CS15B - Concurrent Session 15B: Distributed Curve Number Modeling for Runoff Volume Prediction
Tuesday, September 19, 2023
2:00 PM – 2:30 PM PDT
Location: Madera
Rainfall-runoff modeling is commonly employed by hydrologic engineers with one of the primary goals being the determination of runoff volume. Runoff volume can be calculated from the total depth of rainfall, the overall amount of losses and the watershed area. In hydrologic engineering modeling the Natural Resources Conservation Services (NRCS) Curve Number (CN) method is commonly used to determine the losses from an associated rain event. Traditionally, application of the CN method involved determining one “average” representative CN for a watershed with the underlying assumption that losses occur uniformly over the entire landscape of the watershed. This assumption has obvious flaws and therefore the presumption of one representative CN for a watershed is inherently incorrect and frequently will lead to erroneous predictions of runoff volume when compared to actual data associated with historical rainfall-runoff measurements. Version 6 of the Hydrologic Engineering Center’s River Analysis System (HEC-RAS) allows users to input a spatially distributed CN layer in combination with a precipitation boundary condition for the purpose of simulating the rainfall-runoff phenomenon. By using a distributed CN layer as opposed to one representative value, runoff volume predictions theoretically will prove more reliable. A study was conducted which compared runoff volume results of 45 simulations using an average CN as part of the NRCS methodology and simulations using a spatially varied CN as part of a HEC-RAS simulation, to actual results using historical data obtained from the National Weather Service (NWS) (rainfall data) and the United States Geologic Survey (USGS) (runoff data). The HEC-RAS simulations produced an average runoff volume error of 67% with 21 of 45 simulations overestimating the total runoff volume. The NRCS simulations had an average runoff volume error of 81% with 31 out of 45 simulations overestimating the total runoff volume. A subsequent analysis was performed to compare the theoretical CN based on soil type and land use used as part of the NRCS simulations, to the average CN calculated based on analysis of the runoff hydrograph produced from the HEC-RAS simulations to the event specific CN determined from analysis of the runoff hydrograph and rainfall values associated with the historical data collected from the NWS and USGS. Results showed that using a spatially distributed CN approach yielded better results than the theoretical average CN, when compared to the storm specific CN.