Aplicação de redes neurais artificiais na análise de padrões de variabilidade de mesoescala

Abstract

In this paper we apply a Neural Network (NN) to treat large oceanographic datasets, specifically to study the mesoscale variability of an oceanic boundary current. The main objective is to distill the massive oceanographic datasets down to a new space of smaller dimension, characterizing the essential information contained in the data. Due to the natural nonlinearity of those data, traditional multivariate analysis, like the Principal Component Analysis (PCA), may not represent reality. However, Nonlinear Principal Component Analysis(NLPCA)can be performed by a neural network model. This work presents the methodology associated with the use of a multi-layer NN with a bottleneck to extract nonlinear information of the data. We illustrate its good performance with a set of tests against comparisons using this methodology and classical PCA in the Sea Surface Temperature (SST) satellite images of the southwestern Pacific Ocean.