Characterizing river water quality and correlating it to environmental health can be challenging. We know that bacteria measured as E. coli (Escherichia coli) and sediment have a very complex and symbiotic relationship (Salam et al., 2021). E. coli tends to increase with increasing sedimentation in streams. We also know that in watersheds predominately covered by forest, we find relatively good water quality while in watersheds increasingly plagued by human alteration such as agriculture or urban development, we find degradation (Baker, 2003; Buck et al., 2004; Tong and Chen, 2002). Sediment that is often laden with bacteria moves from deforested or disturbed areas directly into rivers. Sediment can remain for long periods of time moving slowing though these systems becoming legacy sediments (Fleming et al., 2019; Livers and Snyder, 2025). Models to describe this relationship such as the model of impervious cover (Schueler et al., 2009) or the urban cascade (Taylor and Owens, 2009) characterize this problem associating it with increasing impervious surfaces. These models are useful when correlating urban land use to degrading water quality yet difficult when trying to pinpoint direct sources of health contamination.

To pinpoint pollution, E. coli is widely accepted as the standard water health indicator (USEPA, 1986). Yet it lacks many desirable characteristics needed in an ideal indicator organism (Ishii and Sadowsky, 2008). Of particular concern is the ability of E. coli to live and reproduce in the environment becoming part of the naturalized microbial community rather than remaining an independent evaluator of pollution (Jang et al., 2017). Compounding this concern is the strong association of E. coli with sediment and the propensity to associate with stormwater with excessive suspended sediments (Jeng et al., 2005). Hence, sediments that often coat the bottom of streams and become resuspended during increased flow events naturally increasing the concentration of E. coli. This internal source of contamination confounds predictive ability needing greater research to compartmentalize what is naturally occurring and what is indicating environmental health problems.

Land use is another source of pollution. Agricultural land use is loosely associated with declining water quality (Stone et al., 2005; Cuffney et al., 2000). More specifically, agricultural land use is a specific source of river microbial loading (Tong and Chen, 2002; Stein et al., 2008; Petersen et al., 2018; Petersen and Hubbart, 2020). This land use when associated with precipitation events correlates with elevated E. coli concentrations (Haramoto et al., 2006; Pandeya et al., 2012; Rodrigues et al., 2018). Rainfall events directly after the application of manure or upon cattle-impacted lands can cause even greater elevated bacteria in receiving streams (Soupir et al., 2006; Guber et al., 2007). Croplands bring an additional source for elevated bacterial loadings (VanderZaag et al., 2010) but this is dependent upon specific land management strategies. Finally, wildlife in associated riparian forest add an additional source of contamination (Cox et al., 2005).

Considering the urban environment, pathogens (bacteria, parasites, protozoans, viruses) enter rivers via stormwater, wastewater and overflowing or leaking sanitary sewer systems (Olds et al., 2018). Stormwater over impervious surface is another source that correlates with disease risk (Arnone and Walling, 2007). Excessive water quantity erodes river banks while coating river beds with sediment and organic material creating an ideal environment for bacteria such as E. coli to survive until the next storm re-suspension. We know that sediment loading as a direct result of erosion of exposed land or streambank failures contains heavy loads of bacteria (Chen and Lui, 2017). E. coli survives in sediment much longer than in overlying waters (Pachepsky and Shelton, 2011). This sediment acts as a sink with the potential of continued bacterial contamination long after any contamination event (Mallin et al., 2007). Thus, rivers experiencing erosion and sedimentation harbor extensive beds of bacteria, potential pathogens and other pollutants that resuspended continually when waters rise from storms.

Further, rivers have been historically altered by dams (Walter and Merritts, 2008). Each of these structures is characterized by a particular volume of water impounded and sediment stored through decades of impoundment. The current trend of dam removal while admirable and with certain positive attributes is also unleashing legacy sediments into river systems (Livers and Snyder, 2025). This sediment is eventually processed, distributed and washed through into bays and oceans. Hence, adding additional sediment stored behind dams to a river system already burdened by sediment and land use that increases erosion severely compounds an already difficult problem (Shahady and Cleary 2020).

While impact of land use and river contamination can be documented, the translation of bacterial contamination into public health risk and disease is more elusive. Using climate and epidemiological records, Rose et al. (2000) found statistical evidence suggesting a correlation between storm events and disease outbreak in cities. DeFlorio-Barker et al. (2018) estimated recreational waterborne illness on United States surface waters is significant and costly. They estimate 4 billion surface water recreation events occur annually, resulting in an estimated 90 million illnesses with a cost of $2.2- $3.7 billion annually. Illnesses of moderate severity (suggested by a visit to a healthcare provider) are responsible for over 65% of the economic burden while severe illnesses (resulting in hospitalization or death) are responsible for approximately 8% of the total economic burden.

In response, criteria for contamination levels to minimize disease risk are formulated (Kay et al., 2004). In these scenarios, researchers examined how successive increases in fecal bacterial concentrations generated a concurrent risk of illness. Water low in fecal contamination (FC) or a concentration of <40 FC per 100 mL generated a minimal risk or less than 1% chance of contracting some form of illness. Subsequent incremental increases in fecal contamination are studied and quantified. Concentrations greater than the 500 FC per 100 mL are associated with (>10%) potential illness rick.

Translating these established risks into credible regulatory standards produced our current regulatory framework (USEPA, 1986). The United States Clean Water Act calendar month geometric mean standard of 126 E. coli (CFU per 100 mL) is suggested to provide a minimal disease risk of less than 1% (Dufour and Ballentine, 1986). Further research suggested that maintaining E. coli below 200 CFU per 100 mL as a geometric mean and lower than 400 CFU per 100 mL as an instantaneous standard helps keep the rate of gastrointestinal illness from exposure to approximately 1%–2% (Tobin and Ward, 1984). These microbiological standards are not based on robust epidemiological data but represent best estimates of associated risk (Kay and Fawell, 2007). The best preventative approach is to maintain bacterial and pollutant concentrations as low as possible.

While tracking E. coli provides indication of the total microbial loading into a water source and the risk for disease, advances in Microbial Source Tracking (MST) now provides even greater detail into the sources for this contamination. Microbes from the genus Bacteriodes are used to pinpoint human and other animal origins of the contamination (Simpson et al., 2002; Job et al., 2023). This technique provides greater specificity when tracking environmental health concerns but is much more expensive potentially limiting its effectiveness as an approach.

The problem of sedimentation has become so pervasive that river flow is always accompanied by some level of turbidity. The removal of dams, alterations to river water courses, fire, flooding, farming, urbanization and deforestation all contribute to continued sedimentation of water resources (James, 2019). Compounding this are alterations in precipitation driven by changes in climate. This creates what many authors have described as a “wicked” problem (Shortle and Horan, 2017). Our freshwater streams and rivers flow through an altered human landscape driven by an intensifying water cycle (Wohl, 2019).

One approach to tackle such problems is an understanding of sources and sinks along with areas of intensification or “hot spots” (Fleming et al., 2019). Once identified, restoration may occur. Because these hot spots and legacy sediments are pervasive in watersheds, the spreading management expenses and fixes throughout a region may not produce any observable improvements other than money spent. Further, measurable results may take upward of 20+ years to be realized or never realized at all due to the slow movement of these sediments (Melland et al., 2018; Livers and Snyder, 2025). The transport longevity of sediments in river systems is poorly understood and studied (Livers and Snyder, 2025).

Legacy pollutants that exist in streambank soils and throughout the stream bed may be the primary polluting material now observed during storm events. Intense storms work to activate these pollutants by resuspending sediment and pulling deposited material from stream banks back into the water. With each storm, sediment moves down the channel while upstream erosion resupplies more and more easily movable soil. As we know, these sediments harbor bacteria that are activated by water movement. Hence, sediment hotspots in any river may be the primary problem in stream channels causing most of the problems. This disconnect between watershed management practices and observable stream water quality improvement may only be resolved when systems flush existing sediments reaching an equilibrium.

Our investigations were initiated in the summer of 2018 to evaluate water quality conditions in the Pigg River, an important tributary to Leesville Lake. This lake is an important recreational, environmental, and power generation reservoir in Central Virginia. While the initial focus is on water quality, greater focus and scrutiny began in 2019 to determine origins of the bacterial contamination (assessed by E. coli concentrations and later MST) discovered throughout the 2018 study. In 2020, we intensively evaluated water quality in areas of urban, agricultural and forested land use to determine the impact of these as sources of contamination. During 2021–2023, we experimented with bottom sediment disturbance and stormwater flow to better quantify sources of bacteria. Throughout the study, the importance of sedimentation from both dam removal and legacy accumulation are evaluated to determine impact on bacterial contamination.

For this study, we sampled multiple sites along the main stem of the Pigg River (Figure 1) in central Virginia. Each river site is chosen for accessibility to facilitate direct sampling for bacteriological and chemical analyses and various water uses along the river.

In 2018, we conducted a comprehensive sampling of the river establishing stations and analyzing data. In 2019–2020, we continued to monitor the river while adding potential “hot spots” to include Microbial Source Tracking (MST). A hot spot is considered a sampling station where excessive concentrations of E. coli are measured using the Colilert methodology. Excessive is considered over 1,000 counts/100 mL during low flow conditions and in excess of 2,400 counts/100 mL during storm conditions. The excessive classification for hot spots is determined through observations of the data collected in 2018 along with statistical analysis examining clusters in the data.

Additionally, in 2019 we sampled during low flow and storm flow conditions. Low flow is defined as conditions with good water clarity (turbidity <1 NTU). Stormwater flow is defined as conditions where greater turbidity (>10 NTU) and appreciable rain (>2o mm) occurring in the watershed within 24 h of sampling. Low flow sampling is designed to pinpoint exact location of bacterial contamination. The second sampling is designed to identify contamination during a stormwater event.

In 2020, we continued sampling of the river and increased our efforts to identify hot spots near Rocky Mount Virginia. In 2021–2023 we concentrated efforts on the lower Pigg River trying to quantify the potential areas impacted most significantly by sedimentation and bacteria. We also conducted two separate disturbance experiments to determine the impact of bottom sediments on water quality when disturbed.

The Pigg River, located in south central Virginia, is named after an early settler John Pigg who laid claim to 400 acres of land opposite the mouth of Snow Creek (Clement, 1929). Throughout its history, the river has drained predominately forested land but more recently land clearing for agricultural production has increased. By best estimates, the watershed currently consists of 65% forest, 26% agriculture (crops and cattle) and 5% urban development (Ries et al., 2008; Benham et al., 2006).

Each river segment is characterized using aerial maps and quantification of land use adjacent and directly along the river. Designated land use is characterized where land abutted the river out to 100 m. Using the aerial maps, forest vs. pasture or other use is easily identified. Each land use is measured and then quantified using the following criteria to determine a land use type:

Minimal Impact - > 50% land use forested and buffered along reach. Buffer minimum of 50 feet. Minimal impact from pastureland use

The Dam sampling location is at the confluence of Power Dam Road and Pigg River (GPS coordinates: N 36° 59′45.9″, W 79° 51′36.2″) just below the city of Rocky Mount (Figure 1). This site is selected for sampling due to its close proximity to the site of a dam removal project in 2017 (Figure 2) and its accessibility via Power Dam Road. This site directly drains the city of Rocky Mount.

The Chestnut Hill sample site is located at the intersection of Chestnut Hill Road and Pigg River (GPS coordinates: N 37° 00′11.5″, W 79° 49′34.1″). Powder Mill Creek joins Pigg River directly upstream of this site. The 7.9 km river segment upstream of the Chestnut Hill Site has been classified as having major impact due to extensive agricultural use and lack of significant buffer (Table 1). A Power Dam is removed in 2017 above this site.

The Colonial Sampling Site is located at GPS coordinates (N 36°56′24.9″, W 79°46′03.0″). Doe Run Creek, a small tributary, joins Pigg River upstream of this site, potentially impacting water quality. The 14.9-mile river segment upstream of this site is classified as minimal impact (Table 1).

Truevine sampling site is located at the intersection of Truevine Road and Pigg River (GPS coordinates: N 36°56′05.6″, W 79°43′01.2″). The 5.1-mile river segment upstream of this site contributes minimal impact (Table 1). Potential contributors for water quality at this site include Big Chestnut Creek and Walker Creek.

The Snow Creek Sampling Site can be found at the intersection of Snow Creek Rd and Pigg River (GPS coordinates of N 36°56′40.1″, W 79°38′18.1″). The 10.5-mile river segment upstream of this site is known to have mixed impact closer to Truevine and significant impact near Museville (Table 1). This segment is characterized by the presence of numerous tributaries that converge into Pigg River.

The Museville Sampling Site is located off the Museville Bridge (GPS coordinates: N 36°56′02.9″, W 79°35′40.0″). The upstream segment of the river, extending 5.6 km from this location, has been determined to have a significant impact (Table 1).

The Toshes Sampling Site is located at the intersection of Pigg River and Toshes Road (GPS Coordinates: N 36°59′19.9″, W 79°30′56.1″). The 17.5 km upstream segment of the Pigg River from this site has been determined to have been significantly impacted. This section of the river is characterized by the presence of numerous large tributaries that converge into the Pigg River.

Toler Sampling site is located near the confluence of the Pigg River and Leesville Lake (GPS Coordinates: N 37°00′30.2″, W79°28′41.4″). The 5.6 km upstream segment is determined to have low impact (Table 1). This site has road access closest to the mouth of the Pigg River and is utilized to estimate the quality of Pigg River water entering Leesville Lake.

Water is collected from each bridge crossing either directly by filling water bottles and placement of the YSI multiprobe into the water or collection from the bridge using a collection bottle. After collection, water for laboratory analysis is immediately transferred to acid washed bottles (nutrients) and sterilized bacteriological bottles and stored in a cooler until analysis in the laboratory. E. coli samples are immediately processed upon return to the laboratory while nutrient samples remained refrigerated until analysis within 30 days of collection. Remaining water is analyzed using a YSI multiprobe and Turner Turbidimeter.

Water quality data is obtained using a YSI 556 multiprobe meter (Xylem, Yellow Springs, Ohio) following pre and post calibration QA/QC procedures in accordance with EPA protocols (EPA, 2017). E. coli is quantified using Colilert-18 (IDEXX, Westbrook, Maine) meeting all EPA standards for testing (Warden et al., 2011). This methodology uses MPN to quantify E. coli, reported as MPN/100 mL of sample in accordance with federal and state standards. Total phosphorus (TP) samples are collected in acid ished Nalgene bottles and analyzed analytically using an EasyChem auto analyzer (Systea Analytical Technologies). The EasyChem analysis is compatible with Ascorbic Acid Total Phosphorus Analysis detailed in Standard Methods for Analysis of Water and Wastewater (Baird and Bridgewater, 2017).

For the disturbance experiments, stream segments believed to be impacted by the dam removal and laden with sediments are studied under 3 types of flow conditions; baseline, disturbed bottom and stormwater. Under baseline flow, water samples are collected directly from the stream surface under low flow conditions avoiding any sediment. After the baseline collection, several individuals are deployed above this same sample location and began kicking sediment dislodging it creating flow conditions similar to stormwater flow until turbidity is greater than 50 NTU. Samples are then collected with similar methodology as baseline flow yet these samples contained the bottom disturbed sediment flowing in the river. Stormwater samples are collected during a defined bankfull (Rosgen, 1996) storm event.

Water for bacterial source tracking is collected into sterile bottles and shipped overnight in cooled ice chests for analysis (Source Molecular Corporation, Miami Lakes, FL). There, water samples are filtered through 0.45 micron membrane filter(s) then placed in a separate, sterile 2 mL disposable tube containing a unique mix of beads and lysis buffer. The sample is homogenized for 1 min and the DNA extracted using the Generite DNA-EZ ST1 extraction kit (GeneRite, NJ), as per manufacturer’s protocol.

Amplifications to detect the target gene biomarker are run on an Applied Biosystems StepOnePlus real-time thermal cycler (Applied Biosystems, Foster City, CA) in a final reaction volume of 20 µL sample extract, forward primer, reverse primer, probe and an optimized buffer. Primers employed are for generation of amplicons for human, ruminant and cattle markers. All assays are run in duplicate. Quantification is achieved by extrapolating target gene copy numbers from a standard curve generated from serial dilutions of known gene copy numbers. For quality control purposes, a positive control and a negative control, are run alongside the sample(s) to ensure a properly functioning reaction and reveal any false negatives or false positives.

The code of Virginia (promulgated through EPA and the Clean Water Act), 9VAC25-260-170 outlines the following standards for the waters of Virginia. Bacteria criteria (counts/100 mL) shall apply to protect primary contact recreational uses in surface waters. In freshwater, E. coli bacteria shall not exceed a geometric mean of 126 counts/100 mL and shall not have greater than a 10% excursion frequency of a statistical threshold value (STV) of 410 counts/100 mL, both in an assessment period of up to 90 days.

Geometric means shall be calculated using all data collected during any calendar month with a minimum of four weekly samples. If there is insufficient data to calculate monthly geometric means in freshwater, no more than 10% of the total samples in the assessment period shall exceed 1173 E. coli CFU/100 mL. If there is insufficient data to calculate monthly geometric means in transition and saltwater, no more than 10% of the total samples in the assessment period shall exceed 519 enterococci CFU/100 mL.

The general trends observed in the study are initially displayed using descriptive statistics, including means and standard errors for all observations. The number of observations varied depending upon samples taken in the river and Leesville Lake. In general, the Pigg River is sampled three times each year (2018–2020) and twice per year (2021–2023) primarily late summer to early fall.

Two types of regression analysis are used to create a best fit model to determine predictors for all of the data collected. A partial least squares regression analysis examined the best fit or best predictors for E. coli and ruminant markers based upon all collected parameters. From these findings, two linear regressions are created from identified parameters as the final statistical tool to model relationships in the watershed.

It is clear from the collected data that the Pigg River is significantly impaired (Table 3). Our bacteriological results suggest the river continually contains concentrations of E. coli in exceedance of 9VAC25-260-170 throughout the 6-year study period. During each sampling event, the Geometric Mean (GM) exceeded the 126 CFU/100 mL standard for a violation rate of 100%. Violation of the STV standard of 410/100 mL occurred at a rate of 75% in our study where the standard calls for 10% or less. In instances where the river enters flood stage, the concentrations of E. coli exceed 100x the allowable standard. The river is severely impaired by bacterial contamination.

To examine possible host sources for bacteria, we analyzed genetic markers contained in Bacteroides (Table 3). We found a majority of Bacteroides bacteria found across all sites are ruminant in origin and the predominant source most notably storm events. Bovine and human markers are variable but became a greater proportion of the samples when E. coli measures are lower at base flow conditions. Looking all data over all years, bacterial increases are considerable during storm events and predominantly from ruminant sources.

To determine how MST measures are associated with other water quality parameters, both E. coli and the ruminant concentrations are correlated with all of the other parameters (Figures 3, 4). This analysis correlates all parameters to one selected variable to determine how each is related to the other variables. It is generally accepted that a value greater than 1 is significantly related to the tested parameter. Hence, with this analysis we can look further into the relationship between all of the parameters for the prediction of water quality.

The PLS-VIP analysis using E. coli suggested turbidity, TP and Bovine MST are significantly correlated to the concentrations of E. coli (Figure 3). This supports the idea that stormwater runoff is a significance source of the E. coli found in the river. While turbidity and TP may have multiple origins in this watershed, bovine markers are specific to agriculture production and by association from this analysis suggestive of impact on E. coli measured. What is not understood from this analysis is the low correlation with ruminant markers. These markers while significantly increasing in abundance during storm event are not associated with other parameters such as turbidity and TP.

To further explore what may be generating the high ruminant markers during storm events, a PLS-VIP analysis is conducted comparing ruminant measures with the other parameters. Results suggested that Bovine MST concentrations and conductivity are the only related parameters (Figure 4). As Bovine MST concentrations are a subset of the ruminant MST, and conductivity unrelated to the other parameters, it is interpreted that ruminant MST concentrations did not provide any interpretive value at least for a specific source in the analysis. This measure appears to be too generalized to pinpoint a source of bacterial contamination.

To determine the importance of land use throughout the watershed and interpreting the potential impacts outlined in Table 1, significant parameters identified in PLS-VIP are averaged and plotted along an axis from Toshes Road to Chestnut Hill (Figure 5). These stations begin closest to Leesville Lake (Toshes Road) up to Rocky Mount (Chestnut Hill). This analysis strongly suggests that greater concentrations of contamination occur in the upper portions of the river near Rocky Mount, and primarily in the Colonial section of the river. Only turbidity measures do not directly support the conclusion that Colonial is the most contaminated site among those sampled yet all other parameters did suggest contamination is greater in this section of the river.

The concentration of pollutants at the Colonial station is suggestive of internal loading in the river exacerbating observed conditions. Chestnut to Colonial is of mixed impact based on land use (Table 1) and nothing in the surrounding area is suggestive of increased pollution when compared to any other area in our analysis. To further understand this idea, we conducted the disturbance experiments. This hypothesis tested water in the river at baseline flow and then after it is manually disturbed to mimic stormwater conditions. Additionally, stormwater (precipitation event that elevated river flow and turbidity) is also measured.

The disturbance experiments supported the hypothesis that bedload or legacy sediments that coat the bottom of the river carry a pollutant load that when disturbed change water quality (Figure 6). In all instances, we increased levels of measured pollutants represented by TP and E. coli when we disturbed the sediment. This suggests a flow event that disturbs the bottom sediments adds pollutants to the water irrespective of surrounding land use or incoming water flow. Additionally, it is apparent that surrounding land use adds even greater pollutant loads beyond what we could simulate with our disturbance experiments. While bedload sediment adds pollution though river flow and mechanical disturbance, overland flow and streambank disturbance are important contributors as well as seen during storm events.

Additional evidence supports the idea that legacy pollutants and sediments can create “hot spots” that pollute the river at greater rates during storm events. It is clear from this experiment that the disturbance of sediment on the stream bed produced higher concentrations of E. coli at the Colonial site during both years we conducted the experiment (Figure 6). Expectations based on land use suggest Chestnut Hill should be the most impacted site. The removal of the Power Dam and subsequent sedimentation in this section of the river is the suspected source of this contamination due to river flow and deposit of those sediments.

The bacterial characterization of samples from the experiments demonstrated stark differences among identified markers (Figure 7). Human, ruminant and bovine markers are found in all of the samples. Human markers are found in every sample and are in greatest concentration at Chestnut Hill site. This is consistent with land use and it is likely from the houses that line the river and wastewater discharge from the treatment facility for the city of Rocky Mount.

Ruminant markers dominated observations with greatest concentrations in the Colonial area. As we could not get an increase in these ruminant markers when we disturbed the sediment yet stormwater flow caused significant increases the observation suggests this marker is watershed driven and from storm events. Bovine markers are only found in much lower concentrations and only during storm events at both Colonial and Chestnut sites. This suggest this measure is of minimal importance only occurring in isolated areas and during rain events.

Further, greatest observations for ruminant and bovine markers during storm events occurred at the Colonial site. As we could not stimulate any of the markers in our disturbance experiments, it is assumed markers are derived from external flow. Land use not identified in our study may be the source at the Colonial site. But, we did see E. coli increases in our experimentation (Figure 6) consistent with the idea that internal loading is causing increases in this parameter and other pollution parameters. This suggests, that bedload sediment driven by various flow rates in rivers may drive concentrations of E. coli in river water a majority of the time. And the greater the concentration in the bed-load sediments the higher the elevation in overlying water. This seems to exist during most flows with the exception of storm events that exceed 1–2 years events.