Streamflow records in the southeastern United States are comparatively short (<100 years), resulting in higher uncertainties associated with estimates of extreme flood frequencies. Extreme floods, for the purposes of this thesis, include floods having an annual exceedance probability of 0.01 or smaller (i.e., a recurrence interval greater than or equal to 100 years). Uncertainty of flood frequency models can be reduced by including paleoflood hydrologic data, in which flood data are reconstructed from analyses of sediment deposits or tree-ring records. In this thesis, a paleoflood chronology was developed for the French Broad River, a major tributary of the Tennessee River, based on sampling and analyses conducted at an upstream floodplain (DT3A) site and a downstream rock shelter (DRS) site. A sediment core taken from DT3A was analyzed using laser granulometry to determine particle size of each centimeter of the core. Large flood events were identified based on the volume of sand and statistical analyses were used to eliminate sedimentological changes stemming from factors other than floods, using end member modeling and change point analysis. LOESS regression modeling was used to identify discrete, large magnitude paleofloods based on peaks (positive residuals) in sand volume. Paleoflood discharges were estimated using step-backwater modeling in HEC-RAS based on paleostage indicators found at DRS. Stratigraphic descriptions and floods ages (determined with radiocarbon dating) were used to identify three paleoflood events preserved at two different topographic levels at DRS. The paleoflood discharges and age information were combined with the modern gauge record and a new flood record for the French Broad River was developed that extends an additional 250+ years. Comparing the reconstructed chronology of flood events, to similar records in the southeastern United States, it appears that extreme flood events coincide with climate transitions.