FLOOD05157 - Laserfiche WebLink

Home Browse Search

}6 TECHNICAL MANUAL ACCURACY OF RESULTS The accuracy of a regression equation is usually expressed as the stand- ard error of estimate. It is a measure of how well the observed data agree with estimates from the regression equations. The standard error of estimate, in percent, is the range of error found at about two-thi rds of the sample sites. Stated another way, two out of every three observations for a large sample of data would fall within the specified percentage. The standard error of estimate for each equation is I isted in' tables 1-4. Accuracy can also be expressed as equivalent years of record (Hardison, 1971) that represent the number of actual years of streamflow record needed to provide an estimate of equal accuracy as that of the regression estimate. Va 1 ues of equ I va I ent yea rs of record were computed for each peak d i scha rge re- gression equation listed in tables 1-4, but the computed values for the Plains Region were extremely large and the values for the Mountaiin Region and South- ern Plateau Region were abnormally small. The U.S. Water Resources Council (1975) recommends that a value of 10 years be assumed for equivalent years of record. For cons i stency, a va 1 ue of 10 yea rs \Vas used for equ i va 1 en t yea rs of records for all peak discharge regression equations in the four flood regions. ANALYTICAL DEVELOPMENT OF METHODS The methods described in this manual were developed from a flood- frequency analysis of gaging-station data and a multiple-regression analysis of flood characteristics and basin and cl imatic parameters of 25B gaged basins in Colorado and adjacent States. These flood data are given in tables 5-B. The length of record at all gaging stations used was at least 10 years, with 99 stations having records equal to or greater than 25 years in length. Each station record was screened for significant effects of diversion or regulation. Several of the stations which are now regulated have natural-flow parts of record of 10 years or more. These unregulated parts of record were used to develop natural-flow flood-frequency characteristics for the streams. This infomation, although useful in defining regional flood characteristics, is not applicable for the gaged sites under present-day conditions. A foot- note is provided in tables 5-B for each gaging station in this category. FLOOD-FREQUENCY RELATIONS For a stream where gaging-station records are available, a flood- frequency relation can be defined by fitting the array of annual maximum dis- charges to a theoretical statistical distribution. The U.S. Water Resources Council (1967) recommended a uniform technique for determining flood..flow frequencies by fitting the logarithms of the annual maximum discharges to a log-Pearson Type III distribution. This procedure, generally accepted by most Federal and State agencies, is referred to as a log-Pearson Type I I I frequency analysis.