Seasonal variation in indicator organisms infiltrating from permeable pavement parking lots at the Edison Environmental Center, New Jersey

Abstract

Four types of permeable pavements were monitored at the Edison Environmental Center in Edison, New Jersey, for three water quality indicator organisms consisting of fecal coliform, enterococci, and Escherichia coli. This study expands a previously published result based on less than a year of available data. The current study reflects nearly 5 years of data collection with efforts focusing on collection of data in all four seasons to analyze seasonal effects and to understand the effects of pH on infiltrate concentrations. All three indicators were detected in infiltrates from all four permeable surfaces and as well as asphalt and roof runoff. Seasonally, the infiltrate during winter had fewer detections and lower enumerations and was most often significantly different than surface infiltrate and runoff for the other seasons. More significant concentration reductions were observed in summer and fall, and the lowest reduction was observed in winter. Pervious Asphalt treatment removed the most microorganisms for all three indicator organisms. A permeable interlocking concrete pavement (PICP) that was a replacement for pervious concrete during the study performed better than the original PICP most likely due to smaller gap spacing (8 mm compared to 12.7 mm) and correspondingly smaller specified surface aggregate compared to the original PICP. Percent concentration removal reductions based on geometric means were 89% or greater for PC, PA, and PICP for fecal coliform; 75% or better for PC, PA and PICP for E. coli; and 95% or greater for PC and PA for enterococci, while there were no annual removals for enterococci for original or new PICP nor removals for E. coli for original PICP and minimal removal for fecal coliform for original PICP. The major sources of fecal indicators in the stormwater runoff were most likely from the feces of deer, geese, and other wild animals.

INTRODUCTION

Stormwater runoff continues to be a major cause of water pollution in urban areas as it discharges microbial contaminants along with nutrients, sediments, and other toxic chemicals to receiving waters (USEPA, 2009). These pollutants can cause water quality degradation in receiving waters, thereby impairing the health of aquatic life, as well as contaminating drinking water resources (USEPA, 2000a). Stormwater control measures (SCMs), such as wet ponds, wetlands, bioretention areas, dry detention basins, permeable pavements, rain gardens, and proprietary devices, are used to treat stormwater runoff. SCMs are being incorporated onsite as low impact development or as green infrastructure in the municipal right of way. Permeable pavement is a form of low impact development or green infrastructure infiltration system designed to treat and manage stormwater near the source as the stormwater runoff infiltrates into the ground through a permeable pavement or other stabilized surfaces (Eisenberg et al., 2015; Field & Sullivan, 2003). Permeable pavement usually diverts stormwater runoff into an underground stone reservoir before gradually exfiltrating out of the stone reservoir into the subsoil (Field & Sullivan, 2003), though there are also systems that have a limited storage reservoir for various reasons (e.g., high groundwater and significant underground infrastructure) that discharge to the nearest conveyance system or surface water.

For management of urban stormwater runoff, permeable pavements are considered a good SCM application as these pavements improve water quality as well as restore the predevelopment hydrology (Borst & Brown, 2014). Possible pollutant removal mechanisms for 13 metals including arsenic, antimony, aluminum, cadmium, copper, chromium, lead, manganese, and zinc in permeable pavements were filtration, adsorption, biological degradation, transformation, and volatilization (Liu & Borst, 2018). Others (Clausen & Gilbert, 2003; Ellis et al., 2004; Gilbert & Clausen, 2006; James & Thompson, 1997; Rushton, 2001) have also shown permeable pavement systems can improve stormwater runoff quality after infiltrating through the system.

There are a variety of permeable pavements available for both commercial and residential installations. The EPA’s Office of Research and Development along with EPA Region 2 constructed a functional, 0.4 ha, 110-space parking lot in 2009 at the Edison Environmental Center (EEC) in Edison, New Jersey, that is used daily by EEC staff and visitors during workdays. This study initially chose to evaluate three pavement systems: permeable interlocking concrete pavers (PICP), porous concrete (PC), and pervious asphalt (PA). PC and PA are variations of the typical concrete mixture and typical hot mix asphalt, respectively. The mix for PC omits fine aggregate and the slurry is tamped or rolled in place, whereas the mix for PA omits the fine portion of the aggregate (Dietz, 2007). PICP uses impermeable pavers placed in beds of aggregate to infiltrate water between the pavers.

There is evidence in the literature that microorganism concentrations in stormwater varies seasonally. Selvakumar and Borst (2006) reported that concentrations of microorganisms in urban stormwater runoff were significantly affected by the season during which the samples were collected. The lowest concentrations were observed during winter except for Staphylococcus aureus. Hathaway et al. (2010) monitored an urban watershed in Raleigh, North Carolina, for Escherichia coli, enterococci, and fecal coliform and found statistically significant differences during seasons. A study in Nashville, Tennessee (Young & Thackston, 1999), had an order of magnitude higher fecal coliform counts in summer than winter. These results were consistent with Nationwide Urban Runoff Program data where fecal coliform densities in urban runoff during the warmer months of the year were approximately 20 times greater than those found during cooler periods (USEPA, 1983). Earlier, Geldreich et al. (1968) showed stormwater from city streets, suburban business district storm drains and a wooded hillside adjacent to a city park had peak total coliform, fecal coliform, and fecal streptococci densities occurring in either summer or autumn. Similarly, Evans et al. (1968) collected stormwater samples in the residential and light-commercial areas of the Mt. Washington community in greater Cincinnati, Ohio, and found that total coliform, fecal coliform, and fecal streptococci densities were greatest in summer and lowest in winter. A study in two northwest Arkansas streams found significant seasonal influences on fecal coliform and fecal streptococci concentrations, with the highest concentrations occurring in summer (Edwards et al., 1997).

The above literature shows that indicator organism measurements vary seasonally with increases during warmer periods of the year that coincides with maximum use of surface waters for recreational activities such as fishing and swimming. Therefore, it is important to understand the seasonal performance of SCMs in removing indicator organisms from stormwater runoff. Few studies have assessed the effectiveness of SCMs on the seasonal removal of microorganisms (Hathaway & Hunt, 2012). Hathaway and Hunt (2012) observed greater reduction of indicator organisms in SCMs during the non-swimming season (beginning of November to end of March) although the systems that performed well overall reduced concentration of indicator organisms throughout the year. They concluded more data are needed to strengthen the postulation that SCM effectiveness for indicator bacteria may vary throughout the year. Li and Davis (2009) observed the highest E. coli and fecal coliform concentrations in runoff during the summer though SCM removal efficiency was not correlated to the temperature. Tata-Maharaj and Scholz (2010) evaluated the efficiency of 12 permeable pavement systems constructed in bins and compared the performance for removal of microorganisms between March 2008 and May 2009 and found that removals of microorganisms did not appear to be affected by temperature variations due to seasonal changes. In a literature review of permeable pavements, Drake et al. (2013) indicated microbial populations had not yet been explicitly studied and there continued to be need for long-term water quality performance studies for permeable pavements.

Abdollahian et al. (2018) captured stormwater samples that infiltrated through articulating concrete blocks/mats (ACBM) to the bottom of a trench and shaft and observed mean reductions of E. coli concentrations of 60% in the trench and 78% in the shaft when compared to samples from runoff. It was also observed that pollutant loadings in the stormwater runoff, as well as pollutant reductions by ACBM, were affected by the intensity of sampled rainfall events.

In an earlier study, Selvakumar and O’Connor (2018) monitored the infiltrate of three types of permeable pavements and asphalt driving lane runoff at EEC, for three indicator organisms: fecal coliform, enterococci, and E. coli and they found PA removed more indicator organisms than PC and PICP. One reason for lower number of organisms in PA infiltrate was thought to be the slower measured infiltration rate of PA (Brown & Borst, 2014), though it was also concluded that the lower count in PA infiltrate could be due to high pH observed in porous asphalt system infiltrate which might be due to asphalt emulsions used in processing (Selvakumar & O’Connor, 2018). The PA infiltrate was also identified as a potential leaching source of SVOCs compared to other permeable pavement infiltrates (O’Connor, 2017), which may be toxic to the organisms. Also, of note, other research at EEC (Borst & Brown, 2014) found that the PA retained deicing chemicals longer through the year than the other permeable surfaces. The previous study found poor correlation to storm size or temperature similar to the findings of other researchers (Selvakumar & O’Connor, 2018).

While the previous studies have indicated that permeable pavement systems may reduce the concentration of various pollutants, there is limited research available currently on removal of indicator organisms and the seasonal performance of the systems. In a report (Clary et al., 2017) of summary statistics for International Stormwater BMP Database, the first identified research need was for additional stormwater control data of performance for removal of EPA recommend indicator organisms. The objectives of this study were to (1) evaluate the seasonal effects on indicator organisms such as fecal coliform, enterococci, and E. coli on infiltrate concentrations from permeable pavement and (2) evaluate the effect of pH on concentrations of indicator organisms to better understand possible causes for the observed lower infiltrate concentrations of indicator organisms in PA.

Fecal indicator microorganisms can be found in feces from both human sources and domestic and wild animals (Whitlock et al., 2002). The Edison Environmental Center is an 89 ha (220 acre) complex with many grassed and wooded areas, and there are estimated to be between 50 and 100 deer on the property. Canada geese also roam the property, particularly Atlantic (Branta canadensis) which are considered to be “Resident Canada geese” (https://njaes.rutgers.edu/fs1214/) living year found in New Jersey. The previous research (Selvakumar & O’Connor, 2018) established collected runoff concentration of CFU ranges within established literary values.

EXPERIMENTAL METHODS

As noted above, the parking lot (Figure 1) had three different permeable pavement types: PICP, PC, and PA. The three head-to-head parking rows of permeable pavement systems are 8.5 m (28 ft) m wide by 42.7 m (140 ft) long, while the 7.6 m (25 ft) wide travel lanes are paved with traditional impervious hot mix asphalt. There is a 1.6% surface slope so that each permeable surface receives runoff from the adjacent travel lanes to the north. All surfaces were constructed over an open-graded subbase reservoir of recycled concrete aggregate (RCA) crushed on site to the size of American Association of State Highway and Transportation Officials (AASHTO) No. 2 size aggregate. Five sections of each permeable pavement system allow water to infiltrate into the underlying soil, while four of the sections have an impermeable liner 40 cm below each permeable surface that allows infiltrate to be collected for sampling and analysis. The thickness of each permeable surface varies depending on structural needs for that application. The PA is 8 cm thick, and PC was 15 cm thick and was placed directly on the RCA reservoir. The individual pavers, which are 8 cm thick, were placed on top of two bedding layers comprised of a top 5 cm layer of AASHTO No. 8 aggregate and a bottom 10 cm layer of AASHTO No. 57 that was placed over the common RCA reservoir. The AASHTO No. 8 aggregate also filled the 12.7-mm (0.5-in) wide gap between pavers.

FIGURE 1.

Plan view of parking lot

The initial infiltration capacities of all three surfaces were very large (USEPA, 2010) with the infiltration rate of PC approximately twice that of PICP, and both PICP and PC were more than one order of magnitude larger than PA. Although the surface infiltration rates vary by more than an order of magnitude, even the mean infiltration rate (200 cm/h) (USEPA, 2010) of the slowest surface, that is, porous asphalt, exceeds the local 100-year, 5-min peak rainfall intensity (20.9 cm/hr) (Brown & Borst, 2014). A more detailed description of the liners, permeable surfaces and drainage piping is provided in Brown and Borst (2014).

In the summer of 2016, the PC treatment was replaced with PICP (referred to as “new PICP”) as the concrete started raveling, which is the deterioration of the pavement surface caused by the dislodging of aggregate particles. This failure was attributed to deicing application. Recommendations for concrete are that “Deicing chemicals should not be used … in the first year” (NRMCA, 2015), but this may not be practical on a “working” parking lot.

The surface dimensions of the original PICP blocks (ECO Paver, EP Henry, Woodbury, New Jersey) are 15.5 cm (6.1 in.) by 21.6 cm (8.5 in.) with a thickness of 7.94 cm (3.125-in.) and there is a 12.7-mm (0.5-in) wide gap between blocks which yields a surface open area of approximately 11% for water to infiltrate through. The newer PICP blocks (Belgard 80 mm AQUAROC Natural) have surface block dimensions of 11 cm (4.5 in.) × 23 cm (9 in.) and are just as thick as the original PICP but have an 8 mm (0.3 in.) gap between pavers yielding a smaller surface open area of approximately 9%. This system also used the AASHTO Nos. 57 and 8 aggregate over RCA for PICP but used AASHTO No. 9 for gaps between pavers.

Flow-weighted samples were collected using programmable automatic samplers. Samples were collected from the drainage pipes to the collection tanks for lined sections 1 and 3 (Figure 1) for each permeable surface. Samples of runoff from impermeable asphalt surface driving lane and parking spaces were collected at two curb cuts (CC) (rain gardens number 4 and 5) at the south end of the parking lot. Samples from these locations have been shown to be representative of urban runoff and are used for comparison to and calculation of removals from infiltrate values through the permeable surfaces. Both dissolved and total metal concentrations for Cr, Cu, Ni, Zn, As, Cd, and Pb in asphalt surface runoff (Liu & Borst, 2018) were within one order magnitude of the median values in the National Stormwater Quality Database (Pitt et al., 2004) indicating that the surface runoff component was representative of typical stormwater runoff. Roof runoff (DB) from a nearby building was sampled as well. Rainfall was measured using a 0.1 mm tipping bucket rain gauge and recorded with a Campbell Scientific CR1000 data logger (Logan, Utah) set at 10-minute intervals. The rain gauge was located in the field adjacent to the collection tanks to the east of the parking lot. All storms had at least 2.54 mm (0.1 in.) of rainfall as per NPDES guidance (USEPA, 1992). Details of the overall water quality sampling efforts of parking lot and roof runoff were described in Borst and Brown (2014) and O’Connor and Amin (2015), respectively.

A total of 42 sampling events (15 in summer, 11 in fall, 9 in winter, and 7 in spring) were conducted between July 2015 and April 2020. Collected samples were transported to the onsite laboratory, and analyses were initiated within the standard holding time of 6 h. Samples were analyzed for three indicator microorganisms: fecal coliform, enterococci, and E. coli. Fecal coliform and E. coli were enumerated using Colilert and enterococci was enumerated using Enterolert (IDEXX Laboratories, Inc., Westbrook, Maine). Colilert® and Enterolert® are commercially available enzyme-substrate liquid-broth mediums (IDEXX Laboratories, Inc., Westbrook, Maine). All enumerations were performed using Quanti-tray 2000 trays and following provided instructions, which use a most probable number (MPN) based protocol with a quantitation range from less than 1 colony forming unit (cfu)/100 ml to 2,419.6 cfu/100 ml without sample dilution. Each sample was analyzed with and without dilution; infiltrate was analyzed at 10 times dilution (observable range of 10 to 24,196 cfu/100 ml) and runoff samples were initially analyzed at 10 times dilution but were increased to 20 times dilution (observable range 20 to 48,392 cfu/100 ml).

Event results were the average of the undiluted and diluted analysis. The following rules were used in calculating the sample concentrations when samples either exceeded upper threshold limit or fell below detection. When observed concentrations resulted in a non-detection for both the undiluted and diluted analysis, a non-detection value was used for analysis. When the diluted sample resulted in a non-detection, then only the undiluted value was used. When undiluted was >2,419.6 cfu/100 ml and diluted has a value, then the diluted value only was used. If both undiluted and diluted samples exceed numeric threshold of 2,419.6 cfu/100 ml, then the dilution value (10 or 20) times 2419.6 was used. If the undiluted test resulted in a non-detection and diluted was >1, then the diluted value was averaged with a zero value for non-diluted result. Duplicate samples were averaged according to rules above and resulted sometimes in values less than 1.

In addition, a bench-scale study was conducted to determine the effect of pH on concentrations of microorganisms. Duplicate samples were divided in 100 ml bottles. Experiments were conducted at six different pH values (7–12) in duplicate. The pH of the samples was adjusted using 10, 1, and 0.1 M sodium hydroxide solution as the pH of the surface runoff was nearly neutral. The study was conducted during six different sampling events between April 2017 and July 2018, and asphalt surface runoff samples were collected at the two CC numbers 4 and 5 (Figure 1). Linear regression models were built to investigate if the concentrations of fecal coliform and enterococci changed with pH. Data analysis was performed in R project for Statistical Computing (R Development Core Team, 2014) at significance level of α = 0.05.

DATA ANALYSIS

As previously tested by the Shapiro–Wilk W test using Satistica 10 (StatSoft, 2011) (Selvakumar & O’Connor, 2018) and reconfirmed for larger data set, indicator microorganism concentrations for each permeable surface were not normally distributed. These data contained numerous non-detects; therefore, ProUCL Version 5.1.002 (Singh & Singh, 2015) was used to calculate geometric mean, median, and Kaplan–Meier (KM) mean, and other statistics. As a common practice, non-detects were assigned a value of 1 for geometric mean calculations (Costa, 2021) and values below the detection limit of 1 that were used in the calculation of geometric means were not normalized to 1 as this would have increased the value of the geometric mean. The Kaplan–Meier (KM) estimation method is a non-parametric statistical approach that uses a distribution function estimate and adjusts for non-detects; when there were no non-detects, the KM mean reverted to an ordinary mean.

As per ProUCL, the lower limit for comparing geometric mean and statistical threshold value (STV) was to have at least eight observations and four non-detects and over 10% detections (Singh & Singh, 2015). These minimum requirements were exceeded for general statistics for testing of each surface (i.e., infiltrate of permeable pavement or runoff from surfaces) with one exception for PA which had <10% detection for E. Coli. There were more cases of insufficient sampling size for general statistics for each surface when evaluating seasonal effects. This was due to smaller observation sizes (>8 samples) for porous concrete infiltrate, its replacement surface with PICP; and roof runoff (which only had one sample per event) or insufficient number of detected samples as for PA.

ProUCL (Singh & Singh, 2015) has several nonparametric tests capable of comparing datasets with left censored data or numerous non-detects. The Tarone-Ware (TW) comparison test which tests for differences in medians of left-censored datasets (Singh & Singh, 2015) was used for two-sample hypothesis statistical testing because of the ability to calculate differences when there were many samples below the detection limit (<1) and the data set also incorporated fractional values below the DL. Testing of two-sample hypotheses testing sampling sizes was maintained above the lower limitation of eight observations and minimum of four detections and greater than 10% detections as per recommendation of ProUCL (Singh & Singh, 2015); below, this limits’ analysis was not performed. EPA guidance (USEPA, 2000b) suggests that analysis of proportions of detections through KM estimation method and two-sample hypothesis statistical testing by TW comparison test (as performed through ProUCL) are appropriate for data containing 10–50% detections of all collected data. The TW comparison test first used a two-tailed test of significance with a null hypothesis of equality and an alternative hypothesis of “not equal.” If results were determined to be not equal, then Sample 1 > Sample 2 or Sample 1 < Sample 2 was tested. All tests were performed at 95% confidence.

For comparison of surface infiltrate, only similar event time periods were compared, that is, July 2015 to July 2016 sampling events for runoff versus PC comparison, and September 2016 to April 2020 for runoff versus new PICP as new PICP replaced PC in August 2016. The latter time frame was also used for original PICP to new PICP analysis.

The Mann Whitney test (StatSoft, 2011) was used for a small fraction of data set comparisons that had no non-detects or had very few nondetects (<10) with over 50% detections and greater than 20 observations. The value for non-detect was assigned below either the detection limit of 1 or lowest fractional result, as per guidance in ProUCL (Singh & Singh, 2015). The results of comparison testing by Mann Whitney were similar to TW comparison testing, such that when Mann Whitney test registered a statistical difference so did the TW analysis, and vice versa for observation of no difference; therefore, only the TW comparison tests were reported as they were directly applicable to a greater proportion of the datasets.

Treatment by infiltration through the pavement surfaces was determined by calculating percent removal of infiltrate concentration of each surface in comparison to asphalt surface runoff the samples collected at the two CC.

RESULTS AND DISCUSSION

Summary statistics for the complete sampling of events conducted between July 2015 and April 2020 are presented in Tables 1–9 for fecal coliform, E. coli, and enterococci, for all four surfaces and runoff, respectively. All three organisms were detected in infiltrates from all four surfaces and runoff as opposed to earlier results summarizing the sampling events conducted between July 2015 and February 2016 where all three organisms were detected only in infiltrates from PC and PICP. For PA infiltrate, previously enterococci had the most detections, while fecal coliform was detected only once, and E. coli was never detected (Selvakumar & O’Connor, 2018).

TABLE 1.

Annual and seasonal fecal coliform runoff concentrations

	Asphalt runoff					Roof runoff
	Annual	Winter	Spring	Summer	Fall	Annual	Winter	Springa	Summer	Fall
Number storm events	41	9	7	14	11	42	9	7	15	11
Number of valid data	79	18	14	25	22	42	9	7	15	11
Non-detects (%)	1 (1.3)	1 (5.6)	0	0	0	1 (2.4)	0	1 (14)	0	0
Exceed max (>2,419.6)	8	0	0	5	3	0	0	0	0	0
Geometric mean	317.8 b	13.3	70.53	3,055 b	849.5 b	153.6	5.975	60.28	739.7 b	465.7 b
KM mean	6,383	55.44	792.9	12,725	7,912	862.2	21.08	552.4	1,637	690.6
KM standard of deviation	12,768	107.4	2,286	17,013	13,181	1,780	43.71	878.8	2,763	665.7
KM coefficient of deviation	2.000	1.937	2.883	1.337	1.666	2.065	2.074	1.591	1.688	0.964
Median	302.7	18.53	116.8	3,795	1,525	356.9	5.588	236.3	746.7	419.6
Minimum (>non-detect)	1	1	1	81.95	7.5	1	1	2	65.9	122.5
Maximum	51,987	427	8,704	48,392	51,987	11,199	136.3	2,649	11,199	2,315
90th percentile	24,196	134.4	585	44,926	24,196	1,893	45.21	1,356	2,570	1,358

^aSampling size below suggested minimum of eight observations to draw conclusions, results italicized.

^bExceeds geometric mean of 200 MPN/100 ml (EPA, 1976), result bolded.

TABLE 9.

Annual and seasonal enterococci concentrations for pervious concrete and porous asphalt

+	PC					PA
	Annual	Winter	Springa	Summer	Fall	Annual	Winter	Spring	Summer	Fall
Number storm events	18	4	1	5	8	42	9	7	15	11
Number of valid data	36	8	2	10	16	84	18	14	30	22
Non-detects	3 (8.3)	0	2 (100)	0	1 (6.3)	38 (45)	9 (50)	6 (43)	14 (47)	9 (41)
Exceed max (>2,419.6)	0	0	0	0	0	0	0	0	0	0
Geometric mean	15.5	27.89	1	8.594	23.39	2.1	2.555	2.158	1.964	2.012
KM mean	67.8	75.79	NA b	27.28	97.51	5.5	7.01	8.218	3.813	5.148
KM standard of deviation	122.1	95.52	NA	47.47	160.5	10.3	10.91	15.8	6.635	8.613
KM coefficient of deviation	1.80	1.26	NA	1.74	1.646	1.9	1.556	1.922	1.74	1.673
Median	14.94	60.51	1	6.4	18.98	1	1	1	1	1
Minimum (>non-detect)	0.833	0.833	NA	1	1		1	0.5	0.5	0.5
Maximum	562.6	296.1	NA	152.8	562.6		38.95	54.93	29.2	25.85
90th percentile	218.3 c	151.2 c	1	65.77	336 c	21.6	22.09	28.51	9.998	22.93

^aSampling size below suggested minimum of 8 observations to draw conclusions, results italicized.

^bNot applicable.

^cExceeds 90% of statistical threshold value (STV) of 130 cfu per 100 ml (EPA, 2012), result bolded.

The geometric mean for fecal coliform exceeded the old bathing water quality standards (BWQS) of 200 MPN/100 ml (USEPA, 1976) in PICP infiltrate and asphalt runoff. The geometric mean of E. coli did not exceed the recreational water quality criteria (RWQC) (USEPA, 2012) of 126 MPN/100 ml in any of the infiltrates. The geometric mean for enterococci exceeded the RWQC (USEPA, 2012) of 33 MPN/100 ml in both new and old PICP infiltrates and asphalt runoff. Similar trend was noticed in sampling events conducted between July 2015 and February 2016 (Selvakumar & O’Connor, 2018).

Results of TW comparison tests are summarized in Tables SM-1, SM-2, and SM-3 for fecal coliform, E. coli, and enterococci, respectively. For both fecal coliform and E. coli, asphalt runoff (the source for permeable infiltrate testing), is larger than new PICP, PC and PA (PA not tested for E. Coli due to <10% detection) indicating that these surfaces reduce indicator organisms, while the original PICP does not show removal. The roof runoff is not statistically different than asphalt runoff for fecal coliform and E. coli, but enterococci in asphalt runoff is statistically greater than roof runoff. For enterococci, only PA infiltrate shows removal, while the new PICP and PC reveal no statistical difference to asphalt runoff and PICP is statistically larger than the asphalt runoff, which would indicate the PICP could be thought of as a source for enterococci pollution.

Seasonal study

Statistical analyses were performed to evaluate the seasonal effect on infiltrate concentrations. The equinox and solstice dates were used to divide the year into four seasons, and they are defined as (Winter = December 22 – March 21; Spring = March 22 – June 21; Summer = June 22 – September 21; and Fall = September 22 – December 21).

The results of seasonal concentrations for asphalt and roof runoff are summarized in Tables 1–3 for fecal coliform, E. coli, and enterococci, respectively. The spring roof runoff for enterococci concentration exceeds the 90% STV, though there is the distinct possibility that the sample size was too small as seven observations falls below the minimum of eight observations and there is the influence of a maximum value of 1117 MPN/100 ml which was observed during this season which is an order of magnitude larger than observed maximum for other seasons. Results cannot be stated with confidence.

TABLE 3.

Annual and seasonal enterococci runoff concentrations

	Asphalt runoff					Roof runoff
	Annual	Winter	Spring	Summer	Fall	Annual	Winter	Springa	Summer	Fall
Number storm events	41	9	7	14	11	42	9	7	15	11
Number of valid data	79	18	13	26	22	42	9	7	15	11
Non-detects (%)	1 (1.3)	0	0	1 (3.8)	0	5 (12)	1 (11)	2 (29)	1 (6.7)	1 (9.1)
Exceed max (>2,419.6)	1	0	0	0	1	0	0	0	0	0
Geometric mean	39.0 b	11.74	28.8	54.85 b	83.3 b	11.7	13.26	7.909	10.24	16.43
KM mean	1,182	30.6	42.89	2,267	1,515	73.2	62.69	167.4	42.88	63.08
KM standard of deviation	6,044	57.73	40.63	9,278	5,229	184.1	119.2	387.8	71.21	89.81
KM coefficient of deviation	5.114	1.887	0.947	4.093	3.452	2.52	1.902	2.317	1.661	1.424
Median	23.9	12.8	28.45	24.93	43.3	9.225	8.6	6	13.1	9.25
Minimum (>non-detect)	1	1	6.7	6.2	2	1	3.1	1	1	2
Maximum	48,392	246.1	126.7	48,392	24,196	1,117	384.8	1,117	242.2	312.8
90th percentile	613.6 c	62.96	108.7	2,251 c	922.7 c	183.5 b	173.2 c	469.5 c	145.4 c	121.3

^aSampling size below suggested minimum of eight observations to draw conclusions, results italicized.

^bExceeds geometric mean of 35 colony forming units (CFU per 100 ml; USEPA, 2012), result bolded.

^cExceeds 90% of statistical threshold value (STV) of 130 cfu per 100 ml (USEPA, 2012), result bolded.

The asphalt runoff for enterococci implies a peak in summer and fall as both the geometric mean and STV exceeds 35 and 130 CFU per 100 ml while the roof runoff exceeds the 90% STV for winter, spring (limitation as noted) and summer.

The results of seasonal concentrations for the original and new permeable interlocking concrete pavers are summarized in Tables 4–6 for fecal coliform, E. coli, and enterococci, respectively. The PICP infiltrate exceeds the geometric mean and 90% STV for enterococci (USEPA, 2012) for all seasons for the original PICP. The new PICP exceeds the geometric mean and 90% STV threshold during spring and summer, while it is below the respective 35 CFU and 130 CFU per 100 ml thresholds for winter and results are inconclusive for fall due to insufficient sample size.

TABLE 4.

Annual and seasonal fecal coliform concentrations for permeable interlocking concrete pavers

	PICP					New PICP
	Annual	Winter	Spring	Summer	Fall	Annual	Winter	Spring	Summer	Falla
Number storm events	42	9	7	15	11	24	5	6	10	3
Number of valid data	83	18	14	29	22	47	10	12	19	6
Non-detects (%)	3 (36)	0	0	2 (6.9)	1 (4.6)	23 (49)	4	8 (67)	8 (42)	3 (50)
Exceed max (>2,419.6)	2	0	0	1	1	0	0	0	0	0
Geometric mean	285.3 b	48.63	118.7	625.1 b	753.8 b	2.4	2.327	3.116	2.235	1.663
KM mean	2,280	459	612.8	3,140	3,696	13.4	4.595	20.96	17.1	2.342
KM standard of deviation	4,868	1,448	766.4	5,729	6,060	41.6	6.079	41.6	54.99	2.229
KM coefficient of deviation	2.135	3.154	1.251	1.824	1.64	3.11	1.323	1.985	3.216	0.952
Median	420.1	31.45	208.1	1,057	1,432	1	1	1	1	1
Minimum (>non-detect)	1	5.1	1	22.9	22.9	0.5	1	5	0.5	1
Maximum	24,196	6,131	2,187	24,196	24,196	247.3	21.25	128.5	247.3	7.05
90th percentile	5,670	506.2	1,783	9,925	10,659	21.4	8.875	88.18	24.53	5.025

^aSampling size below suggested minimum of eight observations to draw conclusions, results italicized.

^bExceeds geometric mean of 200 MPN/100 ml (EPA, 1976). result bolded.

TABLE 6.

Annual and seasonal enterococci concentrations for permeable interlocking concrete pavers

	PICP					New PICP
	Annual	Winter	Spring	Summer	Fall	Annual	Winter	Spring	Summer	Falla
Number storm events	42	9	7	15	11	24	5	6	10	3
Number of valid data	83	18	14	29	22	47	10	12	19	6
Non-detects (%)	1 (1.2)	0	0	1 (3.5)	0	1 (2.1)	1 (10)	0	0	0
Exceed max (>2,419.6)	2	0	0	2	0	1	0	0	1	0
Geometric mean	150.3 b	35.7 b	307.3 b	271.1 b	142.1 b	50.1 b	2.7	79.2 b	225.5 b	21.8
KM mean	1,210	119.4	1,148	2,454	502.8	863.8	4.598	235.6	1,976	29.84
KM standard of deviation	3,892	173.3	1,318	6,261	985.8	3,519	4.356	254.2	5,486	24.08
KM coefficient of deviation	3.216	1.451	1.148	2.551	1.96	4.075	0.947	1.079	2.776	0.807
Median	177	20.36	655	196	163.2	29.78	1.75	172.6	270.6	21.85
Minimum (>non-detect)	1	3.1	6.675	12.7	1	1	1	3.1	10.4	7.875
Maximum	24,196	659.3	3,777	24,196	4,523	24,196	12.63	738.2	24,196	63.88
90th percentile	2,106 c	314 c	3,059 c	3,988 c	1,088 c	1,189 c	11.07	589 c	2,716 c	59.39

^aSampling size below suggested minimum of 8 observations to draw conclusions, results italicized.

^bExceeds geometric mean of 35 colony forming units (CFU per 100 ml; EPA, 2012), result bolded.

^cExceeds 90% of statistical threshold value (STV) of 130 cfu per 100 ml (EPA, 2012), result bolded.

The results of seasonal concentrations for porous concrete and porous asphalt are summarized in Tables 7, 8, and 9 for fecal coliform, E. coli, and enterococci, respectively. Neither PC or PA exceeds geometric mean criteria for enterococci, though interestingly winter and fall PC infiltrate exceed the STV of 130 cfu per 100 ml (USEPA, 2012), with winter at the minimal sampling threshold of 8 observations for PC infiltrate.

TABLE 7.

Annual and seasonal fecal coliform concentrations for pervious concrete and porous asphalt

	PC					PA
	Annual	Winter	Springa	Summer	Fall	Annual	Winter	Spring	Summer	Fall
Number storm events	18	4	1	5	8	42	9	7	15	11
Number of valid data	36	8	2	10	16	84	18	14	30	22
Non-detects	7 (19)	2 (25)	1 (50)	3 (30)	1 (6.3)	55 (65)	13 (72)	9 (64)	16 (53)	17 (77)
Exceed max (>2,419.6)	0	0	0	0	0	0	0	0	0	0
Geometric mean	34.3	18.14	3.504	82.29	36.4	2.9	2.076	2.905	4.203	2.225
KM mean	227.5	81.39	6.638	550.3	126.4	54.3	6.822	28.69	111.3	32.14
KM standard of deviation	373.7	149.8	5.638	515.3	212.1	214.1	12.85	71.87	338.2	91.4
KM coefficient of deviation	1.643	1.84	0.849	0.936	1.678	3.95	1.884	2.505	3.04	2.844
Median	48.5	24.08	6.638	553.1	58.6	1	1	1	1	1
Minimum (>non-detect)	12.28	13.95	12.28	50.8	5.9	0.5	2	2	0.5	0.5
Maximum	1,384	473.9	12.28	1,384	840.7	1,624	44.85	270.3	1,624	357.3
90th percentile	867.7	187	11.15	1,202	335.6	44.6	20.99	76.08	170.7	41.65

^aSampling size below suggested minimum of eight observations to draw conclusions, results italicized.

TABLE 8.

Seasonal concentrations of E. coli for pervious concrete and porous asphalt

	PC					PA
	Annual	Wintera	Springb	Summera	Fall	Annual	Wintera	Springa	Summera	Falla
Number storm events	18	4	1	5	8	42	9	7	15	11
Number of valid data	36	8	2	10	16	84	18	14	30	22
Non-detects	25 (69)	5 (60)	2 (100) c	9 (90) c	9 (56)	78 (93)c	18 (100) c	12 (86)	28 (93) c	20 (91)c
Exceed max (>2,419.6)	0	0	0	0	0	0	0	0	0	0
Geometric mean	1.3	1.504	1	1.283	1.282	1.2	1	1.098	1.127	1.429
KM mean	1.6	2.606	NA d	2.105	1.289	2.2	NA	0.993	1.603	6.427
KM standard of deviation	2.8	3.868	NA	3.315	1.123	10.7	NA	1.777	3.069	19.86
KM coefficient of deviation	1.7	1.484	NA	1.575	0.871	4.8	NA	1.79	1.914	3.09
Median	1	1	1	1	1	1	1	1	1	1
Minimum (>non-detect)	0.5	1	NA	12.05	0.5	0.5	NA	0.5	2	27.55
Maximum	12.8	12.8	NA	12.05	4.1	93.85	NA	7.4	18.1	93.85
90th percentile	2.8	5.275	1	2.105	2.8	1	1	1	1	1

^aDetected observations below minimum of four detectable observations to draw conclusions, results italicized.

^bSampling size below suggested minimum of eight observations to draw conclusions, results italicized.

^cGreater than 90% of samples were considered non-detects.

^dNot applicable.

As for general surface statistical results, all seasonal results of E. Coli are below geometric mean of 126 CFU/100 ml and the STV of 410 CFU/100 ml with the caveat that certain seasonal observations did fall below threshold for minimum observations or detectable observations (as noted in Table 2 for Spring roof runoff, Table 5 for Fall new PICP and 8 for spring PC) and indications that results are below the criteria for all seasons should be used with caution.

TABLE 2.

Annual and seasonal E. coli for runoff concentrations

	Asphalt runoff					Roof runoff
	Annual	Winter	Spring	Summer	Fall	Annual	Winter	Springa	Summer	Fall
Number storm events	41	9	7	14	11	42	9	7	15	11
Number of valid data	80	18	14	26	22	42	9	7	15	11
Non-detects (%)	21 (26)	8 (44)	4 (29)	7 (27)	2 (9)	15 (36)	5 (56)	2 (29)	4 (27)	4 (36)
Exceed max (>2,419.6)	1	0	0	0	1	0	0	0	0	0
Geometric mean	5.3	1.674	3.053	6.07	16.39	3.5	1.983	6.959	4.803	2.242
KM mean	899.5	2.166	31.63	699.6	2,422	10.6	2.8	21.06	15.46	3.859
KM standard of deviation	4,501	1.691	96.61	3,327	7,555	22.1	2.292	30.35	27.65	5.066
KM coefficient of deviation	5.004	0.781	3.055	4.756	3.119	2.07	0.819	1.441	1.788	1.313
Median	2	1	1.5	2.55	7.775	2.1	1	6	4.1	2
Minimum (>non-detect)	0.167	1	0.5	0.5	0.167	1	2.2	3	1	1
Maximum	28,272	6.3	378.6	17,329	28,272	111.5	6	92.2	111.5	18.7
90th percentile	141.1	4.486	27.63	163.1	230.5	25.05	6	52.33	32.17	7.55

^aSampling size below suggested minimum of 8 observations to draw conclusions, results italicized.

TABLE 5.

Annual and seasonal E. coli concentrations for permeable interlocking concrete pavers

	PICP					New PICP
	Annual	Winter	Spring	Summer	Fall	Annual	Wintera	Springa	Summera	Fallb
Number storm events	42	9	7	15	11	24	5	6	10	3
Number of valid data	83	18	14	29	22	47	10	12	19	6
Non-detects (%)	23 (28)	4 (22)	5 (36)	13 (45)	1 (4.6)	40 (85)	8 (80)	9 (75)	19 (100) c	4 (70)
Exceed max (>2,419.6)	0	0	0	0	0	0	0	0	0	0
Geometric mean	5.8	4.087	7.558	3.874	11.27	1.2	1	1.977	1	1.217
KM mean	317.7	43.49	65.63	797.5	70.28	4.2	1	13.29	NA d	1.375
KM standard of deviation	1,948	150.4	114.7	3,235	146.1	19.4	0	36.92	NA	0.839
KM coefficient of deviation	6.130	3.459	1.747	4.057	2.078	4.6	NA	2.779	NA	0.61
Median	3	3.6	3.05	1	7.275	1	1	1	1	1
Minimum (>non-detect)	0.167	0.167	1	0.5	1	1	1	2	NA	1
Maximum	17,329	662.6	407.7	17,329	549.7	135.3	1	135.3	NA	3.25
90th percentile	196.9	28.24	193.5	261.1	314.4	1	1	12.06	1	2.125

^aDetected observations below minimum of four detectable observations to draw conclusions, results italicized.

^bSampling size below suggested minimum of eight observations to draw conclusions, results italicized.

^cGreater than 90% of samples were considered non-detects.

^dNot applicable.

Comparison of seasons

A preliminary evaluation using TW comparison tests indicated that when winter differed from all other seasons combined, winter was likely to also be different from at least some of the individual seasons; but when testing indicated no difference between winter and other seasons combined, then there was no difference between winter or other seasons. So first, if hypotheses test indicated winter was different than other seasons combined (not winter), then all individual seasons were tested by TW comparison tests; if test indicated that winter was not different than other seasons combined, testing ceased. This approach may potentially miss some seasonal differences but does identify the most prominent trend.

The number of tests per season available for comparison testing for a particular surface was much less than available tests for comparison testing between surfaces. The number of available tests was even less for PC which ended in 2016 and roof runoff that only had a single sample per event. There was insufficient amount of winter tests (<10) being nine and eight for PC and roof runoff, respectively; therefore, only PICP, parking lot runoff, PA and the new PICP could be tested for seasonality, with other limitations of sampling sizes noted.

The results of TW comparison test for seasonal differences are summarized in Tables SM-4, SM-5, and SM-6 for fecal coliform, E. coli, and enterococci, respectively. Table SM-4 shows fecal concentrations in parking lot runoff during summer are larger than any other season and fall is larger than spring while summer PICP infiltrate is larger than winter infiltrate.

Table SM-5 shows that nonwinter runoff is different than winter runoff and, specifically, fall runoff contains statistically larger counts of E. coli than in winter runoff. No statistical test could be performed for E. coli on PA as there were no detection for E. coli for winter and there were less than four detections in winter for the new PICP. While it is obvious that there is a difference between winter and non-winter with six detections for PA, other individual seasons could not be tested with only two detections for each.

Table SM-6 shows that enterococci concentrations were larger for all tested seasons than winter for PICP, new PICP and runoff.

These results generally agree with literature that the concentration of organisms is lowest in winter (Selvakumar & Borst, 2006) and higher in warmer months compared to cooler months (Badge & Rangari, 1999; Edwards et al., 1997; Evans et al., 1968; Geldreich et al., 1968; Hathaway et al., 2010; Young & Thackston, 1999;).

Potential reduction in concentration of indicator organisms

Table 10 presents the estimated percent seasonal and overall concentration reductions of microorganisms for each permeable pavement type in comparison to the asphalt surface runoff. Percent removals were calculated using the geometric means.

TABLE 10.

Percent reduction of permeable parking surface infiltrate

	Permeable surfaces
	PICP			New PICP			PC			PA
Season	FC	E. C.	Ent	FC	E. C.	Ent	FC	E. C.	Ent	FC	E. C.	Ent
Summer	80	36	NR	100	84	NR	97	79	84	100	81	96
Fall	11	31	NR	100	93	74	96	92	72	100	91	98
Winter	NR	NR	NR	83	40	77	NR	10	NR	84	40	78
Spring	NR	NR	NR	96	35	NR	NS	NS	NS	96	64	93
Annual	10	NR	NR	99	77	NR	89	75	60	99	77	95

Abbreviations: E. C., E. coli; Ent, enterococci; FC, fecal coliform; NR, no reduction; NS, no samples.

Concentration reductions greater than 90% were observed in PA infiltrate except during winter for all the organisms and summer, spring and all four seasons for E. coli. Greater concentration reductions were observed in summer and fall and the lowest were observed in winter. All surfaces had lower removals in winter though there were correspondingly lower observations in runoff during winter. The new PICP performed better than the older PICP. The older PICP infiltrate had the least removal performance for all three organisms in all four seasons. All seasons’ percent removal is higher than winter and spring and lower than summer and fall for all the surfaces and all three organisms. Previous study (Selvakumar & O’Connor, 2018) indicated the indicator organisms has low coefficient of determination values for temperature, rainfall depth, and rainfall intensity (30-minute). Regressions were performed on the calculated removals for 33 events, i.e. (geometric mean of runoff-geometric mean of surface) divided by geometric mean of runoff, versus temperature, rainfall depth, and rainfall intensity. These regressions had similarly poor R², i.e. <0.2 for old PICP, PA, and PC. However, the newer PICP surface had three R² that exceeded 0.2 twice for temperature (R² = 0.57 for fecal coliform and R² = 0.24 for enterococci) and once for rainfall intensity (R² = 0.40 for enterococci). There was a smaller data set for newer PICP, 15 events only, but the decreased removal at lower temperature may be actual effect as shown in Figure 2, while the higher R² for enterococci may be spurious result of negative removals (Figures SM-1 and SM-2).

FIGURE 2.

Calculated Fecal Coliform Removal vs Temperature

EFFECTS OF PH, OTHER RUNOFF CONSTITUENTS AND POROSITY

The bench-scale study of the effect of different pH on concentrations of microorganisms used samples collected from six sampling events of asphalt runoff. Fecal coliform and enterococci were detected in runoff from all the storm events while E. coli was detected in only one event and at very low concentrations There was no trend for the models for fecal coliform and enterococci (Figures 3 and 4, respectively) indicating poor association between pH and indicator organism concentrations. As can be seen in the figures, fecal coliform concentrations were affected beyond a pH of 11, while concentration of enterococci were not affected by pH.

FIGURE 3.

Effect of pH on concentration of fecal coliform

FIGURE 4.

Effect of pH on concentration of enterococci

Initially, it was suspected that the high pH from asphalt might be a reason for low counts of indicator organisms in the PA infiltrate. However, the analysis of pH effects indicated minimal if any effect on fecal coliform or enterococci, with exception of fecal coliform above pH 11. As stated earlier, the PA infiltrate was shown to contain leached SVOCS (O’Connor, 2017) and therefore toxicity of PA infiltrate may play a role. Slower infiltration rates and greater sorption by the asphalt may also be the cause for high removal. Brown and Borst (2014) showed that salt binded more easily to asphalt during winter application and leached out during other seasons. Regressions of several other infiltrate constituents and porosity versus median indicator organisms concentrations for PICP, PC and PA were performed. Mean chloride (Borst & Brown, 2014) had low R² ranging from 0.14 for E. coli to 0.2 for fecal coliform. Samples mean pH (Brown & Borst, 2015) had R² ranging from 0.44 for E. coli to 0.52 for fecal coliform while mean total organic carbon (Razzaghmanesh & Borst, 2019) had a similar range for R² with a low 0.45 for E. coli and a high of 0.52 for fecal coliform. The porosity of all four surfaces were measured (old PICP = 0.464, PC = 0.223, PA = 0.209, and new PICP = 0.405) and regression was performed for the median indicator organisms concentrations. Porosity values of the PICP were derived from the smallest aggregated used in the construction. Fecal coliform had a R² of 0.42 while E. coli had a R² of 0.53. The highest R² was for enterococci at 0.73 and the regression is shown in Figure 5. While not conclusive, the physical properties of the type of permeable pavement may be the most predictive of the reduction of indicator organism concentrations.

FIGURE 5.

Measured Porosity vs Enterococci Median Concentration

CONCLUSIONS

Four types of permeable pavements were monitored at EEC in Edison, New Jersey, for indicator organisms fecal coliform, enterococci, and E. coli. A total of 42 sampling events (an additional 26 events over that of the original study) were conducted between July 2015 and April 2020. All three organisms were detected in infiltrates from all four treatments (PICP, new PICP, PC, and PA) and asphalt and roof runoff. Presence of all three indicator organisms is a significant change for PA infiltrate from earlier sampling events conducted between July 2015 and February 2016 when only one fecal coliform was detected and E. coli was never detected (Selvakumar & O’Connor, 2018).

There was a statistically significant difference between asphalt runoff and infiltrate from PA, PC and new PICP permeable surfaces for both fecal coliform and E. coli. For enterococci, only PA and roof runoff were statistically significant different from asphalt runoff. New PICP infiltrates for all three organisms tested were statistically less than PICP infiltrates. There was a statistically smaller concentration for roof runoff than parking lot asphalt runoff for enterococci only.

Statistical analyses were performed to evaluate the seasonal effect on infiltrate concentrations. There was a statistically significant difference between winter and non-winter seasons for fecal coliform for runoff and PICP infiltrate; summer parking lot runoff fecal concentrations were larger than any other season, and fall was larger than spring and summer PICP infiltrate was larger than winter PICP infiltrate. For E. coli, there was a statistically significant difference between winter and non-winter seasons for runoff only. For enterococci, there was a statistically significant difference between winter and other seasons for runoff and infiltrates from both PICP and new PICP.

Higher concentration reductions of microorganisms from all the permeable surfaces were observed in summer and fall and the lowest reductions were observed in winter. PA infiltrate had the greatest calculated reductions for all three organisms. The new PICP had more and larger calculated reductions than the older PICP, and this was attributed to the lower surface open area for water to infiltrate through the newer PICP. The porosity of the surfaces may be the best indicator for removal of the indicator organisms.

Practitioner Points.

• The infiltrate during winter had fewer detections and lower enumerations and was most often significantly different than surface infiltrate and runoff for the other seasons.

• More significant concentration reductions were observed in summer and fall, and the lowest reduction was observed in winter.

• Pervious Asphalt treatment removed the most microorganisms for all three indicator organisms.

DATA AVAILABILITY STATEMENT

All the data are publicly available and provided upon request.