Evaluating the Optimal Design for Kozak’s Tree Taper Model

Tree taper models are essential tools for forest managers, providing accurate information on current and future growing stock. However, there is no universally standardized design procedure for collecting tree taper data exist, leading to the use of suboptimal designs in practice. This study evaluates the performance of the Ds-optimal design proposed by Berhe and Arnoldsson (2011) against two ordinary designs by Berhe and Arnoldsson (2008) and Leites and Robinson (2004). Data were collected from 70 Cupressus lusitanica trees at Wondo Genet College of Forestry and Natural Resources (WGCFNR), Ethiopia, using the three taper designs. Kozak’s (1988) taper model was fitted to the datasets, and performance indicators—coefficient of determination (R²), bias (B), standard error of estimate (SEE), mean absolute error (MAE), and prediction error sum of squares (PRESS)—were used for comparison. Results indicated that all three designs yielded similar performance for the Kozak model. Notably, the Ds-optimal design proved to be the most efficient, reducing the number of required measurements by nearly half compared to ordinary designs without compromising model accuracy. This study recommends adopting optimal designs for tree taper models and other forestry applications to enhance data collection efficiency.