Abstract
LiDAR is an established technology that is increasingly being used to characterise spatial variation in stand metrics used in forest inventory. As the cost of LiDAR acquisition markedly declines with LiDAR pulse density, it is useful to identify how far pulse density can be reduced without compromising the precision of relationships between LiDAR and stand metrics. Using plot measurements and LiDAR data obtained from highly stocked and unthinned Douglas-fir plantations (Pseudotsuga menziesii [Mirb.] Franco), the objective of this research was to characterise the precision of regressions between LiDAR metrics and stand metrics (mean top height, Hm, volume, V and mean diameter, D) under a range of pulse densities using Digital Terrain Models (DTMs) representing two common scenarios. Under the first scenario, which represents an initial acquisition, the point cloud was sequentially culled and used for creation of a DTM and corresponding LiDAR cloud metrics. In the second scenario, which represents a subsequent acquisition, a DTM generated at high pulse density (10 pulses m−2) was used for the creation of the corresponding LIDAR cloud metrics. Models describing the precision of regressions between LiDAR metrics and stand metrics were developed at 10 pulses m−2. LiDAR data were culled to pulse densities ranging from 10 to 0.01 pulses m−2 and the impact of culling on the precision of these regressions was examined under the two scenarios. For the scenario with the culled DTM, precision of the three models remained stable until densities of 2 – 3 pulses m−2 were reached. Below this threshold, there was a gradual decline in precision to pulse densities of 0.7 – 1 pulses m−2 at which point the R2 was 95% of the maximum values. Further culling of the data resulted in a sharp decline in model precision for all three regressions. For the scenario where the DTM was held at a high pulse density, little change in the precision of the regressions was found until pulse densities of 0.04 to 0.2 pulses m−2 were reached. There was a sharp decline in precision below pulse densities of 0.04 pulses m−2 for all three models. This study was undertaken in highly stocked unthinned Douglas-fir stands located in areas with complex topography. Consequently, the pulse density thresholds described here are likely to be conservative and could be used to guide acquisition of high-quality LiDAR datasets for this species.
Highlights
LiDAR is an established technology that is increasingly being used to characterise spatial variation in stand metrics used in forest inventory
In the simulations where the same culling rates were applied to both the ground and above-ground points, the maximum R2 was reached between 4 and 10 pulses m−2 (Figure 3; Table 4). Both the R2 and root mean square error (RMSE) remained reasonably static at values close to this maximum R2 for all metrics until pulse densities of between 2 and 3 pulses m−2 were reached (Figure 3; Table 4)
The maximum R2 were found at relatively low pulse densities (0.3 pulses m−2) for two of the three metrics, little variation in precision of the culled metrics was noted from 10 pulses m−2 to a threshold of 0.04 pulses m−2 for D and 0.2 pulses m−2 for Hm and V (Figure 3; Table 4)
Summary
LiDAR is an established technology that is increasingly being used to characterise spatial variation in stand metrics used in forest inventory. In the second scenario, which represents a subsequent acquisition, a DTM generated at high pulse density (10 pulses m−2) was used for the creation of the corresponding LIDAR cloud metrics. There are many ways to increase the pulse density (e.g. reducing aircraft altitude and speed), this variable is related to flight time (Jakubowski et al 2013) and acquisition cost increases with pulse density. It is, of considerable interest to understand how far pulse density can be reduced without unduly compromising the accuracy of inventory information obtained from LiDAR data
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
