Concerns of microbial contamination of waters have increased as a result of documented waterborne disease outbreaks (Leclerc et al., 2002). Impairment of water quality due to the presence of pathogens is typically assessed by monitoring for microbial indicators such as total coliforms, fecal coliforms, and Escherichia coli. Indicators have been associated with enteric pathogens present in water from human or animal fecal contamination (e.g., Hörman et al., 2004). However, the reliability of the use of indicators to predict pathogens has been challenged by some studies (Hegarty et al., 1999; Schets et al., 2005). As such, researchers have been developing new methods for the direct detection of pathogens in water samples (e.g., Guy et al., 2003) and have been studying pathogen fate and transport characteristics (e.g., Jiménez et al., 1989).
Most investigations into pathogen occurrence in watersheds have focused only on the water column and have seldom considered interactions with sediments. Sediments can affect the transport and survival of microorganisms in natural waters in several ways: (1) sediments at the bottom of rivers, lakes, and estuaries may serve as reservoirs of viable indicator bacteria and enteric pathogens; (2) sediments and sediment particles can affect concentrations and transport processes of pathogens in the water column – they can be adsorbed to sediment particles and settle out of the water column more rapidly. These are important processes for microbial transport and removal in waters.
The types of particles have a great impact on the settling velocity. Generally, denser inorganic particles will settle out more quickly than others. The microbial adsorption to settleable particles is different for different types of microorganisms and the adsorption behavior of each organism changes between dry and wet weather (Characklis et al., 2005). The release of bacteria during sediment resuspension caused by storms, flood, tides, or strong winds will result in high concentrations of bacteria in the water column (Muirhead et al., 2004; Jamieson et al., 2005). Lee et al. (2006) reported that the peaks of E. coli and enterococci levels in water and sediment were consistent with storm activity in the beaches which were exposed to fecal contaminants. Studies have shown that indictor organisms have a tendency to survive longer in sediments than in natural water (Burton et al., 1987; Davies et al., 1995). The persistence of pathogens in sediments may create a longer period of risk to public health.
The development of computer-based simulation models are needed for the characterization of the variability of emerging microbial contaminants in relation to watershed conditions. Recently, many process-based models for fecal indicator bacteria and/or pathogens have been created and include hydrologic (Ferguson et al., 2005; Hayden and Deletic, 2006, 2007), hydrodynamic (McCorquodale et al., 2004; Kashefipour et al., 2006; Bai and Lung, 2006) and simple reservoir models (Antenucci et al., 2005). Most models have been process-based, although stochastic approaches have also been proposed (Yeghiazarian et al., 2006). Identified research needs for an improved understanding of microbial contaminant fate and transport include: (1) the acquisition of experimental data, (2) better quantification of potential sources of microbial contaminants, (3) improved understanding of the effects of particulates, (4) characterization of the relation of microorganism fate and transport parameters to environmental variables, (5) the evaluation of tracers with similar transport behaviors, and (6) the development of process-based models (Pachepsky et al., 2006).
In this study, we investigated the microbial contamination in sediment and water samples from the Blackstone River Watershed. The objective was to illustrate the role of sediments in the fate and transport of potential pathogens during wet weather events. The specific goals were: (1) to quantify E. coli densities and identify E. coli associated with bed sediments as sources of downstream bacterial contamination, and (2) to improve predictions of microbial fate and transport by including sediment resuspension in a hydrologically based microbial contaminant fate and transport model. The research approach involved wet weather and dry weather sampling, identifying E. coli ribotypes present in samples, and augmenting the WATFLOOD model to include processes of sediment resuspension and loading from urban, impervious surfaces.
Watershed and Land Use Characteristics
The area of this study is the Upper Blackstone River Watershed (Figure 1), a subwatershed of the Blackstone River encompassing an area of 186.7 km2 (72.1 mi2). The Blackstone River originates in the Worcester hills in central Massachusetts, and flows southeasterly into Rhode Island, discharging eventually into Narragansett Bay. It is known as the “Birthplace of the American Industrial Revolution.” With a steep gradient [the river drops 134 m (438 ft) over 74 km (46 miles)] the hydraulic potential of the river was first tapped by Samuel Slater who built a mill at the outlet of the Blackstone in 1793. Others followed, which led to one dam for nearly every mile of the river (Shanahan, 1994; BRNHC, 2006). The industrial history and population density have contributed to numerous water quality issues that include urban runoff in the headwaters where the population density is greatest, and contaminants such as nutrients, toxics, and heavy metals trapped in sediments behind the many impoundments.
Another potential source of water quality impairment is the Upper Blackstone Water Pollution Abatement District, a 212 million liter per day (56 million gallons per day) activated sludge treatment plant serving the city of Worcester and surrounding communities. The facility treats on average approximately 146 million liters per day (38.5 million gallons per day) with average reported effluent concentrations of 5.0 mg/l of biochemical oxygen demand and 7.2 mg/l of suspended solids (UBWPAD, 2006).
Most of the Blackstone River Watershed is covered by till and sand and gravel (glacial outwash) deposits. The upland areas are predominantly covered by till, which covers approximately 71% of the watershed and stream valleys are generally underlain by stratified, well-sorted sand and gravel deposits (Barbaro and Zarriello, 2006). Land use in the Upper Blackstone River Watershed is 41.5% forest, 29.9% residential, 8.7% open land, 4.4% industrial, 4% water, 3.6% commercial, 3.6% transportation, 3.5% agricultural, and 0.8% wetland. The entire Blackstone River Watershed is urbanized with a population density of approximately 386 people/km2 (1,000 people/mi2) based on the 2000 census data (USCB, 2002; Mangarillo, 2006).
Hydrologic and Water Quality Information
Monthly mean discharge in the Blackstone River at the outlet of the Upper Blackstone River Watershed (at the USGS gauging station 01109730 in Millbury, Massachusetts) (lat 42°11′20′′N long 71°45′56′′W) for the period of study (September 2005 to November 2006) ranged from 2.09 m3/s [73.9 cubic feet per second (CFS)] in September 2006 to 16.9 m3/s (596.9 CFS) in October of 2005. Peak streamflow occurred on October 15, 2005 where gauge height reached 3.584 (11.76 ft) and the daily mean discharge was 127 m3/s (4,500 CFS). The Blackstone River at the Millbury, Massachusetts station has an average depth of approximately 1 m, channel width of 18 m, and average velocity of 0.76 m/s (USGS, 2008).
Sediment and surface water samples were collected from three sites – MR02, BS04, and BS01 (Figure 1) for five wet weather events (November 2005, May 2006, June 2006, October 2006, and November 2006) and three dry events (September 2005, September 2006). Total suspended solids (TSS) for the dry weather events ranged from below the detection limit (2.50 mg/l) to 4.60 mg/l at Site BS04. For wet weather events at Site MR02, TSS ranged from 4.0 mg/l to a maximum of 17 mg/l with a mean of 8.3 mg/l. At Site BS04, TSS ranged from 3 mg/l to 46 mg/l with a mean of 13.7 mg/l. Site BS01 had TSS samples that ranged from below the detection limit to 106 mg/l with a mean of 18.8 mg/l.
A wastewater treatment plant is located immediately downstream of Site BS04. At times, sediment samples were not collected at specific sites due to excessive flooding. Composite sediment samples were collected up to a depth of 3 cm using long glass pipettes (May 2006 event) and subsequently using sterile syringes (September 2006, October 2006, and November 2006 events) and placed in small sterile bottles. The composite samples were created by combining five samples for each site to account for spatial variability. All samples were placed on ice in a cooler and immediately sent for analysis. Analysis was done within four hours of sample collection.
Bed sediments were characterized at the three sites with regards to particle size and were determined to be predominantly coarse sand (0.5-1 mm) to very fine gravel (2-4 mm) (Wu, 2007). E. coli was enumerated using the Colilert method (IDEXX, Atlanta, Georgia). Water samples were diluted with PBS buffer (phosphate buffered saline pH = 7.2) by 10 and 100 times except for the samples collected on May 12, 2006, which were diluted by 100 and 1,000 times. The Colilert media and 100 ml of the diluted samples were placed into vessels and mixed. The mixtures were poured into Quanti-Tray®/2000, sealed in a Quanti-Tray® Sealer (IDEXX, Atlanta, Georgia) and incubated for 24 hours at 35 ± 0.2°C. Sediment samples were analyzed according to the method presented by Craig et al. (2002). Twenty-five grams of sediments (wet weight) were placed into 75 ml of 0.1% peptone solution. Sediments were shaken and resuspended in solution to separate bacteria from the particles. Ten milliliters of the resuspended solution was placed into vessels and diluted 10 and 100 times. An additional 10 ml were used for measurement of both wet and dry weights. The E. coli densities in the sediments are reported per dry weight throughout this study. During the experiments two control strains were used: (1) E. coli ATCC 29194 (Remel Europe, Ltd., Dartford, United Kingdom), as a positive control, and (2) Klebsiella pneumoniae ATCC® 33495TM (Quality Technologies Ltd, Newbury, California) as a negative control.
To isolate E. coli from the Quanti-Trays, a syringe was used to remove 0.5 ml liquid from the fluorescent wells. The liquid was transferred to a tube with 10 ml PBS buffer, mixed, and serially diluted three times at 10−1, 10−2, and 10−3, as described by Eckner (1998). Then, 0.1 ml liquid was spread onto the Luria-Bertani agar (LB-agar) plates. After half an hour, plates were inverted and placed in an incubator at 35°C for 12-24 hours (depending on their growth status, allowing for a colony to grow to at least 1 mm in diameter for inoculation). A colony from the plate was visually selected and placed in a test tube containing inverted Durham tubes with EC broth medium as described by Feng and Hartman (1982). The tubes were put into an incubator at 35°C for 24 hours. The tubes with gas production were then observed under UV light at 365 nm, to identify tubes with fluorescence to confirm the presence of E. coli. Subsequently, 0.1 ml of liquid was extracted from the positive tubes and diluted 10 and 100 times in PBS buffer. The diluted liquid (0.1 ml) was then spread on an LB agar plate. The plates were incubated at 35°C for 12-24 hours. A single colony was used to inoculate an LB agar tube. After inoculation, the LB agar tube was incubated at 35°C for 12 hours until E. coli growth was visible with the naked eye. The tube (E. coli slant) was removed from the incubator, sealed with paraffin to prevent microbial contamination and stored at 4°C.
The ribotyping was carried out by the New York City Department of Environmental Protection (NYC DEP) using the RiboPrinter® Microbial Characterization System (DuPont Qualicon, Inc., Wilmington, Delaware). This system uses Riboprint pattern similarities generated by restriction fragment length polymorphisms of 16S ribosomal RNA genes from sampled E. coli and compares them with patterns of other E. coli from known hosts in the DuPont Identification library and a custom library. Upon arrival at the NYC DEP E. coli isolates were sub-cultured into Brain Heart Infusion agar. Escherichia coli was heat lyzed to make the sample available for processing and the DNA was digested by either one or two restriction enzymes. In this study, restriction enzymes EcoR1 and PvuII were used. DNA was separated by size on pre-cast agarose gels via electrophoresis. The DNA and fragments were subsequently transferred to, and immobilized on, a nylon membrane. On each membrane, denatured DNA fragments were hybridized with chemically labeled E. coli rRNA. The membrane was then rinsed and treated with blocking buffer and an anti-sulfonated DNA antibody/alkaline phosphatase conjugate. Unbound conjugate was removed and ultimately a chemiluminescent substrate was used. The procedure allowed each electrophoretic band containing the rRNA genes’ genetic information, to be observed by the custom camera in the RiboPrinter® system. Images from the membrane were captured, digitized, and then transferred to the workstation where it is processed through a series of proprietary algorithms (RiboPrinter® Data Analysis User’s Guide, 1999). The ultimate output was a set of RiboPrint® patterns for each isolate. The bacteria were ultimately identified based on how similar the sample patterns were to the source specific patterns in the Custom Identification library. Similarity is assumed according to band position, weight, and intensity. Isolates were only categorized into the same ribogroup if their genetic (riboprint) pattern matched other riboprints by 90% or greater similarity (Price et al., 2002).
Microbial fate and transport modeling was performed using modeling code developed by Dorner et al. (2004, 2006). The fate and transport model was linked with the WATFLOOD/SPL9 modeling system for simulating watershed hydrology. The distributed hydrologic model uses a grouped response unit approach for watershed representation and simulation. A detailed description of WATFLOOD/SPL9 is available elsewhere (Kouwen and Mousavi, 2002; León et al., 2001). The microbial fate and transport model included routines for simulating overland transport, movement through soil layers, and channel routing and sedimentation of free and attached microorganisms. The model was developed for simulating and predicting peak pathogen concentrations and was first tested on a predominantly agricultural watershed (Dorner et al., 2006). Dorner et al. (2006) hypothesized that including the process of sediment resuspension would improve the correlations between simulated and observed results because the rapid rise of E. coli during storm events were most likely from in-stream sources.
The Dorner et al. (2006) was therefore upgraded to improve simulations of the processes governing sediment resuspension during periods of increased streamflow. Resuspended microorganisms were estimated using a simple power function
where R is the density of resuspended microorganisms (#/m3), Q is streamflow (m3/s), and a and b are calibration parameters.
Sediment sources were assumed to be unlimited based upon the sediment monitoring results and thus resuspension was modeled as a function of streamflow and not sediment concentration. Fate and transport processes and model parameters (see Table 1) are described in greater detail by Dorner et al. (2006); however, the fraction of E. coli sorbed to flocs was changed to 50%, which is in accordance with modeling studies by Ferguson et al. (2007). In addition, a base loading function was added to simulate the base load during dry weather. The land surface load was selected based upon estimates obtained by Dorner et al. (2004, 2006) and the observation that an urban watershed had similar E. coli densities as an adjacent agricultural watershed (Dorner et al., 2007).
|Inactivation in soil, ks (l/h)||5 × 10−3||E. coli inactivation in soil for spring and fall temperatures|
|Inactivation in water, kw (l/h)||2 × 10−2||E. coli inactivation in water for spring and fall temperatures|
|Erodibility, d (#/joule)||4 × 10−3*||Number of microorganisms detached from the soil surface as a result of energy from raindrops and overland flow|
|Free-floating settling velocity, vs (m/h)||1 × 10−4||Settling velocity of unattached E. coli|
|Attached settling velocity, vsa (m/h)||2 × 10−1||Settling velocity of E. coli that are attached to flocs|
|Fraction of attached microorganisms, Fa (percent)||50%||Fraction of E. coli that are attached to flocs|
|a, b||100; 4.5*||Parameters for estimating E. coli resuspended from bed sediments|
|α1, α2, α3 (unitless)||8 × 10−3; 8 × 10−3; 8 × 10−3*||Parameters describing flow of E. coli through the soil, with interflow or artificial drainage systems and to deeper ground water|
|Base load, Lb (#)||3 × 108*||Number of E. coli that are added to the base flow in each grid for the time step (time step = 1 hour)|
|Wastewater load, Lww (#)||1 × 101||Number of E. coli that are added to the stream below the wastewater treatment plant each time step (time step = 1 hour)|
|Land surface load, Ls (#/ha)||1 × 1018||Number of E. coli that are deposited on the land surface in the watershed|
ArcGIS 9.2 (ESRI, 2006) software was used for GIS analysis for preprocessing of the hydrologic model input data files. The base maps (shapefiles), including the Blackstone River Watershed were acquired from MassGIS. The elevation data needed to simulate the water balance were acquired from Digital Terrain Model downloaded from MassGIS website (http://www.mass.gov/mgis/). The studied area was divided into 226 grids, each with an area of 1 km2. In each grid, the elevation of the center was assumed to represent the elevation of the grid. The slope, contour, and aspect were calculated using the Spatial Analyst tools in ArcGIS. The flow direction was calculated using the hydrology function in the Spatial Analyst tools and used to create input files for the WATFLOOD/SPL9 model.
Land use data for the Blackstone River Watershed was also obtained from MassGISA total of 17 distinct land uses exist within the modeling area. The 17 land uses were then classified into the following categories each having unique hydrologic parameters for the model: urban impervious, water, wetland, forest, cropland and pasture, urban open. Hydrologic parameters in the WATFLOOD/SPL9 model are primarily a function of land class.
The hydrological component of the model was calibrated first using the month of November 2005 for calibration and other months for validation. Hourly hydrometric data from USGS gauging station 01109730 at Millbury (lat 42°11′20′′N long 71°45′56′′W) were used for model calibration and validation. Watershed hydrology was simulated for months that included events sampled for E. coli and other water quality parameters. Hourly precipitation and air temperature data were available from the National Climate Data Center for Worcester airport, located within the upper Blackstone River Watershed (lat 42°16′N long 71°53′W). Model calibration was performed by trial and error, varying parameters one at a time to reduce the error between observed streamflow and model predictions. The hydrologic component of the model was calibrated prior to the calibration of the pathogen transport model. Parameters for the microbiological model were then also calibrated for the November 2006 event and validated using results from other events.