Contribution of topographic features and categorization uncertainty for a tree species classification in the boreal biome of Northern Ontario

Variations within local topography can effectively impact the location of tree species naturally forested areas. Furthermore, uncertainty prediction for classification vastly differ amongst and overlying groupings. This study investigated supplementation a suite topographic features corresponding to morphometry hydrological considerations, in addition multispectral imagery other LiDAR-derived features, at fine (2 m) spatial resolution pixel-based region boreal biome northern Ontario, Canada. The area conforms Abitibi River Forest (ARF) consists black spruce (Picea mariana), balsam fir (Abies balsamea), trembling aspen (Populus tremuloides), poplar balsamifera), tamarack (Larix laricina), white glauca), eastern cedar (Thuja occidentalis). Random forest (RF) support vector machines (SVMs) were implemented classification. Topographic specifically those conforming channel base level, valley depth, multi-resolution bottom flatness (MRVBF), among most important predictors. RF SVM methods trained on pixels pure stands (composed 70%+ same species) groupings, which split by site level. Modelling accuracies both pixel level reported, with best model attaining an overall accuracy Cohen’s kappa score 0.79 0.69 classification, respectively. Entropy maps generated characterize prediction, substantiate that regions lowest correspond wetlands, are dominated mariana). A modified entropy map was calculated from normalized top two probabilities groupings predicted per pixel, so as better highlight uncertainty. second most-likely also computed, supports presence balsamea) secondary throughout ARF region.