Polynomial models are most commonly used in Brazilian forestry for taper modeling due to its straightforwardly fitting and precision. The use of nonlinear regression classic models, like Gompertz, Logistic and Weibull, is not very common in Brazil. Therefore, this study aimed to verify the best nonlinear and linear models, and among these the best model to describe the longitudinal tree profile. The comparison measures were: R², syx, R²adjusted, residual graphics and fitting convergence. The results pointed out that among the non-linear models the best behavior, in general, was given by the Logistic model, although the Gompertz model was superior compared with the Weibull model in terms of residual standard error (syx). Among the linear models, the polynomial by Pires and Calegario proved to be better than the others. When comparing non-linear models with linear models, the Logistic model was better mainly because of the behavior of the data, the low correlation between the parameters and in meaning, facilitating convergence and adjustment.
Stochastics models; taper; Eucalyptus spp
