ANALYSIS OF T, S-CHARACTERISTICS OF THE ATLANTIC WATER MASS IN THE EURASIAN BASIN USING THE CLUSTER METHOD

Cluster analysis was performed using CTD data from transects across continental slope of the Eurasian Basin collected within the NABOS (Nansen and Amundsen Basins Observing System) project in 2002–2015. The most commonly used k-means clustering method was applied for calculations. Two cases were considered: division into two and three clusters. The identified clusters which united points on the θ, S-diagrams (θ, S are potential temperature and salinity, respectively) satisfactorily corresponded to the θ, S-values of the Fram and Barents branches of the Atlantic water. For each cluster the thermohaline characteristics of centroids (cluster centers) were calculated, that is, the average values at the same time of temperature, salinity and potential density of each water mass. The thermohaline characteristics of centroids, estimated from the available data obtained in different years of observation, were presented in θ, S-coordinates and θ, σ-coordinates (σ is potential density). Additionally, dependences of temperature, salinity, and potential density of centroids on the year of observation were analyzed. The final results made it possible to obtain estimates of the average thermohaline characteristics of Atlantic water in different years of observation and the variability of these estimates from year to year. In particular, it was found that the values of the average thermohaline characteristics of the AW indicated a strong warming and salinization of the AW in 2006–2009 (sections along 126° E), and the response to warming was observed in all AW clusters. An interpretation of the obtained results was given. Approaches to the choice of the most optimal method of cluster analysis were considered. The results of estimating the influence of temperature and salinity variability ranges in the analyzed CTD data on the accuracy of centroids evaluations were presented.