Estimating neighborhood-based stand structure parameters using sampling simulation

This paper compares two sampling methods to assess forest stand structure, the,structural group of four" (VG) (fig. 1) and a modified cluster sampling (KG) (fig. 4). The spatial distribution of trees, small scale differences in species mingling, and size differentiation are used to describe stand structure. The target parameters of the inventory are, therefore, the mean uniform angle index (M) over bar, the mean mingling index ($) over bar, the mean measure of dominance (M) over bar, and the mean single tree basal area g (for definitions see fig. 1). As experience shows the estimates of basal area and number of trees per ha are biased when applying VG. We also show that stand structure parameters are not in every case independent of a tree's selection probability which is linked to the tree's growing space (tab. 1). Estimates of the target parameters without bias and a higher precision are expected from the new inventory method KG. Applying sampling simulation KG is compared to VG in six generated stands (fig. 2) and four trial stands (fig. 3). Statistical measures to judge the methods' performances are the frequency of estimates within a defined confidence region rho, the relative bias (eq. 1), and rRMSE (eq. 2). The sampling simulation results are presented in figures 5 to 8. KG has a clear advantage over VG in estimating the uniform angle index (fig. 5) and the measure of dominance (fig. 7) in the stands investigated. Both inventory methods show similar precision in estimating the mingling index (fig. 6). For estimating the mean single tree basal area (g) over bar with equal precision as VG the modified method KG is more time consuming (fig. 8). To enhance the efficiency of KG adjustments of the method by including diameter measurements of neighboring trees seem promising.