Sistema de equações aditivas e estimadores por razão de biomassa de Pinus spp. no sul do Brasil
Resumo
Resumo: A modelagem independente tem sido amplamente utilizada em estudos visando obter estimativas de biomassa florestal. Contudo, este tipo de modelagem não atende a premissa de aditividade em seus componentes. Nesse cenário, surge o presente estudo objetivando aplicar os procedimentos de modelagem simultânea e estimadores por razão na estimativa de biomassa de Pinus spp. no Sul do Brasil, sob a hipótese de que, em função das restrições impostas aos coeficientes nos procedimentos NSUR e razão, as equações resultantes serão aditivas contrariamente a modelagem clássica, resultando portanto em estimadores que melhor representem a condição biológica das árvores. Foram selecionadas 70 árvores de Pinus spp. distribuídas nos intervalos de idade (2 - 24 anos) e diâmetro a 1,30 m do solo (1,91 - 40,27 cm), buscando-se sempre selecionar indivíduos médios das condições locais. A amostragem deu-se na forma destrutiva onde foi realizada a quantificação da biomassa dos seguintes componentes: raízes, fuste, casca, galhos, acículas e total. Para cada componente e para o total, vários modelos não lineares foram ajustados pelo método dos mínimos quadrados tomando como variáveis independentes várias combinações das seguintes variáveis biométricas: diâmetro a 1,30 m do solo (dap), altura total (ht), diâmetro da copa (dc) e altura da copa (hc). Foram usados como critérios de seleção dos melhores modelos alométricos o R2 aj. (%), o Syx (kg), o CV (%), o AIC, o MAE (kg), o MPE (%) e análise gráfica dos resíduos. Com as equações selecionadas procedeu-se o ajuste simultâneo. Para a estimativa da biomassa de Pinus spp. por meio de estimadores por razão foi considerada a biomassa dos componentes (raízes, fuste, casca, galhos e acículas) e biomassa total como variáveis dependentes e o volume total das árvores estimado pela equação derivando do ajuste do modelo de Schumacher e Hall na sua forma linearizada como variável independente. O modelo da razão empregado para cada componente . Tendo sido detectada heteroscedasticidade, os respectivos modelos foram reajustados com ponderação por meio da modelagem da estrutura da variância, nos procedimentos de estimadores por regressão e estratificação em dois estágios (classes de dap e classes de complementados com análise de discriminantes, no procedimento de estimadores por razão. Numericamente observou-se pouca alteração das estatísticas de ajuste com aplicação da modelagem simultânea em detrimento da modelagem independente, porém, este procedimento trouxe consistência biológica nas estimativas ao garantir que a soma das estimativas da biomassa dos componentes corresponda à estimativa da biomassa total. Comparativamente aos estimadores WNSUR, os estimadores da razão melhoraram em quase todos os componentes de biomassa a exatidão das estimativas e asseguraram a premissa de aditividade dos componentes de forma natural. Palavras-chave: Ajuste independente. Ajuste simultâneo. Heteroscedasticidade. Aditividade. Abstract: Independent modeling has been widely used in studies aiming to obtain estimates of forest biomass. However, this type of modeling does not meet the premise of additivity in its components. In this scenario, the present study aims to apply the simultaneous modeling procedures and ratio estimators in the estimation of biomass of Pinus spp. in Southern Brasil, under the hypothesis that, due to the constraints imposed on the coefficients in the NSUR and ratio procedures, the resulting equations will be additive unlike classical modeling, thus resulting in estimators that best represent the biological condition of the trees. 70 trees of Pinus spp. distributed in age range (2 - 24 years) and diameter at 1.30 m height (1.91 - 40.27 cm) were selected, always seeking to select average individuals from local conditions. Sampling occurred in the destructive form where the biomass quantification of the following components was performed: roots, stem, bark, branches, leaves and total. For each component and total biomass, several nonlinear models were adjusted by the least squares method using as independent variables several combinations of the following biometric variables: dbh, total height (ht), crown diameter (dc) and crown height (hc). The following criteria were used to select the best models: R2 aj. (%), Syx (kg), CV (%), AIC, MAE (kg), MPE (%) and graphical analysis of residues. With the selected equations the simultaneous adjustment was carried out. For the estimation of the biomass of Pinus spp. by ratio estimators method was considered the biomass of the components (roots, stem, bark, branches and leaves) and total biomass as dependent variables and the total volume of trees estimated by derived from the Schumacher and Hall model fit in its linearized form as an independent variable. The ratio estimators model used for each component was . When heteroscedasticity was detected, the respective models were readjusted with weighting by modeling the variance structure in the procedures of regression estimators and stratification in two stages (classes of dbh and classes of ?? ?? ??) complemented with discriminant analysis in the procedure of ratio estimators. Numerically, there was little change in the adjustment statistics with simultaneous modeling, but this procedure brought biological consistency in the estimates by ensuring that the sum of the biomass estimates of the components corresponds to the total biomass estimate. Compared to WNSUR estimators, the ratio estimators improved in almost all biomass components the accuracy of the estimates and ensured the additive premise of the components in a natural way. Keywords: Independent adjustment. Simultaneous adjustment. Heteroscedasticity. Additivity.
Collections
- Dissertações [482]