Modeling uncertainties in sodium spatial dispersion using a computational intelligence-based kriging method

Classical and geostatistical methods have been used to create continuous surfaces from sampled data. A common geostatistical method is kriging, which provides an accurate estimation based on the existing spatial structure of sample points. However, kriging is sensitive to errors in the input data, the dispersion of the sample points, and the fit of the model to the variogram. The purpose of this research is to develop a new method to address the uncertainties resulting from the input data and choice of model in the kriging method. In our approach, the existing uncertainties in the input data are modeled by fuzzy computations, and the variogram variables are optimized by a genetic algorithm. To test this new hybrid method, sodium contamination values in the Zanjan aquifer were used. The results show a general improvement in accuracy compared with the ordinary kriging method. Consideration of all equations and values in fuzzy computations highlights the complexity of the computation. Herein, the integration problems experienced by other researchers when trying to use fuzzy kriging are resolved.