Statistical Guidelines for Curve Number Generation

The accuracy of runoff curve numbers (CNs) is unknown, and empirical evidence has suggested that with the current CN table, hydrological systems are being over designed—which some believe is due to a variable known as the initial abstraction coefficient $\left(\lambda \right)$ . Therefore, guidelines to produce a new CN table are needed. In order to develop this set of guidelines, the following objectives were met: Determination of CN and $\lambda$ sensitivity, assessment of the accuracy of alternative methods for determining CNs, examination of the effect of the range of data, and evaluation of the potential change in hydrologic design due to a new table. Using measured and simulated data, the methods of estimating the CN were evaluated and assessed for accuracy. A new method using concepts of lognormal frequency was developed and found to be more accurate than the current practices. It was also found that $\lambda$ is an insensitive variable when compared to the CN, and therefore, may complicate the optimal fitting of the CNs. Finally, it was determined that developing a new CN table would affect the estimation of peak discharge rates, and thus hydrologic designs. Therefore, it may be advantageous to develop a new CN table based on peak discharge measurements rather than depths of rainfall and runoff. Guidelines that should lead to a revised CN model with improved accuracy are provided.