Abstract
Airborne lidar point clouds of vegetation capture the 3-D distribution of its scattering elements, including leaves, branches, and ground features. Assessing the contribution from vegetation to the lidar point clouds requires an understanding of the physical interactions between the emitted laser pulses and their targets. Most of the current methods to estimate the gap probability ( P gap ) or leaf area index (LAI) from small-footprint airborne laser scan (ALS) point clouds rely on either point-number-based (PNB) or intensity-based (IB) approaches, with additional empirical correlations with field measurements. However, site-specific parameterizations can limit the application of certain methods to other landscapes. The universality evaluation of these methods requires a physically based radiative transfer model that accounts for various lidar instrument specifications and environmental conditions. We conducted an extensive study to compare these approaches for various 3-D forest scenes using a point-cloud simulator developed for the latest version of the discrete anisotropic radiative transfer (DART) model. We investigated a range of variables for possible lidar point intensity, including radiometric quantities derived from Gaussian Decomposition (GD), such as the peak amplitude, standard deviation, integral of Gaussian profiles, and reflectance. The results disclosed that the PNB methods fail to capture the exact P gap as footprint size increases. By contrast, we verified that physical methods using lidar point intensity defined by either the distance-weighted integral of Gaussian profiles or reflectance can estimate P gap and LAI with higher accuracy and reliability. Additionally, the removal of certain additional empirical correlation coefficients is feasible. Routine use of small-footprint point-cloud radiometric measures to estimate P gap and the LAI potentially confirms a departure from previous empirical studies, but this depends on additional parameters from lidar instrument vendors.
Highlights
Lidar remote sensing encompasses a broad range of technologies and applications [1]
For a fixed leaf area index (LAI), estimations with infinitesimal footprint size converge to Pgap for all PNB methods, which tends to verify the capability of terrestrial laser scan (TLS) to capture Pgap accurately [79]
LPIboth balances the underestimation of LPIfirst and the overestimation of LPIlast, the results do not indicate that LPIboth is more accurate
Summary
Lidar remote sensing encompasses a broad range of technologies and applications [1]. Most remote sensing lidar devices use the time-of-flight technique to generate precise range measurementsRemote Sens. 2020, 12, 4; doi:10.3390/rs12010004 www.mdpi.com/journal/remotesensingRemote Sens. 2020, 12, 4 based on the reflected signals of outgoing laser pulses. Full-waveform lidars have been explored in depth to estimate gap fraction, leaf area index (LAI) profiles, and biomass [mid-to-large footprint [11,12,13,14,15], small-to-mid footprint [16,17], and TLS [18,19]. Most of these physically based approaches are capable of accurate estimation of gap probability (Pgap ) and effective LAI (eLAI = ω · LAI, where ω is the clumping index [20]) without calibration with field measurements. Within-crown leaf area density variation was not completely addressed
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have