Estratificação de florestas de eucalipto com base na forma do fuste das árvores

• Ano de defesa: 2014
• Autor(a) principal: Santos, Jeangelis Silva
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Espírito Santo; BR; Mestrado em Ciências Florestais; Centro de Ciências Agrárias e Engenharias; UFES; Programa de Pós-Graduação em Ciências Florestais
• Palavras-chave em Português: Artificial neural network; Cluster sampling; Forest inventory; Levantamentos florestais; Amostragem (Estatística); Análise por agrupamento; Inteligência artificial; Redes neurais (Computação); Mensuração florestal; Amostragem estratificada; Inventário florestal; Recursos florestais e engenharia florestal; 630
• Link de acesso: http://repositorio.ufes.br/handle/10/5051

The forest inventory is the main method to obtain quantitative and qualitative information on forests. However, when working with large areas, there is the inconvenience of the variables present great heterogeneity, being necessary to adopt a higher sampling intensity. In these cases, an alternative for the realization of forest inventories is the stratification of the area in more homogeneous subpopulations as the variable of interest, ensuring more accurate estimates with a lower sampling intensity. This study aimed to stratify eucalyptus forests considering variables that describe bole form. For this purpose, we used a database containing information of 47.770 ha with clonal Eucalyptus stands. The stands consisted of fourteen clones with three different management regimes (high forest, divided into areas of first and second rotations, and coppice) and four spacings (6, 9, 10 and 16 m2 per plant), aged four to six years. Initially the area stratification was performed, yielding forty strata, in which were performed the scaling and forest inventories. Then, were applied the clustering methods of profile similarity, principal component analysis, class of form quotient, class of form factor and artificial neural networks, generating new sampling strata. For comparison, were also considered sampling without stratification, the complete stratification (40 initial strata), stratification considering the age and spacing and stratification by age only. Then was conducted the calculation of population estimators for forest inventory considering each stratification method presented, as well as the cost of conducting a forest inventory and scaling. Among the methods proposed to stratify the stands, the ones that showed the best results in accuracy, was the clustering by artificial neural networks and clustering by class of form quotient (K0,5H). Regarding costs, the clustering method by artificial neural networks has also achieved best results, followed by clustering by profile similarity method. 8 By analyzing precision and cost, among all methods, the use of artificial neural networks proved to be the most efficient alternative to the stratification of forests