Abstract

A stem profile model was developed for black spruce (Picea mariana (Mill.) B.S.P.) trees in Alberta, Canada using a nonlinear mixed model approach. The model included two random parameters to capture between-subject variation and a general covariance structure to model within-subject residual autocorrelation. After evaluating various covariance structures, the 4-banded toeplitz and the spatial power structures were chosen for further evaluation. The 4-banded toeplitz structure provided a better fit. The model was further evaluated using an independent data set to examine its validation accuracy. Model validation results showed that the model was able to accurately predict stem diameters at the population and subject-specific levels. Both covariance structures produced reliable model predictions, but the spatial power structure was superior to the 4-banded toeplitz structure. One to four stem diameters were used to predict random parameters and to subsequently generate subject-specific predictions. At least three stem diameters were needed to achieve better subject-specific predictions than population-average predictions.

Keywords

nonlinear mixed model, black spruce, autocorrelation

Creative Commons License


This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Share

COinS
 
Apr 19th, 12:30 PM

A STEM PROFILE MODEL CALIBRATED BY NONLINEAR MIXED-EFFECTS MODELING

A stem profile model was developed for black spruce (Picea mariana (Mill.) B.S.P.) trees in Alberta, Canada using a nonlinear mixed model approach. The model included two random parameters to capture between-subject variation and a general covariance structure to model within-subject residual autocorrelation. After evaluating various covariance structures, the 4-banded toeplitz and the spatial power structures were chosen for further evaluation. The 4-banded toeplitz structure provided a better fit. The model was further evaluated using an independent data set to examine its validation accuracy. Model validation results showed that the model was able to accurately predict stem diameters at the population and subject-specific levels. Both covariance structures produced reliable model predictions, but the spatial power structure was superior to the 4-banded toeplitz structure. One to four stem diameters were used to predict random parameters and to subsequently generate subject-specific predictions. At least three stem diameters were needed to achieve better subject-specific predictions than population-average predictions.