Regressão não linear quantílica para classificação de sítio em povoamentos de Eucalyptus sp.

Abstract

The productive capacity is conventionally estimated by anamorphic or polymorphic calculated site curves; however, these curves have some limitations. Due to these limitations, the objective of this work was to evaluate the creation of growth curves for site classification by the nonlinear quantile regression method and to compare them with the curves generated by the algebraic difference method. The database was collected in two mesoregions of the state of São Paulo (Paulista and Bauru Metropolitan Macro), with 460 measurements in 98 plots with distinct periods in Eucalyptus sp. The site curves were adjusted according to the Chapman-Richards model using the algebraic difference method for anamorphic and polymorphic curves and nonlinear quantum regression (RNLQ). The results of this work allowed to conclude that the site curves adjusted by the RNLQ method are superior to the curves generated by the algebraic difference method by graphical analysis of the residuals with smaller ranged of the errors and greater homogeneity and residual standard error with lower values in the classes in general.