| description abstract | Recent studies propose different metrics for hydrologic design under nonstationarity, such as the effective return level, the expected waiting time (EWT)-based return level, the expected number of events (ENE)-based return level, the design life level (DLL), and the minimax design level (MDL). In this study, we formalize a method to test the credibility of such metrics in (1) developing precipitation intensity-duration-frequency relationships, (2) at-site design flood estimation, and (3) regional flood frequency analysis. The test relies on asymptotic normality assumptions and applies to the mean of the estimated return levels. Our results show that, based on historical records, point estimates or means of nonstationary design quantiles in all three applications are not significantly different from their traditional stationary counterparts when the associated uncertainties are considered. For example, in the application of at-site design flood estimation, although the estimated stationary 100-year flood is 32% and 29% lower in magnitude than the EWT- and ENE-based nonstationary 100-year return level, respectively, such a difference is not statistically significant. Further, enhanced model complexity is found to result in increased uncertainty in design levels under nonstationarity to at least twice the range obtained from a stationary analysis. | |
