Evaluation of Bitterlich angular count sampling in different spatial patterns and diametric distributions by the means of simulations

Abstract

The precision and accuracy of Bitterlich angular count (BAC) for the estimation of basal area density could depend on diametric distribution since the inclusion probability of the trees increases with their basal area. Trees spatial pattern can also influences since point-tree distances determine their inclusion in the sample. In this study, the precision and bias of BAC under different diametric distribution and spatial patterns were evaluated by the means of simulations. Normal and lognormal diametric distributions were simulated with basal area density ranging between 30-33 m² ha^-1. These diametric distributions were simulated in three spatial patterns: regular, random and clumped yielding six combinations. A seventh class was simulated with bimodal diametric distribution and a mixed spatial pattern: clumped for small trees and random for large trees. Bias was less than 0.10 m² ha^-1 except for clumped pattern with lognormal diametric distribution and mixed pattern. In these cases bias tended to be negative (-0.15 a -0.25 m² ha^-1). The lowest mean absolute deviation (MAD) was observed for the regular pattern 3-7 %. In the clumped pattern it ranged 7-10 % and 8-11 % in the mixed. MAD tended to be slightly lower in the normal distribution than in lognormal in the regular spatial pattern but the reverse trend was observed in the clumped pattern. In practical terms BAC yields null bias and low MAD with a low intensity sampling for a wide variety of stand structures.