Bulletin 17B Research Articles

Risk Factors for 100-Year Flood Events in the Mid-Atlantic Region of the United States.

Anecdotal information indicates that streams in the Mid-Atlantic region of the United States experience more extreme flood events than might be expected. This leads to the question of whether this is an unfounded perception or if these extreme events are actually occurring more than should be expected. If the latter is true, is this due solely to randomness, or alternately to characteristics that make certain watersheds more prone to repeated events that may be defined as 100-year or greater floods? These questions are investigated through analysis of flood events based on standard flood frequency analysis. 100-year streamflow rates for stream gages were estimated using Bulletin 17B flood frequency analysis methods, and the probability of the annual peak flow record for each gage was calculated. These probabilities were compared to a set of synthetic probabilities to evaluate their distribution. This comparison indicates that for the Mid-Atlantic region as a whole, the Bulletin 17B method does not systematically over or underestimate flood frequency. A Random Forest model of probability of actual flood record (PAFR) versus watershed and stream gage characteristics was developed and used to understand if certain characteristics are associated with PAFR. This analysis indicated that unexpected numbers of large flood events in a stream gage period of record can be attributed primarily to randomness, but there is some correlation with watershed and gage characteristics including weighted skew, drainage area, and mean annual peak discharge. The results indicate that watersheds with high values of these characteristics may warrant advanced flood frequency methods.

Bias correcting instantaneous peak flows generated using a continuous, semi‐distributed hydrologic model

Flood risk analysis in the United States follows the guidelines recommended in Bulletin 17B, which inherently ties the probability of exceeding a selected peak flow to the historical record. However, insufficient streamflow data, land use change, or climate change can render these recommended techniques impractical, forcing alternative measures. One such alternative is hydrological models which use climate and landscape data to simulate streamflow. Daily time steps or finer, however, are required to extract the necessary peak flow series. Accurately simulating flows at a daily time step can be challenging, and bias in prediction of hydrological extremes can be large. In this analysis, we compare three types of statistical bias‐correction measures for peak flow series data extracted from simulations using the Soil and Water Assessment Tool (SWAT), a widely used continuous, semi‐distributed hydrologic model. Results show that a correction method that only adjusts the mean of the peak flow series was the most robust. This efficacy is likely a result of persistent bias in watershed simulation due to uncertainty in calibration parameters and simplified watershed processes, suggesting that correcting for the mean of the peak flow series may be sufficient when applying bias corrections for flood risk analysis using hydrological models.

Examination of the Spencer-McCuen Outlier-Detection Test for Log-Pearson Type 3 Distributed Data

AbstractIdentification of outliers in flood records can be an important step in a robust flood frequency analysis procedure. Bulletin 17B includes the Grubbs-Beck test (with α=10%) as an objective criterion of whether the smallest observations in a flood record are outliers. The Spencer-McCuen test extends the Grubbs-Beck test to consider explicitly whether the three smallest observations are outliers in log-Pearson Type 3 (or equivalently Pearson Type 3) distributed samples with log-space skew coefficients between [−1,1], and three significance levels [1, 5, and 10%]. Presented here are Monte Carlo experiments evaluating the performance of the Spencer-McCuen test. When that test relies on the sample skew coefficient as an estimate of the population skew, the test generally fails to achieve the nominal significance levels. The same is true when used with a generalized (weighted) skew coefficient. Thus, the test will often be inappropriate if used as originally proposed. More fundamentally, when a proposed...

If Stationarity is Dead, What Do We Do Now?1

Galloway, Gerald E., 2011. If Stationarity Is Dead, What Do We Do Now? Journal of the American Water Resources Association (JAWRA) 47(3):563-570. DOI: 10.1111/j.1752-1688.2011.00550.x Abstract: In January 2010, hydrologists, climatologists, engineers, and scientists met in Boulder, Colorado, to discuss the report of the death of hydrologic stationarity and the implications this might have on water resources planning and operations in the United States and abroad. For decades planners have relied on design guidance from the Interagency Advisory Committee on Water Data Bulletin 17B that was based upon the concept of stationarity. After 2½ days of discussion it became clear that the assembled community had yet to reach an agreement on whether or not to replace the assumption of stationarity with an assumption of nonstationarity or something else. Hydrologists were skeptical that data gathered to this point in the 21st Century point to any significant change in river parameters. Climatologists, on the other hand, point to climate change and the predicted shift away from current conditions to a more turbulent flood and drought filled future. Both groups are challenged to provide immediate guidance to those individuals in and outside the water community who today must commit funds and efforts on projects that will require the best estimates of future conditions. The workshop surfaced many approaches to dealing with these challenges. While there is good reason to support additional study of the death of stationarity, its implications, and new approaches, there is also a great need to provide those in the field the information they require now to plan, design, and operate today’s projects.

Flood Frequency and Design Flood Estimation Procedures in the United States: Progress and Challenges

Design flood estimation procedures in the United States have traditionally focused on two primary methods: frequency analysis of peak flows for floodplain management and levee design; and deterministic, probable maximum flood (PMF) estimates for design of dams and nuclear facilities. Federal Agencies in the United States, including the Bureau of Reclamation, US Geological Survey and Army Corps of Engineers, are currently examining potential changes to these standard flood hydrology procedures. This paper presents overviews of some ongoing investigations and data collection studies to support potential changes in design flood estimation. For floodplain management, the current guideline is Bulletin 17B, which specifies the use of an LP3 distribution, method of moments and regional skew information. Potential improvements to Bulletin 17B currently under consideration are: (i) use of historical and paleoflood information; (ii) adjusting for low outliers; (iii) improved plotting positions; and (iv) confidence intervals. Ongoing testing results are presented, highlighting the expected moments algorithm. In contrast to well-established, deterministic (PMF) extreme flood estimates for dam safety, agencies are now moving toward risk-based techniques. The Bureau of Reclamation has developed and applied several methods in order to estimate extreme floods and probabilities for large dams. Techniques used to date are summarised, along with those being considered by other US agencies. Improvements to extreme flood databases that provide inputs, including extreme storms and probable maximum precipitation estimates, precipitation frequency and paleofloods, are ongoing. Some challenges to updating design flood methods and data, including institutional effects, national scale, research to operations and use of new technologies, are described.

Record Length Requirements for Annual Maximum Flood Series

A basic problem in evaluating the accuracy of an annual maximum series is knowing the sample size that will provide a specific level of relative accuracy. Current methods are inadequate and do not consider all of the necessary variables needed to describe the relationship between return period, record length, skew, and accuracy. The objective of this work was to provide a graphical approach that will enable a user to assess the expected accuracy in 2-, 10-, and 100-year magnitudes expected from a log-Pearson Type III analysis. A Monte Carlo simulation was used to determine the relative errors for the upper 80, 90, and 95% levels of confidence for the 2-, 10-, and 100-year floods. Analyses of the graphs show that accuracy decreases with increases in return period and skew. The results show good agreement to an example problem from Bulletin 17B.

Closure to “Regional Flood Estimations in Red River Using L-Moment-Based Index-Flood and Bulletin 17B Procedures” by Yeo H. Lim and Donovan L. Voeller

Closure to “Regional Flood Estimations in Red River Using L-Moment-Based Index-Flood and <i>Bulletin 17B</i> Procedures” by Yeo H. Lim and Donovan L. Voeller

Discussion of “Regional Flood Estimations in Red River Using L-Moment-Based Index-Flood and Bulletin 17B Procedures” by Yeo H. Lim and Donovan L. Voeller

Discussion of “Regional Flood Estimations in Red River Using L-Moment-Based Index-Flood and <i>Bulletin 17B</i> Procedures” by Yeo H. Lim and Donovan L. Voeller

Regional Flood Estimations in Red River Using L-Moment-Based Index-Flood and Bulletin 17B Procedures

L-moment-based index-flood method (LMIF) is used in a regional flood frequency analysis of the Red River of the North basin. The results are appraised with respect to the traditional methods based on the 1982 Interagency Advisory Committee on Water Data (Bulletin 17B) procedure. The L-moments approach detects the necessity of subdividing the basin into three main homogeneous hydrologic regions through hierarchical clustering techniques, regionalization, distribution fittings, and simulations. Each region has a unique index-flood growth curve. The dominant distributions include the Pearson Type III, the generalized Pareto, and the generalized extreme value. For a certain number of gauged sites, the quantiles obtained by the LMIF differ significantly from those obtained by pursuing the Bulletin 17B-regression procedures. Generally, the latter gives lower quantiles for most sites. Overall, the LMIF is effective in identifying homogeneous regions, coherent in statistical distribution fitting, and free from state boundary issues. The investigation also illustrates the flexibility of the method in allowing flood estimations for a special regulated region and a main stem region. Further studies on the use of the LMIF on a national or global scale are recommended.

Closure to “Log-Pearson Type 3 Distribution and Its Application in Flood Frequency Analysis. II: Parameter Estimation Methods” by V. W. Griffis and J. R. Stedinger

By extensive Monte Carlo analysis, the authors show that “a method of mixed moments estimator is the best estimator not employing regional skew information and, remarkably, is competitive with the Bulletin 17B estimator that employs reasonable skew information.” The discusser stumbled upon the idea that the log-Pearson type 3 LP3 distribution parameters could be estimated without using the skew coefficient by mixing the moments of log and real space after reading Bobee’s paper 1975 on the LP3 distribution in real space. With the freshly acquired academic background in hydrological frequency analysis Rao 1978 , writing the required FORTRAN programs and performing various evaluations were completed in a relatively short period. The study was a part of a wider investigation of frequency distributions for

Log-Pearson Type 3 Distribution and Its Application in Flood Frequency Analysis. III: Sample Skew and Weighted Skew Estimators

The accuracy of flood quantile estimates is constrained by the data available at a site. To improve the accuracy of quantile estimators, Bulletin 17B recommends combining the station skew with a regional skew using the inverse of their mean square errors (MSEs) as weights. While these weights can yield the minimum MSE skewness estimator, they do not provide the minimum MSE quantile estimators except when the true at-site skew is zero. In this paper, optimal weights which provide minimum MSE quantile estimators are derived. A Monte Carlo experiment illustrates the value of different weighting schemes and the value of using an informative regional skew. For reasonable values of the regional skew, the MSE of quantile estimators is reduced when the sample skew is combined with an informative regional skew. Modest improvements in the MSE of quantile estimates are obtained using optimal quantile weights rather than the MSE-skew weights. When the regional skew is actually very informative, there is a large loss ...

Flood Frequency Analysis in the United States: Time to Update

A large portion of the U.S. population, infrastructure, and industry is located in flood-prone areas. As a result, floods cause an average of nearly 140 deaths and cost roughly $6 billion annually excluding flooding caused by Hurricane Katrina, which cost $200 billion alone www.usgs.gov/hazards/floods/; USGS 2006 . The 1993 Midwest flooding along the Mississippi and Missouri Rivers caused $20 billion in damages. Furthermore, these estimates neglect the real costs associated with loss of personal possessions and shattered lives and communities. With Hurricane Katrina August 2005 and recent floods in the Northeast June 2006 , Ohio winter 2007 , Texas–Oklahoma June 2007 , and central England July 2007 , it seems like damaging floods have been in the news daily over the last few years. While engineers cannot stop floods from occurring, they should seek structural and nonstructural strategies to reduce the risk of large economic losses, social vulnerability, environmental damage, and loss of life IFMRC 1994 . Development of economically efficient and rational plans requires good estimates of the risk of flooding. In the United States, that computation is done following guidelines in Bulletin 17, for which the latest update, Bulletin 17B, was published in 1982 IACWD 1982 . That update includes the skew map published 30 years ago in the original Bulletin 17 WRC 1976 , and a list of areas needing additional research. Bulletin 17B has served the nation for over 30 years; it is a remarkable document that has withstood the test of time and use. However, given long-standing problems listed in the document, recent advances that address those problems, and the current national interest in flood risk, the time has arrived to update Bulletin 17B to maintain the statistical credibility of the guidelines and to provide accurate risk and uncertainty assessments.

Log-Pearson Type 3 Distribution and Its Application in Flood Frequency Analysis. II: Parameter Estimation Methods

A number of parameter estimation methods for the log-Pearson type 3 distribution have been explored in the hydrologic literature, including the method of moments (MOM) in both log space and real space, maximum likelihood estimators (MLEs), and the method of mixed moments (MXM). Several studies have compared MLE and MXM estimators to a so-called Bulletin 17B MOM estimator, but only marginal gains in accuracy are reported and the conclusions are often conflicting. This paper resolves these discrepancies. The observed performance of MLEs can depend critically on the convergence criteria and parameter constraints. Reasonable constraints on parameters can also improve the performance of log space MOM estimators. The method of mixed moments does very well in comparison to MOM and MLE in the absence of regional information. However, a Monte Carlo analysis demonstrates that log space MOM estimators with regional skew information as recommended by Bulletin 17B are likely to be more attractive, and the more precise...

Evolution of Flood Frequency Analysis with Bulletin 17

The current methodology recommended for flood-frequency analyses by U.S. Federal agencies is presented in Bulletin 17B. Bulletin 17 was first published in 1976, minor corrections were made in 1977 resulting in Bulletin 17A, which was later succeeded by Bulletin 17B published in 1982. Bulletin 17B sought to resolve ambiguities in the recommended procedures, including the treatment of low outliers, and the use of both regional skew and historical flood information. As the 25th anniversary of the publication of Bulletin 17B approaches, this paper provides a discussion and evaluation of the procedures that document recommends. The fields of hydrology and flood frequency analysis have substantially evolved since Bulletin 17 was first published. New techniques are now available which should become part of these standard procedures. A comparison is provided which demonstrates how the standard and weighted Bulletin 17B quantile estimators perform relative to alternative LP3 quantile estimators that also make use of regional information.

Log Pearson type 3 quantile estimators with regional skew information and low outlier adjustments

The recently developed expected moments algorithm (EMA) [Cohn et al., 1997] does as well as maximum likelihood estimations at estimating log‐Pearson type 3 (LP3) flood quantiles using systematic and historical flood information. Needed extensions include use of a regional skewness estimator and its precision to be consistent with Bulletin 17B. Another issue addressed by Bulletin 17B is the treatment of low outliers. A Monte Carlo study compares the performance of Bulletin 17B using the entire sample with and without regional skew with estimators that use regional skew and censor low outliers, including an extended EMA estimator, the conditional probability adjustment (CPA) from Bulletin 17B, and an estimator that uses probability plot regression (PPR) to compute substitute values for low outliers. Estimators that neglect regional skew information do much worse than estimators that use an informative regional skewness estimator. For LP3 data the low outlier rejection procedure generally results in no loss of overall accuracy, and the differences between the MSEs of the estimators that used an informative regional skew are generally modest in the skewness range of real interest. Samples contaminated to model actual flood data demonstrate that estimators which give special treatment to low outliers significantly outperform estimators that make no such adjustment.

Comparisons of two moments‐based estimators that utilize historical and paleoflood data for the log Pearson type III distribution

The expected moments algorithm (EMA) [Cohn et al., 1997] and the Bulletin 17B [Interagency Committee on Water Data, 1982] historical weighting procedure (B17H) for the log Pearson type III distribution are compared by Monte Carlo computer simulation for cases in which historical and/or paleoflood data are available. The relative performance of the estimators was explored for three cases: fixed‐threshold exceedances, a fixed number of large floods, and floods generated from a different parent distribution. EMA can effectively incorporate four types of historical and paleoflood data: floods where the discharge is explicitly known, unknown discharges below a single threshold, floods with unknown discharge that exceed some level, and floods with discharges described in a range. The B17H estimator can utilize only the first two types of historical information. Including historical/paleoflood data in the simulation experiments significantly improved the quantile estimates in terms of mean square error and bias relative to using gage data alone. EMA performed significantly better than B17H in nearly all cases considered. B17H performed as well as EMA for estimating X100 in some limited fixed‐threshold exceedance cases. EMA performed comparatively much better in other fixed‐threshold situations, for the single large flood case, and in cases when estimating extreme floods equal to or greater than X500. B17H did not fully utilize historical information when the historical period exceeded 200 years. Robustness studies using GEV‐simulated data confirmed that EMA performed better than B17H. Overall, EMA is preferred to B17H when historical and paleoflood data are available for flood frequency analysis.

Seasonal flow frequency analysis

Numerous hydrologic designs require seasonal discharge estimates. Given the seasonal variation in the distribution of rainfall, significant differences can exist between the annual maximum and seasonal maximum T-yr return period discharges. Recommendations for making seasonal flood frequency analyses for gauged and ungauged locations are presented. When performing a seasonal frequency analysis of gauged data, missing discharge data will generally be a problem. Two potential solutions to the problem of incomplete records are introduced. First, a maximum likelihood approximation that replaces missing discharges below a threshold is developed and tested. Results of simulations indicate that it is more accurate than the method demonstrated in Bulletin 17B. Second, the ratio of the measured instantaneous to mean daily discharge is regionalized. This provides a method of replacing missing, below threshold, discharges. The ratio of measured instantaneous discharges above a threshold and the corresponding mean daily discharges can be used for predicting missing instantaneous discharges in seasonal records. Once regionalized, the method can be used for developing seasonal frequency curves at ungauged locations with either the USGS regression equations or a rainfall-runoff model. The Eastern Coastal Plain data are used to demonstrate the development of regional index ratios.

Data-based comparisons of moments estimators using historical and paleoflood data

This paper presents the first systematic comparison, using historical and paleoflood data, of moments-based flood frequency methods. Peak flow estimates were compiled from streamflow-gaging stations with historical and/or paleoflood data at 36 sites located in the United States, Argentina, United Kingdom and China, covering a diverse range of hydrologic conditions. The Expected Moments Algorithm (EMA) and the Bulletin 17B historical weighting procedure (B17H) were compared in terms of goodness of fit using 25 of the data sets. Results from this comparison indicate that EMA is a viable alternative to current B17H procedures from an operational perspective, and performed equal to or better than B17H for the data analyzed. We demonstrate satisfactory EMA performance for the remaining 11 sites with multiple thresholds and binomial censoring, which B17H cannot accommodate. It is shown that the EMA estimator readily incorporates these types of information and the LP-III distribution provided an adequate fit to the data in most cases. The results shown here are consistent with Monte Carlo simulation studies, and demonstrate that EMA is preferred overall to B17H. The Bulletin 17B document could be revised to include an option for EMA as an alternative to the existing historical weighting approach. These results are of practical relevance to hydrologists and water resources managers for applications in floodplain management, design of hydraulic structures, and risk analysis for dams.

Generalized Flood Skew: Map versus Watershed Skew

The skew coefficient is an important statistic in flood estimation at gaged sites, but sample values are subject to considerable sampling variation, and the map skew lacks both a conceptual basis and acceptable accuracy. While weighting the sample skew with the Bulletin 17B map skew can temper the variation inherent to sample skew coefficients, the inaccuracy of the map suggests that the weighted value of skew may still not have either the desired accuracy or the accuracy that would be possible if a more realistic generalized skew were used in place of the map skew. A new generalized skew coefficient is proposed herein, namely watershed skew. It has a conceptual basis in its relationship to the skew of the rainfall that generates the floods and the watershed and channel storage that transform rainfall skew to the skew of the runoff. Weighting the sample skew with the watershed skew will provide an estimate of the skew coefficient that incorporates the effects of the physical processes inherent to measured...

Confidence intervals for expected moments algorithm flood quantile estimates

Historical and paleoflood information can substantially improve flood frequency estimates if appropriate statistical procedures are properly applied. However, the Federal guidelines for flood frequency analysis, set forth in Bulletin 17B, rely on an inefficient “weighting” procedure that fails to take advantage of historical and paleoflood information. This has led researchers to propose several more efficient alternatives including the Expected Moments Algorithm (EMA), which is attractive because it retains Bulletin 17B's statistical structure (method of moments with the Log Pearson Type 3 distribution) and thus can be easily integrated into flood analyses employing the rest of the Bulletin 17B approach. The practical utility of EMA, however, has been limited because no closed‐form method has been available for quantifying the uncertainty of EMA‐based flood quantile estimates. This paper addresses that concern by providing analytical expressions for the asymptotic variance of EMA flood‐quantile estimators and confidence intervals for flood quantile estimates. Monte Carlo simulations demonstrate the properties of such confidence intervals for sites where a 25‐ to 100‐year streamgage record is augmented by 50 to 150 years of historical information. The experiments show that the confidence intervals, though not exact, should be acceptable for most purposes.