Introduction. This work is aimed at solving the problem of phytoplankton dynamics in the coastal environments using the example of the Azov Sea. This takes into account the transformation of forms of phosphorus, nitrogen and silicon, as well as the aquatic medium motion, the distribution of temperatures and salinities over the sea area. River flow, varying in volume and chemical composition, affects significantly the variability of hydrophysical and biogeochemical parameters of the processes occurring in the coastal environment. This explains the need for statistical processing of the data from long-term observations over the river flow characteristics. Materials and Methods. The mathematical model of biogeochemical cycles is based on a system of non-stationary equations of the convection-diffusion-reaction of parabolic type with nonlinear functions of sources and lower-order derivatives, to which the corresponding initial and boundary conditions are added. In the course of statistical analysis of the series of long-term observations over river flows, the values of the following indicators were found: skewness coefficient, degree of kurtosis, variance and standard deviation, coefficient of variation, autocorrelation coefficient, Neumann ratio, and Anderson criterion. Results. The statistical analysis of the series of long-term observations over the hydrochemical indicators of the Don river suggests heterogeneity of the field data. This is due to the stochasticity of nutrient inputs and the volume of freshwater flow to the sea as a result of natural and anthropogenic factors. Field data should be correlated with seasonal changes in the aquatic environment temperature. This paper presents the results of a computational experiment to model the dynamics of phytoplankton populations in summer season, when temperatures are favorable for their reproduction and growth. The proposed mathematical model considers the spatially inhomogeneous distribution and transformation of forms of phosphorus, nitrogen, and silicon, as well as changes in salinity, temperature, and motion of the aquatic environment. Discussion and Conclusions. The multispecies mathematical model of the dynamics of phytoplankton populations is considered with account for the transformation of forms of phosphorus, nitrogen, and silicon in the coastal environments. The analysis of data from field observations, for which its major statistical parameters are calculated, is carried out. As a result, it is concluded that data of the long-term observations are significantly variable. This is due to two reasons. Random nature of the input of nutrients and the volume of river flow as a result of anthropogenic factors is the first reason. The second reason includes the alternation of relatively high-water and low-water periods for fresh flow over the last 12-15 years. The hydrological regime is changing mainly due to the reduction of the average annual freshwater flow of the Don and partly of the Kuban. This trend is likely to increase due to climate changes, as well as with further regulation of the Don river flow after the Bagaevsky hydroelectric installation start-up. Numerical experiments based on the field data confirmed the predictive validity of the developed models and programs. They can be used to predict change in the composition and abundance (concentrations) in the Azov sea core planktonic populations, which define, on the one hand, food resources, and, on the other hand, the aquatic environment in terms of the ongoing sea salinization.