For the past 36 years, Bulletin 17B, published by the Interagency Committee on Water Data in 1982, has guided flood-frequency analyses in the United States. During this period, much has been learned about both hydrology and statistical methods. In keeping with the tradition of periodically updating the Bulletin 17B guidelines in light of advances in our understanding and methods, the Hydrologic Frequency Analysis Work Group (HFAWG) was charged by the Subcommittee on Hydrology (SOH) of the Advisory Committee on Water Information (ACWI) to consider possible updates to Bulletin 17B.
The purpose of this report is to consider the statistical performance of possible revisions to Bulletin 17B procedures. Of particular interest are procedures designed to accommodate more general forms of flood information. The concern is how the proposed procedures would affect the precision, accuracy and robustness of flood-frequency estimates. The investigations reported here focus on techniques for the following:
- incorporating information related to historical flooding that occurred outside the period of systematic streamgaging; and
- identification of potentially influential low floods (PILFs).
The proposed changes, which mostly involve generalizing Bulletin 17B’s method-of-moments procedures by using the Expected Moments Algorithm (EMA), are relatively modest, at least in the sense that they would not affect the main features of Bulletin 17B. The proposed methods include the following:
- continued use of the log-Pearson Type 3 (LP3) distribution;
- continued use of the Method-of-Moments fitting method applied to the logarithms of annual-peak-flow data; and
- a generalization of the Grubbs-Beck test used in Bulletin 17B to identify low outliers. The new multiple Grubbs-Beck test is sensitive to multiple PILFs.
The hydrological literature already provides extensive support for the theory behind the proposed changes. The remaining question is practical: How well do the proposed methods perform under typical and realistic conditions and, specifically, with difficult records occasionally encountered in practice? In order to answer these questions, the HFAWG commissioned the work reported here. The following four major sets of results are provided:
- Monte Carlo simulations of fitting procedures employing data drawn from simulated LP3 populations;
- Monte Carlo simulations of fitting procedures employing data drawn from non-LP3 populations that were selected to reflect likely deviations of flood series from LP3 distributions, based on the experience of HFAWG members;
- a direct comparison of the fitted LP3 distributions for 82 real “test sites” identified by an independent data group as both “typical” and “challenging” for flood-frequency estimation; and
- simulations of fitting procedures using records obtained by resampling with replacement from the longest of the 82 test-site records.
Collectively, these studies provide a reasonably comprehensive, valid, and robust assessment of the properties of the Bulletin 17B methods and proposed alternatives. The experiments and analysis indicate that the flood quantile estimators, proposed as a revision of Bulletin 17B, do the following:
- perform generally as well as, and in some cases much better than, Bulletin 17B estimators in terms of the mean square error of flood quantiles estimates;
- allow for incorporation and efficient statistical treatment of broader classes of flood-frequency data and information, including historical information, binomial data and interval data; and
- generally confirm studies and the theoretical findings reported in the hydrological literature that would support use of updated estimation procedures that have been developed since Bulletin 17B was published.
|Title||Evaluation of recommended revisions to Bulletin 17B|
|Authors||Timothy A. Cohn, Nancy A. Barth, John F. England, Beth A. Faber, Robert R. Mason,, Jery R. Stedinger|
|Publication Subtype||USGS Numbered Series|
|Series Title||Open-File Report|
|Record Source||USGS Publications Warehouse|
|USGS Organization||Office of Surface Water|