FLOOD07616 - Laserfiche WebLink

Home Browse Search

, Table 4. AMS Data for Harvard Gulch Year Maximum Flow (cfs) 1981 785 1982 214 1983 488 1984 191 1985 81 1986 104 1987 372 1988 597 1989 181 1990 222 1991 471 1993 424 1994 746 1995 589 1996 1100 1997 776 1998 573 1999 257 Note: No AMS data available for 1992 There is an abundance of frequency analysis methods available to hydrologists. The most common is probably the Gumbel method. However, frequency statistics often do not fit the Gumbel regression. Therefore, the AMS data for Harvard Gulch were plotted using Gumbel frequency factors (Gumbel and log-Gumbel), exponential frequency factors (semi-log and log-log), and normal frequency factors (normal and log-normal). It was found that Gumbel frequency factors with a standard y-axis (flow) best described the AMS data (see graph 1) with a correlation (r) of 0.9667. The log-normal distribution of the AMS data was close with a r value of 0.9536 (see graph 2). The 2-, 5-, 10-, 50-, and 100-year 1-hour precipitation values published by the UDFCD were input into the CUHP program and the peak flow was plotted against the AMS data. The CUHP input variables were then varied to find the best correlation between the AMS actual regression and the CUHP predicted regression. While it was found that the AMS data, correlated well with the Gumbel distribution, the CUHP data produced from the UDFCD 1-hour rainfall data did not fit as well. The peak flow data based upon the UDFCD return period precipitation appears to have a distinctly exponential character to it. When plotted with log-normal distribution, the correlation coefficient for the 'predicted' CUHP model produced a very good fit of 0.9982 (see graph 3). Comparisons of all frequency types can be found in appendix A. Based on these findings, Application and Evaluation of CUHP Page 9 of 52