California's fire suppression policy has dramatically changed Sierra Nevada forests over the last century. Forests are becoming more dense and homogenous, leading to fire regime changes that increase the potential of stand-replacing wildfires over large, continuous areas. To mitigate this problem on public lands, the US Forest Service has proposed to implement strategically placed forest fuel reduction treatments. These treatments have been proved effective in modeled and simulated environments, but their efficacy and impact in real forests is not known. The research described in this dissertation is part of a large multidisciplinary project, known as the Sierra Nevada Adaptive Management Project (SNAMP), that aims to evaluate strategically placed landscape area treatments (SPLATs) in two forests of the Sierra Nevada mountains. Specifically, in this thesis, I investigate the feasibility of using an airborne light detection and ranging (lidar) system to gain accurate information about forest structure to inform wildfire behavior models, forest management, and habitat mapping.

First, I investigate the use of lidar data in predicting metrics at the landscape level, specifically to derive surface fuel models and continuous canopy metrics at the plot scale. My results in Chapter 2 indicate that using lidar to predict specific fuel models for FARSITE wildfire behavior model is challenging. However, the prediction of more general fuel models and continuous canopy metrics is feasible and reliable, especially for metrics near the top of the canopy.

It is also possible to derive canopy parameters at the individual tree level. In Chapter 3, I compare the ability of two processing methods--object-based image analysis (OBIA) and 3D segmentation of the lidar point cloud--to detect and delineate individual trees. I find that while both methods delineate dominant trees and accurately predict their heights, the lidar-derived polygons more closely resemble the shape of realistic individual tree crowns.

Acquiring remotely sensed data at high resolution and over large areas can be expensive, especially in the case of lidar. In Chapter 4, I investigate the ability of lidar data to reliably predict forest canopy metrics at the plot level as the data resolution declines. I show that canopy metrics can be predicted at a reasonable accuracy with data resolutions as low as one pulse per squared meter. These findings will be useful to land managers making cost benefit decisions when acquiring new lidar data.

Collectively, the results of this dissertation suggest that remote sensing, and in particular lidar, can reliably and cost-effectively provide forest information across scales--from the individual tree level to the landscape level. These results will be useful for the fire and forest management community in general, as well as being key to the goals of the SNAMP program.