Nonlinear Representation of the Quasi-Biennial Oscillation

A nonlinear principal component analysis (NLPCA) is applied to monthly mean zonal wind observations from January 1956 through December 2007 taken at seven pressure levels between 10 and 70 hPa in the stratosphere near the equator to represent the well-known quasi-biennial oscillation (QBO) and investigate its variability and structure. The NLPCA is conducted using a simplified two-hidden layer feed-forward neural network that alleviates the problems of nonuniqueness of solutions and data overfitting that plague nonlinear techniques of principal component analysis. The QBO is used as a test bed for the new compact model of NLPCA. The two nonlinear principal components of the dataset of the equatorial stratospheric zonal winds, determined by the compact NLPCA, offer a clear picture of the QBO. In particular, their structure shows that the QBO phase consists of a predominant 28.3-month cycle that is modulated by an 11-yr cycle as well as by longer cycles. The differences in wind variability between westerly and easterly regimes and between Northern Hemisphere winter and summer seasons and the tendency for a seasonal synchronization of the QBO phases are well captured.