114
3.5 Weather regimes, singular vectors and sensitivity patterns
The concept of weather regimes is a long standing one (cf Grosswetterlagen; Hess and Brezowsky, 1977), and is based on the notion that the
large-scale flow may evolve around various recurrent configurations. This
notion was made more precise in modelling studies (Reinhold and Pierrehumbert, 1982) who related the onset, maintenance and decay of regimes
to interactions of the large-scale flow with synoptic-scale variability. The
existence of such weather regimes in the real atmosphere has been inferred
through observational studies (eg Hansen and Sutera, 1986, 1995; Mo and
Ghil, 1988, Molteni et al , 1990, Cheng and Wallace, 1993, Kimoto and
Ghil, 1993), though the existence of unambiguous multimodality is still a
matter of debate (Wallace et al, 1991). In many observational studies, (eg
Yang and Reinhold, 1991; Dole and Gordon, 1983 and Toth 1992), it is
suggested that baroclinic instability sets the timescale for the transition
process betweeen regimes. This timescale is much shorter than a typical
residence timescale (on the order of weeks). This two-timescale behaviour
is consistent with the regime structure in the 3-component Lorenz model
(see Fig 19 below).
Fig 11 shows two of the large-anomaly cluster centroids found by Mo
and Ghil (1988) (l1a and 11b) and by Molteni et al (1990) (l1c and 11d).
Despite different clustering algorithms, Figs 11a and c correspond to one
another quite well (as do Fig 11b and d). The regime centroids have significant projection onto opposite phases of the Pacific North American (PNA)
pattern (Wallace and Gutzler, 1981), though also have structure over the
Atlantic and EurAsia (cfsection 6 on climate change). By convention, the
PNA index of the fields in Fig 11a and c is positive, the PNA index of
the fields in Fig 11b and d is negative. For future reference, Molteni et al
(1990) refer to Fig 11c as cluster 2, and Fig 11d as cluster 5.
Regimes can also be found in atmospheric model integrations. Fig 12a
shows the PDF from a 100 consecutive winter sample of a 1200 perpetualwinter integration of a 3-level T21 quasi- geostrophic model (Marshall and
Molteni, 1993; Corti, 1994; Palmer et al, 1994). The PDF is estimated in
a phase-space plane spanned by two of the dominant empirical orthogonal
functions of the model (shown in Fig 12b). During this chosen 'century',
the PDF is bimodal along an axis that corresponds to fluctuations in the
North Atlantic Oscillation.
As shown in Palmer (1988), low-frequency intraseasonal variability is
3.5 Weather regimes, singular vectors and sensitivity patterns
The concept of weather regimes is a long standing one (cf Grosswetterlagen; Hess and Brezowsky, 1977), and is based on the notion that the
large-scale flow may evolve around various recurrent configurations. This
notion was made more precise in modelling studies (Reinhold and Pierrehumbert, 1982) who related the onset, maintenance and decay of regimes
to interactions of the large-scale flow with synoptic-scale variability. The
existence of such weather regimes in the real atmosphere has been inferred
through observational studies (eg Hansen and Sutera, 1986, 1995; Mo and
Ghil, 1988, Molteni et al , 1990, Cheng and Wallace, 1993, Kimoto and
Ghil, 1993), though the existence of unambiguous multimodality is still a
matter of debate (Wallace et al, 1991). In many observational studies, (eg
Yang and Reinhold, 1991; Dole and Gordon, 1983 and Toth 1992), it is
suggested that baroclinic instability sets the timescale for the transition
process betweeen regimes. This timescale is much shorter than a typical
residence timescale (on the order of weeks). This two-timescale behaviour
is consistent with the regime structure in the 3-component Lorenz model
(see Fig 19 below).
Fig 11 shows two of the large-anomaly cluster centroids found by Mo
and Ghil (1988) (l1a and 11b) and by Molteni et al (1990) (l1c and 11d).
Despite different clustering algorithms, Figs 11a and c correspond to one
another quite well (as do Fig 11b and d). The regime centroids have significant projection onto opposite phases of the Pacific North American (PNA)
pattern (Wallace and Gutzler, 1981), though also have structure over the
Atlantic and EurAsia (cfsection 6 on climate change). By convention, the
PNA index of the fields in Fig 11a and c is positive, the PNA index of
the fields in Fig 11b and d is negative. For future reference, Molteni et al
(1990) refer to Fig 11c as cluster 2, and Fig 11d as cluster 5.
Regimes can also be found in atmospheric model integrations. Fig 12a
shows the PDF from a 100 consecutive winter sample of a 1200 perpetualwinter integration of a 3-level T21 quasi- geostrophic model (Marshall and
Molteni, 1993; Corti, 1994; Palmer et al, 1994). The PDF is estimated in
a phase-space plane spanned by two of the dominant empirical orthogonal
functions of the model (shown in Fig 12b). During this chosen 'century',
the PDF is bimodal along an axis that corresponds to fluctuations in the
North Atlantic Oscillation.
As shown in Palmer (1988), low-frequency intraseasonal variability is