NONPOINT SOURCE POLLUTION ASSESSMENT AND MANAGEMENT OF AIR, SURFACE- AND GROUNDWATER IN RHODE ISLAND

Nonpoint source (NPS) pollution is one of the major sources of environmental contamination in the United States. Depending on the contaminant and its origin, NPS pollution can be introduced through atmospheric deposition or in the aqueous phase, for example through stormwater runoff. Here, both the atmospheric deposition of 16 EPA priority polycyclic aromatic hydrocarbons (PAHs) was measured at six different sites in Rhode Island as well as their removal, including other contaminants, from stormwater runoff. For the first part of this study seasonal and spatial variation in atmospheric PAH deposition rates were quantified along with investigating the source of the PAHs. The atmospheric deposition of PAHs followed an urban-to-rural gradient with deposition rates significantly higher in urban areas (up to 2261 μg m -2 yr -1 ∑PAH16) compared to rural areas (as low as 73.6 μg m -2 yr -1 ∑PAH16). In fall and winter, PAH deposition was up to 10 times higher compared to summer/spring. In part two of this study a stormwater best management practice (BMP) was evaluated for its contaminant treatment efficiency, with specific attention paid to PAH removal and microbial load reduction. Because many BMPs are not designed to treat bacteria specifically, a novel engineered material, amended red cedar wood chips with different loadings of two antimicrobial compounds: 3-(trihydroxysilyl)propyldimethyloctadecyl ammonium chloride (TPA) and silver nanoparticles (AgNP), was developed and tested in the laboratory to target microbial load reduction and removal of stormwater cocontaminants, including PAH. Laboratory tests showed that E. coli inactivation, at 25°C, can be achieved with log10 removal values (LRV) up to 3.71 ± 0.38 (mean ± standard error) for TPA-red cedar and 2.25 ± 1.00 for AgNP-red cedar, while unmodified red cedar only achieved 0.45 log10. Similarly, PAH removal from the aqueous phase was 68.9 times higher using TPA modified red cedar compared to the conventional shale used in the tree filter BMP. After the screening under laboratory conditions, bioactive TPA amended red cedar wood chips were tested in a pilot-scale field project using a tree filter BMP. The performance of the amended wood chips was compared to a commercially available tree filter BMP that contains a mixed sand/shale filtration matrix. Under pilot test conditions, the two filters performed similarly with regards to bacteria removal. But, the tree filter containing amended red cedar was significantly more effective at removing PAH. As demonstrated in the first part of this study, PAH influxes to stormwater are not limited to local sources, e.g. automobiles, but also have an atmospheric deposition component. Also, because PAH are just one of many contaminant classes present in stormwater runoff, it is therefore important that innovative BMP technologies have multi-contaminant treatment capabilities that can effectively prevent contaminants like PAHs, heavy metals, and/or bacteria from entering the environment. While this study has shown that tree filters can be modified to meet that expectation, more work is necessary to develop, improve and demonstrate a full-scale BMP device.


ACKNOWLEDGMENTS
This research has been funded by the Rhode Island Department of Transportation, the University of Rhode Island Transportation Center, and four years of support from the Enhancement of Graduate Research Awards from the URI graduate school.
First and foremost, I would like to thank my major professor Dr. Thomas Boving. Tom was instrumental in guiding me through the process of obtaining a PhD; from sharing his passion for concrete mixing to expanding my knowledge in the field of hydrology and giving me the chance to present at various conferences in the field.
Tom's support was unconstrained and was only enhanced through our common mother tongue, German. I would also like to thank two very powerful women on my committee, Dr. Dawn Cardace and Dr. Vinka Craver, for all their invaluable input on scientific advice and mentoring and proving that success in the science and engineering world is not tied to a gender. My thanks also go to Jose Amador and Rainer Lohmann, both of whom have played important roles as mentors and teachers.
Overall, the combined knowledge of my committee has been encouraging and fruitful for my professional advancement and academic growth.
Without Kevin Broccolo's help in constructing research equipment and his wide creativity, several portions of this project would have been nearly impossible to carry out.
I want to recognize the ongoing support I received from the faculty and students of the URI Geoscience department from the day I stepped foot in Rhode Island. All of you have made me feel at home for the past five years and I could not have asked for a better support network. Simple encounters at URI, such as being classmates (E.  Table 3.1.Summary of winter (November through March) and summer (April through September) average ± standard deviation influent concentrations and ranges in parentheses along with reported literature values for the following parameters: pH, specific conductance (SC, µs/cm), chloride (mg/L), nitrate(mg/L), phosphate (mg/L), total suspended solids (TSS, mg/L), E. coli (CFU/100 ml), copper (µg /L), nickel (µg /L), lead (µg /L), zinc (µg/L), and ∑PAH 16

INTRODUCTION
Nonpoint source (NPS) pollution is a type of contamination that is introduced into the environment from a non-traceable or several sources, i.e. not a single spill or leak. Therefore, NPS pollution is neither limited spatially nor temporally. It is also not restricted to one type of contamination and contaminant phase (e.g. soil, water, or air).
Any location ranging from urban areas to coastal regions and mountainous terrain may be impacted by one or more types of NPS pollution (Carrera et al., 2001;Zehetner et al., 2009;. This can include, but is not limited to atmospheric deposition of pollutants and runoff from agricultural, industrial areas, or roadways Kaushal et al., 2011). Because NPS pollution impacts all environmental compartments it is important to understand contaminant fluxes so that managing and prevention of NPS pollution results in improved water and air quality as well as ecosystem health.
Even though NPS pollution encompasses a wide range of contaminants, such as nutrients, heavy metals, suspended solids, organic contaminants, and pathogens, the emphasis of this study lies in one type of organic contaminants, namely polycyclic aromatic hydrocarbons (PAH), and one type of indicator bacteria that is commonly used to monitor water quality, specifically Escherichia coli (E. coli).
As one of the major groups of persistent organic pollutants, PAHs are typically not produced at large quantities for any specific industrial uses. Rather, PAHs are byproducts of many combustion processes from which they get released into the atmosphere Lima et al., 2005). PAHs are suspected carcinogens and their presence in the environment poses a health risk to humans ; however, PAHs are part of the human lifestyle. Common combustion processes that produce PAHs can include wood burning, residential heating, automobile combustion, just to name a few (Ravindra et al., 2008;. Once emitted through combustion processes, atmospheric transport of PAHs causes diffuse pollution of whole landscapes as PAHs settle out of the atmosphere via dry particle deposition, gaseous deposition, and wet deposition (Bidleman 1988). This can result in the contamination of surface waters, soils, and eventually groundwater, even in rural areas Lohmann et al., 2000).
As discussed in Manuscript 1, the exact emission source of PAHs in the atmosphere can rarely be identified, but general emission source types can be fingerprinted when atmospheric PAH samples are collected. Several studies have identified PAH profiles that correspond to specific emission sources, such as wood, gasoline, diesel, and oil combustion (Ravindra et al., 2008;Rogge et al., 1993;Motelay-Massei et al., 2007;. Using these fingerprints and statistical source attribution models allows for identification of emission sources that contribute to PAH contamination in a specific region. In addition, PAHs can also be found in roadway runoff, one of the major nonpoint sources within the United States . When present in runoff, PAHs commonly stem from gas and oil drippings, leaks, spills, or coal-tar sealed impervious structures such as parking lots 3 . Stormwater runoff containing contaminants including PAHs introduces high loads of contaminants into soils along road sides or nearby surface water bodies and estuaries during the first flush of rain events . The physicochemical properties of PAHs lead to accumulation of these compounds in soils and sediments, significantly impacting the ecosystem (Zehetner et al., 2009;Krein and Schorer, 2000).
Besides PAHs, pathogens are another prevalent contaminant class in stormwater runoff. Within that contaminant class, E. coli and fecal coliforms are typically monitored in watersheds, as their presence is considered an indicator of pathogenic microbial contamination (EPA, 2012). Microbial contamination is often traced back to stormwater runoff and its uncontrolled release into the environment can cause public health concerns. For instance, shell fishing bans or precautionary beach closures are often implemented in the interest of public health Soller et al., 2014;. Hence, bacteria contamination can have significant economic repercussions in areas like Rhode Island that rely strongly on fisheries and the tourism industry .
Manuscripts 2 through 4 focus on bacteriological contamination of stormwater runoff and how it, together with other pollutants (mainly PAH) can be possibly attenuated by innovative structural best management practices (BMP). While stormwater runoff itself cannot be prevented, the contaminant flux associated with it can be remediated through the use of structural BMP. Several states in the U.S. have started implementing structural BMPs to improve the quality of stormwater runoff in 4 addition to other benefits like flood mitigation (Bakeman et al., 2012;MDE, 2009;PA DEP, 2006;VT ANR, 2002;. Rhode Island has gone further and set minimum contaminant removal standards that structural BMPs should comply with . However, many BMPs are not designed to treat bacteria. Building on research focused on point-of-use drinking water filtration methods that rely on bioactive compounds, such as nano-silver or quaternary ammonium compounds, this research investigated if these bioactive compounds can find use in the enhanced treatment of pathogens in stormwater runoff. This project was undertaken with the goal to identify an improved filtration matrix that can be implemented in stormwater BMPs .
As this and related research shows, the contaminant influx through atmospheric deposition of PAHs can be orders of magnitude lower than that of stormwater. Hence, ATM deposition of PAH can be detected everywhere and anytime, even in remote areas and during times of little rainfall. Conversely, stormwater runoff impacts areas like New England more significantly because frequent precipitation events result in the quick spreading of locally derived contaminants. Therefore, the deposition of atmospheric pollutants along with the management of locally derived pollutants associated with stormwater runoff must be incorporated in the design of sustainable management strategies for nonpoint pollution sources. 8

INTRODUCTION:
Atmospheric non-point sources, such as automobile exhaust and by-products of other combustion processes, introduce gaseous and particulate contamination to both urban and rural areas. Even though contamination is highest in regions closest to emission sources, often urban and industrial areas, atmospheric deposition can take place far away from the original emitting source(s) and can lead to diffuse contamination of whole landscapes (Lohmann et al., 2000;Lima et al., 2005;Gocht et al., 2007a;. With fossil fuel consumption increasing in response to a growing population and vehicle density, urban areas experience increased loadings of atmospheric contaminants, especially polycyclic aromatic hydrocarbons (PAH) . Many of the fossil fuel derived contaminants introduced into the atmosphere, including PAHs, are toxic and suspected carcinogens . Thus, increasing levels of PAHs in the environment result in health and safety concerns of the general public .
There are local and regional sources of PAH. A prominent local contamination source is linked to traffic. Although stricter environmental regulations over the past years decreased the emissions from individual cars, much of the gain is outweighed by the increase in the number of registered automobiles as well as urban sprawl, an increased number of roadways, and the construction of parking lots and other (coal tar) sealed surfaces that volatilize petroleum hydrocarbons to the atmosphere Lima et al., 2005;. Another prevalent local source of PAH in winter results from wood combustion 9 Khalili et al., 1995;Lohmann et al., 2000). In New England where wood burning is a very popular means of heating homes during the cold season, this source can locally contribute upwards of 70% to atmospheric PAH contamination Kroetz and Friedland, 2008).
These regional inputs from combustion and burning processes are linked to atmospheric deposition and include dry particle and gaseous deposition, as well as wet deposition Gocht et al., 2007a). The combined contribution of gaseous and particulate phase deposition, including precipitation derived compounds is referred to as bulk deposition.
Besides reflecting actual differences in the deposition rates over space and time, the influence of different PAH deposition sampling techniques cannot be ignored. Liu et al. (2007) reviewed sample collection, pretreatment and analytical methods for quantifying PAH in airborne particulates and the gaseous phase and concluded that even though passive and active sampling techniques are available, neither technique is optimal for bulk (gaseous + particle bound) sampling of PAHs. This is because active sampling techniques employ high volume air filters that trap particle bound, mostly heavy molecular weight PAHs on a filter that is later extracted. Passive sampling systems rely on sorptive processes; therefore trapping mostly more volatile, low molecular weight compounds (Liu et al., 2007). In Europe, researchers have tested and verified a passive sampling technique that can capture the time-integrated long term atmospheric bulk deposition rates of PAH and other compounds (German Industrial Standard DIN 19739-2, 2003).
Studies in Europe and the eastern United States quantified deposition rates of PAH to land surfaces using both, passive and active sampling systems, often report 2-6 times higher deposition rates for particulate and gaseous phase samples in proximity to urban areas relative to rural sites that are considered distant from potential PAH sources Gocht et al., 2007a;. For instance,  investigated the atmospheric PAH concentration in the particulate and gas phase using an active sampling system in the metropolitan area of New Jersey. In the urban area of New Brunswick gas phase ∑PAH 36 concentrations were on the order of 3.5 to 84 ng m -3 while particulate ∑PAH 36 concentrations were 0.38 to 11.6 ng m -3 compared to 2.8 to 42 ng m -3 and 0.15 to 4.0 ng m -3 for particulate and gas phase ∑PAH 36 concentrations at a coastal site in Sandy Hook, NJ, respectively. This resulted in bulk annual deposition rates of up to 34.8 µg m -2 yr -1 for the urban area and 16.8 µg m -2 yr -1 for the coastal area . Gocht et al. (2007a)   . Overall, the PAH deposition rates from studies on both sides of the Atlantic differ by an order of magnitude.
Deposition of PAHs does not only vary spatially, but also temporally. For instance, in study areas in Europe and the eastern United States, 2-3 times higher deposition rates were observed during the cold season and attributed to increased 11 residential heating with fossil fuels Motelay-Massei et al., 2003;Gocht et al., 2007b). In addition, particle bound PAH concentrations are often elevated during the cold months compared to the warmer months because of the association of PAH with atmospheric soot particles with combustion origins .
The advantage of using a passive sampling system is that natural, seasonal bulk deposition rates can be easily determined because no maintenance or electrical power is required to operate the sampler. Also, while active sampling techniques can capture atmospheric deposition over shorter periods of time by pumping air through a filter at a specific rate , they may produce artificially high deposition rates due to temporarily high or low emissions sources in the vicinity of the measuring station. This can possibly lead to misrepresenting longer term depositional patterns .
Another important factor in comparing PAH studies lies in the reporting of results. Many studies, including this one, use the sum of the 16 priority PAHs identified by the U.S. EPA (∑PAH 16 ;EPA, 2006) to quantify PAH deposition rates (Liu et al., 2007). However, some studies report the sum of 30 or more PAH Okuda et al., 2002;Naumova et al., 2003; (Liu et al., 2007); however, the sources of contamination origins cannot be specified in this manner.
To identify individual combustion sources in atmospheric PAH samples, ratios of environmentally stable PAH isomers that are characteristic in source combustion PAH signatures can be employed Ravindra et al., 2008;. These isomer ratios are beneficial for assigning general emission sources, however their use has been criticized because differential reactivities of the PAH isomers may skew the emission source signature . A multivariate factor analysis, such as Positive Matrix Factorization (PMF) Larsen and Baker, 2003; circumvents this problem when apportioning sources of airborne particulate matter in the atmosphere Paatero and Tapper, 1994;Larsen and Baker, 2003;Ravindra et al., 2008).
Herein, PMF was adopted to evaluate PAH sources and receptor locations in addition to the IP/(IP+BgP) and Fla/(Fla+Pyr) isomer ratios. The goal of this study 13 was to identify sources that contribute to PAH deposition in urban, peri-urban, and rural areas in Rhode Island (RI), a small coastal state on the east coast of the United States and determine whether state recommendations for atmospheric PAH load reductions are required for wood burning stoves, which introduce a high concentration of PAHs to New England .
More specifically, the aim was to 1) quantify the seasonal atmospheric deposition of PAH rates at six different locations in RI using a contamination index, 2) trace and fingerprint the PAH sources using two methods: isomer ratios and PMF analysis, and 3) compare atmospheric deposition patterns in RI to similar latitudes in Europe where PAH bulk deposition rates were quantified with the same passive bulk deposition sampling method. By evaluating bulk PAH deposition rates and patterns in their spatial and temporal context, the basis is provided for determining the potential effects of PAH contamination in Rhode Island.

Site Descriptions
All study sites were located in Rhode Island (RI), which is a coastal state in New England that has 2710 km² of land mass and has just over one million residents with a population density of 388/ km². RI is the smallest state in the United States and more than one hundred times smaller than Germany. Like many Northeastern states, RI is characterized by densely developed urban and peri-urban areas concentrated along major transportation arteries, while large parts of the state are covered by forest interspersed with small rural settlements. The study sites were chosen to reflect this 14 rural-to-urban gradient (Figure 1.1). The urban site is located in Providence, the capital of RI, which has >100,000 inhabitants. The peri-urban sites are located in East Greenwich and Hope Valley. The rural sites are located over a kilometer away from main roads in a forested area in the north (Pascoag) and a coastal area in the south (Charlestown). The latter site is located approximately 350 meters from the Atlantic Ocean. The actual sampling spots were chosen to minimize the influence of trees or any other obstructions.

Atmospheric Deposition Sampling
Atmospheric deposition of PAH was measured using a passive sampling system that was tested and validated in previous studies (Figure S1.1) (Martin, 2000;Gocht et al., 2007a;Gocht et al., 2007b;. Sampling collection and processing is described in the supplemental information. All samples were analyzed for the 16 EPA priority PAH with a Shimadzu Gas Chromatograph/Mass Spectrometer QP2010 and following a modified EPA method (Method 610; Manoli and Samara, 1999 where V dep is the deposition velocity, which is approximated to be 0.5 cm/s as reported in the literature Esen et al., 2008;Castro-Jiménez et al., 2012). This value describes the dry deposition of particle 15 distributions that are dominated in size by particles less than 1 µm of mass median diameter, as is appropriate for urbanized and industrialized regions Zufall et al., 1998;Gigliotti et al. 2002;.

Contamination Index, Source Assignment, and Source Apportionment
To quantify the PAH deposition based on both, concentration and the number of compounds detected, a contaminant index for a specific sample i was defined as: where ∑PAH is the sum of PAH concentrations detected in a sample in ng/m 2 d, n is the number of compounds detected in the sample, and N is the number of analytes, gasoline combustion sources. It is not possible to discern between diesel and heating fuel sources using this method due to chemical similarities in combustion products of the two fossil fuels. In case of Fla/(Fla+Pyr), typical biomass combustion indices are greater than 0.5 and petroleum combustion ratios are less than 0.5 Ravindra et al., 2008).
To further identify sources of PAH contamination and link them to their spatial distribution in RI, positive matrix factorization (PMF) was carried out (further details on PMF in supplemental information). For that, the PAH deposition datasets were split up to represent a warm and a cold season: spring/summer and fall/winter. Four combustion source factors were defined (gasoline, diesel, oil, wood; is not possible (Larsen and Baker, 2003;). The PMF model was run 100 times to achieve a final solution for each data set. Afterward, the results were optimized by running the model with an fpeak value of -0.5. Results from the base model and fpeak model were compared to determine the robustness of the model.

RESULTS AND DISCUSSION
Variation in Spatial and Seasonal Deposition Patterns.
As expected, the annual sum of the 16 priority PAH (∑PAH 16 ) deposition throughout RI followed a rural-to-urban gradient. The sites in Charlestown (coastal) and Kingston (peri-urban) had the lowest annual deposition rates (∑PAH 16 73.64 ± 9.88 µg yr -1 m -2 and ∑PAH 16 92.0 ± 4.07 µg yr -1 m -2 , respectively), followed by the peri-urban and rural sites in East Greenwich and Pascoag (∑PAH 16 154.8 ± 28.8 µg yr -as a peri-urban site, because even though it is located in a more rural area of RI, it is in close proximity to I-95, has a 4.6 to 2.7 times higher annual PAH deposition rate (∑PAH 16 428.3 ± 106.2 µg yr -1 m -2 ) compared to Kingston and East Greenwich, the two peri-urban sites ( Figure 2). The sampling site with the highest annual PAH deposition rate is located in the center of urban Providence (∑PAH 16 2261.3 ± 1050.8 µg yr -1 m -2 ).
Over the survey duration and when compared by seasons, the highest PAH deposition rates in RI usually occur in the winter season followed by fall, while the lowest seasonal deposition rates occur in spring and summer (Figures 1.2&1.3). This pattern is consistent with other studies  and can be attributed to more heating fossil fuels and wood biomass burning during the heating season . In addition, shorter daylight hours as well as lower temperatures reduce the photo-oxidation and chemical destruction of PAHs, potentially leading to higher concentrations of PAHs relative to the summer months Motelay-Massei et al., 2003).
Even though no PAH specific air quality standard exists for Rhode Island, these data can be evaluated against regional air quality standards. were exceeded due to wood burning in the winter time (Lohmann et al. 2000 and.
Comparing the results from this study to studies conducted in rural southwestern Germany during 2001(Gocht et al., 2007a, the deposition rates in RI's peri-urban sites are similar in magnitude. Differences exist between the rural German sites, which has a deposition of ∑PAH 16 222 µg yr -1 m -2 and the RI coastal site, which has a slightly lower deposition rate of 73.6 ± 9.9 µg m -2 yr -1 . Also, the deposition rate at the RI urban site (2261 ± 1020 µg m -2 yr -1 ) is ~10 times higher than at the German rural sites. (Figure 1.3).   found that higher PAH deposition occurred when air masses originated in the Great Lakes area before passing over Pennsylvania, New Jersey, and New York before moving into the New England coastal area compared to air masses originating in the Great Lakes area, then passing over Canada, Maine, and the Atlantic and into New England.
An analysis of RI's predominant wind directions ( Figure S1.2) indicates that the air masses moving into the state are a mix of continental and oceanic air. As shown 19 by others

Source Identification using Isomer Ratios
The IP/(IP+BgP) isomer ratio was applied to distinguish between biomass, diesel, and gasoline combustion, while Fla/(Fla+Pyr) distinguishes between biomass and petroleum combustion Ravindra et al., 2008). Correlated against each other, the distribution of data points can be used to identify the predominant sources of PAH deposition at a specific site and season (Figure 1.4).
Spring/Summer. The Fla/(Fla+Pyr) isomer ratios at all sites indicate that petroleum combustion sources dominate during the warmer season. Based on the IP/(IP+BgP) ratios, the site data from Pascoag suggests a dominant gasoline combustion source (Figure 1.4). All other sites are dominated by diesel combustion rather than gasoline. Because these ratios were measured during the time of year when heating fuels are not widely used, it is likely that these combustion sources represent fuels used in transportation, i.e. diesel or marine fuels.

21
Fall. The isomeric signatures at all sites point to an increasing amount of biomass burning in the fall season (Figure 1.4). This shift coincides with the beginning of the heating season in Rhode Island when many households start burning wood and/or use other heating fuels.
Winter. The PAH signature derived from the two PAH isomer ratios is almost exclusively that of biomass burning, which is explicable by the number of homes in this region that are heated with wood stoves or fireplaces during the cold season . In New England, wood combustion is and has been a major source of heating, especially in rural areas, where up to 50 % of households use wood burning stoves . Thus, an increase in biomass combustion markers during the cold period points to PAH emissions resulting from wood combustion sources (Lohmann et al., 2000). The results from the two most rural areas, Pascoag and Charlestown confirm this trend, whereas the more urban sites have a slightly lesser biomass fingerprint and maintain a gasoline and diesel PAH signature (Figure 1.4). Only the peri-urban site in Hope Valley continues to carry a IP/(IP+BgP) diesel fuel signature, albeit with a greater Fla/(Fla+Pyr) ratio, which indicates additional contributions from biomass burning during winter. The continued prominence of diesel fuel combustion derived PAH is likely attributed to the site's proximity to the busy interstate highway I-95 (about 3 km to the west) ( Figure S4).
The problem of identifying unique sources of PAH emission showcases the limitations of PAH isomer ratios.

Positive Matrix Factorization
To further elucidate the contribution of specific combustion sources to the PAH profile, a PMF analysis was carried out using the spring/summer data sets and the combined winter/fall data sets. The datasets for each period and site were analyzed based on four factors: combustion of gasoline, diesel, oil, and wood (Table 1.1).
Again, oil denotes emissions from heating oil and marine and aviation fuel combustion, which carry a different combustion signature than diesel fuel emissions from automotive sources.
Spring/Summer. According to the PMF analysis the East Greenwich site is dominated by a high percentage of diesel emissions (62%) (Figure 1.5a). This finding correlates with the site's PAH isomer ratio results (Figure 1.4). The sites with the highest gasoline source percentages were Pascoag (73%) and Providence (66%) ( Figure 1.5a). In Pascoag, the PMF signature is in line with the PAH isomer fingerprints. In Providence, the PAH isomer source ratios indicate that the main PAH source was diesel, rather than gasoline combustion (Figure 1.4&1.5a).
The PMF gasoline and diesel source percentages at the Hope Valley site were 43% and 33%, respectively. The dominance of both of these automotive fuel combustion fingerprints correlates with the site's proximity to an interstate highway ( Figure S1.4).
The summer PMF profiles of the Charlestown and Kingston sites are both dominated by oil combustion (73% and 64%, respectively). In both cases, the PMF analysis does not agree fully with PAH isomer ratios, which indicated that diesel combustion was the main PAH source (Figure 1.4). The locality of the Kingston and   23 Charlestown sites, however, support the link to oil combustion sources. At the coastal site in Charlestown, the deposition of oil combustion products is likely related to marine fuel oil  and possibly originated from the nearby Point Judith harbor and the general boat traffic along the coast ( Figure S1.4).
The source of oil combustion at the Kingston site is not as obvious, but could be at least partially related to heavy fuel oil burning freight trains that run on the Amtrak line, which is located less than 1 km to the northwest of the sampling site (Melymuk et al., 2013) (Figure S1.4). Or, the oil signature could reflect the heating oil combustion emissions from the University of Rhode Island power plant that is located less than 1 km to the east of the sampling site.
These problems identifying the exact source of those oil combustion products illustrate the limitations of using PAH signatures in combination with a PMF analysis. This is because disseminating oil combustion signatures can be complicated by differences in the properties and composition of oils (density, aromaticity), the nature and effectiveness of the combustion processes, ranging from oil burning in home heating boilers compared to marine and aviation engines, and the combustion temperature together with the oxygen percentage in the burner's intake air   Figure S1.5). These residual percentages were not included in Figure 1 and Charlestown (36%) sites. It is noteworthy that the Kingston site also has a significant oil combustion signature (38%) (Figure 1.5b), likely from a nearby heating oil power plant operating at capacity in the winter time .
The influence of wood combustion at the East Greenwich and Pascoag sites was similar; 28% and 29%, respectively. During the winter, the East Greenwich site still had a high diesel combustion source PMF profile (31%). The relative high percentage of diesel related deposits at the sampling station cannot be readily explained without further investigations of major emission sources in this area, including diesel-like fuels from transportation sources, such as kerosene emissions from two nearby airports ( Figure 1.5b &S1.5).
During winter in rural Pascoag, a strong oil and wood combustion source is prevalent (28% and 29%, respectively) and can be attributed to residential heating with heating oil along with wood burning Kroetz and Friedland, 2008).

CONCLUSIONS
A passive atmospheric bulk deposition sampling system developed and tested in Europe (Martin, 2000;Gocht et al., 2007a;Gocht et al., 2007b) was successfully employed to measure the bulk atmospheric deposition of PAH in RI. The results show that the PAH deposition follows an urban-to-rural trend with higher deposition rates closer to urban area. These results confirmed deposition trends in other studies, several of which used the same passive sampling technique as was used in this study (Martin, 2000;Gocht et al., 2007a;Gocht et al., 2007b; and others, which used active sampling devices . Even though results obtained using different sampling techniques cannot be compared directly, the spatial and temporal trends observed can be used as a basis for comparison (Liu et al., 2007;. In addition, seasonal trends in PAH deposition patterns were identified, i.e. higher deposition rates in winter and fall contrasts with lower rates during spring/summer. These findings corroborated bulk PAH atmospheric deposition measurements in France Motelay-Massei et al., 2003;Motelay-Massei et al., 2007), and Germany (Gocht et al., 2007a;, and studies that took measurements of gaseous and particle phase PAH deposition in the United States  and China . Because the bulk atmospheric PAH deposition sampler used in this study was identical to that of samplers used in studies at German locales, the findings between the two studies were compared (Martin, 2000;Gocht et al., 2007a;Gocht et al., 2007b;. Relative to German sites, the PAH deposition in RI was comparable to the peri-urban sites; however, the coastal sites received up to 3 times less atmospheric PAH deposition, whereas the urban site received up to 50 times higher PAH deposition. A major difference between RI and Germany appears to be related to the prevailing wind directions that transport the PAH from their sources to the sampling station. RI is influenced to a greater extent by cleaner oceanic air masses during the summer, whereas more continental and thus more polluted air masses prevail in Germany (Figure S1.2&S1.3). This explanation is consistent with previous measurements of PAH deposition in the Northeastern United States.

Methods -Positive Matrix Factorization (PMF)
Also referred to as receptor modeling, PMF uses input data consisting of several analytes and their concentrations to produce two matrices: factor contributions and factor profiles (Paatero and Tapper, 1994). The user specifies the number of factor profiles that describe the sample and then identifies profiles based on known emission source information (     Figure S1.5. Seasonal combustion sources contributing to PAH deposition in the six sampling areas determined by PMF.
The four sources described are gasoline combustion, diesel combustion, oil combustion, and wood combustion, where fingerprints typical of wood combustion are detected only in the winter time, and other unidentifiable sources are detected in the summer time, described by the category "residuals" in the caption.

ABSTRACT
Stormwater runoff can contain high concentrations of pathogens that may impair water bodies used for drinking or recreational purposes. Currently, stormwater best management practices (BMP) are typically not designed to treat bacteria. In this study, we amended red cedar wood chips with different loadings of two antimicrobial compounds: 3-(trihydroxysilyl)propyldimethyloctadecyl ammonium chloride (TPA) and silver nanoparticles (AgNP) for use in stormwater treatment. After exposing Escherichia coli (E. coli) suspensions to the modified wood for three hours, the culturable bacteria in the aqueous phase and ones attached to the red cedar were determined. Results showed that E. coli inactivation, at 25°C, can be achieved with log 10 removal values (LRV) up to 3.71 ± 0.38 (mean ± standard error) for TPA-red cedar and 2.25 ± 1.00 for AgNP-red cedar, while unmodified red cedar only achieved 0.45 log 10 . Similar trends were obtained when a temperature of 17.5°C was used. At even lower temperature (10°C) there was no statistically significant difference in the level of bacteria inactivation between modified and unmodified cedar wood. The stability of the amended materials, in terms of efficiency, was also tested through storage under saturated conditions for 120 days, (equivalent to ~4.25 years of projected field use). We found that the level of bacteria inactivation for concentrations of 10 6 CFU/100ml using 6 mg/g TPA modified red cedar was reduced from 2.20 log 10 to 1.62 log 10 over the storage period, which was still elevated compared to unmodified red cedar at 0.42 log 10 . This study demonstrated that cedar wood chips amended with antimicrobial compounds could find application in structural BMP systems, such as tree filters. Other, chemically less complex compounds suggested for removing bacteria include iron oxide or copper Ren et al. 2009). Prior studies investigated various media, including wood chips and shale to enhance stormwater treatment of BMPs Li, Deletic and McCarthy 2014). However, these studies focused primarily on the amendment procedures and did not evaluate these modified materials for their antimicrobial properties.

Keywords
Particularly, the use of 3-trihydroxy silylpropyldimethyloctadecyl ammonium chloride solution (TPA) and silver nanoparticles (AgNP) antimicrobials in the reduction of microbial load in stormwater treatment applications has not been documented.
The scope of this research was to develop materials with antimicrobial properties than can effectively treat stormwater runoff contaminants with emphasis on the reduction of microbial load. Red cedar wood was chosen as the base material and AgNP and TPA were the antimicrobial agents. Red cedar has a high affinity for organic contaminants and heavy metals ) and is resistant to degradation (Grohs et al. 1999). Based on our prior research, both amendments show enhanced removal of organic and inorganic contaminants and are expected to provide antimicrobial properties ).
The antimicrobial performance of a material in a complex environment, e.g.
stormwater BMP, depends on the amount of antimicrobial matter in the filtration matrix, microbial loading, temperature, hydraulic loading (contact time), water chemistry and other factors. Here, we hypothesize that antimicrobial properties of TPA and AgNP can be transferred to cedar wood and that the resulting material can significantly enhance the microbial treatment of stormwater runoff under a range of 57 environmental conditions simulated under laboratory conditions. Using a batch experiment approach, the specific objectives of this study were to 1) perform kinetic studies and model the disinfection performance to achieve a desired bacteria inactivation given a specific disinfectant concentration and exposure time ; 2) determine the microbial inactivation efficiency of red cedar modified with varying loadings of AgNPs and TPA, 3) evaluate the bacteria removal efficiency of the three most effective materials at lower, environmentally relevant temperatures, and 4) evaluate the long term saturated storage impacts on the inactivation efficiency of the three most effective materials.

Untreated red cedar (RC) wood, obtained from Liberty Cedar and New
England Wood Products (both West Kingston, RI), was chipped, thoroughly mixed, and sieved. Wood chips passing through a 10 mm sieve, but retained on a 3.3 mm sieve were collected and used for the experiments. Wood chips were soaked in deionized water for a minimum of two weeks to leach out soluble matter and remove fines. Biosafe (Pittsburgh, PA) provided a 5% solution of TPA (EPA product Reg.No. 83019-2). The TPA molecule differs from other organosilane quaternary ammonium compounds previously studied (e.g. Torkelson et al. 2012) as the structure is a polymer consisting of a mixture of condensed 3-(trihydroxysilyl) propyldimethyloctadecyl ammonium chloride molecules ( Figure S2.1). AgNPs were synthesized in the laboratory via the Tollens method as described by , using polyvinylpyrrolidone (PVP, average molecular weight: 29,000 g/mol, Sigma Aldrich) as a stabilizer. To achieve various loadings of TPA and AgNPs, red cedar wood chips were amended following procedures described in Kasaraneni et al.
(2014) ( Table 2.1). Briefly, wood chips were submerged in TPA or AgNP solutions of 1.3 mmol/L ionic strength (NaCl) and were allowed to equilibrate for four to six days on a shaker to achieve a range of loadings that include the maximum possible loading of each antimicrobial ( Table 2.1) (more details on modification and desorption in  in Appendix I).
A nonpathogenic wild strain of E. coli (IDEXX, Westbrook, ME, USA) was used to test the antimicrobial performance of the filter matrix. E. coli was chosen because it is commonly used as an indicator bacteria for fecal contamination in drinking water and surface waters and monitoring it is typically required in stormwater regulations (Edberg et al. 2000;Wheeler et al. 2003;. In addition E. coli is a widely used reference pathogen when testing filter media for drinking water treatment .

Experimental Methodology
The first part of the study involved batch experiments to test the inactivation kinetics of TPA and AgNP modified red cedar (RC) at 25°C and establish whether exposure time or disinfectant loading were more important in inactivating bacteria.
This was followed by testing the disinfection performance of each material at 25°C and establishing bacteria inactivation efficiency at three different temperatures (25°C, 59 17.5 °C, and 10°C) for the most effective materials. In the last part of the study, the amended materials were submerged in water to test for possible changes in bacteria inactivation efficiency over 120 day period.
During all experiments the experimental solution ratio was 1 g of sorbent to 25 ml of solution containing the desired E. coli concentration. This ratio was based on a previous study that tested the performance of contaminant removal from an aqueous solution using the wood chips . All experiments were carried out in triplicates. Controls, buffer blanks, and NaCl blanks were analyzed for E. coli to establish microbial contamination, quality control, and natural decay of bacteria over the experimental period. All solutions, glassware and other materials used for the experiments were sterilized through autoclaving. All work surfaces were sterilized with a 70% isopropyl alcohol solution.

Bacteria Culturing and Enumeration
Bacteria were cultured in LB broth (10 g/L sodium chloride, 10 g/L tryptone, and 5 g/L yeast extract, Sigma Aldrich) at 37.5 °C for 13 hrs. After culturing, the bacteria were removed from the LB broth, washed and stored in phosphate buffer solution ( To enumerate bacteria attached onto the solid phase, samples were carefully sonicated (QSonica, Q125, Newtown, CT, USA) twice at 20% amplitude in 25 ml phosphate buffer solution for 10 minutes each and filtered following the above procedure. The sonication procedure was previously established by   , further tests were performed to determine the number of cycles required and the effect of sonication on E. coli (data not shown). The test results indicated that two cycles for 10 min is sufficient to remove attached E. coli from the wood chips and the sonication did not inactivate E. coli.

Inactivation Kinetics
The experiments were conducted with solutions containing E. coli at 10 (Tchobanoglous et al. 2003).
The parameterized model (details in SI) was tested against the experimental data and a percentage difference was calculated to determine whether the model overor underestimated inactivation kinetics.

Inactivation Efficiency of Amended Red Cedar
To test the bacteria inactivation efficiency as a function of antimicrobial amendment loading, batch experiments with E. coli solutions at five different concentrations ranging from 10 2 to 10 6 CFU/100ml were prepared in 1.3 mM NaCl solution. These solutions were exposed to unmodified, TPA modified (loading range: 3.6 mg/g to a maximum loading of 9.3 mg/g TPA) and AgNP modified (0.33 mg/g to a maximum loading of 0.68 mg/g AgNP) red cedar (from now on test materials are abbreviated as shown in Table 2.1) at 25 °C in a rotisserie incubator set to 5 rpm (Major Science, Saratoga, CA, USA). The exposure time was 180 minutes. This time was based on a preliminary kinetic study in which a ~50% reduction of E. coli was measured when exposed to unmodified red cedar (UM) at 25 °C (data not shown). The inactivation efficiency of the materials was calculated as log 10 removal value (LRV): Where C 0 is the initial E. coli concentration and C t is the final E. coli concentration, which is the sum of the E. coli that survived in both the aqueous phase and the E. coli attached to the unmodified or modified RC material. NB: Although all the materials were test in the same set of experiments, the inactivation efficiency data for unmodified red cedar, 6TPA and 0.6 Ag were published as part of a proof of concept paper ) and the data for these three materials was supplemented with four additional amended materials in this paper to compare the whole spectrum of materials ( To observe the differences in active E. coli present on the surface of unmodified and 6mg/g TPA modified red cedar, scanning electron microscope (SEM) images were taken. Each of the wood chips were put in contact with 10 6 CFU/100 ml for three hours in a rotary incubator at 25°C set to 5 rpm (Major Science, Saratoga, CA, USA) and prepared for SEM analysis (details in SI and Pathan et al. (2013)).
Control images on silicon wafers were prepared as well to document differences only due to the antimicrobial ( Figure S2.2)

Temperature Effect on Inactivation Performance
From the seven materials tested, the three best performing amended materials (6TPA, 9TPA, and 0.6AgNP; abbreviations explained in Table 2.1) along with unmodified red cedar (UM) were selected for testing possible effects of temperature on bacteria inactivation. The materials were exposed to an E. coli solution containing 10 6 CFU/100 ml in 1.3 mmol/L NaCl, at temperatures of 25°C, 17.5°C, or 10°C for 180 minutes in a temperature controlled rotisserie incubator set to 5 rpm (Major Science, Saratoga, CA, USA). Bacteria that survived the antimicrobial treatment were counted in the aqueous phase and the solid phase (E. coli culturable after attachment to the solid phase are measured as mentioned in Section "Bacteria Culturing and Enumeration") to determine removal through attachment and inactivation.

Long-term storage of modified materials and its impact on inactivation efficiency
Tests were conducted to determine how long-term storage impacts the inactivation performance of the TPA and AgNP modified materials. Three amended RC sorbents (0.6AgNP, 6TPA, and 9TPA) were stored in a dark at room temperature without mixing under saturated conditions in sterile, deionized water for 120 days at the same ratio as the batch experiments (1:25). The reason deionized water was chosen as the medium for storage was that it provides the greatest concentration gradient from the modified red cedar wood chip into the aqueous phase. Therefore, if any desorption should take place, the gradient is favorable. The 120 day length of storage is equivalent to 4.25 years of an installed tree filter (TF) BMP assuming that the filter matrix in such a BMP would be saturated for a maximum of 48 hours after a rain storm of ≥25 mm, which on average occurs 14 times per year in the Northeast United States, based on a 10-year data set obtained from a NOAA operated meteorological station in Providence, RI (NOAA 2014). The materials were stored under saturated conditions to test the maximum possible degradation over this time period because antimicrobials have been observed to degrade over time when stored under saturated conditions . It is assumed that unsaturated or dry storage will have lesser impact on inactivation efficiency over time as degradation occurs more slowly, however this has not been validated.

64
The inactivation efficiency was quantified immediately after modification (0 days), 60 days (equivalent to ~2.1 years), 94 days (equivalent to ~3.3 years), and 120 days (equivalent to ~ 4.25 years) after modification. At those times, subsamples of the stored material were exposed to 10 6 CFU/100 ml E. coli solutions (worst case scenario for stormwater loading ) in 1.3 mM NaCl. Exposure time was, once again, 180 minutes. The exposure temperature was set to 25°C in a temperature controlled rotary incubator that rotated at 5 rpm (Major Science, Saratoga, CA, USA). All modified materials were compared to unmodified red cedar. After 120 days, the storage solutions were tested for desorbed AgNPs or TPA. The storage solution containing AgNP modified red cedar was subsampled and digested using 2% nitric acid and filtered through a 0.45µm filter before analysis in a Perkin Elmer 3100 XL ICP-OES with a method detection limit of 0.01 mg/l for Ag. To test the TPA storage solution, a subsample was prepared following Hach Method 8337 and was analyzed on a Hach DR 2800, which had a method detection limit of 0.2 mg/L for TPA.
The disinfection efficiency at a specific time was modeled using Equation 2.3: where LRV 0 is the initial LRV that was achieved by the freshly amended material, LRV Final is the final LRV achieved after the material has been stored for 120 days, k is the disinfection efficiency rate loss constant, and t is the time of storage.

Statistical Analysis
To compare the overall inactivation efficiency of the different materials for all bacteria concentrations tested, a one-way ANOVA was performed. Post-hoc comparisons were carried out using Tukey's Honest Significant Difference. An ANCOVA was carried out to analyze the fraction of culturable attached bacteria and the fraction of culturable bacteria in the aqueous phase. The data were compared based on log 10 removal at different temperatures using log-normalized data. In this ANCOVA, the culturable attachment E. coli concentrations were the response variable and the explanatory variables were the exposure temperature and the type of material.
The model was simplified using a stepwise model simplification and AIC (Akaike's Information Criterion, Akaike 1998) was used as a criterion for the minimal adequate model. The difference in disinfection performance using total E. coli survival as the reference at the three different temperatures (10°C, 17.5°C, and 25°C) were compared with a two-way ANOVA, where log 10 removal values (LRV) was the response variable and exposure temperature and material type were the factors.
To compare the inactivation kinetics of the different materials, an ANCOVA was carried out where LRV was the response variable and exposure time was the explanatory variable for each different type of red cedar treatment. The comparison between the LRV of the amended materials before and after long term storage was tested using a two-tailed t-test for each material. All statistical analyses had the alpha value set to 0.05 and were carried out using R version 3.0.3.

Inactivation Kinetics
The kinetic studies data were evaluated with the Chick-Watson model (Haas and Karra 1984) (Tchobanoglous et al. 2003). The model can be used to describe the inactivation kinetics of the amended materials using first order reaction kinetics.
Verifying the predictions of the model with the experimental data shows that there is good agreement between the modeled and observed data, as the predicted LRVs for both TPA and AgNP modified RC were within 25% of the observed LRV (Table 2.2).
In the parameterized model, the coefficient of dilution, n, indicates that the loading of TPA is more important than exposure time (n = 1.99 and k' = -0.03); whereas for AgNP modified red cedar exposure time is key for its disinfection performance (n =0.58 and k' = - 1.18). The model predicts that with 6TPA a 99% reduction of total number of culturable organisms of E. coli can be achieved within 3.7 hrs, while it is 5.9 hrs with 0.6AgNP (Table 2 membrane. Therefore, reduction of total number of culturable organisms by TPA relies heavily on the TPA loading, i.e. the more TPA molecules, the better the chance of contact .

Inactivation Efficiency of Amended Red Cedar
Comparing all seven amended materials demonstrates that highly modified materials (0.6AgNP, 6TPA, 9TPA) are overall significantly more effective at 68 deactivating E. coli over a three hour time period compared to unmodified and low modified materials (0.3AgNP, 3TPA, 4TPA; Figure 2.1).
The LRV of unmodified RC was less than 1 and similar results were obtained using low loading on RC (3TPA, 4TPA and 0.3 AgNP;  Both, TPA and AgNP modified RC were successful at deactivating E. coli but in order to archive significantly higher LRV, a minimum loading of ~6 mg/g is required for TPA and 0.6 mg/g for AgNP. Based on these results, the materials with higher antimicrobial agent loading (6TPA, 9TPA, and 0.6AgNP) were chosen for testing the effect of lower exposure temperatures and storage time on inactivation efficiency.

Temperature Effect on Performance of Amended Red Cedar
The performance of the highly amended materials (6TPA, 9TPA, and  Figure   2.4b).
As the more highly modified materials have a greater concentration of antimicrobial agent on the surface of the red cedar, the likelihood of the bacteria coming in contact with an antimicrobial agent during attachment were greater than for unmodified and low modified materials. As the temperature increased, there are significantly fewer bacteria that survive once they attach to the modified red cedar surface (p<0.001, DF=2, F-value=49.92, ANOVA table in Table S2.2b). This trend was clearly detected when comparing results at 17.5°C and 25°C (Figure 2.4b), whereas no difference in bacteria survival after attachment to the three amended materials was detected at 10°C. However, at 10°C survival after attachment for unmodified red cedar was two orders of magnitude higher compared to modified materials ( Figure 2.4b). These results are consistent with previously reported studies, where lower temperatures, such as those used in this study, allowed for higher bacteria survival (Gayán et al. 2011;Raffellini et al. 2011). It is proposed that at lower temperatures the movement of the bacteria in solution is reduced and bacteria generally attach to surfaces in fewer numbers , resulting in fewer E.
coli interacting with the antimicrobial agent. The reduced bacteria motility and attachment, along with a change in their cell membrane structure, making them more rigid at low temperature, (Shivaji and Prakash 2010) contribute to an overall reduced E. coli inactivation performance at low temperatures.

Long-term reduction of the disinfection efficiency of modified materials
The three most effective sorbent materials (0.6AgNP, 6TPA, and 9TPA) were tested to quantify if and how saturated bench storage impacts the inactivation performance of the materials. These materials showed changes in efficiency over time that resulted in a 26% to 38% decrease in LRV (Table 2 However, even though no total silver or silver ions were present in the storage solution when analyzed on the ICP-OES, the concentration of total silver or silver ions may be elevated at the wood chip-water interface, but may not be detected in the bulk sample (Unwin and Bard 1992;Pfeiffer et al. 2014). This phenomenon may not limit the antimicrobial effectiveness of AgNPs, as the concentration of Ag + ions can still be elevated at the interface of the nanoparticle (Pfeiffer et al. 2014).
To test whether eroded red cedar fines might have settled on the wood surface after storage and thus limited access to the antimicrobial coating, all materials were agitated by ultrasound for 10 minutes. Subsequently, the inactivation efficiency of all material was tested again and compared to the non-sonicated materials' performances.
The LRV for the sonicated and non-sonicated materials was the same, therefore the covering of actives sites by wood-derived fines could be ruled out as a reason for decreased inactivation efficiency after storage.
The measured decrease in inactivation performance over time could potentially be due to chemical degradation of the antimicrobial material. However, storing aqueous TPA solutions under the same conditions as the amended wood did not decrease the agent's concentration or its inactivation performance. Thus, degradation of the TPA antimicrobial is also not responsible for a decrease in inactivation performance. Similarly, PVP stabilized AgNPs in aqueous solutions have been shown to be stable as well . Rather than degrading, the antimicrobial amendments may diffuse from the wood's surface into its interior over time. This would make it more difficult for the bacteria to come in contact with the antimicrobial agent. Existing analytical models for diffusion into wood cannot be applied to describe the diffusion processes of these compounds (Burr and Stamm 1947;Behr et al. 1953) because the TPA and AgNP sorption processes to amend the red cedar do not behave in a linear fashion . It would therefore be necessary to develop a complex numerical model as suggested by Rügner et al. (1999)

Limitations
The results in this study indicate that the inactivation performance of amended materials is significantly higher than unmodified materials. However, further testing of these materials under dynamic flow conditions is required to confirm these results.
Stormwater is a complex mixture, which can include organic contaminants, heavy metals, inorganic salts, humic substances, along with several species of microorganisms. A study determining the impact of the complex stormwater chemistry conditions and different species of microorganism on the inactivation performance of the amended materials is necessary prior to use of these materials in the field.

Implications and Benefits to Using Antimicrobial Filter Materials
Red cedar wood chips amended with either TPA or AgNP have been shown to increase the removal of E. coli from water. Even though similar results could be achieved with chlorine or UV disinfection, both of which occur within minutes (Gayán et al. 2011;Tchobanoglous et al. 2003), these two approaches have not found application in stormwater treatment because of complexity and cost of such treatment (Tchobanoglous et al. 2003). Relative to these technologies, the passive wood filtration approach described herein does not produce disinfection byproducts, as often observed when using chlorine. As shown by the saturated storage experiments, the antimicrobials also do not leach from the red cedar into the environment, suggesting that the proposed filter media might be safe for use in the environment. Further, using a passive approach of disinfection requires less energy; in particular compared to UV disinfection, which also involves frequent maintenance to ensure proper disinfection (Gayán et al. 2011;Tchobanoglous et al. 2003).
Based on these properties, the red cedar wood chips amended with either TPA or AgNP could find use in stormwater treatment systems. An example is a Tree Filter BMP, which treats conventional stormwater pollutants, such as heavy metals or petroleum hydrocarbons, but could be outfitted with a layer of amended wood chips to provide additional antimicrobial treatment functions. Such a BMP system could potentially remove a much higher percentage of E. coli and possibly other bacteria from stormwater runoff than most currently available treatment systems such as  (Grohs 1999). However, a further study of how long the wood chips can remain in the BMP is required.
Even though both antimicrobials are strongly sorbed to the red cedar, there are benefits to using TPA over AgNP. This is because AgNP amended materials rely on the slow, but continuous release of Ag + ions (Kim et al. 2007) and inactivation is timedependent, as was demonstrated herein. Therefore the antimicrobial treatment of AgNP amended wood is slower and the efficiency is expected to decrease eventually due to a depletion of Ag + ions. In comparison to AgNP and other antimicrobials, such as copper (Ren et al. 2009) or other nanoparticles, TPA does not release ions from its matrix into solution. Hence, TPA should retain its antimicrobial efficiency for longer 77 than AgNP (Shi et al. 2007;Saif et al. 2008;Ferreira and Zumbuehl 2009;). In addition, a cost benefit analysis of TPA versus AgNP modified materials  shows that TPA is more cost effective than AgNP.
If the proposed amended red cedar were to be implemented as a filtration matrix in structural best management practices, the 6TPA modified red cedar would be favorable compared to the currently used matrix that consists of a shale-sand mix, or even unmodified red cedar. This is because not only does 1 gram of this material (i.e. 6 mg of TPA) achieve >2 log 10 removal after three hours of exposure at room temperature, but it also is effective at removing bacteria at decreased temperatures that could be found in New England (17.5°C). These lower temperatures are closer to a natural BMP operating temperature in most temperate climates. To decrease the effective treatment time to less than 30 minutes, which many BMPs are designed for (RIDEM 2010), an optimal wood:solution ratio would have to be determined before implementing the material in the field. However, introducing any type of bioactive material into the environment will require additional field studies to ensure that no unanticipated consequences result from the well-intended use of antimicrobial wood for stormwater treatment.  (3), 691-699.  Figure 2.1. Comparison of (average ± standard error) LRV to increased exposure times of the six modified materials and the unmodified red cedar. A steeper slope indicates a higher inactivation rate, thus a more effective antimicrobial material. Comparison of log 10 removal values (LRV) (average ± standard error) for the unmodified red cedar wood chips and the six modifications at different E. coli concentrations ranging from 10 2 to 10 6 CFU/100ml. All materials were exposed for 180 minutes. Natural decay data shows the fraction of E. coli die-off that is not due to antimicrobials or removal by attachment. In order to compare the performance of all the materials data for unmodified red cedar, 6TPA, and 0.6 AgNP was obtained from our previously published work ).  A B Figure 2.5. Average log 10 removal value ± standard error for the three most effective materials (6TPA, 9TPA, and 0.6Ag) compared to unmodified red cedar at 25°C. The LRV was determined immediately after modification and at 60, 90, and 120 days after being stored in water. An initial drop in the antimicrobial performance over the first 60 days was followed by a much slower decrease in inactivation performance afterwards.  where y is the y-intercept and N t /N o is the desired ratio of E. coli concentration at time t relative to the initial E. coli concentration, in this example 99%.

Methods -Sample Preparation for SEM analysis
Briefly, the samples were fixed in 2.5% glutaraldehyde solution for 30 minutes. After fixation, the samples were washed three times with phosphate buffer and then dehydrated for 20 minutes using sequentially increasing ethanol solutions (increments of 10%) containing 40% to 100% ethanol. The samples were exposed to 100% ethanol three times before being chemically dried using Hexamethydisilazane  Tables   Table S2.1a -Performance -ANOVA Table. ANOVA table for one-way comparison of log 10 removal values for the six modified red cedar wood chips and unmodified red cedar wood chips using LRV as the response variable and material as the factor.  In A) the E. coli at 104 CFU/100 ml were not exposed to TPA and depict live organisms, whereas in B) E. coli at 106 CFU/100ml were exposed to TPA for three hours before SEM imaging, thus depicting damaged organisms.

INTRODUCTION
Several contaminant groups, e.g. heavy metals, organic contaminants, such as polycyclic aromatic hydrocarbons (PAH), pathogens, nutrients, and pesticides accumulate on impervious surfaces during dry periods  2000; .
These environmental compartments act as sinks and can accumulate some of the contaminants found in runoff, such as heavy metals and organic contaminants (Benfenati et al. 1992;Zehetner et al. 2009).
Other contaminants, such as deicing salt or bacteria do not accumulate, but instead are flushed out and have the potential to contaminate surface water bodies after rain events Zehetner et al. 2009). The presence of these pollutants impacts groundwater and surface water resources  and is threatening sensitive species that inhabit these ecosystems . Due to its small size, a TF generally takes up less than 1% of its maximum treatment area of up to 0.75 acre (3035 m 2 ) of paved surface. TF are meant to be installed in urban areas along city streets or within parking lots ). TFs contribute to the "greening" of a city and can easily be implemented in landscape architecture designs.
In this study, the contaminant removal efficiency was tested for a TF unit to determine whether the performance complies with the requirements specified in the Rhode Island Stormwater Manual (RISM) requires 85% removal of TSS, 30% removal of nitrate and phosphate, and 60% removal of bacteria. In addition to the contaminants listed in the RISM (nutrients, TSS, bacteria), the field parameters (temperature, pH, and electrical conductivity), chloride (Cl -), polycyclic aromatic hydrocarbons (PAH), and heavy metals were analyzed. The objective of this study was to establish whether tree filters meet water quality treatment standards set by RISM under conditions that are typical for Rhode Island.

Site Description
In August 2013 a tree box filter was installed at the Rhode Island Stormwater Treatment Demonstration Facility (Figure 3.1). The RISTDF is located on the main campus of the University of Rhode Island. The TF was integrated into a conventional asphalt cover parking lot north of the university that has 458 parking spaces covering 4.9 acres (19830 m 2 ; Figure 3.1a). The drainage area contributing to the TF is 0.06 acres (242.8 m 2 ). The site is located over the Chipuxet Aquifer, which is a major source of drinking water in southern Rhode Island. The aquifer is overlain by a thin veneer of loess. Below follow a series of complexly interbedded lenses of sand and gravel with smaller contributions of silt and silty sands (Dickerman 1984). These glacio-fluvial deposits were deposited during the Pleistocene epoch (Dickerman 1984). The stratified material is ~60 m thick and overlies fractured granitic bedrock.
The depth to groundwater at the site is approximately 7 m below ground surface (bgs) .  logger was placed adjacent to the buried sampling pan. The data loggers were accessed from the surface via a data transfer cable buried adjacent to the Teflon tubing. The data loggers were set to take measurements every 10 minutes. Data from the loggers was downloaded using Decagon Devices' ECH2O Utility Software.

Meteorological data was obtained from a National Climatic Data
During sampling events, a portable peristaltic pump (Environmental Sampler, Masterflex, Vernon Hills, IL) was connected to the Teflon tubing and water collected in the buried sampling pan was pumped to the surface (Figure 3.2b).
After the TF concrete structure was placed into the pit, the remaining space was filled with approximately 4 m 3 engineered media composed of 75% coarse sand, 15% expanded shale (Φ=-4; 16 mm), and 10% sphagnum peat to achieve a treatment depth of 0.9m (Figure 3.2c). According to the TF manufacturer, the infiltration rate of the filtration matrix is 0.21 m/hr with the first effluent reaching the sampling port in 30 minutes when the system is saturated.

110
A red maple tree (Acer rubrum) was planted inside the TF (Figure 3.2d).
Although there are many other choices of trees, Red Maple was chosen to match surrounding trees in the parking lot. A grate was cut to accommodate the maple tree and placed on top of the concrete structure. Surface runoff can the TF through a curbside opening of 66 cm by 15 cm. Any asphalt removed during the TF installation was patched with commercial black top mix.
The drainage area contributing to the tree box filter is approximately 243 m 2 .
The drainage area of the TF unit lies close to the entrance of the parking lot and is heavily trafficked every day as the parking facility is a student commuter lot ( Figure   3.1b).

Sampling and Analysis
To determine whether the TF meets the standards of RISM, the unit installed at the RISTDF was monitored weekly for eight months starting in November 2013.
Runoff that accumulated in the catch basin during prior rain events (influent), grab samples during seven rain events and effluent from the pan were sampled and analyzed for specific conductance, temperature, chloride (Cl -; range 0-1000mg/L), standards were also used to correct for PAH extraction recovery. A method detection limit (MDL) study was conducted and PAHs with molecular weight of 228 g/mol (including Chrysene) the MDL was 5 pg/µl and 10 pg/µl for compounds with molecular weight 252 g/mol and above (Benzo(b)fluoranthene and above).

RESULTS AND DISCUSSION
Stormwater runoff concentrations found in this study are comparable to those found in other studies (Table 3.1). Generally, the pH of the stormwater influent (pH range 6.4 to 11.3) was elevated compared to the effluent (pH range from 5.9 to 7.9).
Also, influent pH tended to be higher during winter than summer (Table 3.1). The higher pH of the influent is likely linked to the concrete and the time the influent water is in contact with it. That is, while water remains stagnant in the TF catch basin between storms, the partial dissolution of minerals in the concrete raises the pH of the solution. This effect is most pronounced in new concrete structures, but is expected to diminish over time (Kakade 2014).
Average annual specific conductance in the influent was 1612±3467 µS/cm. It

116
However, when samples were taken from the catch basin, the removal was lower than 20%, therefore highlighting the importance of sampling the first flush of a rain event.
In general phosphate removal is associated with the attachment to particles and is therefore well correlated with removal of total suspended solids (Berretta and Sansalone 2011). However, the phosphate content of the TSS particles were not investigated in this study. With regard to dissolved phosphate, however, several field studies have shown that the makeup of the biofiltration matrix can have an impact on the removal efficiency because higher organic matter content generally results in higher phosphate removal ). The TF filter matrix used in this study contained 10% sphagnum peat, which represents an organic matter content higher than in most top natural soils but less than organic soils (Troeh and Thompson 2005).  linked phosphate removal (up to 80%) in biofiltration systems to sorption to clay and silt particles. Particularly clays are not common in the formerly glaciated New England area and their general absence might further limit the performance of the TF investigated herein. But adding clay particles to the TF filter matrix is not desirable because it would likely lower the infiltration rate of this BMP.
In this study, the dissolved phosphate removal correlates well with the total suspended solids removal (R 2 =0.73) during the abovementioned storm events ( Figure   S3.1).

117
Total suspended solids were grab-sampled during seven storm events; two winter and five summer storms. The influent TSS concentration in the grab samples ranged from 5 mg/L to 46 mg/L whereas the effluent ranged from 2.2 mg/L to 18.2 mg/L (Table 3.2). Overall, the concentrations of TSS were higher in the grab samples compared to those in the catch basin. The catch basin is designed to collect the influent stormwater and allow the associated particles to settle out, thus the TSS concentration likely decreases as after a storm event. coli removal in wetland and bioretention systems that underwent changes in soil moisture content due to intermittent stormwater influent was most effective . Kibbey et al. (1978) found significantly higher bacteria die-off rates in soils at field capacity compared to saturated soils and attributes this finding to a low tolerance for dryness. It is suspected that high soil moisture content allows 119 pathogens to survive in the subsurface for extended periods of time and also get remobilized, possibly being exported during runoff events ).
In the TF studied here, the system's high removal percentage can be attributed to the variable moisture content, allowing for less favorable bacteria survival conditions and increased straining in the subsurface ).

Heavy Metals
The dissolved influent concentrations of heavy metals in the catch basin in winter were as follows: 23.3±39.4 µg/L for copper, 4.9±18.2 µg/L for nickel, 19.5±39.9 µg/L for lead, and 312±710 µg/L for zinc (Table 3.1, Figure 3.6). In summer the concentrations were lower overall. Copper and nickel were not detected, lead was on average 5.6±6.2 µg/L and zinc was 4.4±8.6 µg/L. The metal removal percentages are shown in Table 3.2. In general, much of the results are inconclusive because of at or below detection limit concentrations (i.e. nickel) or sample contamination issues related to the presence of brass fittings attached to the stainless steel sampling pan. Brass has been shown to leach copper, zinc and lead  and for that reason likely caused higher lead, copper, and zinc effluent concentration in both, summer and winter. The leaching assumption is further corroborated by the stormwater runoff pH, which decreased from an average influent range as high as 8.5 to as low as 5.9 in the effluent. At lower pH, a higher portion of zinc and copper is soluble and mobilized, resulting in higher effluent concentrations ).

Polycyclic Aromatic Hydrocarbons
During winter the average PAH influent (catch basin) concentration was 4.1±6.8 µg/L with a range of compounds below the detection limit ( Petroleum hydrocarbons, such as PAH, have been removed successfully in field studies of a tree filter system, with ~65% removal in summer and 99% removal in winter ). Laboratory column studies show that PAH removal likely occurs through sorption to organic matter and that this process is 80-95% successful . In this study removal of PAH was 47.1% in the winter compared to 7% in the summer. While the overall removal percentages are lower, the removal trend is similar to that of the UNHSC test site, where summer removal was lower than in winter. The reason for the overall low removal percentages is unclear .
Since no grab samples were collected for PAHs and the influent concentration was based on catch basin water only, it is very likely that PAH in the dissolved phase have been scavenged and settled out with fine particles by the time the catch basin was sampled. Therefore, the dissolved PAH concentration in the catch basin may be lower in the analyzed samples and not representative of the actual stormwater PAH concentration.

Recommendations and Suggestions for Improvements of the Tree Filter Technology
Based on the constituents analyzed, the tree filter unit performs well with respect to the removal of nitrate (EPA 2014). The TF also performs well in E. coli removal and exceeds RISM bacteria standards throughout the year. Metal removal in this filter unit was not quantifiable or inconclusive for most of the heavy metals.
To improve the nitrate removal from stormwater runoff, the TF system would likely improve by providing a longer residence time for stormwater runoff and/or the creation of an anoxic zone where denitrification could occur. A longer residence time could be achieved through the addition of finer grained material, such as clay or fine organic material. An addition of clay or organic material could also result in a higher percentage of phosphate and PAH removal through attachment/sorption. When designing such a system however, it is important to maintain a high infiltration rate so that the incoming stormwater runoff does not overwhelm the system and bypass it due to ponding and slow infiltration. Determining an optimal mixture of the TF engineered media and structural design demands further research.
To create an anoxic zone that has the potential to improve nitrate removal, continuous saturation of the TF unit is recommended. A structural design change 122 allowing the catch basin to slowly drain the stagnant water into the TF unit would maintain higher soil saturation between storm events and prevent the microbial community responsible for denitrification from decomposing. To do this, the catch basin could have a low permeability membrane on the side facing the infiltration box that allows the water to slowly drain into the engineered media.
The slow drainage of the catch basin could have multiple advantages, ranging from improved TSS removal to fewer mosquito breeding grounds in the spring. An empty catch basin can better capture the first flush of the stormwater influent containing the highest load of TSS, therefore allowing the particles to settle out instead of the influent spilling directly into the infiltration area of the TF. As the catch basin fills up, the particles can settle out and the TSS removal is potentially improved.
Dorfman, M., and Rosselot, K. (2012).  (1), 125-132.  *Grab samples were not analyzed for PAH and heavy metals; b.d. below method detection limit References: a= , b= , c= , d= , e= , f= , g= , h= ), i= , j= , k= , l= , m= (Chong et al. 2013), n= ), o= , p= , q= , r= ), s=    *Grab samples were not analyzed for PAH and heavy metals; ^ Potential metal release from brass fitting on sampling pan therefore not quantifiable; +Nickel release reported by authors in study (z); b.d. concentration is below detection on the instrument. References: h= ), i= , j= , k= , l= , m= (Chong et al. 2013), n= ), o= , p= , q= , r= ), s= , t= ), u= (UNHSC et al. 2007, v= , w= , x= ), y= , z= , aa=        have not yet been introduced on a federal level, however increasingly more states require at least primary treatment of stormwater runoff to prevent water quality degradation of surface waters. In this study we compare the results of an eight month field monitoring of two tree filter (TF) systems. A TF is a type of a structural best management practices (BMP) for road runoff treatment. One conventional TF, containing a sand/shale mix, was compared with an innovative unit that contained red cedar wood chips amended with the antimicrobial 3-(trihydroxysilyl)propyldimethyloctadecyl ammonium chloride (TPA

INTRODUCTION
Stormwater runoff mobilizes microorganisms and other contaminants into surface and ground water bodies, potentially impairing their quality ) and compromising ecosystem and human health EPA 2012 There may be several BMP technologies that can treat TSS very effectively, for example, stormwater bioretention systems ).
But only a few commercial systems advertise effective bacterial removal capacities.
Examples of such systems include BactoLoxx (Filtrexx, Goffstown, NH), which relies on proprietary flocculation agents that result in settling out bacteria ), and Bacterra (Filterra Bioretention Designs, Ashland, VA), which relies on physical filtration and predation in a biomat layer . While the manufacturer supplied bacteria removal efficiency of these two stormwater BMPs reaches up to 99%, these treatments are not geared towards the simultaneous removal of other contaminants. For that purpose, supplementary filter media must be added ). Other BMPs, such as biofiltration systems, that simultaneously target more than one class of contaminants, e.g. heavy metals and PAHs, have limited bacteria removals capacities, ranging from 0-80%.
From a public health perspective, this is not considered sufficient EPA 2001;).
Besides overall low removal efficiencies, recent studies of 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 and not deactivation . This is because sorbed pathogens can remain viable during attachment ) and therefore can be remobilized or detached during intermittent flow conditions . Bacteria remobilization or detachment is especially relevant in areas with high water tables, because insufficient soil depth may prevent effective pathogen treatment, therefore increasing the risk of bacterial contamination of groundwater resources . New approaches are therefore needed to enhance the inactivation of pathogenic organisms in stormwater runoff, possibly together with other co-contaminants, and implement these technologies in the next generation of BMPs.
Herein, we investigated if innovative stormwater treatment approaches can be based on antimicrobial compounds that have found wide-spread application in many consumer products or water filtration materials. Several studies have shown that These findings suggest that it might be possible to amend antimicrobial properties to 145 the matrix of existing BMP systems, such as tree filters (Schifman et al. in prep a.; Schifman et al. in prep. b), thereby achieving simultaneous treatment of chemical and bacteriological contaminants.
The objective of this pilot-scale study was to compare a conventional sand/expanded shale based tree filter BMP to an innovative one amended with TPA/Red Cedar wood chips. The primary focus was on quantifying the removal of bacteria from stormwater runoff. It was hypothesized that the antimicrobial media, which has been extensively tested in the laboratory for its ability to simultaneously remove heavy metals, PAHs, and E. coli, will show an increased pollutant and bacteria removal capability in the field compared to the conventional filter Schifman et al. in prep. a;Schifman et al. in prep. b). The results from this study are expected to shed light on the potential of this simultaneous treatment approach for possibly implementation in full scale stormwater runoff BMP systems in the future.

Site Description
A conventional asphalt covered parking lot on the northern end of the University of Rhode Island, Kingston campus holds 458 parking spaces and is heavily used by commuting students. This parking lot covers 4.9 acres (19830 m 2 ; Figure   4.1a) and, together with adjacent roadways, it is part of the Rhode Island Stormwater Demonstration Facility (RISTDF). The test area is located over the Chipuxet Aquifer in southern Rhode Island. Beneath is a thin layer of loess, a Pleistocene glacio-fluvial 146 aquifer contains complexly interbedded lenses of sand and gravel with smaller contributions of deposits of silt and silty sands (Dickerman 1984). The stratified material is about 60 m thick and overlies fractured granitic bedrock. The depth to groundwater at the site is approximately 7 m below ground surface (bgs) ).

Installation
In August

Red Cedar TF Preparation
A quaternary ammonium compound, 3-(trihydroxysilyl) propyl-dimethyloctadecyl ammonium chloride (TPA), was the antimicrobial agent used in this study.
In prior laboratory studies, TPA was identified as a stable amendment to red cedar wood chips Schifman et al. in prep. a). Although TPA loadings of up to 9 mg per gram wood were found possible , for cost reasons the loading selected for this pilot experiment was 6 mg/g. The red cedar chips were prepared for TPA loading by first washing the wood to remove fines and leaching out soluble matter. For that, the wood was soaked in frequently exchanged clean water for a minimum of 2 weeks. The wood chips were then mixed with an aqueous solution containing 75 mg/L TPA ( Figure S4.1). The ratio of the red cedar mass to TPA solution was 1g to 25 ml, as based on previous experiments. The mix was stirred for 9 days to reach the 6 mg/g TPA load as suggested by . Afterwards, the wood chips were dried, weighed, and stored in a dark and dry location before being added to the tree filter. A total of 30 kg of TPA amended Red Cedar was mixed into the top 0.5 m of sand/shale mix already inside the tree filter.

Sampling and Analysis
From  standards were also used to correct for PAH extraction recovery. The method detection limits was 5 pg/µl for PAHs with molecular weight of 228 g/mol (including Chrysene) and 10 pg/µl for compounds with molecular weight 252 g/mol and above (Benzo(b)fluoranthene and above).
A two tailed t-test was carried out using R software version 3.0.3 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.

Controlled Field Tracer Experiments
Field tracer experiments were carried out on each TF using a solution containing NaCl as a conservative tracer and a green fluorescent protein ( The two tracer tests were carried out on separate days one week apart in September 2014. No major rain events were observed immediately prior to or during the tracer test period. In both instances clean, non-chlorinated water was pumped into the TF at about 3 L/min (circa 200 L) to create similar initial conditions in both TF filters. The water was pumped into the TF directly, bypassing the catch basin using an electric submersible pump and garden hose (Figure 4.2). The water flow was metered with a 2.5 m 3 /hr ISTEC Flowmeter (Sparta, NJ, USA). After approximately 70 minutes of pumping clean water, the tracer solution containing NaCl and GFP E. coli was injected. The NaCl concentration was 888 mg/L (428 liters) and 969 mg/L (392 liters) for the conventional and the innovative TF experiment, respectively (Table 4.2).
The initial concentration (C o ) of GFP E. coli in the solution was 6.90*10 5 CFU/100 ml and 8.32*10 6 CFU/100 ml for the conventional and the innovative tree filter, respectively (Table 4.2). The tracer solutions were pumped over the course of 120 minutes and 145 minutes, respectively (Table 4.2). Afterwards, an additional 207 L of clean, non-chlorinated water was pumped through each of the two TF for approximately 70 minutes.
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. Every 5 minutes, the effluent was metered for specific conductance using a YSI Professional Plus Multimeter logger (Yellow Springs, OH, USA) ( Figure S2). Every 15 minutes samples were collected for bacteria analysis and immediately processed at the field laboratory at the RISTDF site.

Catch Basin Water Quality
Overall, both tree filters had similar influent contaminant concentrations during storm events and catch basin samples were similar to grab samples, except TSS, where concentrations in the grab samples were elevated compared to the catch basin (Table 4.3). As a result, all further discussion on TSS is based on grab samples as the influent. For the remaining constituents the catch basin water samples are used as a base comparison as these concentrations did not differ from the grab samples.
All constituents that were analyzed occurred at higher concentrations during winter months except for PAHs, for which concentrations were higher during the summer (Table 4 Nitrate, similar to chloride, is very mobile. Studies suggest nitrate concentrations in forests spike at the end of the winter period . While there are no forests at the site, there are leaves and other tree residues accumulating on the parking lot surface. These organic materials are incorporated into snow piles during winter. The nitrates associated with these organic residues appear to be exported during snowmelt, therefore contributing to higher loads during the late winter sampling period (Figure 4.3b) . Once the snow cover has melted, the nitrate loading from stormwater runoff decreased, which correlates with the chloride concentrations (Table 4.3, Figure 4.3).
Total suspended solids (TSS) display similar trends in winter, where higher loads of particles are transported through runoff during melting or rain events ( Figure   4.3c); this is apparent in both, catch basin and grab samples (Table 4.3). The grab samples did exhibit higher concentrations of TSS, likely because most particles are removed in the catch basin after the stormwater enters the TF unit.
In part, the high concentrations of TSS in the winter time can be attributed to the parking lot being treated with a sand/salt mix (Pers. comm. Joseph Paradise, URI).
Even though the concentration of TSS decreases in summer, elevated loads of TSS can still be observed during rain events following longer antecedent dry periods.

Water Quality and chloride
Both tree filters were evaluated for their ability to remove contaminants from stormwater. In both TFs, the pH in the effluent decreased relative to the catch basin in summer and winter. The chloride effluent concentrations in winter were lower by 51.1% in the conventional TF compared to the influent. In the innovative TF, chloride effluent concentrations decreased by 23.5% during the summer (Table 4.4). It is unclear why or how chloride was apparently removed because this ion is not expected to sorb or otherwise interact with the TF filter matrix.

Total Suspended Solids
TSS analysis is based on seven grab samples taken during rain events because the catch basin is not a representative sample for this analysis (Table 4.3). During these events the TSS removal was higher in summer for both filters (84.9% and 90.3% for the conventional and innovative filter, respectively) compared to winter (75.4% and 80.2%, for the conventional and innovative filter, respectively). Neither filters met the RISM standard of 85% TSS removal. However more frequent grab sampling during storm events is needed to validate these results. Comparing these two systems statistically shows that there is no significant difference in TSS removal in winter (p=0.3), while there is a significant difference in TSS removal in summer (p<0.05).

Nitrate
Nitrate removal was higher in winter, where the conventional TF achieved removal of 38.6% and the red cedar TF removed 62.6% (Table 4. 4). During winter, the catch basin nitrate concentrations were also significantly higher compared to summer, during which the EPA MCL for nitrate (10 mg/L) was rarely exceeded (Table 4 Although in this study only nitrate was measured, other organic and inorganic nitrogen species are often found in stormwater runoff . When these other nitrogen species are present in stormwater runoff and accumulate in BMPs, they can be converted to nitrate through biologically mediated nitrification during dry periods. This is because dry periods reduce the water content in the tree filter and conditions become more aerobic, favoring nitrification . When this occurs nitrate can be released from the system during the next storm event.  tested seven materials i.e., alfalfa, leaf mulch compost, newspaper, sawdust, wheat straw, wood chips, and elemental sulfur, for their ability to remove nitrate in laboratory column studies. It was found that the materials that were effective electron donors, such as wood chips, newspaper, and sawdust outperformed 157 washed filter sand. Nitrate reduction was observed during steady state flow conditions as well as intermittent flow conditions . On this background, during more frequent rain or melt events, as is observed at this test site, the filter may turn anaerobic, creating conditions that are favorable for denitrification .
When this occurs wood chips in the innovative filter can act as effective electron donors and denitrification can occur, resulting in more effective nitrate removal in the innovative tree filter system. Because the soil oxygen content was not measured in this study, a definite statement cannot be provided if conditions inside the TF were aerobic or anaerobic.

Phosphate
Phosphate removal, though only studied in the summer months, was more effective in the innovative filter (46.5% removal, p=0.08) compared to the conventional sand/shale TF unit (31.5% removal) ( Table 4.4). Similar results have been reported in other studies where organic based filter media, such as biochar, peat, or mulch outperformed a sand mixture in the removal of phosphates . In general, phosphate removal in other BMP studies indicated that not just the filter media but also the initial phosphate content of the filter media may play a role in the removal efficiency of phosphate ). 158

Heavy Metals
For heavy metals, the contaminant removal was not quantifiable for copper and zinc because during both sampling seasons these two metals apparently leached from the brass fitting connected to the stainless steel sampling pan .
Nickel aqueous phase concentrations were at or near the method detection limit (10 µg/L) during the entire sampling campaign, particularly during summer. During winter, nickel removal was 9.7% removal in the conventional TF compared to 97.1 % in the red cedar tree filter (Table 4.4). Because of at or near the method detection limit concentrations during summer, nickel removals for that period were not calculated.
For lead, the effluent concentrations during both, winter and summer, were elevated compared to the catch basin influent in both filters. Leaching tests were conducted using shale and red cedar samples placed into deionized water. These tests indicated that lead was not leaching from either material (data not shown).

Polycyclic aromatic hydrocarbons
The red cedar TF was significantly more effective (p-value<0.05) than the conventional shale TF at removing ∑PAH 16 from runoff in both seasons ( Using wood as a matrix for stormwater filtration, in particular for treating PAH, has been previously tested in the field and in the laboratory . In a laboratory study by , it was shown that red cedar amended with 6 mg/g TPA had a 69.8 times higher sorption capacity for the PAH compound fluorene compared to expanded shale. Similarly, for acenapthene, TPA amended red cedar had a 58.1 times higher affinity than expanded shale. A field study by  found that Aspen wood filters deployed in stormwater detention ponds removed on average between 18.5% and 35.6% of PAHs.
Even though ∑PAH 16 is being used as the reference for the much large class of PAH compounds, not all 16 target PAH were detected each time and the relative distribution of compounds also varied from sample to sample. In the winter, when PAHs were present in the runoff an average of 42% of the time, 4.8 compounds were detected compared to 9.6 compounds in the summer time, when PAHs were detected 74% of the time. Commonly the compounds detected in the winter were low molecular weight compounds, i.e. those with 3 or fewer benzene rings (naphthalene through anthracene), compared to the summer where a greater number of high molecular compounds, those with 4 or more benzene rings, were detected. This shift can be attributed to runoff and ambient temperatures that influence the solubility of PAHs that already have limited solubility . This shift is also supported by a study conducted in Rhode Island that showed that the PAH atmospheric deposition profile in the winter time has a strong signature of low 160 molecular compounds, whereas in summer the fraction of higher molecular weight PAH compounds increased (Schifman and Boving in review).

E. coli
Overall, both tree filters achieved E. coli removal that exceeds the RISM standard of 60%. Similar removal capabilities have been observed in other studies that investigated bacteria straining and removal through filtration ).
Comparing the E. coli removal by season, lower removal percentages were observed in winter compared to summer ( There are several possible explanations for why no significant differences between the conventional and amended TF were found during this pilot scale study. Foremost, only a comparably small amount of amended cedar wood (30 kg) was applied. This amount represents less than 1% of the TF volume. Greater amounts of amended wood are required to ensure greater removal. Second, because of the small amount of amended wood material, the contact time between the infiltrating stormwater and the bioactive matrix was insufficient. In prior laboratory experiments by Schifman et al. (in prep a), it was determined that a three hour contact time using TPA modified red cedar resulted in 2.2 log removal. In the pilot scale test described herein, the residence time was 20 minutes, therefore the residence time in the layer containing the antimicrobial wood chips is likely less than 10 minutes. Third, the TPA loading to the cedar wood was 6 mg/g, but up to 9 mg/g can be achieved as shown by . Higher TPA loading should result in greater bacteria removal percentages. Finally, the TPA amended red cedar was only applied on the top 0.5 m of the TF system instead of being homogenously mixed into the entire filter media. It is expected that better mixing would increase the contact of the infiltrating bacteria with antimicrobially active wood chips, possibly leading to more efficient bacteria removal.

CONCLUSIONS
Two tree filter units, one conventional with a sand/shale mix and one innovative containing antimicrobially amended red cedar wood chips, were compared for their contaminant removal capabilities in a field pilot study. Over the course of an 8 month period, the water in the TF catch basin and the effluent was sampled for twelve constituents. Additional stormwater grab samples were collected during some storm events. Overall, the innovative TF exhibited higher nitrate, phosphate, TSS, and ∑PAH 16 removal, However, the E. coli removal was insignificantly different from the conventional systems.

162
Even though E. coli removal by TPA amended red cedar wood chips was extensively and successfully tested in the laboratory Schifman et al. in prep. a), this pilot scale field study demonstrated that the laboratory results were difficult to replicate in the field. Possible reasons for this are 1) co-contaminants, such a high salt load during winter, may reduce the antimicrobial effectiveness of the TPA, 2) insufficient amounts of reactive material were added to the TF system, preventing sufficient contact time of the infiltrating stormwater with the matrix and, 3) the amended red cedar should have been mixed with engineered matrix more evenly, allowing for more contact between the bacteria and TPA red cedar during treatment.
At the time of the installation of the TF, the units were filled entirely with conventional sand/shale matrix. Later on, this prevented the replacement with TPA amended wood chips or better mixing of the filter matrix afterwards. If a similar study is to be repeated, it is recommended to add much more TPA amended red cedar wood chips to the system and mix it more thoroughly with the conventional sand/shale matrix before packing the TF unit with the mix. Also, the ratio of wood chips to conventional sand/shale matrix should be increased to realize contact times longer than the approximately 10 minutes in this study. It is very likely that these improvements will show that TPA amended tree filters can lower the bacteria concentrations more significantly than conventional TF systems.   Reported are the average ± standard deviation for the runoff collected in the TF catch basin. TSS values were determined from grab samples of storm water entering the TF and were collected during rain events (n=7).

INTRODUCTION
Stormwater runoff contains polycyclic aromatic hydrocarbons (PAH), heavy metals, and pathogens that are discharged into natural surface and ground water bodies, impairing ecosystems and compromising human health [1][2][3] . During and after precipitation events, these contaminants commonly exceed the maximum contaminant level (MCL) standards in runoff and receiving water bodies 1, 4, 5. High concentrations of heavy metals and petroleum hydrocarbons, such as PAHs 3 can compromise the ability to use stormwater for recharging aquifers or apply it in grey-water operations.
The concentrations of Escherichia coli (E.coli) in runoff can exceed 10 4 colonyforming units (CFU) per 100 ml 1, 6 making stormwater runoff one of the major contributors of pathogens into surface and coastal waters 5,7 . The sources of pathogens in stormwater runoff are attributed to wildlife or pets 8 and, to some degree to human fecal contamination 7 . Exposure to pathogens can lead to serious illness, such as gastroenteritis or cholera 5,9 .Ideally, structural BMPs should simultaneously attenuate organic, inorganic, and microbiological contaminants. However, most stormwater BMPs are only effective in treating heavy metals and petroleum hydrocarbons [10][11][12] through filtration and sorption [11][12][13] . They are largely ineffective in treating pathogens [14][15][16][17] . Previous studies showed that materials such as organoclays amended with quaternary ammonium compounds exhibit higher sorption for PAHs and metals 18-21 . 3-(trihydroxysilyl) propyldimethyloctadecyl ammonium chloride (TPA) is a quaternary ammonium polymer used as a disinfectant material in environmental and medical applications such as ceramic filters 22 and prosthetic devices 23 . Silver nanoparticles (AgNPs) are another well-known antimicrobial agent. Impregnating filters media with AgNPs effectively removes pathogens from aqueous solutions 24,25 .
To our knowledge these nanoparticles and the TPA polymer have never been used for stormwater runoff treatment.
One commonly used BMP are tree filters, which consist of a subsurface biofiltration system that combines filtration, sorption, and phytoremediation processes The liquid phase was then analyzed by the aforementioned method.

Experimental methodology
The laboratory experiments were divided into two phases: during the first phase, the loading capacity and stability of TPA and AgNPs amendments on RC and ES were tested. This was achieved through batch sorption isotherms (loading capacity) followed by sequential desorption for a week (stability). In the second phase, building on the isotherm results obtained during the first phase, RC was modified with different TPA and AgNPs loadings and evaluated in batch experiments for sorption capacity of PAHs and heavy metals, and also E.coli disinfection performance.

Nanoparticle and polymer loading capacity of sorbents
Batch isotherms were carried out in triplicates to determine the loading capacity of TPA and AgNP onto the sorbent materials. For all experiments, a 1.3 mmol/L ionic strength medium using sodium chloride (NaCl) was used as a background solution in order to mimic surface water conditions in Rhode Island 27 .
Detailed experimental procedures are in the supporting information Text S1. After 191 conducting the sorption experiment the samples were decanted and dried at 60°C.
Desorption of TPA and AgNP from modified sorbent materials was determined through sequential desorption in DI water for a minimum of 24 hrs and up to 168 hrs, i.e. until aqueous concentrations were below the method detection limit of 0.2 mg/L for TPA and 0.01 mg/l for Ag. The method detection limit for Ag is below EPA MCL of 0.1mg/l, while for TPA no EPA MCL limit is defined. TPA is categorized as slightly toxic with oral LD 50 above 5000 mg/l. The results from the desorption experiment were reported as total silver and TPA, respectively. The fraction of desorbed AgNPs and Ag ions were also measured and reported in supporting information Table S2.

Organic and inorganic contaminant removal efficiency of the modified sorbents
Using the results obtained in Phase I, RC was modified at different loadings of TPA and AgNPs to assess the ability to remove organic, inorganic, and microbiological contaminants as a function of loading. This process was required to adapt the lab-scale amending procedure to a large scale due to the large amount of the material required to perform all tests; details of this procedure is presented in the supplemental information Text S2. Due to disintegration of ES when agitated for modification, this material was no longer tested. However, unmodified ES was used as baseline comparison for modified RC.
Batch tests were carried out to determine the sorption capacity of the modified and unmodified sorbent materials for PAHs and heavy metals. A synthetic stormwater runoff stock solution, containing PAHs, metals and inorganic salts, was prepared (Table S3). For this part of the study all experimental solutions were adjusted to a pH of 5.5. Details of batch test are described in the supporting information Text S3.
Visual MINTEQ Ver.3.0, an equilibrium speciation software, was used to determine the speciation of metals in the synthetic runoff.

Microbiological contaminant removal efficiency by the modified sorbents
Two of the modified RC sorbents (0.6 mg/g TPA-RC and 0.6 mg/g AgNP-RC) along with unmodified RC and ES were selected to test bacteria deactivation efficiency. Five different initial E. coli concentrations relevant to stormwater runoff, ranging from 10 2 -10 6 CFU/100ml were used in the experiments 6,7 . Detailed experimental procedures are provided in supporting information Text S4.

Cost analysis and limitations
The cost for the preparation of modified red cedar was calculated based on material cost incurred from lab scale modifications. The total initial cost calculation for incorporating modified red cedar in a full scale tree filter BMP system was based on a 5 cm thick layer of modified RC for a tree box filter with dimensions 1.22m x 1.83m and drainage area of 0.5 acre.

Material Characterization
The average hydrodynamic size of AgNPs was 38.5±3.5 nm. The surface area of the modified RC increased with increasing loading of AgNPs and TPA for example, the surface area increased from 4.89±0.092 m 2 /g for UM-RC to 7.98 ±0.42 m 2 /g and 6.01±0.23 m 2 /g for 0.9 TPA-RC and 0.6 AgNP-RC respectively. The contact angle of deionized water for modified RC increased with loading of TPA and AgNPs (Table   S4). The contact angle was greater than 90 o for, 0.9TPA-RC and was greater than 70 o for 0.6TPA-RC 0.6AgNp-RC indicating that modification with TPA and AgNPs made the material more hydrophobic.

Nanoparticle and polymer loading capacity of sorbents
The TPA batch sorption on RC showed that between 37% and 98 % of the initial mass of TPA in the aqueous phase was sorbed to the RC, with a maximum loading of 0.93 ± 0.03 mg/g. For expanded shale (ES) the total mass of TPA sorbed was in the range of 30% to 75%, with a maximum loading of 0.30 ± 0.03 mg/g.
During the sequential desorption study, only 0.54 ± 0.30% of the TPA initial mass sorbed was released from RC, while for ES it was 1.13 ± 0.60%.
The higher sorption of TPA to RC is likely due to stronger interaction with the wood's molecules, such as lignin and cellulose 28 . The non-linear shape of the sorption isotherm model ( Figure S1a) supports that TPA predominantly sorbs onto RC rather 194 than diffusing into it, which would have been by linear sorption isotherm, indicative of, Fickian transport processes. In comparison, ES contains little or no organic matter due to the extreme heating during expansion process 29 . This likely resulted in fewer sorption sites for TPA. The linearity of the ES isotherm ( Figure S1 a) indicates that TPA may be partitioning into to porous spaces in the shale.
The amount of AgNP sorption to RC depended on the initial concentration of the AgNP solution. That is, RC sorbed up to 97% when the initial concentrations of AgNP were below 21 mg/L. However, the AgNP sorption to RC decreased to 75% and 10% when initial concentrations were at 52 and 104 mg/L, respectively.
Maximum loading of AgNPs to RC was 0.63 ± 0.10 mg/g.  (Figure 1b). Depending on the initial aqueous phase concentration of AgNP, between 88% and 98% of the mass desorbed is in the form AgNPs, while the remainder is in form of Ag ions (Table   S2). For ES, desorption of AgNP was <0.01% relative to the initial mass of AgNP sorbed, regardless of the initial aqueous phase AgNP concentration or mass sorbed ( Figure 1b). The fraction of AgNPs and Ag ions desorbed from ES were below method detection limit. The uptake of AgNPs can be due to trapping of AgNPs in the 195 pores of the sorbent material and sorption due to van der Waal forces of attraction between AgNPs and the sorbents.
At concentrations below 52 mg/L, the hydrodynamic size of AgNPs is ≤40 nm, whereas at higher concentrations (104 mg/l), the average hydrodynamic size of AgNPs increased to >80 nm, due to aggregation. This phenomenon was confirmed by TEM imaging ( Figure S2). The larger size could have hindered the sorption of AgNPs to the micropores in wood chips, whereas it did not reduce the uptake by ES, due to the large pores present in expanded shale 29 .

Organic and inorganic contaminant removal efficiency of the modified sorbents
The results of the batch isotherm experiments for PAHs and heavy metals onto unmodified ES as well as unmodified and modified RC are nonlinear ( Figure 2).
Assuming that the number of sorption sites is limited, the results were fitted to the Langmuir model: where q is the amount of solute sorbed (µg/g), Q is the Langmuir maximum amount of solute that can be absorbed by the sorbent (µg/g), b is the Langmuir adsorption coefficient, and C e is the equilibrium aqueous solute concentration (µg/l

Polycyclic Aromatic Hydrocarbons
The Langmuir maximum sorption capacity of PAH for ES is much lower compared to both unmodified and modified RC (Table 2), which may be a result of the vitrification process the expanded shale has undergone during thermal expansion, making it more inert by removing potential organic sorption sites 29 . In case of RC, the value of Q for acenaphthene and fluorene increased with increased loadings of TPA and AgNPs ( Figure 2). For example, for acenaphthene, Q increased from 48.6 µg/g for UM-RC to 1703.5 µg/g for 0.9TPA-RC and to 1429.2 µg/g for 0.6AgNP-RC (Table 2).
Finally, the modification with TPA and AgNPs increased the hydrophobicity of the RC as confirmed by contact angle measurements (Table S4). While the C 18 chain of the TPA molecule is enhancing the hydrophobicity of RC, in case of AgNP-RC it is the PVP coating on AgNPs. The increase in surface hydrophobicity of RC resulted in increase of PAH sorption after the amendment of RC with both AgNP and TPA. This is due to hydrophobic interactions between PAH and modified RC . In addition PAHs can also partition on to organic phase created by TPA and PVP on modified RC.

Heavy Metals
The results obtained from the Visual MINTEQ model show that at pH 5.5, the metal speciation in the synthetic stormwater is dominated (73.5 % to 87.6 %) by the divalent cations (Pb +2 , Cd +2 ,Zn +2 ,Ni +2 , and Cu +2 ) and otherwise consists of predominantly sulfate salts of e.g. PbSO 4 ( Table S6). The percentage of metals available for sorption was taken into account for batch isotherm calculations and the Langmuir isotherm model was used for quantitative comparison of the sorption capacity between the unmodified and modified materials.
Of all sorbents tested, the unmodified ES exhibited the lowest Q for all metals (Table 2). In case of RC, Q for Pb increased from 85.14 µg/g for unmodified RC to 625 µg/g for 0.9TPA-RC and 153.8 µg/g 0.6AgNP-RC (Table 2). For Cu, the TPA and AgNPs amendments enhanced the sorption of Cu compared to unmodified RC.
However, no correlation could be identified between amendment loading and Cu sorption ( Table 2). The modification of RC with TPA and AgNPs did not have an impact on Q for Cd and slightly decreased in case of Zn and Ni. ( Table 2).  (Table S4) likely provided more sorption sites for these metals.
However, the reasons for the slight decrease in Q for Zn and Ni (Table 2) is unclear.
No correlations with physiochemical properties or interactions with the amendments were found to explain the different sorption behavior of these metals to modified RC.
Therefore, further studies are required to gain insight on the sorption behavior of these metals onto modified RC.

Microbiological contaminant removal efficiency of the modified sorbents
Two modified sorbents (0.6TPA-RC and 0.6AgNP-RC) were tested for bacteria deactivation efficiency and compared to unmodified RC and ES. The disinfection performance of the materials was calculated as log removal value (LRV): LRV = log (initial E. coli concentration)log (final E. coli concentration)

Equation 2
The final E. coli concentration is the total E. coli present in the aqueous phase and sorbed to the sorbents. The average LRV for the unmodified RC and ES was always below 1 ( Figure 3). For the modified materials, the average LRVs ranged from 1.74 ± 1.04 log to 2.90 ± 0.50 log for 0.6TPA-RC. In case of 0.6AgNP-RC, the average LRV ranged from 2.03 ± 0.60 log to 2.10 ± 0.90 log. Overall, the deactivation performance shows that the modified materials are significantly more effective at deactivating bacteria compared to unmodified ES and RC (p<0.001) (Figure 3). Previous studies showed that the impregnation of TPA and AgNPs onto surfaces enhanced the deactivation of E. coli 22 Several mechanisms have been suggested for disinfection using AgNPs. These include damage to bacteria by pitting the cell membrane, lysis of cells caused by silver ion release, or damage of the cell by the reactive oxygen species formed on the surface of the AgNPs. The biotoxicity of TPA has been explained mainly by the positively charged quaternary amine groups attracting E. coli and C 18 groups piercing membrane and causing cell disruption 31 .

Cost analysis and limitations
The contaminant removal performance and bacterial deactivation of both, TPA and AgNP modified RC compare well and significantly enhanced water quality compared to unmodified RC. However, the selection of which antimicrobial agent to use for modification will be driven by the costs of amending the materials. The initial modification cost for RC using AgNP is higher compared to TPA (Table S7). For instance, to modify 1 kg RC at similar loadings (~0.6 mg/g), an AgNP modification costs $57.90/kg for versus $10.30/kg for TPA (Table S7). When installing these amendments in a field scale BMP, such as a tree filter, less than 8% of the total initial cost incur in case of TPA amendment compared to 35 % increase for AgNP. The toxicological data as determined by previous studies also suggests that TPA is less toxic compared to AgNPs (Table S8). Further, the antimicrobial performance of AgNPs primarily depends on silver ion release. Over time, the silver ion release from AgNP may decrease, and thus could result in the reduction of antimicrobial efficiency.

200
Overall, TPA has a cost advantage over AgNPs in addition to lower toxicity and better contaminant removal performance. The long-term performance of TPA is currently unknown and requires further studies. These need to include investigations of TPA behavior at different water chemistry conditions.
In summary, this study reveals that conventional materials such as unmodified expanded shale and red cedar have very limited pathogen treatment capabilities.
However, RC amended with TPA and AgNPs both increased antimicrobial properties by orders of magnitude. The amendments also enhanced PAH sorption while not hindering the sorption of most metals. These findings show that modified sorbent materials can enhance the stormwater treatment efficiency of BMP filters. The high cost involved in amending the sorbent with AgNP and its higher toxicity may prevent the use of AgNP modified materials in BMPs. Conversely, TPA modified RC has the advantage of lower cost, lower toxicity and higher contaminant and pathogen removal performance, making it the overall better choice as a filter medium for use in BMPs, such as tree filters. However, the long term performance and effectiveness of TPA modified RC under natural environmental conditions has yet to be studied.

OVERVIEW
Objective: Water quality parameters and runoff contaminants sampled 1/week at all three tree filters. Contaminants are measured in the catch basin, the influent, and effluent (sampling pan at bottom of tree filter unit) of the tree filter matrix.  (1) Liter amber glass bottle should be filled with one liter of stormwater runoff collected from the catch basin of each tree filter. The amber bottle collecting the influent sample (attached to the tree filter) should be replaced with an empty bottle. Connect portable pump to hose(s) inside TF and pump contents of the (buried) sampling pan into a one (1) L amber glass bottle.
For each sample record the water quality parameters (pH, EC, temperature) on attached field sampling data sheet. Return all samples to Dr. Craver's lab within 6 hrs. Use cooler to transport samples and to protect them. Monthly Sampling Procedure: Pump the catch basin dry after taking your sample using the sump pump. Collect the accumulated sediment/solids from the catch basin in buckets (available in the shed). Weigh the accumulated sediment in the lab to determine the amount of build up over time. Data sheet 11x1L amber glass bottles labelled appropriately prior to sampling (Example: TF type & sample type (influent, pan, or catch basin), initials, date). 3x1L clean amber glass bottles to replace the influent bottles at each tree filter Peristaltic pump "Environmental Sampler" with flexible tubing (Ensure that the pump is charged before going into the field) Calibrated pH Meter pH 4 buffer, pH 7 buffer, pH 10 buffer solutions to calibrate Calibrated Multimeter with Electrical Conductance, Nitrate, and Chloride Probes 1413 μS calibration solution Thermometer (may be included in a field meter) Analytical balance, capable of weighing to 0.001 g Computer and communication cable for Decagon field data loggers Procedure Calibrate all water quality meters according to the manufacturer's instructions before going into the field to ensure proper functionality. Fill out the data sheet (Date, personnel, weather, air temperature, last precipitation date and type, precipitation amount since last sampling) Collect one liter of the catch basin sample by submerging the bottle in the catch basin. Collect the installed influent bottle and replace it with a clean bottle. Collect one liter of the sample in the submerged pan, by attaching the flexible tubing to the tubing in the monitoring tube below the tree filter grate. Insert the flexible tubing in the peristaltic pump and pump out the pan into the appropriate bottle until dry. Take the electrical conductivity, temperature, pH, Chloride, and Nitrate readings for all samples. Allow all readings to stabilize and record the results in the appropriate data sheet. To download the data from the data loggers plug the connection chord into the communication port in the data logger and the USB end into the computer. Open the ECH2O Utility software and connect to the device. Select the data you want to download and save the data in an excel file. Save the file on dropbox under >FlaggRoad>Data>DataLogger>FilterTYPEXXXX>YYYYMMDD Check to see if the data was downloaded properly before you delete the data on the datalogger, tell it to start collecting data again, and disconnect from the software.

213
Bacteria (Standard Method 9222.D -Membrane Filtration using m_ColiBlue24 Culture Media and Standard Method 9230 -Fecal Streptococcus and Enterococcus Groups) Supplies 46 autoclaved 250 mL polyethylene bottles (2 dilutions per sample per test (2 tests), 11 samples, 2 phosphate buffer blanks) 46x 100 mL Autoclaved phosphate buffer solution 11.2 g/L dibasic potassium phosphate 9.6 g/L monobasic potassium phosphate 0.02 g/L Ethylenediaminetetraacetic acid (EDTA) Micropipette capable of measuring 1mL and 10mL Autoclaved graduated cylinder 23x 2 mL ampoules of m-ColiBlue24 (for E.coli and total coliform) 46x Sterile 0.45 μm filters with 47 mm diameter 23x 47 mm petri dishes with absorbent pad 23x 47 mm petri dishes with 5 mL m Enterococcus Agar (for enterococci) Suspend 42 g of the powder in 1 L of purified water. Mix thoroughly. Heat with frequent agitation and boil for 1 minute to completely dissolve the powder. DO NOT AUTOCLAVE. Cool to 45-50°C and dispense into 50 x 9 mm Petri dishes to a depth of 4-5 mm (approximately 4-6 mL). Incubator set to 44.8°C Incubator set to 37°C Thermometer Forceps Hot plate 1 liter beakers filled with water for sterilization of filter funnels on hot plate Vacuum filtration manifold for membrane filtration Isopropyl alcohol (ACS grade) Analytical balance, capable of weighing to 0.001 g Procedure (follow the same procedure for both pathogens until step 8, then be careful to follow the separate instructions for E. coli and enterococci sample preparation) At the lab, sterilize the workbench and filtration manifold using isopropyl alcohol. Ensure that no isopropyl alcohol residual is present. Sterilize the filter funnels by placing them in boiling water for 15 minutes Label XX autoclaved polyethylene bottles appropriately (Pathogen (E. coli or enterococci), Sample (TF type & sample type (influent, pan, or catch basin)), personnel, date). Dilute all samples twice (once for E. coli and once for Enterococci) in the following manner into labeled bottles filled with 100 ml phosphate buffer each: A: 10 mL of sample into 100 mL of phosphate buffer solution (dilution factor: 10) B: 1 mL of sample into 100 mL of phosphate buffer solution (dilution factor: 100) Prepare one blank containing 100 mL of phosphate buffer solution only.
Prepare the manifold by placing a 0.45 μm filter paper on the filtration disks using sterilized forceps Place the filtration funnel on the manifold and allow to cool. Pour the sample through the filter and turn on the vacuum filtration to allow all water to drain. E. coli preparation Prepare the petri dishes for E. coli with adsorbent pads by adding one ampoule of m-ColiBlue24 broth to each petri dish and carefully place the filter into the petri dish using forceps. Pat the petri dish on the counter to ensure that the filter paper is placed on the adsorbent pad without air bubbles and is covered in culture medium. Immediately incubate the petri dishes in the incubator at 44.8°C for 24 hours. After 24 hours of incubation count the bacteria using the colony counter pen and record the results, ensuring to adjust for the dilution factor. Blue colonies represent E. coli and red colonies represent total coliform. Enterococci preparation Obtain petri dishes for enterococci containing 5 mL m Enterococcus Agar (without adsorbent pads, prepared previously) Carefully place the filter into the petri dish using forceps after the sample has passed through the vacuum filtration. Pat the petri dish on the counter to ensure that the filter paper is placed on the culture medium without air bubbles. Allow the petri dishes to stand for 30 minutes on the bench top. Invert the petri dish and place into the incubator at 37°C for 24 hours. After 24 hours of incubation read the bacteria using the colony counter pen and record the results, ensuring to adjust for the dilution factor.
Total Suspended Solids (EPA Method 160.2) Supplies Gelman A/E or Whatman 943 AH glass fiber filter paper (0.45 μm pore size) Gooch crucible or similar and suction flask attached to vacuum filtration unit with glass flask to capture filtrate Graduated cylinder Drying oven, 103-105°C Desiccator Forceps Aluminum weighing dishes (numbered) Analytical balance, capable of weighing to 0.1 mg TSS Filtration file in Dropbox Procedure Clean a Gelman A/E or Whatman 943 AH glass fiber filter paper with 20 ml of deionized water three successive times using the filtration unit. ***Only handle filter papers with forceps from now on*** Place the filter paper into an aluminum weighing dish and dry in the oven at 103-105°C until dry. Store the weighing paper in an aluminum weighing dish in a desiccator until further use. Before use, weigh the dry filter paper on a balance and record the weight. Place the filter on the vacuum filtration unit and wet it with some deionized water to seat it against the fritted support. Shake the sample to ensure proper mixing and filter 250 ml for an influent sample and 200 ml for a pan sample (record how much you filtered in the TSS filtration sheet in Dropbox) through the pre-weighed glass fiber filter under vacuum and collect the filtrate in a glass flask. Ensure that you collect more than 10 mg of solids. Carefully remove the filter paper with the collected solids using forceps from the vacuum filtration set up and dry in an oven at 103-105°C for at least one hour or until a constant weight is reached. Capture the filtrate from the flask in a pre-labelled bottle for Heavy Metal and PAH analysis. Rinse the flask several times with deionized water before proceeding with the next filtration. Once the filter paper is dry, weigh the filter paper and calculate the non-filterable residue as follows: ( ) = ( ℎ ( ) − ℎ ( )) * 1000 ( ) Heavy Metals (Standard Method 3120) Supplies 15 mL falcon tubes labelled appropriately (Sample (TF type & sample type (influent, pan, or catch basin)), personnel, date) Nitric Acid (HNO 3 , Trace Metal grade) Inductively Coupled Plasma -Optical Emission Spectrometer Procedure Take 10 ml of the filtrate from the TSS sample preparation and place it into a 15 ml plastic falcon tube. Add nitric acid to the sample so the solution now contains 2% nitric acid (0.2 mL for 10 mL) Place the sample into the fridge and allow the sample to digest overnight. If not analyzed within 2 days, freeze one day after digestion to store the sample. Turn on the ICP-OES (located in Bliss 114) and allow to warm up (~70 minutes) Check all gases to ensure that enough resources are available to complete the analysis Ensure that enough 5% nitric acid solution is present in the eluent container. If not, prepare more using deionized water and Nitric Acid. Prepare a blank sample containing deionized water only. Place samples into the autosampler starting at position 9 Open the WinLab software on the desktop Create a new Sample Info File Record the sample names according to their positions in the auto sampler. Select the "Stormwater" method. Save the Sample Info File using the following format: YYYYMMDD Click Analyze, Rebuild List, and Analyze Samples Record all results in the excel sheet on Dropbox. Dispose of the samples in the designated hazardous waste accumulation container.
Adjust the vial number according to the sample locations in the autosampler Select the most recent method that was used to calibrate the instrument (dated) Select the most recent tuning file that was created during the last calibration (dated) Set the injection amount to 2 Set the data file name Right click the "Data File" tab. Click the "Folder" tab and select "Place in designated folder" and navigate to C/GCMSSolution/Data/Varun/Data/FlagRoad/YYYY-MM/MM-DD Click the "Data File Name" tab and select "Create name from XXXX" In the selection menu remove the Sample ID and Sample Description from the list, keep Sample Name in the list. Press OK to save. The Data File fields should now appear yellow. Delete the following fields from the table: level, internal standard amount, sample report, sample description Save the batch file in the following format YYYMMDD_Runoff under GCMSSolutions>Data>Varun>Data>Flag Road>YYYY-MM>MM-DD> Close the Analysis Editor and open GCSolutions' Real Time Analysis Close the Method window and select Open… Navigate to your folder where you saved the batch file: GCMSSolutions>Data>Varun>Data>Flag Road>YYYY-MM>MM-DD> and make sure you can open batch file in the browser window (make sure you are able to see batch files in the windows explorer window by selecting the batch file option from the drop down list in the windows explorer menu) Open your batch file Check if all parameters are correct (especially if all vial number positions correspond to the vial placement in the autosampler) Click the green start button and start the run Each run is set to take ~ 20 minutes to analyze the samples, and the column has to cool back down again, so you can calculate about 40 minutes per run. This means you can come back the next day when you analyze the metal samples and read the bacteria to note down the results from the GCMS run. Dispose of the remaining liquid from the extractions and the samples in the designated hazardous waste accumulation container.
Phosphate (USEPA1 PhosVer 3® (Ascorbic Acid) Method) Supplies 10 mL glass sample cells for Hach Spectrophotometer PhosVer® 3 Phosphate Reagent powder pillows Hach Spectrophotometer Kimwipes Procedure Turn on the Hach Spectrophotometer. Select the "Stored Program" button on the touch screen and select search by Program Number Enter 490 and hit enter Select the 490 P React PP Method. ***The following steps have to be followed for every sample (i.e. every sample has a blank and an unknown) *** Prepare the sample: Take 10 ml of the filtrate from the TSS sample preparation and place it into the sample cell. Add the contents of the PhosVer 3 Phosphate Reagent Powder Pillow to the sample cell. Swirl and shake the sample cell to dissolve the powder, taking care not to spill the sample. Allow the reaction to complete for two minutes. In the meantime, prepare the blank by filling the sample cell with 10 mL of sample. Wipe the sample cell containing the blank to remove finger prints and water drops using a Kimwipe Insert the sample cell containing the blank into the machine with the fill line facing towards the right. Hit the "Zero" button. Remove the sample cell and dispose of the blank. Wipe the sample cell containing the sample to remove finger prints and water drops using a Kimwipe Insert the sample cell containing the sample into the machine with the fill line facing towards the right. Hit the "Read" button. Record the results immediately, as this instrument does not store your data. Dispose of the samples in the designated hazardous waste accumulation container. WHAT TO DO IF SOMETHING GOES WRONG Safety Make yourself aware of the safety features in all labs. This includes the eye wash stations, fire extinguishers, emergency showers, exits, and so forth. Always wear personal protective equipment (PPE) when working in the lab. This includes googles and latex gloves at minimum. Work in a fume hood, especially when working with Nitric Acid and Organic Solvents. Know where the MSDS sheets are kept in case you have a question about the chemical you are working with or you need to access it in the case of an emergency. Always work in the lab when other people are around. This ensures that someone is there to help if a question arises or there is an emergency. Prevent emergencies by working during day light hours and when you are well rested. If you start to feel tired and notice that your attention is fading, take a break or better yet, stop your work and continue the next day.

Emergency
An emergency constitutes something like injury in the lab (cut, burn, chemical spill/burn). If this happens please call the campus emergency number at 401.874.2121. If a chemical is involved have the MSDS information available and rinse the affected area with copious amounts of water.
Disposal of Hazardous Waste You will generate hazardous waste during sample preparation. Collect the waste in a labelled container ("Hazardous Waste") and dispose of it in the designated 5 gallon polyethylene containers in 111. When you dispose of the waste, make sure that the compounds that are in your sample are also listed on the hazardous waste label that is placed on the bottle. If the compound is not yet listed, add it to the label. I screwed up in the procedure If you think you screwed up in the sample preparation, don't fret. Try to think about what you did and where the mistake occurred. Write down exactly when you realized you made a mistake and write down what you did wrong. Notify either Laura at laura.schifman@gmail.com or Varun at vkasaraneni@my.uri.edu. I don't know how to proceed If you are stuck and don't know how to proceed, please don't hesitate to ask! Either Laura or Varun should be in the lab to help.

ORDERING INFORMATION AND HOW TO RE-ORDER
If supplies are running low, meaning there are enough supplies for only two weeks of sampling left, please copy the desired items along with their vendor information, catalog number and weblink, and pricing information into an email and send it to Tom Boving at boving@uri.edu and cc both, Laura and Varun at laura.schifman@gmail.com and vkasaraneni@my.uri.edu.