APPLICATION OF SCS-CN TECHNIQUE TO RAINFALL, RUNOFF AND SEDIMENT YIELD DATA FROM PLOT SCALE WATERSHED

Abstract

The hill slopes in agro-forestry watersheds are characterized by high degree of soil macro porosity and during the monsoon seasons such areas frequently receive extreme rainfall which result into rapid overland and subsurface flows causing devastating flash floods in the rivers. The huge soil loss due to overland runoff and saturation-excess overland flow severely affects the soil health, agricultural productivity and livelihood of several dependent societies. The main objective of this study was to understand the process of runoff and soil erosion in plot scale watersheds having different slope, land use and land cover using the observed rainfall, runoff and sediment data using SCS-CN based runoff and sediment yield models. The study was carried out at the plot scale watershed located in Toda Kalyanpur, Roorkee, Dist. Haridwar, Uttarakhand, India. It lies in alluvial plain of River Ganga and is a fertile agricultural field. Primarily, three main plots having different slopes of 8%, 12% and 16% were prepared in this experimental field. Each plot was sub-divided into three sub-plots of 12mx3m size having different land uses such as: Maize, Finger Millet and Fallow Land. The types of soil of the field was determined by conducting double ring infiltrometer and were found to fall under respective hydrologic soil group ‘A’ and soil texture was determined with sieve analysis and found as sandy soil for all slopes and land uses. Daily rainfall, associated runoff, sediment concentration and soil moisture data were recorded for different land uses and slopes and were plotted to derive the best fit equations among these variables. A good linear relationship was found between rainfall and runoff, rainfall and soil loss and runoff and soil loss across different slopes and land uses. This experiment revealed significantly worst correlation between runoff and sediment concentration with R2 value in 8, 12 and 16% slopes as 0.001, 0.005 and 0.001 in Maize crop; 0,0.05 and 0.074 in Finger Millet; and 0.154,0 and 0.091 in Fallow Land. But at the same time, good correlation was found between runoff and soil loss with R2 value in 8, 12 and 16% slopes as 0.815, 0.641 and 0.594 in Maize crop; 0.895, 0.52 and 0.807 in Finger Millet; and 0.705, 0.64 and 0.964 in Fallow Land. This shows that the sediment concentration varies with other variables like rainfall intensity, runoff rates, soil type and moisture content. The result of regression analysis shows that the relative contribution of slope for runoff and soil loss is higher than soil moisture and rainfall iii amount in all land uses. Thus the effect of the rainfall on runoff and soil losses varies with land uses and slopes. This study evaluated the impact of three different slopes, viz., 8%, 12% and 16% on runoff generation process using different forms of SCS-CN based rainfall runoff models developed by Mishra et al. (2006) coupled with the slope adjusted models of Sharpley and Williams (1990) (SAF-1) Huang et al.2006 (SAF-2) and Ajmal et al. (2016) (SAF-3) using data from experimental field plots for varied initial abstraction coefficients λ, viz., 0.05, 0.1, 0.15, 0.2, 0.25 and 0.30 by using NSE, R2 and RMSE as a goodness of fit statistics. The slope-wise dataset shows that for rainfall-runoff model R3 in 8% slope highest value of NSE 0.87 was obtained at λ=0.25 for existing SCS-CN model and Huang et al. (2006) (SAF-2) and at λ=0.30 for Sharpley and Williams (1990)(SAF-1) and Ajmal et al.(2016) (SAF-3). Similarly, for 12% slope, highest efficiency 0.86 was found for existing SCS-CN and SAF-2 at λ=0.15 and 85% for SAF-1 and SAF-3 at λ=0.20 at the same time for 16% slope 0.80 efficiency found for existing SCS-CN and SAF-2 and 0.81 for SAF-1 and SAF-3 at λ=0.05.For R4 highest efficiency was obtained at different value of λ 0.30, 0.25 and 0.05 for 8, 12 and 16% respectively. An increase in initial abstraction coefficient λ results into decrease in R2, and vice versa for both models R3 and R4 for all slope-adjusted models at all slopes. Furthermore, lower value of R2 and higher value of RMSE found in 16% slope required further investigations in higher slope for precise estimation of runoff. Among different slope adjustment formulations, the performance of slope adjusted CN from Huang et al. (2006) was found better for both runoff models R3 and R4 for dataset of overall plot at λ=0.10 and 0.15with highest value of NSE and R2 and lower value of RMSE as 0.84, 0.92 and 6.13 for R3 and 0.84, 0.92 and 6.20 for R4 respectively. Huang et al. (2006) SAF-2 and Sharpley and Williams (1990) SAF-1 for rainfall runoff model R3 and R4 at λ=0.05 and 0.15 with least runoff coefficient variation 0.60% and -0.60% was found better among different slope adjustment formulations and λ. For rainfall runoff model R3 and R4 and low variation on runoff coefficient was found in runoff model R4 than R3 for λ=0.15 to 0.25. Finally, sediment yield was modeled using the SCS-CN based sediment yield models developed by Mishra et al. (2006) using the observed rainfall, runoff and sediment yield data. Similar to the runoff computations, here in this case also, the slope adjustments were made using the models of Sharpley and Williams (1990), Huang et al. (2006), Ajmal et al. (2016). The goodness of fit statistics was evaluated in terms of NSE R2 and iv RMSE. The values of NSE, R2 and RMSE during validation are found to be vary from 0.30-0.56, 0.57-0.82, and 0.21-0.16; 0.32-0.58, 0.57-0.82, 0.21-0.16; 0.28-0.57, 0.56- 0.82, 0.21-0.16; 0.30-0.58, 0.57-0.82, 0.21-0.16, respectively for the models of Sharpley and Williams (1990), Huang et al. (2006), Ajmal et al. (2016) for plot slope of 8%. The study also found that beyond 8% watershed slope, the applicability of all the four slope adjusted CN models in sediment yield simulations is not very promising and hence some improved models are still required for CN slope corrections in sediment yield applications. The result from plot wise data a good degree of correlation up to 12% slope, i.e., R2 value 0.68 and 0.89 was also found between the CNs derived from the rainfall-runoff models R2 and R5 and rainfall sediment yield models S2 and S5 but after slope adjustment good correlation was found for all slopes and Ajmal et al. (2016) was found better with R2 value 0.758. The initial abstraction ratio λ, slope and antecedent moisture content play an important role for sediment prediction. The study also shows that as λ is decreased from 0.30 to 0.05, the efficiency of sediment yield models is increased for all the models, and vice versa. Similar inferences were also drawn by Hawkins et al. (2001) and Singh et al. (2008). Similarly, the R2 values and RMSE are found to increase and decrease, respectively, for the models S3 and S4 for a decrease in λ value for all the eighteen storm events taken in this study.