Using three-dimensional point clouds to improve characterizations of forest structure across spatial and temporal scales in mixedwood forest stands

Sustainably managing the world’s forests requires detailed inventories of the resource at varying spatial and temporal scales. The structural and compositional diversity of the boreal mixedwood forest, one of Canada’s largest forest types, provides valuable timber resources and ecological services. However, the extent and complexity of this forest type poses challenges for inventories. The objective of this dissertation was to develop and assess the utility of three-dimensional remote sensing techniques for enhancing forest inventories by characterizing forest structure in boreal mixedwood forests. These technologies are scalable and adaptable for use in forest inventory as they provide consistent spatial and temporal detail. Digital terrestrial photogrammetry from spherical cameras at known locations was used to model individual tree stems and sample plots. For individual trees, stem diameters at different heights were estimated very accurately (RMSE < 1 cm for stem heights below 10 m), which matched or exceeded the accuracy of conventional ground-based inventories. Plot-level point clouds based on a relatively small set of images were used to locate and model trees on sample plots to an accuracy that was comparable to other studies on homogeneous plots (mean 72% detection and 19% RMSE of diameter at breast height). At broader scales, airborne laser scanning (ALS) was used to characterize forest structure by estimating stem size distributions (SSD) across a large forest management unit. First, ALS was used to differentiate unimodal and bimodal stands. Next, parameters of functions describing the SSDs were estimated with ALS metrics (r2 = 0.5) and the resulting functions were more accurate in characterizing field-measured SSDs than without differentiating stands by modality. For assessing temporal patterns of forest structure, photo-interpreted polygons of fire and harvesting were used with the derived SSDs to characterize structural development following stand-replacing disturbance. It was determined that stands that had burned had significantly more trees in larger diameter classes than harvested stands (at a = 0.05). This dissertation outlined the methods required when applying three-dimensional remote sensing technologies to enhancing forest inventories in mixedwood stands and demonstrated the utility of these technologies for deriving information to inform responsible decision-making for management of these forests.Forestry, Faculty ofGraduat