Neuro-fuzzy method of processing hydrochemical data for river flow

Production and socio-environmental requirements for the quality of land waters have determined the need to create a network of hydrochemical observation posts, and the variability of controlled indicators – the need to perform routine chemical analytical studies. The standard (rigid) statistical methods of processing measurement results common in analytical chemistry, as a rule, underestimate the specifics of studying noisy (fuzzy) experimental data, which are the series of values of the impurity concentration of a river stream in space and time. It is shown that in this case, alternative soft computing tools designed to process exactly such data based on neuro-fuzzy hybrid algorithmic structures related to the ANFIS architecture are appropriate. The arrays of chemical analytical data on copper and zinc analyzed in this way on the Volga River, depending on water flow at different distances from the shore and depths, made it possible to identify the complex oscillatory behavior of concentrations of both substances in the water stream. It is concluded that the neuro-fuzzy scheme for processing monitoring results provides an opportunity for in-depth study of poorly understood processes of hydrochemical dynamics in systems far from thermodynamic equilibrium, which include natural watercourses.