Comment on “Assessing flood hazard on dynamic rivers”

N. Pinter, R. Thomas, and J. H. Wlosinski (Eos, 31 July 2001) propose an alternative to the U.S. Army Corps of Engineers' methodology for estimating stage-frequency curves in the Upper Mississippi River System Flow Frequency Study (UMRSFFS). They suggest that the stage-frequency relationship can be obtained through direct analysis of the period of record stages adjusted for trends and present a stage-frequency curve for St. Louis (Fig. 5, p.339). However, the direct analysis of stages proposed by these authors produces erroneous results, and their results will undoubtedly differ significantly from the stage-frequency curves that will be produced by the UMRSFFS. Their methodology is not acceptable as an alternative for flood risk estimation because it: (1) uses an at-site regression estimator of the flood distribution, which is subject to significant sampling error and cannot easily be used to infer the underlying flood population; (2) does not account for the effect of reservoir regulation and other human activities on the flood flow record; (3) assumes that the stage-discharge data have been uniformly recorded; (4) assumes the stage non-stationarity can be represented by a simple linear trend model that assumes that channel changes have occurred uniformly over time; and (5) ignores many factors that cause a large natural variability in the stage-discharge relationship.