Silvilaser 2019 - Poster Presentations »
Detection of Sub-Canopy Forest Structure Using Airborne LiDAR.
Knowledge on forest structural is vital for making sound forest management decisions. Currently, airborne LiDAR data has been well established as a tool to delineate and analyse the structure of forest canopies worldwide. However, what remains less well known is information on the forest sub-canopy. Sub-canopy structure consists of regenerating saplings, shrubs, herbs, snags and coarse-woody-debris. With the increasing density of LiDAR footprints, new opportunities exist to describe these sub-canopy structural components in forests that were previously overlooked by other remote sensing technologies. In this research we use discrete return airborne LiDAR acquired at a density of 22/ppm to estimate sub-canopy forest structure for 50,000 hectares of forest area in northern British Columbia, Canada. In order to do this we first segmented the forest into canopy and sub-canopy based on Lorey’s mean height. Lorey’s height is a weighted average based on basal area and provides a more stable representation of stand height compared to an unweighted mean height as it is less affected by large numbers of smaller trees. Once Lorey’s height was established we defined the threshold between canopy and sub-canopy as 10% less than Lorey’s height. Both ground truth forest inventory data and the LiDAR point cloud were then segmented into canopy and sub-canopy components. A mixture of standard height based and density based LiDAR metrics were then computed to develop models to predict sub-canopy component of the stands. Models were cross-validated through a forward stepwise regression with the strongest predictors being P95, leaf area density and vertical rumple. The vertical rumple in particular describes the complexity of the vertical structure of the stand. Using the selected metrics, linear regression models were developed that predicted the basal volume and basal area of sub-canopy trees with R-squared = 0.72 and 0.63 respectively. We then applied these models over the entire study area by generating 20m wall-to-wall metrics to estimate stand-level sub-canopy basal volume and area. These results can be used by forest managers to determine future timber supply and candidate locations for selective logging. Additionally, theses attributes can be used to identify fuel loads and potential presence of ladder fuels for fire susceptibility assessments.