Due to their particularly good mechanical and self-healing properties combined with exceptionally efficient cation adsorbents and exchanger capacities, clay minerals and clay rock formations are considered as suitable geological barrier for radioactive waste disposal. The Middle Jurassic Opalinus clay formation has been identified as a potential host rock”. Logging Data were measured at the Benken borehole drilled through this formation in Northern Switzerland. The paper presents a statistical methodology to improve the description of the physical properties of the clay rock based on the well-log data. The methodology involves the classification of a set of local statistics, calculated from a reduced number of principal components computed from well-log properties. The use of kernel-based method to calculate local statistics allows the analysis of spatial variability at different scales, and scale effects. The first-order layering is found to be robust and independent of kernel size, i.e. observation scale, while preserving small scale heterogeneities useful for further interpretation. The log-units can be more clearly interpreted in terms of stationary or transitional log-units, depending on the behaviour of local statistics. Finally, the derived spatial variability of the log-units property is compared with earlier lithological descriptions and stratigraphic data.