Urban Flood Depth Estimate with a New Calibrated Curve Number Runoff Prediction Model

Abstract

The 1954 Soil Conservation Services (SCS) runoff predictive model was adopted in engineering designs throughout the world. However, its runoff prediction reliability was under scrutiny by recent studies. The conventional curve number (CN) selection methodology is often very subjective and lacks scientific justification while nested soil group catchments complicate the issue with the risk of inappropriate curve number selection which produces unreliable runoff results. The SCS CN model was statistically invalid (a = 0.01 level) and over predicted runoff volume as much as 21% at the Sungai Kerayong catchment in Kuala Lumpur, Malaysia. Blind adoption of the model will commit a type II error. As such, this study presented a new method to calibrate and formulate an urban runoff model with inferential statistics and residual modelling technique to correct the runoff prediction results from the SCS CN model with a corrected equation. The new model out-performed the Asymptotic runoff model and SCS CN runoff model with low predictive model bias, reduced sum of squared errors by 32% and achieved high Nash-Sutcliffe efficiency value of 0.96. The derived urban curve number is 98.0 with 99% confidence interval ranging from 97.8 to 99.5 for Sungai Kerayong catchment. Twenty-five storms generated almost 29 million m3 runoff (11,548 Olympic size swimming pools) from the Sungai Kerayong catchment in this study. 75%-94% of the rain water became runoff from those storms and lost through the catchment, without efficient drainage infrastructure in place, the averaged flood depth reached 6.5 cm while the actual flood depth will be deeper at the flood ponding area near to the catchment outlet. © 2013 IEEE.