Nonlinear principal component analysis (NLPCA) is a generalization of tradi
tional principal component analysis (PCA) that allows for the detection and
characterization of low-dimensional nonlinear structure in multivariate da
tasets, The authors consider the application of NLPCA to two datasets: trop
ical Pacific sea surface temperature (SST) and tropical Indo-Pacific sea le
vel pressure (SLP). It is found that for the SST data, the low-dimensional
NLPCA approximations characterize the data better than do PCA approximation
s of the same dimensionality. In particular. the one-dimensional NLPCA appr
oximation characterizes the asymmetry between spatial patterns characterist
ic of average Fl Nino and La Nina events, which the 1D PCA approximation ca
nnot. The differences between NLPCA and PCA results are more modest for the
SLP data. indicating that the lower-dimensional structures of this dataset
are nearly linear.