Mapping plant area index of tropical evergreen forest by airborne laser scanning. A cross-validation study using LAI2200 optical sensor

[Departement_IRSTEA]Territoires [TR1_IRSTEA]SYNERGIE [Axe_IRSTEA]TETIS-ATTOS

Английский. Leaf area index estimates in dense evergreen tropical moist forest almost exclusively rest on indirect methods most of which being of limited accuracy or spatial resolution. In this study we examine the potential of full waveform Aerial Laser Scanning (ALS) to derive accurate spatially explicit estimates of Plant Area Index (PAI). A discrete representation of the forest canopy is introduced in the form of a 3D voxelized space. For each voxel (elementary volume, typically one cubic m) a first estimate of local transmittance of vegetation is computed as the ratio of the sum of energy exiting a voxel to the sum of energy entering the same voxel. A spatially hierarchical model is subsequently applied to refine estimates of individual voxel transmittance. Plant area density (PAD) profiles are then computed from the local transmittance values by applying Beer Lambert's turbid medium approximation. PAI values are obtained from vertical integration of PAD profiles. The model is shown to be robust to low sampling intensity and high occlusion rates. We further compared simulated values of gap fraction obtained by ray tracing for 5 angular sectors with in situ LAI2200 measurements taken at 135 positions in a 0.5ha forest plot located in the center of the scene. The overall patterns of simulated and measured values (average value per inclination and pattern of variation along a 70m transect line) were highly consistent. A slight but systematic discrepancy was observed along the inclination gradient, gap fractions derived from ray tracing in the voxelized scene being slightly lower than the measured values. This difference might be the consequence of multiple reflections which have been found to bias gap fractions estimates produced by LAI2200. PAI estimates derived from LAI2200 measurements (either simulated 6.8 or observed 5.9) are much lower than the PAI derived from vertical integration of local PAD (13.6). This large difference reflects the fact that distribution of foliage is strongly spatially structured and that this structural information is not properly accounted for in PAI estimates derived from mean gap fraction per elevation angle. After adjusting local transmittance to match mean LAI 2200 profiles the PAI at plot level was found to be 13.2m2·m−2. We conclude that Aerial Laser Scanning can produce accurate maps of Plant Area Index over large areas with unmatched efficacy, accuracy and ease. This should be of major relevance for many forest ecological studies.