Análise comparativa de modelos segmentados e não segmentados na estimativa do afilamento de Khaya spp. | Comparative analysis of segmented and non-segmented models in the estimation of taper for Khaya spp.

Taper modeling is crucial for the accurate estimation of diameters along the tree stem, being essential for forest management and the sustainable use of timber resources. However, the lack of studies on taper models for the genus Khaya, known as African mahogany, especially in regions with water deficits in Brazil, limits the understanding of strategies for characterizing the stem profile of stands of this genus. This study aims to fill this gap by evaluating the fit of segmented and non-segmented taper models for African mahogany, contributing to the development of effective strategies in forest management of stands. The non-segmented models of Hradetzky, Schöepfer, and Kozak, as well as the segmented models of Clark, Max & Burkhart, and Parresol, were evaluated to estimate the diameter along the tree stems in a Khaya spp. stand. During the rigorous measurement of sample trees, diameters were measured at the following positions along the stem: 0.3 m, 0.7 m, 1.3 m, 2 m, and at 1 m intervals until the minimum diameter of 6 cm or the end of the commercial stem (insertion of the first branches of the crown). The precision of the models was evaluated using Pearson's correlation statistics and the root mean square error (RMSE). To determine the accuracy of the models, bias (D), standard deviation of the differences (SD), sum of squared relative residuals (SSRR), and residual percentage (RP) statistics were evaluated, and a ranking was performed considering the results of these statistics at each evaluated stem position. Additionally, the two one-sided test (TOST) procedure was applied to evaluate diameter estimates in the following stem portions: basal, intermediate, and tip. Under the conditions in which this research was conducted, both non-segmented and segmented models, except for Parresol, were accurate, with estimation errors below 10%. The best fit and precision results were observed in Hradetzky’s non-segmented model and Clark’s segmented model.

FAPEMIG - Fundação de Amparo à Pesquisa do Estado de Minas Gerais