Revisiting Design Criteria for Stormwater Treatment Systems, Part 5

This is the last in a series of articles examining design criteria for stormwater treatment systems. The first three articles focused on the removal of settleable sediment with attached (particulate) pollutants by basins , fine-media filters , and flow-through swales . The fourth article considered removal of dissolved pollutants in wet basins. This article discusses removal of dissolved pollutants by fine-media filters and briefly flow-through swales. This article discusses removal of dissolved pollutants by fine-media filters and briefly flow-through swales.

The previous article established that three regulated metals are primarily found in a dissolved state: zinc, copper, and cadmium. Nitrogen is dissolved or may become dissolved in the system by bacteria. While dissolved phosphorus is generally less than 50% of the total, this fraction is nonetheless important given its bioavailability.

Fine-Media Filters
Fine-media filters have several advantages over wet basins. Filters can be placed beneath the surface allowing for continued use of the land. They do not freeze if placed below the frost line. The original media can be replaced with more appropriate media if performance objectives change. The performance of filters does not depend on and is not affected by the vicariants of plants. Performance may therefore be more reliable and consistent. Filters do not attract waterfowl, which adds to the pollutant load. Sand is the most commonly used fine medium. A second medium may be included to enhance removal of soluble pollutants.

Sand
One would think sand is inert, incapable of removing dissolved pollutants. Nonetheless, removal of dissolved phosphorus and metals (Caltrans 2004, City of Portland 2003) has been observed, but not consistently. It has not been observed in laboratory studies. Firm conclusions are hampered by sampling too few storms and the low influent concentrations, usually close to the respective detection limits with the exception of dissolved zinc. Removal of dissolved pollutants may occur by one or a combination of three mechanisms: precipitation, biological growth, and sorption to organic litter.

Precipitation of dissolved phosphorus can occur with iron, aluminum, manganese, and calcium phosphate. The surface chemistry of the sand is likely a significant factor in whether removal occurs by this mechanism. One study found sands with sorption capability vary significantly with the source (Arias, Bubba, and Brix 2001). Calcium phosphate may form on the sand if the hardness, alkalinity, and pH are appropriate (Aulenbach and Meisheng 1988), particularly if the sands are calcareous. The possible mechanism of precipitation of dissolved metals is with iron and/or aluminum oxide on the sand, either original to the sand or from aluminum and/or iron in the stormwater. Sorption of zinc and other metals to aluminum-iron oxide is known to occur in soils including wetlands (Minton 2002).

Reduction of dissolved metals by biological growth has been observed (Anderson et al. 1997), primarily by a bacterial biofilm that grows on the sand media and/or on the sediment removed by the filter. Growth likely occurs primarily on the bed surface and within the top few inches of the sand (Horner and Horner 1990). Removal may be due to the growth of the biomass and/or sorption to polymeric materials extruded by the bacteria in the case of dissolved metals (Johnson et al. 2003).

Dissolved metals sorb to humic organics in wetponds and wetlands. Humic organics are available for sorption through the degradation of leaves and other organic matter that enter the filter.

Deducing from existing field and laboratory studies the primary mechanism(s) and the conditions under which each sorption process may occur is not possible. There are too few studies and most are limited in scope. Often field studies are limited to the first year of operation. Field studies of a mature filter showing apparent continued removal after many years may be misleading. Periodic loss may not be observed as typically not all storms are sampled. Comprehensive laboratory and field studies are needed, covering the array of different sands and stormwater chemistries.

It would seem that as a field filter matures, removal by biomass reaches equilibrium: The amount of organic matter produced is equaled by the die-off or sloughing of biomass. After a period of time this mechanism would no longer result in net removal. However, should the interevent time between storms be significant and/or the filter be in an area with a distinct dry season, the bed may dry, causing desiccation of the biomass. The previously removed pollutants may be lost in the first storm of the following wet season. The biofilm reestablishes as the wet season proceeds.

One study (DeBusk et al. 1997) suggests that metal-oxide-metal or metal-phosphorus complexing occurs but that the mechanism is short-lived, likely a year or less in wet climates. This suggests that iron and aluminum in the stormwater do not play a role. The aggregate concentration of dissolved aluminum and iron can be substantial in stormwater–more than a milligram per liter. It is possible that the metal-oxide complex mechanism is dependent on the chemistry of the particular sand. The loss of both iron and phosphorus has been observed during what was believed to be an anaerobic incident, although monolayer absorption tests of the sand suggested that ferric phosphate precipitation was a minor contributor to high phosphorus removal (Bell 1995). The mechanism if it occurs may function longer in semi-arid climates with fewer storms and therefore a lighter loading rate.

The few laboratory studies of this question have generally found no removal of either dissolved phosphorus or metals, with one exception. However, neither lab nor field studies have examined the chemistry of the sand prior to use. Given the dichotomy between field and laboratory studies, the limited evidence might suggest that the primary removal mechanism with dissolved metals is humic organics provided by the degradation of organic litter. However, laboratory studies have generally not observed the removal of dissolved metals or lent credence to this mechanism.

Table 1 shows performance data for five sand filters (Caltrans 2004). The table shows a decline in performance from year two to year three. Data from the first year are not shown because there were too few samples. The differences in removal efficiencies between the filters and the decline in performance appear unrelated to influent concentrations. Despite the decline, the efficiencies remained high in the third year. One filter was considerably less effective than the other filters for unknown reasons, with negative performance in the third year. The decline of all filters may reflect stabilization of the bacterial mass, but may also reflect a decline in sorption capacity. It is possible that after bacterial mass stabilization and saturation of the soprtion capacity inherent to the sand, continued removal is limited to surface coating by incoming ferric-aluminum-manganese oxides, which may be minor or non-existent in most cases.

Experience With Additive Media
The removal of dissolved pollutants may be enhanced by modifying the media, either by the addition of a second media or by modification of the surface of the sand. The mechanism of sorption to metal-oxide coatings (iron, aluminum, or manganese) has been exploited for many years with metal-laden wastewaters (Edwards and Benjamin 1989) and more recently with stormwater (Sansalone 1999). Oxide-coated sand removes dissolved phosphorus (Benjamin 1995). Dissolved metals and phosphorus removal over a three-year period has been observed in sand filters to which calcite or iron filings were added (Shapiro and Associates and the Bellevue Utilities Department 1999). Organic media of various types are known to remove dissolved metals including peat, leaf compost, and soybean hulls (Johnson et al. 2003, Tobiason 2004). Additionally, leaf compost is used in a manufactured stormwater filter. A multitude of other organic materials have been found to remove metals (Minton 2002). Calcite in various forms, including crushed seashells, removes dissolved phosphorus (Leszczynska, Dzurik, and Michalowski 2002).

Organic media, as well as sand, remove petroleum hydrocarbons, given their extremely hydrophobic nature. Organic media, including activated carbon, remove pesticides. However, laboratory and field experience is limited.

Stormwater Chemistry
The chemistry of stormwater is complex and as yet little understood. Removal by any of the mechanisms previously described requires the metal or phosphate ion to be in the ionic form, not bound to dissolved humic organics or fine colloid sediment. As discussed in the fourth article of this series (Stormwater, May/June 2005), “dissolved” does not mean the pollutant is in the ionic form. Dissolved metals bind to humic organics and sorb to very fine colloidal sediments (inorganic and organic). They complex with carbonates and ferric-aluminum oxide complexes. The later forms sorb to media. Humic-bound metals do not sorb to media, apparently not even to activated carbon. Small colloidal-metal complexes will not sorb and are too small to be removed by straining. Research suggests that most of the zinc and cadmium remains in the ionic form, but that most copper binds to colloids (Grout, Wiesner, and Bottero 1999; Johnson et al. 2003; Morrison et al. 1984). Copper is particularly attracted to humic organics as well.

Reducing the concentration below receiving-water standards–particularly that of copper with its very low standard–may not be possible in many situations. A study (Caltrans 2004) of sand filters suggests a lower effluent limit of about 10 µg/L for dissolved copper and about 20 µg/L for dissolved zinc, similar to observations with zeolite and organic media (Johnson et al. 2003).

Organic media are general-purpose removers. They remove all heavy metals to varying degrees irrespective of whether we want the particular metal removed. But each type of organic medium has a hierarchy of preference. The results of one study are shown in Table 2 (Johnson et al. 2003). Note the interest in iron, a pollutant not typically of concern. Organic media also remove aluminum.

The above observations suggest that when conducting performance studies the stormwater samples need to be analyzed for constituents other than just the pollutants of immediate interest. Analytes to include are aluminum, iron, and dissolved organic carbon. Both aluminum and iron are present in substantial concentrations, exceeding the aggregate concentrations of the remaining metals.

Drawdown Time
Manuals differ with respect to the specified drawdown time: 24 to 48 hours. The choice of design drawdown time affects the size of the filter: The greater the drawdown time, the smaller the surface area of the filter. Doubling the drawdown time halves the filter area. The choice of drawdown time is therefore not trivial.

The most common reasons given for specifying a short drawdown time (relative to the interevent time between storms) are avoidance of temporary anaerobiosis at the bottom of the filter and/or clogging by the bacteria biofilm. Bacteria consume dissolved organic compounds and nitrify ammonia with the passage of stormwater through the filter bed. The oxygen available in the stormwater may be consumed by the time it reaches the lower part of the filter. This may occur during each storm or it may only occur periodically during a period of extended wet weather. Continued wetness sustains the bacterial biomass. With time its thickness may be sufficient to inhibit if not terminate flow.

The evidence for either concern is anecdotal. Clogging of fabric by bacteria has been observed (Johnson et al. 2003) in submerged sidewall sand filter-berms in wetponds (Horner and Horner 1990). To what extent algae from the pond contribute to the later observation is not known. Anaerobiosis has been documented in one filter (Shapiro and Associates and the Bellevue Utilities Department 1999). However, the filter had a bed thickness of 36 inches, a drawdown time of 72 hours, and iron filings mixed with the sand, and the filter was located in a wet climate: conditions that provide the maximum potential for anaerobic conditions. As anaerobiosis was limited to the lower 24 inches, substantial removal of dissolved metals and phosphorus still occurred.

A laboratory study (Johnson et al. 2003) suggests that drying of the upper part of a sand filter requires several days and that the lower remains wet regardless. It is likely that clogging is driven by the accumulation of sediments, possibly in concert with the biofilm, and that focusing on just the biofilm is unwarranted. However, definitive studies under field conditions are needed.

Whether anaerobic conditions are inherently undesirable has not been established. A concern is the loss of previously removed dissolved phosphorus or metals. But where neither is a focus, the concern for anaerobiosis is overstated. Loss of phosphorus or metals previously removed is limited to that portion complexed with iron or manganese. In an anaerobic condition these two oxides disassociate, losing sorbed metals. Aluminum oxide-metal complex and calcium phosphate are unaffected. The question has not been evaluated under conditions representative of field conditions and is hampered by a lack of understanding of each mechanism.

Anaerobic conditions in the bottom of the filter are desired if the objective is nitrogen removal. Ideally, nitrification occurs in the upper part of the filter and denitrification in the bottom. Denitrification may be enhanced by purposely flooding the lower portion of the bed. This is achieved by placing the outlet several inches above the bottom of the bed. The bed is configured to direct the water downward through the bed before exiting. However, nitrification-denitrification may be limited by the low concentration of dissolved organic matter in the stormwater. As most dissolved organic matter is likely removed in the upper aerobic zone of the filter, the remainder may be insufficient to sustain denitrifiers in the bottom of the filter.

A final possible benefit of a short drawdown time is temporary improvement in the filtration rate during the next storm (Johnson et al. 2003).

Operating Capacity
How long a medium will effectively remove a pollutant is determined by its operating capacity (Minton 2002). The operating capacity is the amount of a dissolved pollutant that is removed to the point where the concentration of the pollutant in the effluent of the filter reaches the specified allowable concentration. Such a specification does not occur with stormwater unless a total maximum daily load has been established for the particular receiving water.

The operating capacity is presented visually in Figure 1. A medium is placed in a column. Water with the pollutant of interest is passed downward through the column. The effluent concentration is monitored. With time the medium becomes saturated with the pollutant. The “saturation front” moves downward through the medium. The concentration of the pollutant begins to increase, dramatically increasing when the saturation front nears the bottom. When the allowable effluent concentration is reached, the medium is replaced. This value is identified at a relatively low concentration in Figure 1. It more often may be located farther to the right where the curve rises dramatically. In this case the medium should be replaced when the slope of the curve changes abruptly, called the breakpoint. The y-axis, C/C°, is the ratio of the effluent to influent concentration.

Operating capacity is not the same and is considerably less than the maximum saturation capacity. Operating capacity is a function of the concentration of the pollutant “seen” by the medium. The lower the operating concentration, the lower the operating capacity in proportion to the maximum capacity.

The maximum capacity is determined by mixing different known quantities of water with the pollutant of interest and with the medium. The mixture is then mixed for several hours. The residual concentration of the pollutant is determined, and the difference between the initial and final pollutant concentration in water gives the maximum capacity.

It is imperative that column and batch tests be performed within the influent concentration range found in stormwater. If metals are the pollutant of interest the tests should be done with a mixture of the metals known to be present in stormwater, including iron and aluminum. It is also preferable that the test water be stormwater, not potable water to which salts are added. The chemistry of stormwater is complex as previously noted. The medium should be tested for the leaching of regulated pollutants, which may occur at the low concentrations present in stormwater (Johnson et al. 2003).

Operating capacity is affected by the size and porosity of the medium. A more porous medium has more sorption sites, as do smaller media. Smaller media also enhance contact of the pollutant with the media. However, using a smaller medium is limited by hydraulic conductivity considerations.

Given the rather low concentrations of dissolved pollutants in stormwater, it follows that the operating capacities of most, if not all, media are very low. The sizing of media filters must be based on the operating capacity, not the maximum capacity. The medium with the highest capacity does not necessarily have the best performance (Johnson et al. 2003).

Which Comes First: Clogging or Capacity Exhaustion?
What will occur first: clogging by sediments or exhaustion of the sorption capacity? Likely the former will (Johnson et al. 2003), despite the relatively low operating capacities. The maintenance solution is to periodically remove the sediment and the top inch (2.5 centimeters) of sand to return its hydraulic capacity, with less frequent replacement of the entire bed.

Design Criteria
The limited number of studies of either natural or amended sand prevents definitive conclusions regarding design criteria for sorptive filters. Bed surface area matters for the removal of sediment. Bed volume matters for dissolved pollutants. Volume defines both performance and long-term capacity. In the second article of this series it was proposed that a bed thickness of 18 inches is likely excessive if the sole objective is the removal of sediment with attached pollutants. A depth of 18 inches or more may be appropriate for the removal of dissolved pollutants. In short-term laboratory studies, 15 minutes were sufficient for maximum removal of dissolved metals, suggesting a shallow bed is satisfactory. However, the need for volume and therefore operating capacity may require otherwise deeper beds than suggested by removal rates. The exception would be the removal of dissolved metals by organic litter as it is likely a bed surface phenomenon.

The appropriate mix of amendment media with sand depends on the particular amendment medium and its operating capacity. Best performance may be obtained with two or more amendments. There is not necessarily a correlation between performance and capacity. Best results may be obtained with two amendments: one with a relatively high capacity but lower performance, and another that exhibits the reverse (Johnson et al. 2003).

Few studies of operating capacity have been conducted, preventing the definition of mixture percentages at this time. Capacities likely differ considerably between media of the same generic type: for example, zeolites. Consequently, tests should be foregone only if the particular medium has been previously evaluated. Manufacturers of filter systems should test their media under realistic conditions and should develop appropriate specifications to ensure continuity of media quality. Tests should be performed on sands available in each region to ascertain if removal of dissolved phosphorus or metals occurs and, if so, to determine their capacity. Local jurisdictions can specify the most effective sand in their region. Test procedures are well established (Minton 2002).

Flow-Through Swales
The distinction of flow-through swales is made from infiltration swales in which essentially all of the stormwater escapes the swale by infiltrating the ground. Here the focus is with swales in which the majority of the stormwater passes to the far outlet of the swale, continuing as surface flow to the receiving water. The soils in these swales have a relatively poor infiltration rate. Care is advised when considering the benefit of infiltration. Keep in mind that infiltration does not result in 100% removal of the pollutants from the infiltrated stormwater. Furthermore, the higher the infiltration rate, the less effective the soil at removing pollutants, because there is less contact time.

Grass and other rooted vegetation do not remove dissolved pollutants directly from stormwater as it passes through the swale. Plants obtain nutrients for growth from the pore water in the soil. Hence, dissolved pollutants to be removed must contact the soil. This is not likely for the majority of the stormwater, although turbulence may provide the opportunity with sufficient residence time. The question therefore arises as to potential effectiveness of vegetated swales absent infiltration.

Dissolved metals, pesticides, and petroleum hydrocarbons may be removed by dead organic litter that has accumulated at the base of the grass stems. Free petroleum hydrocarbons are so hydrophobic that they sorb directly to grass. Studies of bioretention systems have established that a surface layer of mulch provides high reductions in metals; however, all of the stormwater passes through the mulch. This is not the case with flow-through swales.

The removal of dissolved metals has been observed but with a loss of dissolved phosphorus, likely from fertilization and/or fall dormancy (Caltrans 2004; Cammenmeyer, no date).

Final Observations
An understanding of the specific mechanisms of removal in both swales and sand filters is needed. Lacking is an understanding of whether and why removal capacity differs among sands and decreases with time and influent concentration. Classic sorption capacity tests used for decades in wastewater and potable water treatment are applicable to stormwater treatment (Minton 2002). The tests should be performed on each medium, sand or otherwise, before placement in a filter. This is particularly important when determining the amount of an amendment to add to sand. The amount of bacterial mass that grows in sand filters is unknown and therefore its potential effects on the concentrations of dissolved metals and nutrients are also unknown.

Final Observations on the Series
The time of simply measuring what is coming into and out of a treatment device to measure performance is long past. Such studies were useful early in the development of our understanding of treatment systems. It is, however, no longer of value to know only how well a system performs. It is also necessary to know how a system performs. This requires the analysis of constituents not of direct concern, such as iron, aluminum, and sulfate, and both the particulate and soluble species. It requires a definition of influent and effluent particle size distributions. It requires inclusion of gross solids, such as leaves and litter. It requires sampling of most if not all storms during the monitoring period, the continuous monitoring of flow, and the analysis of mass balances.

Such understanding of how a system performs also requires analysis of soils in wetponds and wetlands before flooding as to chemistry and sorption capacity for pollutants of interest and a similar analysis of media used in filtration and infiltration devices, including seemingly simple sand and soil. It requires analysis of the chemistry of wetland vegetation. It also requires an understanding of all of the factors that affect performance and a full description of the test facility in reports so that engineers are able to consider the relationship of specific design criteria to performance.