COMPARATIVE EVALUATION OF CN DETERMINING METHODS AND NRCS-CN BASED MODELS FOR RUNOFF ESTIMATION

Abstract

United States Department of Agriculture (USDA) Soil Conservation Service (SCS) (now NRCS) Curve Number (CN) procedure is a simple, popular and single parameter based enduring, ubiquitous, versatile and widely used method to compute event wise rainfall-runoff depths from small agricultural watersheds (Hawkins et al., 2014). The spatial and temporal variability of rainfall and qualitative observed or measured data play important roles in deciding Curve Numbers. CN is also dependent on runoff-producing watershed characteristics i.e. soil type, land use, watershed slope, hydrologic condition and antecedent soil moisture condition (AMCs) (Mishra et al.,1999 & 2006). Computing accurate CNs is always an important aspect (Singh et al.2010; Mishra & Singh 2006) for having improved runoff estimates using NRCS-CN method. The watershed slope and initial abstractions play vital roles in the improved prediction of runoff depths from rainfall can also reduce the predictive capability of a CN based runoff model, if not incorporated in the model processes. In this study, three mathematical models (M1, M2 & M3) with different architecture were developed for computing accurate CNs incorporating watershed slope as well as initial abstraction and a comparative evaluation of the existing slope adjusted models of Ajmal et al. (2016), Sharpley and Williams (1990) and Huang et al. (2006) was performed considering the event wise rainfall-runoff data of experimental catchment. For this, 36 different combinations of watershed slope, initial abstraction and land use were formulated. The experimental watershed was having three slopes, i.e., 16%, 12% and 8% and three type of land uses: Fallow, Raagi and Maize. The initial abstraction coefficient was allowed to vary as 0.05, 0.10, 0.20, and 0.30. The optimized parameters values of the proposed models were estimated using Goal-seek and Solver programming in MS-EXCEL. The NRCS-CN models based on the proposed CN estimation models performed much better than the slope based existing NRCS-CN models. In all the 36 applications, model M3 is found to perform best followed by M6 and other models and the model developed by Ajmal et al. (2016) is found to be the least performing. The watershed response behavior iii to the storm rainfall depth was also investigated using asymptotic CN estimation method and was found to be Complacent in nature. NRCS-CN based Soil and Water Assessment Tool (SWAT) model was also applied to simulate runoff for Ong sub-basin of Mahanadi basin. Total 900 simulations were executed for the 9 parameters to decide their new limits. The new values of the parameters were then utilized for the validation part. Model calibration and uncertainty analysis were performed with SUFI-2 (Sequential Uncertainty Fitting Ver. 2) programming using SWAT-CUP. The model was found to perform satisfactorily in calibration and validation processes for runoff simulation with the coefficient of determination (R2) value of 0.51 and 0.60 respectively. The parameters like CN, GWQMN, ESCO, SOL_AWC, GW_DELAY and GW_REVAP were found more sensitive for Ong sub-basin than Alpha_BF, RECHRG_DP and REVAPMN.