<p>This study explores how to construct three-dimensional block models of cone penetration test (CPT), Menard modulus and limit pressure, consistent with an underlying stratigraphic model, on a nuclear power plant site under development. From a stratigraphic viewpoint, the site is on recent sediments, and there is a thin silty stratum (thickness between 0.5 and 2.5&#xa0;m), whose geotechnical behaviour differs from the adjacent sandy layers. For instance, the CPT is lower in the silty stratum, raising some geotechnical concerns. Following an exploratory data analysis and study of the spatial stationarity, several geostatistical approaches are designed and tested for stratigraphic and geotechnical modelling. Based on the results, the main conclusions and recommendations for a successful geotechnical modelling can be summarized as follows: (i) On a site under investigation, stratigraphic information is usually more abundant than geotechnical information. Geotechnical parameters depend on stratigraphic units. Therefore, the underlying stratigraphic model should be used while generating a geotechnical model. It is highlighted that stratigraphic modelling is not only important in generating a robust and coherent geotechnical model, but also necessary in correct evaluation of the model uncertainty, quantified by kriging prediction standard error or weighted entropy by kriging efficiency. (ii) The weighted entropy is proposed to quantify the uncertainty of a geotechnical model, which is anchored to an underlying stochastic stratigraphic model. (iii) The necessity of using a flattening technique to reduce the artefacts of geotechnical modelling in folded strata is also illustrated.</p>