In this dissertation I present three studies incorporating lidar data into different aspects of forest restoration. All studies use lidar individual tree detection as source data, in part to enable making measurements of tree spatial patterns in terms of tree clumps and canopy openings. This common focus exists because spatial patterns of trees influence fire and insect behavior, snow retention, tree regeneration, and other key ecosystem functions and services for which humans manage forests. In Chapter 1 I sought to provide this dataset by asking these questions: (1) What is the geographic and environmental distribution of restored active-fire forest patches in the Sierra Nevada mixed-conifer zone? (2) What are the ranges of variation in structure and spatial patterns across restored patches? (3) How do density, tree clumping, and canopy opening patterns vary by topography and climate in restored patches? I analyzed fire history and environmental conditions over 10.8 million ha, including 3.9 million ha in the Sierra Nevada mixed-conifer zone, and found that the 30,379 ha of restored patches were distributed throughout the range but were more abundant on National Park lands (81% of restored areas) than National Forest lands and were positively correlated with lightning strike density. Furthermore, 33% of restored areas were located in western Yosemite National Park and met our criteria for inclusion in this study only after being burned at low and moderate severity in the 2013 Rim Fire. Lidar-measured ranges of variation in reference condition structure were broad, with density ranging from 6-320 trees ha−1 (median 107 trees ha−1), basal area from 2-113 m2 ha−1 (median 21 m2 ha−1), average size of closely associated tree clumps from 1 to 207 trees (median 3.1 trees), and average percent of stand area >6 m from the nearest canopy ranging from 0% to 100% (median 5.1%). These ranges matched past studies reporting density and spatial patterns of contemporary and historical active-fire reference stands in the Sierra Nevada, except this study observed longer tails on distributions due to the spatial completeness of lidar sampling. Reference areas in middle-elevation climate zones had lower density (86 vs. 121 trees ha-1), basal area, (13.7 vs. 31 m2 ha-1), and mean clump size (2.7 vs. 4.0 trees) compared to lower- and higher-elevation classes, while ridgetops had lower density (101 vs. 115 trees ha-1), basal area (19.6 vs. 24.1 m2 ha-1), and mean clump size (3.0 vs. 3.3 trees) but more open space (7.4% vs. 5.1%) than other landforms. In Chapter 2 I developed new methods for integrating lidar data into silvicultural planning at treatment unit and project area scales, with a focus on dry forest restoration treatments. At the stand scale my objective was to delineate the decision space for prescription parameters like density, basal area, and spatial patterns given the soft constraints of reference conditions and the hard constraints of possible transitions given current structure. At the landscape scale my objective was to provide a framework for selecting from available treatment options, stand by stand, to meet different landscape-level goals. I applied the new methods to a case study area in the Lake Tahoe Basin, California and asked in this context: How do structural departures from reference conditions and associated treatment prescriptions vary with topographic position and aspect? I found that ridges and southwest-facing slopes in the study area had a greater degree of departure from the reference envelope and required more density reduction compared to valleys and northeast-facing slopes. In Chapter 3 I used pre- and post-Rim Fire data from the 25.6 ha Yosemite Forest Dynamics Plot (YFPD) to build a model of tree mortality predicted from lidar individual tree detection structural metrics. I calculated metrics at the scale of lidar-detected trees (termed tree-approximate objects, TAOs), at the scale of 0.1 ha plots centered on each TAO, and at the 90×90 m neighborhood scale. I used these to predict TAO mortality at the neighborhood scale and TAO mortality class – immediate or delayed mortality – at the TAO scale. I also tested the inclusion of a set of topoedaphic and burn weather predictors as well as a cross-scale interaction term between the TAO mortality model and the neighborhood-level mortality model. I asked these questions: (1) How does mortality progress 1-4 years post-fire in terms of rates, demographics, and agents? (2) What elements of forest structure and pattern predict immediate and delayed post-fire mortality at scales from TAOs to neighborhoods? (3) How does the prevalence of different mortality agents vary with changes in the important fine-scale predictors of fire mortality? I found that smaller trees were killed in the first year with a 40% mortality rate and the average diameter of killed trees increased each subsequent year while the mortality rate decreased. The topoedaphic and burn weather predictors as well as the cross-scale interaction improved model fit and parsimony, but that the improvement was directional, i.e., including neighborhood-level information improved the TAO-level model but not vice-versa. Important predictors fell into categories of fuel amount, fuel configuration, and burning conditions. Amounts of crown damage for immediately killed trees were higher for TAOs shorter than 51 m and in 0.1 ha areas where mean clump sizes was less than 21 TAOs. The amount of delayed mortality that was directly fire-related was higher when TAO crown base heights were less than 28 m and TAO density in 0.1 ha areas was greater than 170 TAOs ha-1. Crown base heights over 18 m and local TAO density of less than 180 TAOs ha-1 had more beetle kill and less rot. The model performed similarly well on an independent validation dataset of 48 0.25 ha plots spanning the footprint of the Rim Fire within Yosemite as on the YFDP training data, indicating that the model is widely applicable.