Scholarlink

<p>The spatial and temporal variability of ocean drives its mixing which in turn drives the ocean heat uptake and circulation. Determining Indian Ocean spatial structures has relied on ad hoc combinations of its physical, chemical, and dynamic properties. In this study we apply an unsupervised classification technique (Gaussian Mixture Modelling, or GMM) to Indian Ocean Argo float temperature and salinity profiles as a step toward an alternative method for defining spatial variations. GMM automatically distinguishes many spatially coherent groups influenced by temperature and low/high salinity water masses without using any spatial position (latitude or longitude) detail. In addition, GMM detects outliers among the various Argo Floats. GMM can be used to classify constructs in both observational and model data sets because it is stable, systematic, and automatic, making it a valuable supplement to existing classification techniques. This is noteworthy because the climate science and ocean sciences communities need a new generation of methods, including the machine learning techniques described in this work, to deal with a vast and growing amount of observational and computer model data.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A machine learning approach to identify upper ocean water masses in the Indian Ocean

  • 期刊论文

  • 发表日期：

    2025年7月29日

查看全文
  • Shikha Singh,
  • Mayur Sapkal

摘要

The spatial and temporal variability of ocean drives its mixing which in turn drives the ocean heat uptake and circulation. Determining Indian Ocean spatial structures has relied on ad hoc combinations of its physical, chemical, and dynamic properties. In this study we apply an unsupervised classification technique (Gaussian Mixture Modelling, or GMM) to Indian Ocean Argo float temperature and salinity profiles as a step toward an alternative method for defining spatial variations. GMM automatically distinguishes many spatially coherent groups influenced by temperature and low/high salinity water masses without using any spatial position (latitude or longitude) detail. In addition, GMM detects outliers among the various Argo Floats. GMM can be used to classify constructs in both observational and model data sets because it is stable, systematic, and automatic, making it a valuable supplement to existing classification techniques. This is noteworthy because the climate science and ocean sciences communities need a new generation of methods, including the machine learning techniques described in this work, to deal with a vast and growing amount of observational and computer model data.