Identification of boreal mixedwood forest structure cohorts in Northwestern Ontario using Ontario's Forest Resource Inventory, Abitibi-Bowater's Continuous Forest Inventory and stepwise discriminant function analysis

A model was developed to determine if Ontario's Forest Resource Inventory (FRI) could be used to classify boreal mixedwood stands in northwestern Ontario into structural cohorts as proposed in the Multi Cohort Forest Management (MCFM) concept. Successful methods developed to-date for determining cohort status rely on Weibull functions of tree diameter distributions. This study proposed an alternative method for multi-cohort classification based on the range of stand attributes found in the FRI. Weibull classified stands were analysed using FRI data to determine if a new model of cohort classification was possible. Results from this analysis showed that FRI data alone were insufficient to predict cohort status. An analysis was subsequently conducted using the richer data set found in the Continuous Forest Inventory (CFI). Stepwise discriminant function analysis of the CFI data revealed three diameter based variables capable of significantly discriminating between cohorts: Diameter Class Richness, Coefficient of Variation and Evenness. The model built using CFI data correctly classified mixedwood stands into the cohort classes in 75% of the cases. Validation tests of this predictive model using Permanent Sample Plot (PSP) data revealed the model capable of discriminating between the Cohorts in 79.2% of the cases. The model was able to successfully classify Cohort 1 and Cohort 2 stands, but was less able to classify Cohort 3 stands. Model accuracy was improved when a two-cohort approach was applied, using Cohort 1 (even-aged cohort) and a combination of Cohorts 2 and 3 (complex structured cohort). This study shows that diameter class data is essential in predicting cohort class in boreal mixedwood stands.;Key Words: mixedwoods, boreal, cohorts, multi cohort forest management, Weibull functions, diameters, discriminant function analysis.