Runoff prediction errors with conjugate curve number

Abstract

The curve number (CN) rainfall-runoff model is widely adopted for runoff prediction. However, its practitioners seldom verify its validity according to the rainfall-runoff dataset. Conjugate CN was often used by mistake in the CN runoff predictive model for runoff prediction. This study compared runoff predictions of a calibrated CN model to the existing CN modelling approach and simulated the runoff prediction results from the inaccurate use of the conjugate CN. The calibrated CN model outperformed the other 2 models with the highest Nash–Sutcliffe model efficiency index of 0.917 without runoff over and under prediction tendency. When conjugate CN was wrongly used for runoff prediction, the model had the largest runoff depth under-prediction concern. Contrarily, the existing CN approach over-predicted runoff amount when it is not calibrated. On average, the model over-predicted runoff volume by 5 million liters/km2 while the inaccurate use of conjugate CN under-predicted runoff volume by nearly 7.6 million liters/km2 in Peninsula Malaysia.