CLUSTER AND DISCRIMINANT ANALYSES FOR STEM VOLUME MODELLING OF TREE SPECIES GROUPS IN AN AMAZON RAINFOREST

Abstract

The diversity of species in native tropical forests causes difficulty in the interpretation of data that support their management and conservation. Species grouping, based on characteristics of interest, reduces significantly the number of volume equations and helps solve the problem of undersampling rare species. This study aims to group 32 Amazonian trees species of commercial interest based on regression coefficients of the Schumacher and Hall’s model and their fit statistics. To accomplish this, we employ a two-stage approach, in which we first applied cluster analysis to classify species with higher sampling intensity (n > 30). This phase allowed us to allocate poorly sampled species (n < 30) to groups created by discriminant analysis, resulting in the second stage. This proposed approach has proven adequate for grouping timber species in the Amazon forest, and so the stem volume can be modelled on consistent groups of species. The grouping of Amazon rainforest commercial species, based on the regression coefficients and fit statistics, performs better in aggregation for the stem volume modelling, providing stabilisation of estimation error and supplying few equations for the evaluation of standing stock.