Disconnect between Capture Areas and Stormwater Runoff: A Pre- and Postinstallation Model of Two Stormwater Control Measures

Abstract

Overland runoff was modeled before and after the installation of two stormwater control measures (SCMs) in a 23-ha urban catchment. The location of adjacent SCMs, a berm and bioswales, on the same hillslope provided an opportunity to compare stormwater capture in an urban catchment, and to evaluate whether SCMs would have measurable effects on a headwater stream. A physically based model (GSSHA: gridded surface/subsurface hydrologic analysis) was constructed using high-resolution LiDAR data collected before and after the berm and bioswale installation. Field data to support the model included water level loggers along with subsurface 1-L bottles to capture first flush runoff. Modeled tracers placed uphill of the berm and bioswales, as an analog for runoff contribution, indicated that runoff volume from the berm’s catchment varied between 0% and 50% of runoff volume from the bioswales’ catchment whereas the expected ratio was 14% based on the size of the berm and bioswales’ capture areas. Thus, the capture area did not predict contribution to streamflow. The berm’s capture area contributed more to total runoff during high intensity storms, when runoff from semi-pervious grassy areas was more prevalent. For the eight storm events modeled, a decrease in discharge was observed only for small storms with little or no change for more intense storms. Uphill-downhill paired samples of dissolved nitrate and total suspended sediment showed a decrease over the slope, but the decrease was observed in both pre- and post-SCM samples. The upslope runoff concentrations varied - casting doubt on whether urban overland runoff geochemistry can be representative given variable runoff generation and heterogeneous land uses. This study points out the challenges in evaluating runoff and pollutant fluxes to stormwater control measures and suggests that capture areas and flow paths can be better assessed through distributed modeling complemented by sampling and data loggers.