This research aimed at evaluating the adjustment of Schumacher and Hall volumetric model by different algorithms and the application of artificial neural networks to estimate the volume of wood of eucalyptus according to the diameter at breast height (DBH), total height (Ht) of the clone. For such, 21 scalings of stands of eucalyptus clones were used with DBH ranging from 4,5 to 28,3 cm and total height ranging from 6,6 to 33,8 m. The Schumacher and Hall volumetric model was adjusted linearly and nonlinearly with the following algorithms: Gauss-Newton, Quasi-Newton, Levenberg-Marquardt, Simplex, Hooke-Jeeves Pattern, Rosenbrock Pattern; Simplex, Hooke-Jeeves, and Rosenbrock, used simultaneously with the Quasi-Newton method and the principle of Maximum Likelihood. Different architectures and models (Multilayer Perceptron - MLP and Radial Basis Function - RBF) of artificial neural networks were tested and the networks that best represented the data were selected. Estimates of the volumes were evaluated by graphics of estimated volume according to the observed volume and by the L&O statistical test . It was concluded that the adjustment of the Schumacher and Hall model can be used in its linear form, with good representation and without presenting bias of the data; the Gauss-Newton, Quasi-Newton and Levenberg-Marquardt algorithms were effective in the adjustment of Schumacher and Hall volumetric model. The artificial neural networks showed good adequacy to the problem and are highly recommended to perform production prognoses of planted forests.
Volume equations; nonlinear regression; neural networks
