Résumé
Recent studies have shown the promising potential of terrestrial laser scanner (TLS) for theestimation of tree-and plot-level biovolumes. However, accurate and precise conversion of these volumes into AGB remains a big challenge. Here, we examined the implications of vertical variations of WSG on AGB estimations thanks to the destructive sampling of 821 trees from 51 species distributed across the Congo basin forests. We performed a principal component analysis (PCA)on the WSG of wood samples collected vertically along trees to characterize WSG vertical patterns. We further quantified errors associated to the use of species-level WSG extracted from Global Wood Density database (GWSG) when converting tree volume to biomass, and developed a simple correction model accounting for vertical variations. The two first PCA axes captured 78% of the variance in samples WSG and revealed the existence of increasing, constant and decreasing patterns of variation from tree stump tobranches that were well conserved within species. Vertical variation patterns could be predicted from WSG from tree stump as well as from GWSG. Using GWSG values to convert volumes to mass led to errors ranging from -40.31% to 72.77% at the tree level andca. 9% across trees. A simple model based on GWSG andtree diameter at breast height allowed drastically reducing the error (ca. 1.5% across trees). This study is an important step towards the standardization of a reliable AGB estimation method from TLS technology