Terrestrial lidar remote sensing of forests: Maximum likelihood estimates of canopy profile, leaf area index, and leaf angle distribution

Abstract

Terrestrial laser scanning (TLS) swings a tiny-footprint laser to resolve 3D structures rapidly and precisely, affording new opportunities for ecosystem studies, but its actual utility depends largely on efficacies of lidar analysis methods. To improve characterizing forest canopies with TLS, we forged a methodological paradigm that combines physics and statistics to derive foliage profile, leaf area index (LAI), and leaf angle distribution (LAD): We modeled laser–vegetation interactions probabilistically and then developed a maximum likelihood estimator (MLE) of vegetation parameters. Unlike classical gap-based algorithms, MLE explicitly accommodates laser scanning geometries, fully leverages raw laser ranging data, and simultaneously derives foliage profile and LAD. We evaluated MLE using both synthetic lidar data and real TLS scans at sites in Everglades National Park, USA. Estimated LAI differed between algorithms by an average of 26%. Compared to classical gap analyses, MLE derived foliage density profile and LAD more accurately. Also, MLE has a rigorous statistical foundation and generated error intervals better indicative of the true uncertainties of estimated canopy parameters—an aspect often overlooked but essential for credible use of lidar vegetation products. The theoretical justification and experimental evidence converge to suggest that classical gap methods are sub-optimal for exploiting tiny-footprint lidar data and MLE offers a paradigm-shifting alternative. We envision that MLE will further boost confident use of terrestrial lidar as a versatile tool for environmental applications, such as forest survey, ecological conservation, and ecosystem management.