From 2005 to 2006, the U.S. Geological Survey worked cooperatively with the South Carolina Department of Transportation in Beaufort and Colleton Counties, South Carolina, to assess the performance of four different structural devices that served as best management practices (BMPs). These structural devices were installed to mitigate the effects of stormwater runoff on waterways near State roads. The South Carolina Department of Transportation is required to address the quality of stormwater runoff from State-maintained roadways as part of the National Pollutant Discharge Elimination System stormwater program mandated in the Clean Water Act.

The performance assessment of the four structural best management practices was based on stormflow measurements and chemical analyses of stormwater-quality samples collected during a 20-month period from March 2005 through October 2006, which represented a range of seasons and rainfall intensities. A total of 49 sample sets that included stormwater from the inlet and outlet of each of the four structural devices were collected as flow-weighted composites to provide event-mean concentrations of suspended sediment, nutrients, and trace metals. In addition, each set included grab samples that were collected to provide the first flush concentrations of oil and grease and fecal-indicator bacteria.

A tiered statistical approach was used in the data analysis. Performances of the four structural BMPs were assessed individually based on how well the BMPs were able to reduce the selected constituents. Descriptive statistics and nonparametric Wilcoxon signed rank tests were applied to event-mean concentrations and loads in the water entering the inlet and in the water leaving the outlet of each BMP for each constituent to identify if significant reductions occurred. If significant reductions occurred, the BMP was considered efficient at reducing the constituent. To quantify efficiency, a simplistic approach was applied to compute mean and geometric mean efficiency ratios for the significantly reduced constituents in each BMP. Each BMP performance was ranked based on its computed efficiency ratios, however, the computed efficiency ratios were not sufficient to determine if statistical differences occurred among the performances of the four BMPs. Consequently, a more complex approach was used to apply statistical comparison tests to reduction percentages computed for individual storms (a modified removal efficiency of individual storm-load approach) to determine if differences in event-mean concentrations, loads, and reduction percentages for significantly reduced constituents occurred among the four structural BMPs.

Overall, the four BMPs were efficient in reducing suspended-sediment event-mean concentrations and loads in the stormwater entering the inlets of the BMPs to significantly lower event-mean concentrations before discharging the stormwater from the outlets. The cumulative suspended-sediment event-mean load in stormwater entering the BMPs from the storms sampled during the data-collection period was 1,026 kilograms (1.13 tons). The BMPs removed a cumulative suspended-sediment load of 558 kilograms (0.62 ton). The BMPs tended to preferentially trap the sand-size fraction of the sediment, thereby releasing a greater percentage of fine-grained (silt and clay) sediment in the water discharging from the outlet. The preferential trapping of fine-grained sediment by the BMPs could explain, at least in part, why the BMPs were not successful at significantly reducing these constituents.

In general, the four BMPs were not successful at significantly reducing fecal bacteria, nutrients, and total organic carbon (including associated properties of biochemical oxygen demand and chemical oxygen demand). Three of the four BMPs significantly lowered oil and grease concentrations before the stormwater discharged from the outlet. Additionally, only one BMP was effective at reducing all total and particulate trace-metal event-mean concentrations and particulate trace-metal event-mean loads in stormwater entering the inlet. With respect to trace-metal event-mean concentrations, however, minimal or no improvement in outlet water quality was observed for the four BMPs, and the majority of the outlet concentrations were above the established acute and chronic aquatic-life criteria by the South Carolina Department of Health and Environmental Control.

No statistical differences among the removal-efficiency of the four BMPs were determined for suspended-sediment event-mean concentrations, total suspended solids event-mean concentrations, or oil and grease concentrations. These statistical findings indicated that differences among the mean efficiency ratios were not significant among the BMPs for these properties. Additionally, one BMP generally had statistically greater removal efficiency for total and particulate cadmium, copper, lead, and zinc than one or more of the other three BMPs.

Statistical correlation tests were unable to identify a single major factor that would explain the high variability in inlet and outlet water concentrations and in removal efficiencies estimated by reduction percentage. Highly variable inlet and outlet concentrations for each BMP that produced highly variable reduction percentages were probably the result of multiple interacting factors, particularly rainfall intensity, the amount of rainfall between sampling events, traffic density, and the period of time since the last maintenance (clean out) of the BMP.