The study of spatial patterns is a helpful approach to formulating hypotheses about plant population dynamics. Our aim was to search for variation in spatial patterns of Q. grandiflora height classes at different scales within and among fragments of different cerrado physiognomies. Spatial patterns of the number of plants in subplots (5 x 5 m) were described and compared among six 0.5-ha plots placed in four fragments. We performed Moran's I correlogram analysis to describe spatial patterns and tested spatial correlations of height classes by partial Mantel tests. We found 18 to 319 plants 0.5 ha-1. Larger plants (height >1.5 m) had a clumped pattern in all plots, varying in numbers and magnitudes of scales independently of the physiognomy. Correlogram analysis for smaller plants was possible in only three plots, where patterns were random. In two of these cases there was significant positive correlation of size classes but in the third it was not significant. Abundance and spatial patterns were as different within the same fragment as in different ones. These results show that Q. grandiflora spatial patterns cannot be predicted based on overall characteristics of cerrado physiognomy. The random-to-clumped transition with size and the spatial association of classes suggest that density dependent mortality does not seem to be an important driver of Q. grandiflora dynamics, and that spatial structure of suitable sites for recruitment to maturity may be more important.
correlogram; density dependence; partial Mantel test; recruitment; savanna
