Optimum plot and sample sizes for carbon stock and biodiversity estimation in the lowland tropical forests of Papua New Guinea

Abstract

National forest inventories (NFIs) have to be properly designed in order to yield statistically representative data and make best use of limited resources. A key element to accomplish this is the knowledge of the local variability in the sampled forest stand features. In this study, we estimated the variability of tree taxonomic richness and carbon (C) stock in logged-over and unlogged lowland tropical forests of Papua New Guinea, to define the optimal plot and sample size needed to estimate these characteristics in the context of the forthcoming implementation of the new NFI. We used data from 133 one-hectare permanent sample plots to calculate the coefficient of variation (CV) of C stock and taxonomic richness at different simulated plot sizes. CV was seen to rapidly decrease with increasing plot size up to 0.2-0.3 ha for both features following an inverse-exponential trend. Optimum plot size ranged between 0.08 and 0.32 ha, with 75-164 plots needed to achieve an estimation within 5 per cent of the true mean (95 per cent confidence), depending on the stand feature and the previous silvicultural treatment. We concluded that the establishment of a network of 319 permanent sample plots between 0.2 and 0.3 ha in size would represent an efficient sampling scheme in lowland forests for the new NFI.