this background continuum:
the negative correlations between its primary and remote "centers of action" are significantly stronger than their counterparts in maps for typical
gridpoints.
Just how many teleconnection patterns stand out above the background
continuum depends upon one's criterion for "stand out". Barnston and
Livezey (1987) assigned triacronymic labels to no fewer than eight patterns, Wallace and Gutzler (1981) labeled five, whereas Kushnir and Wallace (1987) deemed only two patterns worthy of a label. All three studies
concur that the PNA pattern and Walker and Bliss's (1932) so called North
Atlantic Oscillation (NAO), documented in detail by van Loon and Rogers
(1978) clearly qualify as teleconnection patterns. The NAO, shown in Fig.
10, dominates the leading EOF of the sea-level pressure field (Kutzbach
1970, Rogers 1981). There is a hint of it in both the leading EOF's of the
500-mb height field (Fig. 8), but in order to see it clearly it is necessary
to rotate (Le., to form linear combinations of) the EOFs in a manner that
tends to simplify their spatial structure (HoreI1981; Barnston and Livezey
1987; Kushnir and Wallace 1987).
In contrast to EOF patterns, which may assume a full range of amplitudes of either polarity, the patterns identified by cluster analysis techniques refer to anomalies of a specified amplitude and sign. Figure 11
shows total fields and anomaly fields for the three dominant clusters of the
10-day lowpass filtered 500-mb height field, as determined by Cheng and
Wallace (1993) on the basis of a hierarchical clustering algorithm. Kimoto
(1987, 1989) and Kimoto and Ghil (1993) obtained similar patterns using
a method of analysis designed to identify local maxima in the probability
density function. The "Alaska" cluster resembles the negative polarity of
pattern be featureless, apart from the inevitable bull's eye surrounding the
reference gridpoint? This question has been addressed by Wallace and
Gutzler (1981) and, in a somewhat different context, by Kushnir and Wallace (1987), who showed that anisotropic patterns with multiple centers of
action far removed from the reference gridpoint are, in fact, observed on
virtually all such maps. The zonal and meridional scales of the features
on different maps tend to be similar, though the intensity and the position of the features relative to the reference grid point varies from map to
map. Hence, the interannual variability of the wintertime circulation is
characterized by a continuum of anisotropic (zonally elongated) structures
qualitatively similar to the ones that have just been examined. But the
pattern in Fig. 9 stands out clearly above this background continuum:
the negative correlations between its primary and remote "centers of action" are significantly stronger than their counterparts in maps for typical
gridpoints.
Just how many teleconnection patterns stand out above the background
continuum depends upon one's criterion for "stand out". Barnston and
Livezey (1987) assigned triacronymic labels to no fewer than eight patterns, Wallace and Gutzler (1981) labeled five, whereas Kushnir and Wallace (1987) deemed only two patterns worthy of a label. All three studies
concur that the PNA pattern and Walker and Bliss's (1932) so called North
Atlantic Oscillation (NAO), documented in detail by van Loon and Rogers
(1978) clearly qualify as teleconnection patterns. The NAO, shown in Fig.
10, dominates the leading EOF of the sea-level pressure field (Kutzbach
1970, Rogers 1981). There is a hint of it in both the leading EOF's of the
500-mb height field (Fig. 8), but in order to see it clearly it is necessary
to rotate (Le., to form linear combinations of) the EOFs in a manner that
tends to simplify their spatial structure (HoreI1981; Barnston and Livezey
1987; Kushnir and Wallace 1987).
In contrast to EOF patterns, which may assume a full range of amplitudes of either polarity, the patterns identified by cluster analysis techniques refer to anomalies of a specified amplitude and sign. Figure 11
shows total fields and anomaly fields for the three dominant clusters of the
10-day lowpass filtered 500-mb height field, as determined by Cheng and
Wallace (1993) on the basis of a hierarchical clustering algorithm. Kimoto
(1987, 1989) and Kimoto and Ghil (1993) obtained similar patterns using
a method of analysis designed to identify local maxima in the probability
density function. The "Alaska" cluster resembles the negative polarity of