Oscar García
Forest Research Institute
Rotorua, New Zealand
Spatial correlation of tree diameters causes diameter distributions to vary with the extent of land considered. In particular, distributions derived from plot data may not be representative of the distribution for a whole stand or compartment. A review of the literature on spatial correlations in forests suggests that positive spatial correlations, attributable to slow changes in microsite (soil/topography) with distance, are common. Often the effect of microsite is overriden by competition, causing negative correlations at shorter distances. Predominantly positive correlations between the trees in a plot would cause the variance within a plot to be smaller than the variance for the stand, while negative correlations would have the opposite effect. It is shown that the stand variance can be expressed as a sum of two terms, one involving the within-plots variance, and the other the variance of the plot means. Numerous studies of the relationships between plot size and variance of plot means in forest inventories, have found that the variance decreases more slowly with size than what would be the case in a random sample. This indicates positive correlations over distances comparable to the size of the plots tested. Variance partitioning is used to derive a method for estimating stand variances from inventory data. Using this method in three radiata pine plantations, it was found that the plots diameter variance underestimated the variance of the stand distribution by 3, 15, and 20\%. Mapped data from an unthinned radiata pine stand, remeasured over a number of years, is used to illustrate short-range spatial correlations. Correlations are initially positive, indicating a microsite effect, but change to negative as competition intensifies. The implications of the area-dependency of tree size distributions for forest information systems and for growth model development are discussed. Although the practical impact on forest management decisions may not be large, at least the highly detailed data produced by distribution-based systems should be viewed with caution. Ignoring microsite correlations in individual-tree models can produce unrealistic and misleading results.