Robust National Flood Frequency Guidelines: What Is an Outlier?

The Interagency Advisory Committee on Water Data is finalizing proposed updates to the current national flood frequency guidelines, known as Bulletin 17B. Among the updates is a new multiple Grubbs-Beck test (MGBT) for low outliers. It is a generalization of the Grubbs-Beck (GB) test recommended in Bulletin 17B. Bulletin 17B defines low outliers as unusually small observations "which depart significantly from the trend in the rest of the data." Such unusually small values might better be labeled as Potentially Influential Low Flows (PILFs). PILFs can have a large influence on estimates of extreme flood quantiles; that is an undesirable situation if flood frequency procedures are to be robust. This paper reports on Monte Carlo analyses of the performance of low-outlier/PILF tests, including the Bulletin 17B GB test and variations of a MGBT. It also considers the accuracy of log-space method-of-moments flood quantile estimators that use different PILF identification procedures.