Microbial Water Quality Conditions Associated with Livestock Grazing, Recreation, and Rural Residences in Mixed-Use Landscapes

Contamination of surface waters with microbial pollutants from fecal sources is a significant human health issue. Identification of relative fecal inputs from the mosaic of potential sources common in rural watersheds is essential to effectively develop and deploy mitigation strategies. We conducted a cross-sectional longitudinal survey of fecal indicator bacteria (FIB) concentrations associated with extensive livestock grazing, recreation, and rural residences in three rural, mountainous watersheds in California, USA during critical summer flow conditions. Overall, we found that 86% to 87% of 77 stream sample sites across the study area were below contemporary Escherichia coli-based microbial water quality standards. FIB concentrations were lowest at recreation sites, followed closely by extensive livestock grazing sites. Elevated concentrations and exceedance of water quality standards were highest at sites associated with rural residences, and at intermittently flowing stream sites. Compared to national and state recommended E. coli-based water quality standards, antiquated rural regional policies based on fecal coliform concentrations overestimated potential fecal contamination by as much as four orders of magnitude in this landscape, hindering the identification of the most likely fecal sources and thus the efficient targeting of mitigation practices to address them.


Introduction
Contamination of surface waters with pathogens from fecal sources is a significant human health issue [1][2][3]. The concentrations of predominantly non-pathogenic fecal indicator bacteria (FIB), such as Escherichia coli (E. coli) and fecal coliforms (FCs) are commonly monitored and regulated in waterways as technologically simple and economical proxies to safeguard public health from pathogens of concern such as Cryptosporidium spp., Giardia spp., and E. coli O157:H7 [4][5][6]. Ideally, there is a high correlation between the presence of FIB at certain concentrations (i.e., microbiological water quality standards), direct fecal contamination, waterborne pathogens, and human health risk. However, this assumption has become increasingly challenged by a growing body of research (e.g., [4,5,7]), and the scientific search for improved proxies to detect microbial health risks continues with a focus on molecular and other methods (e.g., [8][9][10]). Despite their shortcomings, FIB-based microbial water quality standards remain the primary policy approach to microbial water quality protection [11,12]. Contemporary research demonstrates that E. coli is a superior FIB to FCs as a proxy for fecal pollution from warm-blooded This cross-sectional, longitudinal survey of FIB concentrations in surface waters was conducted in the summer of 2016 across the upper reaches of three mixed land use rural watersheds traversing the northern Sierra Nevada and southern Cascade regions of California ( Figure 1). The upper Stanislaus River Watershed, upper Susan River Watershed, and upper East Walker River Watershed reflect the diversity of annual precipitation, range and forest ecosystems, geology, and mixed uses common to the region. Table 1 provides details on the attributes of each watershed enrolled in this survey. Overall, the study area ranged from 40°34′ to 38°5′ N latitude and 121°9′ to 119°7′ W longitude, with an elevation range from 323 to 3754 m. The climate across the study area is Mediterranean with cool, wet winters and warm, dry summers. In these watersheds, the majority of precipitation falls as snow between December and April, with peak snowmelt typically occurring in May and early June. Following peak spring flows, streamflow declines rapidly to base flow conditions during summer (July through September). Numerous streams in the area are intermittent with flows ceasing during mid and late summer. Situated on the west slope of the Sierra Nevada, the upper Stanislaus River Watershed receives the greatest annual precipitation with uplands dominated by mesic conifer forests. The other watersheds are located on the drier eastern slopes of the Sierra Nevada (East Walker River Watershed) and the Cascade Range (Susan River Watershed) with xeric mixed conifer forests in the higher elevations transitioning to xeric shrublands in the lower elevations. Riparian areas associated with streams and springs are found throughout all watersheds, but represent less than 5% of the land area. Higher elevation forests across all watersheds are primarily National Forest lands administered by the U.S. Forest Service. Lower elevation shrublands on the two east-side watersheds are primarily public rangelands administered by the U.S. Bureau of Land Management. Privately owned land parcels are embedded throughout these larger public holdings.

Sample Sites and Fecal Sources
Seventy-seven stream water sample collection sites were established across all three watersheds in June 2016 (Table 1). Sample sites were selected immediately downstream of near-stream public lands key grazing areas (n = 31 or 40% of all sample sites), near-and in-stream public recreation areas (n = 40 or 52% of all sample sites), and near-stream rural residences (n = 6 or 8% of all sample sites). Sites were selected which could (1) be physically and legally accessed to sample and (2) logistically allow compliance with requirements to process samples at the laboratory within 8 h of collection. Recreational uses included near-stream trailheads used by hikers and recreational horse riders (i.e., pack stock), near-stream campgrounds, and in-stream swimming and bathing areas. Rural residence sites were near-stream small clusters (4 to 12) of rural residences with on-site waste treatment systems (septic systems). Key grazing areas on public lands were near-stream meadows and riparian areas where livestock graze and occupy frequently or for extended periods throughout the grazing season [46,47]. Grazing livestock were commercial beef cattle across all watersheds, with a minor component of commercial sheep grazing on some rangeland areas in the upper East Walker River Watershed. Public lands grazing was extensive in nature with livestock densities less than 0.05 animal unit/ha on individual grazing management units in excess of 10,000 ha. An animal unit is the grazing equivalent of a 450 kg cow with or without calf, assumed to consume 11.8 kg of forage per day on a dry weight basis. Grazing seasons were from June through September.

Sample Collection and FIB Determination
During the 1 July through 30 September summer season of 2016, stream water samples were collected across all three watersheds every 9 to 13 days. E. coli and FC concentrations were determined for each sample collected. All flowing sites within each watershed were grab sampled during each sampling bout via depth-integrated collection of 250 mL of stream water from the channel thalweg into a sterile container. Samples were immediately stored on ice and processed within 8 h of collection. Following standard method SM9222D, FC and E. coli concentrations as colony forming units (cfu) per 100 mL of water (cfu/100 mL) were enumerated via one step direct membrane filtration (0.45 µm nominal porosity filter) and incubation (44.5 • C, 22-24 h) on selective agar [48]. Agar used for FC was Difco mFC Agar, Becton, Dickinson and Company, Spars, MD, U.S. Agar used for E. coli was CHROMagar E. coli, ChromAgar, Paris, France. Two 100 mL negative control samples (autoclaved, double distilled phosphate buffered saline solution) as well as two positive control samples (negative control spiked to~100 cfu/100 mL) were randomly included in every laboratory run for both E. coli and fecal coliforms.

Data Analysis and Interpretation
To address our first study objective, we calculated descriptive statistics for FIB concentrations observed by potential fecal source-extensive public lands grazing, recreation, and rural residencesand over the entire dataset (i.e., overall across all sources). FIB concentration data generated in this Sustainability 2020, 12, 5207 5 of 16 study were count-based, suggesting either a Poisson or negative binomial regression analysis approach to achieve our second objective. Preliminary analysis of standard Poisson and post-hoc goodness of fit analysis, as well as standard negative binomial regression and post-hoc likelihood ratio tests for over-dispersion (alpha) both indicated that negative binomial regression was superior to Poisson regression due to over-dispersion. The data also had potential autocorrelation among samples collected from the same sample site. Thus, we utilized mixed effects negative binomial regression analyses with sample site identity set as a random intercept term to examine relationships between FIB concentrations (dependent variables) and the following independent variables: (1) streamflow condition (intermittent, perennial); (2) fecal source (grazing, recreation, rural residences); and (3) fecal source and seasonal progression (Julian Day). Likelihood ratio tests indicated that the random intercept mixed models were superior to standard negative binomial regression analysis in all cases. E. coli and fecal coliform concentrations were set as dependent variables and sample site identity was set as a random intercept in the following mixed effects regression analyses. Analysis one included FIB (one model each for E. coli and fecal coliform) and the independent variable streamflow condition (intermittent, perennial) using data from all 77 sample sites (595 samples) to examine differences in FIB between these conditions. Analysis two included FIB (one model each for E. coli and fecal coliform) and the independent variable fecal source (grazing, recreation, rural residence) using data from all 77 sample sites (595 samples) to examine differences in FIB between these sources. Analysis three included FIB (one model each for E. coli and fecal coliform) and the independent variables fecal source, Julian Day, the quadratic term for Julian Day, and the interaction between fecal source and Julian Day to examine expected effects of seasonal progression on FIB [46]. In analysis three, we only included data from perennial sample sites with sustained streamflow throughout the entire summer season (60 sample sites, 521 samples)-given our interest here in seasonal progression. Independent variables were considered significantly related to FIB concentrations at p-Value < 0.10. All statistical analyses were conducted in Stata/SE 13.1 [49]. For our final study objective, we compared observed FIB concentrations to benchmark E. coli (100, 320) and FC (20,40,200,400) concentrations (cfu/100 mL) that form the basis for national, state, and regional FIB microbial water quality standards relevant to the study area [12,[43][44][45]. These relevant water quality standards are described in detail in the results and discussion, for ease of comparison.

Precipitation, Streamflow, and Samples Collected
Precipitation realized for the hydrologic year (1 October 2015 through 30 September 2016) encompassing the study period (July through September 2016) was 97% (126 cm), 113% (78 cm), and 97% (59 cm) of the 30-year average for the upper Stanislaus River, Susan River, and East Walker River watersheds, respectively (Table 1). There were 595 stream water samples collected with concentrations determined for E. coli and fecal coliform. Sample sizes for grazing, recreation, and rural residence sites were 239, 309, and 47, respectively. Mean sample collection frequency was 11 days per site-ranging from 9 to 13 days. The number of samples collected per site ranged from 3 to 11 with a mean of 7.8 and a median of 8 samples collected per site, respectively. Fewer sample collections were associated with intermittent flow conditions observed at 17 sample sites (12 sites in the upper Stanislaus River watershed, 5 sites in the upper Susan River watershed). The number of sample sites and, thus, sample collections decreased as the season progressed and intermittent streams ceased flowing due to summer dry down. Sixty of the original 77 sample sites were perennial and sustained stream flow throughout the entire summer study period with 80%, 74%, and 67% of recreation, grazing, and rural residence sample sites maintaining perennial streamflow, respectively. Sample sizes for intermittent and perennial flowing sites were 74 and 521, respectively.

FIB Concentrations, Streamflow Conditions, Fecal Sources, and Season
Mixed effects negative binomial regression coefficients for fecal source indicated that concentrations of E. coli (p-Value = 0.003) and fecal coliforms (p-Value = 0.090) were higher for sample sites with intermittent streamflow compared to sites with perennial stream flow (Table 2, Figure 2). In previous research on National Forest lands in the study region, we observed mean E. coli concentrations of 114 cfu/100 mL for samples collected under low flow (<2 L/sec) conditions compared to 35 cfu/100 mL for samples collected under flowing conditions [46]. In that study, we also observed mean FC concentrations of 216 cfu/100 mL for low flow sample collections compared to 72 cfu/100 mL for flowing conditions. When intermittent streams are sampled as they approach cessation of flow, relatively high FIB concentrations are realized [50][51][52] due to a suite of interacting physical and biological factors such as temperature, nutrient concentrations, and biological growth rates [53]. Table 2. Results of mixed effects negative binomial regression analysis to examine differences in fecal indicator bacteria (E. coli and fecal coliform) concentrations (cfu/100 mL) between stream sample sites with intermittent and perennial flow conditions for 595 water samples collected across 77 stream sample sites throughout the summer sampling period (July through September of 2016).

E. coli
Fecal Coliform

Independent Variables
Streamflow recreation, grazing, and rural residence sample sites maintaining perennial streamflow, respectively. Sample sizes for intermittent and perennial flowing sites were 74 and 521, respectively.

FIB Concentrations, Streamflow Conditions, Fecal Sources, and Season
Mixed effects negative binomial regression coefficients for fecal source indicated that concentrations of E. coli (p-Value = 0.003) and fecal coliforms (p-Value = 0.090) were higher for sample sites with intermittent streamflow compared to sites with perennial stream flow (Table 2, Figure 2). In previous research on National Forest lands in the study region, we observed mean E. coli concentrations of 114 cfu/100 mL for samples collected under low flow (< 2 L/sec) conditions compared to 35 cfu/100 mL for samples collected under flowing conditions [46]. In that study, we also observed mean FC concentrations of 216 cfu/100 mL for low flow sample collections compared to 72 cfu/100 mL for flowing conditions. When intermittent streams are sampled as they approach cessation of flow, relatively high FIB concentrations are realized [50][51][52] due to a suite of interacting physical and biological factors such as temperature, nutrient concentrations, and biological growth rates [53]. Table 2. Results of mixed effects negative binomial regression analysis to examine differences in fecal indicator bacteria (E. coli and fecal coliform) concentrations (cfu/100 mL) between stream sample sites with intermittent and perennial flow conditions for 595 water samples collected across 77 stream sample sites throughout the summer sampling period (July through September of 2016).

E. coli Fecal Coliform Model Parameter Coefficient (S.E.) p-Value Coefficient (S.E.) p-Value Independent Variables
Streamflow   Concentrations of both E. coli and fecal coliforms were lowest at recreation sample sites followed by sites associated with grazing, and were highest downstream of rural residences. Table S1 reports outcomes of mixed effects negative binomial regression analysis examining differences in FIB Sustainability 2020, 12, 5207 7 of 16 concentrations between fecal source coefficients for all 77 stream sample sites (595 sample collections) (Figure 3). Table 3 reports outcomes of the mixed effects negative binomial regression analysis examining differences in FIB concentrations between fecal source coefficients while accounting for seasonal progression (Julian Day) for the 60 perennial flowing stream sample sites (521 sample collections). Results in both analyses (Table S1 and Table 3) were congruent on FIB concentration differences between fecal sources. As Table 3 indicates, E. coli concentrations at recreation (p-Value = 0.016) and rural residence (p-Value = 0.050) sample sites were significantly different from grazing sites. Post-hoc pairwise comparison indicated that recreation and rural residence sites had significantly different coefficients for E. coli concentrations (p-Value = 0.001). E. coli concentrations at recreation (p-Value = 0.088) and rural residence (p-Value = 0.005) sample sites were significantly different from grazing sites. Post-hoc pairwise comparison indicated that recreation and rural residence sites had significantly different coefficients (p-Value < 0.001) for fecal coliform concentrations (Figure 3).    The relatively low magnitude of E. coli and fecal coliform concentrations we observed for recreation followed by extensive public lands grazing activities are in agreement with our previous findings in similar watersheds [46], and with other studies. For example, Tiedemann et al. [50] and Gary et al. [54] report similar magnitudes of increase in fecal coliform concentrations associated with extensively grazed cattle pastures on Colorado's Front Range and Oregon's Cascade Mountains, respectively. Forrester et al. [55] and Clow et al. [56] both report comparable values for E. coli in several studies of water quality conditions associated with backpacker and pack livestock trail stream crossings, and near-stream pack stock grazing in national parks in the south Sierra Nevada. Pandey et al. [37] found mean E. coli concentrations to be less than 10 cfu/100 mL for sample sites on the mid and upper reaches of the "wild and scenic" designated Merced River Watershed draining Yosemite National Park to the south of the Stanislaus River Watershed, providing some reference for concentrations in limited use watersheds in the region.
On-site waste management systems (i.e., septic systems) are one likely reason for the relatively elevated FIB concentrations observed below near-stream rural residences (Figure 3). Research in other rural regions have demonstrated the potential for near-stream septic systems to contribute fecal bacteria and other pollutants to surface and ground water. For example, Verhougstraete et al. [30] identified septic systems to be a primary driver of in-stream fecal bacteria concentrations across mixed-use rural watersheds in Michigan's lower peninsula. Lapointe et al. [57] found dissolved nutrients in ground and surface waters to be concentrated near antiquated septic systems contributing to eutrophication of waters in southeast Florida, and Habteselassie et al. [58] report elevated FIB concentration in ground and surface waters near failing septic systems in rural North Carolina. Based upon our literature searches, there appears to be limited published research on septic systems as a microbial pollutant source for surface or ground water quality in rural, mountainous regions of California. Williams et al. [59] and Izbicki et al. [60] both identified septic systems as one source of elevated nitrate concentrations in ground water in semi-arid and arid mountain watersheds in rural areas of southern California. Given our preliminary assessment of FIB concentrations downstream of a Sustainability 2020, 12, 5207 9 of 16 limited number (6) of rural residences there is need for further examination of the potential magnitude and scope of this land use in the study region.
We found significant relationships between seasonal progression (Julian Day) and FIB concentrations; however, there were no significant (p-Value > 0.10) interactions between seasonal progression and fecal source (Table 3, Figures 4 and 5). Overall, FIB concentrations are lowest early and late in the summer season, with a pronounced increase as mid-summer approached. Roche et al. [46] documented this summer seasonal rise-peak-decline in FIB on similar watersheds within the study region, as have studies in other systems (e.g., [61][62][63]). This mid-summer peak corresponds to peak livestock and recreationalist numbers, peak migratory wildlife numbers, and peak environmental growth conditions in the region as discussed in reference [46]. Sustainability 2020, 12, x FOR PEER REVIEW 10 of 17 source of elevated nitrate concentrations in ground water in semi-arid and arid mountain watersheds in rural areas of southern California. Given our preliminary assessment of FIB concentrations downstream of a limited number (6) of rural residences there is need for further examination of the potential magnitude and scope of this land use in the study region. We found significant relationships between seasonal progression (Julian Day) and FIB concentrations; however, there were no significant (p-Value > 0.10) interactions between seasonal progression and fecal source (Table 3, Figures 4 and 5). Overall, FIB concentrations are lowest early and late in the summer season, with a pronounced increase as mid-summer approached. Roche et al. [46] documented this summer seasonal rise-peak-decline in FIB on similar watersheds within the study region, as have studies in other systems (e.g., [61][62][63]). This mid-summer peak corresponds to peak livestock and recreationalist numbers, peak migratory wildlife numbers, and peak environmental growth conditions in the region as discussed in reference [46].  Finally, it is important to note that E. coli and fecal coliform concentrations observed during the study were over-dispersed with a majority of observations at low concentrations, and relatively few observations at high concentrations. This distributional attribute and associated large variability in FIB is common to FIB datasets reported for rural mixed-use landscapes (e.g., [46,50,55,64]). In our study, with the exception of rural residence sites, the median and 75th percentile concentrations were lower than mean concentrations for both E. coli and fecal coliforms (Figures 2 and 3). This indicates a substantial preponderance of low concentration observations in the dataset with relatively few high concentrations disproportionately influencing mean values, and diminishing the utility of the mean as an accurate descriptor of the data. This over-dispersion was relatively more pronounced for recreational and grazing sites (i.e., means influenced by few high observations) than for rural residence sample sites (i.e., means less influenced by few high observations). Sustainability 2020, 12, x FOR PEER REVIEW 11 of 17 Finally, it is important to note that E. coli and fecal coliform concentrations observed during the study were over-dispersed with a majority of observations at low concentrations, and relatively few observations at high concentrations. This distributional attribute and associated large variability in FIB is common to FIB datasets reported for rural mixed-use landscapes (e.g., [46,50,55,64]). In our study, with the exception of rural residence sites, the median and 75th percentile concentrations were lower than mean concentrations for both E. coli and fecal coliforms (Figures 2 and 3). This indicates a substantial preponderance of low concentration observations in the dataset with relatively few high concentrations disproportionately influencing mean values, and diminishing the utility of the mean as an accurate descriptor of the data. This over-dispersion was relatively more pronounced for recreational and grazing sites (i.e., means influenced by few high observations) than for rural residence sample sites (i.e., means less influenced by few high observations).

FIB Concentrations, Fecal Sources, and Water Quality Benchmarks
Six FIB microbial water quality standards were relevant to the study area in the timeframe of this study and publication. In order to provide context with regard to applicable national, state, and regional recommendations for recreational water quality protection, we compared our results to the key benchmark E. coli and fecal coliform concentrations upon which these six water quality standards are based (Tables 4 and 5). For an individual sample site, the United States Environmental Protection Agency (USEPA) and California's State Water Quality Control Board (State Water Board) jointly recommend two microbial water quality standards. The first standard uses a benchmark that the geomean of E. coli concentrations for samples collected over the sample period is less than 100 cfu/100 mL. The geomean is calculated by (1) transforming sample concentrations by log10, (2) calculating the mean of those transformed concentrations, and then (3) raising the transformed mean by the power of 10 [12]. The second standard is based upon a statistical threshold value (STV) benchmark that no more than 10% of samples collected at a site over the sample period exceed an E. coli concentration of 320 cfu/100 mL [12,43]. The upper Stanislaus River Watershed is located within the jurisdiction of California's Central Valley Regional Water Quality Control Board (Central Valley Board), while the upper Susan River and East Walker River Watersheds are located within the jurisdiction of the

FIB Concentrations, Fecal Sources, and Water Quality Benchmarks
Six FIB microbial water quality standards were relevant to the study area in the timeframe of this study and publication. In order to provide context with regard to applicable national, state, and regional recommendations for recreational water quality protection, we compared our results to the key benchmark E. coli and fecal coliform concentrations upon which these six water quality standards are based (Tables 4 and 5). For an individual sample site, the United States Environmental Protection Agency (USEPA) and California's State Water Quality Control Board (State Water Board) jointly recommend two microbial water quality standards. The first standard uses a benchmark that the geomean of E. coli concentrations for samples collected over the sample period is less than 100 cfu/100 mL. The geomean is calculated by (1) transforming sample concentrations by log 10 , (2) calculating the mean of those transformed concentrations, and then (3) raising the transformed mean by the power of 10 [12]. The second standard is based upon a statistical threshold value (STV) benchmark that no more than 10% of samples collected at a site over the sample period exceed an E. coli concentration of 320 cfu/100 mL [12,43]. The upper Stanislaus River Watershed is located within the jurisdiction of California's Central Valley Regional Water Quality Control Board (Central Valley Board), while the upper Susan River and East Walker River Watersheds are located within the jurisdiction of the Lahontan Regional Water Quality Control Board (Lahontan Board). California's regional water boards are semi-autonomous and FIB water quality standards have varied among them. At the time this analysis was conducted, the Central Valley Board was in the process of adopting USEPA and State Water Board E. coli-based geomean and STV recommendations to replace existing FC-based geomean (200 cfu/100 mL) and STV (400 cfu/100 mL) standards [44]. At the same time, the Lahontan Board was employing FC-based geomean (20 cfu/100 mL) and STV (40 cfu/100 mL) standards [45], and was considering the adoption of USEPA and SWRCB recommendations. Table 4 reports the percentage of individual sites (n = 77) whose geomean concentration exceeded relevant FIB benchmarks. A comparison of individual sample concentrations to benchmark concentrations used in statistical threshold value (STV) microbial water quality standards is displayed in Table 5. Table 4. Exceedances based upon the study period geomean concentration (cfu/100 mL) standard for E. coli and fecal coliform (FC) at each sample site by specific fecal sources (recreation, grazing, and rural residences) across the entire summer study period (July through September of 2016). Results in Tables 4 and 5 demonstrate the potential for substantial discord, and policy-driven confusion, in assessments of microbial conditions using FC compared to E. coli-based water quality standards as indicators of fecal pollution by warm-blooded mammals. While by no means perfect [4], E. coli-based water quality standards have clearly been determined to be biologically and statistically superior to standards based on FC at any benchmark concentration (e.g., [11,12,16,43]). Thus, we will accept the overall exceedance rates observed for E. coli concentrations as the best estimates of actual microbial water quality pollution and human health risk across the study area and period (Tables 4 and 5). Compared to the sample site geomean E. coli benchmark (14% of sample sites with geomean > 100 cfu/100 mL E. coli), our results suggest FC-based standards overestimated potential microbial pollution and human health risk at sample sites by as much as 336% (61% of sites > 20 cfu/100 mL FC) and 278% (53% of sites > 200 cfu/100 mL FC) ( Table 4). Compared to the individual sample STV E. coli benchmark (13% of samples > 320 cfu/100 mL E. coli), our results would suggest FC-based standards overestimated potential microbial pollution and human health risk across all samples by as much as 538% (83% of samples > 40 cfu/100 mL FC) and 92% (25% of samples > 400 cfu/100 mL FC) ( Table 5). We observed similar discrepancies in the assessment of microbial water quality in the region in a previous study of 155 sample sites and 743 sample collections [46]. It is likely that the elevated fecal coliform concentrations relative to E. coli are driven by the fecal coliform assay's greater detection of non-fecal, thermotolerant environmental coliform bacteria creating false-positive indications of fecal contamination and erroneous estimates of health risk (e.g., [14][15][16]). This flaw in FC as an accurate, selective indicator of fecal contamination is among the various reasons international [11], national [12], and state [43] environmental and human health protection agencies recommend the use of E. coli-based microbial water quality standards.

Percent of Sample
Rates of observed exceedances for geomean E. coli water quality benchmarks were similar for recreation and grazing sample sites ( Table 4). Rates of observed exceedances for grazing sites were higher than recreation sites for STV E. coli water quality benchmarks (Table 5). These results indicate that microbial conditions were relatively good in surface water across the study site proximate to these activities, with approximately 5% to 16% of recreation and grazing sites exhibiting concentrations in exceedance of state and national E. coli standards for recreational waters, respectively. Site-specific fecal source assessments and implementation of appropriate best management practices have a high probability of reducing potential fecal contributions from these locations once identified [64][65][66][67]. Fifty to 83 percent of rural residence associated sample sites (Tables 4 and 5) exceeded state and national E. coli benchmarks. As discussed above, on-site septic systems are a likely reason for elevated E. coli concentrations and exceedances below rural residences (e.g., [30,57,58]). Our sample size for this land use is relatively small, so there is an important need for additional research into this potential source across the region to provide a clearer picture of the potential mechanisms and magnitude of E. coli contributions from this land use.

Implications
Overall, we found that 86% to 87% of stream sample sites across the study area met E. coli-based microbial water quality standards. Fecal indicator bacteria concentrations were lowest at recreation sites, followed closely by extensive livestock grazing sites. The highest levels of FIB concentrations observed in this study were at sites associated with rural residences, and at intermittently flowing stream sites. Given our small sample size for rural residences, additional research is required to determine if elevated FIB concentrations are broadly associated with this activity across the study region. Based upon the application of a structured spatial and temporal watershed-scale assessment of E. coli concentrations, we were able to identify relative magnitudes of potential fecal contribution among these sources over the course of the summer grazing and recreation season. The efficient identification of fecal contamination sources in these mixed-use rural watersheds is essential to guide the effective deployment of limited resources for on-the-ground mitigation of microbial pollution to safeguard human health. Our results demonstrate that policies reliant on antiquated fecal coliform-based water quality standards overestimated potential fecal contamination by as much as four orders of magnitude in this landscape. Such policies hinder the identification of the most likely fecal pollution sources and thus the successful targeting of mitigation practices to address them. They also misrepresent actual microbial water quality conditions to stakeholders, resources managers, and policy makers, hindering science-based decision making. We strongly recommend the application of the most current fecal indicator bacteria water quality standards to inform water quality improvement efforts and safeguard human health in these vast landscapes.
Supplementary Materials: The following are available online at http://www.mdpi.com/2071-1050/12/12/5207/s1, Table S1: Results of mixed effects negative binomial regression analysis to examine relationships between fecal source (recreation, grazing, and rural residences) and fecal indicator bacteria (E. coli and fecal coliform) concentrations (cfu/100 mL) for 595 water samples collected across 77 stream sample sites sustaining perennial streamflow throughout the entire summer sampling period (July through September of 2016).