A comparative study of the estimation methods for NRCS curve number of natural arid basins and the impact on flash flood predications

Abstract

There are various methods to estimate the curve number (CN) for flood studies. In the ungauged basins, hydrologists rely on the use of an NRCS-CN table called (CNdesign). The CNdesign, in this study, is estimated using remote sensing techniques and geographical information systems based on alluvium-rock-vegetation classification of natural basins. However, in gauged basins, it is common to use rainfall-runoff data through the application of the least-squares method (LSM) to get the best CN value (CNobs), or the asymptotic fitting method (AFM) to obtain asymptotic CN (CN∞). A comparison between these methods is made under the effect of changing both the coefficient of abstraction ratio, λ, and the effect of data sorting techniques to find out the best estimation of CN for reliable prediction of floods. A methodology has been developed to convert the NRCS-CN table values at λ = 0.2 to λ = 0.01 for arid basins. The relationship between the observed CN and the NRCS-CN table shows that estimating runoff using λ = 0.2 is best made by CN of the impervious area (CNimp = 90) instead of 98 (for urban areas) used in the literature. The highest value of CN between the methods is the CNdesign, then CNobs. CN∞ shows the lowest value. Therefore, for a safe design of the hydraulic structures, it is recommended to use CNdesign. However, for the simulation of the rainfall (P)-runoff (Q) process in the natural basins, it is recommended to use CNobs at the natural sorting of data pairs (P: Q). The root mean square error (RMSE) of CN is reduced from 11 at CNimp = 98 to 7 at CNimp = 90. This value reflects the infiltration process in the rocks due to the high density of fractures and fissures in the mountains in the area. The developed NRCS-CN table at λ = 0.01 reduces the RMSE of the estimated runoff depth by 57%.