A methodology to understand the wind-driven circulation in semi-enclosed bays and its connectivity with the open sea at the southern coast of France

By applying an unsupervised neuronal network (NN) to sea velocity profiles and wind data, it was possible to determine the main wind-driven circulation patterns in the Toulon bay. In addition, the NN outputs were utilized to perform a conditional averaging to High-Frequency radar surface current data (HFR) and the atmospheric AROME model, in order to understand the connectivity between the inner Toulon bay circulation features and the offshore marine-atmospheric conditions. For instance, upwelling scenarios are observed under strong westerly winds, whereas the downwelling is present under easterly wind conditions. Additionally, a barotropic system is observed when weak-mid wind blows for long time periods, and first baroclinic modes occur under strong wind events. Up to date, few studies have presented a clear connectivity between semi-enclosed bays and the offshore conditions, particularly in the northwestern Mediterranean Sea. Thus, this methodology presents great advantages when trying to study the interaction between semi-enclosed bays and the open sea by means of a combination of several in situ meteo-marine information.