No abstract

Background: In fast-growing forests such as Eucalyptus plantations, the correct determination of stand productivity is essential to aid decision making processes and ensure the efficiency of the wood supply chain. In the past decade, advances in remote sensing and computational methods have yielded new tools, techniques, and technologies that have led to improvements in forest management and forest productivity assessments. Our aim was to estimate and map the basal area and volume of Eucalyptus stands through the integration of forest inventory, remote sensing, parametric, and nonparametric methods of spatial prediction. Methods: This study was conducted in 20 5-year-old clonal stands (362 ha) of Eucalyptus urophylla S.T.Blake x Eucalyptus camaldulensis Dehnh. The stands are located in the northwest region of Minas Gerais state, Brazil. Basal area and volume data were obtained from forest inventory operations carried out in the field. Spectral data were collected from a Landsat 5 TM satellite image, composed of spectral bands and vegetation indices. Multiple linear regression (MLR), random forest (RF), support vector machine (SVM), and artificial neural network (ANN) methods were used for basal area and volume estimation. Using ordinary kriging, we spatialised the residuals generated by the spatial prediction methods for the correction of trends in the estimates and more detailing of the spatial behaviour of basal area and volume. Results: The ND54 index was the spectral variable that had the best correlation values with basal area (r = − 0.91) and volume (r = − 0.52) and was also the variable that most contributed to basal area and volume estimates by the MLR and RF methods. The RF algorithm presented smaller basal area and volume errors when compared to other machine learning algorithms and MLR. The addition of residual kriging in spatial prediction methods did not necessarily result in relative improvements in the estimations of these methods. Conclusions: Random forest was the best method of spatial prediction and mapping of basal area and volume in the study area. The combination of spatial prediction methods with residual kriging did not result in relative improvement of spatial prediction accuracy of basal area and volume in all methods assessed in this study, and there is not always a spatial dependency structure in the residuals of a spatial prediction method. The approaches used in this study provide a framework for integrating field and multispectral data, highlighting methods that greatly improve spatial prediction of basal area and volume estimation in Eucalyptus stands. This has potential to support fast growth plantation monitoring, offering options for a robust analysis of high-dimensional data.

No abstract

RESUMOObjetivou-se avaliar e aplicar as redes neurais artificias (RNA) para estimar o diâmetro (di) ao longo do fuste (hi) em um plantio de Eucalyptus sp., e com isso comparar com o método de regressão linear por um polinômio do quinto grau (PQG). Foram cubadas 74 árvores pelo método absoluto e destrutivo, coletando diâmetros (di) a várias alturas (hi). Para realizar a modelagem de predição dos diâmetros, a base de dados foi dividida em um conjunto de treinamento e outro de teste. O PQG foi ajustado por meio do software estatístico R considerando o método dos mínimos quadrados ordinários como procedimento de ajuste. As variáveis utilizadas para estimar os diâmetros (di) das árvores pelo método do PQG foram: Dap (diâmetro a 1,30m), di, hi e Ht (altura total). A RNA do tipo perceptron de múltiplas camadas foi implementada no software Scilab com o auxílio do ANN toolbox. As variáveis utilizadas para o treinamento da RNA foram obtidas por meio de combinações com diferentes operações matemáticas nos dados de Dap, hi e Ht. As estatísticas MAPE, MAD, MSD, RSME (%) foram aplicadas nos dados estimados com a finalidade de analisar os desvios frente aos dados observados e realizar a comparação entre os métodos utilizados. Considerando uma comparação aplicada dos métodos, foi calculado o volume e a quantificação do sortimento por árvore, usando como base as estimativas geradas. A RNA em geral apresentou boas estatísticas e um melhor gráfico residual. Conclui-se que ambas as metodologias mostraram-se eficientes para alcançar os objetivos propostos, assim as RNA podem ser consideradas como uma boa alternativa de uso e aplicação. PALAVRAS-CHAVE: Afilamento, Inteligência Artificial. Manejo florestal ENCICLOPÉDIA BIOSFERA, Centro Científico Conhecer -Goiânia, v.11 n.22; p. 2015 2420 FORECAST DIAMETERS ALONG THE STEM BY ARTIFICIAL NEURAL NETWORKSABSTRACT This study aimed apply and evaluate the artificial neural networks (ANN) to estimate the diameter (di) along the stem (hi) in a plantation of Eucalyptus sp., and after compare with the linear regression method for the fifth polynomial degree (QGP). Seventy four trees were cubed by the absolute and destructive method. The diameters (di) were collect at different heights (hi). To model the prediction of diameters, was necessary divided the database into a training set and other for test We adjusted the QGP and used the statistical software R considering the method of ordinary least squares as adjustment procedures. The variables used to model by regression QPG were: Dap (diameter 1,30m), di, hi and Ht (overall height). The ANN perceptron type multilayer was implemented in Scilab software with the help of ANN toolbox. The variables used for the training of ANN were obtained by an combinations with different mathematical operations on data Dap, hi and Ht.The statistics, MAP,MAD, MSD, RSME (%) were applied on the estimated data for the purpose of analyzing the deviations compared to the observed data and compare the methods used. Considering a apply comparison of the methods, the volum...

ABSTRACT:Modeling of the ecological niche of vegetal species is useful for understanding the species-environment relationship, for prediction of responses to climate changes and for correct reforestation programs and establishment of plantation's recommendation. The objective of this work was to establish a model for the distribution of four tree species (Casearia sylvestris, Copaifera langsdorffii, Croton floribundus and Tapirira guianensis), widely used in reforestation projects in the state of Minas Gerais, Brazil. In addition, we analyzed the relationship between environmental characteristics and the occurrence of species and tested the performance of Random Forest and Artificial Neural Networks as modeling methods. These methods were evaluated by their overall accuracy, sensitivity, specificity, Kappa, true skill statistic and the area under the receiver operating curve. The results showed the species Casearia sylvestris, Copaifera langsdorffii and Tapirira guianensis widely occurring in the state of Minas Gerais, including a broad range of environmental variables. Croton floribundus had restricted occurrence in the southern state, showing narrow environmental variation. The resulting algorithms demonstrated greater performance when modeling restricted geographic and environmental species, as well as species occurring with high prevalence in data. The algorithm Random Forest performed better for distribution modeling of all species, although the results varied for each metric and species. The maps generated had acceptable metrics and are supported by and ecological information obtained from other sources, constituting a useful tool to understand the ecology and biogeography of the target species. MODELAGEM DO NICHO ECOLÓGICOS DE ESPÉCIES ARBÓREAS EM UMA ÁREA TROPICAL BRASILEIRARESUMO: A modelagem de nicho ecológico de uma espécie é útil para a compreensão da relação espécie-ambiente, para a previsão do comportamento frente às alterações climáticas e para a indicação correta em reflorestamentos e estabelecimento de plantações. O objetivo foi modelar a distribuição de quatro espécies arbóreas amplamente utilizadas em projetos de reflorestamento no estado de Minas Gerais (Casearia sylvestris, Copaifera langsdorffii, Croton floribundus e Tapirira guianensis). Como complemento, o objetivo foi analisar a relação entre as características ambientais e a ocorrência de espécies e testar o desempenho das técnicas random forest e redes neurais artificiais como métodos de modelagem. Estes métodos foram avaliados pelas métricas de acurácia global, sensibilidade, especificidade, kappa, true skill statistic e área sob a curva. Verificou-se que as espécies Casearia sylvestris, Copaifera langsdorffii e Tapirira guianensis apresentaram ampla área de ocorrência no estado Minas Gerais, cobrindo ampla gama de variáveis ambientais. Já Croton floribundus demonstrou ocorrência restrita do sul do estado, mostrando estreita variação ambiental. Os resultados dos algoritmos demonstraram maior desempenho na modelagem de espécies geograf...