Amostras virtuais no monitoramento da produção florestal
Carregando...
Arquivos
Data
2018-02-23
Autores
Lima, Natália da Silva
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
A elevada produtividade do eucalipto ocasiona a necessidade do domínio de técnicas e ferramentas precisas acerca dos povoamentos florestais, como o inventário florestal, procedimento fundamental para o monitoramento da produção que permite o conhecimento do máximo potencial das florestas, além disso, é essencial para o planejamento de atividades de corte, colheita e suprimento de madeira. As técnicas convencionais para o levantamento da produção florestal utilizam-se de métodos tradicionais de estatística (estatística clássica), considerando apenas a adoção de valores médios para a tomada de decisões, não explorando as correlações espaciais que possam existir entre as parcelas amostrais. Então, um tipo de interpolador geoestatístico que permite definir a estrutura de dependência espacial dos dados é a krigagem, que juntamente com a metodologia de amostras virtuais, pode-se tornar uma alternativa viável para obtenção de um variograma com modelagem aceitável. Desta forma, o objetivo geral deste estudo foi empregar a metodologia de amostras virtuais no planejamento da malha amostral do inventário florestal, a fim de possibilitar a determinação de um número de parcelas que possam ser estabelecidas na área para obtenção de produtividade, sem que haja perda de precisão ou aumento de custos. A área de estudo abrange uma área plantada de 287,66 ha, pertencente a empresa Eucatex S/A, localizada em Itatinga, no estado de São Paulo. Os dados dendrométricos de volume (m³.ha-1) foram obtidos por meio de inventários realizados pela própria empresa, utilizando-se 32 parcelas retangulares permanentes, para os inventários florestais contínuos (IFC) e 32 parcelas retangulares permanentes mais 66 temporárias, para o inventário pré-corte (IPC). Os seguintes passos foram realizados com os dados de IFC para espacialização dos dados originais, no programa ArcGIS: aplicação de uma estatística univariada sobre a variável volume/ha; geração do variograma experimental; ajuste do variograma; validação do modelo teórico; aplicação da krigagem; geração de mapas de produtividade ano a ano. Posteriormente, 47 amostras virtuais foram inseridas na malha amostral dos inventários contínuos, considerando apenas as parcelas permanentes (32 parcelas), sendo que 10 parcelas foram selecionadas previamente para a validação externa; a análise geoestatistica se deu com as amostras virtuais mais as amostras permanentes restantes, totalizando 69 parcelas. Comparando-se os métodos, observou-se pela validação cruzada que o erro médio e a correlação entre valores observados e estimados melhoraram adicionando-se amostras virtuais, possibilitando um melhor ajuste do variograma e obtenção de estimativas favoráveis. A comprovação da eficácia da metodologia de inserção das amostras, chamada de validação externa, deu-se pelo cálculo da média do erro médio, que foi igual a 6,8% em volume, em relação ao inventário real. Desta forma, a técnica de inserção de amostras virtuais pode ser utilizada, mostrando-se satisfatória para o planejamento da próxima malha para posteriores análises geoestatísticas, em plantios equiâneos.
The high productivity of eucalyptus trees makes it necessary to master accurate techniques and tools for forest, such as the forest inventory, a fundamental procedure for monitoring production that allows the knowledge of the maximum potential of forests, and is essential for planning of harvesting and supply of timber. Conventional techniques for surveying forest production use traditional statistical methods (classical statistics), considering only the adoption of average values for decision making, not exploring the spatial correlations that may exist between the sample plots. So, a kind of geostatistical interpolator that allows defining the structure of spatial dependence of data is kriging, which together with the methodology of virtual samples, can become a viable alternative to obtain a variogram with acceptable modeling. In this way, the aim of this study was to use the methodology of virtual samples in the planning of the sampling of the forest inventory, in order to allow the determination of a number of parcels that can be established in the area to obtain productivity, without losing accuracy or increasing costs. The study area covers a planted area of 287.66 ha, belonging to Eucatex S / A, located in Itatinga, in the state of São Paulo. The volume dendrometric data (m³.ha-1) were obtained through inventories carried out by the company and were estimated in 32 permanent rectangular plots for continuous forest inventories (CFI) and 32 permanent rectangular plots plus 66 temporary plots for the pre-cut inventory (PCI). The following steps were performed with the CFI data for spatialization of the original data in the ArcGIS program: application of an univariate statistic on the variable volume. ha-1; generation of the experimental variogram; variogram adjustment; validation of the theoretical model; application of kriging; generation of productivity maps year by year. Subsequently, 47 virtual samples were inserted into the sampling network of continuous inventories, considering only the permanent plots (32 plots), 10 plots were previously selected for external validation; the geostatistical analysis occurred with the virtual samples plus the remaining permanent samples, totaling 69 plots. Comparing the methods, it was observed through cross-validation that the mean error and the correlation between observed and estimated values were improved by adding the virtual samples, allowing better adjustment of the variogram and obtaining favorable estimates. The validity of the methodology of insertion of the samples, called external validation, was verified by means of the average error, which was 6.8% in volume, in relation to the original forest inventory. In this way, the technique of insertion of virtual samples can be used, proving to be satisfactory for the planning of the next mesh for later geostatistical analysis, in plantations that have the same age.
The high productivity of eucalyptus trees makes it necessary to master accurate techniques and tools for forest, such as the forest inventory, a fundamental procedure for monitoring production that allows the knowledge of the maximum potential of forests, and is essential for planning of harvesting and supply of timber. Conventional techniques for surveying forest production use traditional statistical methods (classical statistics), considering only the adoption of average values for decision making, not exploring the spatial correlations that may exist between the sample plots. So, a kind of geostatistical interpolator that allows defining the structure of spatial dependence of data is kriging, which together with the methodology of virtual samples, can become a viable alternative to obtain a variogram with acceptable modeling. In this way, the aim of this study was to use the methodology of virtual samples in the planning of the sampling of the forest inventory, in order to allow the determination of a number of parcels that can be established in the area to obtain productivity, without losing accuracy or increasing costs. The study area covers a planted area of 287.66 ha, belonging to Eucatex S / A, located in Itatinga, in the state of São Paulo. The volume dendrometric data (m³.ha-1) were obtained through inventories carried out by the company and were estimated in 32 permanent rectangular plots for continuous forest inventories (CFI) and 32 permanent rectangular plots plus 66 temporary plots for the pre-cut inventory (PCI). The following steps were performed with the CFI data for spatialization of the original data in the ArcGIS program: application of an univariate statistic on the variable volume. ha-1; generation of the experimental variogram; variogram adjustment; validation of the theoretical model; application of kriging; generation of productivity maps year by year. Subsequently, 47 virtual samples were inserted into the sampling network of continuous inventories, considering only the permanent plots (32 plots), 10 plots were previously selected for external validation; the geostatistical analysis occurred with the virtual samples plus the remaining permanent samples, totaling 69 plots. Comparing the methods, it was observed through cross-validation that the mean error and the correlation between observed and estimated values were improved by adding the virtual samples, allowing better adjustment of the variogram and obtaining favorable estimates. The validity of the methodology of insertion of the samples, called external validation, was verified by means of the average error, which was 6.8% in volume, in relation to the original forest inventory. In this way, the technique of insertion of virtual samples can be used, proving to be satisfactory for the planning of the next mesh for later geostatistical analysis, in plantations that have the same age.
Descrição
Palavras-chave
Inventário florestal, Geoestatística, Produtividade florestal, Forest inventory, Geostatistical, Forest productivity