Affiliation: College of Arts and Sciences, Curriculum in Environment and Ecology
Forest modification, from local stress to global change, has given rise to efforts to model, map, and monitor critical properties of forest communities like structure, composition, and diversity. Predictive models based on data from spatially-nested field plots and LiDAR-hyperspectral remote sensing systems are one particularly effective means towards the otherwise prohibitively resource-intensive task of consistently characterizing forest community dynamics at landscape scales. However, to date, most predictive models fail to account for actual (rather than idealized) species and community distributions, are unsuccessful in predicting understory components in structurally and taxonomically heterogeneous forests, and may suffer from diminished predictive accuracy due to incongruity in scale and precision between field plot samples, remotely-sensed data, and target biota of varying size and density. This three-part study addresses these and other concerns in the modeling and mapping of emergent properties of forest communities by shifting the scope of prediction from the individual or taxon to the whole stand or community. It is, after all, at the stand scale where emergent properties like functional processes, biodiversity, and habitat aggregate and manifest. In the first study, I explore the relationship between forest structure (a proxy for successional demographics and resource competition) and tree species diversity in the North Carolina Piedmont, highlighting the empirical basis and potential for utilizing forest structure from LiDAR in predictive models of tree species diversity. I then extend these conclusions to map landscape pattern in multi-scale vascular plant diversity as well as turnover in community-continua at varying compositional resolutions in a North Carolina Piedmont landscape using remotely-sensed LiDAR-hyperspectral estimates of topography, canopy structure, and foliar biochemistry. Recognizing that the distinction between correlation and causation mirrors that between knowledge and understanding, all three studies distinguish between prediction of pattern and inference of process. Thus, in addition to advancing mapping methodologies relevant to a range of forest ecosystem management and monitoring applications, all three studies are noteworthy for assessing the ecological relationship between environmental predictors and emergent landscape patterns in plant composition and diversity in North Carolina Piedmont forests.