Imagery Method. Although a LiDAR model was not constructed, for reasons of imprecise accuracy comparison, a cruder estimate was created from imagery to show some capability that exists with this approach. Some field data was readily available to gauge accuracy, making a provisional analysis more feasible. Canopy percent cover was calculated by summarizing the surface area of all individual tree crown objects, computing the ratio of this over an entire plot, and comparing that against field-crew estimations of canopy cover. Although we did not explore the relationships between densitometer and LiDAR, we did explore the optical data and field data relationships as this is commonly done in remote sensing research.
Imagery Method. Like many of the other metrics, stand density is dependent on the existence of ITOs. Because of varying tree composition throughout the project area, one single TDA does not have the flexibility to address disparate scenarios. Ideally, the algorithm would first evaluate the area of interest, attempt a preliminary classification of composition types, and then execute the respective delineation function that best accommodates the spectral and spatial composition of that type. Types would vary according to presence of conifer/deciduous trees and maybe even species. However, the single TDA is still showing promising results for the majority of sample plots within the project area.
Imagery Method. Conifer/Deciduous classification, Species classification, and Basal Area are all predictors that, with imagery interpretation, rely much more on the statistical analysis of pixel content of individual trees. While previous metric predictions are still influenced by spectral analysis within the TDA, estimating more granular qualities of the tree, like species, requires further investigation of color compositions.
Imagery Method. Basal Area predictions were based on a linear regression model, with all independent variables being respective signatures of ITOs. The literature review for this project informed which larger set of variables to be used within the initial model. Then, this model was gradually narrowed down to only the most influential components show below.
Imagery Method. It is possible to use pixel-based statistics for each ITO in order to identify tree species. Known trees and corresponding ITCs can be used as training sites which will form the unique signatures for each tree species that a classification algorithm will use. The input for each ITC consists of a single multispectral vector, containing signatures for: Mean intensity value The Standard Deviation of this mean Mean-lit value [ave. of all pixels in an ITO that have a pixel value above the mean intensity of all pixels in the object] Tree top value, which is the brightest pixel in the ITO
