Abstract

Tree density, species richness, and composition drive vegetation patches identified from remotely-sensed data in a semi evergreen tropical forest. A proposal for characterizing habitat of forests, obtained from an object-oriented classification of RapidEye multiespectral imagery, based on dissimilarity matrices of vegetation structure, species diversity and composition is presented. The study area is a forested landscape mosaic after slash and burn agriculture (Ac: 8-23 years ago), selective logging (Fs: 43-53 years ago), and selective logging and forest fire (Fc: 21-28 years ago). The site is located in the central part of Quintana Roo, México, where three vegetation patches were delineated according to remotely sensed multiespectral imagery. Mean differences between vegetation structure properties of each vegetation patch were obtained through a permutational multivariate analysis of variance (P < 0.001). Species richness, stem density per hectare, and the axis-1 scores of the non-metric multidimensional scaling ordination of specific composition were identified as the vegetation attributes more relevant to differentiate the vegetation patches by a multinomial logistic model. Fc vegetation patch is characterized by the greatest mean values on Shannon-Wiener index, species richness, and stem density. The Fs has the greatest mean values of canopy height, basal area, and biomass at 80 percentile, and the Ac vegetation patch has the lowest values of all mentioned metrics. The species with the greatest relative importance value were: Ac: Bursera simaruba and Psidia psipula, Fs: Gymnanthes lucida and Manilkara zapota, Fc: G. lucida and B. simaruba. The uncertainty associated with the metrics assessed by vegetation patch was smaller than the uncertainty of the whole area, because of the efficient variability aggregation of the field data. We conclude that multiespectral information is a reliable tool for distinguishing vegetation patches with specific features, as stem density, specific composition, and species richness.

Keywords: RapidEye, image segmentation, relative importance value, permanova, multinomial model, tropical forest, habitat characterization