Reservoir release patterns are determined by a number of purposes, the most fundamental of which is to manage water resources for human use. Managing our water resources means not only controlling the water in reservoirs but also determining the optimum release rate taking into account factors such as reservoir stability, power generation, water supply for domestic, industrial, and agricultural uses, and the river ecosystem. However, riverbank stability has generally not been considered as a factor, even though release rates may have a significant effect on downstream riverbank stability. Riverbank retreat not only impacts land properties but also damages structures along the river such as roads, bridges and even buildings. Thus, reservoir releases need to also take into account the downstream riverbank stability and erosion issues. The study presented here investigates the riverbank stability and erosion at five study sites representing straight as well as inside and outside channel meander bends located on the lower Roanoke River near Scotland Neck, North Carolina. Extensive laboratory and field experiments were performed to define the hydraulic and geotechnical properties of the riverbank soils at each site. Specifically, soil water characteristic curves were determined using six different techniques and the results compared to existing mathematical models. Hydraulic conductivity was estimated using both laboratory and in situ tests. Due to the wide range of experimentally obtained values, the values determined by each of the methods was used for transient seepage modeling and the modeling results compared to the actual ground water table measured in the field. The results indicate that although the hydraulic conductivities determined by in situ tests were much larger than those typically reported for the soils by lab tests, numerical predictions of the ground water table using the in situ values provided a good fit for the measured ground water table elevation. Shear strengths of unsaturated soils were determined using multistage suction controlled direct shear tests. The test method was validated, and saturated and unsaturated shear strength parameters determined. These parameters, which were determined on the basis of results from both laboratory and field measurements, and the associated boundary conditions, which took into account representative flow rates and patterns including peaking, drawdown and step-down scenarios, were then utilized for transient seepage analyses and slope stability analyses performed using SLIDE, a software package developed by Rocscience. The analyses confirmed that the riverbanks are stable for all flow conditions, although the presence of lower permeability soils in some areas may create excess pore water pressures, especially during drawdown and step-down events, that result in the slope becoming unstable in those locations. These findings indicate that overall, the current reservoir release patterns do not cause adverse impacts on the downstream riverbanks, although a gradual drawdown after a prolonged high flow event during the wet season would reduce unfavorable conditions that threaten riverbank stability. / Ph. D.