Nonlinear canonical correlation analysis of the tropical Pacific climate variability using a neural network approach

Recent advances in neural network modeling have led to the nonlinear genera lization of classical multivariate analysis techniques such as principal co mponent analysis and canonical correlation analysis (CCA). The nonlinear ca nonical correlation analysis (NLCCA) method is used to study the relationsh ip between the tropical Pacific sea level pressure (SLP) and sea surface te mperature (SST) fields. The first mode extracted is a nonlinear El Nino-Sou thern Oscillation (ENSO) mode, showing the asymmetry between the warm El Ni no states and the cool La Nina states. The nonlinearity of the first NLCCA mode is found to increase gradually with time. During 1950-75, the SLP show ed no nonlinearity, while the SST revealed weak nonlinearity. During 1976-9 9, the SLP displayed weak nonlinearity, while the weak nonlinearity in the SST was further enhanced. The second NLCCA mode displays longer timescale f luctuations, again with weak, but noticeable, nonlinearity in the SST but n ot in the SLP.