Submitted to: Journal of Irrigation and Drainage Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: June 6, 1996
Publication Date: N/A
Interpretive Summary: Engineers and hydrologists worldwide use the runoff-volume-estimating procedure called the "curve number method". Curve numbers (CN) reflect the ability of a watershed to generate runoff, with high CN associated with high runoff-producing watersheds, and vice versa. There is no standard method for developing a CN from measured rainfall and runoff data. A method is proposed that uses frequency distributions of measured data to develop CN ("derived-distribution method", DD). The DD method was shown to yield less variable estimates than 2 other methods in a Monte Carlo simulation, which generated "data" with known characteristics. The DD method gave CN estimates, using measured data, that were in agreement with a third method. The third method shows the trend in CN with high rainfall values for the two known types of watershed responses to rainfall. Prior knowledge of the type of watershed response is needed with the third method, while the DD method does not need this information. However, the DD method can also provide the watershed response type. A set of guidelines was established for practical usage based on the study results. The results are useful for developing CN where data are scanty, as well as for determining CN from measured data. The results can be used by scientists, engineers, the NRCS, and other hydrologic practitioners for practical estimation of CN.
Technical Abstract: Curve numbers (CN) are developed from measured rainfall and runoff data, yet there is no standard method for determining CN. The original method of CN determination used the maximum annual events. Subsequent development treated measured P and Q data as frequency distributions. In this paper, derived frequency distributions are evaluated as another method for determining watershed curve numbers from watershed data treating P and Q data as separate frequency distributions. A Monte Carlo simulation showed the derived-distribution method gave less variable estimates of CN for a wide range of sample sizes than two other methods for CN estimation. CN estimates using measured data with the derived-distribution method agreed well with CN estimated by the asymptotic method for the watershed types tested. The derived-distribution and asymptotic methods were in agreement while the traditional methods gave lower CN estimates for the "violent" watershed type. CN was indeterminate for the derived-distribution and asymptotic methods for "complacent" watersheds. The derived-distribution method has potential for determining curve numbers when there are limited P and Q data.