Marking outliers using available methods for identifying such observations should be a standard practice in the database management process. The research aimed to adapt universal data quality control tools and tests to their applicability in the southern Baltic Sea by setting new limit values, enabling the detection of erroneous or suspicious data, which can be subjected to expert verification at a later stage. This verification stage may include analysing current conditions and processes and determining the values measured at a given time and space. Our research has proven that using global algorithms requires adapting the limit values based on regional conditions. The global quality control tests, such as the spike test, Dixon’s 4(σ), Q-Dixon, Hampel, quartile, and gradient tests for oceanographic data, were examined, and their application in the research area was verified. As part of the task, a few tests were conducted in various areas of the Baltic Sea, and they were modified to adapt them to the research results of the southern Baltic Sea. Depending on the methodology adopted, verification tests result in the selection of suspicious observations, enabling their expert assessment and the final qualification of measurements that may be considered outliers.

This study demonstrates changes in the hydrodynamic regime associated with climate change in the southern Baltic over more than 70 years. The analysis of long-term data about sea level, the occurrence of ice cover, waves, and storm surges in the southern Baltic enabled the identification of spatiotemporal variability, including the detection of changes in intensity, frequency, and repeatability of these phenomena. The sea level in the southern Baltic rose by approximately 1 cm/decade from 1886 to 1955. Then, from 1956 to 2019, intensification was observed, and the sea level rose by 1.6 cm/decade and 1.9 cm in the western and eastern parts, respectively. The most intense decadal sea level change in 1955–2019 occurred in March (3.1 cm) and January (2.5 cm), while from July to December, it was at 0.8–1.3 cm. Statistical direct correlation analyses using Spearman’s rank method showed a weak but statistically significant relationship between the mean daily sea level with water temperature and air temperature measured at the same stations. An increase in the frequency of storms in individual decades and a decrease in the number of days with ice was demonstrated. There was no clear trend in the wave conditions regime during the period covered by the analysis in 1980–2021.

This study demonstrates the potential of advanced machine learning techniques, combined with Google Earth Engine (GEE) and JavaScript, for detecting and analysing barchan dunes from satellite imagery. The proposed approach enables accurate dune identification and subsequent estimation of migration rates and directions. Climatic factor analysis revealed conditions that accelerate dune migration and identified which dunes are more sensitive to environmental changes. Temporal patterns of dune migration show distinct peaks in movement speed, strongly associated with increases in average wind velocity. The propagation direction of barchan dunes correlates closely with the prevailing wind direction. Spatially, the highest migration rates occur in the northern part of the erg, where unobstructed wind flow promotes higher speeds. Furthermore, isolated dunes migrate faster than clustered dunes, where inter-dune distances are below 300 m. Dunes within clusters exhibit greater stability, likely due to mutual aerodynamic interactions that limit movement. Wind speed emerged as the dominant driver of dune migration, while soil and air moisture exert a stabilising effect—migration slows under humid conditions and accelerates during dry periods. Elevated sediment temperatures also favor migration, likely due to surface desiccation. Overall, dune migration is a cyclic process influenced by climatic variability, with increased movement during dry, windy years. These findings highlight the importance of integrating remote sensing, machine learning, and climatic data to improve understanding of dune dynamics and their response to environmental forcing.