Ecosite-based predictive modeling of black spruce (Picea mariana) wood quality attributes in boreal Ontario

Abstract

Enhanced forest resources inventory systems delineate and define polygons based on fundamental ecological units such as ecosites, which are standard combinations of vegetation and substrate types. Our study objective was to model wood quality characteristics of individual black spruce (Picea mariana (Mill.) B.S.P.) trees across a representative boreal forest landscape in northeastern Ontario, Canada, based on relationships to ecosite and other stand-level variables. A total of 127 large (12 mm) increment core samples were extracted at breast height from dominant or co-dominant black spruce trees in forest stands representing a gradient from dry sandy to wet mineral and organic ecosites. Sample cores were prepared, processed, and analyzed using standard SilviScan protocols. Hierarchical classification models were then fitted using Random Forests to predict density and latewood percentage for black spruce stems at a reference age of 50 years. These models each explained over 32% of variance, with estimated root mean squared errors of 40.4 kg·m−3 and 5.6% for density and latewood percentage, respectively. Among tree-, site-, and stand-level covariates, ecosite group was the most important predictive variable. Knowledge of ecosite – wood quality relationships could support efficient planning for black spruce management by including an indication of potential use as a modeled variable in a forest inventory system.