This paper presents a probabilistic approach to the analysis of non-stationary sea level measurement series in the southern Baltic Sea based on tide gauge data for Swinoujscie, Kolobrzeg, Ustka, Wladyslawowo and Gdansk stations. Harmonic analysis (HA), Continuous Wavelet Transform (CWT), AR(1) autoregressive model and Monte Carlo uncertainty propagation were applied to identify trends, multiscale variability and the stochastic structure of the measurement data. The results indicate a spatially consistent sea level rise trend of 1.8–2.2 mm/yr, modulated by multiscale periodic variability and short-term stochasticity. The model used allows for a probabilistic forecast of sea level changes. In addition, the analysis of extremes using the Gumbel distribution indicates an increase in the probability of extreme sea levels along the southern Baltic coast. The proposed methodology extends conventional sea level analysis by integrating probabilistic interpretation, classical uncertainty estimation, and multiscale signal analysis, thereby providing a useful tool for coastal hazard and flood risk assessment and climate change adaptation in coastal regions.