Emprego da análise discriminante de Fischer para classificar tipos de florestas no bioma Pampa, RS, Brasil

Abstract

Fisher’s discriminant analysis (FDA) is a dimension reduction method which finds an optimal linear transformation that maximizes the independents variables between-class scatter and minimize the within-class scatter. Was used FDA to improve an existing classification scheme by identifying those structural vegetation atributes that can jointly differentiate among a set of forest in the Pampa biome, southern of Brazil. The study were conducted in data base on 5 ha tree-plots of Forest’s Inventory of Rio Grande do Sul State. Before analysis all premisses for analysis was garanted. In forward stepwise analysis all variables contributes for created the Fisher discriminant functions (FDF). The FDF1 accounts 74,7% of the discriminating ability of the discriminanting variables and FDF2 accounts 14%. The canonical correlation associated with the two function was > 90. In FDF1 implying a relation with atributes: species richness, commercial height and total height of trees. The FDF2 the relationship are with atributes basal área and maximum DBH in plot. All of 50 cases werw successfully classificated by FDA. These methodology and results should provide useful information for studies conducted in the classification forests in Rio Grande do Sul State and may be applicable to classification forests the other biomes of Brazil with properties measurement issues comparable to those Invetory.