In order to understand and to model soil runoff erosion as well as to optimize sampling schemes, improved understanding of spatial variability of clay and some other soil erosion parameters is needed. For this purpose, two complementary approaches to the study of spatial variability of silty topsoil characteristics, the pedological approach and the fractal approach applied to geostatistics, were carried out in the context of soil erosion within the intensive cereal agriculture zone of northwestern Europe. Fractal geometry provides one synthetic key to the description of classical geostatistical tools such as variograms. Spatial structures of soil properties of each of the three topsoil pedological units were mostly characterized along the slope by the ranges of the fractal one-dimensional space domains and their scale invariants: the fractal dimensions. Results suggest that, within each topsoil pedological unit, these scale invariants are relevant qualifiers of the intrinsic topsoil variability, which can be modelled as a fractal Brownian process and should be incorporated in simple recursive or complex network soil erosion models. Different surface fractal dimensions, in a one-dimensional space, have been found within these three topsoil units for each soil parameter studied (multifractals). All the data have been aggregated within the whole one-dimensional slope transect in order to obtain both possible general scale laws on clays and other soil characteristics and possible evidence concerning underlying soil erosion mechanisms by particle runoff.