purely deterministic approach in geotechnical engineering usually is chosen despite there is enough knowledge about the uncertainties and spatial variability of the soil parameters. The geotechnical design and recommendations are based on some field observations, measurements and calculations with homogenized material and simplified models. These affect the results and important decisions in the design process. With the knowledge of the spatial variation and autocorrelation of the parameters, statistical methods can be applied in geotechnical design as well as in the risk assessment of the earthen structures. The Paper delivers statistical data and methods to handle these, which are investigated systematically in a river dike. The importance of the spatial variation and autocorrelation of the grain size distribution is demonstrated regarding suffusion phenomena.