Freshwater Sources in the Global Ocean Through Salinity-δ18O Relationships: A Machine Learning Solution to a Water Mass Problem

Abstract

Changes in the hydrological cycle can affect ocean circulation and ventilation. Freshwater enters the ocean as meteoric water (MW; precipitation, river runoff, and glacial discharge) and sea ice meltwater (SIM). These inputs are traced using seawater salinity and stable oxygen isotopes in seawater, ${\delta }^{18}O$. We apply a self-organizing map, a machine learning technique, to water mass properties to estimate the global distribution of the isotopic signature of MW $\left({\delta }^{18}{O}_{MW}\right)$ by characterizing distinct salinity-${\delta }^{18}O$ relationships from two comprehensive data sets. The inferred ${\delta }^{18}{O}_{MW}$ is then used in a three-endmember mixing model to provide a globally coherent MW and SIM contributions to the extratropical ocean freshwater budget. Through the use of ${\delta }^{18}O$, our results show the role of MW and SIM in dense water formation and the resulting interhemispheric asymmetry in the freshwater sources that fill the interior ocean freshwater budget. Trends drawn in $\theta$-S space show a significant decrease in sea ice formation driving the freshening of Antarctic bottom water for the 1980–2023 period, whereas SIM is significantly increasing in parts of the Arctic halocline. The different roles of sea ice in dense water formation has implications for future ocean circulation under climate change, where machine learning techniques applied to ${\delta }^{18}O$ have been proven to have utility in detecting such changes.