Geostatistical Modeling Resistivity of Cohesionless Soil Using RCPTU Data

Resistivity of soils has been widely used for analyzing both geotechnical and geoenvironmental problems. Mapping the distribution of resistivity provides significant information for site characterization including soil properties and contamination in the subsoil. In this research, the distribution of resistivity of a silty sand is investigated using resistivity piezocone penetration test (RCPTU) based on geostatistics. First, to achieve stationary Gaussian random field, the trend removal and Box-Cox transformation were used to convert non-stationary and non-Gaussian resistivity measurements into weakly stationary normal random field. Then semivariogram was used to describe the spatial variability of the residuals including variance, correlation length and nugget. The maximum likelihood (ML) estimator was applied to estimate the theoretical semivariogram, revealing that anisotropy exists in the spatial correlation structure. Ordinary Kriging was then used to provide the best linear unbiased estimates of residuals at unsampled locations of the random field. Finally, cross-validation, or removing some known observations from the database and predicting the known results, was used to validate the geostatistics model. It was shown that the geostatistics is a powerful tool to map the resistivity of subsoil by considering the spatial variability characteristics.