coli and Enterococcus loads in a coastal Texas watershed

Pathogens are the principal cause of water body impairment for 303(d) listed waters in Texas and across the United States with 10,654 impairments nationally (TCEQ 2012; USEPA 2013). In Texas, 45% of 568 total impairments are caused by elevated bacteria levels (TCEQ 2012). Models such as the Soil and Water Assessment Tool (SWAT) and Hydrological Simulation Program-FORTRAN (HSPF) have been used for assessing bacterial sources and loading. Other simplistic microbial models, such as the Potential Nonpoint Pollution Index (PNPI), Spatially Explicit Delivery MODel (SEDMOD), and Spatially Explicit Load Enrichment Calculation Tool (SELECT), have been developed to rank potential pollution impacts from nonpoint sources and identify critical areas primarily using land use and geomorphology.


Estimating E. coli and Enterococcus loads 34
Terms used in paper

INTRODUCTION
Pathogens are the principal cause of water body impairment for 303(d) listed waters in Texas and across the United States with 10,654 impairments nationally (TCEQ 2012;USEPA 2013).In Texas, 45% of 568 total impairments are caused by elevated bacteria levels (TCEQ 2012).Models such as the Soil and Water Assessment Tool (SWAT) and Hydrological Simulation Program-FORTRAN (HSPF) have been used for assessing bacterial sources and loading.Other simplistic microbial models, such as the Potential Nonpoint Pollution Index (PNPI), Spatially Explicit Delivery MODel (SEDMOD), and Spatially Explicit Load Enrichment Calculation Tool (SELECT), have been developed to rank potential pollution impacts from nonpoint and point sources and identify critical areas primarily using land use, geomorphology, and potential sources in the watershed (Borel et al. 2012;Fraser et al. 1998;Munafo et al. 2005;Riebschleager et al. 2012;Teague et al. 2009).
The SELECT methodology was developed to characterize E. coli sources from point and nonpoint pollution in watersheds where watershed protection plans or total maximum daily loads (TMDLs) are developed to address bacterial contamination (Teague et al. 2009).Automated SELECT contains a graphic user interface within ArcGIS, whereby the user can adjust project parameters to develop watershed-specific pollutant loading scenarios using source and area characteristics (Riebschleager et al. 2012).

STUDY WATERSHED
The Copano Bay watershed (Figure 1) is located in the San Antonio-Nueces Coastal Basin and contains 3 water bodies impaired for bacteria: the tidal classified segments of the Aransas and Mission rivers and Aransas Creek, an unclassified water body.Both the rivers discharge to Copano Bay.This 485,073-hectare (1,198,641 acre) rural watershed contains both fresh and tidal waters.The Aransas River watershed totals 217,068 hectares (536,387 acres), of which 45% is used for cultivated crops (Figure 2).The largest municipality within the Aransas River watershed is Beeville with a population of 13,101 (USCB 2013).The Aransas River watershed encompasses portions of Bee, San Patricio, and Refugio counties with less than 1% of the watershed within Aransas and Live Oak counties.The Mission River watershed is predominantly comprised of range and pasture land -73% shrub/scrub and pasture hay out of the total 267,807 hectare (661,765 acre) land area (Figure 2).The watershed contains only 2 munici-

SPATIAL ANALYSES
The spatially explicit methodology, developed at Texas A&M University by the Department of Biological and Agricultural Engineering and Spatial Sciences Laboratory, was used to identify contributing potential bacteria sources and to estimate daily potential from indicator bacteria, E. coli and Enterococcus, loads in the Aransas River and Mission River watersheds.
All birds and mammals are potential sources of bacteria, and those present in the watershed contribute E. coli or Enterococcus to the tidal and freshwater portion of the Aransas and Mission rivers.However, each watershed is different and not all sources are likely to contribute significant amounts of bacteria to the water bodies.Additionally, sufficient information on species populations, E. coli and Enterococcus concentrations, and feces production rates are often unavailable, thus precluding the ability to effectively assess potential E. coli or Enterococcus contributions from respective sources.Data from government agencies and local stakeholders on the number and distribution of contributing sources, in combination with National Land Cover Database 2006 (NLCD 2006) land use classification data were entered into a GIS software program.Each watershed was also divided into 4,047 to 16,187 hectares (10,000 and 40,000 acres) 12-digit hydrologic unit code (HUC) subwatersheds acquired from U.S. Geological Survey (USGS) and U.S. Department of Agriculture-Natural Resources Conservation Service (USDA-NRCS) (2012).Bacterial sources were examined at this subwatershed level, as opposed to the entire watershed of each river, to identify and target the areas in each subwatershed where the sources were likely to impact water quality.
The 2 factors considered when determining the potential contribution for each source were the abundance of a particular source in the watershed and whether sufficient information is available to effectively predict bacteria loadings from that source.The methodology developed by Wagner and Moench (2009) to estimate animal population was applied here to update population densities.Wagner and Moench (2009) included cattle, horses, goats, sheep, domestic hogs, poultry, deer, and feral hogs in their estimation protocol.With the exception of domestic hogs and poultry, all of these sources, with the addition of on-site sewage facilities (OSSFs), wastewater treatment facilities (WWTFs), and pet dogs, warranted modeling and had sufficient information to model using SELECT.

POTENTIAL E. COLI AND ENTEROCOCCUS LOAD ESTIMATION
The analysis was performed at a 30 meter by 30 meter resolution.Each source was distributed to appropriate land uses in the watershed, and then the potential E. coli and Enterococcus loads were calculated using the equations in Tables 1 and 2, respectively.E. coli and Enterococcus bacte-ria are indicator bacteria and are generally not pathogenic but indicators of potential fecal contamination.The daily fecal coliform production rates for the livestock, wildlife, and feral hog sources were estimated per animal unit (AU) from Wagner and Moench (2009).The highest fecal coliform production values in U.S. Environmental Protection Agency (USEPA) guidelines document (2001) were used for OSSF and dog sources.E. coli and Enterococcus loads were calculated    (TCEQ 2010).
After the potential E. coli or Enterococcus loads were calculated, the results were aggregated to the subwatershed level to identify areas of potential concern.

Cattle
Cattle were uniformly applied according to 4 separate stocking rates derived from 2004 to 2008 Texas Agricultural Statistics (USDA-NASS 2004-2008) and USDA-NRCS estimates from Wagner and Moench (2009).The 4 stocking rates were: 20 acre/AU for the land use classifications of deciduous forest, evergreen forest, and mixed forest, 30 acre/AU for the shrub/ scrub land use classification, 15.4 acre/AU for the land use classification grasslands/herbaceous, and 7.7 acre/AU for the pasture/hay classification.A total of 3,152 AUs were evenly distributed to all of the forested lands.13,153 AUs of cattle were uniformly applied over shrub/scrub classifications.3,148 AUs were evenly distributed over grassland/herbaceous land use classifications and 40,052 AUs were distributed over pasture/ hay lands.The total cattle potential loads were estimated by adding together the results from the 4 separate stocking rate distributions.A fecal coliform production rate of 8.55 x 10 9 colony forming units (CFU) per AU per day (Wagner and Moench 2009) was used in the model and converted from fecal coliform to E. coli using a conversion ratio.The total potential E. coli and Enterococcus loads for cattle were estimated using the distributed cattle density production rate, fecal coliform production rate, and conversion ratios.

Horses
A total of 2,772 AUs of horses were evenly distributed over developed open space, grassland/herbaceous, and pasture/ hay.This number was derived from the 2007 USDA Census of Agriculture (USDA-NASS 2007) county estimates multiplied by the percentage of the county in the Aransas River and Mission River watersheds and the AU conversion of 1.25 (Wagner and Moench 2009).The percentage of a county in a watershed was adjusted from Wagner and Moench (2009) because the watershed boundary differed from the original report, causing the estimated animal populations to adjust as well.The fecal coliform production rate used in the model was 2.91 x 10 8 CFU per AU per day (Wagner and Moench 2009).The total potential E. coli and Enterococcus loads for horses were estimated using the distributed horse density, fecal coliform production rate, and conversion ratios.

Goats
A total of 565 AUs of goats were evenly distributed over developed open space, shrub/scrub grassland/herbaceous, and pasture/hay.Wagner and Moench (2009) estimated the goat numbers by using the 2005-2008 Texas Agricultural Statistics for Bee, Goliad, and Karnes counties and district numbers for Aransas, Refugio, and San Patricio counties.The numbers were updated from Wagner and Moench (2009) by using an adjusted percent in watershed number because the Aransas River and Mission River watersheds boundaries differed from the original report.The fecal coliform production rate used in the model was 2.54 x 10 10 CFU per AU per day (Wagner and Moench 2009), and total potential E. coli and Enterococcus loads for goats were estimated using the distributed goat density, fecal coliform production rate, and conversion ratios.The fecal coliform production rate used in the model was 2.90 x 10 11 CFU per AU per day (Wagner and Moench 2009) and total potential E. coli and Enterococcus loads for sheep were estimated using the distributed sheep density, fecal coliform production rate, and conversion ratios.

Deer
A total of 9,951 deer AUs were evenly applied over the entire watershed.This is the population estimate produced by applying Wagner and Moench (2009) county densities calculated from Texas Parks and Wildlife Department (TPWD) surveys.Deer were evenly distributed across the Aransas River and Mission River watersheds to best reflect the surveying techniques used by the TPWD.The densities were multiplied by the number of acres of the county in the watershed and the AU conversion of 0.112 to determine the number of deer AUs in each county.The fecal coliform production rate used was 1.50 × 10 10 CFU per AU per day (Yagow 2001; Cox et al. 2005;Wagner and Moench 2009), and total potential E. coli and Enterococcus loads for deer were estimated using the distributed deer density, fecal coliform production rate, and conversion ratios.Estimating E. coli and Enterococcus loads

Feral hogs
A total of 4,198 feral hog AUs were applied uniformly across deciduous forest, evergreen forest, mixed forest, shrub/scrub, grassland/herbaceous, pasture/hay, cultivated crops, and woody wetlands.This population estimate was derived by Reidy (2007) using a density of 33.3 acre/hog and an AU conversion of 0.125 (Wagner and Moench 2009).The fecal coliform production rate used was 1.21 × 10 9 CFU per AU per day (Cox et al. 2005;Mukhtar 2007;Wagner and Moench 2009), and total potential E. coli and Enterococcus loads for feral hogs were estimated using the distributed feral hog density, fecal coliform production rate, and conversion ratios.

On-site sewage facilities
OSSFs were modeled using spatially distributed point data of each household obtained from residential 911 address data gathered from the Coastal Bend Council of Governments and the Golden Crescent Regional Planning Commission.2010 census data (USCB 2010) was used for Karnes County because 911 address data was unavailable for this county.Households within Certificate of Convenience and Necessity areas (PUC 2013) were removed to exclude households being serviced by a WWTF.The total number of households with OSSFs in the watershed was estimated to be 10,047, and the average persons per household for a census block were calculated by using 2010 Census data (USCB 2010).A fecal coliform concentration of raw sewage 10 x 10 6 CFU per 100 milliliters (USEPA 2001) was used to model failing OSSFs with a constant sewage discharge of 70 gallons per person per day.OSSF failure rate was estimated by applying the soil drainfield limitation classes as follows: very limited with 15% failing, somewhat limited with 10% failing, not limited with 5% failing, and not rated with 15% failing (USDA-NRCS 2004).The percentage of E. coli and Enterococcus contributing to the Aransas River and Mission River watersheds due to OSSF failures were calculated by multiplying the OSSF household densities, average person per household, fecal coliform concentration of raw sewage, sewage discharge, failure rate, and conversion ratios.

Dogs
A dog density of 1 dog per household was an updated density as reported by the American Veterinarian Medical Association and used in the Geronimo Creek watershed analysis (AVMA 2002;Geronimo and Alligator Creeks Watershed Partnership 2012).The density was applied to the residential 911 addresses, resulting in an estimated dog population of 10,065.The fecal coliform production rate of 5 x 10 9 CFU per dog per day (USEPA 2001) multiplied by the conversion ratios was used to determine the potential E. coli and Enterococcus loads resulting from dogs.

Wastewater treatment facilities
There are 12 WWTFs in the Aransas River and Mission River watersheds.Three WWTFs are within the Mission River watershed: town of Refugio, town of Woodsboro, and Pettus Municipal Utility District with permitted discharges of 0.576, 0.25, and 0.105 million gallons per day (MGD) respectively.Nine WWTFs are within the Aransas River watershed including 2 for the city of Beeville with permitted discharges of 3 and 2.5 MGD as well as, 2 for the city of Sinton with permitted discharges of 0.015 and 0.8 MGD.The remaining WWTFs in the Aransas River watershed are: city of Taft, Skidmore Water Supply Corporation, St. Paul Water Supply Corporation, Tynan Water Supply Corporation, and Texas Department of Transportation with permitted discharges of 0.9, 0.131, 0.05, 0.045, and 0.00038 MGD, respectively.Each WWTF was modeled at its daily maximum permitted discharge and, if applicable, its effluent limitation of either an E. coli concentration of 126 CFU or most probable number per 100 milliliters or an Enterococcus concentration of 35 CFU per 100 milliliters, to estimate the potential bacteria loads from WWTFs.
In total, 9 potential sources were modeled using SELECT in both the Aransas River and Mission River watersheds.Since the 2 watersheds are part of the entire Copano Bay watershed, the potential source contribution population densities applied were the same for both the Aransas River and Mission River watersheds.

RESULTS AND DISCUSSION
The watershed spatial analysis performed by SELECT highlights subwatersheds within the individual Aransas River and Mission River watersheds that have the highest potential to contribute E. coli and Enterococcus based on land use distribution and potential source contributions.The analysis highlights subwatersheds of concern for particular sources and for total potential E. coli and Enterococcus loads, taking into account all of the potential sources modeled.Conclusions can be made about which sources have the highest potential to contribute E. coli and Enterococcus and pinpoint subwatersheds where those contributions are in the Aransas River and Mission River watersheds by using SELECT results.

Spatially explicit E. coli and Enterococcus load estimation
Tables 3 and 4 illustrate the source specific E. coli and Enterococcus ranges used to estimate the contribution of each source

Estimating E. coli and Enterococcus loads 40
for both the Mission River and Aransas River watersheds.The source-specific ranges for both watersheds are identical or similar because sources were distributed similarly.The source exhibiting highest daily potential E. coli and Enterococcus loads in both the Aransas River and Mission River watersheds was cattle, while OSSFs, dogs, and deer were the next highest contributors, respectively.Sources with moderate potential contributions were sheep, feral hogs, and goats.The lowest contributors were horses and WWTFs.
Figures 3 and 4 illustrate the total potential E. coli and Enterococcus loads, or combined load, which includes the loading potentials of all of the modeled sources applied in the Mission and Aransas River watersheds.Subwatersheds in red indicate areas with the highest potential for E. coli contributions to the river, while the darkest green represents areas with the lowest potential.The highest contributing subwatersheds either had a dominant land use of pasture/hay or urban land uses generally due to the high numbers of cattle, OSSFs, and dogs associated with these land uses.The lowest contributing subwatersheds contained areas of mostly cultivated crop land.

Potential Issues
SELECT provides a daily snapshot of fecal and bacterial deposition based on conditions inputted into the model, and as such, fecal buildup or bacteria die-off and re-growth are not taken into account.Fecal buildup over the land surface before a rainfall event as well as bacteria die-off can cause the actual E. coli or Enterococcus production rates to vary widely compared to those in the model.
SELECT does not take into account direct deposition of fecal material into the stream.Direct deposition significantly affects the bacteria concentrations present in a water body more so than land deposition.Larsen et al. (1994) found that manure deposited 0.6 meters (2 feet) from a stream contributed 83% less bacteria and manure deposited at 2.1 meters (7 feet) contributed 95% less than manure deposited directly into a stream.The timing of fecal deposition is also not taken into account.If fecal matter is deposited shortly before a rain event, then the bacteria will more likely end up in the water body   via surface runoff.A significant factor found by Wagner et al. (2012) is when runoff occurred more than 2 weeks following grazing (and fecal matter deposition), E. coli levels were 88% lower compared to runoff during or soon after grazing.SELECT does not take into account the distance of the fecal deposition from the water body.As found by Larsen et al. (1994), bacteria from fecal deposition have a much higher potential to impact water quality when deposited at closer distances compared to farther distances from a water body.
In addition, the animal densities used in the model can vary.Animal densities can change drastically from season to season and from year to year, particularly in response to worsening drought conditions that often affect many areas of Texas.Further, with the exception of feral hogs and deer, estimates of wildlife numbers are impracticable to evaluate.These issues can impact the watershed planning process, particularly because SELECT results show cattle as the top contributors of bacteria into the water bodies.In comparison, bacterial source tracking conducted in the Buck Creek, Little Brazos River, and Big Cypress Creek watersheds suggest that wildlife contributions range from 42-65%, while cattle and other domestic animal contributions range from only 14-29% (Di Giovanni et al. 2013).The fecal material from cattle may not be reaching and contaminating the water body, but other sources could be contaminating the water more directly or not integrated into the model.As a result, the best management practices (BMPs) chosen to be implemented would be influenced by these issues with the model, and their overall impact of projected loadings would be greatly affected.

CONCLUSIONS
The SELECT methodology was applied to the entire Copano Bay watershed, comprised of the Aransas River and Mission River watersheds, to calculate potential E. coli and Enterococcus loads occurring in the watersheds and identify priority areas for implementing management practices.The SELECT methodology was adapted to the Copano Bay watershed to include the perceived potential contributors with data availability.The model is unable to reflect the true potential loading of the watershed as not all contributing sources are taken into account due to the availability of data.Once data is collected, the SELECT model can be adapted and additional sources can easily be added to the model.In both the Aransas River and Mission River watersheds, cattle were determined to be the largest potential contributor.This suggests that BMPs supporting good grazing management will yield the most improvements and be the most effective at lowering the bacteria contamination in the water body.The SELECT methodology was able to highlight areas of highest concern, which provides guidance for individuals and entities that implement BMPs where practices would be the most effective.The SELECT methodology can easily be adapted and applied to watersheds to reflect stakeholder concerns.
The next steps for the SELECT methodology are to account for sources that are currently not able to be modeled, such as mesomammals, birds, and other background sources.Potential integration of bacterial source tracking results with SELECT could also be evaluated to address these issues.Future SELECT methodology could also include fecal buildup and bacteria die-off and re-growth to improve the model.SELECT outputs could be combined with another model to determine the E. coli and Enterococcus loads reaching the water bodies either through surface runoff or through soils.

SheepA
total of 111 AUs of sheep were evenly distributed over developed open space, shrub/scrub, grassland/herbaceous, and pasture/hay.This number was derived from the 2007 USDA Census of Agriculture county estimates multiplied by the percentage of the county in the watershed and the AU conversion of 0.2 (USDA-NASS 2007; Wagner and Moench 2009).

Figure 4 .
Figure 4. Total daily potential Enterococcus loads from all considered sources in the Mission River and Aransas River watersheds.

Figure 3 .
Figure 3.Total daily potential E. coli loads from all considered sources in the Mission River and Aransas River watersheds.

Table 1 .
Calculation of potential E. coli loads from various sources.

Table 2 .
Calculation of pote2004l Enterococcus loads from various sources.fromfecalcoliformloadsusing a ratio of 0.63 fecal coliform to E. coli and 0.175 fecal coliform to Enterococcus, based on 2010 Texas Surface Water Quality regulatory standards in recreational waters.The geometric mean regulatory standard for primary contact recreational use for freshwater(Segments  2002 and2004) is 126 organisms per 100 milliliter for E. coli, and for high saline inland waters(Segments 2001 and 2003)35 organisms per 100 milliliter for Enterococcus

Table 3 .
Daily potential E. coli load ranges.