Modelagem do crescimento em nível de árvores individuais utilizando redes neurais e autômatos celulares

• Autor(a) principal: Castro, Renato Vinícius Oliveira
• Data de Publicação: 2011
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/3070
• topic: Eucalipto; Modelo de árvore individual; Autômatos celulares; Índice de competição; Eucalyptus; Modeling the individual trees; Cellular automata; Competition index; CNPQ::CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL::MANEJO FLORESTAL

Accurately model the stand growth and yield is important to improve the forest resource management. As the management has increase demanded more details about the forest stand growth and dynamics, especially those for the wood or multiproduct high quality production, diametric distribution model and individual trees model have been more demanded. Thus, this study aimed to model the eucalyptus growth and production, at the individual trees level, using regression, artificial neural networks (ANN) and Cellular Automata (CA). Data from permanent plots were used to equations adjust to predict the dendrometric variables (diameter and height) growth and the tree mortality probability, as well as for ANN training for these same variables. The estimates obtained by ANN were more accurate than those estimated by regression models. The best networks, herewith an CA architecture, were used to validate the model, using independent data of those used to adjust the regression models and ANN training. Projecting the structure of these independent plots (diameter, height and tree mortality) from 24 months to 72 months, it was observed that the ANN CA structure provided accurate estimates at the individual trees level, diameter classes and for area unit (hectares).