Content area
Full Text
ABSTRACT. We computed and compared the statistical properties of the estimators for the number of trees/ha (density) for fixed-radius plot and n-tree distance sampling. In forests with random spatial patterns, n-tree distance sampling density estimators are at least as precise as those of plot sampling if the fixed-radius plot size is less than the ratio of (n - 2) and the expected density, where n is the number of trees included at an n-tree location. A similar result holds for the clustered forest, where the ratio is multiplied by a factor involving a constant of heterogeneity. If the expected number of trees per plot and the plot sizes are the same for both the random and clustered spatial patterns, the variance of the plot sampling density estimator for the clustered pattern will always be greater than for that of the random spatial pattern. FOR. Sci. 48(1):1-6.
Key Words: Density-adapted sampling, spatial pattern.
N-TREE DISTANCE SAMPLING, referred to as "density-- adapted" sampling by Jonsson et al. (1992), provides an alternative forest inventory method to the familiar fixed-radius and variable-radius plot sampling methods. To conduct n-tree distance sampling, measurements are taken on the n trees nearest the sampling location. The distance from the plot center to the center of the nth tree forms the radius of a circular plot. Although the number of trees remains fixed for each sample, the distribution of the size of the circular plot varies with the spatial pattern of the trees.
A number of empirical studies have compared estimates obtained by n-tree distance sampling with those obtained by fixed-radius plot sampling (Jonsson et al. 1992, Lessard et al. 1994, 1995, Lynch and Rusydi 1999). Jonsson et al. (1992) found that bias and precision of n-tree distance sampling estimates were comparable with those from 10 m fixed-- radius plots (about 0.03 ha). Lessard et al. (1994) found that n-tree distance sampling was cost-competitive with the more traditional point and plot sampling in estimating the number of trees/ha for three forest types. Lessard et al. (1995) found that the basal area estimates were generally unbiased (within 2% of the true values) for point and plot sampling; n-tree distance sampling generally underestimated the true mean by 2-10% in nearly random and clustered stands and...