In geotechical design, one of the greatest challenges is the uncertainty in soil properties. Due to inherent variability of the soil, scatter in the values of soil parameters occur even in relatively homogenous soil layer. The uncertainty caused by variability of the soil parameters can be quantified by using coefficient of variation (COV). COV is an important parameter in reliability based design (RBD). In Finland, RBD-methods are not generally used, but in other countries this approach has been adopted more widely. However, the usage of RBD-methods requires reliable information on COV values of different soil properties. The problem is that in Finland, research on inherent variability has been practically nonexistent.

In this thesis, the inherent variability of Perniö clay was studied via COV values. In the analysis, results of laboratory tests in Perniö were used. Based on these COV values, settlement caused by hypothetical railway embankment was estimated using probabilistic approach. As such, the probability of the total settlement after the construction exceeding the maximum allowable value could be estimated. Similarly, the probability of the change of slope caused by differential settlements exceeding the maximum value could be defined. In addition, the effect of the variability of each soil parameter on the settlement value was studied by using sensitivity analysis. It was observed that not only compressibility properties but also the variability in effective overburden pressure and uncertainty in modeling the preconsolidation pressure affect the result of settlement calculation significantly.

As a part of the study, a database on the laboratory test results from not only Perniö but from other Finnish clay deposits was constructed. Based on the database, correlation between clay properties was studied. The database was further used in an analysis, in which typical values of clay properties were estimated. A priori distributions were fitted for both clays with relatively low water content and for clays with higher water content. As such, expected values and typical ranges for different soil properties were acquired. In addition, the effect of testing method and sampler on soil parameters and sample quality was studied. Based on the results, testing method clearly has a significant effect but the effect of sampler seemed to be minor. Since the effect of sampler diameter on sample quality has been observed in several studied, these results might simply be due to special features of Perniö clay.