Sampling and stratification hypsometric relation data in forest inventories of savanna

Abstract

This study aimed at evaluating the application of linear hypsometric equations in savanna located in the state of Tocantins, Brazil. Data were obtained in 46 plots of 1,000 m² each, and 13 plots were used to simulate six different systems of data sampling to fit hypsometric models. The remaining 33 plots were reserved for an application test of generated hypsometric equations. In this assessment, to decide the sampling system and model, in addition to the residual graphic analysis, the following statistic criteria were adopted: residual standard error and multiple linear correlation. After this decision, a model identity test was applied to determine the best way to prepare the database aiming at the hypsometric models adjustment. We concluded that the linear functional relation: Ln(h)=f[Ln(N); Ln(dq/d); 1/d] fitted with data from the first 50 trees of each plot and, independently, from diameter class, was the most appropriate to characterize the hypsometric relation in savanna in Tocantins.