Models for predicting product recovery using selected tree characteristics of black spruce

Publisher Website
Google Scholar
Abstract
The artificial neural network (ANN) model and five traditional statistical regression models were used to predict four parameters of simulated product recovery (lumber volume, lumber value, chip volume, and total product value) from the stud mill simulation based on three basic tree characteristics of black spruce (i.e., diameter at breast height (DBH), tree height, and tree taper). The ANN model (i.e., the three-layer perceptron with error back-propagation algorithm) performed as well as or better than the five statistical regression models in terms of statistical criteria such as R2, root mean square error, and mean absolute error of predictions. The second-order polynomial with both DBH and tree height predicted the four product recoveries as accurately as the ANN model. This study showed that the ANN model, the second-order polynomial function, and the power function were suitable for the prediction of product recovery using the selected tree characteristics. The models developed in this study allow the estimation of the product recovery of individual trees and of a forest stand before it is harvested. It is evident that these models would be valuable tools for forest resource managers.