1 Value of Regional Information using Bulletin 17B and LP3 Distribution Veronica W. Griffis 1 and Jery R. Stedinger 2 1 Assistant Professor, Department of Civil & Environmental Engineering, Michigan Technological University, 1400 Townsend Drive, Houghton, MI 49931-1295; Phone: 906-487-1079; Fax: 906-487- 2943; Email: vgriffis@mtu.edu 2 Professor, School of Civil & Environmental Engineering, Cornell University, Hollister Hall, Ithaca, NY 14853-3501; Phone: 607-255-2351, Fax: 607-255-9004; Email: jrs5@cornell.edu To improve the accuracy of quantile estimators, Bulletin 17B recommends a number of procedures to improve at-site estimators using regional information. Several procedures, including two recommended by the Bulletin, are considered here. Because the data available at a site is generally limited, the skewness estimator can be particularly unstable. When fitting the LP3 distribution, Bulletin 17B recommends combining the station skew with a regional skew using the inverse of their mean square errors as weights. Previous studies have demonstrated the impact of a more precise regional skewness estimator on quantile estimator precision. To improve quantile estimates computed using short records, the Bulletin also suggests combining the at-site quantile estimate with a regional quantile estimate using their effective record lengths as the weights. Potential problems with this weighted estimator are discussed here. Two examples compare the precision of the Bulletin 17B weighted quantile estimator to several alternative estimators which employ different combinations of at-site and regional information, including an index flood procedure which did poorly. The simple Bulletin 17B weighting of at-site and regional regression quantile estimates performs nearly as well as more complex alternatives, and for short records provides a substantial improvement in quantile accuracy. However, when the regional standard deviation and skew are very informative and the regional mean estimator is relatively imprecise, more accurate estimates can be obtained by weighting each of the three sample moments separately with regional estimators of those same statistics. Introduction A large portion of the U.S. population, infrastructure, and industry is located in flood prone areas. Floods cause an average of nearly 100 deaths and cost roughly $2.3 billion annually in the U.S. Accurate estimates of the magnitude and frequency of flood flows are needed for the design of water-use and water-control projects, for floodplain definition and management, and for the design of transportation infrastructure such as bridges and roads. Unfortunately, the accuracy of flood quantile estimates are constrained by the data available at a site: record lengths are often limited to 100 years, and are typically less than 30 years. This paper considers the use of regional skew information and regional quantile estimates to improve the accuracy of at-site quantile estimates. World Environmental and Water Resources Congress 2007: Restoring Our Natural Habitat © 2007 ASCE