Sangeeta Verma

This paper further modifies soil conservation service curve number (SCS-CN) based on the concept of adjusting the rainfall in accordance with rain duration and considering the initial abstraction (Ia) as a fraction of rainfall for runoff estimation. The former yields Model M3 and its explicit form with constant parameter λ = 0.2 is designated as Model M4. Model M5 couples both the concepts and thus all these models are the advanced versions. The applicability of all the five models is tested using a large number of rainfall-runoff events (25,502) derived from 53 U.S. Department of Agriculture-Agricultural Research Service watersheds. Models M3–M5 performed better than Models M1 and M2. Model performance is evaluated by employing six statistical measures, namely, root mean square error, mean absolute error, normalized root mean square error, Nash–Sutcliffe coefficient (%), percent bias, RSR, n(t), and several grading criteria. Results show Model M5 to have performed the best of all in both calibration and validation largely due to its incorporating the impact of rain duration and allowing Ia to vary with rainfall, which is close to reality and not accounted for in any other models considered in this study.

The need of incorporating storm intensity or duration in Soil Conservation Service Curve Number (SCS-CN) methodology for improved direct surface runoff estimation for a watershed has been highlighted by many engineers and hydrologists since long and despite this fact, it is still poorly explored. Therefore, this study aims to present storm duration-based improved SCS-CN models for estimating more accurate direct surface runoff from rainfall events. The accuracy and consistency of improved models are tested on a large rainfall-runoff dataset (18,660 rainfall events) derived from 39 watersheds in the USDA-ARS. Furthermore, the quantitative model’s performance is also evaluated employing six widely accepted statistical measures viz. root mean square error (RMSE), mean absolute error (MAE), normalized root mean square error (NRMSE), Nash-Sutcliffe efficiency (NSE), percent bias (PBIAS), observations standard deviation ratio (RSR), and several grading criteria. These models are compared with the original SCS-CN model (M1) and its simple form (M2) with traditionally fixed initial abstraction ratio (λ) as 0.2. The resulting lowest values of RMSE, NRMSE, MAE, NSE, PBIAS and highest values of RSR and ranking grading system (RGS) for the proposed models (M3-M5) reveal that their performance is better than M1 and M2 models. The proposed M5 model incorporating both storm duration and varying initial abstraction (Ia) as a certain percentage of rainfall, performed the best followed by M3 incorporating only storm duration. According to RGS, M5 also ranked first with the highest marks (195) followed by M3 (140). Due to high accuracy in predicted runoff, M5 can be recommended for both small and large watersheds as it overcomes the following issues: fixed λ (=0.2), assumption of constant rainfall intensity (time-independent), fixation of Ia at 2% of rainfall and applicability to only small watersheds, restricting the application of original SCS-CN and its modified versions.

Drought is one of the most widespread global environmental problems leading to low water availability for plants, which causes a significant loss in growth, productivity and finally their yields.