The use of carbon-utilization profiling to determine sources of fecal contamination in a central Texas watershed

Abstract

The purpose of this research was to characterize the relationship of Escherichia coli isolates from various animals, sewage, and water based on carbon substrate utilization patterns, and to use these patterns to determine the dominant contributors of nonpoint source fecal pollution to a central Texas reservoir from its watershed. We collected 1028 fecal samples from cattle, companion animals, goat, horse, poultry, sewage, sheep, and wildlife, collectively. From these, 1915 E. coli fecal isolates were analyzed. We collected 100 water samples throughout the North Bosque watershed; 910 E. coli water isolates were analyzed. The Biolog system was used to generate a carbon-utilization pattern for each isolate. A dendrogram constructed from the carbon-utilization patterns demonstrated that isolates from the same source category usually clustered together, excluding water isolates that were spread over many clusters. A bacterial-source tracking library was constructed from the carbon-utilization data obtained from the fecal and sewage isolates and analyzed for internal accuracy. Rates of correct classification for the library ranged from 12.8 to 78.6%. The average rate of correct classification for the library was 45.8%, and specificity values were high, ranging from 75 to 99%. When water isolates were submitted to the library for identification, 43 were classified as originating from cattle, indicating cattle were the dominant source of fecal pollution in the watershed. This was followed by sewage at 27%. Based on these data, our first recommendation for decreasing bacterial pollution in this watershed is to implement strategies that can reduce the contribution of fecal contamination from these 2 sources.

Key words:

• bacteria
• bacterial source tracking
• carbon substrate utilization
• Escherichia coli
• nonpoint source pollution

Monitoring bacteria in our environment, specifically the harmful ones, has been a long-standing goal of organizations such as the United States Environmental Protection Agency (USEPA). However, monitoring the levels of fecal pollution in our environment and controlling the levels of fecal pollution in our environment are quite different. Most of the methods used previously to monitor the potentially harmful organisms are quantification methods only, relying on the detection of Escherichia coli or other intestinal bacteria. Unfortunately, because many of these organisms are ubiquitous bacteria found in the intestines of all warm-blooded animals and a few cold-blooded animal species, their presence in water or other environmental samples does not specifically differentiate the sources of the pollution (Wiggins 1996). However, with continued advances in bacterial source tracking (BST), dominant, nonpoint sources (NPS) of contamination can be uncovered (Myoda et al. 2003). Determining the dominant contributor(s) of NPS pollution to waterways is specifically important for developing lake and reservoir management strategies. If the dominant source(s) of NPS fecal pollution can be determined, remedial processes that target these specific polluters can be implemented if deemed necessary.

Bacterial source tracking refers to a variety of techniques currently under development and testing for use in identifying NPS bacterial pollution in the environment (Scott et al. 2002). Several methods have been tested for their use in BST studies, but at this time there is no consensus on the best field application method(s) (Simpson et al. 2002, Casarez et al. 2007). Historically, however, the most widely used BST methods are library-based methods (Stewart et al. 2003), which require the collection of bacterial samples from known host sources (such as cattle or horses), followed by analysis of the isolates with the selected BST method(s). Data from the known source isolates are then compiled and used as a “reference” for determining the origins of bacterial isolates from unknown sources of contamination (such as bacterial isolates obtained from water). Continued research in the area of BST is expected to help this branch of science become widely used and explored as scientists, regulatory agencies, and the public gain confidence in the accuracy of the results obtained with these methods (Straub and Chandler 2003).

A library-based BST method, carbon substrate utilization profiling, was used in this study. The purpose of this research was 2-fold: (1) to determine the carbon substrate utilization profiles of E. coli isolates from various host organisms, sewage, and water; and (2) to use these carbon substrate utilization patterns to determine the dominant contributor(s) of NPS fecal pollution to a central Texas reservoir from its watershed and devise management strategies for these sources of pollution.

Methods

Study area

Fecal samples were collected from 6 counties spanning central Texas, and sewage samples were collected from 4 different wastewater treatment plants (WWTP) located throughout the study area (Fig. 1).

Figure 1 Fecal and sewage sampling locations. Counties where fecal and sewage samples were collected are in gray.

Water samples collected for this study were obtained from within the boundaries of the North Bosque watershed and Lake Waco located in central Texas (Fig. 2). The North Bosque watershed encompasses approximately 1210 mi² (TIAER 2007), and its most prominent feature is the 97-mi long North Bosque River. The North Bosque River flows south from Erath County, cutting through numerous other counties including Somervell, Hamilton, Bosque, Coryell, and McLennan. It is joined by other branches of the North Bosque River before reaching Lake Waco in McLennan County.

Figure 2 North Bosque River Watershed. Letters denote designated water sampling sites.

The water in the North Bosque watershed is relatively free-flowing, and the vegetation surrounding the river includes mainly hardwoods, conifers, and grasses (The Handbook of Texas online, accessed 2008). Major agricultural practices are present throughout the watershed, including the use of land for dairy farming, range cattle, forage hay, peanuts, and pecans (Keplinger et al. 2003). The North Bosque River has been included on Texas's impaired water list since 1992 for excessive aquatic growth due to phosphorus overload (Keplinger et al. 2003).

Mean flow rate from the North Bosque watershed into Lake Waco is approximately 290 cfs. Lake Waco is located within the Waco city limits just off Texas Highway 6 on the Bosque River. The lake covers 35.3 km² with a maximum depth of about 22 m and a mean depth of 6.4 m (Filstrup and Lind 2010). Lake Waco, as well as the area around the lake, serves as a recreation area for many central Texans. Around Lake Waco there are multiple camping sites as well as numerous hiking and biking trails (Texas Parks and Wildlife online, accessed 2008; Conry 2010). Aside from its use as a recreation area, pollution management strategies are specifically important for Lake Waco because it is the main source of public drinking water for more than 200,000 Central Texas residents (Keplinger et al. 2003). The North Bosque watershed is a main contributor of water and, likely, pollutants to Lake Waco. Because of the lake's importance as a major source of drinking water for central Texas, and because the North Bosque River is a waterway that is known to be impaired, the North Bosque watershed was targeted as the study area for this research.

Study design

Escherichia coli were isolated from fecal, sewage, and environmental water samples and were analyzed for their carbon source usage. Fecal and sewage samples were collected over the course of one year, from January 2005 through November 2005, with one additional sampling in November 2006. Water samples for this study were collected from February 2005 through June 2005 from 20 sampling sites located within the boundaries of the North Bosque watershed (Fig. 2).

Fecal and sewage sample collection

Fecal samples (about 10–25 g each) were obtained from 8 different animal categories. The 5 agricultural animal categories included in the study were beef and dairy cattle, goat, sheep, and poultry. Additionally, fecal samples were collected from companion animals (domestic cat and dog), horse, and wild deer. Fecal samples were collected from freshly voided droppings using a sterile swab or a gloved hand and were only collected if the source of the droppings could be identified. The majority of the cattle samples were collected directly through rectal palpation under the supervision of a veterinarian. Samples were placed into labeled, sterile Whirl-Pac bags for transportation back to the lab. Sewage samples were collected from the influent at various WWTP in 100-mL sterile plastic bottles.

Water sample collection

Water samples for this study were collected from 20 designated sampling sites spanning the North Bosque watershed (Fig. 2). Samples were collected on 5 collection dates during 2005; February 2, February 23, March 9, May 5, and June 17. On each collection date, one water sample was collected from each of the 20 sampling sites, totaling 100 water samples collected for the study. Water samples were collected in 100-mL sterile plastic bottles from each designated site and stored on ice until reaching the lab.

Sample processing and E. coli isolation

Fecal samples and sewage samples were diluted in 100 mL of sterile phosphate-buffered dilution water (PDW) and mixed thoroughly to create a homogenous solution before processing. All samples were processed according to USEPA method 1604 for isolation of E. coli. Briefly, samples were vacuum-filtered through 0.45-μm gridded membrane filters (Fisher Scientific, Waltham, MA, USA), filters were placed on MI agar (VWR, West Chester, PA, USA), and plates were incubated for 24 h at 35 C. Escherichia coli exhibit a fluorescent blue halo after incubation when grown on MI agar. Water isolates exhibiting a fluorescent blue halo were enumerated to determine the E. coli concentrations for each water sample site. Additionally, isolates chosen for BST analysis were confirmed as E. coli with standard biochemical testing (negative citrate permease and negative urease tests). Any isolates that did not meet these criteria were omitted from further analysis.

Carbon substrate utilization profiling

Approximately 2–4 isolates from each fecal and sewage sample and 10 isolates from each water sample were randomly selected for carbon profiling using the Biolog Microbial Identification System (Biolog, Hayward, CA, USA). Carbon-utilization profiling was conducted following the manufacturer's instructions for gram-negative organisms. In summary, selected E. coli isolates that passed confirmation were streaked onto Biolog Universal Growth (BUG) agar supplemented with defibrinated sheep's blood and incubated overnight at 34C. Bacteria were then suspended in inoculating fluid, pipetted into each well of a Biolog GN2 Microplate and incubated at 34C for 14–16 h in the dark and without agitation. The Biolog GN2 Microplate is a 96-well microplate that contains 95 different carbon substrates (one substrate per well) and a negative control well (no carbon substrate). All wells contain a tetrazolium dye. Following inoculation and during incubation of the plates, a respiratory burst occurs in wells where the carbon source is utilized, reducing the tetrazolium dye in that well and resulting in a purple color change. An automated 96-well plate reader from Biolog (Molecular Devices, Sunnyvale, CA, USA) was used to read color formation at 590 nm in each well of the inoculated 96-well GN2 plate.

Quality control

Approximately 100 E. coli were selected at random from the database of isolates and run in triplicate to validate the reproducibility of the carbon-utilization patterns. The average reproducibility for carbon-utilization patterns across all groups was 91.3%.

Data analysis

Raw Biolog plate reader data were exported into spreadsheet format and values were corrected for their corresponding negative control well value. Any resulting negative values were changed to zero. To characterize the relationship of the E. coli isolates to each other based on carbon-utilization patterns, similarity values were calculated from the corrected Biolog data using Pearson's product-moment correlation coefficient, and the resulting values were used to construct an Unweighted Pair Group Method with Arithmetic Mean (UPGMA) dendrogram in Bionumerics version 4.0 (Applied Maths, Austin, TX, USA).

BST library analysis

To create the library of isolates for BST, carbon-utilization data from fecal and sewage isolates were imported into Bionumerics, and carbon profiles were compared using Pearson's product-moment correlation coefficient. Prior to inclusion of the carbon sources into the BST library, stepwise discriminant analysis was performed to determine which carbon sources (if any) did not contribute to the differentiation of the source categories. This resulted in a reduction in the number of carbon substrates used in the creation of the BST library from the original 95 carbon substrates to 64. The BST carbon-utilization library was evaluated using Jackknife analysis after clonal isolates were removed from the dataset. Jackknife analysis consists of removing one isolate at a time from the library and resubmitting that isolate back to the library as an unknown for identification. The rate of correct classification (RCC), also called sensitivity, for each source in the library was determined by calculating the percent of isolates correctly identified as belonging to their source category. Rates of correct classification were then averaged across all sources to determine the average rate of correct classification (ARCC) for the entire BST carbon-utilization library. Additionally, specificity values, the ability of a particular method to discriminate between source categories, were calculated for the library, as outlined in the USEPA Microbial Source Tracking Guide Document (USEPA 2005).

Identification of sotential sources of pollution in the watershed

To identify the potential source(s) of unknown (water) isolates, carbon-utilization data for each water isolate were imported into Bionumerics and compared to library carbon-utilization data constructed from the fecal and sewage isolates. Pearson's product-moment correlation coefficient was used with the maximum similarities approach to determine the source category in the library that the water isolate most closely resembled. To reduce the rate of misclassification of water isolates into certain source categories, a match of a water isolate to a specific source category was only accepted if the water isolate had a minimum similarity of 70% to the source category in the library.

Additionally, the classification of water isolates into a specific source category was only accepted if the source identification was given a quality factor in Bionumerics of C or better. Water isolates assigned to a source category with a quality factor below C were categorized as “unidentified.” A quality factor is an identification score, assigned by Bionumerics, that is based on the confidence of the assignment. As outlined by Genthner et al. (2005), a quality factor score of A is a very good assignment, B is a good assignment, and C is still a faithful assignment, whereas a score of D or E is considered a doubtful or unlikely assignment. Although these extra precautions increased the number of unidentified water isolates for the BST carbon-utilization library, they allowed us to be relatively certain that the source identification assignments accepted for the water isolates were accurate. Perhaps the most obvious benefit of this is the decreased likelihood of falsely blaming a potential polluter as the contamination source.

Results

Number of isolates collected and analyzed

The number of samples collected and the final number of E. coli isolates analyzed from each source category and water were summarized (Table 1). We collected 1028 fecal samples from cattle, companion animal, goat, horse, poultry, sewage, sheep, and wild deer, collectively. To correctly represent the source populations in the watershed, the numbers of fecal samples collected from each source category were proportional to the sources occurrence in the watershed. We obtained 8905 E. coli isolates from these fecal samples and attempted to obtain carbon-utilization data from 2056 of the isolates. From these 2056 fecal isolates, approximately 7% of the isolates and their data were omitted due to either lack of adequate results or the isolates’ clonal nature. In total, 1915 fecal E. coli isolates were used for analysis in this study. We collected 100 water samples from designated sample sites spanning the majority of the North Bosque watershed (Fig. 2); from these 100 water samples, 910 E. coli isolates were successfully analyzed.

Table 1 Number of samples collected and number of isolates analyzed from each sample type.

Number of Samples Source	Number of E. coli Collected	Number of E. coli Isolated	Included in Study
Cattle	382	3606	816
Companion	66	573	107
Goat	103	945	161
Horse	21	182	39
Poultry	30	295	580
Sewage	342	2493	55
Sheep	30	292	56
Wild Deer	54	519	101
Total	1028	8905	1915
Water	100	943	910

Relationship of E. coli isolates based on carbon substrate utilization data

A UPGMA dendrogram was created based on values calculated using Pearson's coefficient to determine how the approximately 2000 fecal and water E. coli isolates would cluster together based on similarity. Because of the large sample size, a condensed dendrogram using a 45% similarity cutoff value was constructed, reducing overall tree length and dividing the isolates into 8 groups (Fig. 3). Although the various source categories did not exclusively cluster together in the dendrogram, most were predominantly found in 1 or 2 clusters. For example, cluster A was large with a mixture of isolates, but almost half the total sheep isolates (46%) were grouped closely in this cluster. The total wild deer isolates (30%) were classified to cluster B, and the majority of poultry isolates (67%) were found in either cluster A or BCluster E was comprised of almost entirely sewage isolates (82%), and almost half of the isolates in cluster G were beef (44%). Cluster H had the largest percentage of total dairy and sewage isolates (33 and 25%, respectively). To some extent, E. coli isolates from water were spread across all clusters (excluding cluster E).

Figure 3 Unweighted Pair Group Method with Arithmetic Mean (UPGMA) dendrogram of E. coli isolated from various source categories based on carbon substrate utilization patterns (Bionumerics). Because of the large sample size (n = 2825), a condensed dendrogram using a 45% cut-off value is presented to show the 8 major groups.

Analysis of the carbon-utilization library for BST

The BST library for this study was constructed from the carbon-utilization data of the fecal and sewage isolates as described above, and the library was evaluated using Jackknife analysis (Table 2). The RCC for each source category, also called sensitivity (in bold) indicates the percentage of isolates from a specific source category that was correctly identified as from that source. For example, 62.8% of the sewage isolates were correctly identified as originating from sewage. Overall, RCC values were good for most categories, with the highest RCC in the cattle category at 78.6% and the lowest RCC in the horse category at 12.8%. The ARCC for the carbon-utilization library, at 45.8%, was calculated by averaging the RCC. The specificities for the carbon-utilization library by source class (Table 3) show that specificity values were high for all source categories, and all source categories exceeded the 80% ideal cutoff recommended by the USEPA (2005), except for the cattle category at 75%.

Table 2 Jackknife analysis of the carbon substrate utilization library.

	Classification (%)
Source Category	Cattle	Comp	Goat	Horse	Poultry	Sewage	Sheep	Deer
Cattle	78.6 *	1.6	5.0	0.4	1.6	7.2	2.1	3.6
Comp	25.9	42.6	3.7	0.9	2.8	18.5	2.8	2.8
Goat	28.6	1.9	55.3	1.2	1.2	10.6	1.2	0.0
Horse	33.3	5.1	2.6	12.8	5.1	23.1	2.6	15.4
Poultry	40.0	10.9	0.0	1.8	29.1	1.8	1.8	14.6
Sewage	21.4	5.3	4.7	1.0	0.7	62.8	0.7	3.5
Sheep	28.6	1.8	12.5	1.8	3.6	8.9	37.5	5.4
Deer	28.7	6.9	4.0	0.0	2.0	9.9	1.0	47.5
ARCC	45.8

Table 3 Specificities for the carbon-utilization library (%), by source category.

Source Category	Specificity
Cattle	75
Companion	97
Goat	95
Horse	99
Poultry	98
Sewage	91
Sheep	98
Wildlife	96
Average Specificity	94

Another important aspect to consider in BST studies is the number of unknown (in this case water) isolates left unidentified after analysis using the library. For this study, if the identification of the unknown isolate to a specific source category did not meet the stipulations outlined above (minimum 70% similarity, quality factor of C or better), these unknown isolates were labeled as unidentified. If a large percent of water isolates are left unidentified in a BST study, it is possible a potential source of contamination in the watershed has been overlooked and not included in the library construction. However, the percent of water isolates left unidentified using the carbon-utilization library was only 11%, even with the strict stipulations for accepting identifications.

E. coli concentrations in water samples

The concentration of E. coli in individual water samples was determined based on the number of E. coli colonies present on MI agar plates following USEPA method 1604 and the volume of water that was filtered. These values were used to calculate the geometric mean concentration of E. coli for each sample site over the entire water sampling period (Fig. 4).

Figure 4 Geometric mean concentrations of E. coli at each sample site for the entire sampling period. Black dashed line represents the USEPA Long Term Geometric Mean Limit for Contact Recreation(126 E. coli/100 mL).

Source identification of water isolates using the carbon-utilization library

The BST carbon-utilization library was used for the identification of 910 unknown (water) isolates. The percentage of water isolates (of the total isolates that were identified by each library) classified to each source category using the library were outlined (Fig. 5). Overall, the carbon-utilization library identified cattle as the dominant source of contamination in the watershed. Isolates identified as originating from cattle accounted for 43% of the total isolates classified. As suggested by Ram et al. (2007), values above 40% were considered highly indicative of major sources of contamination. The next main contributor of fecal contamination in the watershed was sewage; however, isolates identified as originating from sewage only accounted for 27% of the unknown isolates classified, much less than the number of unknown isolates classified as originating from cattle.

Figure 5 – Identification of water E. coli isolates using the carbon substrate utilization library.

Discussion

The research presented here served 2 main purposes. First, it allowed us to determine the relationship of E. coli isolates from various source categories and water based on carbon-utilization profiles. Second, it applied this information to determine the sources of fecal contamination in a central Texas watershed so that management practices could be considered, if necessary. This research represents one of the most comprehensive studies to date investigating the metabolic characteristics (carbon utilization) of naturally occurring E. coli populations and the subsequent application of these data for BST purposes. In addition to the creation of a large database of isolates, the study was conducted over a time frame of approximately one year, used standardized isolation techniques for all isolates prior to method analysis, and ensured that the carbon-utilization patterns generated were reproducible. Finally, fecal samples were specifically collected from a variety of hosts in the watershed to ensure the most likely potential polluters were not overlooked. For instance, our companion animal category, which included cat as well as dog feces, is one of only a relatively few BST studies to include cat samples, despite the fact that data in at least one study suggested that cats may be a more significant source of environmental fecal contamination than dogs in some areas (Ram et al. 2007).

The relationship of E. coli isolated from the various source categories was determined by constructing a UPGMA dendrogram (Fig. 3). For our dendrogramatic analysis, we anticipated that isolates from a specific source category (such as horse) would tend to be most similar to other isolates from that same source category. As mentioned previously, although isolates from each source category did not form discrete clusters, isolates from the same source did tend to clump together and were usually predominately present in only 1 or 2 clusters.

The overall separation of E. coli isolates into clusters based on source is very important information for BST. For BST to be successful at discriminating between different sources of bacterial pollution, the measured characteristic should be similar between isolates from the same source category but different between isolates obtained from the different source categories. For example, ideally, we want human isolates to be most similar to other human isolates in their carbon-utilization patterns. However, we also want human isolates to be very different from isolates obtained other source categories such as sheep or goat. This separation of groups allows isolates to be distinguished from one another based on source category. The clumping of isolates into clusters based on source in our dendrogram implies that carbon-utilization profiling may be proficient at distinguishing between source categories.

Interestingly, water isolates, although they comprised almost half of cluster H, were otherwise spread across numerous clusters and usually did not tend to clump together. This spread of water isolates over numerous clusters, which usually contained predominately isolates from other source categories, especially cattle and human isolates, could signify that some of these water isolates originated from these other source groups, as we suspected.

The second goal of this study was identification of the dominant NPS of fecal contamination in the North Bosque watershed. Previously reported blind tests have demonstrated that although the identification of all sources of contamination in environmental samples may not be possible, BST methods can reliably determine the dominant NPS of contamination (Myoda et al. 2003). Determining the dominant NPS of contamination is especially important for this watershed and to the local community because the North Bosque watershed drains into Lake Waco. Lake Waco serves as the source of drinking water for more than 200,000 Central Texans, in addition to being a site for many recreational activities (Conry 2010).

In terms of the concentration of E. coli in our water samples, E. coli geometric means of the monthly samples for all sample sites were above the USEPA and State of Texas Water Quality Standard (WQS) long-term geometric mean for contact recreation (126/100 mL; Fig. 4; USEPA 1986, TNRCC 2002). Although some of the sampling sites, such as Site N, only slightly exceeded the WQS, other sites, such as Sites C, F, and H, exceeded the WQS for this sampling period by 10-fold or greater. Sampling sites that contained the highest E. coli concentrations were all located in Erath County. In general, the E. coli concentrations decreased as we moved downstream in the watershed and toward Lake Waco. Given that long-term geometric means were all above the WQS, these sample sites are inappropriate for contact reaction. Water Quality Standards are based on epidemiological data that correlate bacterial levels with risk of contracting gastrointestinal illnesses from exposure to those bacterial levels. The elevated levels seen here suggest a potential risk for pathogen exposure, specifically in the north-westernmost portion of the North Bosque watershed. Potential recreational uses impacted by these exceedances are boating, swimming, and fishing.

Note, however, that the water sampling dates for this study only spanned 5 months. It would be useful to determine if these high concentrations of E. coli occur in the watershed year round. Additionally, although bacterial levels were lower near the inlet of the North Bosque River into Lake Waco, no water samples were taken from the interior of the lake. When considering possible impacts on drinking water quality, it would be useful to look more closely at bacterial concentrations within the reservoir. Excessive bacterial concentrations can result in increasing use of disinfection chemicals, increasing treatment costs, and potentially increasing generation of disinfection by-products. A previous study, conducted in our laboratory and spanning February 2003 to September 2004, indicated that, in the reservoir, the long-term geometric mean concentrations of E. coli did not exceed the USEPA/State of Texas WQS. However, E. coli concentrations noted in the previous study were lower overall, and the geometric mean concentrations did not exceed WQS. Therefore, it is possible that the increased E. coli concentrations that we see here are the result of increased bacterial loading into the watershed since the time of the previous study. If this is the case, the concentration of E. coli in the reservoir could have increased as well. It is also possible the high concentrations of E. coli reported here are due to other variables such as increased rain. Weather pattern plays a major role in the delivery of particulates, including bacteria, to Lake Waco (Filstrup and Lind 2010). Regardless, further investigation would be beneficial to determine if the high E. coli concentrations seen in this study occur often, if they occur consistently, and how these levels impact the E. coli concentrations in the reservoir.

For BST purposes, Jackknife analysis of the carbon-utilization library that was created from the fecal isolates (Table 2) yielded rates of correct classification within the range reported in previous published studies (Johnson et al. 2004, Moore et al. 2005, Casarez et al. 2007, Jiang et al. 2007). Additionally, the overall ARCC value calculated for this library was similar to values reported previously (Harwood et al. 2003, Johnson et al. 2004, Moore et al. 2005) but not as high as some libraries created in more recent BST studies. However, previous research has shown that ARCC values tend to be high for small scale studies where bacterial diversity is low (Wiggins 1996, Wiggins et al. 2003). It is possible that the mid-range ARCC value reported for this library is due to the large number of E. coli isolates analyzed. An important point to remember is that although the large sample size in this study likely resulted in a decrease in the ARCC value, the large sample size allowed us to better represent in our library the true diversity that is likely present in these bacterial populations. Finally, the specificity values reported for this library (Table 3) were much higher than many previously reported values (Harwood et al. 2003).

When water isolates were submitted to the library for identification, overall, cattle were determined to be the dominant source of contamination in the watershed (Fig. 5). Sewage also appeared to be a major contributor of fecal contamination, although the percent of water isolates identified as originating from sewage (27%) was less than the percent of isolates identified as originating from cattle (43%). Two additional BST studies, using different BST methods but conducted on this same set of fecal, sewage, and water isolates, confirmed that cattle were the dominant source of contamination in the watershed (results not shown). Finally, a recent study concluded by Moussa and Massengale (2008) found similar results in regard to cattle contamination in the South Bosque watershed, a smaller watershed located adjacent to the North Bosque watershed.

Although a detailed land-use analysis incorporating these data is still underway, when one considers land-usage in general around the watershed, it is not surprising that cattle were identified as the dominant category responsible for nonpoint sources of pollution. One of the major agricultural practices in the watershed includes concentrated animal feedlot operations (CAFOs), specifically housing dairy cattle (TNRCC 1999). Additionally, approximately 2% of the area in the watershed is used as dairy cattle waste application fields, with the majority of these present in the uppermost portion of the watershed (TIAER 1999), possibly explaining the increased E. coli concentrations noted in this area.

When water isolate identifications were sorted based on sampling site (results not shown), cattle were again the dominant source of contamination at each individual site excluding Site D and Site H, where sewage was the dominant source of contamination (results not shown). Site D may be partially impacted by discharge from the Stephenville wastewater treatment plant located in the area, resulting in increased contamination by sewage.

From a lake and reservoir management perspective, the ultimate goal of a BST study is to apply the information obtained regarding the potential source(s) of fecal contamination and use it to implement mitigation, best management, or remediation processes to help control or decrease contamination from major NPS. Ultimately, the goal is to increase the quality of our water.

Because E. coli concentrations in our water samples exceeded WQS and identification indicated cattle as the major contributor of fecal contamination, our first recommendation for decreasing bacterial pollution in this particular watershed would be to implement strategies to reduce the contribution of fecal contamination from this source. Some suggestions include revising rules and regulations outlined for CAFOs, including issuing and amending permits. Inspecting these facilities and enforcing compliance is also crucial. Establishing, rebuilding, and/or monitoring the riparian zone between cattle operations and waterways may help prevent or at least decrease contamination due to run-off. Education and outreach, specifically to the individuals directly involved and to the associated communities in general, will likely benefit our waterways. In terms of decreasing contamination from sewage, ensuring that wastewater treatment plants are properly operating and sufficiently upgraded could offer additional protection. Further investigation is needed to understand if there are fluctuations in the dominant source(s) of pollution across a larger range of sampling dates and, if so, if these are correlated with additional factors such as rainfall, temperature, or flow rates.

When we consider bacterial source tracking studies as a whole, we know that in the past decade a significant amount of time, research, and money has been invested into validating various BST methods. More recently, these various methods and combination of methods (sometimes called a “toolbox approach”) have been used in waterways across the United States to assist in determining the potential sources of contamination. It is now important to begin applying our findings to management strategies. Individuals included in developing management strategies have a lot to consider when it comes to BST studies. Those involved must understand that most BST methods have inherent limitations and that study design, sample size, and the sources included can all affect the representativeness and specificity of a BST study. Additionally, studies are still being conducted to assess the accuracy of BST libraries over geographical space and time and to determine their usefulness to other areas and at future dates. The introduction of library-independent BST methods has introduced yet another variable. When considering BST results, hasty management decisions should, of course, be avoided. Bacterial source tracking results, the limitations of the methods that were used, and the overall study design should be scrutinized so that valuable resources are not wasted targeting incorrect contributors of NPS pollution. However, useful information can be gained from sound BST studies, and, with the tremendous amount of research that has been dedicated to this area of science, management officials can have reasonable confidence in the results. These results can greatly assist in determining where efforts should be focused when it comes to watershed and reservoir management and remediation.

Acknowledgments

We wish to thank the numerous individuals who assisted with the collection of field samples and with the carbon-utilization profiling. This tremendous amount of data could not have been collected in a reasonable timeframe without your help.

Notes

*RCC's are shown in boldface.