Neural Networks in Analysing 137Cs Behaviour in the Air in the Belgrade Area

Abstract

The application of the principal component analysis and artificial neural network method in forecasting Cs-137 behaviour in the air as the function of meteorological parameters is presented. The model was optimized and tested using Cs-137 specific activities obtained by standard gamma-ray spectrometric analysis of air samples collected in Belgrade (Serbia) during 2009-2011 and meteorological data for the same period. Low correlation (r = 0.20) between experimental values of Cs-137 specific activities and those predicted by artificial neural network was obtained. This suggests that artificial neural network in the case of prediction of Cs-137 specific activity; using temperature, insolation, and global Sun warming does not perform well, which can be explained by the relative independence of Cs-137 specific activity of particular meteorological parameters and not by the ineffectiveness of artificial neural network in relating these parameters in general.