Improved Runoff Estimation Using Event-Based Rainfall-Runoff Models

Abstract

Event-based rainfall-runoff models are effective tools in operational hydrological forecasting and preparedness for extreme events. In the current study, the popular Natural Resources Soil Conservation curve number (NRCS-CN) model and the proposed simple nonlinear models were employed for runoff estimation. The runoff prediction capability of the NRCS model for the CN values obtained from tables was very poor in comparison to those calculated from the measured rainfall-runoff (storm-events) data. The proposed models were calibrated based on the rank-order, measured rainfall-runoff data (1,005 events) from 25 watersheds and validated in six watersheds for runoff estimation (170 events). The quantitative models' performances were evaluated and compared based on the root mean square error (RMSE), Nash-Sutcliffe efficiency (NSE), and percent bias (PBIAS). Using tabulated CNs, the NRCS model exhibited comparatively insignificant results in the maximum number of watersheds (high RMSE, low NSE, and statistically poor PBIAS values). Using storm-event based calibrated CNs, the NRCS model showed improvement for runoff estimation. Furthermore, the proposed models without the CN concept were superior (with comparatively low RMSE, high NSE, and statistically significant PBIAS values) for depicting improved performance in almost all of the watersheds.