106
Chapter 6: Analysing the Boreal Summer Relationship
When two time-series have serial correlation, then the effective sampIe size
is reduced from the number of time points, and should be considered when
assessing statistical significance (see Section 9.4). In this chapter, crosscorrelations are calculated between series that are filtered into HF and LF.
The HF components have near zero serial correlation (Table 6.1), so reductions in degrees of freedom are trivial and hence ignored, assuming n - 2
degrees of freedom throughout, where n is the number of time points in the
correlation. This is a particularly useful by-product of filtering the data.
If two time-series have lag-one serial correlation even as low as ab out 0.2,
the degrees of freedom can be substantially reduced. So serial correlation
should always be checked for when assessing the statistical significance of
correlations. The LF data have very high serial correlation (Table 6.1), so
cross-correlations between LF series have very few degrees of freedom indeed.
No attempt has been made to assess significancej the maps should be viewed
as purely descriptive.
6.4.2 Patterns in the Marine Atmosphere Associated
with EOF p2 .
In the tropical central and eastern Pacific, the correlations with SST a2(t)
(Figure 6.2) suggest that the atmosphere is responding to warming and cooling of the SST in a way similar to that predicted by the LN model. Warm
events accompany lower sea-Ievel pressure, anomalous atmospheric fiow down
the pressure gradient, defiected by the Coriolis force, and anomalous convergence over the maximum in SST anomalies straddling the equator at
10° S - 10° N. At 10 - 20° N there are strong correlations with anomalous divergence.
Near and to the south of the equator in the western Pacific, SST a2(t)
correlates with high sea-Ievel pressure and near-surface divergence. Some EI
Niiios and La Niiias strongly modulate the gradient of SST into this region
of the western Pacific (Fu et al. , 1986; Ward, 1994; Ward et al. , 1994), and
it is these events that give rise to the strongest variability in the western
Pacific. Correlations with near-surface convergence in the China Sea may
refiect a direct anomalous circulation driven from the Indonesian region to
the south (see Gill, 1982, p. 466). Anomalous meridional overturning over
these longitudes has been found to be an important component of variability
in aversion of the UK Met. Office's GCM (Rowell, 1991; Ward et al., 1994).
Connections between SST p 2 and the marine atmosphere are found in
locations remote from the direct SST forcing in the Pacific. For example,
there are correlations with high sea-Ievel pressure in the North Indian Ocean
and reduced southwesterly monsoon fiow. In the tropical Atlantic, there
are correlations with high er sea-Ievel pressure near West Africa and reduced
southwesterly monsoon fiow near West Africa at 10° N - 20° N.
SST p2 is also related to some atmospheric variability in the extratropics
though to prove that the associations are significantly more numerous and
Chapter 6: Analysing the Boreal Summer Relationship
When two time-series have serial correlation, then the effective sampIe size
is reduced from the number of time points, and should be considered when
assessing statistical significance (see Section 9.4). In this chapter, crosscorrelations are calculated between series that are filtered into HF and LF.
The HF components have near zero serial correlation (Table 6.1), so reductions in degrees of freedom are trivial and hence ignored, assuming n - 2
degrees of freedom throughout, where n is the number of time points in the
correlation. This is a particularly useful by-product of filtering the data.
If two time-series have lag-one serial correlation even as low as ab out 0.2,
the degrees of freedom can be substantially reduced. So serial correlation
should always be checked for when assessing the statistical significance of
correlations. The LF data have very high serial correlation (Table 6.1), so
cross-correlations between LF series have very few degrees of freedom indeed.
No attempt has been made to assess significancej the maps should be viewed
as purely descriptive.
6.4.2 Patterns in the Marine Atmosphere Associated
with EOF p2 .
In the tropical central and eastern Pacific, the correlations with SST a2(t)
(Figure 6.2) suggest that the atmosphere is responding to warming and cooling of the SST in a way similar to that predicted by the LN model. Warm
events accompany lower sea-Ievel pressure, anomalous atmospheric fiow down
the pressure gradient, defiected by the Coriolis force, and anomalous convergence over the maximum in SST anomalies straddling the equator at
10° S - 10° N. At 10 - 20° N there are strong correlations with anomalous divergence.
Near and to the south of the equator in the western Pacific, SST a2(t)
correlates with high sea-Ievel pressure and near-surface divergence. Some EI
Niiios and La Niiias strongly modulate the gradient of SST into this region
of the western Pacific (Fu et al. , 1986; Ward, 1994; Ward et al. , 1994), and
it is these events that give rise to the strongest variability in the western
Pacific. Correlations with near-surface convergence in the China Sea may
refiect a direct anomalous circulation driven from the Indonesian region to
the south (see Gill, 1982, p. 466). Anomalous meridional overturning over
these longitudes has been found to be an important component of variability
in aversion of the UK Met. Office's GCM (Rowell, 1991; Ward et al., 1994).
Connections between SST p 2 and the marine atmosphere are found in
locations remote from the direct SST forcing in the Pacific. For example,
there are correlations with high sea-Ievel pressure in the North Indian Ocean
and reduced southwesterly monsoon fiow. In the tropical Atlantic, there
are correlations with high er sea-Ievel pressure near West Africa and reduced
southwesterly monsoon fiow near West Africa at 10° N - 20° N.
SST p2 is also related to some atmospheric variability in the extratropics
though to prove that the associations are significantly more numerous and