READ Delambert2021

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Science of the Total Environment 775 (2021) 145738

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Microbial pathogens and contaminants of emerging concern in


groundwater at an urban subsurface stormwater infiltration site
Jane R. de Lambert a,⁎, James F. Walsh a, Deanna P. Scher a, Aaron D. Firnstahl b, Mark A. Borchardt c
a
Minnesota Department of Health, 625 Robert St. N, St. Paul, MN 55164, United States
b
U.S. Geological Survey Upper Midwest Water Science Center, 2615 Yellowstone Drive, Marshfield, WI 54449, United States
c
U.S. Department of Agriculture-Agricultural Research Service, Environmentally Integrated Dairy Management Research Unit, 2615 Yellowstone Drive, Marshfield, WI 54449, United States

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Infiltration to control stormwater runoff


is common but may degrade groundwa-
ter.
• Contaminants measured beneath storm
water infiltration sites at multiple depths.
• Rain a significant driver in microbial var-
iation across sampling events.
• Pathogens present at all depths but low-
est below fine-grained sediment.
• Infiltrating urban stormwater may
threaten geologically sensitive aquifers.

a r t i c l e i n f o a b s t r a c t

Article history: Urban stormwater may contain a variety of pollutants, including viruses and other pathogens, and contaminants
Received 28 September 2020 of emerging concern (pharmaceuticals, artificial sweeteners, and personal care products). In vulnerable geologic
Received in revised form 29 December 2020 settings, the potential exists for these contaminants to reach underlying aquifers and contaminate drinking water
Accepted 4 February 2021
wells. Viruses and other pathogens, as well as other contaminants of emerging concern, were measured in
Available online 10 February 2021
stormwater and groundwater at an urban site containing a stormwater cistern and related subsurface infiltration
Editor: Jay Gan gallery, three shallow lysimeter wells, and a monitoring well. Five of 12 microbial targets were detected more
than once across the eight rounds of sampling and at multiple sampling points, with human-specific Bacteroides
Keywords: detected most frequently. The microbial and chemical contaminants present in urban stormwater were much
Water microbiology lower in the water table monitoring well than the vadose zone lysimeters. There may be numerous causes for
Zoonotic pathogens these reductions, but they are most likely related to transit across fine-grained sediments that separate the
Contaminant water table from the vadose zone at this location. Precipitation amount prior to sample collection was signifi-
Hydrogeology cantly associated with microbial load. A significant relation between microbial load and chloride-bromide ratio
Water quality
was also observed. The reduction in number and concentrations of contaminants found in the monitoring well
Runoff infiltration
indicates that although geologically sensitive aquifers receiving urban stormwater effluent in the subsurface
may be prone to contamination, those with a protective cap of fine-grained sediments are less vulnerable.
These results can inform stormwater infiltration guidance relative to drinking water wells, with an emphasis
on restricting infiltration near water supply wells finished in geologically sensitive aquifers to reduce public
health risks.
© 2021 Elsevier B.V. All rights reserved.

⁎ Corresponding author at: Minnesota Department of Health, 625 Robert St. N, St. Paul, MN 55164, United States.
E-mail address: jane.de.lambert@state.mn.us (J.R. de Lambert).

https://doi.org/10.1016/j.scitotenv.2021.145738
0048-9697/© 2021 Elsevier B.V. All rights reserved.
J.R. de Lambert, J.F. Walsh, D.P. Scher et al. Science of the Total Environment 775 (2021) 145738

1. Introduction serve as indicators of elevated microbial load. We also assessed the rela-
tion between precipitation amount prior to sampling and microbial
The high demand for water resources has led stormwater capture concentration to determine if targeting sample collection after rainfall
and reuse to become a popular practice in many communities across events increases the likelihood of finding a detectible or above-normal
the United States, particularly those affected by droughts (USEPA, level of microbial contamination.
2019). Population growth, changes in climate, and greater irrigation
needs can stress the demand for high quality water, even in water- 2. Methods
rich states such as Minnesota (Interagency Workgroup on Water
Reuse, 2018). Capturing stormwater for irrigation reduces the use of po- 2.1. Site description
table water sources and minimizes pollutant loads reaching surface wa-
ters, including vulnerable wetland systems. The study site is located at Upper Villa Park in Roseville, Minnesota,
Despite its utility in providing an alternative use to potable water USA, in Ramsey County. This park houses a baseball field and is
sources, urban stormwater has been widely found to carry human surrounded by developed land to the north and west and deciduous for-
enteric pathogens and other microbiological and chemical contami- est, wetlands, and a small, unnamed lake to the east (Fig. 1). The
nants (Sauer et al., 2011; Sercu et al., 2009; Sidhu et al., 2012, 2013). stormwater infrastructure at the site consists of a concrete cistern (ap-
Little is known about the fate and transport of these constituents in proximately 18 m × 14.5 m and 2.5 m high) and related subsurface in-
the subsurface during stormwater infiltration and the resulting ef- filtration gallery. The cistern was built to harvest and hold stormwater
fects on groundwater quality (NAS, 2016). From 2016 to 2017, the entering from a municipal storm sewer, and then release to a 1700
Minnesota Department of Health (MDH) conducted a monitoring cubic meter underground infiltration gallery located beneath an asphalt
study to characterize viruses, other pathogens, and chemical con- parking lot and constructed of perforated, corrugated metal pipe. Runoff
taminants of emerging concern in stormwater at varying depths in to the storm sewer is primarily from roadways and residential proper-
the vadose and phreatic zones beneath a subsurface stormwater in- ties. The infiltration system automatically releases water from the cis-
filtration gallery. tern to the gallery after a rain event. A portion of the infiltrated water
The main goals of this work were to increase understanding of the is pumped out of the cistern and used to irrigate the baseball field in
transport and fate of stormwater contaminants in the subsurface and the park; the remaining water in the cistern travels to the infiltration
improve assessment of groundwater contamination risk related to this gallery at a depth of approximately 10 to 15 ft below ground surface.
practice. We also sought to improve upon the study design of larger- Three non-weighing lysimeters were installed at the site and finished
scale, future monitoring efforts through evaluation of this pilot study. at depths of 3, 5, and 7 ft below the infiltration gallery. These will be re-
To target future sampling, we looked for the relation between microbial ferred to as Lysimeter 3, Lysimeter 5, and Lysimeter 7, respectively. Ad-
concentration and general water quality parameters to determine ditionally, a monitoring well was installed at the park and is located
whether any of these routinely collected and lower-cost analytes may between the infiltration system and the receiving wetland (Fig. 1). A

Fig. 1. Upper Villa Park site location.

2
J.R. de Lambert, J.F. Walsh, D.P. Scher et al. Science of the Total Environment 775 (2021) 145738

municipal sanitary sewer runs from west to east along the length of through the vadose zone following rain events. It is estimated that the
the park. flush of stormwater released to the infiltration gallery following a
storm event passes by the lysimeters within as little as 40 h (Capitol
2.2. Site geology and hydrogeology Region Watershed District, 2017). In contrast, the water in the monitor-
ing well has an estimated recharge age of approximately 1 year based
The surficial geology of the study site consists of approximately on sulfur hexafluoride (SF6) analysis (Sharon Kroening, Minnesota Pol-
200 ft of unconsolidated Pleistocene sediments overlying bedrock of lution Control Agency, written communication, 2019). As such, the
the Ordovician-age St. Peter Sandstone (Meyer and Swanson, 1992). lysimeters represent a geologically sensitive portion of the groundwater
The water table at the site is approximately 30 ft below the land surface, system, whereas the monitoring well is drawing from a less sensitive
and groundwater flow is southeasterly, towards McCarrons Lake, into portion due to its greater depth and presence of relatively low-
which the unnamed lake directly east of the site flows. The generalized permeability sediments above the well screen (Minnesota DNR, 1991).
geologic soil profile inferred by the site boring logs indicates 2 to 6.5 ft of
fill, consisting mostly of lean clay, overlying native soils. These are gen- 2.3. Stormwater collection system
erally composed of coarse alluvium to depths of approximately 23 ft and
represent glacial meltwater sediments of the Grantsburg sublobe Stormwater flow through the cistern and infiltration gallery is
(Meyer and Swanson, 1992). Finer-grained sediments (silty till and guided by the Opti-RTC system. Opti-RTC constantly checks forecasted
fine alluvium) are noted locally below 23 ft, especially near the monitor- precipitation 48 h into the future. If a rain event that is greater than
ing well, until coarse alluvium is again encountered from 31 to 43 ft. 0.05 in. and greater than 60% probability is forecasted, the Opti-RTC sys-
Mixed silty till and fine alluvium are noted below that depth. The sedi- tem will open the valve between the cistern and the infiltration gallery
ment surrounding the cistern, three lysimeter wells, and the infiltration (generally 24 h prior to the forecasted rain event), emptying the stored
gallery is primarily native fill material that was replaced after excava- cistern water into the infiltration gallery. The stormwater from the fore-
tion and installation of the infiltration system. Given that the lysimeters casted event refills the cistern. Therefore, the water being collected in
are completed at depths of 3, 5, and 7-ft below the base of the infiltra- the three lysimeters is water that was stored in the cistern from a previ-
tion gallery, which is itself buried approximately 15 ft below grade, ous event. In some circumstances, the cistern is evacuated, and then the
they are completed in coarse alluvium with an estimated infiltration forecasted event does not occur. Alternately, the forecasted event could
rate of 6–7 in/h (American Engineering and Testing, Inc., 2013). The be much larger than predicted causing the cistern to reach the overflow
monitoring well, screened from 33 to 43 ft below grade, is open to a level, whereby stormwater from the current event overflows into the
mix of coarse alluvium and mixed silty till and fine alluvium that is sep- infiltration gallery. In this case, sampling efforts after a cistern release
arated from the lysimeters by intervening silty or clayey sediments with and event overflow could represent multiple storm events (Sarah
an estimated infiltration rate of less than or equal to 1.4 in/h (Fig. 2). The Wein, Capitol Region Watershed District, written communication,
lysimeters capture stormwater and groundwater percolating rapidly 2020). Additionally, precipitation from the predicted rain event also

Fig. 2. Upper Villa Park geologic cross-section.

3
J.R. de Lambert, J.F. Walsh, D.P. Scher et al. Science of the Total Environment 775 (2021) 145738

may reach the lysimeters through rapid recharge from the ground procedures described in Stokdyk et al. (2016). Hydrolysis probes
surface. were used for quantification, and standard curves were created
from gBlocks® and Ultramer® oligos (Integrated DNA Technologies,
2.4. Sample collection Coralville, Iowa). Forward and reverse primer sequences, probe se-
quences, and standard curve metrics for all qPCR assays are reported
MDH collected eight rounds of samples between November 2016 in Supplemental Material (SM). Following Gibson et al. (2012),
and November 2017 at the cistern, three lysimeters, and monitoring lambda phage DNA and hepatitis G virus armored RNA were used to
well. Samples were collected on the following dates: November 9, evaluate all samples for inhibition of qPCR and reverse transcription-
2016; and May 24, June 16, July 14, July 28, August 22, September 12, qPCR, respectively. Negative controls were included at all processing
and November 21 in 2017. Sampling events occurred at or greater steps (secondary concentration, nucleic acid extraction, reverse tran-
than 48 h following a release of the cistern (whether manual or auto- scription, and qPCR) and must exhibit no fluorescence above the base-
matic) to ensure that the stored storm water had been emptied into line for data to be acceptable. Modified live virus vaccines (Zoetis Inc.,
the infiltration gallery by the time of sample collection. The cistern re- Kalamazoo, Michigan) were used for DNA (bovine herpes virus) and
leases were generally timed to occur 24 h prior to forecasted precipita- RNA (bovine respiratory syncytial virus) extraction positive controls,
tion events; therefore, samples contained stormwater from a mix of with the latter serving also as the reverse transcription positive control.
recent rain events, infiltration gallery recharge, infiltrating recharge
from the ground surface not associated with the infiltration gallery, 2.6. Chemical parameter sample collection and analysis
and in the case of the monitoring well, groundwater from outside the
infiltration gallery area. The sanitary sewer present onsite was also sam- Samples were collected once during November 2016 from the
pled one time by MDH staff in November 2016. sewer, cistern, Lysimeter 5, and the monitoring well. MDH shipped sam-
Samples were collected at the cistern and monitoring well using a ples on ice to the Water and Environmental Analysis Laboratory (WEAL)
Geotech submersible pump. Mean and range sample volumes for the cis- at the University of Wisconsin, Stevens Point. WEAL analyzed the sam-
tern and monitoring well were 339 L (range 57–693 L) and 91 L (range ples for a suite of pharmaceuticals, artificial sweeteners, and personal
1.9–144 L), respectively. The three lysimeter wells were pre-equipped care products that are considered indicative of wastewater (Table II)
with individual Proactive Mini-Typhoon Purging pumps which were using high performance liquid chromatography and mass spectrometry
used for sample collection. All tubing was replaced at the beginning of (ESI-LC/MS/MS). Prior to solid phase extraction, Benzoylecgonine-D3, a
each sampling event. At the cistern and monitoring well, microbial sam- surrogate standard, was added to all samples, blanks, and lab spikes. Du-
ples were collected using dead-end ultrafiltration using Hemodialyzer plicate, blank, and lab-spiked samples were prepared and analyzed for
Rexeed-25 s filters (Asahi Kasei Medical MT Corp., Oita, Japan). Due to approximately 10% of samples. A seven-point calibration was used for
the presence of sediment and organic material, pre-filters were used each analyte. Deuterated analogs of target analytes were used as inter-
during five of eight sampling events at the cistern beginning in July nal standards.
2017 to reduce ultrafilter fouling and risk of filter breakage.
At the three lysimeter wells, 1-l samples for microbial analyses were 2.7. Water quality parameters
collected in high-density polyethylene (HDPE) bottles. Microbial sam-
ples were shipped on ice and analyzed at the Laboratory for Infectious After microbial samples were collected, MDH collected additional
Disease and the Environment in Marshfield, Wisconsin, by qPCR (quan- samples to measure general water quality parameters (Table III), unless
titative polymerase chain reaction). the available volume of water was insufficient for collection. Samples
were brought to the MDH Public Health Laboratory (PHL) in St. Paul,
2.5. qPCR analyses Minnesota for analysis.

The 12 microbial targets analyzed are as follows: adenovirus groups 2.8. Statistical methods
A-F and C,D,F, enteroviruses, hepatitis A virus, human polyomavirus,
norovirus genogroups I and II, pepper mild mottle virus (PMMV), rotavi- Descriptive statistics on the individual microbial, chemical, and
rus group A (NSP3) gene, human-specific HF183 Bacteroides (hereafter water quality parameter results are reported by sampling location. We
HF183 Bacteroides) and Bacteroidales-like Hum M2, and Cryptosporidium also created an indicator of total microbial load by summing all micro-
spp. Upon arrival to the laboratory, ultrafilters were back flushed follow- bial targets with results available for the eight sampling events (herein
ing Smith and Hill (2009) and 1% of desiccated beef extract (mass/ referred to as “microbial sum”). Because several microbial targets were
volume) was added to backflush eluate. Pre-filters were transferred to not detected in any of the five sampling locations, and Cryptosporidium
plastic bags and soaked in 200 mL of 3% beef extract solution containing and Bacteroidales-like Hum M2 were not analyzed during the first round
0.05 M glycine at a pH of 9.5 for 15 min. Liquid was then squeezed from of sample collection, microbial sum ultimately encompassed HF183
pre-filters and transferred to a sterile 500-mL bottle and pH adjusted to Bacteroides, rotavirus group A, and pepper mild mottle virus. The rela-
7.0. Ultrafilter and pre-filter eluates were further concentrated by poly- tion between the amount of precipitation prior to the sampling dates
ethylene glycol precipitation (Lambertini et al., 2008; Stokdyk et al., (National Weather Service Station ID: GHCND:USC00218450, approxi-
2016). One-liter samples were supplemented with desiccated beef ex- mately 5.2 km from the studied site) and total microbial sum were eval-
tract (1% mass/volume) and then concentrated using the polyethylene uated using the total precipitation amount in the 7 days prior to the
glycol precipitation method. All final concentrated sample volumes sample collection date. Microbial sum was also assessed in relation to
(FCSV) from ultrafilter and 1-l samples were stored at −80 °C until anal- the concentration of the following water quality parameter and bacte-
ysis. Nucleic acid extraction was performed with QIAcube® and QIAamp rial indicators (selected due to previously reported association with
DNA blood mini kits with buffer AVL (Qiagen, Valencia, California). To sewage, wastewater, and/or stormwater (USEPA, 1986, Tiefenthaler
obtain sufficient template, two 280 μL aliquots of FCSV were extracted, et al., 2008, NAS, 2016): boron, sulfate, potassium, chloride, the mass
each with a final elution volume of 120 μL. RNA gene targets were ratio of chloride to bromide (Cl\\Br ratio), nitrate, Escherichia coli
reverse-transcribed using random hexamers (ProMega, Madison, (E. coli), and total coliform. The relation between microbial sum and
Wisconsin) and SuperScript® III reverse transcriptase (Invitrogen Life precipitation, and microbial sum and water quality parameters overall
Technologies, Rockville, Maryland) following procedures described in and within each sampling location, were assessed visually. Due to the
Stokdyk et al. (2016). qPCR was performed using a LightCycler® 480 non-normality of the data, non-parametric tests were used to assess
instrument (Roche Diagnostics, Mannheim, Germany) following the strength and direction of correlation between total microbial sum

4
J.R. de Lambert, J.F. Walsh, D.P. Scher et al. Science of the Total Environment 775 (2021) 145738

Table I
Microbial results.a

Microbial target Concentrations in genomic copies per liter (gc/L)

Cistern Lysimeter 3 Lysimeter 5 Lysimeter 7 Monitoring well

HF183 Bacteroides # Positive: 5/8 # Positive: 6/8 # Positive: 7/8 # Positive: 7/8 # Positive: 7/8
p50: 8 p50: 1008 p50: 1920 p50: 1456 p50: 38
p75: 121 p75: 1350 p75: 7406 p75: 3425 p75: 141
Max: 96,429 Max: 2132 Max: 10,314 Max: 46,496 Max: 367
Rotavirus group A (NSP3 gene) # Positive: 2/8 # Positive: 0/8 # Positive: 1/8 # Positive: 1/8 # Positive: 0/8
p50: ND p50: ND p50: ND
p75: 5 p75: ND p75: ND
Max: 11,213 Max: 14,428 Max: 5141
Pepper mild mottle virus (PMMV) # Positive: 5/8 # Positive: 2/8 # Positive: 1/8 # Positive: 1/8 # Positive: 0/8
p50: 17 p50: ND p50: ND p50: ND
p75: 212 p75: 10,507 p75: ND p75: ND
Max: 21,013 Max: 43,027 Max: 218,783 Max: 9635
Cryptosporidium spp. # Positive: 2/7 # Positive: 4/7 # Positive: 2/7 # Positive: 0/7 # Positive: 0/7
p50: ND p50: 887 p50: ND
p75: 0.8 p75: 1330 p75: 1677
Max: 6 Max: 15,260 Max: 5930
Bacteroidales like Hum M2 # Positive: 1/7 # Positive: 0/7 # Positive: 1/7 # Positive: 1/7 # Positive: 1/7
p50: ND p50: ND p50: ND p50: ND
p75: ND p75: ND p75: ND p75: ND
Max: 55 Max: 513 Max: 9805 Max: 6
a
ND = no detection. p50 = 50th percentile, p75 = 75th percentile. Table excludes adenoviruses, enterovirus, hepatitis A, human polyomavirus, and norovirus genogroups I and II, as all
non-sewer results were negative. Cryptosporidium and Bacteroidales like Hum M2 were not analyzed during the first round of sample collection in November 2016 (n = 7).

and water quality parameters (Kendall's tau (τ)) as well as differences in the cistern (63%). The highest median concentrations of HF183
in total microbial sum between ordinal categories of precipitation and Bacteroides were found in the lysimeters. Bacteroidales-like Hum M2
water quality parameters (Kruskal-Wallis test). Two-sided p-values was detected only once at all sampling sites except Lysimeter 3; the
<0.05 were considered significant. Due to a skewed and wide distribu- highest concentration was found in Lysimeter 7. Cryptosporidium spp.
tion, microbial sum was log transformed to aid visual interpretation; a was detected most frequently and at the highest concentration in Ly-
small value (0.01 gc/L) was assigned to microbial sum non-detections simeter 3 (57% positive), with 29% of results positive from the cistern
in visual plots. All analyses were conducted using SAS 9.4 (SAS Institute and Lysimeter 5. It was not detected in Lysimeter 7 or the monitor-
Inc. Cary, North Carolina). ing well. Rotavirus group A (NSP3 gene) was detected in 25% of
samples from the cistern and in 13% of samples from both Lysime-
3. Results ters 5 and 7. It was not detected in Lysimeter 3, the monitoring
well, or the city sewer. Pepper mild mottle virus (PMMV) was de-
3.1. Microbial results tected in varying concentrations at all sampling sites except for
the monitoring well. The highest frequency of detection was in
Enterovirus, human polyomavirus, and norovirus genogroup II were the cistern.
detected at high concentrations in the sewer (1.3 × 105 gene copies L-1,
6.8 × 105 gene copies L-1, and 1.6 × 107 gene copies L-1, respectively), 3.2. Chemical parameters
but were not found in any other sampling location throughout the re-
mainder of the study. The microbial results from the sanitary sewer All chemicals except two pharmaceuticals were detected in the san-
are included in the SM. At the other sampling locations (cistern, lysim- itary sewer (Table II). Only cotinine and caffeine were detected in the
eters, monitoring well), five of 12 microbial targets were detected more cistern sample, with the former also being detected in the sample col-
than once across the study period and at multiple sampling points lected from Lysimeter 5. The monitoring well had no detections of phar-
(Table I). HF183 Bacteroides was the most frequently detected microbial maceuticals, artificial sweeteners, wastewater indicators, or personal
target, with 80% positive samples overall. Percent detection was lowest care products.

Table II
Chemical description and results.a

Chemical Description Group MRL (ng/L) Concentration (ng/L)

Sewer Cistern Lysimeter 5

Acesulfame Artificial sweetener Artificial sweeteners 5 31,735 <MRL <MRL


Sucralose 25 46,642 <MRL <MRL
Saccharin 25.0E 15,480 <MRL <MRL
Cotinine Nicotine metabolite Wastewater indicators 3 1911 5.2 5.1
Caffeine Stimulant 12 118,388 18.0 <MRL
Paraxanthine Caffeine metabolite 5 31,127 <MRL <MRL
Carbamazepine Antiepileptic Pharmaceuticals 2 1213 <MRL <MRL
Acetaminophen Analgesic 35.0E 204,774 <MRL <MRL
Trimethoprim Human antibiotic 5.0E 10.9 <MRL <MRL
Sulfamethazine Bovine antibiotic 1 <MRL <MRL <MRL
Sulfamethoxazole Human antibiotic 5.0E <MRL <MRL <MRL
Venlafaxine Antidepressant 5.0E 257 <MRL <MRL
Triclosan Antimicrobial Consumer products 75 411 <MRL <MRL
a
<MRL = less than method reporting limit. The monitoring well was also sampled – all results were < MRL. One sample per location.

5
J.R. de Lambert, J.F. Walsh, D.P. Scher et al. Science of the Total Environment 775 (2021) 145738

Table III
Water quality parameter results.a

Water quality parameter MRL Cistern Lysimeter 3 Lysimeter 5 Lysimeter 7 Monitoring well

Bromide (mg/L) 0.005 # detects: 7/7 # detects: 7/7 # detects: 8/8 # detects: 8/8 # detects: 8/8
Min: 0.02 Min: 0.01 Min: 0.02 Min: 0.02 Min: 0.05
p50: 0.03 p50: 0.03 p50: 0.03 p50: 0.02 p50: 0.16
p75: 0.06 p75: 0.04 p75: 0.04 p75: 0.04 p75: 0.25
Max: 0.06 Max: 0.04 Max: 0.05 Max: 0.04 Max: 0.28
Chloride (mg/L) 0.500 # detects: 8/8 # detects = 7/7 # detects: 8/8 # detects: 8/8 # detects: 8/8
Min: 49.5 Min: 32.9 Min: 55.0 Min: 57.0 Min: 65.9
p50: 94.9 p50: 56.3 p50: 69.7 p50: 73.2 p50: 77.6
p75: 117.5 p75: 75.9 p75: 92.2 p75: 87.3 p75: 96.4
Max: 124 Max: 92.1 Max: 129.0 Max: 116.0 Max: 130.0
Chloride-bromide ratio N/A Min: 1452 Min: 1097 Min: 1536 Min: 1462 Min: 234
p50: 2368 p50: 2317 p50: 2829 p50: 2954 p50: 638
p75: 2987 p75: 2583 p75: 3294 p75: 3242 p75: 1690
Max: 3691 Max: 4572 Max: 4553 Max: 4413 Max: 2476
Nitrate + Nitrate Nitrogen (mg/L) 0.05 # detects: 7/7 # detects: 3/4 # detects: 4/4 # detects: 4/4 # detects: 6/7
Min: 0.3 Min: <MRL Min: 0.35 Min: 1.1 Min: <MRL
p50: 0.62 p50: 0.11 p50: 2.2 p50: 1.8 p50: 0.63
p75: 0.89 p75: 0.44 p75: 4.4 p75: 4.3 p75: 0.93
Max: 1.2 Max: 0.74 Max: 6.3 Max: 6.2 Max: 1.6
Escherichia coli (MPN/100 mL) 1 # detects: 6/6 # detects: 6/6 # detects: 5/6 # detects: 5/6 # detects: 0/6
Min: 2 Min: 52 Min: <MRL Min: <MRL
p50: 20 p50: 845 p50: 4.0 p50: 60.0
p75: 41 p75:4400 p75: 8.0 p75: 1200
Max: 190 Max: >MRL (>24,000) Max: 41.0 Max: 1700
Total coliform (MPN/100 mL) 1 # detects: 5/5 # detects: 5/5 # detects: 5/5 # detects: 5/5 # detects: 4/5
Min: 610 Min: >MRL (>24,000) Min: 180 Min: 2500 Min: <MRL
p50: 13,000 p50: >MRL (>24,000) p50: 990 p50: >MRL (>24,000) p50: 160
p75: 24,000 p75: >MRL (>24,000) p75: 8200 p75: >MRL (>24,000) p75: 640
Max: >MRL (>24,000) Max: >MRL (>24,000) Max: >MRL (>24,000) Max: >MRL (>24,000) Max: 2400
Potassium (mg/L) 0.50 # detects: 8/8 # detects: 8/8 # detects: 8/8 # detects: 8/8 # detects: 8/8
Min: 2.2 Min: 3.9 Min: 3.2 Min: 3.0 Min: 1.2
p50: 4.0 p50: 5.4 p50: 3.5 p50:4.6 p50: 2.7
p75: 4.4 p75: 7.5 p75: 4.7 p75: 5.7 p75: 3.4
Max: 5.5.5 Max: 8.52 Max: 5.4 Max: 8.5 Max: 15.0
Sulfate (mg/L) 0.50 # detects: 8/8 # detects: 5/5 # detects: 5/5 # detects: 5/5 # detects: 7/7
Min: 3.83 Min: 1.6 Min: 4.4 Min: 17.1
Min: 5.1
p50: 9.0 p50: 3.2 p50: 5.9 p50: 6.4 p50: 38.2
p75: 12.7 p75: 6.3 p75: 8.3 p75: 8.9 p75: 113.0
Max: 13.5 Max: 7.5 Max: 16.3 Max: 15.3 Max: 114.0
Boron (mg/L) 20 # detects: 5/8 # detects: 7/7 # detects: 5/8 # detects: 5/8 # detects: 7/7
Min: <MRL Min: 21.9 Min: <MRL Min: <MRL Min: 33.1
p50: 24.9 p50: 28.3 p50: 21.4 p50: 23.4 p50: 123.0
p75: 29.6 p75: 45.1 p75: 25.0 p75: 32.7 p75: 155.0
Max: 30.8 Max: 51.4 Max: 35.8 Max: 81.9 Max: 194.0
a
<MRL = less than method reporting limit. p50 = 50th percentile, p75 = 75th percentile.

3.3. Precipitation 3.4. Water quality parameters

Overall, a positive relation between total microbial sum and precip- Boron, chloride, Cl\\Br ratio, nitrate, potassium, sulfate, E. coli, and
itation was both visually apparent and statistically significant (Fig. 3a). total coliform were assessed in relation to total microbial sum. When
Within sampling locations, higher microbial sum was generally found combining results from all sampling locations, significant correlation
at higher precipitation amounts for the lysimeters only, but relations was found between microbial sum and Cl\\Br ratio (highest correlation
are more difficult to discern due to small sample size (Fig. 3b). coefficient of about 0.42), E. coli (0.26, borderline significant), and sul-
fate (−0.31, negative correlation) (Table IV). Microbial sum by tertiles
of these three general chemistry parameters are shown in Fig. 4; differ-
ences between the tertiles are the most obvious for Cl\\Br ratio. When
Table IV stratified by sample location, no strong or consistent monotonic or lin-
Kendall's Tau correlation coefficients for microbial sum vs. water quality parameters (all
ear relations between total microbial sum and assessed water quality
sampling locations combined).a
parameters were visually apparent in scatterplots (Fig. S1). However,
Parameter # Correlation coefficient (p-value) small sample sizes restricted the ability to visualize location-specific
Boron 38 −0.2093 (p = 0.0717) relations.
Chloride 39 −0.1075 (p = 0.3386)
Chloride-Bromide ratio 38 0.4175 (p = 0.0002)
4. Discussion
Nitrate 26 0.0373 (p = 0.7910)
Potassium 39 0.2000 (p = 0.0750)
Sulfate 30 −0.3120 (p = 0.0165) Stormwater infiltration practices have the potential to promote
Escherichia coli 30 0.2582 (p = 0.0513) recharge to groundwater on the urban landscape, reduce runoff, and
Total Coliform 25 0.2398 (p = 0.1100) limit the travel of pollutants to surface waters and wetland systems
a
Dependent variable = log microbial sum. Sampling location entered as random effect. (Weiss et al., 2008). However, our understanding of the effects of

6
J.R. de Lambert, J.F. Walsh, D.P. Scher et al. Science of the Total Environment 775 (2021) 145738

Fig. 3. Scatterplots of log microbial sum results by precipitation category: a) overall (n = 40) and b) by sampling location (n = 8)a. aDiamond represents mean value. Transparency is set so
that darker data points represent multiple results. K-W test = Kruskal-Wallis test.

infiltration on underlying aquifers is limited. Several factors affect the PMMV, and Cryptosporidium spp. were not detected in the monitoring
potential for groundwater pollution, including land use practices, infra- well at any sampling event. HF183 Bacteroides was detected in the mon-
structure design, and the hydrogeologic setting (Edwards et al., 2016; itoring well at multiple sampling events, but in concentrations much
Nieber et al., 2014). Previous studies have found contamination from lower than those measured at the lysimeters or cistern. The decreases
stormwater infiltration in groundwater and drinking water sources or absences of microbes in the monitoring well relative to the lysimeters
when infrastructure is poorly managed (Edwards et al., 2016). Other overall may be attributable to reductions via transit across the low per-
studies have observed chemical changes in groundwater below infiltra- meability sediment that lies atop the sand in which the monitoring well
tion basins (Datry et al., 2004). The results of this monitoring study may is finished, rather than reductions related to percolation through the
help address data gaps in the understanding of stormwater infiltration sandy vadose zone. However, the consistent decrease in Cryptosporid-
and groundwater interaction, particularly for microbial contaminants. ium spp. detections with lysimeter depth, combined with the fact that
Aside from the sanitary sewer, the highest frequencies of detection protozoa are one and two orders of magnitude larger in size than bacte-
and highest concentrations of microbial targets were observed in the ria and viruses, respectively (Blaschke et al., 2016), indicate that the re-
cistern and three lysimeters, with lower frequencies and concentrations duction may be attributed to movement through the coarse sand of the
of compounds observed in the monitoring well (Table I). Rotavirus, vadose zone. In addition, the groundwater well has a screen that is open

Fig. 4. Vertical boxplots of log microbial sum by chloride-bromide ratio, Escherichia coli (E.coli), and sulfate tertilesa. aK-W test = Kruskal-Wallis test. Diamond = mean, horizontal line =
median.

7
J.R. de Lambert, J.F. Walsh, D.P. Scher et al. Science of the Total Environment 775 (2021) 145738

to a 10-ft profile of the aquifer approximately 3-ft below the water though it is unclear if such leakage occurred in the past or from a dif-
table; samples from the groundwater well may have water recharged ferent source, or if such leakage may have infiltrated the storm water
from the infiltrating stormwater but also may have recharge from piping at some point in space or time. In addition, extrapolation of
non-infiltration gallery areas. Thus, lower microbial concentrations are study results to settings containing active pumping wells would
also likely related to dilution of vertical recharge from the infiltration need to be cautioned, as this study design could not account for influ-
gallery with water recharged from non-infiltration areas. ences caused by well pumping, such as downward vertical hydraulic
Similar to the microbial results, chemical parameters reflective of gradients within a pumping well's zone of influence and mixing of
wastewater were observed in the single samples from the cistern and waters from within its zone of contribution. A further discussion of
Lysimeter 5, but none were detected in the sample from the monitoring the study limitations is included in the SM.
well (Table II). It is likely that as infiltrating stormwater moves down-
ward through the vadose zone, eventually reaching the water table, mi- 5. Conclusions
crobiological contaminants are lagged by relatively slow transit through
the clay-rich sediments that occur between the base of the lysimeters Stormwater infiltration projects have several environmental bene-
and the monitoring well screen. Even then, some presence of the dis- fits, including the capturing of total phosphorus, total suspended solids,
tinctive stormwater chemical parameters would have been expected. heavy metals, petroleum products, and other contaminants before they
However, when the well is sampled stormwater infiltration is expected enter nearby lakes and wetland systems. They can also conserve mil-
to mix with groundwater recharged from non-infiltration gallery areas. lions of gallons of potable water each year by reusing stormwater for ir-
In the process, particulate and dissolved contaminants, including mi- rigation and aquifer recharge. However, the compounds present in
crobes and chemical compounds, are reduced due to dilution, as well as stormwater have the potential to affect surrounding areas when infil-
sorption, predation and dispersion that occurred during travel through trated. Several microbial contaminants were detected throughout the
the vadose zone. study period at multiple sampling points, with the lowest number of de-
Microbial concentrations varied across sampling events and differ- tections and microbial concentrations observed in the monitoring well.
ences in concentration by precipitation amount indicate that rainfall Wastewater indicators were detected in samples from the cistern and
events are one driver in this variability. It has been observed that low from Lysimeter 5. A positive relation between microbial sum and pre-
ionic strength water, typical of rainfall, may enhance virus transport cipitation was found to be statistically significant (Fig. 3a). Additionally,
(Hunt and Johnson, 2016). Larger studies examining the role of precipita- a significant correlation was observed between microbial sum and
tion and recharge events on microbial contamination of both stormwater Cl\\Br ratio. The results of this study indicate that subsurface in compa-
and groundwater could be conducted to further assess this connection. In rably sensitive geologic settings may be susceptible to contamination by
addition, a more in-depth analysis of the travel of contaminants through microbes and other wastewater-related contaminants when in proxim-
the vadose zone will be necessary in understanding the complete picture ity to subsurface stormwater infiltration and especially during periods
of fate and transport in these systems. Because extensive microbial sam- of substantial recharge.
pling is time consuming and expensive to implement, we also examined In contrast, aquifers overlain by fine-grained sediments offering
whether water quality parameters that are frequently part of routine moderately protective geology are expected to be less susceptible
analysis may serve as predictors of microbial contamination. Although to these contaminants. However, contaminants still have the po-
sparse data created limitations, we found a significant, positive relation tential to reach these aquifers, perhaps at a longer time of travel
between high Cl\\Br ratio and microbial load (Table IV). This may reflect and in reduced concentrations. For example, John and Rose
on the relative contribution of different sources of salinity within the (2005) described inactivation rates for a variety of microbes in
storm watershed. Ratios in excess of 250–300 are reflective of human- the 0.02–0.1 log 10day−1 range, indicating months may be required
caused sources, such as human or animal waste, water softener effluent to inactivate these constituents in some groundwater settings. As
and road salt, with higher ratios like those observed at the cistern and ly- geologic protection increases, the time it takes water to travel
simeters likely related to increasing amounts from the latter two sources from the stormwater discharge point to the aquifer is expected to
(Mullaney et al., 2009). This relation and other potential relations be- increase, therefore reducing the risk of contamination when com-
tween microbial concentrations and standard water quality parameters pared to shallower, more vulnerable aquifers.
could be evaluated further in future studies with larger sample sizes In sum, when in proximity to subsurface stormwater infiltra-
and higher sampling frequencies to determine whether any may serve tion, geologically sensitive aquifers may be susceptible to contam-
as a reasonable indicator of microbial contamination risk from urban ination by the same microbes and wastewater compounds found in
stormwater, although Hunt et al. (2010) showed that the extreme sensi- storm water, particularly following rain events or an increase in re-
tivity of the pathogen analytical methods generally makes them superior charge. Further, Cl\\Br ratio may be a useful indicator of microbial
tracers. load, in the absence of more advanced analytical technologies. The
There are important study limitations. The sanitary sewer line presence of low permeability sediments above an aquifer can be an
present through the site presents an obstacle to definitive conclu- important barrier for reducing contaminant loads. These results
sions about the source of microbiological contamination. Many of may help inform siting decisions relative to urban storm water in-
the contaminants found in urban stormwater are identical to those frastructure and nearby drinking water supply sources and their
found in human sewage (Fairbairn et al., 2018). Without further in- protection areas, although the water quality effects that might be
vestigation, it cannot be determined with certainty whether the san- altered by the pumping effects of water supply wells could not be
itary sewer is leaking and infiltrating contaminants to nearby evaluated in this study design.
groundwater, although a recent video inspection of the line showed
no obvious flaws or leaks (City of Roseville, written communication, CRediT authorship contribution statement
2016). Moreover, the fact that three microbial contaminants (en-
terovirus, human polyomavirus, and norovirus genogroup II) were Jane R. de Lambert: Conceptualization, Methodology, Investiga-
detected in high concentrations in the sanitary sewer but not any tion, Writing – original draft. James F. Walsh: Conceptualization,
other sampling points indicates the sewer was not leaking. The Supervision, Writing – original draft. Deanna P. Scher: Formal anal-
Cl\\Br ratios observed in samples from the monitoring well largely ysis, Visualization, Writing – original draft. Aaron D. Firnstahl:
fell in the range of waters affected by human or animal waste Methodology, Investigation, Writing – review & editing. Mark A.
(Mullaney et al., 2009). This indicates that leaking sewers may Borchardt: Conceptualization, Validation, Resources, Writing –
have generally affected groundwater in the vicinity at one time, review & editing.

8
J.R. de Lambert, J.F. Walsh, D.P. Scher et al. Science of the Total Environment 775 (2021) 145738

Declaration of competing interest Hunt, R.J., Borchardt, M.A., Richards, K.D., Spencer, S.K., 2010. Assessment of sewer source
contamination of drinking water wells using tracers and human enteric viruses. En-
viron. Sci. Technol. 44 (20), 7956–7963. https://doi.org/10.1021/es100698m.
The authors declare that they have no known competing financial Interagency Workgroup on Water Reuse. 2018. Advancing Safe and Sustainable Water
interests or personal relationships that could have appeared to influ- Reuse in Minnesota. Retrieved July 16, 2020, from https://www.health.state.mn.us/
communities/environment/water/docs/cwf/2018report.pdf.
ence the work reported in this paper.
John, D.E., Rose, J.B., 2005. Review of factors affecting microbial survival in groundwater.
Environ. Sci. Technol. 39 (19), 7345–7356. https://doi.org/10.1021/es047995w.
Acknowledgements Lambertini, E., Spencer, S.K., Bertz, P.D., Loge, F.J., Kieke, B.A., Borchardt, M.A., 2008. Con-
centration of enteroviruses, adenoviruses, and noroviruses from drinking water by
use of glass wool filters. Appl. Environ. Microbiol. 74, 2990–2996.
This work was supported by the Clean Water Fund of the Minnesota
Meyer, G.N., Swanson, L., 1992. C-07 Geologic Atlas of Ramsey County, Minnesota. Minne-
Clean Water, Land, and Legacy Amendment. The authors thank Capitol sota Geological Survey Retrieved from the University of Minnesota Digital Conser-
Region Watershed District, especially Britta Belden, Forrest Kelley, and vancy. Available:. http://hdl.handle.net/11299/58233.
Sarah Wein, and the city of Roseville for their cooperation and partner- Minnesota DNR (Department of Natural Resources). 1991. Criteria and Guidelines for
Assessing Geologic Sensitivity of Ground Water Resources in Minnesota. Accessed
ship on this project, and Walter Eshenauer of SRF Consulting Group, Inc. 03/04/20, Available: https://files.dnr.state.mn.us/waters/groundwater_section/map-
for bringing this project to our attention and facilitating its creation. We ping/sensitivity/docs/assessing_geologic_sensitivity.pdf.
also wish to thank the Minnesota Pollution Control Agency for agreeing Mullaney JR, Lorenz DL, Arntson AD. 2009. Chloride in groundwater and surface water in
areas underlain by the glacial aquifer system, northern United States: U.S. Geological
to install the monitoring well at this site and for assisting with sample Survey Scientific Investigations Report 2009–5086, (41 p).
collection during the study and sharing of analytical data. Finally, thanks NAS (National Academies of Sciences, Engineering, and Medicine), 2016. Using Graywater
to Luke Pickman for his help with figure creation. Any use of trade, firm, and Stormwater to Enhance Local Water Supplies: An Assessment of Risks, Costs, and
or product names is for descriptive purposes only and does not imply Benefits. The National Academies Press, Washington, DC https://doi.org/10.17226/
21866.
endorsement by the U.S. Government. Nieber JL, Arika C, Lahti L, Gulliver JS, Weiss PT. 2014. The Impact of Stormwater Infiltra-
tion Practices on Groundwater Quality. St. Anthony Falls Laboratory. Retrieved from
Appendix A. Supplementary data the University of Minnesota Digital Conservancy, http://hdl.handle.net/11299/
169456.
Sauer, E.P., Vandewalle, J.L., Bootsma, M.J., McLellan, S.L., 2011. Detection of the human
Supplementary data to this article can be found online at https://doi. specific Bacteroides genetic marker provides evidence of widespread sewage con-
org/10.1016/j.scitotenv.2021.145738. tamination of stormwater in the urban environment. Water Res. 14, 4081–4091.
https://doi.org/10.1016/j.watres.2011.04.049.
Sercu, B., Van De Werfhorst, L.C., Murray, J., Holden, P.A., 2009. Storm drains are sources of
References human fecal pollution during dry weather in three urban southern California water-
sheds. Environ. Sci. Technol. 43 (2), 293–298.
American Engineering and Testing. 2013. Lake McCarrons Subwatershed BMP Clean
Sidhu, J.P., Hodgers, L., Ahmed, W., Chong, M.N., Toze, S., 2012. Prevalence of human path-
Water Partnership Project, Villa Park, Dale Street and County Road B, Roseville, Min-
ogens and indicators in stormwater runoff in Brisbane, Australia. Water Res. 46 (20),
nesota. AET Project No. 03-04730.
6652–6660. https://doi.org/10.1016/j.watres.2012.03.012.
Blaschke, A.P., Derx, J., Zessner, M., Kirnbauer, R., Kavka, G., Strelec, H., Farnleitner, A.H.,
Sidhu, J.P., Ahmed, W., Gernjak, W., Aryal, R., McCarthy, D., Palmer, A., Kolotelo, P., Toze, S.,
Pang, L., 2016. Setback distances between small biological wastewater treatment sys-
2013. Sewage pollution in urban stormwater runoff as evident from the widespread
tems and drinking water wells against virus contamination in alluvial aquifers. Sci.
presence of multiple microbial and chemical source tracking markers. Sci. Total Envi-
Total Environ. 573, 278–289. https://doi.org/10.1016/j.scitotenv.2016.08.075 (Epub
ron. https://doi.org/10.1016/j.scitotenv.2013.06.020 463-464:488-96.
2016 Aug 26).
Capitol Region Watershed District. 2017. Upper Villa Exfiltration Monitoring Standard Smith, C.M., Hill, V.R., 2009. Dead-end hollow-fiber ultrafiltration for recovery of diverse
Operating Procedure. Internal report. microbes from water. Appl. Environ. Microbiol. 75, 5284–5289.
Datry, T., Malard, F., Gibert, J., 2004. Dynamics of solutes and dissolved oxygen in shallow Stokdyk, J.P., Firnstahl, A.D., Spencer, S.K., Burch, T.R., Borchardt, M.A., 2016. Determining
urban groundwater below a stormwater infiltration basin. Sci. Total Environ. 329 the 95% limit of detection for waterborne pathogen analyses from primary concentra-
(1–3), 215–229. https://doi.org/10.1016/j.scitotenv.2004.02.022. tion to qPCR. Water Res. 96, 105–113.
Edwards, E.C., Harter, T., Fogg, G.E., Washburn, B., Hamad, H., 2016. Assessing the effec- Tiefenthaler, L., Stein, E.D., Schiff, K., 2008. Origins and mechanisms of watershed and land
tiveness of drywells as tools for stormwater management and aquifer recharge and use based sources of fecal indicator bacteria in urban stormwater. Southern California
their groundwater contamination potential. J. Hydrol. 539, 539–553. https://doi.org/ Coastal Water Research Project 2008 Annual Report. Costa Mesa, CA, pp. 153–161.
10.1016/j.jhydrol.2016.05.059. USEPA (U.S. Environmental Protection Agency). 1986. Ambient Water Quality for Bacte-
Fairbairn, D.J., Elliott, S.M., Kiesling, R.L., Schoenfuss, H.L., Ferrey, M.L., Westerhoff, B.M., ria, 1986. Retrieved from https://www.regulations.gov/document?D=EPA-HQ-OW-
2018. Contaminants of emerging concern in urban stormwater: Spatiotemporal pat- 2007-0808-0001. (Accessed September 17, 2016).
terns and removal by iron-enhanced sand filters (IESFs). Water Res. 145, 332–345. USEPA (U.S. Environmental Protection Agency). 2019. December 18. Water Reuse Re-
Gibson, K.E., Schwab, K.J., Spencer, S.K., Borchardt, M.A., 2012. Measuring and mitigating search. Retrieved July 16, 2020, from https://www.epa.gov/water-research/water-
inhibition during quantitative real time PCR analysis of viral nucleic acid extracts reuse-research.
from large-volume environmental water samples. Water Res. 46 (13), 4281–4291. Weiss, P.T., LeFevre, G., Gulliver, J.S., 2008. Contamination of Soil and Groundwater Due to
Hunt, R.J., Johnson, W.P., 2016. Pathogen transport in groundwater systems: contrasts Stormwater Infiltration Practices, A Literature Review. St. Anthony Falls Laboratory
with traditional solute transport. Hydrogeol. J. 25, 921–930 (2017). https://doi.org/ Retrieved from the University of Minnesota Digital Conservancy. http://hdl.handle.
10.1007/s10040-016-1502-z. net/11299/115341.

You might also like