Abstract

The ocean's structure, characterized by the volume and organization of water masses, is closely linked to climate dynamics. Paleoceanography relies on the idea that water mass geometry is encoded in isotope ratios and trace element concentrations in seawater, and then preserved in sediments. Chemical and paleo oceanographers use their intuition about the positions of water masses, developed from studying modern property distributions, to interpret new data. It is therefore important to accurately and objectively label water masses, particularly if we hope to track their evolution over time. Here we apply various clustering algorithms, a form of unsupervised machine learning, to active and passive tracer distributions from the World Ocean Atlas 2018 (Boyer et al., 2018) and the most modern slice of C-iTRACE (Gu et al., 2021) model output. We compare the geographic distributions of clusters identified with active …

Date

January 1, 1970

Authors

Jordan Landers, Julien Emile-Geay, Deborah Khider

Journal

AGU Fall Meeting Abstracts

Volume

2023

Issue

1762

Pages

OS41C-1762