BMP Effectiveness for Nutrients, Bacteria, Solids, Metals, and Runoff Volume

Introduction
In response to existing and evolving regulations regarding nutrients, fecal indicator bacteria, solids, metals, and runoff volume, urban stormwater professionals around the United States need reliable performance data for best management practices (BMPs). A few examples of current regulatory drivers include (1) the USEPA urging states to adopt numeric nutrient criteria, (2) steadily increasing use of numeric action levels and numeric effluent limits for stormwater municipal and industrial discharges, (3) new initiatives for reducing bacteria levels in waterbodies used by the public for recreation, (4) total maximum daily loads (TMDLs) that identify waste load allocations for stormwater discharges, and (5) a major emphasis on runoff volume reduction at all levels of government. As a result, public and private sector stormwater dischargers are and will be spending substantial sums of money on these issues, primarily in the form of planning, designing, constructing, and maintaining structural and nonstructural BMPs. But which BMPs should be used and why, and how reliable will they be? To help answer these questions, the International Stormwater BMP Database project has recently completed a comprehensive stormwater BMP performance analysis technical paper series relying on data contained in the International Stormwater BMP Database.

The BMP Database is a long-term, publically available research database that contains results of stormwater BMP studies independently conducted and provided by researchers throughout the US and several other countries. Along with the Water Environment Research Foundation (WERF), cosponsors of the International Stormwater BMP Database are the American Society of Civil Engineers/Environmental & Water Resources Institute (ASCE/EWRI), the Federal Highway Administration (FHWA), the USEPA, and the American Public Works Association (APWA). The BMP Database was initiated by ASCE/EWRI Urban Water Resources Research Council (UWRRC). Original principal investigators included Ben Urbonas, Jonathan Jones, and Eric Strecker.

The BMP Database now contains performance data for nearly 500 BMP studies. In 2011, the WERF published a “Research Digest” that provides a summary of the results of performance analysis contained in a technical paper series completed over the past two years, providing information on BMP performance for these stormwater parameter categories:

• Nutrients
• Solids
• Metals
• Fecal indicator bacteria
• Runoff volume

In addition to BMP performance results for each parameter of interest, each technical paper provides a discussion of the current regulatory context for the constituent, a summary of data included in the database, a detailed assessment of adequacy of the data for purposes of analysis, discussion of unit treatment processes relevant to the constituent, and statistical analysis and evaluation of BMP performance for the data sets. Each technical paper concludes with recommendations regarding BMP selection, limitations on use of the data, and recommendations for additional research. Key summary statistics are provided in the technical papers, and appendices provide additional statistical analyses organized by BMP type. Underlying data spreadsheets are also provided for download to allow researchers to conduct independent analyses.

The performance summaries included in the papers are intended to provide technically sound analysis, presented in a straightforward manner that is usable by a wide range of entities such as municipal stormwater managers, regulatory agencies, university researchers, students, and other stormwater professionals. This article provides highlights of some of the findings contained in the WERF Research Digest and underlying technical summaries for various pollutants. For more detailed information regarding analysis methods and underlying data sets, see www.bmpdatabase.org or www.werf.org.

BMP Performance Findings for Solids
More than 6,200 waterbodies across the country are listed as sediment-impaired (USEPA 2010). Sediment impairments can result from external sediment loading to the stream, as well as from in-channel erosion. Sediment is naturally present to varying degrees in receiving waters and runoff; however, both urban and agricultural human activities can increase sediment loads to levels that impact aquatic life and other beneficial uses of waterbodies. Effective removal of sediment from urban runoff by stormwater BMPs is determined by both the unit treatment processes present in the BMP and the characteristics of the sediments in urban runoff. Dominant removal mechanisms for suspended solids include sedimentation and filtration. The effectiveness of these mechanisms is affected by multiple factors such as particle size distribution, particle density, BMP unit treatment processes (e.g., detention times, flow velocities in the BMP), temperature, and other factors. General findings from the analysis include:

• Total suspended solids (TSS). All of the BMPs included in the analysis generally performed well with respect to TSS, both in terms of statistically significant pollutant removal and relatively low effluent concentrations. Median effluent concentrations for all categories were below 25 mg/L. Bioretention, detention basins, media filters, retention ponds, and wetland basins showed particularly good performance with median effluent concentrations on the order of 10 mg/L.
• Total dissolved solids (TDS). Reduction in TDS was generally not statistically significant at the overall BMP category level. Some BMP categories (i.e., filter strips, media filters, and retention ponds) showed increases in TDS effluent concentrations.
• Turbidity. Findings similar to TSS were present for BMPs with turbidity data available for analysis. Overall, the turbidity data set was relatively limited.

This analysis suggests that stormwater managers have a broad range of options for reducing TSS concentrations. BMPs that provide sedimentation and filtration processes and are well designed, installed, and maintained are expected to provide good removal of TSS. In general, these mechanisms are anticipated to be more effective as the hydraulic residence time increases. Hydraulic residence can be increased in wetlands and ponds by increasing flow paths through the use of berms, baffles, and dense vegetation. In media filters and bioretention, increasing bed thickness and evenly distributing flows via the use of outlet control may be helpful. For filtration- and infiltration-oriented BMPs, maintenance is critical to prevent clogging from sediment.

An important note related to the TSS analysis, as well as analysis for other constituents in the BMP Database, relates to manufactured devices. The overall manufactured device BMP category included in the BMP Database includes a broad range of unit treatment processes that may result in widely varying performance for individual devices within this broad category. For example, some manufactured devices rely on hydrodynamic gravitational separation only, some provide filtration, others provide peak attenuation, and some provide a treatment train of multiple unit processes. Therefore, results for the “manufactured device” category provide only a gross characterization of the range of performance within this broad category. More refined analysis based on finer segmentation by unit treatment processes is necessary to draw conclusions for a particular type of device. As a result, the primary use of the data summaries at the broad category level is only for general information related to ranges of effluent concentrations potentially achievable with this general category of BMPs. The online BMP Database search tool can be used as a follow-up to refine such broad, general observations and further evaluate how individual manufactured devices
perform.

BMP Performance Findings for Nutrients
As of 2010, EPA identified more than 14,000 water bodies across the country as impaired for nutrients, organic enrichment, algal growth, and/or ammonia. The nutrient analysis for the BMP Database focused on phosphorus and nitrogen. The extent of available data varies among BMP-constituent combinations, resulting in more robust analysis for some BMP-constituent combinations than others. Nonetheless, the analysis of BMP performance data aligns relatively well with observed urban runoff concentration characteristics and theoretical background of unit treatment processes and transport mechanisms for phosphorus and nitrogen (WERF 2005). Key findings based on the BMP Database performance analysis follow for phosphorus and nitrogen constituent groups.

• Phosphorus. Total and dissolved phosphorus and orthophosphate were included in the analysis data set for phosphorus. Table 1 and Figure 1 provide a summary of analysis results for total phosphorus. Phosphorus in stormwater runoff is generally particulate-bound. As a result, BMPs with unit processes for removing particulates (i.e., sedimentation and filtration) will generally provide good removal for total phosphorus. In particular, BMPs with permanent pools appear to be effective at reducing the major forms of phosphorus. Leaching of phosphorus from soils/planting media and resuspension of captured particulate phosphorus may be a cause of phosphorus increases observed in vegetated BMPs such as bioretention, swales, and filter strips. Vegetated BMPs should be designed with adequate inlet protection, dense vegetation, and drop structures or check dams to minimize resuspension of particulates. The use of virgin compost or chemical fertilizers should be avoided and planting media within BMPs should be tested for phosphorus content (Hunt et al. 2006), especially if phosphorus is a constituent of concern.
• Nitrogen. Total nitrogen, total Kjeldahl nitrogen (TKN), and nitrate (reported as nitrate/nitrite or nitrate) were included in the analysis data set for nitrogen. BMPs with permanent pools (i.e., retention ponds and wetlands) appear capable of reducing nitrate concentrations, but may increase organic nitrogen. The opposite appears to be true for biofilters and media filters. Based on engineering and scientific theory of unit processes and knowledge of the nitrogen cycle, it is hypothesized that retention ponds and wetlands sequester nitrogen in vegetation during the growing season and then release nitrogenous solids during vegetation (including algae) die-off periods. Depending on factors such as climate, algal growth may contribute to ongoing particulate-bound nitrogen discharge year-round. For wetlands in particular, nitrification/denitrification processes can result in conversion of nitrates to nitrogen gas, which is then lost to the atmosphere. As indicated by the relatively high TKN removal and low nitrate removal for media filters, inert filtration appears capable of capturing nitrogenous solids, but the conditions are not as conducive for significant denitrification or nitrogen uptake as compared to bioretention or BMPs with permanent pools (retention ponds and wetland basins). Bioretention designs with pore storage above and below the underdrain may provide aerobic and anaerobic zones for nitrification/denitrification processes. Harvesting of vegetation and removal of algal mats and captured sediment may also be key maintenance practices for reliable removal of nitrogen.

EPA has identified metals as a leading cause of water quality impairment in the US (USEPA 2011). Many heavy metals in urban stormwater originate primarily from automobile-related activities and the exposure of building materials to rain. Elevated concentrations of some naturally abundant metals such as iron may be associated with erosion of soils and groundwater interflow. Atmospheric deposition of metals may also be an issue, particularly in the case of mercury, as a result of air emissions from coal-fired power plants, waste incinerators, certain manufacturing facilities, and other sources (USEPA 2005).

BMP Performance Findings for Metals

The analysis for metals included total and dissolved forms of arsenic, cadmium, chromium, copper, iron, lead, nickel, and zinc. Overall, most BMP categories provided good pollutant removal for most total metals. Findings regarding
dissolved metals are less clear, due in part to detection limit issues associated with the dissolved metals data set; however, it is also recognized that many conventional BMP types do not provide unit processes expected to be effective for dissolved metals. When reviewing the summary of findings for metals, it is important to be aware that if hypothesis test results show no statistically significant difference between influent and effluent data sets, this does not necessarily mean that the BMP is not capable of reducing pollutant concentrations. Instead, it may mean data set limitations do not allow a conclusive determination to be drawn (e.g., small data sets with large percentages of non-detects, or both influent and effluent data sets have low concentrations). Specific findings for several of the metals evaluated include:

• Large percentages of values below detection limits hamper the statistical performance analysis for some BMP-constituent combinations, particularly for dissolved cadmium, chromium, lead, and nickel, as well as total cadmium. Total and dissolved copper and zinc are least affected by non-detects. As an overall BMP category, porous pavement is most affected by large percentages of non-detects.
• Cadmium. Bioswales, filter strips, media filters, and retention ponds demonstrated reductions in total and dissolved cadmium concentrations. Analyses for both forms of cadmium were hampered by large numbers of non-detects for several BMP categories.
• Copper. All BMP categories demonstrated significant reductions in effluent total copper concentrations, with the exception of wetland channels. (Median effluent total copper concentrations for wetland channels were lower than the influent, but not at a statistically significant level. As noted above, this finding may be affected by characteristics of the data set, such as relatively low influent concentrations of total copper at wetland channels in this data set.) Effluent concentrations for total copper ranged from approximately 4 to 11 µg/L. Detention basins, filter strips, and retention ponds showed reductions for dissolved copper. Statistically significant reductions of dissolved copper for other BMP types was not demonstrated.
• Zinc. All BMP categories analyzed demonstrate reductions in total zinc, with most effluent concentrations in the 15- to 30-µg/L range. Figure 2 provides a summary of analysis results for total zinc. The broad category of manufactured devices showed higher effluent concentrations at approximately 60 µg/L, although removal of total zinc was still evident. (As previously noted, review of individual studies or unit treatment process-based groups of manufactured devices is needed to draw meaningful conclusions related to manufactured device performance.) Bioswales, filter strips, media filters, porous pavement, retention ponds, and wetland basins all showed reductions for dissolved zinc, as well. Detention basins and wetland channels had relatively low influent concentrations of dissolved zinc, which may help to explain why differences between median inflow and outflow concentrations were not evident for this BMP category.

BMP Performance Findings for Bacteria
Pathogens are one of the top causes of stream impairments nationally, with more than 10,000 stream segments identified as impaired, typically due to elevated concentrations of fecal indicator bacteria in waterbodies. Fecal indicator bacteria in urban runoff can originate from both natural and human sources. Based on the performance data available to date in the BMP Database for fecal indicator bacteria, only general inferences regarding BMP selection are appropriate at this time. General recommendations as a result of the analysis include:

• The majority of conventional stormwater BMPs in the BMP Database do not appear to be effective at reducing fecal indicator bacteria concentrations to primary contact stream standards, which is the ultimate target of TMDLs. Because the data are limited, both in the number of data points and the representativeness of the data, rigorous statistical conclusions cannot be drawn based on available data. Significantly more studies are needed for all BMP types to increase the confidence of performance estimates with regard to bacteria.
• In terms of reducing overall bacteria loads to receiving waters, site designs and individual BMPs that reduce runoff volumes should reduce bacteria loading from urban runoff and can reduce the frequency of discharges if volume losses are significant enough. However, even if there are significant volume losses, discharged surface flows may still cause receiving water exceedances.
• At the BMP category level, retention (wet) ponds and various types of media filters may help to reduce bacteria concentrations, although not necessarily to instream standards. Individual bioretention studies also appear to reduce bacteria concentrations, but more studies are needed for this category of BMP to draw category-level conclusions.
• In general, grass swales/strips and detention basins do not appear to provide meaningful reduction in bacteria concentrations and often show higher indicator bacteria concentrations in effluent. These BMP types may require enhancements to improve specific additional treatment processes such as filtration and sedimentation. However, it should be noted that volume reductions may be significant, so these BMPs may be effective at reducing overall bacteria loadings to receiving waters.
• The manufactured devices in the BMP Database include a range of unit treatment processes, requiring case-by-case evaluation of performance. As an overall category, the individual studies currently included in the BMP Database do not demonstrate significant fecal indicator bacteria removal, regardless of the unit treatment process.
• Various individual BMPs may provide reductions in bacteria that may not be reflected in BMP category-level analysis presented in this summary. Representative examples in the BMP Database where reductions in indicator bacteria were identified include individual bioretention studies, a wetland basin, and a few detention basins. Care should be taken to understand both site-specific and BMP design characteristics in these studies before assuming that similar performance will occur at other locations.

BMP Performance Findings for Volume Reduction
The hydrologic performance of BMPs is an important factor in the overall effectiveness of BMPs in reducing potential adverse impacts of urbanization on receiving waters. In addition to providing water-quality data, the BMP Database also provides watershed characteristics and monitoring results for precipitation and flow conditions, enabling assessment of the hydrologic performance of BMPs when such data are provided by the researcher. BMP performance metrics recommended since the inception of the BMP Database project have included:

• Fraction of long-term runoff volume managed by the BMP (i.e., capture efficiency),
• Fraction of the captured volume that is lost by the BMP (i.e., via infiltration and/or evapotranspiration) and does not discharge to surface water (i.e., surface runoff volume reduction), and
• Level of treatment provided for water that discharges from the BMP (effluent concentration characteristics).

The approach used in the volume reduction analysis relied upon performance metrics developed over the previous two years as revisions to the database were implemented to better integrate low-impact development (LID) techniques (Geosyntec and WWE 2009a and 2009b). Prior to conducting analysis, a substantial data screening and evaluation effort was completed. The screening effort resulted in an important observation that, over time, the objectives of BMP monitoring data have changed. Specifically, over the last several years, greater emphasis has been placed on management of stormwater runoff volumes (USEPA 2009). As a result, recent studies tend to have more complete monitoring of hydrology. In general, the volumetric data contained in the BMP Database reflect this trend. Older studies for which volume reduction was not a study objective may theoretically contain greater sources of potential bias and error than newer studies. Conversely, a greater portion of recent studies were conducted specifically to quantify volume reduction along with water-quality characteristics. For this reason, volumetric analysis based on the BMP Database must be carefully conducted, and limitations of the datasets should be understood. With appropriate consideration for reliability of the datasets and screening of data to remove unreasonable studies, meaningful results can be obtained from volumetric analyses of the BMP Database for various BMP categories, when considered in proper context and with appropriate caveats.

Although a variety of different performance metrics are applied and discussed in the volume reduction analysis technical paper, the metric identified as “BMP Categorical Analysis of Relative Volume Reduction by Study” is helpful for those interested in a generalized, big-picture perspective on BMP categories likely to provide volume reduction benefits. In this analysis, all storm events with paired inflow and outflow were summed within studies and relative volume reduction was calculated for each study. These study results were then pooled by BMP category (e.g., bioretention, detention basin) and summarized. Table 2 provides summary statistics of study-based relative volume reduction by category for “normally dry” BMP categories comparing inflow and outflow volumes. Normally dry BMPs are those that are not designed to maintain a permanent pool of water.

As shown in Table 2, the BMP categories considered in this analysis suggest significant surface runoff volume reduction. Variability in study performance is relatively wide. These numeric estimates may be useful at a planning level with consideration of the reliability of input datasets and the potential role of design criteria and site conditions on volume reduction performance. These results may be useful in estimating the range of performance that could be expected within a BMP category over a range of conditions and design standards; however, these generalized estimates are not appropriate to predict the volume reduction performance of a specific BMP designed to specific design criteria for a specific set of site conditions.

Based on the performance data available to date in the BMP Database, the authors recommend only general inferences regarding BMP selection with regard to volume reduction at this time. These general recommendations include:

• Normally dry vegetated BMPs (filter strips, vegetated swales, bioretention, and grass lined detention basins) appear to have substantial potential for surface runoff volume reduction, on the order of 30% for filter strips and grass-lined detention basins, 40% for grass swales, and greater than 50% for bioretention with underdrains. Therefore, these BMPs can be an important part of an overall strategy to manage site hydrology and control pollutant loading via volume reduction.
• Normally dry vegetated BMPs also tend to provide better volume reduction for smaller storms, which tend to occur more frequently than larger storms; this can lead to reduced frequency of discharges or much smaller discharge volumes.
• Both of the above results would tend to reduce the frequency of water-quality-standards exceedances. Developers of BMP design and performance criteria may want to consider the role of BMP volume reduction in reducing pollutant loadings when developing design requirements.
• Retention ponds and wetland basins and channels do not appear to provide substantial volume reduction, on average, and should not be selected to achieve volume reduction objectives.
• Variability in volumetric performance between studies indicates that design attributes and site conditions likely play key roles in performance. Therefore, when using BMP category-level analysis results to select BMPs to maximize volume reduction, it is also important to ensure that design features to promote volume reduction are explicitly included in design and the site characteristics are conducive to allow volume reduction.

Conclusion
Based on performance analysis completed for the International Stormwater BMP Database, urban stormwater BMPs are demonstrated to be a vital component of strategies to reduce pollutant loads to receiving waters. Pollutant loads may be reduced by a combination of treating runoff effluent quality and/or by reducing runoff volumes. Stormwater BMPs can also play a valuable role in limiting the adverse hydrologic effects that often accompany urbanization. Based on the unit treatment processes present in the BMP, some BMP categories appear to be more effective than others for improving effluent water quality for various constituents. The BMP performance analysis findings presented in the BMP Database pollutant category summaries are a resource for informing better selection and design of BMPs by the stormwater community. Continued growth of the BMP Database with high-quality studies will be helpful in enabling more robust findings regarding BMP performance and ultimately selecting BMPs targeted to site-specific conditions and treatment objectives. Project co-sponsors and project team members have identified and prioritized follow-up analyses to conduct in each major topic area. See each technical paper for more information on additional research needs for specific BMP-constituent combinations.

For More Information
The WERF Research Digest can be purchased from www.werf.org, and the individual technical summaries and statistical appendices can be downloaded from www.bmpdatabase.org. The BMP Database project is focusing on two new areas in 2012, including agricultural BMPs and support for nutrient-related efforts in the Chesapeake Bay area. Information on these new efforts can also be obtained from the project website.