Estimativas do Volume de Árvores de Eucalipto com Diferentes Técnicas de Modelagem e Tamanho do Banco de Dados

• Ano de defesa: 2022
• Autor(a) principal: JEAN DE JESUS DA SILVA
• Orientador(a): Gileno Brito de Azevedo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Fundação Universidade Federal de Mato Grosso do Sul
• País: Brasil
• Palavras-chave em Português: Estimativas do volume, Eucalipto, Técnicas de modelagem, Banco de dados.
• Link de acesso: https://repositorio.ufms.br/handle/123456789/4732

The quantification of the volume of trees is a fundamental activity in the management of forest resources. However, it is a very expensive activity to obtain directly. As an alternative, there is the use of indirect methods such as regression models and artificial neural networks (ANNs). Therefore, the objective was to evaluate the precision of the estimates of the volume of eucalyptus trees using different modeling techniques and database size. The specific objectives were: (i) to assess whether the reduction in the number of trees used in volume modeling influences the accuracy of these estimates; (ii) assess whether the performance of ANNs is superior to that of regression models; and (iii) to assess whether the inclusion of categorical variables in the ANNs contributes to improving the estimates. The study data were obtained in eight eucalyptus clonal plantations (four clones x two rotations), implanted in the municipality of Ribas do Rio Pardo, State of Mato Grosso do Sul. 465 trees were rigorously cubed to obtain the volume. Of this total, 20% were used to validate the estimates and 80% to adjust the models and train the ANNs. To assess the effect of the size of the database on the accuracy of the estimates, training was performed with different fractions of the training data (10% to 100%). For each fraction of the database, the volumetric model of Schumacher and Hall was adjusted and 500 ANNs of the Multlayer Perceptron type were trained. By reducing the size of the database used to model the volume of eucalyptus trees, using regression models and ANNs, it was possible to maintain the precision of the estimates. Regardless of the size of the database, the accuracy of volume estimates generated by ANNs was slightly higher than that of regression models. The inclusion of qualitative explanatory variables in the ANNs provided slightly higher volume estimates than those that did not use these variables.