USGS Publications Warehouse

Abstract

Techniques are presented for generalizing the skewness coefficient of log-Pearson Type III distributions of annual maximum discharges and for flood magnitudes that have selected recurrence intervals from 2 to 100 yr. A weighted least-square (WLS) regression model was used to generalize the coefficients of station skewness that resulted in a root-mean-sq error of prediction of 0.35 compared to 0.55 for the skewness map published in Bulletin 17B of the U.S. Water Resources Council. Estimates of generalized skewness were computed for each of 245 streamflow gaging stations with a minimum of 10 years of record and a contributing drainage area of < 20,000 sq mi. The WLS regression model also was used to develop equations for estimating flood magnitude for selected recurrence intervals for ungaged stream locations by using data from 218 of the 245 streamflow gaging stations that had contributing-drainage areas of less than 10,000 sq mi. The errors of prediction of the most reliable WLS equations ranged from 28 to 42%. The WLS equations were compared statistically to previous developed equations and were determined to be different and more accurate than previously published equations. Flood magnitudes and frequencies for 245 streamflow gaging stations, based on data collected through the 1983 water year, are presented along with a summary of the seasonal distribution of annual maximum discharges and an analysis of the maximum observed discharges. (Author 's abstract)