Biomass modeling in a mixed plantation of Pinus taeda L. and Pinus elliottii Engelm

Abstract

The objective of this study was to compare three approaches to fit models to estimate forest biomass in a mixed plantation of Pinus taeda and Pinus elliottii with 16 years of age. The data came from the biomass of 60 trees sampled, 30 trees of Pinus taeda and 30 trees of Pinus elliottii. The aerial biomass was estimated through the regression analysis (independent adjustment and simultaneous adjustment) and the artificial intelligence method with the nearest neighbor techniques. The models were selected and compared based on the quality of the statistical indicators: adjusted determination coefficient (Raj), standard error of the relative estimate (Syx%), Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Graphical analysis of residuals (%) and the ranking of the models. In the three approaches there were differences in total and component biomass estimates, being this difference associated with the heterogeneity of species (genetic variability by seminal origin), their components, and the lowest correlation of weight with diameter and height (allometric relationships). The technique of the nearest neighbor did not present satisfactory results, its use being recommended for a larger data base. Simultaneous adjustment was similar to the independent fitting method. However, the simultaneous equation has the advantage that when adding the biomass of the components, the result is compatible with the total biomass, which is more satisfactory for the estimation of total biomass.