A stem taper equation for Eucalyptus camaldulensis in northeast Nigeria

Abstract

Eucalyptus (Eucalyptus camaldulensis Dehnh.) is an important exotic species in northern Nigeria commonly used for poles and timber. Sustainable management of this resource would require quantifying its volume. Stem taper equations are one of the main and most efficient methods for estimating stem volume to any merchantable limit of a species. There is currently no taper equation for Eucalyptus species in Nigeria. Therefore, this study developed taper equations for E. camaldulensis in northern Nigeria. Data for this study were obtained from a private plantation in Jalingo Local Government Area, Taraba State, Nigeria. 68 trees were felled and sectioned into 1-m bolt across the stem to a merchantable limit of 5 cm, which were used as the fitting dataset. An additional 22 trees were felled and used to validate the taper equations for stem volume estimation. Seven taper equations were initially fitted to the dataset using nonlinear least squares. The best taper equation was then refitted using a nonlinear mixed-effects approach and calibrated using diameters of one to five sections from the butt end. The taper equations were numerically integrated to obtain the stem volume, which was compared with empirical volume equations. The result shows that the Kozak (Can J For Res 27(5): 619–629. 10.1139/x97-011, 1997) equation, which included eight parameters, provided the best fit for predicting section diameters for under and over bark. The mixed-effects taper equation (NLME-TE) explained most stem diameter variations in the fitting dataset (pseudo-R2: 0.986–0.987; RMSE: 0.547–0.578 cm) without substantial residual trends. The validation showed that the prediction accuracy of the integrated NLME-TE improved as the number of sectional diameter measurements increased, with at least a 35% reduction in volume estimate error. For practical implementation, two calibration sectional diameter measurements taken from the butt end per tree are recommended. This approach would reduce measurement effort and cost while improving model performance.