Effects of voxel size and sampling setup on the estimation of forest canopy gap fraction from terrestrial laser scanning data

Assessments of forest canopy structure remain a challenge and are most often conducted using indirect techniques limited to a two-dimensional perspective. Using terrestrial laser scanner (TLS) technology, a three-dimensional (3D) approach to study canopy structure was conducted by modeling forest scenes from three broadleaved forest stands with different canopy features. Field TLS data were collected from each stand using a phase based FARO LS880 laser scanner on four sampling setups. The capability of TLS-derived data to represent canopy structure was evaluated by comparing gap fraction estimates from the 3D models with gap fraction values from digital hemispherical photographs (DHP). Firstly, the collected 3D point clouds were processed to obtain fully representative voxel-based models of the forest canopy. Secondly, ray tracing algorithms were applied on these models to simulate hemispherical views and estimate gap fraction. Finally, a sensitivity analysis was done using different voxel sizes and the four sampling setups on the simulations, in order to assess their impact on the gap fraction estimates derived from TLS data. Results of TLS-derived gap fraction showed that combining nine scans produced better results in all forest stand. Similarly, the dimension of voxels have a marked influence on these results. Voxel sizes of 1 cm, 2 cm and 4 cm were found to have less error when compared with real gap fraction values derived from DHP for young, intermediate and mature forest stands, respectively (RMSE ranging from 9% to 16%). However, substantial differences in gap fraction were observed and described at different zenith angles. These results suggest that specific TLS sampling setup and processing are required depending upon the forest type under analysis. Overall, this research indicates that phase based TLS data can be used for objective calculation of gap fraction. (C) 2014 Elsevier B.V. All rights reserved.