Self-organizing map–derived regimes as a basis for basin-integrated diagnostics of Baltic Sea hydrographic change (1993–2023)

Abstract

The Baltic Sea has experienced substantial hydrographic changes over the past three decades, including warming, increasing deep-water salinity, and intensified vertical stratification. While these changes are well established, their temporal evolution and transitions between distinct hydrographic states remain less well understood. In this study, we use a Self-Organizing Map (SOM), a neural network–based unsupervised classification approach, to assess monthly temperature and salinity profiles from a 1993–2023 physical ocean reanalysis. This approach identifies six distinct hydrographic regimes, representing a progression from colder and fresher to warmer and saltier states, interspersed with transitional phases. The timing of these regimes is consistent with known episodes of major inflow from the North Sea and associated variability in basin-scale exchange. We further quantify changes in internal heat and salt content, vertical stratification strength, and the relative roles of temperature and salinity in shaping density gradients. Results show that, alongside strengthened stratification during the study period, salinity remains the primary control on density structure. However, recent regimes exhibit an increasing thermal contribution to the density gradient relative to earlier years. This pattern indicates a system that remains largely haline-controlled but is becoming increasingly influenced by temperature, with implications for future stratification, vertical mixing, deep-water renewal, and ecosystem functioning in the Baltic Sea.