Comparison of Lumped and Quasi-Distributed Clark Runoff Models Using the SCS Curve Number Equation

Abstract

The Clark synthetic unit hydrograph and the Soil Conservation Service (SCS) curve number method has been used to simulate the rainfall and runoff behavior of a watershed for many years. Methodologies like Clark generally rely on the use of lumped or average rainfall and runoff parameters defined for the watershed, even though such parameters are spatially variable. In an attempt to leverage spatial parameters derived from geographic information, a modified Clark (ModClark) method or quasi-distributed model was developed for HEC-HMS. The ModClark method was initially developed to use the national network of WSR-88D radar (NEXRAD) rainfall data but few has been published on its application which is likely because of the difficulties in obtaining usable and reliable radar rainfall data and because of a lack of despisal preprocessing tools required to parameterize a ModClark simulation. While the original implementation and testing of the ModClark method required the use of NEXRAD data in specific formats, this study shows that it is possible to use any real or synthetic rainfall data whether it is spatially distributed or not. By not restricting the use of the distributed ModClark method to the use of spatially varying rainfall, distributed loss methods such as the commonly used SCS curve number can vary spatially over a grid and the effects of distributed watershed loss parameters can be analyzed with or without distributed rainfall. The implementation of the ModClark method in HEC-HMS is validated by comparing results to the Clark method using identical CN values. Further tests and examination of the SCS equation demonstrate that the runoff computed from distributed CN is always greater than the runoff computed from the traditional composite or area-averaged CN for ordinary ranges of rainfall depths. Moreover, by allowing a relatively fine grid resolution, the ModClark method determines the overall runoff from the watershed using a discharge weighted approach as opposed to weighted CN, which as reported in the National Engineering Handbook Part 630 is more accurate.