Geostatistics were used for describing and modelling spatial patterns of gene frequencies at eight allozyme loci in 93 populations of sessile oak well distributed over the natural range of the species. Spatial patterns were summarized by the variogram, a measure of gene frequency variance as a function of distance. They were of three kinds: stationary patterns, clines and random distributions of allele frequencies. The main directions of variation of gene frequencies over Europe were estimated using directional geostatistics. They were shown to reflect postglacial migration routes as well as a genetic divergence among populations from different glacial refugia. A method of spatial interpolation, called kriging, was used to draw synthetic maps for allele frequencies and for multilocus patterns of variation using two canonical variables. It appears from this study that geostatistics are useful tools for the description of complex patterns of genetic variation, and therefore also for the management of natural genetic resources in forest trees.