| Contemporary forest management on public land incorporates a focus on restoration and maintenance of ecological functions through silvicultural manipulation of forest structure on a landscape scale. Incorporating reference conditions into restoration treatment planning and monitoring can improve treatment efficacy, but the typical ground-based methods of quantifying reference condition data---and comparing it to pre- and post-treatment stands---are expensive, time-consuming, and limited in scale. Airborne LiDAR may be part of the solution to this problem, since LiDAR acquisitions have both broad coverage and high resolution. I evaluated the ability of LiDAR Individual Tree Detection (ITD) to describe forest structure across a structurally variable landscape in support of large-scale forest restoration. I installed nineteen 0.25 ha stem map plots across a range of structural conditions in potential reference areas (Yosemite National Park) and potential restoration treatment areas (Sierra National Forest) in the Sierra Nevada of California. I used the plots to evaluate a common ITD algorithm, the watershed transform, compare it to past uses of ITD, and determine which aspects of forest structure contributed to errors in ITD. I found that ITD across this structurally diverse landscape was generally less accurate than across the smaller and less diverse areas over which it has previously been studied. However, the pattern of tree recognition is consistent: regardless of forest structure, canopy dominants are almost always detected and relatively shorter trees are almost never detected. Correspondingly, metrics dominated by large trees, such as biomass, basal area, and spatial heterogeneity, can be measured using ITD, while metrics dominated by smaller trees, such as stand density, cannot. Bearing these limitations in mind, ITD can be a powerful tool for describing forest structure across heterogeneous landscape restoration project areas. |
