The tempo-spatial variability of the Black Sea surface circulation was investigated using satellite data from the Archiving, Validation, and Interpretation of Satellite Oceanographic data (AVISO) project. Self-Organizing Maps (SOMs) were utilized, a neural network-based method known for its unsupervised learning capabilities, to analyze and categorize the circulation patterns, for the very first time. To assess the clustering precision achieved by the SOM algorithms the Davies-Bouldin Index (DBI) was employed as an internal validation metric. Through this comprehensive analysis, six distinct spatial patterns were identified, each exhibiting unique temporal variabilities and occurrence rates. Pattern 1: Characterized by the Sevastopol Cyclonic and Batumi Dipole Eddies, occurring 21 % of the time; Pattern 2: Defined by the Cyclonic RIM Current and Anticyclonic Batumi Eddy, with a 16 % occurrence rate; Pattern 3: Consisting of Anticyclonic Sevastopol and Batumi Eddies, occurring 17 % of the time; Pattern 4: Featuring the Cyclonic RIM Current and Cyclonic Batumi Eddy, also with a 21 % occurrence rate; Pattern 5: Marked by the Anticyclonic RIM Current and Batumi Dipole Eddies, with a 15 % occurrence rate; Pattern 6: Characterized by the Anticyclonic RIM Current and Multi Eddies, occurring 10 % of the time. To further validate the identified patterns, their relevance for predicting the hydrodynamics of the Black Sea was examined. This was achieved by exploring potential correlations between these patterns and major climatological indices, such as the North Atlantic Oscillation (NAO), the East Atlantic/West Russian (EAWR) oscillation, and the El Niño-Southern Oscillation (ENSO). These indices are known to influence large-scale atmospheric and oceanographic conditions, and understanding their relationship with the identified patterns can enhance predictive models of Black Sea dynamics. The findings from this study provide valuable insights into the complex circulation patterns of the Black Sea and their temporal behaviors. The use of advanced neural network techniques such as SOMs, combined with rigorous validation methods like the DBI, underscores the robustness of the analysis. Moreover, the established connections with climatological indices offer a promising avenue for improving long-term forecasts and understanding the broader climatic impacts on the Black Sea's surface circulation.
Read full abstract