Evaluating techniques for measuring the volume of large woody debris on erosion-prone cutovers

Abstract

Background: This study compares two remote sensing methods to a field-based line intersect sampling (LIS) method for measuring Large Woody Debris (LWD) on 27 ha of recently clear-felled plantation cutover in North Canterbury, New Zealand. The remote sensing methods tested were: 1) machine learning-based LWD detection on high-resolution orthophotography; and, 2) virtual line intersect sampling via photogrammetric point clouds.

Methods: The site was flown at 40 m above ground level with a 20 megapixel RGB camera, and 80% image overlap, achieving a ground sampling distance of 1.1 cm. The resulting point cloud and orthophoto were used for remote sensing measures, compared to the ground-based volume measure. The ground-based line intersect sampling used 28 x 50 m, slope-corrected L-shaped transects arranged in a grid pattern across the site. LWD diameter was measured at the transect intersection and used to derive volume using Van Wagner’s (1968) LIS formula.

Results: The ground-based method measured a mean LWD volume of 31.0 ± 10.2 m³/ha. The photogrammetry-based method measured a lower mean volume of 13.6 ± 3.8 m³/ha, with an r² value of 0.61, indicating moderate correlation with the ground-based method. The machine learning method measured a mean LWD volume of 14 m³/ha and had a moderately weak r² value of 0.39, in line with other published models. The machine learning method positively correlated with both ground-based and photogrammetric LWD volume, making it most useful for identifying high-density areas.

The ground-based method consistently returned higher volume measures. Occluded residues could not be captured by the remote sensing methods, but were positively observed in the field. Accuracy in establishing and measuring to the in-field transect remains critical also, regardless of topography challenges. Plot-wise interpretation of the orthophoto revealed 11% of field-measured pieces fell outside the transect. While challenging in steep topographies, field measures remain the only way to capture occluded volume in high-density areas.

Conclusions: Remote sensing techniques based on RGB photography offer a safer, more efficient alternative that can reliably identify high-density areas, even if the estimate is likely low. Strategic application of these methods, considering costs, benefits, and limitations, will enhance plantation owners’ ability to ensure alignment with new cutover standards.