Characterization of Oceanographic Fields in the Gulf of Tonkin and Their Relationship with ENSO Using Cluster Analysis

Abstract

In this study, the Self-Organizing Maps (SOM) method, combined with the K-means algorithm, was employed to analyze the seasonal variability of oceanic physical fields in the Gulf of Tonkin by grouping similar patterns of surface currents, temperature, and salinity into clusters. Analyzing the transitions between these patterns over time enables the identification and assessment of climate characteristics in these physical fields, utilizing a substantially larger amount of information compared to traditional climatological averages. SOM serves to simplify complex nonlinear features into easily observable two-dimensional relationships, while K-means assists in recognizing and highlighting significant spatial characteristics of the clusters. The results indicate that this approach provides a clearer and more intuitive depiction of the seasonal differences and variability of the physical fields, particularly improving the identification of the timing of current reversals and sea surface temperature changes. Furthermore, the study reveals that some representative patterns occur more frequently during El Niño years, whereas others are more prominent in La Niña years. These findings offer a novel approach, contributing to a better understanding of the relationship between oceanic physical fields in the Gulf of Tonkin and ENSO.