The EOF ( Empirical Orthogonal Functions ) analysis is used to identify
the dominant spatial and temporal patterns present in the OA anomaly. It
is also known as Principal Component Analysis or Eigenvector Analysis.
It has been applied initially to time series of meteorological data . (
Lorenz,1956 ; Kutzbach, 1967 )
In this report, the EOF method is applied to analyze the interannual
anomalies of the Sea Surface Temperature ( SST ), the depth of the 20 Degrees
isotherm ( D20 ) and the Dynamic Height 0/400 ( DHT ).
The main spatial patterns are illustrated in the first
3 spatial EOF's. (eigenvectors). The temporal EOF's (coefficients
) show the time variation of the spatial patterns. The spatial
and the corresponding temporal EOF's are illustrated in Figures
28 to 33.
NOTE :
In order to calculate the covariance matrix, the missing anomaly values
were replaced with zeros.
For line PX-2, due to the large number of missing values obtained when
calculating the dynamic height anomaly (
Fig. 27 ) , the first eigenvector (
Fig. 32.1 ) appears biased towards zero, at the western end of the
line. For this reason, the missing temperature and salinitiy values, at
a particular depth, grid point and point in time, were replaced with the
annual mean at that depth and grid point. With the new temperature and
salinity sets, the dynamic height and its anomaly, as well as the EOF of
the anomaly, were recalculated. The eigenvectors and coefficients thus
obtained, are illustrated in Figures
34 and 35.