Longitudinal Poisson Regression To Evaluate the Epidemiology of Cryptosporidium, Giardia, and Fecal Indicator Bacteria in Coastal California Wetlands

Abstract

Fecal pathogen contamination of watersheds worldwide is increasingly recognized, and natural wetlands may have an important role in mitigating fecal pathogen pollution flowing downstream. Given that waterborne protozoa, such as Cryptosporidium and Giardia, are transported within surface waters, this study evaluated associations between fecal protozoa and various wetland-specific and environmental risk factors. This study focused on three distinct coastal California wetlands: (i) a tidally influenced slough bordered by urban and agricultural areas, (ii) a seasonal wetland adjacent to a dairy, and (iii) a constructed wetland that receives agricultural runoff. Wetland type, seasonality, rainfall, and various water quality parameters were evaluated using longitudinal Poisson regression to model effects on concentrations of protozoa and indicator bacteria (Escherichia coli and total coliform). Among wetland types, the dairy wetland exhibited the highest protozoal and bacterial concentrations, and despite significant reductions in microbe concentrations, the wetland could still be seen to influence water quality in the downstream tidal wetland. Additionally, recent rainfall events were associated with higher protozoal and bacterial counts in wetland water samples across all wetland types. Notably, detection of E. coli concentrations greater than a 400 most probable number (MPN) per 100 ml was associated with higher Cryptosporidium oocyst and Giardia cyst concentrations. These findings show that natural wetlands draining agricultural and livestock operation runoff into human-utilized waterways should be considered potential sources of pathogens and that wetlands can be instrumental in reducing pathogen loads to downstream waters.

INTRODUCTION

Fecal pathogen contamination of waterways is an important public health consideration in multiuse ecosystems. Sources of pathogens may include agricultural runoff containing livestock feces, sewage outfalls containing human and pet feces, and storm runoff containing feces of domestic and wild animals (12, 36). Zoonotic waterborne protozoa, specifically Cryptosporidium and Giardia spp., commonly cause diarrheal disease that can be severe or even fatal in immunocompromised individuals (1, 8, 11, 26). As surface waters move down a watershed, pulses of pathogen contamination from upstream sources contribute to downstream loads that affect recreational uses. Additionally, environmental persistence of Cryptosporidium oocysts and Giardia cysts, and their low infectious dose, contribute to the health risks posed to humans and animals who are in contact with surface waters (9, 13, 28).

The importance of wetlands for reducing select agricultural and chemical pollutants and pathogens in impaired surface water has been previously studied (21) and supports wetland use for improving water quality downstream from agricultural lands, livestock operations, and urban communities. However, most studies to date have focused on wetlands receiving sewage effluent (7, 18, 27), and little is known about how natural wetlands function with respect to transport and persistence of Cryptosporidium oocysts and Giardia cysts from upstream sources. In California and elsewhere, identifying ecologically sustainable management practices to improve water quality is of particular interest to resource managers and agencies.

To evaluate the distribution and transport of protozoa and fecal indicator bacteria (FIB) in natural wetlands, microbial concentrations among three wetland types along the central California coast were compared: (i) a medium-sized, tidally influenced slough that connects urban and agricultural runoff to Monterey Bay, (ii) a small seasonal wetland adjacent to a dairy that receives runoff from cattle, and (iii) a small, unidirectionally constructed wetland that filters water from an adjacent slough dominated by agricultural runoff. Associations between fecal protozoa and indicator bacteria were evaluated to determine the effectiveness of FIB as a protozoal proxy in this setting. Multiple sites within each wetland were monitored during a 2-year period to evaluate how environmental factors, such as rainfall and water quality measures, are associated with pathogen detection and transport from land to sea.

MATERIALS AND METHODS

Study sites.

The three central California coast wetland study sites (Fig. 1) were sampled repeatedly between April 2008 and June 2010. The tidal wetland was originally an estuary that drained into a multiuse harbor and recreational area. Degradation of the wetland due to landscape conversion to agriculture, including livestock operations, has altered the region such that area waters remain brackish to hypersaline in most years. For water collection, seven sites spaced approximately 500 m apart, ranging from 4,000 m upstream to the harbor tide gates, were sampled seasonally (Fig. 1A). The tidal wetland received runoff from the dairy wetland described below. A total of 38 samples were collected for fecal protozoan analysis, and 25 samples were also tested for fecal indicator bacteria (FIB). The dairy wetland was a seasonally occurring wetland that received runoff from cattle pastures and eventually drained into the tidal wetland described above. Surface water was present in the dairy wetland predominantly during the wet season (November to April) and was sampled at five sites, spaced approximately 150 m apart along a 1,000-m track from the dairy down to the outflow into the tidal slough described above (Fig. 1A). A total of 26 samples were collected for protozoal analysis, and 21 samples were also tested for FIB.

Fig 1.

Maps of field sampling locations. (A) Tidal and dairy wetlands. (B) Constructed wetland.

The constructed wetland (Fig. 1B) was built as a field research site with water pumped in from an adjacent slough that drains a large expanse of agricultural fields. It has an upper snake-like channel section that is characterized by unidirectional flow and was sampled at locations spaced approximately 100 m apart. Additional samples were collected from the lower section of the constructed wetland, where a large floodplain permits multidirectional flow depending on tidal and weather conditions, as well as from the source water in the adjacent slough. California bulrush (Schoenoplectus californicus) lined the banks of the upper snake-shaped channel, and the same plants formed 10 two-meter-wide vegetative buffers, or berms, that were perpendicular to the main water channel throughout this upper section. The lower floodplain section consists of a variety of plants, including California bulrush and slough sedge (Carex obnupta). Water samples were collected from the constructed wetland once weekly over four consecutive weeks seasonally, resulting in a total of 87 samples for protozoal analysis and 76 samples tested for FIB.

At each sampling location in all 3 wetlands, 10 liters of surface water was collected in a sterile plastic container for protozoal analysis along with 100 ml of water for bacterial analysis, as described below. Water quality parameters were also recorded at each wetland on each sampling day, including water temperature (°C), turbidity (nephelometric turbidity units [NTU]), salinity (parts per thousand [ppt]), total dissolved solids (mg/liter), dissolved oxygen (mg/liter), and pH.

Matrix spikes were prepared periodically during the study to determine percent recovery for protozoal detection methods. Easy Seed (BTF Bio, Pittsburgh, PA) aliquots containing 100 inactivated Cryptosporidium oocysts and 100 inactivated Giardia cysts were added to 10 liters of surface water from the wetland. These matrix samples were processed in the same way as the unspiked-field wetland water samples.

Microbial detection.

Samples for protozoal analysis were processed according to EPA Method 1623 (33) for detection of Cryptosporidium parvum and Giardia lamblia in water through filtration using Envirochek cartridges, followed by immunomagnetic separation (IMS) for parasite purification using Dynabeads GC-Combo (Invitrogen Life Sciences, Carlsbad, CA) and direct fluorescent antibody tests (DFA) for oocyst or cyst identification using EasyStain (BTF Bio, Pittsburgh, PA) and quantification using microscopy under both fluorescein isothiocyanate (FITC) and DAPI (4′,6-diamidino-2-phenylindole) epifluorescence. Particles with apple-green fluorescence in an oval or spherical shape (3 to 7 μm in diameter) with bright, highlighted edges under FITC and light blue internal staining with a green rim or up to four distinct, sky-blue nuclei under DAPI were classified as Cryptosporidium oocysts. Particles with apple-green fluorescence in a round to oval shape (6- to 15-μm in diameter) with bright, highlighted edges under FITC and light blue internal staining with a green rim or up to four distinct, sky-blue nuclei under DAPI were classified as Giardia cysts. Results were expressed as the oocyst or cyst count/10 liters of water sampled (33).

To compare protozoal concentrations with those of the more commonly measured indicator bacteria, samples for FIB analysis were examined using Colilert-18 (IDEXX Laboratories, Inc.), a standard EPA-approved method for detection of total coliforms and Escherichia coli in surface water samples. Briefly, water samples were mixed with Colilert reagent packs and incubated at 35°C in a Quanti-Tray/2000 (IDEXX Laboratories, Inc.) for 18 h, with total coliform concentrations estimated by the number of yellow wells and E. coli counts estimated by the number of fluorescent wells. The results were expressed as the most probable number (MPN) per 100 ml.

Statistical analyses.

A variety of wetland characteristics and environmental factors were evaluated as risk factors for associations with protozoal and indicator bacterial concentrations. Risk factors included wetland type (tidal, dairy, constructed), sample location within each wetland, and season (wet versus dry). Along the central California coast, the rainy season extends from November through April and the dry season extends from May through October. Rainfall events were described as rain occurring within the 12 h prior to sampling or within the preceding 4 days, as determined by the nearest weather station. Turbidity, total dissolved solids, and pH were evaluated as continuous variables. Water temperature and dissolved oxygen values were defined as “increased” or “decreased” with respect to the mean of the collected values for each wetland. Salinity values of less than 1 ppt were defined as “freshwater,” while samples with salinity greater than or equal to 1 ppt were defined as “brackish.” Categories for high FIB contamination of surface water were defined as E. coli concentrations greater than 400 MPN and total coliform concentrations greater than 10,000 MPN. Escherichia coli and total coliform counts falling below these established cutoffs were deemed to have low to moderate FIB contamination.

The four primary outcomes of interest were the surface water concentrations of Cryptosporidium, Giardia, E. coli, and total coliforms. Descriptive statistics were calculated to compare these pathogen groups, with geometric mean concentrations calculated for each protozoal species and regression on order statistics (ROS) calculated as a measure of central tendency for indicator bacteria. The ROS method was used to take into account censored data for samples with concentrations below the method detection limits (1 organism/100 ml). Initial analyses revealed that the protozoal means were equal to the variance, supporting the use of Poisson regression models for subsequent statistical procedures. Log-linear longitudinal Poisson regression (14) was used to determine the effect of each independent variable on each study outcome in order to identify risk factors that were significantly associated with higher protozoal or bacterial counts. Poisson regression is a form of statistical regression analysis used to model count data, which are assumed to have a Poisson distribution. One assumption of this model is that there is no overdispersion, meaning that the variance does not exceed the mean. This assumption was not violated with the study data. Longitudinal regression was used to account for repeated sampling at the same sites over time; specifically, sampling sites were treated as random effects. Nonparametric Spearman's rho and chi-square tests were calculated to determine potential colinearity between the parameters of interest. Factors significant in bivariate analyses were then used to build a multiple Poisson model to test several predictor variables simultaneously. Additionally, Spearman's rho correlation coefficient was calculated for protozoal and bacterial counts. Statistical analyses were completed using SAS software, version 9.2 (SAS Institute Inc., Cary, NC), and P values of <0.05 were considered statistically significant.

RESULTS

Microbial detection.

A total of 151 samples were tested for fecal protozoa and 122 samples were tested for fecal indicator bacteria (FIB), with all target organisms detected in all wetland types and sampling locations. The prevalence, mean concentration, and concentration range of Cryptosporidium, Giardia, total coliforms, and E. coli in water samples collected from the three wetlands are shown in Table 1. Within the tidal wetland, 39% of water samples were positive for Cryptosporidium oocysts while 21% were positive for Giardia cysts. The dairy wetland exhibited the greatest prevalence of Cryptosporidium oocysts and Giardia cysts, with 69% and 50% of water samples testing positive for each, respectively. The constructed wetland had a 44% Cryptosporidium oocyst prevalence and a 13% Giardia cyst prevalence. Furthermore, surface water collected from the dairy wetland exhibited the highest geometric mean concentrations of both protozoa (168 oocysts and 85 cysts per 10 liters), followed by water collected from the tidal wetland (31 oocysts and 36 cysts per 10 liters) and the constructed wetland (5.5 oocysts and 3.2 cysts per 10 liters). Matrix spikes from eight surface waters showed that Cryptosporidium recovery ranged from 2 to 26%, with an average of 13% (standard error [SE], 3.4%), and Giardia recovery ranged from 0 to 30%, with an average of 15% (SE, 4.6%).

Table 1.

Prevalence and concentrations in coastal California wetlands of Cryptosporidium oocysts and Giardia cysts using geometric means and of Escherichia coli and total coliforms using regression on ordered statistics

Wetland and characteristica	Value for each protozoan or FIB
	Cryptosporidium	Giardia	Total coliforms	E. coli
Tidal wetland
Prevalence (% [no. of positive samples/total no. of samples])	39 (15/38)	21 (8/38)	100 (25/25)	100 (25/25)
Mean concn	31	36	8,571	316
Concn range	0–389	0–15,000	100–41,060	46–1,890
Dairy wetland
Prevalence (% [no. of positive samples/total no. of samples])	69 (18/26)	50 (13/26)	100 (21/21)	100 (21/21)
Mean concn	168	85	173,590	870
Concn range	0–54,044	0–36,417	2,000–2,419,600	200–4,840
Constructed wetland
Prevalence (% [no. of positive samples/total no. of samples])	46 (40/87)	13 (11/87)	100 (78/78)	100 (78/78)
Mean concn	5.5	3.2	155,973	2,913
Concn range	0–148	0–14	1,320–2,420,000	4–46,400

^aConcentrations are oocysts or cysts per 10 liters for Cryptosporidium and Giardia and MPN per 100 ml for total coliforms and E. coli.

As expected, 100% of samples from all three wetlands tested positive for E. coli and total coliforms across all sample points and time periods. Surface water collected from the constructed wetland exhibited the highest mean concentrations by ROS of E. coli (2,913 MPN/100 ml), followed by water from the dairy wetland (870 MPN/100 ml) and then water from the tidal wetland (316 MPN/100 ml). For total coliforms, the dairy wetland exhibited the highest concentrations (173,590 MPN/100 ml), followed by the constructed wetland (155,973 MPN/100 ml) and the tidal wetland (8,571 MPN/100 ml).

The concentrations of Cryptosporidium, Giardia, E. coli, and total coliform by sample location for each wetland are shown in Fig. 2, with the “distance downstream” representing the distance from the most-upstream sampling location to each sampling location in the wetland. Microbe levels in the tidal wetland were strongly influenced by sample location, particularly with regard to where the outflow from the dairy wetland joined the tidal wetland. For both protozoa and bacteria, microbe concentrations at the tidal wetland were highest at sites immediately upstream and downstream of the dairy wetland outflow and then decreased thereafter. Results from the dairy wetland showed that Cryptosporidium, Giardia, and FIB concentrations were significantly higher (P value < 0.0001) in the site closest to the dairy than in sample locations further downstream. Although very low concentrations of protozoa were detected in the constructed wetland, FIB enumeration demonstrated a similar trend, with the highest concentrations at the sample location in the source water from the adjacent slough (sample distance of 0 m) and lower counts in water sampled from the subsequent downstream sites within the channel portion of the constructed wetland.

Fig 2.

Mean concentrations of Cryptosporidium oocysts, Giardia cysts, total coliforms, and Escherichia coli by site in the tidal wetland (A and B), dairy wetland (C and D), and constructed wetland (E and F). Arrows indicate where the dairy wetland joins the tidal wetland. For A, C, and E, the solid line represents the Cryptosporidium oocyst concentration and the dashed line represents the Giardia cyst concentration. For B, D, and F, the solid line represents the total coliform concentration and the dashed line represents the E. coli concentration.

Risk factors.

Analysis of associations for Cryptosporidium and Giardia concentrations with environmental factors was examined using Poisson regression, as the protozoal means were not overdispersed. Without adjusting for wetland effects, rainfall in the 12 and 96 h prior to sample collection was associated with detection of significantly higher concentrations (P value < 0.0001) of both Cryptosporidium oocysts and Giardia cysts. The odds ratio (OR) for rainfall in the prior 12 h with Cryptosporidium counts as the outcome was 30, meaning that in water samples collected from sites that received rainfall within the 12 h prior, the Cryptosporidium oocyst counts were likely to be 30 times greater than those for water samples collected when there was not rainfall in the preceding 12 h. For Giardia, the association between rainfall and cyst enumeration was even more pronounced, with an odds ratio of 217. Thus, for every Giardia cyst in a water sample when there was no rainfall in the 12 h prior to collection, there were likely to be 217 times as many Giardia cysts in water samples collected when rainfall had occurred just previously. The wet season was also associated with significantly higher protozoal counts compared to those of dry season sampling, with odds ratios of 45 for Cryptosporidium and 1,510 for Giardia detection in wet-season surface waters (P value < 0.0001).

Microbial detection and concentrations were significantly associated with most of the measured water quality parameters, including dissolved oxygen, water temperature, total dissolved solids, and salinity (P value < 0.05), but not turbidity and pH. Increased dissolved oxygen was negatively associated with protozoal and bacterial counts. Increases in total dissolved solids were also negatively associated with protozoal counts (P value < 0.0001). Warmer water temperature was inversely associated with protozoal counts (P value < 0.0001) but positively associated with total coliform values (P value < 0.005). When stratified by wetland type, water samples with higher salinity were negatively associated with both Cryptosporidium (P value < 0.05) and Giardia (P value < 0.0001) counts, meaning that water samples from sites with lower marine influence were more likely to have high protozoal concentrations. Mean salinity levels for the tidal wetland tended to be brackish, at 11.1 ppt, whereas the dairy and constructed wetlands exhibited lower salinity, at 1.3 ppt and 2.6 ppt, respectively.

Protozoal relationship with fecal indicator bacteria.

When Spearman's rho correlation coefficient was calculated to compare protozoal counts to indicator bacterial counts, no FIB-protozoa combinations were found to be significant. However, when a Poisson model approach was taken to look at the association between protozoal concentrations with “high” or “low” FIB levels, significant associations were observed. Using this approach, E. coli counts greater than 400 MPN/100 ml were associated with higher protozoal counts; for example, the odds ratio of having high levels of E. coli with increased Giardia cyst counts in water samples was 210 (P value < 0.0001). Thus, if a sample was found to have E. coli counts greater than 400 MPN, the sample would be predicted to have a count of 210 times more Giardia cysts than if there was an E. coli concentration of below 400 MPN. Conversely, concentrations of total coliforms exceeding 10,000 MPN were associated with lower protozoal oocyst or cyst concentrations.

Regression models.

Longitudinal Poisson regression was used to further assess relationships between protozoal counts in surface water and various environmental and water quality parameters and FIB measures. In these models, Cryptosporidium and Giardia concentrations in surface water samples were best predicted by simultaneously considering wetland type, FIB levels, and rainfall variables. After adjusting for the difference in protozoal count by wetland type, rainfall was assessed to account for both the overall effect of rain and the difference in rainfall between the three wetlands through the use of interaction terms in the model. These results are tabulated in Table 2, and the associated predictive models are summarized in Fig. 3.

Table 2.

Odds ratios from multiple Poisson regression models for significant variables predicting concentrations of protozoa and fecal indicator bacteria in surface waters from coastal wetlands^a

Predictor variable	n	OR and P value of concn (per 10 liters) of:				OR and P value of MPN (per 100 ml) of:
		Cryptosporidium		Giardia		Escherichia coli		Total coliform
		OR	P value	OR	P value	OR	P value	OR	P value
Wetland type
Tidal	38	1		1		1		1
Dairy	26	12.13*	0.0020	3.93	0.1565	0.86	0.7146	32.97*	<0.0001
Constructed	104	0.07*	0.007	0.08*	0.0105	2.85*	0.0189	4.37*	<0.0001
Indicator bacteria
Total no. of coliforms (below 10,000 MPN)	86	1		1		1
Total no. of coliforms (above 10,000 MPN)	82	0.01*	<0.0001	0.12	0.0035	2.51*	<0.0001
E. coli (below 400 MPN)	102	1		1				1
E. coli (above 400 MPN)	66	4.75*	0.0096	265.39*	<0.0001			1.94*	0.0340
Salinity
Freshwater	76					1
Brackish water	92					0.34*	<0.0001
Rainfall (mm)
Total rainfall on sample date	155	0.73*	0.0254
Rainfall × tidal	38	1
Rainfall × dairy	21	1.75*	0.0252
Rainfall × constructed	96	1.42*	0.0226
Rainfall occurrence
No rainfall on sample date	127			1
Rainfall on sample date	41			103.86*	<0.0001
Rainfall × tidal	7			1
Rainfall × dairy	5			0.68	0.7473
Rainfall × constructed	29			0.01*	0.0082
Rainfall (mm)
Total rainfall within prior 96 h	172					0.99	0.5339	0.99	0.5904
Rainfall × tidal	38					1		1
Rainfall × dairy	26					1.11*	0.0172	0.73*	0.0122
Rainfall × constructed	108					1.04*	0.0255	1.06*	0.0353

^aAsterisks indicate a P value of <0.05. OR, odds ratio.

Fig 3.

Predictive models for Cryptosporidium oocysts (A), total coliforms (B), and E. coli (C) utilizing Poisson regression models in which variables other than rainfall were held stable. Diamonds, tidal wetland; squares, dairy wetland; triangles, constructed wetland.

For Cryptosporidium, recent rainfall exposure was the best predictor of protozoal concentrations when evaluated as the volume of rainfall (in mm) in the 12 h prior to sampling. Rainfall occurring within 12 h prior to water sampling at the dairy and constructed wetland sites was associated with 1.7- and 1.4-fold increases in Cryptosporidium oocyst concentrations, respectively (P value of <0.05 for both). Inclusion of FIB levels into this model also had an impact on Cryptosporidium oocyst detection: “high” E. coli counts of greater than 400 MPN were associated with 4.8-fold increased concentrations of Cryptosporidium oocysts. Conversely, total coliform counts of greater than 10,000 MPN in surface water were associated with lower concentrations of Cryptosporidium oocysts, similar to what was noted in univariate analyses. The predictive model in Fig. 3A shows that as rainfall volume increases, the Cryptosporidium concentration is predicted to increase in the dairy wetland but decrease in the tidal wetland.

For Giardia, the longitudinal Poisson regression model showed trends similar to those of the Cryptosporidium model, with one difference being that increased Giardia counts were associated with the occurrence of rainfall in the 12 h prior to sampling rather than the total volume of rainfall. The odds ratio shows that Giardia counts were 104 times greater in water samples collected when there was preceding rainfall, regardless of wetland type, than when there was no rainfall within the prior 12 h. Additionally, the effect of fecal contamination (classified as water samples with E. coli concentrations above 400 MPN) was much greater for the Giardia model than for Cryptosporidium, with an odds ratio of 265. The Giardia model utilizing rainfall as a continuous variable did not converge, so no predictive model is shown.

The longitudinal Poisson regression models for the bacterial indicators included the risk factors of wetland type, FIB level, and total rainfall within the prior 96 h (Table 2 and Fig. 3). For the E. coli model, salinity was also an important factor, with an odds ratio of 0.3 (P value < 0.0001) for brackish water compared to freshwater. Thus, freshwater streams had greater concentrations of E. coli than brackish water estuaries, as expected if fecal pollution is coming from freshwater sources up the watershed. For total coliforms, both the dairy and constructed wetlands had odds ratios greater than 4 when compared to the tidal wetland (P value > 0.0001), meaning that significant increases in total coliform counts were noted between these two freshwater wetlands and the more marine-influenced tidal wetland. When stratified by wetland type with tidal wetland as the reference category, total rainfall (mm) was negatively associated with levels of total coliforms for the dairy wetland, but a slight positive association was noted for the constructed wetland compared to the tidal wetland.

DISCUSSION

While the majority of studies on protozoal transport have focused on constructed wetlands as a mitigating factor for sewage treatment (15, 18, 32, 35), livestock operations, domestic pets, and wildlife can also contribute substantially to environmental loading (2, 4, 5, 22). All three central California coastal wetlands examined in this study contained the protozoal pathogens Cryptosporidium and Giardia as well as FIB in surface waters, and all three wetlands showed reductions of both protozoal and bacterial concentrations within the wetlands. For the tidal and dairy wetlands in particular, the pathogen concentrations were high enough to be a public health concern for humans utilizing these water sources for recreation, especially considering that these wetlands drain into a harbor and sanctuary that is heavily used for recreation (31). The public health importance of the parasites depends on the particular species of Cryptosporidium or Giardia, as there are host-specific species and genotypes, such as the cattle-specific Cryptosporidium andersoni and G. lamblia assemblage E, which are not commonly linked to disease in humans. However, the transport dynamics are expected to be similar across genotypes, and this study highlights the importance of wetlands in reducing pathogen loads.

The dairy wetland, which forms naturally during the wet season and drains into the adjacent tidal wetland, exhibited the greatest concentrations of Cryptosporidium, Giardia, and FIB. Increased concentrations of both protozoa and bacteria were also noted at the tidal wetland when the dairy wetland was discharging water to sample sites closest to the junction of the two wetlands. Increased livestock density, increased water flow rates, or further degradation of this wetland habitat may lead to increased amounts of pathogens being transported to adjacent recreational waters. It is acknowledged that results may be affected by unaccounted-for variation between study sites, such as differing water volumes, watershed areas, and surrounding land uses between each wetland. Furthermore, the actual concentrations of Cryptosporidium and Giardia in the wetlands are underestimated, as shown by the matrix spikes in which protozoal recovery ranged from 0 to 30% of spiked parasites in wetland water samples. The use of acid dissociation may underestimate the number of parasites compared to that found by heat dissociation (29, 37), but acid was used to be consistent with the standard EPA 1623 protocol for Cryptosporidium and Giardia identification in water (33). Ultimately, this study illustrates the ability of natural and reconstructed wetlands to reduce pathogen loads from a variety of sources.

Protozoal contamination from livestock feces is of particular concern because cattle are potential reservoirs for Cryptosporidium parvum and Giardia lamblia assemblage A (11, 17, 39), both of which are known to infect humans. Previous studies have examined the role of dairies and other intensive livestock operations in pathogen transmission and identified potential risk factors for nearby surface water contamination, including application of manure on agricultural fields (30) and surface runoff from areas with high animal densities, such as dairies and beef cattle operations (23). Consistent with these results, the data from this study demonstrate that the wetland in closest proximity to a livestock operation exhibited the greatest protozoal and bacterial loading. We also observed that a tidal wetland receiving water from this same dairy wetland had the highest microbial concentrations at sites nearest to the outflow from the dairy wetland. Tidal wetland sites immediately upstream and downstream of the junction with the dairy wetland exhibited greater parasite concentrations because this tidally influenced wetland experiences bidirectional flow (tide was not controlled for in the wetland sampling scheme).

While intensive livestock operations have significant potential to contribute to downstream pathogen loads, several best management practices (BMPs) have been proposed to reduce pathogen contamination in runoff flowing from pastures to adjacent receiving waters. First, reduction of cattle density may reduce the spread of infection within the herd and reduce the amount of contaminated manure released into the environment (23, 24). Second, reducing or more carefully controlling field applications of manure and slurry can greatly decrease the number of pathogens released into the environment (16). Finally, incorporation of vegetative filter strips, or “buffer strips,” can reduce the number of oocysts and cysts transported out of intensive livestock operations (3, 16, 23, 24). Implementation of these BMPs at dairies and other livestock operations may considerably reduce their contribution to pathogen loading of adjacent wetlands.

In addition to the close proximity of fecal pathogen sources, rainfall and specific water quality parameters were also associated with Cryptosporidium, Giardia, and FIB counts in surface water. Rainfall events just prior to water sampling were associated with increases in oocyst or cyst concentrations across wetland types. Rainfall occurring within the previous 12 h was associated with increased Cryptosporidium and Giardia counts in the dairy wetland water, a result which is likely due to enhanced runoff from fecally contaminated soils. In livestock operations that utilized vegetative buffer strips to reduce watershed contamination from storm runoff, previous studies have documented reduced pathogen loading of adjacent waterways (6, 23, 24). Reductions in water quality, characterized by increased water temperature, total dissolved solids, and decreased dissolved oxygen, were also associated with higher Cryptosporidium and Giardia concentrations. Runoff from intensive livestock and agricultural operations can adversely affect these water quality measurements (19) and may also directly and indirectly contribute to pathogen transport.

Wetlands can be used to reduce microbial pollutant loads through several processes, including adsorption, sedimentation, and vegetative uptake (20). As with prior studies, our data indicate a general reduction in Cryptosporidium, Giardia, and FIB counts as contaminated water travels downstream through the wetlands. The data from the dairy wetland clearly showed that when protozoa and FIB were detected in upstream samples, samples collected further downstream exhibited lower pathogen concentrations. In a similar pattern, the tidal wetland had increased microbial concentrations nearest to the dairy inlet but downstream samples had returned to the low concentrations seen upstream of the inlet. There was an increase in E. coli concentrations in the last tidal wetland site, but this may be attributed to other potential fecal contamination sources, including live-aboard boats in the adjacent harbor and a substantial wildlife population residing near the downstream site. While the background levels of Cryptosporidium and Giardia were too low to document an effect in the constructed wetland, the total coliform and E. coli counts decreased sequentially throughout the water channel portion of the wetland. In contrast, the lower floodplain section was observed to be periodically inundated with undiluted water from the adjacent slough, perhaps explaining the increase in pathogen concentrations at the last sample site.

Surface water quality-monitoring programs generally rely on FIB levels to represent the risk of pathogen exposure and the associated health risks to humans. This approach is taken in part because FIB testing is less expensive and easier than direct pathogen testing. When FIB were analyzed as predictors of protozoal pathogen concentrations in coastal wetland systems, the detection of E. coli counts exceeding EPA guidelines (greater than 400 MPN) was positively associated with increased concentrations of protozoal oocysts or cysts in water samples. However, detection of total coliform levels exceeding EPA guidelines (greater than 10,000 MPN) was negatively associated with protozoal counts. In general, E. coli is considered a more reliable indicator of fecal contamination than total coliforms (10), the latter of which can vary with environmental conditions and bacterial populations (34). In contrast to FIB that can multiply in the environment, Cryptosporidium oocysts and Giardia cysts are shed in the feces and do not multiply in the environment. Thus, the association of protozoal counts with a more reliable marker for fecal contamination (E. coli counts) rather than the less-specific total coliform counts is not surprising. However, it can be noted that although increased E. coli counts correlated with greater protozoal concentrations, no significant correlations were identified between the absolute numbers of E. coli cells and protozoa detected in the same water samples. This contrasts with results from prior studies that evaluated Cryptosporidium and Giardia contamination of drinking water sources and reported linear correlations between protozoal concentrations and FIB concentrations (25, 38).

In conclusion, given that fecal pathogens can enter waterways through a multitude of routes and sources, natural wetlands may serve as a sustainable BMP to improve water quality in downstream waters. The study results show that protozoal and bacterial concentrations can be reduced as water travels through the coastal wetland and also that rainfall events are important to consider when identifying the times of highest risk for fecal pollution entering waterways. Recognizing the potential of natural and reconstructed coastal wetlands for reducing transport of pathogens downstream holds promise for improving ecosystem health while also conserving natural habitats that are important for wildlife and a variety of recreational uses.