This dissertation examines the relationship between abundances of Roanoke darter, Roanoke logperch, and black jump rock and availability of stream habitat features at three spatial scales in two reaches of the Roanoke River, Virginia. The utility of underwater observation as a method of estimating benthic fish densities is also assessed. Distributions of perpendicular sighting distances indicate models assuming equal sighting probability are not appropriate for benthic species but distance sampling models assuming decreased sighting probability with increased distance from observers provide reasonable alternatives. Abundances estimated using two distance sampling models, a strip transect model, and a backpack electroshocker were strongly correlated. At the microhabitat scale (45 m² cells), differential use of depth, velocity, substrate, and siltation level by all three species during summer low flows was evident. Habitat use characteristics were not transferable, as depths and velocities associated with high fish densities varied between reaches. Univariate and multivariate habitat suitability indices gave similar rankings for combinations of the four habitat variables, but site suitabilities based on these indices were poor predictors of fish abundance. Habitat cells were not selected independently of surrounding habitat characteristics, as fish densities were highest in target cells adjacent to cells with preferred microhabitat characteristics. Roanoke darter and black jumprock abundances were highest at sites where preferred microhabitat patches were non-contiguous while contiguity had no effect on logperch abundance. Multiple regressions showed area of suitable habitat and patch contiguity accounted for 42 %, 34 %, and 33 % of variation in darter, logperch, and jumprock abundances, respectively. Estimates of area of target riffles, area of pools and riffles upstream and downstream of target riffles, and depth, velocity, and substrate characteristics of pools and riffles immediately upstream and downstream of target riffles were obtained. Fish densities were correlated with at least one measure of proximal habitat for all three species. Significant multiple regression models relating fish density to adjacent habitat unit characteristics were also obtained, but the explanatory power of adjacent unit variables varied among small, medium and large riffles and among species. In summary, fish abundance was related to habitat at all spatial scales but explanatory power was limited. / Ph. D.

The current method for estimating wood failure is highly subjective. Various techniques have been proposed to improve the current protocol, but none have succeeded. This research aims to use deep learning/semantic segmentation using SegNet architecture to estimate wood failure in four types of three-ply plywood from mechanical shear strength specimens. We trained and tested our approach on custom and commercial plywood with bio-based and phenol-formaldehyde adhesives. Shear specimens were prepared and tested. Photographs of 255 shear bonded areas were taken. Forty photographs were used to solicit visual estimates from five human evaluators, and the remaining photographs were used to train the machine learning models. Twelve models were trained with the combination of four image sizes and three dataset splits. In comparison to visual estimates, the model trained on 512 × 512 image size with 90/10 dataset split had a mean absolute error (MAE) of 6%, which was the best among the literature.