A comparison of methods for edge-bias compensation
Ignoring the effects of off-plot trees on variables measured within a plot boundary may result in the under estimation problem known as edge bias. A number of edge bias compensation techniques have been proposed in the literature. Four of these were compared with the alternative of ignoring off-plot trees to determine their relative adequacy in modeling crown closure from individual tree crown measurements. Data from a spacing trial of loblolly pine (Pinus taeda L.) were used to make the comparisons. By shrinking the effective size of experimental plots, measurements of "off-plot" data were made available to compare with the results of edge-bias compensation models. Three edge-bias compensation algorithms were found to perform equivalently well: translation; reflection via a reflecting line through the edge trees; and a random arrangement of interior trees around the plot. The ability of the models to compensate for edge bias declined with stand age. Furthermore,the variability of compensation values increased with age and as plot size was reduced.