Areas with shallow water tables are vulnerable to stormwater and agricultural runoff contamination, as high contaminant loads are introduced from untreated surface water runoff during storm events, resulting in water quality issues, beach closings, and impairing the economic activities of (coastal) communities [1
]. Despite the risks associated with stormwater runoff contaminants, very few stormwater treatment regulations exist. In the United States, Rhode Island is at the forefront of stormwater regulations, requiring any new construction to include stormwater BMPs that remove at least 85% of total suspended solids (TSS), 60% of influent bacteria, and 30% of nitrogen and phosphorous [7
The majority of BMP systems are not designed to treat the wide range of contaminants that are typically found in stormwater runoff, e.g., heavy metals, PAHs, and bacteria [9
] and it is unclear how performance is influenced under varying weather conditions, such as those experienced in the Northeastern United States.
A popular type of BMPs is a stormwater retention system. While these BMPs may treat total suspended solids (TSS) effectively [11
], few systems exist that address bacterial removal. Two of such systems are BactoLoxx
(Filtrexx, Goffstown, NH), and Bacterra
(Filterra Bioretention Designs, Ashland, VA, USA). These systems rely on proprietary flocculation agents that result in the settling of bacteria [10
], and on physical filtration and predation from biomat formation [9
], respectively. The stated bacteria removal efficiency for these two stormwater BMPs reaches up to 99%. However, recent studies on bacteria removal in a wide range of structural BMPs concluded that bacteria treatment was ineffective and unreliable, as the removal of pathogens primarily relied on attachment/collection and not inactivation [13
]. This is because sorbed pathogens can remain viable during attachment [15
] and therefore can be remobilized/detached during intermittent flow conditions. The ineffectiveness in inactivating bacteria is especially relevant in areas with shallow water tables, because not enough soil depth might be present to filter out pathogens effectively, therefore increasing the risk of bacterial contamination of groundwater [18
While BMP manufacturers, such as Filtrexx and Filterra offer filter modules that have to be added to a BMP to treat additional contaminant categories, other multi-contaminant BMPs, such as biofiltration systems, remove 0%–80% of bacteria [19
]. This is considered insufficient from a public health perspective [1
]. Therefore, new approaches are required to enhance the inactivation of pathogenic organisms as well as the removal of other contaminants in stormwater and possibly implement these technologies in the next generation BMPs.
In addition, varying climatic conditions such as those experienced in the Northeastern United States, including cold winters, spring snowmelt, and long intermittent dry periods persist in between intense precipitation events in summer, may impact system performance. While studies have carried out BMP evaluations in regions that do not experience extreme temperature variation [19
] it is imperative to evaluate BMP performance in conditions characterized by long dry periods, intense precipitation events and cold temperatures that might freeze the system.
Studies have shown that filter materials amended with antimicrobials can remove bacteria from aqueous solutions by inactivation processes that damage the cell [15
] or permanently fixating bacteria on porous media [16
]. For instance, in a previous study antimicrobially amended wood chips were tested with intentions for use in stormwater BMPs [15
]. The laboratory derived data indicate that red cedar wood chips loaded with 6 mg/g of 3-(trihydroxysilyl)propyldimethyloctadecyl ammonium chloride (TPA) effectively removed E. coli
from aqueous solution by at least two Log10
units (>99%). In addition, studies have shown that wood is capable of almost quantitatively removing aqueous phase PAH [30
], dissolved metals and other inorganic compounds [34
]. While the E. coli
removal rate of the amended wood is substantially greater than required by some states, the performance of the filter material was only tested under laboratory batch conditions [15
]. Additional studies are therefore needed to evaluate the performance of this promising antimicrobial treatment approach under field conditions.
The need for multi-contaminant runoff treatment systems and innovative BMP designs was addressed in this study by modifying a conventional tree filter BMP with an antimicrobial agent, TPA. A tree filter (TF) is a structural BMP that relies on filtration and bioretention in which plant, soil, and microbial activity remove pollutants through a variety of physical, chemical, and primarily aerobic biological processes. A typical TF collects runoff and entrained sand and sediment, which enters the TF via curbside openings. Depending on the manufacturer, the runoff immediately flows into an approximately 100 cm deep “pretreatment” sump (here referred to as the catch basin), located within the interior of the concrete frame of the TF structure. Heavier sediment settles out and is retained in the catch basin. As water continues to enter and fill the catch basin, the water level rises and ultimately spills over the catch basin wall to enter the main portion of the TF, which is composed of a layer of engineered mix of permeable sand and expanded shale. This matrix has a water holding capacity that supports healthy tree growth and, according to the manufacturer, has chemical and biological treatment capabilities. Finer sediments in solution, which were not captured in the catch basin, are primarily trapped and held within the pore spaces of the sand and shale matrix. Water flows downward and the “cleansed” runoff ultimately infiltrates to the groundwater zone in TF systems with open design (as used in this study). TF systems with closed box design discharge the infiltrating runoff to the public stormwater drainage system.
The objective of this study was to evaluate two TFs (1) a conventional tree filter with a sand/shale mix as a filtration matrix and (2) an innovative TF with an additional layer of red cedar wood chips amended with an antimicrobial agent, TPA, to address the removal of bacteria, nutrients, heavy metals, and PAH pollutants from stormwater runoff. We hypothesize that the new media will increase the removal of inorganic and organic pollutants and bacteria in the field when compared to the performance of a conventional TF system. The results of this study may aid stormwater and water quality managers in deciding which BMP technology to select for future projects.
2.1. Site Description
The study site was located on a commuter parking lot on the northern end of the University of Rhode Island, Kingston campus. The dominant use of land surrounding the parking lot and the university in general is agricultural. The parking lot has a conventional asphalt cover with 458 parking spaces covering 4.9 acres (19,830 m2
; Figure 1
) and is part of the Rhode Island Stormwater Demonstration Facility (RISTDF) which is situated over the Chipuxet Aquifer in southern Rhode Island. The aquifer is overlain by a thin layer of loess and below it contains complexly interbedded lenses of sand and gravel with smaller contributions of silt and silty sands [35
]. These glacio-fluvial deposits were deposited during the Pleistocene epoch [35
]. The stratified material is approximately 60 m thick and overlies fractured granitic bedrock. The depth to groundwater at the site is approximately 7 m below ground surface (bgs) [36
]. Based on a 20 year data set starting in 1994, the field site receives an average of 1202 mm precipitation per year, with an annual average of 841 mm of snow. During this period an average of 77 days per year received precipitation equal to or greater than 2.54 mm (0.1 inch), 31 days received greater than or equal to 12.7 mm (0.5 inch) of precipitation and 12 days per year received greater than or equal to 25.4 mm (1 inch) [37
The two tree filter systems were provided by Storm-Tree Inc., Providence, RI, USA and installed on the RISTDF site, which was constructed with funding from the Rhode Island Department of Transportation and the URI Transportation Center in August 2013. The TF dimensions and their catchment areas are summarized in Table 1
, with the conventional TF (CTF) being slightly smaller (0.64 m3
) than the innovative TF (ITF) (1.13 m3
). Both TF units contained a mixture of sand and expanded shale and 10% sphagnum peat as the filter media. The ITF unit had an approximately 8 cm (3 inch) layer of 30 kg (0.10 m3
) bioactive red cedar wood chips added to the sand and expanded shale mix (1.13 m3
). Both TF units are open concrete structures with a catch basin. The open structure on the bottom of the TF allowed the filtered runoff to infiltrate into the ground. The catch basin retains sediment and other debris present in runoff. However, the load of sediment captured by the TF systems was not quantified in our study. The contributing area of two filters as well as the units themselves varied slightly in size with the unit containing the amended red cedar wood chips being larger and serving a greater contributing area (Table 1
). A red maple tree (Acer rubrum
) was planted in the units to complete the TF system. Red maple was chosen to match the surrounding trees in the parking lot.
During installation of the TF systems, a stainless steel pan with an area 0.072 m2 was installed at the bottom of the excavation pit approximately 1.5 m (5 ft) below ground surface (bgs), i.e., 71 cm (28 inch) below the infiltration surface of each TF. The pan drained collected water through a central hole connected to a Teflon tube leading to the surface. Water collected in the pan was pumped to the surface using a peristaltic pump. In addition, soil water and temperature sensors were installed at the bottom of the filter (Decagon Devices, Pullman, WA, USA; Figure S1).
Meteorological data was obtained from a National Climatic Data Center station that is located less than 0.6 km from the experimental site.
2.2. ITF Preparation
Following the procedures described by Kasaraneni et al.
, the red cedar wood chips were amended with 6 mg/g 3-(trihydroxysilyl)propyldimethyloctadecyl ammonium chloride (TPA), a quaternary ammonium compound [29
]. Briefly, the red cedar was soaked in frequently exchanged clean water for a minimum of 2 weeks to leach out soluble matter and remove fines. Next, a solution containing 750 mg/L TPA and wood in a ratio of 25 L to 1 kg was stirred for 9 days or until the aqueous concentration of TPA solution did not change and maximum sorption to the wood was achieved (Figure S2). The amended red cedar wood chips were dried, weighed, and stored in a dark, dry location before being added to the tree filter.
2.3. Sampling and Analysis
From November 2013 until August 2014 composite runoff in the catch basin (influent) and the stainless steel pan (effluent) were sampled weekly. The samples collected as part of this sampling effort can be described as composite samples, as the collection pan buried at the bottom of the TF collected an integrated sample of the runoff while the influent collected from the catch basin is a time-integrated composite sample. In addition, six individual rain events were sampled during this period. These samples were collected at least 30 min after the start of a rain or snow melt event and the influent was collected as a grab sample before it entered the catch basin. The effluent was collected 20–30 min after the influent sample (and after the water from the catch basin entered the infiltration area). The lag time between the sample events reflects hydraulic retention time in the BMPs. While the main focus was on changes in E. coli and fecal coliform (as fecal indicator bacteria) concentrations, all samples were analyzed for pH, specific conductance, temperature, chloride, nitrate, orthophosphate (starting in July 2014), total suspended solids (TSS), heavy metals (Cd, Cu, Ni, Pb, Zn) and polycyclic aromatic hydrocarbons (PAH). Details on the analytical techniques used can be found in the supplemental information.
Cadmium was never detected in runoff samples and therefore not further reported. The removal of copper, lead, and zinc could not be quantified because during both sampling seasons, these metals leached from the brass fitting and welds on the stainless steel sampling pan. Similar observations were reported by Schock and Neff [38
For both seasons the long-term mass load for each constituent in the influent and effluent was calculated by first calculating the runoff volume:
is the runoff volume per sampling date, Pi
is the precipitation amount (m), A
is the contributing area (m3
is the runoff coefficient (0.9 for impervious surfaces), L
are losses incurred otherwise (0.5 due to losses to drains) and then multiplying it by the mass load for the influent and effluent for each event:
is the total mass of the pollutant for each event, V
is the runoff volume generated by the precipitation event, and C
is the concentration of the constituent measured in the discreet influent and effluent sample that was collected.
A two tailed t-test was carried out using R software version 3.0.3 with α = 0.05 to statistically compare the storm event removal percentages for TSS, nitrate, phosphate, E. coli, and PAH from the two tree filters. Metals were not statistically analyzed because metal data was inconclusive.
2.4. Controlled Field Tracer Experiments
Two separate tracer tests, using a solution containing sodium chloride (NaCl) and E. coli
, were carried out within one week of each other in September 2014 under similar climatological and initial filter matrix saturation conditions (Table S1). The E. coli
tracer was carried out to investigate microbial transport in the system, while the sodium chloride solution was used as a conservative tracer to estimate the hydraulic characteristics of the system. Both filters were conditioned by pumping clean non-chlorinated water directly into the TF, by-passing the catch basins, at about 3 L/min (circa 200 L) using an electric submersible pump and garden hose (Figure S3). The water flow was metered with a 2.5 m3
/h ISTEC Flowmeter (Sparta, NJ, USA). After approximately 70 min of pumping clean water, the tracer solution containing NaCl and GFP E. coli
was injected. The green fluorescent protein (GFP) strain of E. coli
(BTF 132) (Biomérieux, Hazelwood, MO, USA) was the bacterial tracer. GFP labelled E. coli
are ideal for use as a bacterial tracer [39
] because bacteria colonies glow green under UV light. This allows for differentiating E. coli
added as a tracer from those native to the TF system.
The GFP E. coli were cultured in LB broth (10 g/L sodium chloride, 10 g/L tryptone, and 5 g/L yeast extract; Sigma Aldrich, St. Louis, MO, USA) at 37.5 °C for 13 h. After culturing, the bacteria were removed from the LB broth, washed and stored in phosphate buffer solution (11.2 g/L K2HPO4, 4.8 g/L KH2PO4, and 20 mg/L ethylenediaminetetraacetic acid; all Sigma Aldrich; pH: 7.3).
The NaCl concentration in the tracer solution was 888 mg/L (428 L) and 969 mg/L (392 L) for the conventional and the innovative TF experiment, respectively. The initial concentration (Co) of GFP E. coli in the solution was 6.90 × 105 CFU/100 mL and 8.32 × 106 CFU/100 mL for the conventional and the innovative tree filter, respectively. The tracer solutions were pumped over the course of 120 min and 145 min, respectively. Afterwards, an additional 207 L of clean, non-chlorinated water was added to each of the two TF over a period of approximately 70 min.
Periodic influent samples were taken throughout the tracer test to monitor natural decay of E. coli in the tracer solution. Using a Masterflex Environmental Sampler (Vernon Hills, IL, USA), effluent was pumped constantly from the buried sampling pan inside the TF at a rate of 200 mL/min for the conventional and 225 mL/min for the innovative system (Figure S4). Every 5 min, the effluent was metered for specific conductance using a YSI Professional Plus Multimeter logger (Yellow Springs, OH, USA). Every 15 min samples were collected for bacteria analysis and immediately processed at the onsite RISTDF field laboratory. The changes in the soil moisture content and the breakthrough of the wetting front were automatically recorded by sensors buried in the TF systems (Decagon Devices, Pullman, WA, USA).
Two filter units, one conventional with a sand/shale and peat mix and one innovative filter matrix, containing an additional layer consisting of antimicrobially amended red cedar wood chips were compared for their contaminant removal capabilities in a field pilot study. Over the course of a one year period starting in November 2013, TF influent and effluent was sampled for organic and inorganic constituents, including E. coli. Overall, the ITF exhibited higher nitrate, phosphate, TSS, E. coli and ∑PAH16, removal compared to the CTF. However, the differences in the removal rates between the two TF systems did not differ significantly when compared statistically.
While E. coli
removal was higher in the ITF compared to the CTF, even higher removal was expected based on the results of previous laboratory studies [15
]. However, this can be explained by the TPA loading to the red cedar wood in this field test being 6 mg/g, whereas it could have been as high as 9 mg/g, based on prior studies [15
]. Second, the amended red cedar layer installed inside the TF was 8 cm (3 inch) thick and only facilitated a 2 to 3 min residence time when flooded with stormwater. The amount of contact time between the stormwater and the antimicrobially amended material was insufficient as shown in subsequent laboratory column studies [15
]. Besides these system design elements, it was likely disadvantageous to have added the wood chips as a discrete layer rather than mixing them into the sand/shale matrix. A more homogenous mix could have extended the contact time of bacteria and TPA red cedar during the treatment process. Finally, whereas artificial stormwater was used to prove the concept of E. coli
treatment by TPA amended wood chips in the laboratory, under actual field conditions the composition of the stormwater varied over the seasons, including high loads of dissolved salts, TSS, and other compounds that likely altered the TPA covered surface of the wood matrix. In particular, it must be assumed that fines were deposited on the wood surface and covered the bioactive TPA. In consequence, bacteria in the stormwater were less likely to come in contact with the TPA.
The next generation of the amended TF systems would likely benefit from installing the amended wood chips at the bottom of the filter unit or mixed into the sand/shale matrix rather than on the top of the TF unit, where fines are present in high concentration during infiltration events. Further tests with other pathogens aside from E. coli should be conducted to determine if the amended red cedar has the ability to remove other pathogens from stormwater runoff.
While the bacteria removal of the amended TF during the monitoring study was not as high as expected, the absolute bacteria treatment efficiencies of both TF systems, together with those for TSS and nutrients, exceeded the requirements for BMP systems in Rhode Island. Further, the evidence for inactivation of bacteria during the field tracer experiment in the ITF indicated that the addition of the amended material reduced the amount of culturable cells during pretreatment compared to the CTF. This makes TF technology competitive with other structural BMP systems currently on the market for agricultural and urban runoff treatment and highlights the importance of incorporating antimicrobial amendments into runoff BMPs.