Manifestations of the interaction of riverine and marine waters in the statistical structure of salinity by the observations data at coastal stations of Ukraine

Time series of water salinity observations from 1997 to 2010 at marine hydrometeorological stations on the northern coast of the Northwestern Black Sea from the mouth of the Dnipro-Buh Estuary (Kinburn Strait) to the Gulf of Odesa are analyzed. Empirical probability distribution functions of salinity were constructed for the Ochakiv, Pivdenny and Odesa-port stations for all months and year as a whole. Using the mixture analysis method, one-dimensional clustering of empirical salinity histograms is performed under the assumption that they consist of several normal distribution functions, each of which represents a separate water mass and is characterized by its own indicators - the average value, standard deviation and share (proportion) in the general distribution (mixture). To search for optimal solutions, the maximum likelihood method and the Akaike information criterion were applied. Experimentally, taking into account the peculiarities of the water dynamics in the northern part of the Northwestern Black Sea shelf, it was established that salinity histograms can be approximated by a set of 2-3 Gaussian functions. These functions, as a rule, correspond to waters of river origin, marine origin and intermediate waters as a result of the interaction of the first two. The approximation parameters (mean values, standard deviation and proportion in the mixture) vary in space depending on the distance to the source of desalination or salinization, as well as in time in accordance with the seasonal variability of the processes of supply and interaction of water masses on the North-Western Black Sea shelf. Numerical indicators of the components that make up the water mixture near the northern shores of the North-Western part of the Black Sea have been obtained. They can be used for probabilistic modeling (forecasting) of water salinity at different points both for separate months and for the whole year, for example, by means the method Monte Carlo.