Estimating tropical tree diversity indices from forestry surveys: A method to integrate taxonomic uncertainty

摘要: Analyses of tree diversity and community composition in tropical rain forests are usually based either on general herbarium data or a restricted number botanical plots. Despite their high taxonomic accuracy, both types difficult to extrapolate landscape scales. Meanwhile, forestry surveys provide quantitative occurrence large areas, thus increasingly used for landscape-scale analyses diversity. However, the reliability these approaches has been challenged because ambiguity common (vernacular) names by foresters complexity taxonomy those hyper-diverse communities. We developed tested novel approach evaluate propagate resulting uncertainty estimates several indicators (alpha beta entropy, Fisher-alpha Sorensen similarity). Our is Monte-Carlo processes that simulate communities taking into account expected accuracy names. this method French Guiana, 9 one-hectare plots (4279 trees - DBH ? 10 cm) which standardized determinations were available. then applied our simulation inventories (560 ha) at scale compared indices obtained sites with computed from precise determination situated same localities. found varied 22% (species level) 83% (family Amazonian region. Indices directly raw resulted incorrect values, except Gini-Simpson beta-diversity. On contrary, correction provides more accurate estimates, highly correlated measurements, almost all regional local robust ranking consistent shown inventories. These results show (i) represent significant part information, (ii) relative can be successfully ranked using inventory (iii) valuably contribute detection large-scale patterns when biases well-controlled corrected. The tools we as R-functions available supplementary material adapted parameters forest management conservation issues other contexts. (Resume d'auteur)