Interpolation modeling and mapping of environmental radioactivity using geostatistic methods, artificial neural networks and fuzzy logic approach

In this study, appropriateness of geostatistic analysis techniques (Kriging techniques) and artificial intelligence applications (artificial neural networks and fuzzy logic approach) were examined and comparatively evaluated in order to investigate the environmental radioactivity distribution and ensure its monitoring. Radiological dispersion of Artvin was determined by estimating activity of interpolation regions through each method used in the study. Outdoor gamma dose measurements were performed in 204 outdoor stations, that are considered to be representing Artvin, and soil and water samples were collected from 117 stations. Interpolation estimation models were created with 70% of the activity data obtained from these samples and the remaining 30% of the dataset was separated as the test data to be used to determine performance of the methods. In addition, radiological distribution maps were created for soil (Ra-226, Th-232, K-40, Cs-137), water (gross alpha and gross beta) and air (outdoor gamma dose rate) measurements by using these interpolation estimation methods. The findings obtained indicate that artificial intelligence approaches have shown better performance while determining radiological distribution.