ESTIMAÇÃO DA ALTURA TOTAL DE ÁRVORES DE IPÊ FELPUDO UTILIZANDO MODELOS DE REGRESSÃO E REDES NEURAIS

The aim of this work was to evaluate the accuracy of total height estimation with regression models and artificial neural networks (ANNs) in a pure ipe felpudo plantation. We adjusted four regression models and trained Multilayer Perceptron artificial neural networks, with two training criteria: considering the input variables DBH, age and dominant height, and only DBH. The ANNs and regression models were evaluated by the statistical analysis of bias (V), Mean of Absolute Differences (MD), Root Mean Square Error (RMSE) and Correlation Coefficient (r). The estimation of the total height of the trees with generic regression models and ANN with covariates presented a greater accuracy when compared to the regression models and ANN with DBH only.