Site Characterization in Geotechnical Engineering-Does a Random Field Model Always Outperform a Random Variable Model?

In the site characterization, the spatial trend of a soil property is oftentimes characterized by a polynomial fit, and the residual (between the spatial trend and the on-site 'true' property) may be characterized either by a random variable or a random field. Note although the random field model is sophisticated and theoretical sound, the random variable model is easy to use. The choice between the random variable model and the random field model is not a trivial task; and, this issue becomes more profound in the face of the uncertainty in the characterization of the random variable or random field. The uncertainty of the characterization of the random variable or random field might be attributed to either the limited availability of the test data and the existence of the test error. An illustrative site characterization problem is undertaken, in this study, to investigate the influence of the residual characterization in the site characterization on the bearing capacity prediction of a drilled shaft in clay. The focus is on the comparison of the random variable model with the random field model. With the results presented, it is found that although the random field model may lead to a more precise prediction of the statistics of the soil property, the random variable model could exhibit a greater robustness in the prediction of the geotechnical system performance.