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.
Environmental niche modeling is an increasingly common tool in conservation and management of non-timber species. In particular, models of species’ habitats have been aided by new advances in remote sensing and it is now possible to relate forest structure variables to understory species at a relatively high resolution over large spatial scales. Here, we model landscape responses for a culturally-valued keystone shrub, velvet-leaf blueberry (Vaccinium myrtilloides Michaux), in northeast Alberta, Canada, to better understand the environmental factors promoting or limiting its occurrence, abundance, and fruit production, and to guide regional planning. Occurrence and abundance were measured at 845 and 335 sites, respectively, with both strongly related to land cover type and topo-edaphic factors. However, their influence varied widely, reflecting differences in the processes affecting occurrence and abundance. We then used airborne laser scanning (ALS) to characterize horizontal forest canopy cover for the study area, and related this and other geospatial variables to patterns in fruit production where we demonstrated a five-fold increase in fruit production from closed to open forest stands. We then simulated forest canopy thinning across the study area to identify places where gains in fruit production would be greatest following natural disturbance or directed management (e.g., thinning, prescribed burning). Finally, we suggest this approach could be used to identify sites for habitat enhancements to offset direct (land use change) or indirect (access) losses of resources in areas impacted with resource extraction activities, or simply to increase a culturally-valued resource through management.
Forest inventories in uncertain future economic and environmental conditions require the development of cost-effective measurement techniques to provide robust and accurate information on forests across regional and global scales. Digital terrestrial photogrammetry (DTP) can be used to detect and measure trees on sample plots. In this study, a method was developed which used spherical images taken strategically within plots, and under varying acquisition conditions, to derive forest inventory attributes. Using a set of 102 photos on 400 m2 circular plots achieved a mean detection rate of 72.3% and estimated diameter to an RMSE of 19.0%. This study also explored the sensitivity of detection and estimation accuracy to different field and acquisition conditions. Detection of individual trees was significantly influenced by the tree size and species (p < 0.05 in a regression analysis), while plot-level detection was influenced by size and stem density. Tree size and the distance to the camera significantly influenced the accuracy of estimated attributes. These results are comparable to those of other DTP and terrestrial laser scanning studies in similar forest types while using fewer photos and less time, demonstrating the value of cost-effective methods for DTP estimation of forest attributes.
3 April 2025 in Forestry: An International Journal of Forest Research
Monitoring changes in forest cover is critical to understanding forest dynamics and informing sustainable forest management practices. Both Canada’s National Forest Inventory and satellite-based monitoring programs have consistently reported an increasing trend in treed area over recent decades, despite differences in survey design, spatial resolution, and temporal representation. This study presents a spatially explicit mapping approach that integrates annual satellite-derived land cover data (1984–2022) with historical disturbance records to analyze treed area dynamics at regional and national scales. Using this approach, we assess the spatial and temporal trends in treed areas in Canada’s forested ecosystems over nearly four decades, distinguishing between treed area gains resulting from recent disturbances and those related to older, pre-1984 baseline events. Our analysis revealed an average annual increase in treed area of 0.19% nationally (632 655 ha per year), resulting in a total increase of 24.04 Mha (7.2%) over the 39-year period. This increase is mainly due to post-disturbance tree regrowth and natural expansion of trees into previously non-treed areas, especially along forest edges, in gaps, and in wetlands transitioning to treed vegetation. Tree dynamics varied by ecozone, with northern regions (e.g. Hudson Plains, Taiga Cordillera) experiencing the greatest relative gains in treed area, while southern regions (e.g. Montane Cordillera) showed localized declines due to wildfire and other disturbances. Comparisons with National Forest Inventory data revealed similar trends in treed area increase, with notable differences in the direction of change for some ecozones, such as the Hudson Plains and Montane Cordillera. By providing spatially and temporally detailed insights, this study complements sample-based national statistics and provides annualized, spatially explicit mapping that enhances our understanding of forest dynamics and informs similar applications in other regions.
Predicting habitat for rare species at landscape scales is a common goal of environmental monitoring, management, and conservation; however, the ability to meet that objective is often limited by the paucity of location records and availability of spatial predictors that effectively describe their habitat. To address this challenge, we used an adaptive, model-based iterative sampling design to direct four years of rare plant surveys within 0.25 ha plots across 602 sites in northeast Alberta, Canada. We used these location records to model and map rare plant habitats for the region using a suite of geospatial predictors including airborne light detection and ranging (LiDAR) vegetation structure metrics, land cover types, soil pH, and a terrain wetness model. Our results indicated that LiDAR-derived vegetation structural metrics and land cover were the most important individual factors, but all variables contributed to predicting the occurrence of rare plants. For LiDAR variables, rarity was negatively related to maximum canopy height, but positively related to canopy relief ratio. Rarity was therefore more likely in places with shorter canopy heights and greater structural complexity. This included fens, which overall had the highest rates of rare plant occurrence. Model-based allocation of sampling led to detections of uncommon species at nearly all sites, while the rarest species in the region were detected at an average encounter rate of 8%. Landscape predictions of rare plant habitat can improve our understanding of environmental limits in rarity, guide local management decisions and monitoring plans, and provide regional tools for assessing impacts from resource development.