An ISOMAP Analysis of Sea Surface Temperature for the Classification and Detection of El Niño & La Niña Events

Isometric feature mapping (ISOMAP) is a nonlinear dimensionality reduction method used for extracting features from spatiotemporal data. The traditional principal component analysis (PCA), linear method, measures the distance between two data points based on Euclidean (line segment), which cannot reflect actual in space. By contrast, ISOMAP geodesic distance, more closely reflects by view of tracing along local linearity original structure. Thus, ISOMAP-reconstructed can real structures and be classified accurately than PCA-reconstructed points. Moreover, these cluster emphasizing differences among those PCA. In this study, sea surface temperature (SST) reconstructed using PCA were compared. classification SST was tested Niño 3.4 index, labels El Niño, La Niña, or normal events. mean larger not only helped differentiate different events but also provided better difference measurement belonging to same class (e.g., 82/83 97/98 events). On examining evolution leading three temporal eigen components PCA, especially ISOMAP, we found that trajectories similar Lorenz 63 model phase space figure. This implies NWP perturbations traced measure growing unstable behaviors. Spatial eigenmodes (empirical orthogonal function) determined compared herein.