Characterizing spatial variability of cone penetration testing through geostatistical evaluation

Geostatistical methods are used to quantify the spatial variability in cone penetration test (CPT) tip resistance at a site in southwest Indiana. Spatial variability of soil properties affects all geotechnical projects. Geostatistical tools provide a method to both predict the properties at unsampled locations and to quantify their uncertainty. Universal kriging is a geostatistical method that accounts for a variable mean, such as a varying mean tip resistance in different soil layers, when making predictions. In this paper, universal kriging is applied to a database of CPT results. Because CPT tip resistance shows greater variability with depth than with lateral distance, a method of accounting for greater correlation between measurements in the lateral direction than with depth is presented. Finally, cross-validation, or the removal of measured points from the database and use of the model to predict the known results, is used to validate the geostatistical model.