Distance sampling for forest inventory in Indonesian teak plantations

Abstract

Distance sampling techniques were compared to point (variable plot) sampling and fixed-radius circular plot sampling for inventory of teak ( Tectona grandis) plantations in Indonesia. n-tree distance sampling selects the n trees closest to a sample point located in the field. The distance to the nth closest tree is used as a plot radius. Two n-tree estimators were compared. Prodan (1968)described a technique for n=6, which counts the nth tree as 1/2 for tree counts and tree volume (or other tree attributes). Another approach, here called `the ( n−1)/ n adjusted estimator', uses the bias correction factor ( n−1)/ n. Eberhardt (1967)demonstrated that the ( n−1)/ n adjusted estimator is unbiased for Poisson (random) and negative binomial (clumped) spatial distributions. No proof of unbiasedness is known for either method in uniform distributions which would be characteristic of plantations. n-tree sampling was performed in Indonesian teak plantations for values of n ranging from 3 to 10 for both methods. At the same locations, point (variable plot) samples with metric basal area factors 1 m 2 ha −1 tree −1 and 2 m 2 ha −1 tree −1 and fixed-radius circular plots 0.1 ha in size were installed for comparison with the n-tree distance methods. The time required for each sampling method was recorded at each field location. For each method, an efficiency ratio was calculated as (C 2 k× Time k)/(C 2 0.1 ha × Time 0.1 ha ) where C k is the coefficient of variation for method k, Time k is the average time required for method k, C 0.1 ha is the coefficient of variation for 0.1 ha plots and Time 0.1 ha is the average time required for 0.1 ha plots. The more efficient methods have smaller efficiency ratios. According to this criterion, 5-tree distance sampling using Prodan's estimator was most efficient of the techniques compared that did not show bias for these data. The ( n−1)/ n adjusted estimator showed bias for these trials in teak plantations. This was probably due to the uniform spatial distribution characteristic of plantations. Point sampling with BAF=1 m 2 ha −1 tree −1 and BAF=2 m 2 ha −1 tree −1 and 0.1 ha plot sampling were not as efficient as n-tree sampling methods for these trials.