A CRITICAL INVESTIGATION OF SCS-CN METHODOLOGY FOR RAINFALL-RUNOFF MODELLING

Abstract

Modelling surface runoff is of paramount importance to tackle issues arising due to scarcity as well as abundance of water and to adequately manage the water-dependent societal requirements. The Soil Conservation Service-Curve Number (SCS-CN) method is an age-old, globally popular method for estimating direct surface runoff resulting from incident rainfall. In spite of its widespread applications, the CN method shows certain structural inconsistencies and deficiencies that require attention. It shows abrupt changes during runoff computations due to sudden jumps in CN values for the changing antecedent moisture condition (AMC). Additionally, applying CN look-up tables with a fixed initial abstraction coefficient (λ) and without slope adjustments results in severe underestimation of surface runoff in steep watersheds. Furthermore, there are areas where the usage of the CN method is limited or unexplored, such as associating CN with the return period and duration of large storms (t > 1 day) to predict design flood and coupling the SCS-CN method with the Universal Soil Loss Equation (USLE) to estimate design sediment yield for designing water and soil conservation systems. Therefore, it is imperative to address the aforementioned issues and develop/modify the new/existing approaches for improving the efficiency of the CN methodology. A simplistic novel approach was developed to account the antecedent moisture in the original SCS-CN method to prevent abrupt jumps in the CN values for runoff estimation and to accurately derive the CN values for watersheds. Using the CNs derived from observed events, the proposed model accurately predicted runoff in 36 US watersheds showing higher NSE (mean = 0.53, IQR = 0.37-0.63) and RMSE (IQR = 5.25 to 9.52 mm) values. Also, the proposed model outperformed the existing four CN models in the Godavari sub-basins when RS-based CNs were used for runoff prediction, achieving the highest NNSE of 0.642 and the lowest RMSE and MAE values of 8.98 mm and 6.64 mm, respectively. Further, the proposed model exhibited the best correlation between CN and its associated rainfall (highest R2 values among the tested models) in 36 US watersheds, substantiating the rationality of the model. The empirical moisture content showed a significant correlation with the in-situ water content indicating the potential of the CN methodology to accurately estimate the soil moisture.