Identifying eastern hemlock (Tsuga canadensis) patches using LiDAR, in New Brunswick forests
University of New Brunswick
Aerial light detection and ranging (LiDAR) has become an increasingly popular method of determining forest stand characteristics. The objective of this study was to create a predictive model to spatially identify small forest patches dominated by eastern hemlock (Tsuga canadensis), using leaf off LiDAR data. Specifically, 46 LiDAR attributes were evaluated for their capability to differentiate hemlock from 8 other patch types using randomForest regression analysis in New Brunswick, Canada. Two final classification models were created both achieving 92% hemlock classification accuracy, but had differing capabilities in classifying hemlock subclass types. Attributes used in the models were found to be strongly related to the canopy structure of hemlock-dominated patches. It is not believed that model accuracy could be improved using leaf on data; however new models could be designed using a similar method and larger sample of Acadian forest stand types to be more general in classification.