Tree-Level Models for Predicting Lumber Volume Recovery of Black Spruce Using Selected Tree Characteristics
Four model forms (linear, polynomial, power, and exponential) and their extensions with different independent variables were studied for their predictive ability for lumber volume recovery from black spruce (Picea mariana [Mill.] B.S.P.) trees at two types of sawmills (a stud mill and a random length mill) using basic tree characteristics (dbh, tree height, and stem taper). Of the three tree characteristics, dbh was identified as having the most predictive ability for lumber volume recovery. We found that all models predicted lumber volume recovery more precisely for small-sized and medium-sized (≤23 cm) trees than for large-sized (>23 cm) trees. In terms of the selected statistical criteria such as adjusted R2, root-mean-square error (RMSE), predicted error sum of squares (PRESS), and prediction error, the best models for predicting lumber volume recovery were considered to be the following forms for both types of sawmills: (1) the second-order polynomial function with dbh alone, (2) the polynomial function with an interaction term of squared dbh and tree height, and (3) the power function with dbh and tree height. For both types of sawmills, the three final fitted model forms accounted for over 91% of the total variation. Model validation using independent data from a stud mill indicated that by using measured tree characteristics the three model forms were able to accurately and precisely predict lumber volume recovery, especially from small- and medium-sized trees.
No Supplementary Data
No Article Media