A Categorical Modelling Approach to Site and Growth of Eucalyptus Stands in Brazilian Eastern Amazon

Abstract

Theoretical framework: Site index cannot be generalized to different eucalyptus clonal stands, since each clone has a distinct growth and yield pattern, in which categorical variables may add site-specific effects to assess model's interregional variability. 
  
 Objective: This study aimed to assess the statistical performance of site index, as well as growth and yield models in different configurations adding categorical variables. 
  
 Method: The study was carried out in eucalyptus stands in Eastern Brazilian Amazon with three clones of different ages and a different number of trees. Traditional Schumacher’s site model was fitted with the addition of categorical clone variable. Beck-Della Bianca’s model was fitted by ordinary least squares (OLS) and two-stage least squares (2SLS), adding dominant height as site variable and including clone variable. 
  
 Results and discussion: Schumacher’s clone model presented lower standard estimate error (9.50%) and higher adjusted coefficient of determination (0.61), correcting the lack of normality and homoscedasticity. 2SLS was more accurate than OLS for Beck-Della Bianca’s model. This model validation resulted in root-mean-squared error of 2.82% and bias of 0.03%. 
  
 Research implications: Additive and multiplicative effects on site index resulted in polymorphism. Clone variable provided more parsimonious and accurate models to estimate site index and forest growth and yield, in which 2SLS was recommended for forest prognosis.