An investigation of the El Niño‐Southern Oscillation cycle With statistical models: 1. Predictor field characteristics

We have developed two sets of linear models for predicting equatorial Pacific sea surface temperatures (SSTs) from the Indo‐Pacific trade wind field and the near‐global sea level pressure (SLP) field. The models were constructed using a combination of extended empirical orthogonal functions (EEOFs) and canonical correlation analysis (CCA), a new approach in geophysical modeling. Our results are of interest both as they show the dominant modes of evolution in the SLP and wind fields through the El Niño‐Southern Oscillation cycle and with respect to the problem of predicting equatorial SSTs. This paper deals with the first issue above and describes some statistical composites that typify the development of features in the predictor fields over periods of years. The results of the EEOF analyses clearly show slowly propagating anomalies in both the near‐global SLP and trade wind fields. The CCA analysis, which highlights the co‐evolution of the two fields, suggests a strong coupling between the two and depicts the anomalous features as migrating centers of divergence and convergence that first appear over the eastern Indian Ocean. These features propagate slowly eastward, amplify as they expand into the western Pacific, decline as they cross the central ocean, then reamplify over the eastern Pacific. As reamplification takes place, new opposing anomalies appear in the Indo‐Pacific. Descriptions of the predictor data sets and details of the statistical techniques used may also be found in this paper. The SST data, model validation techniques, and forecast model results are presented in a companion paper (Graham et al., this issue).