The present paper responds to the challenge of modeling uncertainty in soil strength parameters concerning its spatial variability in a situation of limited soil information. Understanding this uncertainty allows for the management of the risk of geotechnical structure failure. In the present work, this uncertainty is identified based on signals from the cone penetration test (CPT) device. Signals are directly transformed using existing interpretation methods (typically used as a source of mean values of parameters for a given range of depths) to obtain depth-varying effective strength parameters of the soil. The process is performed by incorporating data from two case studies from different locations in similar soil materials. First, Keswick clay from Australia, for which the results of both CPT and laboratory tests are available, is examined. Second, to further verify the obtained results, the soil from Poland called Świerzna clay, for which only CPT signals were available, is also tested. As shown, the variability of the transformed signals can be a good source of information for identifying uncertainty in soil strength. It agrees well with literature data and can be used to identify random fields describing soil parameters.

The design process in geotechnical engineering requires the most accurate mapping of soil. The difficulty lies in the spatial variability of soil parameters, which has been a site of investigation of many researches for many years. This study analyses the soil-modeling problem by suggesting two effective methods of acquiring information for modeling that consists of variability from cone penetration test (CPT). The first method has been used in geotechnical engineering, but the second one has not been associated with geotechnics so far. Both methods are applied to a case study in which the parameters of changes are estimated. The knowledge of the variability of parameters allows in a long term more effective estimation, for example, bearing capacity probability of failure.

Random field theory is currently the most important tool used when modelling spatial variabilities in soil properties. However, there are practical difficulties in indicating spatial correlation parameters, such as horizontal and vertical fluctuation scales. Therefore, the performed analyses touch upon the occurrence of the worst-case effect of the horizontal fluctuation scale. The random finite element (RFEM) approach is utilised to estimate probability distributions of the bearing capacity of a spread foundation. The analyses show that the observed location of the worst-case phenomenon is closely related to the adopted safety factor. Therefore, it is difficult to directly indicate the horizontal fluctuation scale for which the failure probability has the greatest value. Nevertheless, some general guidelines are formulated and proposed. These general rules can be applied to the two-dimensional bearing capacity problem.