Ecological Systems and Devices Annotation << Back Use of artificial neural networks for forecasting pollution of atmospheric air by emergency sources at changing meteorological conditions Savitskaja T.V., Dudarov S.P., Lyovushkina S.A. etc. The models of operative forecasting of pollution of atmospheric air based on neural networks as a result of emissions of dangerous chemical substances at changing meteorological conditions are offered. The problems of forecasting meteorological conditions at each three hour and values of the maximal concentration, length and width of a zone of pollution as a result of accidents for each 30 minutes are solved. The results of training and testing of the offered neural networks are presented.