The Nonlinear Relationships Between SSTA Over Tropical Pacific And 500hPa Height Field Over Pacific Northern American

Based on the monthly average analysis data of SST by NOAA and monthly average 500hPa height reanalysis data by NCEP/NCAR, the nonlinear characteristics of Tropical Pacific SSTA and 500hPa height field over Pacific Northern American are determined with the techniques of NLPCA and NLCCA, and also the relationships between them. NLPCA and NLCCA are based on neural network methods. They are nonlinear generalization of traditional PCA and CCA. An EOF mode can only express the standing oscillation in multivariate datasets. A CCA mode can only express the linear relationships between two datasets. But it's usually impossible to expect the best approximation to a dataset is linear, and also the relationships between the two datasets are linear too. NLPCA and NLCCA are able to detect and characterize the more general low-dimensional structures than EOF and CCA. The results of SSTA based on NLPCA show that the largest positive anomaly during warm events is close to Peru. But the largest negative one during cold events is centered near 150°W. The intensity of positive center (4.5~5℃) is much larger than the negative one(-1.5~-2℃). For the 500hPa height field based on NLPCA, when u takes on maximum value, the 500hPa height anomalies show a weakened jet stream which does not have an equally strong negative counterpart, but is overshadowed by the negative PNA state found during minimum u. Apparently, NLPCA model is related to ENSO only when u is negative. We turn to NLCCA for the relationships between SSTA and 500hPa height field. When u variates from high to low, the reconstructed SSTA is evolving from El Nino to La Nina. The center of La Nina (-1.5—2℃) is 50° westward to the one of El Nino (4.5~5℃) . The spatial structures and the amplitudes of SSTA for ENSO by NLCCA are similar to the conclusions by NLPCA, the correlation is 0.97 and 0.96 respectively corresponding to the warm and cold events. To the 500hPa height field, the nonlinearity is largely due to a phase shift of the stationary wave anomalies. The four anomalies of PNA for El Nino are 20° eastward to the ones of PNA for La Nina event. The conclusions of the two nonlinear methods are basically similar to the composites.