Abstract

The Baltic Sea is a shallow, stratified, brackish water sea with several subbasins. In this work, we analyzed CTD casts from HELCOM monitoring data of the Baltic Sea, data from FMI's glider missions in 2016 and 2017 and profiles from FMI's Argo floats. We used clustering and machine learning algorithms developed for time series analysis to classify vertical profiles. We endeavored to classify profiles into classes with similar shapes. Then, we defined whether it was possible to define the depth of the upper mixed layer and possible halocline depth in each of these classes. Our results show that time series classification algorithm can cluster vertical profiles of CTD temperature and salinity and classify them according to sea area and season.