Modeling wood fiber attributes using forest inventory and environmental data for Newfoundland’s boreal forest

• Models of fiber attributes were developed using an extensive database of measurements taken from tree cores.
• Akaike’s information criterion was used to select predictive models.
• The most influential variables were dependent on the species and fiber attributes being modeled.
• Results demonstrate potential to map fiber attributes for boreal forests in Newfoundland, Canada.

We explore the possibility of predicting wood fiber attributes across Newfoundland for two commercial species: black spruce (Picea mariana (Mill.) B.S.P.) and balsam fir (Abies balsamea (L.) Mill.). Estimates of key fiber attributes (including wood density, coarseness, fiber length, and modulus of elasticity) were derived from measurements of wood cores taken from sample plots representing a wide structural gradient of forest stands. Candidate models for predicting fiber attributes at plot and landscape scales were developed using an information-theoretical approach and compared based on Akaike’s information criterion. The most influential variables were stand age and the presence of precommercial thinning. Other significant explanatory variables included those that characterize vegetation structure (mean diameter at breast height, dominant height), climate (annual precipitation, mean temperature of the growing season), and geography (elevation, latitude) depending on the species and fiber attribute being modeled. At the plot level, model inference gave root mean square errors of 5.3–11.9% for all attributes. At the landscape level, prediction errors were similar (5.4–12.1%), with the added benefit of being suitable for mapping fiber attributes across the landscape. The results obtained demonstrate the potential for predicting and mapping fiber attributes over a large region of boreal forest in Newfoundland, Canada.