Using the Support Vector Machine Classification to Identify Water Masses in the Western Alboran Sea

Quasi-synoptic hydrological data collected in the Western Alboran Sea, during a glider cruise, dedicated to sample the Western Alboran Gyre (WAG) in late fall 2020 have been used to examine the distribution of the water masses in the region. The spatial variability has been addressed by means of a Support Vector Machine (SVM) classification applied on TS diagram through a training set of more than 3000 vertical profiles of temperature and salinity from the World Ocean Database (WOD). The training process is based on a manual labelling data in which the frontiers between the water masses have been defined on the basis of the values of temperature, salinity, and density, inferred from traditional expert analysis of the TS diagram in the area. Compared to conventional clustering analysis methods, this approach is appropriate to distinguish water masses in regions where intense mixing occurs such as the Alboran Sea. This new method highlights the WAG's characterization by recent AW advected from the Strait of Gibraltar, the presence of the Levantine Intermediate Water (LIW) in the northern two-third of the basin and the uplift of the dense Mediterranean Waters in the south, in the sill surroundings, roughly parallel to the Moroccan continental slope where they mix directly with AW.