How many trees and samples are adequate for estimating wood-specific gravity across different tropical forests?

Abstract

Wood density (WD) is a key trait used to determine forest biomass and carbon stocks, but determining WD accurately is logistically demanding and expensive. These challenges also hamper comparisons across studies and different forest types, because sampling intensity within forests and within individual trees often vary across studies. We aimed to evaluate the relationship between WD and forest type using a standardized protocol and to simulate the number of samples required to obtain a representative estimation of WD of trees belonging to different tropical vegetation types representing an increasing order of aridity: rain forest, semideciduous forest, evergreen dry forest, savannah woodland, and seasonally deciduous forest. We measured WD at five vertical profiles along the trunks of 1,671 trees representing 349 species. Using bootstrapping analyses, we modeled WD as a function of the different combinations of samples extracted at the five sampling heights and evaluated the models with the best performance. The lowest and highest mean WD values were found in rain forest and seasonally deciduous forests, respectively, in line with the correspondingly low and high aridity of these habitats. Depending on forest type, sampling approximately 30–60 trees is sufficient for stabilizing the coefficient of variation in WD. Additionally, using samples collected at 25% and 50% height from the base along the vertical profile of each tree is adequate for WD estimations. These insights could be used to develop less destructive methodologies for wood density sampling, and, thus, help to reduce costs of carbon stock inventories in tropical forests.