Drainage area characterization for evaluating green infrastructure using the Storm Water Management Model

Abstract. Urban stormwater runoff quantity and quality are strongly dependent upon catchment properties. Models are used to simulate the runoff characteristics, but the output from a stormwater management model is dependent on how the catchment area is subdivided and represented as spatial elements. For green infrastructure modeling, we suggest a discretization method that distinguishes directly connected impervious area from the total impervious area. Pervious buffers, which receive runoff from upgradient impervious areas should also be identified as a separate subset of the entire pervious area. This separation provides an improved model representation of the runoff process. With these criteria in mind, an approach to spatial discretization for projects using the U.S. Environmental Protection Agency's Storm Water Management Model is demonstrated for the Shayler Crossing watershed, a well–monitored, residential suburban area occupying 100 ha, east of Cincinnati, Ohio. The model relies on a highly resolved spatial database of urban land cover, stormwater drainage features, and topography. To validate the spatial discretization approach, six different representations were evaluated with eight 24 h synthetic storms. With minimal calibration effort, the suggested approach out–performed other options and was highly correlated with the observed values for a two month continuous simulation period (Nash–Sutcliffe coefficient = 0.852; R2 = 0.871). The approach accommodates the distribution of runoff contributions from different spatial components and flow pathways that would impact green infrastructure performance. We found that when all subcatchments are discretized with the same land cover types, instead of using an j × k array of calibration parameters, based on j subcatchments and k parameters per subcatchment, the values used for the parameter set for one subcatchment can be applied in all cases (i.e., just k parameters). This approach not only reduces the number of modeled parameters, but also is scale–independent and can be applied directly to a larger watershed without further amendment. Finally, with a few model adjustments, we show how the simulated stream hydrograph can be separated into the relative contributions from different land cover types and subsurface sources, adding insight to the potential effectiveness of the planned green infrastructure scenarios at the watershed scale.