Bayesian kriging to characterize spatial variability.

In many geotechnical problems, it would be impossible to obtain exhaustive data for every desired point within a volume of soil because of practical and economical constraints. This lack of data is a major source of uncertainty in site characterization. Among techniques which can be used to model spatial variability, Bayesian kriging technique merges the kriging method (useful to evaluate the variability of the soil within a volume) and Bayesian theory. The approach can be used to combine the results of geophysical and geotechnical site investigations and to predict, from the data observed at known locations, the unknown values of different soil parameters within an area or volume of interest. This study presents the implementation of both ordinary kriging and Bayesian kriging to a subsurface characterization problem with quantity of data from a geophysical survey and several observations based on geotechnical borehole measurements. A first approximation was pro-vided by geophysical survey for Bayesian kriging. The results of the analyses in the paper show that estimated parameters of semi-variograms could be different using different methods (i.e. ordinary least squares and max-imum likelihood methods) and a better prediction can be achieved by using Bayesian kriging.