Stormwater Pollution: Getting at the Source
The problem with the end-of-pipe approach for stormwater pollution
Now that MS4 (municipal separate storm sewer system) programs have become a major legal and financial concern for municipal public works directors throughout the country, it is time to reevaluate the role of source reduction, also known as pollution prevention. Greater emphasis on source reduction has the potential to improve reliability, reduce costs, improve equitability, and reduce liabilities of MS4 programs.
When Phase II of the MS4 program—or the National Pollutant Discharge Elimination System (NPDES)—was started, it was natural to rely primarily on end-of-pipe treatments. Our national experience with the Clean Water Act showed that treatment of municipal and industrial wastes could be highly effective. From 1968 to 1996, expansion of municipal wastewater treatment plants reduced the amount of biochemical oxygen demand (BOD) entering the nation’s rivers by 45%, even with a growing urban population, resulting in greatly improved oxygen concentrations downstream of major cities (USEPA 2000). If the construction of end-of-pipe treatment worked so well with municipal sewage, shouldn’t it work with urban stormwater?
Not necessarily. The flow and composition of municipal sewage is well known and predictable, making it relatively easy to design efficient treatment systems. By contrast, the flow and composition of urban stormwater can vary by several orders of magnitude within hours. Moreover, modern sewage treatment systems are highly sophisticated, energy-intensive operations with elaborate feedback controls, whereas stormwater structural best management practices (BMPs) are relatively simple devices with little operational control. Consequently, BOD removal in a modern wastewater treatment plant generally exceeds 95% and nutrient removal rates can exceed 85%, whereas typical removal rates in various stormwater BMPs average 25% to 65% across various BMPs (see Table 1, from Weiss, Gulliver, and Erickson 2007). Treatment efficiencies among BMPs of one type are also highly variable, as reflected in the 67% confidence limits shown in Table 1. For example, wet ponds have an average phosphorus (P) removal rate of 52% + 23%, which means that two-thirds of wet ponds have P removal efficiencies between 29% and 75%. Furthermore, winter performance of stormwater BMPs is poor in cold climates (Novotny et al. 1999).
Another problem with structural BMPs is the long-term fate of pollutants. Some pollutants, such as BOD, are actually degraded and converted to harmless end products (for BOD, the end products are carbon dioxide and water), but most pollutants simply accumulate in the BMP. Ultimate removal requires dredging or other operations to transfer pollutants from the BMP to a suitable disposal site to maintain functionality and sustainability. Some pollutants could accumulate to toxic levels. For example, adsorption of metals in infiltration basins over the period of several decades has been found to result in soil-bound metal concentrations well above safe exposure levels (Deschesne, Barraud, and Bardin 2005).
Finally, stormwater BMPs are expensive. Table 2 shows the total project costs—capital costs plus operations and maintenance costs, but not land acquisition costs—for a wet pond receiving stormwater from 20-hectare watersheds in the Twin Cities, based on the procedure developed by Weiss, Gulliver, and Erickson (2007).
Total project costs translate to $11,000 to $15,000 per hectare of watershed. Depending on the percentage of impervious surface, treatment costs range from $2,000 to $3,140 per ton of suspended solids and $640 to $1,028 per kilogram of phosphorus.
Structural BMPs will always be a necessary part of a stormwater management program, but relying almost entirely on structural BMPs for stormwater pollution control is analogous to “fighting the last war”—not necessarily the right approach for the new situation. End-of-pipe BMPs are not highly efficient or reliable. They accumulate pollutants and are expensive. In this article I suggest greater emphasis on source reduction as an important part of a multiple-barrier approach for reducing urban stormwater pollution. Currently, source reduction is an EPA-mandated component of MS4 programs, but it is often the “poor cousin” to structural BMPs.
Source reduction has several attractive attributes that are often not recognized. First, source reduction can be more effective than structural BMPs for some pollutants. The best example is chloride, an increasingly important urban stormwater contaminant resulting from road deicing. Chloride is not removed by any structural BMPs, which means that source reduction is the only way to reduce chloride concentrations in stormwater. Another example is soluble P, which typically comprises half of the total P in urban stormwater. Soluble P is not effectively removed in wet ponds and other BMPs that rely on sedimentation—hence the low total P removal in sedimentation-based systems (Table 1). Second, source reduction could potentially reduce the overall cost of a stormwater pollution reduction program. For example, onsite erosion and sediment control (ESC) could greatly reduce sediment inputs to stormwater ponds, especially in areas undergoing a lot of construction, thereby extending the maintenance interval for cleanout. Third, source reduction distributes costs more equitably—the “polluter pays” principle. What is fair about lax enforcement of onsite construction ESC measures when taxpayers are compelled to pick up the cost of additional maintenance downstream sedimentation ponds? Finally, greater utilization of source reduction techniques could increase the overall reliability of stormwater pollution programs, moving them toward the multiple-barrier approach now widely employed for drinking-water treatment.
In order to think about a “source,” we need to think about system boundaries. Many engineers implicitly think of the stormwater conveyance system as the system. With this definition, the source is the network of impervious surfaces in the watershed. The often repeated statement that impervious surfaces are the source of pollution is only true if one looks no further than the curb! Although the asphalt and concrete slowly degrade and could theoretically produce pollutants, there is not evidence that decay of streets is a significant source of pollutants to storm drains. Nearly all pollutants that enter stormwater conveyances come from other “upstream” sources, such as pervious landscapes, tree canopies, atmospheric deposition, and added road deicers. Streets are more accurately thought of as conduits of pollution, not the source.
If one envisions the street as the system, one of the main sources of pollutants is the pervious landscapes. For this system, sources include runoff from lawns (soluble and particle-bound pollutants), animal feces, erosion from construction sites and other bare surfaces, and leaves from boulevard trees. Other pollutants enter the street system directly from outside the watershed, which is discussed below.
The third system definition is the whole watershed. In this definition, pollutant sources come from outside the watershed. Some watershed sources enter pervious landscapes and undergo extensive transformations. Some examples include lawn fertilizer; pet food, which enters landscapes via excretion; and polyphosphates added to municipal water systems for corrosion control, which enter landscapes via lawn irrigation. Human food is a watershed input that might be important for stormwater management in residential watersheds with failing septic systems. Other pollutants enter the watershed via the street: copper and zinc from wearing brake pads and tires, road salt and abrasives used for deicing and traction control, and atmospheric deposition. With these system boundaries in mind, a first question about managing stormwater pollutants should be “Where does it come from, and where is it best controlled?”
Managing Pollutant Inputs From Pervious Landscapes
Lawn Management. Lawn runoff is highly polluted, especially with nutrients and suspended solids. Several studies of lawn runoff in the Upper Midwest (Barten and Jahnke 1997; Steuer et al. 1997; Waschbusch, Selbig, and Bannerman 1999) reveal that lawn runoff has higher suspended solids concentrations than treated sewage effluent and higher total P and total nitrogen (N) concentrations than effluent from an advanced secondary treatment plant designed for nutrient removal (Table 3). P concentrations in lawn runoff are typically an order of magnitude higher than needed to cause lake eutrophication.
High concentrations of nutrients and suspended solids are only part of the problem. Lawns are often compacted during or after construction, increasing the runoff coefficient and the fraction of precipitation that becomes runoff. Kelling and Peterson (1975) showed that the percentage of applied fertilizer nutrients that entered runoff increased as the infiltration rate declined. In a study of runoff from shoreline areas of Wisconsin lakes, Graczyk et al. (2003) found that the volume of runoff from lawns was an order of magnitude greater than runoff from adjacent forests. The combination of elevated nutrient concentrations and elevated runoff coefficients means that lawn runoff is probably a major source, if not the major source, of nutrients in watersheds with high percentages of residential land.
Understanding how lawn management affects nutrient export to streets requires some understanding of the ecology of lawns. Figure 2, a simplified view of the P cycle of lawns, can be used to develop concepts about the effect of lawn management on P in runoff. Some fertilizer P immediately becomes runoff. Carefully controlled studies on university experimental turf plots have shown that “watering in” with about a quarter-inch “of irrigation reduces the amount of phosphorus in runoff from subsequent rain events from 10% to 15% of applied fertilizer to less than 5% of applied P fertilizer” (Shuman 2004). Applying fertilizer to already saturated soils greatly increases the amount of P that will enter runoff in a subsequent rain event. In an experiment designed to simulate a “worst case” scenario (applying fertilizer after deliberately saturating soils), 25% of the applied P become surface runoff or leached downward below the root zone (Linde and Watschke 1997).
When grass is cut and left in place (or mulched), it decomposes, releasing soluble P that can infiltrate or become runoff (Figure 2). P that infiltrates is adsorbed by soils—but only to the point at which “saturation” of soil sites occurs. This adsorbed P is considered an “active” pool, readily available for plant growth (Figure 2). The most common soil tests for lawn management measure the P in this active pool, generally using either the “Bray” or “Olsen” test procedures. There is compelling evidence from 20 years of studies on agricultural soils that P losses by runoff and leaching increase as soil P levels increase (for a summary of many of these studies see Vadas et al. 2005). It is becoming widely accepted that an ideal soil test level for the active pool is about 20 to 25 milligrams per kilogram. Although there is no benefit to turf health in having levels higher than this, most suburban lawns have much higher levels, the result of years of over-applying P fertilizer. In a study of Minnesota residential lawns, Barten (2005) reported that two-thirds of 181 Minnesota lawns had “very high” soil P levels. On the other hand, allowing soil test levels to fall below 20 milligrams per liter also has disadvantages. As soil P levels drop much below 20 milligrams per liter, grass becomes “P limited”; as soil P levels decline to very low levels, turf health declines. Some turf specialists postulate that poor turf health caused by low soil test levels may increase erosion. Because there is also a much larger, less active pool of “fixed” P (residual organic P and mineral P), under-fertilization could have the unintended consequence of increasing P lost by runoff through increased erosion.
Improved lawn management would almost certainly reduce export of nutrients and suspended solids to streets. During initial development, improved construction practices could reduce soil compaction, reducing runoff throughout the life of the future lawn. Small but immediate reductions in P export to streets could be achieved by greater adoption of “watering in” and by avoiding fertilization of saturated soils before rain events. Much wider use of soil testing to guide P fertilization would likely gradually reduce P export to streets over a period of many years. Most of these techniques require active participation of homeowners. The potential for changing homeowner behaviors will be discussed later in this article.
Tree Leaves. Tree leaves are a second source of nutrients to streets. Trees can play an important role in stormwater management, intercepting rainfall and reducing runoff. However, leaf fall from boulevard trees can contribute nutrients to fall-season runoff. In our research, we’ve estimated the amount of P in leaf fall from several types of trees as a function of diameter at breast height (DBH; Figure 3). The regression line in Figure 3 shows that leaves from a maple with a DBH of 40 centimeters would contribute about 0.3 kilograms of P per year. Table 4 shows P inputs from tree leaves falling onto 1 kilometer of a city street lined on both sides by maple trees, as a function of tree size and spacing.
For streets with extensive canopy from boulevard trees, P input from tree leaves can be greater than inputs from lawns. The most appropriate source reduction step would be street sweeping, perhaps on several occasions during the fall leaf-fall period, to remove tree leaves before they decompose and release soluble P.
Dog Feces. There are two ways to estimate the nutrient output from dogs: (1) Follow dogs, collect feces, and measure the P content; or (2) estimate the dietary input, and assume that output equals input. In our research, we used the second approach, using recommended feeding levels and typical nutrient concentrations for dog foods (Baker and Hartzheim et al. 2007). These calculations showed that a 20-kilogram dog excretes about 4.2 kilograms of N and 0.9 kilogram of P. For comparison, average humans excrete about 6 kilograms of N per year and about 0.6 kilogram of P per year. Better compliance of “pooper scooper” laws would be effective at removing P from lawns, because most of the P is in the solid feces. They would be ineffective at reducing N inputs to lawns, because most excreted N is in the urine (Wood et al. 2004).
Figure 4 shows a hypothetical scenario for P inputs to 1 kilometer of a tree-lined residential street bordered by 30- by 30-meter lots, with 50% of the lot area contributing to runoff. The example uses a runoff coefficient of 0.1 and a P concentration in lawn runoff of 1.0 milligram per liter. The street is bordered by 30-centimeter DBH maple trees spaced at 10-meter intervals. Sixty percent of the households owned dogs. Dog wastes were not “scooped,” and the P delivery ratio was assumed to be 0.1. The P input rate from atmospheric deposition to the street was 0.25 kilogram per hectare-year (Barr 2004). Figure 4 shows that tree leaves would be the largest source of P to the street, indicating that street sweeping would be an effective P reduction technique for this street. For streets with no boulevard trees, lawn and dog wastes would be the largest contributors.
Will Homeowners Participate in Improved Lawn Management Practices?
A common refrain from the engineering community is that homeowners are not willing to participate in improved lawn management practices, and, therefore, source-reduction efforts are futile. Historical experience with other attempts to change environmental behaviors suggests otherwise. Curbside recycling was regarded as “unrealistic” back in the 1970s. At that time, only 7% of municipal waste was recycled. Today, 32% of municipal waste is recycled nationwide (USEPA 2007). Farmland erosion has declined by 40% since the early 1980s, as the result of ongoing education, augmented by payments for adopting various conservation practices, like taking highly erodible land out of production and encouraging conservation tillage (USDA 2003). Farmers are also using P fertilizer more efficiently: In Wisconsin, the phosphorus “use efficiency” (the ratio of phosphorus removed in the form of crops divided by the amount of fertilizer and manure phosphorus added to cropland) increased from only 50% to 84% between 1975 and 1995 (Bundy 1998). Residential water conservation efforts can also be successful. An example is Pinellas County, FL, where residential water use declined by 39% from 1989 to 2003.
The question, then, is not whether private citizens will change their environmental behaviors—there is overwhelming evidence that they will—but how to construct policies to encourage the change. Effective programs to change environmental behaviors seem to incorporate a mix of education, subsidies, disincentives or restrictions, and social marketing. For example, water conservation programs often include incentives (subsidies for low-flow appliances), disincentives (increasing block rates for water consumption), extensive education programs, and outright restrictions, at least during droughts, all of which contribute to reduced water use (Renwick and Green 2000). Experience from the agricultural sector suggests that effective efforts to improve lawn management should be spatially and socially targeted (Nowak, Bowen, and Cabot 2006). These examples suggest that programs to improve lawn management should develop targeted to specific types of homeowners, with different messages, incentives, and disincentives. For example, a “casual” homeowner might need to be convinced to occasionally add lawn P fertilizer to maintain a healthy turf needed to control erosion. A “perfectionist” homeowner may need to be convinced that he or she can allow soil P test levels to decline to 20 to 25 milligrams per kilogram through decreased fertilization with no negative impact on his beautiful lawn. Convenience might also be an incentive: Simply providing homeowners with instructions and mailing containers for soil samples would likely increase participation in a soil testing program.
Will Lawn Fertilizer Laws Work?
Another tactic to reduce P in urban stormwater has been enactment of lawn P fertilizer laws. Minnesota recently enacted the first statewide restriction on the use of lawn P fertilizers; statewide bills have been introduced or are being advocated in Michigan, Wisconsin, Florida, and Maine. Lawn P fertilizer laws also have been enacted in at least two cities (Madison, WI, and Ann Arbor, MI) and four counties (Dane, WI; Muskegon and Ottawa, MI; and Martin, FL). Lawn P fertilizer laws restrict the use of lawn fertilizers containing P above a specific, low percentage. All of these laws prohibit use of P-containing lawn fertilizers except for establishing new lawns or as indicated by a soil P test. The laws differ with respect to application to golf courses, parks, etc. In a related effort, the lawn care industry, several states, and other organizations have signed a memorandum of understanding to reduce lawn P fertilizer application in the Chesapeake Bay watershed by 50% (Chesapeake Bay Program 2006).
Photo: John Chapman, University of Minnesota ESC Program
|Figure 5: An inspector examining an erosion and sediment control failure
Do they work? For suburban neighborhoods without a lawn phosphorus fertilizer law, lawn fertilizer can be the largest input of phosphorus to residential watersheds (excluding inputs from human food), accounting for 75% or more of the total P input (Baker and Brezonik 2006; Baker and Wilson et al. 2007). Unfortunately, this does not mean that P concentrations in runoff would immediately decline by 75%, for at least two reasons. First, fertilizer laws do not completely halt the sale of lawn P fertilizers. In the first several years since Minnesota’s lawn P fertilizer law was enacted, the sale of lawn P fertilizers declined by only 48%, in part because sales of high-P lawn fertilizer is still allowed and labeling requirements are weak (MDA 2007). Second, because grass “mines” P from the active soil pool (see Figure 2), decaying grass will continue to supply P to street runoff, even with complete cessation of fertilizer P applications. To find out, John Barten and his colleagues studied stormwater coming from paired, residential watersheds. One watershed was in Plymouth, MN, which had a municipal P fertilizer restriction, and the other was in nearby Maple Grove, which did not. Preliminary results suggest that the P restriction reduced soluble P in runoff by 24% to 34% and total P by 12% to 16% (Johnson 2006). These results support the idea that fertilizer P restrictions will cause a slow, steady reduction in stormwater P, not a dramatic, instantaneous reduction. This is consistent with our knowledge of lawn P cycling, embodied in Figure 2.
Erosion and Sediment Control
Another underutilized opportunity for source reduction is erosion and sediment control. In many places, more rigorous ESC programs, especially for construction sites, could reduce the operation and maintenance (O&M) costs associated with cleaning out wet ponds and other sedimentation-based BMPs. The key limitation to effective ESC programs appears to be poor implementation. In a survey of 128 construction sites in North Carolina, Paterson (2000) found that many planned ESC measures were simply not installed, others were installed improperly, and few were properly maintained (Figure 5). For example, only one-third of installed silt fences were adequately maintained. A study in Dakota County, MN, showed that only 2% of construction ESC inspections subject to NPDES stormwater regulations in 2004 were conducted by trained inspectors (Chapman and Isensee 2006). Fortunately, the growth of MS4 programs has prompted better ESC education in several states. More than 4,000 people have attended ESC training sessions at the University of Minnesota since 2002.
In addition to large potential reductions in O&M costs for downstream BMPs, ESC shifts the financial responsibility from the community to the polluter, making the stormwater program not only more effective but also fairer. The perception of fairness may strengthen public acceptance of MS4 programs.
Source Reduction: The Only Solution for Road Salt
Road salt use in the United States has doubled since the 1980s (Figure 6), resulting in widespread salt contamination in the eastern United States (Kaushal et al. 2005). Much of the problem is caused by chloride, which is toxic to aquatic life at high concentrations. In urban areas, peak chloride concentrations sometimes exceed 5,000 milligrams per liter—one-quarter the concentration of seawater. Chloride also moves readily to underlying groundwater. The chloride concentration of groundwater underlying the heavily urbanized Shingle Creek watershed in Minnesota is 10 times higher than background concentrations (Wenck 2005). Chloride may be one of the most important emerging contaminants for urban stormwater.
Because chloride is a conservative pollutant that does not adsorb to particles, precipitate, or undergo biological degradation, structural BMPs cannot remove, or even trap, chloride. The only effective way to reduce chloride contamination is to reduce the source. Because road salt is needed for safe winter driving, blanket reductions or outright bans are not feasible. Fortunately, road salt is ideally suited to adaptive management, using feedback loops to guide road salting strategy under varying conditions, for the following reasons:
- Road salt crews are a relatively small, captive audience, which enables frequent communication.
- Road salt is almost certainly overused, so there is potential for reduction.
- Several technologies can reduce the amount of salt needed to accomplish deicing, including anti-icing (application of deicers before freezing occurs) and use of brines rather than dry salt.
- There are non-chloride alternatives to the most heavily used road deicer, sodium chloride, including acetate-based salts.
- Chloride can be readily measured indirectly, as conductivity—a method that is simple, inexpensive, reliable, and readily automated. This enables real-time monitoring on pavements, in storm sewers, and in streams.
- Reducing salt use could save money, by avoiding unnecessary purchase of salt.
The key components of an adaptive management strategy are illustrated in Figure 7. For each inter-storm event (point 1 on Figure 7), road crew members record salt application rates (now done with computerized trucks) and the type of salt used. Conductivity probes embedded in the pavement and installed in nearby storm sewers and streams provide feedback on chloride concentrations (point 2). For some roads, conductivity probes could be added to existing roadway monitoring systems with very little expense. These data are analyzed by a technical team (point 3), which compiles data for each storm and makes recommendations to the road crews (point 4). Early in the program, these recommendations would be fairly crude, but as the database and knowledge grows, recommendations would become more specific and effective. Periodically (several times a year), the analysis team would meet with road crews to discuss strategies (point 5). Base on experience with managing taste and odor problems in water supplies, these meetings between practitioners and the analysis team produce the most useful management concepts (Baker, Westerhoff, and Sommerfeld 2006). A key aspect of this type of program is that progress toward achieving chloride goals is continuously evaluated.
Summary: Source Reduction as Part of a Multiple Barrier System
Structural BMPs alone cannot provide sufficient pollution reduction to meet many urban water-quality goals. Various forms of source reduction will always be needed to achieve major reductions of pollutant loadings. The ideal time to consider source reduction is during initial development, when improved construction and landscaping techniques can be used to reduce stormwater volumes, and when ESC programs can reduce the huge potential sediment load from construction. Source reduction also has considerable potential for previously developed sites. Much could be learned from prior efforts to change environmental behavior that could be applied to household lawn management as MS4 programs mature, especially where total maximum daily load loading reductions are required.
Much needs to be learned to make source reduction more effective. Improved knowledge regarding lawn P cycling and more refined knowledge of homeowners’ “lawn behaviors” would allow us to move beyond generic recommendations and toward carefully targeted messages, with appropriate incentives and disincentives. Practical research is also needed to understand how adaptive management could be used to reduce contamination from road deicing.
Despite these constraints, source reduction has more potential for MS4 programs than is generally recognized, offering the potential for achieving pollution reduction in ways that are more effective, cheaper, and fairer than end-of-pipe controls alone.
Author's Bio: Lawrence A. Baker, Ph.D., is a senior fellow with the Water Resources Center at the University of Minnesota in St. Paul.