Conditional Maximum Covariance Analysis and Its Application to the Tropical Indian Ocean SST and Surface Wind Stress Anomalies

Abstract

This study introduces the conditional maximum covariance analysis (CMCA). The normal maximum covariance analysis (MCA) is a method that isolates the most coherent pairs of spatial patterns and their associated time series by performing an eigenanalysis on the temporal covariance matrix between two geophysical fields. Different from the normal MCA, the CMCA not only isolates the most coherent patterns between two fields but also excludes the unwanted signal by subtracting the regressed value of each employed field that depends on the unwanted signal. To evaluate the usefulness of the CMCA, it is applied to the tropical Indian Ocean sea surface temperature and surface wind stress anomalies, from which the El Niño?Southern Oscillation (ENSO) signal is removed. Results show that the first mode of the CMCA represents an east?west contrast pattern in SST and a monopole pattern in the zonal wind stress centered at the equatorial central Indian Ocean. The corresponding expansion coefficients are completely uncorrelated with the ENSO index. On the other hand, in the normal MCA, the expansion coefficients are correlated with both the ENSO index and the Indian Ocean east?west contrast pattern index. Thus, the CMCA method effectively detected the coherent patterns induced by the local air?sea interaction without the ENSO signal considered as an external factor, whereas the normal MCA detected the coherent patterns, but the effects of local and external factors cannot be separated.