HOW TO ESTIMATE BLACK WATTLE ABOVEGROUND BIOMASS FROM HETEROSCEDASTIC DATA?

Abstract

This study developed a system of equations for estimating total aboveground and component biomass in Acacia mearnsii De Wild. We analyzed 140 individuals of approximately 10 years of age, measuring the following data: diameter at 1.30 m height (d), total height (h), basic wood density (branches and stem), and biomass (crown, stem, and total). We tested linear and nonlinear allometric models; model selection statistics were R²adj., Syx%, and BIC. We selected nonlinear models to estimate component biomass, using d as an independent variable for crown biomass, and d and h for stem and total biomass. Adding basic density did not significantly improve biomass modeling. The residuals had non-homogeneous variance; thus, the adjusted equations were weighted, with weights derived from a function containing the same independent variables of the adjusted biomass function. Subsequently, we used simultaneous adjustment of equations to ensure that the sum of each component's estimated biomass values was equal to the total biomass values. Simultaneous adjustment improved the performance of the equations by guaranteeing the components' additivity, and weighted regression allowed to stabilize error variance, ensuring the homoscedasticity of the residuals.