ARTIFICIAL NEURAL NETWORKS AND SPATIAL ESTIMATION OF CHERNOBYL FALLOUT

Abstract

The present work continues advanced spatial data analysis of surface contamina-tion by radionuclides after severe nuclear accident on Chernobyl NPP. Feedforward neural networks are used for the Cs137 and Sr90 radionuclides prediction mapping and spatial esti-mations. Neural networks are used to model complex trends over the entire region. Residu-als are analyzed with the help of geostatistical approach within the framework of NNRK (neural network residual kriging)model. Another set of data is used to validate obtained re-suits.