- The portability and reduced price of unmanned aerial systems (UAS) in recent years has led to a broad range of new UAS-enabled scientific inquiries, including for forestry. Small, consumer-grade UAS are advantageous for forest measurements due to their portability, ease and safety of deployment, and notably, they are currently the only remote sensing technology capable of measuring both individual seedlings and individual mature trees from above. Light detection and ranging (lidar) sensors have been increasingly used in forestry research over the past few decades as well, and now models exist which can be integrated with UAS for tree mensuration. Computer vision software known as structure from motion (SfM) can be used to produce analogous data to those produced by lidar, known as point clouds, from still images taken from UAS. The goal of this dissertation is to examine how to use point cloud data to augment tree level estimates for forest inventory. To cover a broad range of dimensional values defining a forest, two different stages in the life cycle of the stand were investigated, the seedling stage, and the mature stage which precedes harvest. For analyzing seedlings, the first manuscript (second chapter) in this dissertation used UAS and multispectral sensors to produce point clouds of southwestern white pine seedlings in common garden boxes. Here, a methodology is presented for estimating seedling sizes from SfM reconstructions and using them to improve the predictive power of seedling size models along with ground measurements from the previous year. Also, I make recommendations for how common garden designs can be designed so as to lengthen the duration of useful UAS surveys. Finally, I present a seedling size variable that performs well both as a predictor and as a response, the product of seedling height and diameter at root collar, or longitudinal area. To address the mature stage of the trees, the second manuscript (third chapter) compared the performance of three platforms that vary greatly in cost, ease of operation, and data processing requirements. One of the platforms was identical to the UAS used in the first manuscript, one was a lidar carried by a larger UAS (UALS), and one was a ground based mobile lidar scanner (MLS). The UAS produced SfM height estimates that were comparable to those by the UALS, though they tended to be underestimates due to smoothing of the SfM reconstruction. Both the UALS and MLS platforms produced sufficient stem returns to locate a majority of the tree stems in the scene, while none could be located from the UAS. Using data from the MLS and the UALS, I showed that using the stem near the base of the crown or the treetop to estimate lean will produce different lean estimates and contend that the MLS is the best platform for estimating the lean of the stems. In the third and final manuscript (fourth chapter), I compared two methods for estimating stem lean from the MLS data. The more conservative lean estimate, which involves using the horizontal distance between the top and bottom of the merchantable portion of the stem, was included as a predictor to improve the fit of existing nonlinear stem taper and volume equations. The results suggest that trees that lean as little as 2° should be modeled differently than those which are vertical. Also, substituting other diameters higher on the stems for DBH impacts the fit of the models for leaning trees differently than for vertical ones, such that leaning trees seem to have a narrower range of optimal diameter heights. As a whole, my dissertation supports the usage of UAS and MLS to improve the quality and efficiency of remotely measuring single seedlings or mature trees forest inventory, while also identifying major limitations of the technology and recommending strategies to contend with them.