A Deep Learning-based Model for Cold Anticyclonic Eddies and Warm Cyclonic Eddies Detection in the Kuroshio Extension

Abstract

Mesoscale eddies are circular flowing currents that can retain and transport salt, heat, and nutrients all around the ocean. Mesoscale eddies can be classified as anticyclonic and cyclonic eddies, which are usually called cold eddies and warm eddies according to the sea surface temperature anomaly within eddies, respectively. However, with the development of oceanic remote sensing technology, abnormal mesoscale eddies, that is, cold anticyclonic eddies and warm cyclonic eddies, are found in many ocean areas. Studying abnormal mesoscale eddies involves various ocean parameters data. However, it is difficult for traditional data mining methods to extract the target features from the massive multi-source remote sensing data rapidly and accurately. Deep learning has advantages in feature learning and sophisticated relationship modeling, which brings opportunities for intelligent analysis and mining of ocean remote sensing big data. Based on the multi-source remote sensing database and the deep learning method, the paper detects abnormal mesoscale eddies in the Kuroshio Extension during 1993–2015. The implementation of this study provides technical support and a theoretical basis for deepening the scientific cognition of ocean mesoscale eddies and optimizing the accuracy of ocean models.