Modelagem não linear da relação hipsométrica e do crescimento das árvores dominantes e codominantes de Eucalyptus sp

Abstract

This work aimed to evaluate linear and nonlinear, classical and generalized models with the addition of covariates for modeling the hypsometric relation and the height growth of dominant and codominant trees of eucalypt clones. Two linear models and two nonlinear models were fitted to estimate the hypsometric relationship and four nonlinear models to classify the site. Regarding the hypsometry, it was used the technique of inclusion of covariates in the model that provided the best statistics in order to obtain more precise estimates. The selection and quality of the fittings was based on the standard error “Syx (%)”, Akaike information criterion (AIC), Bayesian information criterion (BIC) and the test of maximum likelihood (LRATIO), in addition to the graphical analysis of the residuals. For handling, fitting and processing the data, it was used software ‘R’. According to the statistical analysis of the models, for the hypsometric relation the four parameters logistic model was proved to be superior when compared to the other adjusted models. The logistic model obtained better accuracy compared to the classical model. In relation to the estimates of height growth in dominant and codominant trees, the three parameter logistic model obtained the best statistical precision, and therefore it was used for the construction of site index curves.