Accurate modeling of stem taper is essential for forest management decisions, including the definition of cutting cycles, the feasibility of annual harvesting, assortment classification, size and volume estimation, and ensuring sustainable production continuity. This study modeled the stem taper of Araucaria angustifolia (Bertol.) Kuntze stands in southern Brazil using Kozak’s variable-exponent model fitted with nonlinear mixed-effects techniques. Both fixed- and mixed-effects models showed high predictive performance, regardless of calibration. An unstructured (UN) covariance structure was required to reduce autocorrelation. The mixed-effects model improved predictive accuracy by up to 22%, achieved R2 values above 0.99 with RMSE < 0.74 cm, and significantly reduced residual autocorrelation in diameter estimates. The most effective calibration of random effects was achieved using diameter measurements taken at heights between 0.3 and 6.3 m above ground (approximately between 1.3% and 28.3% of the total height, considering the tallest tree as a reference). This research improves the accuracy of volume estimation and the definition of timber assortments for A. angustifolia, thereby supporting forest management decision-making in southern Brazil.
Costa et al. (Mon,) studied this question.