Article pubs.acs.org/est
Dramatic Improvements in Beach Water Quality Following Gull Removal Reagan R. Converse,*,† Julie L. Kinzelman,‡ Elizabeth A. Sams,† Edward Hudgens,† Alfred P. Dufour,§ Hodon Ryu,§ Jorge W. Santo-Domingo,§ Catherine A. Kelty,§ Orin C. Shanks,§ Shawn D. Siefring,§ Richard A. Haugland,§ and Timothy J. Wade† †
U.S. Environmental Protection Agency, 104 Mason Farm Rd., Chapel Hill, North Carolina 27514, United States City of Racine Health Department, 730 Washington Ave., Racine, Wisconsin 53403, United States § U.S. Environmental Protection Agency, 26 West Martin Luther King Dr., Cincinnati, Ohio 45268, United States ‡
S Supporting Information *
ABSTRACT: Gulls are often cited as important contributors of fecal contamination to surface waters, and some recreational beaches have used gull control measures to improve microbial water quality. In this study, gulls were chased from a Lake Michigan beach using specially trained dogs, and water quality improvements were quantified. Fecal indicator bacteria and potentially pathogenic bacteria were measured before and during gull control using culture methods and quantitative polymerase chain reaction (qPCR). Harassment by dogs was an effective method of gull control: average daily gull populations fell from 665 before to 17 during intervention; and a significant reduction in the density of a gull-associated marker was observed (p < 0.001). Enterococcus spp. and Escherichia coli densities were also significantly reduced during gull control (p < 0.001 and p = 0.012, respectively for culture methods; p = 0.012 and p = 0.034, respectively for qPCR). Linear regression results indicate that a 50% reduction in gulls was associated with a 38% and 29% decrease in Enterococcus spp. and E. coli densities, respectively. Potentially human pathogenic bacteria were detected on 64% of days prior to gull control and absent during gull intervention, a significant reduction (p = 0.005). This study demonstrates that gull removal can be a highly successful beach remedial action to improve microbial water quality.
■
INTRODUCTION Fecal contamination of beaches results in numerous health and aesthetic concerns for recreators, and beach advisories posted when fecal pollution is detected can result in local daily economic losses of up to $100 000.1 Identification of the dominant source(s) of fecal pollution is a first step toward improving microbial water quality and reducing beach advisories. Gulls (Larus spp.) are often cited as an important source of fecal contamination to surface waters and beach sand,2−8 and conservative estimates show that a single gull can shed 4.8 × 109 Escherichia coli (E. coli) cells and 2.1 × 108 Enterococcus spp. cells daily.9−11 As a result, fecal indicator bacteria (FIB) densities in surface waters are significantly and positively correlated with gull populations,12,13 and water quality criteria exceedances are more common at beaches with large gull communities.14,15 Human pathogens have also been detected in gull fecal samples.16−20 Salmonella spp. infections have been reported in nearly 9% of gulls and Campylobacter spp. infections in nearly 16%.16 E. coli O157 has also been identified in the feces of gulls that frequent landfills.17 To date, there is no direct © 2012 American Chemical Society
evidence of transmission of these pathogens from gulls to swimmers at beach sites, and the risk from gull contamination to beachgoer health is poorly understood. With gull populations increasing dramatically in urban areas,21 gull population control measures have been undertaken at some beaches and gull feeding and roosting sites in hopes of reducing beach closures and potentially protecting beachgoer health.15,22 These measures include projection of gull distress calls, shotgun noises, use of monofilament nets, introduction of predators such as falcons, and pyrotechnics.23 Although the efficacy of each technique varies, the introduction of gull control measures has coincided with FIB reductions at an inland reservoir.22 However, it was unclear whether the removal of gulls had a statistically significant effect on FIB densities or the presence of potential pathogens. Received: Revised: Accepted: Published: 10206
June 11, 2012 August 21, 2012 August 22, 2012 August 22, 2012 dx.doi.org/10.1021/es302306b | Environ. Sci. Technol. 2012, 46, 10206−10213
Environmental Science & Technology
Article
fecal material was collected and appropriately disposed of immediately after evacuation. Bird Counts. On each water quality sampling day, birds were counted four times a day (at 9:00, 11:30, 15:00, and 17:00) along three sampling transects previously established by the City of Racine Health Department.28 Birds on the ground, in the air, or on the water within 20 m of the sampling transects were counted and classified as gulls, shorebirds, or other birds. These data are presented in Table S1 of the Supporting Information, SI. For statistical analyses, bird counts over the four daily sampling times were averaged as an estimate of daily bird populations. Weather and Hydrologic Conditions. Hydrometeorological conditions, including air and water temperature, ultraviolet irradiation, precipitation, total suspended solids, and wind speed and direction were recorded twice daily. There was little precipitation during the study period (1.68 cm total with less than 0.58 cm occurring in a 24 h period). The mean air temperature was 24.3 °C (standard error = 0.39), and the mean water temperature was 21.2 °C (standard error = 0.474). All hydrometerological data are presented in Table S1 of the SI, along with twice daily counts of beachgoers and bathers. Water Sampling. Sampling was conducted between 9:00 and 9:30 a.m. each day. Three 3-L water samples were collected at 1 m depth from each sampling transect. Samples were stored on ice and transported to the City of Racine Health Department Lab, where they were composited into a single large sample. Sample processing occurred within 1 h of collection. Culture-Based Detection of FIB and Pathogens. Culturable Enterococcus spp. and E. coli were enumerated by membrane filtration (MF) following U.S. EPA Methods 1600 and 1603, respectively.29,30 One-hundred milliliter, 10 mL, and 1 mL sample volumes were analyzed. To detect and enumerate Campylobacter spp., 1 L of water was passed through a 47-mm diameter, 0.6 μm pore-size filter (GE Osmonics, Trevose, IL). Filters were placed on 7% horseblood agar plates (Oxoid Tryptose Blood Agar Base, Basingstoke, Hampshire, U.K.) and incubated at 42 °C for 48 h in a microaerobic atmosphere (Oxoid BR0038 pack). Presumptive Campylobacter colonies were tested for the presence of catalase activity and subjected to hippuratehydrolysis, indoxyl acetate, and nitrate reduction tests.31 To detect E. coli O157, water samples were incubated at 35 °C in Presence-Absence broth (Hardy Diagnostics, Santa Maria, CA) for 18−20 h. Growth from the enrichment culture was streaked on Rainbow agar (Biolog Inc., Hayward, CA) with 0.8 mg/L potassium tellurite and 10 mg/L novobiocin. Plates were incubated at 35 °C for 20−24 h. Presumptive E. coli O157 colonies were black or gray32 and were streaked on trypticase soy agar with yeast extract (TSAYE) agar (Hardy Diagnostics) and grown at 35 °C for 18−24 h. Biochemical confirmation involved the spot indole, Levine’s Eosin-Methylene Blue (LEMB) with 4-methylumbelliferyl-β-D-glucoronide (MUG), and sorbitol fermentation tests.33 To detect Salmonella and Shigella, water samples (1 L, 100 and 10 mL) were filtered onto 2.5 g sterile diatomaceous earth (DE). Fifty milliliters of selenite-cystine broth (Remel, Lenexa, KS) was applied to the DE, and the sample was agitated. The broth was incubated for 24 h at 42 °C.20 Subsamples were streaked onto xylose lysine deoxycholate (XLD) agar (Hardy Diagnostics), and plates were incubated for 24 h at 35 °C. Presumptive Salmonella colonies (red with black centers) and
In this study, we sought to better estimate and quantify the effect of gull removal on water quality. We characterized changes in FIB and potential human pathogen concentrations following removal of a large gull population from a recreational beach. FIB (Enterococcus and E. coli) were measured using both cultivation and qPCR assays. QPCR was evaluated in addition to culture methods because studies have recently evaluated qPCR methods in the context of beach monitoring and notification.24 Fecal source identification approaches were used to identify or rule-out other potential sources of FIB and estimate contributions from gulls. We also used a combination of culture and qPCR methods to measure pathogens, focusing on those that have been previously detected in gull feces at the study beach.20
■
MATERIALS AND METHODS Study Site. The study was conducted at North Beach in Racine, Wisconsin (N42° 44′, W087° 46′), a popular recreational beach stretching approximately 960 m. Prior to 2005, water quality advisories were posted at North Beach for approximately 27% of the swimming season, and many of these advisories occurred during dry weather and when no point sources of pollution could be identified.25 The implementation of improved beach management practices including ecologically appropriate beach modifications, deeper beach grooming, and the redesign of stormwater outfalls has produced significant water quality improvements.26 This has resulted in a reduction in water quality advisories. However, they still occur on an annual basis, and gulls are hypothesized to be an important cause.25 Experimental Design. FIB, fecal source identification genetic markers, and pathogens were measured in beach water from August 1 to August 11, 2011 (11 days). On August 12, specially trained dogs were introduced to chase gulls from the beach. Gull control continued for 16 days. After one week of gull control, microbial water quality sampling resumed until August 27 (9 days). Gull Control. For a period of 16 days (August 12 through 27, 2011), birds were chased from the beach using trained border collies (Wild Goose Chase Inc., LaGrange, IL). We selected this method of gull control because these specially trained dogs have successfully excluded gulls from office parks, community parks, colleges, and another Lake Michigan beach for entire swimming seasons.15 Furthermore, chasing by dogs was less likely to disturb beachgoers than other methods of bird control (shotgun blasts, pyrotechnics, distress call projection) and did not require permanent modification to the beach or existing infrastructure. Chasing also allowed us to target gulls more specifically than measures such as monofilament wires and nets, reducing disturbance of other bird species. Dog handlers were trained to identify birds protected by the Migratory Bird Treaty Act,27 and care was taken to avoid chasing protected species. Before sunrise (approximately 5:45) until after dark (approximately 20:00), one or two border collies with a handler chased ring-billed and herring gulls (Larus delawarensis and smithsonianus, respectively) on the sand and in the air. Beachgoers observed hand-feeding gulls were approached and given informational cards describing the potential negative impact of gulls on water quality. After dark, laser sweeps were made across the beach and water to discourage gulls from roosting on the beach overnight. The dogs were under constant supervision by handlers. To prevent further fecal contamination to the study beach, dog 10207
dx.doi.org/10.1021/es302306b | Environ. Sci. Technol. 2012, 46, 10206−10213
Environmental Science & Technology
Article
Shigella colonies (red) were inoculated into triple-sugar-iron (TSI) slants and streaked onto urea slants.16 Slants were incubated for 24 h at 35 °C. When urea slants were positive (pink), growth from TSI positive slants (red slant with blackening) were subject to a Salmonella slide-agglutination test using Salmonella O antisera for confirmation. When urea slants were negative (yellow), growth from TSI positive slants were subject to a Shigella slide-agglutination test using Shigella O antisera. Identification of presumed pathogenic bacteria was confirmed using API-20E Bacterial Identification Test Strips (bioMerieux, Marcy l’Etoile, France) following manufacturer instructions. Quantitative Real-Time Polymerase Chain Reaction (qPCR). Water samples (100 mL) were filtered onto polycarbonate membranes (47 mm diameter, 0.45 μm poresize; Millipore, Billerica, MA) and stored at −80 °C until processed (up to six months). Frozen filters were transferred to 2 mL semiconical screw-cap tubes loaded with 0.3 g of 0.1 mm zirconia/silica beads (BioSpec, Bartlesville, OK). Ten microliters of 10 μg/mL salmon testes DNA (Sigma, St Louis, MO), which served as a specimen processing control (SPC), and 490 μL of AE Buffer (Qiagen, Valencia, CA) were added to each sample, and tubes were bead-beaten for 2 min at full speed (BioSpec). After 1 min of centrifugation at 12 000 × g, supernatants were transferred to 1.7 mL microtubes and centrifuged again at 12 000 × g for 5 min. Supernatant was transferred to sterile 1.7 mL microtubes, and DNA was extracted using DNA-EZ RW01 kits (GeneRite, New Brunswick, NJ), following manufacturer instructions. DNA extracts were stored at −20 °C until further processing (less than 4 weeks). QPCR was used to measure FIB and pathogen concentrations and to identify or exclude potential sources of pathogens. To that end, human-associated Bacteroidales (HF18334), a marker of gull fecal pollution (Gull2 targeting Catellicoccus marimammalium35), and two cow-associated markers (Cow-M2 and Cow-M336) were measured. The cowassociated markers were included because cows are the most populous livestock in the watershed.37 Dog-specific markers were not included because dog handlers involved in this study immediately collected all border collie waste and dogs are not normally allowed on the beach. Additionally, qPCR was used to measure presumptive Campylobacter jejuni, coli, and lari, which are the most common human pathogens among Campylobacter spp. We included this assay because Campylobacter spp. are difficult to culture from environmental samples.38,39 TaqMan qPCR assays were used to detect Enterococcus spp. (Entero40), E. coli (E. coli42), Cow-M2 and Cow-M3, Campylobacter spp.(Campy), and the SPC (Sketa2241). With the exception of the Campylobacter assay, all assays were performed as described in the original references. For the Campylobacter assay, reaction volumes were reduced to 25 μL, primer concentrations were increased from 300 nM to 1 μM, and 5 μL of DNA template was used. SYBR green-based qPCR assays for Gull2 and HF183 were also used as previously described.34,35 For SYBR green-based qPCR assays, disassociation curves were examined to determine the presence of potential primer-dimer and other nonspecific reaction products. Signal intensity values were recorded for those reactions with one corresponding amplification peak within the disassociation curves. All reactions were performed on a SmartCycler II (Cepheid, Sunnyvale, CA), a StepOnePlus (Applied Biosys-
tems, Carlsbad, CA), or a 7900 HT Fast Real-Time Sequence Detector (Applied Biosystems, Carlsbad, CA). Calibration standards were created by growing E. faecalis (American Type Culture Collection 29212), E. coli (ATCC 25922), and C. jejuni (ATCC 35920) as previously described.34,42 Cell densities were microscopically counted following Noble and Fuhrman.43 One-hundred thousand cells were filtered onto polycarbonate filters, and DNA was extracted along with water samples as described above. Plasmid standards were created as previously described34−36 and used for C. marimammalium, human-associated Bacteroidales, and the cowassociated markers. Standard curves for each qPCR assay consisted of the calibration or plasmid standard and four 10-fold serial dilutions. The standard curves were run in duplicate or triplicate, and all samples, filtration blanks, and no-template controls were run in duplicate. Percent amplification efficiency (E) was derived from respective calibration curves: E = (10−slope − 1) × 100. E, R2 values, and ranges of quantification are given in Table 1. For E. Table 1. QPCR Amplification Efficiencies, Standard Curve R2 Values, And Ranges of Quantification assay
target
cow M2
cow-associated Bacteroidales cow-associated Bacteroidales Campylobacter E. coli Enterococcus C. marimammalium
cow M3 campy E. coli entero gull2 HF183 sketa22
human-specific Bacteroidales salmon testes DNA
amplification efficiency (%)
R2
98
0.99
90
0.99
86 97 105 96
0.99 0.97 0.99 0.99
91
0.99
107
0.99
range of quantification 10 to 105 copies 10 to 105 copies 50 to 105 CCE 10 to 105 CCE 10 to 105 CCE 10 to 106 copies 10 to 105 copies NA
coli, Enterococcus, and Campylobacter spp. assays, calibrator cell equivalents (CCE) were calculated using the comparative threshold cycle (ΔΔCT) method described in Haugland et al.40 Results from other assays were calculated by fitting sample cycle threshold (CT) values to the standard curves, giving copies per reaction. Copies per qPCR reaction were adjusted to copies per 100 mL by multiplying the standard curve result by the total volume of DNA extract divided by the volume of DNA extract used in the reaction. For pathogen and source-tracking assays, all samples yielding CT values below the lowest dilution tested on a respective calibration curve were designated as detected but below the limit of quantification. For statistical analyses of FIB results, qPCR nondetects were assigned a value of 5 CCE/100 mL. No samples showed PCR inhibition, defined as a 3 CT delay in the SPC assay. Data Analyses and Statistical Methods. Statistical analyses were performed using SigmaPlot 11.0 (San Jose, CA) and SAS 9.2 (Cary, NC), and results and relationships were considered significant when p < 0.05. Because FIB densities and gull populations were highly skewed, data were log10 transformed. Although this data reduced skewness in all cases, some data remained non-normally distributed following transformation. Differences in FIB and fecal source identification marker densities and bird populations before and during intervention were calculated using the Mann−Whitney U test.44 Binomial logistic regression models were used to 10208
dx.doi.org/10.1021/es302306b | Environ. Sci. Technol. 2012, 46, 10206−10213
Environmental Science & Technology
Article
Figure 1. Log10 FIB densities before and during gull control. Solid lines represent medians, dotted lines represent geometric means, box boundaries indicate the first and third quartiles, and whiskers show the 95% confidence intervals. Samples outside this range are indicated by dots.
gulls. Gull harassment resulted in a 98% reduction (p < 0.001) in the bird population. FIB Densities. E. coli and Enterococcus spp. densities in beach water were significantly reduced during the period of gull control (Figure 1). This improvement is underscored by the reduction in water quality exceedances for FIB. Before gull harassment, 18% of days (2 of 11) had culturable E. coli densities surpassing Wisconsin’s single sample advisory limit of 235 CFU/100 mL, and 36% of days (4 of 11) had culturable Enterococcus spp. densities above the single sample advisory limit of 61 CFU/100 mL. During the gull harassment sampling period, neither densities of Enterococcus nor E. coli exceeded the culturable single sample advisory limits. Linear regressions between log10 transformed culturable FIB and log10 average bird populations were highly significant (Figure 2). Weaker, although still significant, relationships were observed between log10 average bird populations and log10 FIB qPCR results. Slopes of these relationships were slightly higher for MF results than qPCR results (Table 2), and the corresponding estimated percent reductions in FIB densities given reductions in the gull population were greater for MF results (Table 3). No significant relationships were identified between FIB densities and hydrometeorologic conditions. Fecal Source Identification Markers. The gull-associated marker, C. marimammalium, was detected in high densities throughout the study period (Figure 3), but a significant
describe the relationships between pathogen detection and presence of human fecal contamination and average gull populations. The associations between bird populations and FIB were modeled using linear regression with log10 FIB densities as the dependent variable and log10 gull populations as the independent variable, such that log y = a + (b × log x)
where y represents FIB concentrations, x represents gull counts, and a and b are constants. From the linear regression parameters, we calculated the percent change in FIB relative to the percent change in gull counts by making use of the following equation: %change in y = %change in x
Δy y Δx x
=
Δy x × =b y Δx
This method is commonly used in other applications.45
■
RESULTS Bird Populations. For the eleven days preceding the application of gull control measures, an arithmetic mean of 665 birds (standard error (SE) = 112.79) were on the beach throughout each day, of which 99% were gulls. After one week of gull harassment, an arithmetic mean of 17 birds (SE = 6.7) were on the beach throughout each day, 94% of which were 10209
dx.doi.org/10.1021/es302306b | Environ. Sci. Technol. 2012, 46, 10206−10213
Environmental Science & Technology
Article
Table 3. Reductions in FIB Concentrations Corresponding to Reductions in the Gull Population % reduction in Gulls
% reduction in Enterococcus MF
% reduction in Enterococcus qPCR
% reduction in E. coli MF
% reduction in E. coli QPCR
10 25 50 75 90
6.9 17.8 37.6 61.0 79.1
4.5 11.9 26.3 45.7 63.7
5.1 13.4 29.3 50.0 68.4
4.3 11.4 25.3 44.1 62.0
Figure 3. Log10 C. marimammalium densities before and during gull control. Solid lines represent medians, dotted lines represent geometric means, box boundaries indicate the first and third quartiles, and whiskers show the 95% confidence intervals. Samples outside this range are indicated by dots.
reduction was observed during the period of gull harassment (p < 0.001). Linear regression between log10 C. marimammalium marker densities and log10 average bird counts produced an R2 value of 0.58, p < 0.001, and a slope of 0.65. Human-associated Bacteroidales markers were detected on four days before the onset of gull harassment, but densities were below the limit of quantification. During gull harassment, human-associated fecal markers were not detected. Bovine-associated Bacteroidales were not detected in any water samples. Pathogens. Potentially pathogenic bacteria were detected before (64% of sampling days, 7 of 11 days) but not during gull harassment, a significant reduction (p = 0.005, Fisher’s exact test). Presumptive E. coli O157 was detected in beach water twice (5 colonies isolated each time) and did not coincide with single sample water quality exceedances for E. coli or Enterococcus spp. or detection of the putative human marker, HF183 (Table 4). Culturable Salmonella spp. (5 colonies isolated) was detected once prior to gull harassment on a day with E. coli and Enterococcus spp. densities above their single sample standards. We did not identify the species of Salmonella detected, so we do not know whether the isolates were pathogenic. Culture methods did not detect Campylobacter spp. during the entire study period. However, presumptive C. jejuni, C. coli, and/or C. lari were detected by qPCR on 55% (6 of 11) of sampling days before gull harassment, although densities were below the limit of quantification on one of these days (Table 4). No Campylobacter were detected by qPCR during gull harassment. However, it should be noted that the volume of water used for qPCR analyses was low, and with such low volumes, it is possible that Campylobacter spp. were still present
Figure 2. Linear regressions between log10 indicator densities and log10 average bird counts. Dots represent MF results; triangles represent qPCR results. Solid lines show relationships between culture results and gull counts; dashed lines show relationships between qPCR results and gull counts. Panel A shows Enterococcus spp. results, panel B shows E. coli results, and panel C shows C. marimammalium results.
Table 2. Slopes and Intercepts of Linear Regressions between log10 Average Gull Populations and log10 FIB Densities (MF = Membrane Filtration) dependent variable
slope
intercept
R2
p-value
Enterococcus MF E. coli MF Enterococcus QPCR E. coli QPCR
0.682 0.498 0.440 0.418
−0.681 −0.097 2.018 0.266
0.71 0.48 0.33 0.41
