Optimal tree sampling for ecosystem-specific biomass allometry modeling in Congo Basin forests

Abstract

Allometric equations are fundamental for estimating biomass in forests and their accuracy depends heavily on the quality and  representativeness of the data used to construct them. This study aimed to benchmark tree sampling techniques and determine the  optimal number of sample trees for constructing allometric equations. Ten sampling strategies consisting of the combination of two  allometric models and five sampling techniques were evaluated. Random sampling techniques and four sampling techniques with eight  diameter size-classes based on cumulative frequency distribution were compared. A wide range of sample data was simulated using a parametric resampling method to ensure unbiased sampling and a representative spread of observations. Data were derived from 15  inventory plots in three Congo Basin forest reserves. Results showed that uncertainty due to differences in size class distribution was  minimized by a sampling technique, which effectively represents large trees. High sample sizes were required for precision in the  absence of large trees. Sample sizes uncertainty was influenced by stand characteristics, mainly the shape of the inventory plot and data  distribution. This study reveals that the biomass prediction uncertainty depends on the population’s specific characteristics, the type of   allometric model used, and the representativeness of large trees in the sample.