Section 6.4: Characteristic Features
101
6.3.5 Rotation of EOFs
The patterns of EOFs p(k) with indices k > 3 were more noisy and less global
in scale. Rotation (cf. 13.2) can often yield dearer and more stable patterns,
especially when the patterns of interest are on ascale that is smaller than
the analysis domain. There are various criteria used for rotation, the most
common being VARIMAX (13.12). Here, EOFs pi_p3 were left unrotated
because they are readily interpretable as global scale patterns with physical
significance. So EOFs p4_p13 were VARIMAX rotated. Inspection of the
eigenvalues revealed a sharp drop in variance explained between EOFs p13
and pB, and this guided the choice of exduding EOFs pI< with k ~ 14
(see O'Lenic and Livezey, 1988, for discussion of which EOFs to retain in
rotation).
The first rotated EOF (pA) was a pattern mainly in the extratropical North
Atlantic. It seems to mainly reflect a response to the North Atlantic Oscillation atmospheric variation (not shown here). It is not discussed further in
this chapter.
PA (Figure 6.1c) has strong positive weights in the southern and eastern
tropical Atlantic. There is a slight opposition of weight in the northern
tropical Atlantic. The time coefficients cxf'(t) (Figure 6.1c) mainly measure
the strength of warm events and cold events associated with the Benguela
current in the South Atlantic.
6.4 Characteristic Features in the Marine Atmosphere Associated with the SST Patterns p 2 , p3 and Pk in JAS
6.4.1 Data and Methods
Ship observations of sea-Ievel pressure (SLP) and near-surface zonal (u) and
meridional (v) wind are used to identify the atmospheric variability associated
with each of the three characteristic SST patterns p2, p3 and PA shown in
Figure 6.1. The ship observations of the atmosphere, like the SST, are now
held in computerised datasets. For the analysis here, data have been taken
from the Comprehensive Ocean-Atmosphere Dataset (COADS, Woodruff et
al. , 1987) and processed as described in Ward (1992, 1994). The data are
formed into seasonal anomalies on the 10 0 lat x 10 0 long scale for each JAS
season 1949-88. Like the SST data, it was necessary to apply a correction
to the wind data to remove an upward trend in wind speed that was caused
by changes in the way winds have been measured aboard ships (Cardone et
al. , 1990; correction method described in Ward, 1992, 1994). A dataset of
10 0 lat x 10 0 long near-surface wind divergence has been calculated using the
101
6.3.5 Rotation of EOFs
The patterns of EOFs p(k) with indices k > 3 were more noisy and less global
in scale. Rotation (cf. 13.2) can often yield dearer and more stable patterns,
especially when the patterns of interest are on ascale that is smaller than
the analysis domain. There are various criteria used for rotation, the most
common being VARIMAX (13.12). Here, EOFs pi_p3 were left unrotated
because they are readily interpretable as global scale patterns with physical
significance. So EOFs p4_p13 were VARIMAX rotated. Inspection of the
eigenvalues revealed a sharp drop in variance explained between EOFs p13
and pB, and this guided the choice of exduding EOFs pI< with k ~ 14
(see O'Lenic and Livezey, 1988, for discussion of which EOFs to retain in
rotation).
The first rotated EOF (pA) was a pattern mainly in the extratropical North
Atlantic. It seems to mainly reflect a response to the North Atlantic Oscillation atmospheric variation (not shown here). It is not discussed further in
this chapter.
PA (Figure 6.1c) has strong positive weights in the southern and eastern
tropical Atlantic. There is a slight opposition of weight in the northern
tropical Atlantic. The time coefficients cxf'(t) (Figure 6.1c) mainly measure
the strength of warm events and cold events associated with the Benguela
current in the South Atlantic.
6.4 Characteristic Features in the Marine Atmosphere Associated with the SST Patterns p 2 , p3 and Pk in JAS
6.4.1 Data and Methods
Ship observations of sea-Ievel pressure (SLP) and near-surface zonal (u) and
meridional (v) wind are used to identify the atmospheric variability associated
with each of the three characteristic SST patterns p2, p3 and PA shown in
Figure 6.1. The ship observations of the atmosphere, like the SST, are now
held in computerised datasets. For the analysis here, data have been taken
from the Comprehensive Ocean-Atmosphere Dataset (COADS, Woodruff et
al. , 1987) and processed as described in Ward (1992, 1994). The data are
formed into seasonal anomalies on the 10 0 lat x 10 0 long scale for each JAS
season 1949-88. Like the SST data, it was necessary to apply a correction
to the wind data to remove an upward trend in wind speed that was caused
by changes in the way winds have been measured aboard ships (Cardone et
al. , 1990; correction method described in Ward, 1992, 1994). A dataset of
10 0 lat x 10 0 long near-surface wind divergence has been calculated using the