The Baltic Sea runoff events during 1970-2000 were analysed with non-linear estimation in order to find potential stability and time-varying (non-constancy) properties of time series parameters. In the Baltic Sea area, the wintertime rainfall is expected to increase, especially in the northern parts of the watershed; furthermore, seasonality is expected to decrease, i.e. the runoffs will become more evenly distributed over seasons, mainly by decreasing springtime peaks. We found a general increase in runoffs over the time period studied that consisted of a level shift upwards during the early 1980s. The runoffs considerably resulted from inputs in the months of April to June (which points to the significance of winter precipitation) and are biased to northern parts of the catchment. Our results suggest that this is connected with snow and ice melting in the northern areas of the watershed, especially in the subarctic catchment area of the Neva River (the area includes, for example, three of the four biggest lakes in Europe, Lake Ladoga, Lake Onega and Lake Saimaa, which are annually covered by ice). The yearly variation in runoff is significantly stabilised by large N-S extension of the watersheds of these lakes. The results are generally in accordance with the presented climatic expectations; however, decreasing seasonality was not obvious in the series yet. However, we found potential implications of system change towards decreased seasonality. By better understanding of ongoing processes in the sea, the results can give us an opportunity to better evaluate the system control, predicted areal consequences and ecosystem effects of the climate change. The analysis proved to be a sensitive tool for studying climatic variation in a relatively small area and over a short time span.