Detection of coherent thermohaline structures over the global ocean using clustering

The classification of the ocean in water masses with similar physical and/or biogeochemical characteristics provides an ideal framework for an efficient monitoring of the change in ocean properties. Particularly, the combination of the seawater temperature and salinity set the stratification of the water column and impacts ocean circulation, with important implications for the surface-to-interior propagation of climate signals. The objective of this study is to find spatially coherent thermohaline structures in different regions of the global ocean, as well as their link with regional dynamics. To this end, we apply clustering techniques to identify water masses delimited by coherent thermohaline structures at different spatial scales over the global ocean. The clustering technique known as K-mean was used with a wide range of $k$ values for conservative temperature and absolute salinity profiles of the entire ocean. Our analysis revealed the impact of the main dynamical oceanic structures (such as ocean fronts, currents, or regions of sluggish circulation) on the vertical thermohaline component. We present three cases of study, the California Current System, the Southern Ocean, and the Eastern Tropical Latitudes, where we could identify regions with common thermohaline characteristics (despite being from different ocean basins), as well as to identify seasonal changes and anomalous profiles. This method makes it possible to reduce the dimensionality of the water column, and allows for the establishment of regional limits driven by their vertical thermohaline structure instead of more rigid, and not always appropriate geographical borders. This has potential important applications for the monitoring and prediction of ocean variability in the context of a rapidly changing climate.