Comparison of three modern methods for estimating volume of sample trees using one or two diameter measurements

Abstract

Seven variations of the centroid, importance and control-variate methods for estimating bole volume were compared using four sample tree data sets, viz. Pinus ponderosa (186 trees), P. radiata (114 trees), P. taeda (4578 trees), and American mixed hardwoods (538 trees). The centroid method was the easiest to apply. Generally, it gave the most precise but also the most biased results (the biases were not severe, ranging from −0.2 to −4.1%). In contrast, methods involving importance sampling using either one or two random points, or a random point and an antithetic point, were unbiased (except for P. taeda) but generally much less precise. Invariably, the precision of the estimate was improved using two random points. Replacing one random point by an antithetic point generally improved the precision further but replacement by the centroid point generally introduced bias. The control-variate methods using one or two random points gave unbiased but imprecise estimates, the precision being better for two random points than one, as for importance sampling.