In this paper we present a method used in quantitative description of weakly predictable hydrological, extreme events at inland sea. Investigations for correlations between variations of individual measuring points, employing combined statistical methods, were carried out. As a main tool for this analysis we used a two-dimensional copula function sensitive for correlated extreme effects. Additionally, a new proposed methodology, based on Detrended Fluctuations Analysis (DFA) and Anomalous Diffusion (AD), was used for the prediction of negative and positive auto-correlations and associated optimum choice of copula functions. As a practical example we analysed maximum storm tides data recorded at five spatially separated places at the Baltic Sea. For the analysis we used Gumbel, Clayton, and Frank copula functions and introduced the reversed Clayton copula. The application of our research model is associated with modelling the risk of high storm tides and possible storm flooding.