The introduction of the new 'Waterwet' led to the retesting of a lot of primary flood defences within the Netherlands, sequentially certifying a lot of levees as unsafe regarding inner slope stability. To test or design a levee an adequate amount of soil investigation is required to indicate the presence of different soil types, stratification and to estimate the relevant soil parameters. The spatial spreading and the fact that the number of cone penetration tests (CPT) and boreholes within a levee to determine geotechnical parameters and stratification are limited, means that the estimated parameters always contain uncertainties. Nevertheless current guidelines do not cover the minimum CPT interval size or the uncertainty in undrained shear strength in longitudinal direction. In this thesis it is analysed how the quantity of CPTs in longitudinal direction affect the estimated undrained shear strength, and how this influences the probability of failure regarding inner slope stability.

A 5 km long levee containing a continuous peat and clay layer along the Markermeer is specified as research area. In this area the CPT interval size in longitudinal direction is 100 m. The inner slope stability is analysed with D-Geo Stability (level I reliability approach) and the Probabilistic Toolkit (level II reliability approach). Although simplifications within the model ensure that only layer thickness and undrained shear strength vary in longitudinal direction, the calculated safety factors still show a lot of fluctuation. The probabilistic stability analysis show that in the considered research area, the undrained shear strength has a larger influence on the probability of failure in comparison to layer thickness, unit weight and outside water level.

As the CPT interval size increases in longitudinal direction, the minimum safety factor could be overseen, which can lead to optimistic reliability estimations. The theory of information entropy is used to analyse the effect of changing CPT interval sizes on the estimation of the undrained shear strength. Information entropy measures the amount of uncertainty/information in a dataset. A new partial factor (the ‘spatial factor’) is proposed based on the normalized entropy. The proposed spatial factor accounts for spatial uncertainty in the estimated undrained shear strength in longitudinal direction and assigns additional safety when the CPT interval size is larger than 100 m for the considered research area. This methodology promotes the collection of additional data, which is thought to be the basic starting point for any partial factor that is assigned to cover epistemic uncertainties.