Can estimation of leaf area density from terrestrial LiDAR scanning data be further improved by using additional information provided by full-waveform instruments?

Abstract:

Terrestrial LiDAR scanning (TLS) data has become an increasingly prevalent means for quantifying 3D structure of plant canopies, including leaf area density (LAD) and the leaf angle distribution. Previous work has primarily based calculations on discrete-return data or first returns from full-waveform data. Recent work has taken advantage of multiple returns per beam from full-waveform data in calculations of LAD, based on equal weighting of returns from the same beam. It has been suggested that a more precise weighting of returns based on corrected relative intensity could improve estimates. Such corrected intensities require knowledge or assumptions about distances, angles, and reflective properties of the vegetation to determine the relative fraction of beam energy associated with each return.

In this work, synthetic full-waveform TLS simulations were implemented to quantify potential improvements in LAD estimation associated with precise intensity-based weighting of multiple returns from full-waveform data. Simulated datasets provided knowledge of exact beam fractions interacting with plant material (or no surfaces at all), which allowed for an idealized weighting between multiple returns. Methods using this “exact” weighting of returns were compared with first-return and equal-weighting methods within the simulations for homogeneous voxels and heterogeneous almond tree cases.

Exact knowledge of fractional energy associated with each return did not necessarily lead to more accurate estimates of transmission probability when substantial energy was associated with partial misses. Equal weighting actually tended to perform better than exact weighting in these cases because it partially compensated for partial misses. In cases where partial misses were considered explicitly or were minimal (i.e., TLS beams hit the ground or other surfaces after passing through the voxel), the exact method approached theoretical transmission probability calculated using Beer’s Law. Results indicate that intensity-based weighting may not provide improvement for scans of isolated trees where the scanner is positioned lower than the voxels of interest. Intensity-based weighting does appear to have potential to improve LAD calculations for voxels below the scanner, such as in aerial scanning.