Through the analysis of six enzyme systems (ACP, ALP, CAT, GOT, PPO and PO), with 18 polymorphic loci, allele frequencies of 24 alleles were calculated for 141 adult individuals of a natural population of Cryptocarya moschata (Brazilian nutmeg), from Carlos Botelho State Park (24º03'21" S, 47º59'36" W), São Miguel Arcanjo, SP, Brazil, which were sampled in an area ca. 647 ha. The spatial genetic structure was investigated by the spatial autocorrelation method, using Moran's I coefficients estimated for 10 distance classes. Additionally, for the same distance classes, coefficients r were calculated through a multivariate analysis. Both approaches were qualitatively similar, poiting out that 15 correlograms from Moran's I coefficients were significant at 95% probability level. The total correlogram from r coefficients indicated that there was a significant spatial structure for the distance class of 0 to 150 m. By subdividing this class into 10 m intervals, positive spatial autocorrelation was detected in the majority of classes, suggesting that clusters of C. moschata trees, which were shorter than 50 m apart, were constituted of individuals with greater genetic similarity and were probably related. However, the greatest genetic similarity occurred among individuals being 105 m apart, which could be related to the kind of seed dispersal promoted by Brachyteles arachnoides (Primates - Cebidae). Results showed that for an adequate sampling, aiming to detect genetic diversity, the collection of individuals separated by a distance of 750 m, or more, from each other, would be advisable.
Cryptocarya moschata; isozymes; Moran's I coefficients; multilocus analysis; spatial autocorrelation
