Design metrics for extreme events in a non-stationary world. [Abstract oly]

Society is increasingly concerned with how climate, land use and other anthropogenic influences are impacting their watersheds. Traditional probabilistic approaches for defining risk, reliability and return periods under stationary hydrologic conditions assume that extreme events are serially independent with a probability distribution whose moments and associated parameters are fixed. However, when non-stationary conditions lead to trends in the moments and parameters of extreme value processes, new methods for understanding the impacts of such changes on traditional design metrics are needed to insure sensible planning and design efforts. We document the general behavior of various metrics of return period, risk, and reliability assuming extreme events follow a nonstationary lognormal distribution. Our results provide guidance on the value, application and caveats associated with such metrics for water resources planning in a nonstationary world.