Three-dimensional (3D) data on forest structure have transformed the level of detail and accuracy of forest information. While these 3D data have primarily been derived from airborne laser scanning (ALS), there has been growing interest in the use of 3D data derived from digital aerial photogrammetry (DAP) and image-matching algorithms. In particular, research and operational forestry communities are interested in using DAP data to update existing ALS-derived enhanced forest inventories. Although DAP depends on accurate terrain information provided by ALS to normalize digital surface models to heights above ground, in an inventory update scenario, DAP data currently have cost advantages over repeat ALS acquisitions.

Accurate, reliable, and cost-effective methods of evaluating forest regeneration success are needed to improve forest inventories and silvicultural operations. While traditional surveys are relatively inexpensive and meet current data requirements, their annual coverage of over 1 million hectares in British Columbia alone are operationally and logistically intensive. To improve the efficiency and utility of forest regeneration inventories, the incorporation of multi-temporal monitoring linked to years since planting (YSP) could help improve understanding of rates and characteristics of vegetative succession while providing a means to evaluate the economic and operational success of management actions on public land. In this study, we evaluate the potential of utilizing Unmanned Aerial System (UAS)-acquired very high spatial resolution imagery to provide spatial, spectral, and structural information on forest regeneration in previously clear-cut stands near Nakusp and Quesnel, British Columbia, Canada. Three stands approximately 5, 10, and 15 YSP were chosen at both sites. Using wall-to-wall UAS-acquired red–green–blue (RGB) imagery, dense Digital Aerial Photogrammetric (DAP) point clouds were produced providing forest structure information. Spectral data in the form of Visible Vegetation Indices (VVI) including the Normalized Green Red Difference Index (NGRDI), visible atmospherically resistant index (VARI_g), and Green Leaf Indices (GLI_x) were computed. Spectral and structural information from the VVI and DAP were combined to perform Object-Based Image Analyses (OBIA) facilitating supervised classifications of forest cover into conifer, deciduous, and ground classes. Independent classifications were performed on each stand, yielding high overall accuracies (86–95% for Nakusp; 93–95% for Quesnel). Spectral and structural differences amongst classes and YSP were analysed. Height and area coverage of conifers were found to increase with YSP in both sites (0.7–2.7 m for Nakusp; 0.3–2.2 m for Quesnel), while VVI metrics were shown to be more successful than standard RGB at differentiating forest cover through time. The results of this study indicate that UAS-acquired imagery has a potential niche for quickly, accurately, and reliably providing highly detailed spatial, spectral, and structural information on forest regeneration. Methodology and data products from this study show promise for benefiting silvicultural monitoring and operations while improving multi-temporal forest inventory knowledge.

The increasing use of unmanned aerial vehicles (UAV) and high spatial resolution imagery from associated sensors necessitates the continued advancement of efficient means of image processing to ensure these tools are utilized effectively. This is exemplified in the field of forest management, where the extraction of individual tree crown information stands to benefit operational budgets. We explored training a region-based convolutional neural network (Mask R-CNN) to automatically delineate individual tree crown (ITC) polygons in regenerating forests (14 years after harvest) using true colour red-green-blue (RGB) imagery with an average ground sampling distance (GSD) of 3 cm. We predicted ITC polygons to extract height information using canopy height models generated from digital aerial photogrammetric (DAP) point clouds. Our approach yielded an average precision of 0.98, an average recall of 0.85, and an average F1 score of 0.91 for the delineation of ITC. Remote height measurements were strongly correlated with field height measurements (r² = 0.93, RMSE = 0.34 m). The mean difference between DAP-derived and field-collected height measurements was −0.37 m and −0.24 m for white spruce (Picea glauca) and lodgepole pine (Pinus contorta), respectively. Our results show that accurate ITC delineation in young, regenerating stands is possible with fine-spatial resolution RGB imagery and that predicted ITC can be used in combination with DAP to estimate tree height.

To improve precision management and the cost effectiveness of forest practices, we investigate a pre-harvest airborne laser scanning (ALS) forest inventory with an unmanned aerial vehicle (UAV) acquired post-harvest digital aerial photogrammetry (DAP) inventory to identify the location and residual volume of stands following selection harvesting. ALS data and field measurements collected pre-harvest in 2013 (T₁) and UAV imagery collected post-harvest in 2015 (T₂) were processed to produce analogous point clouds of the study area near Williams Lake, British Columbia, Canada. Tree height, diameter at breast height (DBH), and species were recorded from systematically located variable radius plots subsequent to ALS and DAP collection. Point cloud metrics and field measurements from each data set were used to create T₁ ALS and T₂ DAP predictive volume models. Direct and indirect volume change estimates were created from the difference between T₁ ALS and T₂ DAP model results. The estimated root mean square error (RMSE) for volume was 17.34% and 18.50% for the 2013 ALS and 2015 DAP models, respectively. The indirect and direct models predicting volume change produced errors of 16.65% and 86.56%, respectively. Results achieved from ALS and DAP models indicate strong potential for inventories generated using UAV-acquired DAP to estimate the quantity and location of residual volume after harvest operations, and could be applied in tandem to act as a semi-automated inventory cycling method to improve operational efficiency and cost effectiveness in Canadian forest management.

Analyses characterizing canopy gaps are required to improve our understanding of spatial and structural variations in forest canopies and provide insight into ecosystem-level successional processes. Gap size frequency distributions (GSFD) are indicative of ecological processes and disturbance patterns. To date, GSFD in boreal forest ecosystems have not been systematically quantified over large areas using a single consistent data source. Herein we characterized GSFDs across the entirety of the Canadian boreal forest using transects of airborne laser scanning (ALS) data. ALS transects were representatively sampled within eight distinct Canadian boreal ecozones. Gaps were detected and delineated from the ALS-derived canopy height model as contiguous canopy openings ≥8 m²with canopy heights ≤3 m. Gaps were then stratified by ecozone and forest type (i.e. coniferous, broadleaf, mixedwood, wetland-treed), and combinations thereof, and GSFDs were calculated for each stratum. GSFDs were characterized by the scaling parameter of the power-law probability distribution, lambda (λ) and Kolmogorov-Smirnov tests confirmed that GSFDs for each stratum followed a power-law distribution. Pairwise comparisons between ecozones, forest types, and combinations thereof indicated significant differences between estimates of λ. Scaling parameters were found to be more variable by ecozone (1.96–2.31) than by forest type (2.15–2.21). These results contrast those of similar studies done in tropical forest environments, whereby λ was found to be relatively consistent across a range of site types, geological substrates, and forest types. The geographic range considered herein is much larger than that of previous studies, and broad-scale patterns in climate, landforms, and soils that are reflected in the definition of unique ecozones, likely also influence gap characteristics.