Design floods for most dams and levees typically have an annual exceedance probability (AEP) of 1:100 (1E-2) or less frequent. In the U.S., high hazard dams are designed to pass the Probable Maximum Flood (PMF), which typically has an AEP of 1:10,000 (1E-4) or less frequent. In order to reduce epistemic uncertainties in the estimated AEP for extreme floods, such as the PMF, it is important to incorporate as much hydrologic information into the frequency analysis as reasonably possible. This paper presents a Bayesian analysis framework, originally profiled by Viglione et al. (2013), for combining at-site flood data with temporal information on historic and paleofloods, spatial information on precipitation-frequency, and causal information on the flood processes. This framework is used to evaluate the flood hazard for Lookout Point Dam, which is a high priority dam located in the Willamette River Basin, upstream of Portland, Oregon. Flood frequency results are compared with those from the Expected Moments Algorithm (EMA). Both analysis methods produce similar results for typical censored data, such as historical floods; however, unlike the Bayesian analysis framework, EMA is not capable of incorporating the causal rainfall-runoff information in a formal, probabilistic manner. Consequently, the Bayesian method considered herein provides higher confidence in the fitted flood frequency curves and resulting reservoir stage-frequency curves to be used in dam and levee safety risk assessments.
$15.00
ANCOLD is an incorporated voluntary association of organisations and individual professionals with an interest in dams in Australia.