Modelagem de crescimento e produção de povoamentos de eucalipto empregando mínimos quadrados generalizados

• Ano de defesa: 2012
• Autor(a) principal: Oliveira, Julianne de Castro
• Orientador(a): Ribeiro, Carlos Antônio Alvares Soares
• Banca de defesa: Nogueira, Gilciano Saraiva
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Mestrado em Ciência Florestal
• Departamento: Manejo Florestal; Meio Ambiente e Conservação da Natureza; Silvicultura; Tecnologia e Utilização de
• País: BR
• Palavras-chave em Português: Heterocedasticidade; Autocorrelação; Modelo de Clutter
• Palavras-chave em Inglês: Heterocedasticity; Autocorrelation; Clutter model
• Área do conhecimento CNPq: CNPQ::CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL::MANEJO FLORESTAL
• Link de acesso: http://locus.ufv.br/handle/123456789/3111

This study aimed to evaluate the efficiency and applicability of growth and yieldstatistical modeling of eucalyptus (Eucalyptus sp.)plantations,in northern region of Minas Gerais state,by generalized least squares. Dendrometric variables were collected from 2698 permanent plots in continuous forest inventories. First, were tested Logistic and Gompertz models for dominant height modelling and three alternatives of Clutter model, both by ordinary least squares.The analysis was based on Adjusted R-square (R2ajust.), Root Mean Square Error (RMSE), averagerate error prediction (bias%), sign and significanceof parameters and graphical errors distributions. For the alternative Clutter models were also performed tests for homogeneity of variance, residual autocorrelation and normality of residuals. The logistic model was chosen to model the dominant height. For the alternative model of growth in volume and basal area, the dominant height observed variable can be used to replace the site index in basal area equation and results in improvements. There was no significant difference in the technical harvest's age between the three alternatives. However, were proven violations of statistical assumptions of homoscedasticity and residuals' independence, undermining the statistical consistency of the modeling. Subsequently, were compared the fittingsbetween ordinary least squares (OLS), generalized least squares (GLS), generalized least squares considering variance heterogeneity (GLS_W) and generalized least squares considering heterogeneity and residual autocorrelation (GLS_WC). The analysis was made by RMSE, bias%, sign and significance of parameters, graphical errors distribution, technical harvest s age and their productivities. The adjust methods influence was verified by volumetric prognosis of Increments Annual Average (IAA) with 6 and 7 years.The generalized least squares adjustments were compared by the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and test of maximum likelihood ratio (TMLR). Low values of AIC and BIC and significancy of TMLR demonstrated superiority of Logistic model with generalized least squares adjustment considering the heterocedasticity and autocorrelation. For alternatives of Clutter model, it was demonstrated that the inconsistency α1parameter with OLSalso binds to the occurrence of residual autocorrelation. The frequency distribution in the error classes ±10% and ±20% to MQG_WC was substantially better than the OLS, for alternatives 1 and 2. Regarding bias% RQEM, AIC, BIC and significancy of TRMV the alternatives were more efficient by using the MQG_WC method. The lack of difference between the forecasting methods showed that the present modeling did not resulted in bias due the statistical assumptions, but reaffirms the importance of using the method more efficient statistically how the MQG_WC.