%0 Journal Article %A Rigol-Sanchez, J P %D 2016 %T Spatial Interpolation of Natural Radiation Levels with Prior Information using Back-propagation Artificial Neural Networks %U https://bridges.monash.edu/articles/journal_contribution/Spatial_Interpolation_of_Natural_Radiation_Levels_with_Prior_Information_using_Back-propagation_Artificial_Neural_Networks/4004457 %R 10.4225/03/58006cbdc9dcc %2 https://bridges.monash.edu/ndownloader/files/6486915 %K monash:89909 %K 1959.1/736809 %K SIC 2004 %K Artificial neural networks %K ANN %K Automatic mapping %K Feed-forward back-propagation neural network %K Environmental data %K Spatial interpolation %K Outliers %K Extreme values %K Algorithms %K Evolutionary approach %K Radioactive contamination %K 2005 %K collection(s) Applied GIS %K text %K journal article %K 1832-5505 %X We propose artificial neural networks (ANNs) as a tool for automatic mapping of daily observations of environmental data. A feed-forward back-propagation neural network for estimating daily natural radiation measurements at unsampled locations using prior information was developed. Feed-forward back-propagation networks were trained to learn: (a) the relationship between daily measurements and their spatial coordinates, and (b) the relationship between daily measurements made at one site and measurements made at the six surrounding closest sites. Results of the study indicate that ANNs can be used for automatic mapping of environmental (background) data with moderate success. ANN models for spatial interpolation can successfully incorporate prior information into the estimation process. However, the ANN approach to automatic mapping of environmental data presented here was clearly inappropriate for dealing with outliers. Results obtained suggest that developing two different models for estimating background values and extreme values, respectively, might be a potentially more successful approach to automatic mapping of environmental data. %I Monash University