Heredity 77 (1996) 359—368 Received 25 September 1995 Spatial structure of genetic variability in natural stands of Fagus sylvatica L. (beech) in Italy STEFANO LEONARDI* & PAOLO MENOZZI Istituto di Ecologia, Università di Parma, Viale delle Scienze, 43100 Parma, Italy We report an autocorrelation study of 11 enzyme loci detected by starch gel electrophoresis in 14 populations over the Italian biogeographical range of beech (Fagus sylvatica L.). In line with previous studies of beech and other forest tree species a low level of spatial autocorrelation was detected. No correlation between the amount of microspatial structuring of genetic varia- bility in different populations and environmental (latitude, longitude, altitude), structural (mean and standard deviation of tree size) and genetic characteristics (mean expected hetero- zygosity, mean F1) was found. No significant differences in the amount of spatial structuring seem to exist among loci if low heterozygosity loci are excluded from the analysis. Keywords: allozymes, Fagus sylvatica, forest-tree populations, genetic structure, spatial autocorrelation. Introduction In forest tree populations most genetic variability estimated by enzyme electrophoresis is found at the intrapopulation level (Hamrick & Godt, 1990; Muona, 1990). This partitioning of genetic variation has been linked to life cycle characteristics (extended life span and long generation time) and mating system traits (pollination and seed dispersal by wind, high outcrossing rate) common to mid to high latitude forest tree species (Loveless & Hamrick, 1984). European beech (Fagus sylvatica L.) is no excep- tion to this general pattern (Comps et al., 1990; Müller-Starck et al., 1992). In a survey of 21 Italian populations of beech we confirmed this partitioning of genetic variability and also found that a substan- tial part of the variation is attributable to the within- subpopulation component (Leonardi & Menozzi, 1995). The study of spatial patterns in genetic variability could lead to new insights into the causes that deter- mine the observed partitioning of variation among and within populations. The introduction of auto- correlation techniques to the study of the spatial distribution of genetic variability (Sokal & Oden, 1978; Sokal & Wartenberg, 1983; Sokal et a!., 1989; Sokal & Oden, 1991) stimulated a number of inves- *Correspondence. 1996 The Genetical Society of Great Britain. 359 tigations in forest tree populations. The descriptive power of the technique, that does not depend on the scale of the spatial structure, has never been ques- tioned. A debate on the limits of autocorrelation in inferring from the observed spatial structure the evolutionary forces that generated it, set the useful- ness of the results from this approach within a real- istic framework (Slatkin & Arter, 1991a). Spatial genetic structure can be the result of the action of several evolutionary processes including isolation in small patches, limited pollen and seed dispersal and selection acting at the microhabitat level. Although it may be unrealistic to expect spatial autocorrelation to assess subtle differences among evolutionary processes, the technique seems quite appropriate to describe the main evolutionary forces affecting a population. Spatial structuring affecting most loci is reasonably interpreted as evidence of migration or processes related to gene flow, whereas significant correlograms for only a few genetic traits can be safely interpreted as a result of drift or selection (Sokal & Menozzi, 1982). Few studies on forest tree species have found significant spatial structuring that could be attrib- uted to some specific cause. In Tamarack (Larix lan- cina) two nearby populations showed striking differences in spatial structuring of alleles explained by the events that followed severe anthropogenic disturbance. One population, re-established by outside sources of seeds, showed random spatial