Redes neurais artificiais na intensidade amostral para estimativas de volume comercial de clones de Eucalyptus spp. em Paragominas-PA.

Abstract

The objective of this work was to evaluate the commercial volume of Eucalyptus spp. obtained by artificial neural networks at different sampling intensities in Paragominas-PA. 240 sample trees of Eucalyptus spp. Were felled (Fifteen trees for each clone and class of age) which were distributed in three diametric classes. Por the training and validation of the neural networks, was used de databasis of the procedure described above, evaluating the performance of the same in five sample intensities, which correspond to the realization of these processes with the use of the whole database, and the others, one, two, three and four trees, respectively, of each of the diametric classes. Through the methodology used, it was possible to verify that there was an acceptable performance of all trained networks. The use of only two trees in the training of neural neworks at the Sample Intensity 3, presented good values of r (0.992144), Variance (0.000189), bias (0.000094) and RMSE% (10.8446), which allows to infer that artificial neural networks can be used to reduce costs in forest inventory operations, since in this study they present satisfactory results with a reduced database.