contributed by Willem A. Landman

A CCA model is used to predict near-global sea-surface temperature anomalies. Four 3-month mean near-global sea-surface temperatures (Reynolds and Smith, 1994; Smith et al., 1996) are combined and used as predictors. Up to nine subsequent 1-month near-global sea-surface temperatures are the predictands. Pre-orthogonalisation using standard EOF analysis (Barnston 1994) is performed on the predictor and the predictand field because of the large number of highly correlated variables and few observations (less than 50) contained in these fields. The predictor and predictand data sets are first standardized, so that the EOF pre-orthogonalisation is performed using the correlation matrices. The number of EOF modes to be retained in the CCA eigenanalysis is determined such that about 60% of the variance of both the predictand and predictor field are explained. This value of 60% is justified because 70% is the recommended threshold by the Guttman-Kaiser criterion (Jackson 1991), which normally overselects the number of modes, and Jolliffe (1972) suggested a fraction of the number of modes suggested by this criterion. The truncation for the number of CCA modes retained is determined by using the Guttman-Kaiser criterion. Forecasts are variance-adjusted, generally referred to as forecast inflation, by dividing the forecasts by the cross-validation temporal correlation (Ward and Folland 1991) because linear statistical models underestimate extreme El Niño events (Burgers and Stephenson 1999). The predictand fields, which are a combination of several 1-month fields, are separated after the prediction to obtain forecasts for each 1-month period contained in the combined predictand field. The following schematic illustrates the lead-times involved in making forecasts during early June 2 for June 2000 to February 2001:

The predictand fields, which are a combination of several 1-month fields, are separated after the prediction to obtain forecasts for each 1-month period contained in the combined predictand field. After separation of the predicted fields, 3-month means are calculated to produce forecasts for June-July-August, September-October-November and December-January-February.

Barnston, A. G. 1994. Linear statistical short-term climate predictive skill in the Northern Hemisphere, Journal of Climate, 7, 1513-1564.

Burgers, G. and Stephenson, D. B. 1999. The non-normality of El Niño, Geophysical Research Letters, 26, 1027-1030.

Jackson, J. E. 1991. A Users Guide to Principal Components, Wiley, New York, p. 569.

Jolliffe, I. T. 1972. Discarding variables in principal component analysis. I: Artificial data, Applied Statistics, 21, 160-173.

Reynolds, R. W. and Smith, T. M. 1994. Improved global sea surface temperatures analyses using optimum interpolation, Journal of Climate, 7, 929-948.

Smith, T. M., Reynolds, R. W., Livezey, R. E. and Stokes, D. C. 1996. Reconstruction of historical sea surface temperatures using empirical orthogonal functions, Journal of Climate, 9, 1403-1420.

Ward, M. N. and Folland, C. K. 1991. Prediction of seasonal rainfall in the north Nordeste of Brazil using eigenvectors of sea-surface temperature, International Journal of Climatology, 11, 711-743.

Figure caption

Forecasts of sea-surface temperature anomalies using CCA for June to August 2000 (JJA; top panel), September to November 2000 (SON; middle panel) and December 2000 to February 2001 (DJF; bottom panel). Contour interval is 0.1C.