Data-driven analysis of joint coastal extremes near a large non-tidal estuary in North Europe

The Vistula river estuary, located in the Gulf of Gdańsk, is the largest river mouth in the entire Baltic Sea catchment area. Thus, the assessment of the likelihood of flood events for this region has become a pressing issue. Such analyses should include the worst scenario of joint coastal extremes due to simultaneous high levels of water in the river and sea. The simultaneously recorded daily water levels in the river and sea in the 1961–1989 period have made it possible to employ data-intensive signal processing techniques to address this problem by: (1) identification of the physical background of patterns embedded in both series with the Singular Spectrum Analysis (SSA) and (2) assessment of the degree of coupling between pairs of patterns from both series that are deemed most likely to trigger a joint extreme; this is done with the Canonical Correlation Analysis (CCA). For the Vistula estuary it was found that although the annual cycles encompass the largest portions of variances of both signals, they cannot be held responsible for joint extremes due to a considerable phase shift between them. However, both signals are often disturbed by random deviations from this basic behavior and these deviations emerged as separate patterns. Their magnitudes are to some extent correlated with the growth of water level in the sea and river and therefore they could be assumed to act as a pair of potential drivers of joint extremes. The degree of their coupling was found to be generally low with maximums in November, February and/or March, which indicates the low likelihood of a great flood event due to the combined effect of high sea and river water level. The results are also believed to be at least partly representative of other estuaries in North Europe, where annual cycles in seawater and water in (large) rivers are similar, e.g. Scheldt, Rijn, Elbe.