Moran's I statistic measures the spatial autocorrelation in a random v
ariable measured at discrete locations in space. Permutation procedure
s test the null hypothesis that the observed Moran's I value is no gre
ater than that expected by chance. The spatial autocorrelation of gros
s basal area increment is analyzed for undisturbed, naturally regenera
ted stands in three Georgia forest types: loblolly, shortleaf, and sla
sh pine. The analysis uses 0.4-ha permanent sample plots from a forest
inventory that included two remeasurement intervals (1961-1972 arid 1
972-1982). We present a new statistic for exploratory spatial analyses
, and this statistic revealed an anomalous duster of unusually slow-gr
owing shortleaf pine plots occurred in the mountains 100 km north of A
tlanta. A regression model was used to predict gross basal area increm
ent as a function of variables that describe local stand conditions, a
nd no significant spatial autocorrelations existed in the regression r
esiduals. This result suggests that the anomalous cluster of slow-grow
ing plots can be explained by the spatial distribution of local stand
conditions rather than spatial patterns of other possible causes such
as air pollution, although alternative interpretations are possible.