January-February 2012

Bioretention Filters

Part 2. What we know with respect to hydrologic and treatment performance

Tweet

Part 1 of this series, which appeared in the November/December 2011 issue of Stormwater, focused on the configuration, processes, and design criteria for bioretention filters. The second part focuses on hydrologic and treatment performance.

Hydrologic Performance
Although they were originally viewed only as treatment devices, it is now recognized that bioretention filters also provide substantial benefits as a hydrologic device, namely that they can significantly reduce the peak rate and volume of the discharge via the underdrain system even in tight soils. As such, bioretention filters can contribute to hydrologic flow requirements, typically defined as controlling the post-development discharge rates to predevelopment rates, commonly by temporarily storing stormwater—that is, by detention.

However, as currently designed, bioretention filters do not provide sufficient storage to fully comply with the requirement for the extreme events, namely five-year through 25-year storm events. The peaks and volumes of extreme events may experience little reduction. As the bioretention filter is full of media, it would have to have a volume twice that of a normal flood/bank channel dry basin, which is generally not the case. Nonetheless, peaks of the smaller events are reduced if not eliminated.

The combined infiltration and evapotranspiration (I/ET) of field units studied have varied widely, from 2 to 99%. A few studies have evaluated evapotranspiration directly, finding it to range from 4 to 50%. Why researchers have found a wide range of I/ET and ET volumes is not clear. One researcher found annual I/ET of 88 and 96% for two filters located at the same site, placed in a soil whose infiltration rate ranged from only 0.2 to 0.6 inch per hour (0.5 to 1.2 centimeters per hour).

As expected, a seasonal difference has been observed, suggesting the important role of plants with respect to evapotranspiration. In one study, the percentage reductions were 84% in the spring, 93% in the summer, 87% in the fall, and 46% in the winter. In a follow-up study of the same system, the results were 96%, 96%, and 98% for spring, summer, and fall, respectively. No data were collected during the winter.

There is the question of infiltration in clay soils. Water will not infiltrate downward into clay. However, water may infiltrate laterally in structural clay (but not amorphous clay). This is possible through the inclusion of the lower box shown in Figure 1 in part 1 of this article, in which a portion of each storm is retained, infiltrating into the soil between storms.

Estimation of the evapotranspiration and groundwater recharge can be evaluated through the use of lysimeters, as well as treatment effectiveness. Regardless, making sense out of the information is hampered by the failure of the researchers to identify the infiltration characteristics of the undersoil or to describe the soil, except for one study. Hence the data are of no value for design.

Therefore, the I/ET volume reductions observed should be expected given that 0.5 inch per hour is the common minimum rate for infiltration basins and trenches. Furthermore, the area of bioretention filters is commonly on the order of 5 to 10%, an area larger than for infiltration basins and trenches. A bioretention cell is essentially an infiltration basin, except for the specified media and intentional planting.

Four additional factors affect the observed range of I/ET values: the media specification, the size of the filter, the relationship between filter area and its volume, and the inclusion of a bottom chamber.

Despite its lower void ratio, bioretention media can hold more water than sand between events. Sand holds about 10% water, whereas a loam soil holds about 15 to 25%; others have said as much as 40% of its volume, retaining it for gradual I/ET. Thus this factor is unresolved at the moment. But it is more likely that the lower percentage, perhaps on the order of 25%, is the valid percentage. Also, a filter media containing of 85% sand retains less stormwater than a media containing 50%.
The effect of size is apparent from a study in which two filters were located at the same site and differed only in volume. Both had the same surface area, but one was deeper (3 feet versus 2 feet). The deeper cell had a greater volume reduction (60% versus 40%).

The parent soil was Class D, suggesting that essentially all of the volume reduction was by evapotranspiration (the units were examined from spring through fall). The greater the media volume, the greater the volume of water retained by the media, and therefore the greater the amount of water available for I/ET. The absolute and relative proportions of water loss through evapotranspiration and infiltration can be ascertained by the use of lysimeters.

A final factor to consider is climate. Most of the studies of the hydrologic performance of bioretention filters have been along the Atlantic Coast where most of the rainfall occurs in the summer, coincident to the growth season, and common east of the Rocky Mountains. There have been a few studies in cold climates. There have been none in either semi-arid climates or wet climates where the winter is the dominant wet period.

Water loss in these climatic areas might be expected to be lower than has been observed to date. Hence, for most studies conducted to date, the maximum I/ET period corresponds with the maximum growth period. This is not commonly the case along the West Coast of the United States and Canada and in semi-arid areas.
A few studies have evaluated the change in filtration rate over several years. One researcher measured the hydraulic conductivity or filtration rate (at shallow operating depths the conductivity and the filtration rate are essentially the same) of two filters at the same site after 24 months of operation and found it to be 15 inches per hour (38 centimeters per hour). The initial filtration rate was not measured.

When the same system evaluated about two years later, the researcher found a significant increase in the bypass volumes, suggesting a notable decrease in the filtration rate. The filtration rate of the two filters had decreased to 1.0 and 1.5 inches per hour (2.5 and 3.8 centimeters per hour). Another study found filtration rates of 1.3 to 3.1 inches per hour (3.3 to 7.9 centimeters per hour) after two years of operation. Again, the initial rate was not measured. This suggests that the surface of filters should be cleaned about every five years.

A study of eight rain gardens in Minnesota found the filtration rate to range from 1.1 to 28 inches per hour (3 to 71 centimeters per hour). The ages were not provided for each rain garden; only a range of new to two years was given. A study found that the filtration rate was 8 inches per hour three years after construction of the filter, and 13 inches per hour 10 years later. Filtration rates likely improved due to the maturation of the root system. Another study found the filtration rate decreased from about 0.33 to 0.20 inch per hour over three years.

The above data tend to suggest that the mulch is degrading, allowing the sediment that it has captured to reach the surface of the filter. The data tend to indicate that the filter surface and about 2 inches of filter media (as with sand filters) should be removed about every five years.

A study was conducted of several bioretention units with regard to the hydraulic conductivity. The desirable range was stated to be about 2 to 8 inches per hour (50 to 200 millimeters per hour). Forty-three percent were found to be within this range, 40% below, and 17% above. It was concluded, however, that there was not a problem regarding the units that were below the design range as the bioretention units were significantly oversized.

a. These are dissolved metals. b. Leaching occurred because the relative low concentration of phosphorus in the influent as this was runoff from a roof of a home. c. Leaching occurred due to high P in the filter media.

Mg/L for TSS, TKN, and TP, ug/L for metals

Sizing and configuring bioretention filters lends itself to hydrologic modeling. Many computational models have been developed to simulate the flow through biotention filters and the effect on the flood control requirements of a jurisdiction. In regions where the time between storms is substantial, sizing can be based on single-event modeling using statistical events.

However, in wet climates where storms can occur back-to-back, continuous simulation is appropriate using the historic rainfall record. With either approach, the effect of the design relative to channel protection and flood control requirements can be integrated seamlessly.

Although concentration increases commonly occur in the underdrain discharge, the effect of infiltration is to give a positive reduction when the performance is based on load reduction. Data are provided in Table 1. However, it is stressed that the actual loadings with respect to the total environment, including the receiving water, can be much lower than calculated as the soil does not remove all of the incoming pollutants—in particular, nitrogen.

Bioretention filters are not recommended where nitrogen or phosphorus is the targeted pollutant. A better media for phosphorus removal is activated alumina. A better media for nitrogen removal is sand or zeolite. However, this is impractical because the filter would have to be dug out when the media has reached exhaustion. This of course depends on the capacity of the amendment and what percentage is placed in the filter. The observed negative loadings for phosphorus are due to leaching, a problem that has since been corrected. Negative loading of total suspended solids (TSS) is likely due to the loss of fines from the loam soil and compost.
As noted above, the engineer must use caution in calculating the loading reduction. Researchers commonly calculate the loading reduction with respect to the discharge from the underdrain system. This implies that the calculation represents the loading reduction relative to the receiving water, and that the parent soil into which the infiltration occurs removes 100% of the pollutants not removed by the bioretention filter. This is not the case.

All soils, regardless of their drainage characteristics, are poor removers of nitrogen. Well-drained soils (e.g., glacial outwash) and fractured geology (e.g., karst) are poor removers of other pollutants as well, with a corresponding increase in the concentrations of contaminants in the groundwater. As a result, the actual loading to the stream may substantially greater than implied by a loading reduction with respect to the underdrain discharge.

Field studies should distinguish between stormwater loss to evapotranspiration and to infiltration. The monitoring of the nearby groundwater is appropriate, as well as the use of lysimeters to sample water quality at the bottom of the filter. The effect is to increase the concentration of pollutants in the receiving waters, most notably streams, during base flow, leaving less margin for elevated concentrations during storms.

Treatment Performance
When considering performance, three modalities are relevant: change in the pollutant concentration, loading reduction that considers the effect of I/ET, and the effect of ET on effluent concentration. Field studies have often found an increase in pollutant concentration in the discharge but a loading reduction relative to the underdrain discharge due to I/ET. Assuming about 50% ET in the field, a reasonable upper boundary, the concentration of a pollutant should double from influent to effluent, absent removal by the filter. A summation of general performance is presented in Table 2.

Many laboratory and field studies have been conducted. Generally, laboratory studies have found higher reductions of pollutant concentrations based on percentage removal than field studies, as noted previously.

Field studies frequently find negative reductions, as noted previously with phosphorus, whereas laboratory studies seldom do. Higher percentage reductions in early laboratory studies can for the most part be attributed to the concentrations in the synthetic stormwater being considerably higher than in typical stormwater, thus contributing to a falsely high percentage removal. However, recent laboratory studies using more appropriate concentrations find higher percentage reductions than field studies, suggesting other factors are involved.

A possible factor is the short-term nature of laboratory studies with flow rates and periodic dosing schedules similar to a few weeks to a year of operation in the field, but sometimes compressed so as to finish the study in a timely manner. Hence, field studies should be viewed as truly representative of the capability of bioretention filters, whereas small-scale studies should be looked upon as a means of determining the factors that affect performance and the relative contribution of the unit processes discussed in part 1 of this article.

Given the apparent bias of small-scale testing, this article focuses on the results of field studies. When considering performance, it is important to recognize that we are speaking of net removal. Pollutants in the stormwater removed by the filter are offset by losses from the media. The prevalence of studies in wet climates where the summer is the dominant wet period has been noted.

The initial studies of the late 1990s and early 2000s were wholly laboratory in scope, involving filter columns and mesocosms. Based on these studies, the general view of performance on removals was TSS, 90 to 95%; metals, 80 to 95%; and nitrogen and phosphorus, 50%.

Influent concentrations have a bearing on effluent concentrations in bioretention systems. One researcher found copper concentrations in the effluent to rise from 3 to
10 µg/L with a rise in influent concentration from 40 to 160 µg/L. The effect of concentration on zinc, phosphorus, ammonia, and nitrate was inconclusive. These relationships indicate the need for aggressive source control where high-quality effluent is needed.

Not infrequently, a field study finds the removal efficiency of some pollutants to be negative; that is, the effluent concentration is higher than the influent. Furthermore, the increase in concentration for some pollutants is such as to suggest leaching from the filter media, particularly of phosphorus and nitrogen. However, due to extensive I/ET, upward to 90%, the loading reduction at the underdrain discharge is positive. As previously noted, engineers must use caution when calculating the loading reduction.
The likely lowest median effluent concentrations from bioretention filters are 5 to 10 mg/L for TSS, 15 to 25 µg/L for zinc, 5 to 10 µg/L for copper, and 50 to 125 µg/L for phosphorus. Significant source control must occur at dirty industrial sites to reach these concentrations.

TSS. With a few exceptions, numerous field, column, and pilot-scale indicate bioretention filters consistently reduce influent TSS concentrations to 5 to 10 mg/L. A review of five rain gardens in Minnesota found a range of 15 to 20 mg/L. It is likely that the modest TSS level in the effluent is of small solids leaching from the filter, although colloidal material in the stormwater likely passes through for the most part. Modeling equations have been developed to predict TSS concentrations.

Metals. Data of individual studies indicate that the lower limit of concentration based on current design criteria is on the order to 5 to 10 µg/L for copper and 10 to 20 µg/L for zinc. As previously noted, early laboratory studies commonly found removal efficiencies in excess of 90% due to influent concentrations higher than subsequently experienced in the field. Higher influent concentrations also resulted in slightly higher effluent concentrations as previously noted, indicating the importance of source control at highly polluted sites such as industrial companies and high-volume freeways where particularly sensitive water bodies are involved.

The study mentioned above also found the concentrations to increase over the length of the testing. Concentrations of copper gradually increased from a few µg/L to almost 20 at 800 bed volumes treated. Zinc in the effluent increased from essentially near the detection limit to slightly over 300 µg/L (influent of 640 µg/L) in 900 bed volumes. Assuming a wet climate of 36 inches of runoff per year, with a bioretention filter using 5% of the drainage area and a depth of 18 inches, 900 bed volumes occurs in seven and half years. The importance of the generation of dead organic matter providing new sorption sites is clear.
 
One study raises questions regarding the efficacy of bioretention as rain gardens at individual homes as a treatment device. The influent and effluent concentrations were below the detection limit, but researchers measured the amount of metals in the mulch and filter media at the beginning of the test and 15 months later. They found the mulch removed metals, but the filter media leached metals.

Overall, there was a net positive removal. This work indicates the need to do mass balances in which the pollutant concentration in the various components of the filter are determined before and after the laboratory or field study. It also illustrates that removal of metals is a sorptive process, which means that desorption can also occur when incoming concentrations are low.

While one researcher found a net positive removal when treating stormwater with very low influent concentrations, a study in Idaho did not with respect to copper. The drainage area was a school parking lot. The incoming median concentrations were below the detection limit, but the effluent concentration was 8 µg/L. This finding indicates the need to be cautious about where to require treatment if the leached pollutant is of particular concern.

Metal concentrations in the soil were found to differ throughout a bioretention filter, indicating differential filtration patterns as the stormwater flows into the filter basin. Highest concentrations were not at the inlet but about 3 feet inward, suggesting scour in the immediate discharge point. Regardless, differential filtration rates may be one factor decreasing the removal performance for the more difficult pollutants such as phosphorus and nitrogen.

Phosphorus. The likely lowest median effluent concentration from bioretention filters is 50 to 125 µg/L for phosphorus. The dominant unit process is likely mineral precipitation to iron and aluminum oxides on clays. Sorption to organic matter is a second mechanism. Negative removal has been observed often.
The increased phosphorus concentrations have generally been attributed to excessive phosphorus in the loam soil. As the common source of raw materials for the compost is yard waste, it is likely that the compost contains excessive phosphorus and nitrogen as well.

Disturbing the soil when preparing the filter media can result in temporary leaching of phosphorus irrespective of the initial soil content. The same may occur with nitrogen. Regardless, given the nature of solubility, a filter that leaches temporarily will never come to a point of removing the pollutants.

Although nutrient concentrations may increase, the total mass loading via the underdrain is significantly reduced due to infiltration. However, again the importance of looking at the total system—that is, the parent soil, the groundwater, and the surface water body—must be stressed when calculating system efficiency. Nitrogen is not effectively removed by any type of soil, and phosphorus decreases little in coarse soils.

One must also consider that what is important is dissolved phosphorus (and nitrogen), which is biologically available for algal growth. If less than half of the phosphorus (and nitrogen) is removed by the filter, it is the particulate phosphorus (and nitrogen), which is of little environmental consequence. The dissolved fraction does not change and may in fact increase due to leaching.

A standard soil chemistry test has been proposed, the purpose of which is to ascertain whether the phosphorus content of the loam soil is excessive. The Mehlich test produces an index of phosphorus availability to plants, but also indirectly an indication of the susceptibility of the soil to leach phosphorus. It has been proposed that the Mehlich index of the soil be less than 25. There are several related test procedures. Soil properties affect the selection of the most appropriate test procedure.
A review of 11 rain gardens indicates a range of effluent concentrations from about 0.05 to 0.20 mg/L, with influent concentrations ranging from 0.20 to 0.85 mg/L. There was no relationship between influent and effluent concentrations.

In one study the mean total phosphorus was reduced from 0.90 to 0.49 mg/L while the mean dissolved phosphorus increased from 0.10 to 0.29 mg/L. Another study in the same time period evaluating two facilities found phosphorus to increase from a mean of 0.044 to 0.061 mg/L. In effect, although there was a decrease in total phosphorus, there was an increase in the biologically available dissolved phosphorus.

This indicates the need to also analyze dissolved phosphorus in studies. Regardless, it is recommended that compost not be used where phosphorus is of concern. A better mix for dissolved phosphorus removal is a combination of dolomite (calcium magnesium carbonate), gypsum (calcium sulfate), and perlite.

Research of wastewater wetlands suggests that the cutting of foliage is not needed for biological phosphorus removal if the area loading of dissolved phosphorus is kept below 0.5 to 1 gm/m2/year. This amount of phosphorus is incorporated into the portion of the vegetation that is resistant to decomposition (lignins), particularly in anaerobic soils. However, bioretention filters are typically aerobic, where decomposition rates are much higher. There is need for area loading data versus performance with aerobic soils.

If phosphorus is of particular concern, as with eutrophying lakes or the implementation of total maximum daily load (TMDL) requirements, the appropriate approach is to include an amendment with the specific ability to remove dissolved phosphorus. There is also prepared media in the stormwater market in which the activated alumina is bound to light material like perlite or pumice.

A study found that best removal of phosphorus occurred when a soil of a high hydraulic conductivity was placed over a soil with low conductivity, in contrast to reversing the layers. Higher flow rates were also observed. The authors also stated that evidence suggests long-term reactions once attached to the soil will regenerate active short-term adsorption sites. What is likely happening is that the phosphorus initially sorbs to the relevant soil particles but over time transforms into a precipitate, opening up adsorption sites.

Nitrogen. The possibility of nitrogen leaching has not been examined in the same manner as phosphorus. Nonetheless, the data suggest that bioretention filters are not particularly effective at removing nitrogen. Data from individual studies support the conclusion made here. In some cases bioretention filters are likely exporters of nitrogen. Again, this view is based on the expected ineffectiveness of the parent soil beneath the filter to remove nitrogen, particularly in coarse soils, karst, and cracked basalt. The loading reductions are therefore misleading.

The likely reason for the poor performance within the filter is leachable nitrogen in the loam soil and compost. A review of the data indicates that it is not just the inability of the bioretention filter to remove nitrate, but also the inability to effectively convert ammonia and organic nitrogen to nitrate. Anaerobiosis may occur in the parent soil. However, experience with infiltration basins and rapid infiltration wastewater systems suggests that anaerobiosis does not commonly occur.
Nitrate may leach after a drying period, having converted from to nitrate from ammonia sorbed to clay during a previous storm. However, another study of the effect of drying periods found no difference in performance. In contrast, significant increase in nitrogen concentration following an extended dry period. This suggests nitrate lost, produced during the dry period from the use of ammonia that sorbed to clays during the prior storm. 

If so, at some point the outgoing will match the incoming, and one can only rely on the dilution of the infiltrated nitrogen by groundwater to lower concentrations. However, dilution will not alter the overall loading to the receiving water. Table 5 in part 1 of this article illustrates the possible benefits of plants during the growth season, but more likely the nitrogen-transforming bacteria.

However, in many studies the concentration is more than double the influent, indicating leaching. Possible factors limiting removal of nitrogen:

• Lack of anaerobic conditions in the lower part of the bioretention filter
• Inadequate conversion of ammonia and organic nitrogen to nitrate
• Inability of specialized bacteria to adjust to intermittent storms
• Insufficient dissolved organic carbon (DOC)
• Acidic pH of the stormwater
• Release by vegetation during the winter

Anaerobiosis (the proper term is anoxia) does appear to occur in small pockets within the filter as nitrate is frequently lower in the discharge, particularly when considering the concentration effect of ET.

It has been mentioned that one potential benefit of the lower chamber (or two-staged system) is the conversion of nitrate to nitrogen gas. But the chamber alone does not appear to be sufficient to create the necessary conditions for nitrate conversion. Several studies have found less than 5% difference with or without the chamber. It only takes modest anoxic conditions to experience nitrate conversion.

Regardless, there appear to be no studies of the redox potential in the lower chamber. The slowness of the infiltration from the lower chamber into the parent soil caused the lower chamber to be substantially anoxic. The removal of nitrate by a rain garden was improved with the addition of newspaper. The newspaper becomes a carbon source whose lack likely limits nitrate conversion to nitrogen gas.

There are several factors that may lead to the ineffectiveness of the lower chamber. One is inadequate flushing during each storm. This is unlikely. Regardless, as yet there have been no studies on flow patterns within these chambers. Another consideration is conversion of the nitrogen to nitrate. A review of the data suggests conversion of ammonia and organic nitrogen to nitrate is far from satisfactory, it being the necessary precursor to the eventual removal of nitrogen as a gas. The inadequate conversion suggests poor oxygen conditions n the media during and between storms.

A third consideration is the intermittent nature of storms. Dormancy affects nitrate reduction. Resting periods of 39 and 84 days were evaluated. Upon reexposure to nitrate in synthetic stormwater, full nitrate reduction did not occur for 14.5 and 30 hours, respectively, considerably longer than most runoff events.
It has been suggested that DOC is a limited factor. With wastewater it is common practice to add a liquid source of DOC to the nitrate reduction operation, as carbon is significantly reduced in the unit operations preceding nitrogen removal. The following equation is used.

C = 0.94N + 0.33D
where
C = needed carbon in mg/L
N = influent nitrogen concentration in mg/L
D = dissolved oxygen in mg/L

The influent concentration refers to the concentration entering the lower chamber. The carbon in untreated stormwater is consumed by aerobic bacteria, possibly leaving little for nitrate reduction. However, one would expect the plants through death and decay to provide the necessary carbon; similarly the compost and mulch. The equation suggests that with an influent concentration of 2 mg/L of nitrogen, about 2 mg/L of carbon is needed if the dissolved oxygen is zero.

However, carbon is also needed to consume the dissolved oxygen in the stormwater. Assuming well-oxygenated stormwater on the order of 8 mg/L and a nitrogen concentration of 2 mg/L in the influent, about 4.5 mg/L of DOC is required. It would seem that the overlying plants do not provide sufficient carbon. Nonetheless, the addition of a carbon source appears to be required to transform the nitrate reduction in laboratory experiments. That higher concentrations than 2 mg/L are needed suggests dissolved oxygen is also present in the stormwater as it enters the chamber.

A laboratory study explored added carbon in the form of alterative solid materials. Media amendments evaluated included sawdust, wood chips, alfalfa, mulch, wheat straw, limestone, and newspaper. All significantly increased nitrate conversion, some to 100% at an incoming nitrate concentration of 2 mg/L. Newspaper was identified as the most effective carbon source, although alfalfa, sawdust, wheat straw, wood chips, and limestone were essentially as effective as newspaper.

Nitrate reduction decreased significantly above a nitrogen loading of 6.5 mg/Day-N. The life of the newspaper was not determined, but it has been observed to be very resistant to decay. Absent a carbon source, the inclusion of a lower chamber was found not to improve nitrogen removal, but an improvement in copper removal was found.
Unrelated research supports the benefit of trees. A study of wastewater application to land found young trees reduce leaching of nitrogen through the soil (gravelly, loamy sand) and were more effective than grass in this regard. A partial solution is to use leaf compost, which has been found to be a modest remover of nitrogen. The loam soil, which appears to also be a source of nitrogen, could be totally excluded, limiting the mix to sand and leaf compost and/or Stalite.

In wet and semi-arid climates, the clay content could be on the order of 40%, necessitating somewhat lower filtration rates. More clay means more sorption of ammonia. The ammonia is then transformed to nitrate, making it available for final transformation to nitrogen gas. Regardless, it has already been noted that this author does not recommend the use of bioretention if the targeted pollutant is nitrogen.

Zeolite (chabazite) could be included as a more consistent sorber of ammonia than clay. Zeolite also removes nitrate. The zeolite can be mixed with the sand/compost or used as a layer beneath a sand/compost layer. Zeolite could be put in the lower chamber as well. A study found reductions of nitrate of about 30%, using coarse zeolite and operating at high flow rates. Influent values were 0.8 to 1.2 mg/L.

One study found that although nitrogen removal was positive, the amount of biologically available dissolved nitrogen was higher in the effluent that the influent. The addition of lime may be needed in some instances to give a neutral pH. A finer size of zeolite with slower rates should improve performance
considerably.

Indicator Bacteria. Studies indicate good removal of indicator bacteria with one exception. The researcher could offer no explanation as to why negative removal occurred. There is some evidence that it takes about a year for the filter to become fully effective, perhaps reflecting the time required for the predator population to increase to an effective level.

Americast in the testing of its Bacterra system found that removal was modest and sporadic in the first year. After one year, consistent reductions occurred above 95%. Americast’s Bacterra filter has been evaluated for bacteria removal. A laboratory bench study found in 80 to 99% (median about 90%) removal of fecal coliform at influent counts raging from 2,000 to 100,000 per 100 milliliters. The field results achieved somewhat lower removals: 70 to 99% with a median removal of about 80%.

Mixes with a high percentage of sand perhaps should be avoided due to the higher flow rates, possibly reducing removal. A high clay count, on the order of 40%, should improve bacteria removal. Sphagnum moss is a natural biocide that could be used in lieu of the compost. A ferric oxide coating on the sand can promote killing of bacteria. In column tests, average removals of about 88% were found, ranging from 55 to 99.7% depending on the test. However, influent and effluent counts were not given.
Pesticides. The removal of the herbicides has been found to occur in laboratory experiments.

Temperature. Temperature reductions on the order of 5 to 10°C during the summer months have been observed in western North Carolina. Others have studied this question. A study in Connecticut did not find any change, attributed to the drainage occurring off a roof that faced north.

Long-Term Performance
It can be expected that bioretention filters will remove dissolved metals, dissolved pesticides, and other organic xenobiotics without need to replace the filter media due to the organic matter provided by the plants and mulch. However, this is not the case for phosphorus. As noted previously, bioretention filters are not recommended when phosphorus and/or nitrogen are targeted pollutants.

As frequent inundation was found to decrease the biomass of turf grass in swales, perhaps a similar condition may be found in bioretention filters at least with respect to forbs and grasses.

In a study of actual bioretention filters, it was found that 43% of the units tested had filtration rates less than the recommended minimum of 2 inches per hour (50 millimeters per hour) due either to clogging or inadequate specification of the filter media.

Use With Underdrains
A minimum infiltration rate of 1 inch (2.5 centimeters) per hour is commonly specified. Absent good infiltration rates, an underdrain system is included. Bioretention becomes a sorptive filter. Fabric is typically placed in fine-media filters to separate the media from larger underdrain material, as well as on the underdrain. The purpose is to prevent mixing of the two materials and migration of the finer material into the coarse. Agricultural experience indicates that depth and spacing of soil drains affects nitrogen retention.

Anecdotal information suggests clogging of fabric placed in bioretention facilities, likely due to the migration of clays and silts present in the media. The alternative is to place a pocket of gravel around the drainage pipes. In this case the fines presumably migrate through the rock into the effluent pipe. Clogging is avoided but performance is reduced.


Sizing
The retention volume is divided by the allowable depth to give the surface area. Another method frequently used is that used for sizing sand filters. This makes sense inasmuch as the bioretention filter is essentially a sand filter.

The system is recessed and contoured to contain the volume of the design event. Bioretention systems are placed on soils with a minimum infiltration rate of 1 inch (2.5 centimeters) per hour. However, as the common minimum infiltration rate for standard infiltration devices is 0.5 inch (1.25 centimeters) per hour, there is no reason why this design criterion cannot be used.

Eight methods for sizing were compared. The required surface area was determined assuming a filter depth of 55 centimeters and a hydraulic conductivity of 10 centimeters per hour. Areas varied by a factor of five. The smallest area was quite effective at filtering the water even though the drawdown time was less than 24 hours. The longest was 15.5 hours with a maximum ponding depth of 6 inches (15 centimeters). This suggests that most sizing methods grossly oversize bioretention filters. The percentage of development area ranged from 5 to 24%.

Maintenance
Suggested maintenance tasks are presented in Table 3. Clogging has not been of concern because the mulch protects the surface of the filter. Field studies have shown that hydraulic conductivity and/or infiltration rate remain at desirable levels after several years of operation and sometimes improve. However, other data suggest the surface should be cleaned about every five years. That mulch removes the sediment suggests that the old mulch should be removed prior to placing new mulch.
Trees and shrubs move some phosphorus and nitrogen in their foliage to the woody parts in the fall in temperate areas. However, most of the annual intake by shrubs and trees is returned to the soil surface as leaves and twigs. Consequently, sustainable long-term removal of dissolved phosphorus requires that the litter be removed. Shrubs and trees should be pruned, motivated by aesthetics as well. Harvesting also benefits the removal of nitrogen and metals. But it is the removal of the fallen plant debris that provides a greater benefit.

Trees and shrubs provide more storage capacity than grass for nutrients and metals. Whereas grass likely has to be cropped several times per year to take advantage of its uptake of pollutants, shrubs and trees can be pruned once every few years to gain the same benefit. These observations are conjectural at this time. Research indicates that plant species differ in their uptake of nitrogen and phosphorus, suggesting the basis for specific planting strategies. An advantage of trees is the many years to mature, sequestering several pollutants over an extended period of time and removing the necessity of frequent harvesting, as would be the requirement for turf grass.

Manufactured Filters
There are currently several manufactured bioretention systems: UrbanGreen BioFilter from Contech, Filterra, and POD. These products differ from the public-domain filter in that they have a very high flow rate, on the order to 35 to 100 inches per hour where an underdrain is included. As such, they require very little area, on the order of 0.5% rather than 5% for standard bioretention filters. The high flow rate is achieved because the media is a mix of small gravel, sand, and compost. The products give good performance, particularly considering the high flow rate.

Given their small area, they must be considered a treatment device only, with very little hydrologic benefit. Very little of the stormwater will be infiltrated except in an outwash soil, where the filtration rate of the system must be matched and therefore sized consistently with the infiltration rate of the underlying soil. They therefore can be used where runoff rates and volumes are not of concern, such as discharges to lakes and rivers.

References

A large number of references were used in the preparation of this article, too many to list here. The reader is referred to the book Stormwater Treatment: Biological, Chemical, and Engineering Principles for a complete listing of the references.

Author's Bio: Gary R. Minton, Ph.D., P.E., is an independent consultant on stormwater treatment with Resource Planning Associates. He is the author of the book Stormwater Treatment: Biological, Chemical, and Engineering Principles.