Impact of Streamflow Persistence on Hydrologic Design

Abstract

Conventional methods for estimating the average return period, $E\left(T\right)\text{,}$ and failure risk, R, generally ignore the impact of persistence in annual streamflows on associated probabilistic statements and streamflow statistics. Recent evaluations of streamflow observations indicate statistically significant serial correlations (persistence) associated with annual low flows in the United States. We define the average occurrence interval $E\left(T\right)$ as the expected time to the first event, and we present a method for estimating $E\left(T\right)$ and R in the presence of persistence. We show that for observed ranges of persistence, $E\left(T\right)$ can be nearly 100% greater and R more than 20% lower than conventional estimates. This implies that the expected design life of a system is longer when persistence is taken into account. Also, by ignoring persistence, low flow quantiles may be underestimated by 50% or more. An evaluation of the effect of persistence on drought risk estimation across the U.S. is presented.