Estimating forest stand structure attributes from terrestrial laser scans.

Forest stands are often parameterized by vegetation indices such as the Leaf Area Index (LAI). However, other indices (i.e. stand denseness, espacement, canopy density, canopy cover, foliage cover, crown porosity, gap fraction) may better characterize forest structure. Terrestrial and airborne active sensor data has been used to describe canopy structural diversity and provide accurate estimates of forest structure indices. This study uses Terrestrial Laser Scanner (TLS) to characterize forest structure through the above-mentioned indices. The relationship between all of them was studied to assess the extent to which they relate and their capability to properly describe forest stands. A strong correlation was visible between LAI and the canopy density index (r = 0.87 to 0.91 depending on the extraction methods) despite the underevaluated values of the first. Even though more precise LAI estimates were expected from using co-registered multiple scans, the LAI variability proved to be low and correlations with the remaining indices weakened when compared to a single scan approach. An exception was canopy cover, a structural index that disregards the three-dimensionality of the canopy, with which the LAI obtained from multiple scans maintained a strong correlation. This suggests that multiple scanning leads to an unweighted oversampling of the scene, overshadowing its advantages in removing tree occlusions. Weak correlations were visible between classic forest structural indices (basal area density index, espacement index, denseness index) and the rest of the descriptors. Despite this exception, most of the forest indices showed average to strong correlations in-between each other. Therefore, we conclude that a better description of forest stands structure may be achieved through unsegmented single scan point cloud processing in both 3D and 2D space, optical data from the incorporated digital camera being a plus, but not an essential requirement.