Ecological Systems and Devices Annotation << Back The Method for Neural Network Calculation of Concentrations of Priority Pollutants in the Surface Layer of Atmospheric Air Tunakova Yu.A., Novikova S.V., Shagidullin A.R., Valiev V.S. The development of methods for computational monitoring of atmospheric air pollution plays an important role in ensuring sustainable development on the territories with high level of atmospheric pollution. The development of local models of atmospheric diffusion of impurities is carried out on the basis of taking into account the characteristics of turbulent air fl ows, determined using a system of hydrothermodynamic equations or empirical data (Gaussian models). Such a description of the atmosphere is deterministic, while in the surface layer of atmospheric air, each calculated concentration value at given parameter values corresponds to an infi nite set of observed impurity concentration values. The complication of deterministic calculation models is associated with an increase in the diffi culty and resource intensity of calculations. Therefore, the accuracy of calculations using the normative calculation model can be increased by using artifi cial neural networks. Neural networks are trained on the data of experimental measurements of surface concentrations of pollutants, and then can reproduce the complex effect of meteorological factors, local dispersion conditions and compensate for the error in the initial parameters of stationary and mobile emission sources. The article formulates the method for using artifi cial neural networks to calculate surface concentrations of pollutants in the atmospheric air. The results of designing and training neural networks for calculating the concentrations of common impurities are presented. Keywords: atmospheric pollution, atmospheric emissions, dispersion calculation, artifi cial neural network. DOI: 10.25791/esip.11.2023.1410 Pp. 24-31.