Town Branch TMDL final

Total Maximum Daily Load for Fecal Coliform Bacteria
to Town Branch, North Carolina
Final version submitted to EPA
August, 2002
Cape Fear River Basin
Prepared by:
NC Department of Environment and Natural Resources
Division of Water Quality
1617 Mail Service Center
Raleigh, NC 27699-1617
[phone]
Final version submitted to EPA
1
TABLE OF CONTENTS
1.0 Introduction … … … … … … … … … … … … … … … … … … … … … … … ….2
1.1 Watershed Description … … … … … … … … … … … … … … … … ….4
1.2 Water Quality Monitoring Program… … … … … … … … … … … … …5
1.3 Water Quality Target … … … … … … … … … … … … … … … … … …6
2.0 Source Assessment … … … … … … … … … … … … … … … … … … … … … …7
2.1 Point Source Assessment… … … … … … … … … … … … … … … … …7
2.2 Nonpoint Source Assessment….… … … … … … … … … … … … … …7
2.2.1 Livestock… … … … … … … … … … … … … … … … … … … …7
2.2.2 Miscellaneous Sources… … … … … … … … … … … … … … …8
2.2.3 Failed Septic Systems… … … … … … … … … … … … … … …..8
2.2.4 Urban Development/Sanitary Sewer Overflows… … … … ….9
2.2.5 Wildlife … … … … … … … … … … … … … … … … … … … ….11
2.3 Source Assessment Conclusion … … … … … … … … … … … … … ….12
3.0 Modeling Approach… … … … … … … … … … … … … … … … … … … … …..12
3.1 Model Framework… … … … … … … … … … … … … … … … … … ….12
3.2 Model Setup… … … … … … … … … … … … … … … … … … … … …..13
3.2.1 Instream Decay Rate … … … … … … … … … … … … … … ….14
3.3 Hydrologic Calibration… … … … … … … … … … … … … … … … … …14
3.4 Water Quality Calibration… … … … … … … … … … … … … … … …..17
3.4.1 Calibration Results… … … ….… … … … … … … … … … … …18
3.4.2 Prediction Uncertainty ….… … … … … … … … … … … … ….21
4.0 Total Maximum Daily Load… … … … … … … … … … … … … … … … … … …22
4.1 Reduction Target… … … …. … … … … … … … … … … … … … … …..22
4.1.1 Critical Conditions… … … … … … … … … … … … …...… …...23
4.1.2 Current Conditions… … … … … … … … … … … … … … … …. 25
4.1.3 Margin of Safety… … … … … … … … … … … … … … … … …. 25
4.1.4 TMDL Allocation… … … … … … … … … … … … … … … ….. 27
4.2 Instantaneous Portion of Standard… … … … … … … … … … … … …. 28
4.3 Seasonal Variation … … … … … … … … … … … … … … … … …..28
5.0 Summary and Future Considerations … … … … … … … ….… … … … … …..29
5.1 Monitoring… … … … … … … … … … … … … … … … … … … … … ….29
5.2 Implementation… … … … … … … … … … … … … … … … … … … …..30
6.0 Public Participation… … … … … … … … … … … … … … … … … … … … … …30
References Cited… … … … … … … … … … … … … … … … … … … … … … … … … …31
Appendix I.Observed Data… … … … … … … … … … ….… … … … … … … … …..34
Appendix II.Model Fit Statistics… … … … … … … … … … … … … … … … … … … 39
Appendix III.Model Calibration Information… … … … … … … … … … … … ….41
Appendix IV. Stream Channel Cross Sections for Subwatersheds… … … … … … …42
Appendix V. Modeled Flow and Fecal Coliform … … … … … … … … … … … … … 44
Final version submitted to EPA
2
Total Maximum Daily Load for fecal coliform bacteria to Town Branch
1.0 INTRODUCTION
On the 2000 North Carolina 303(d) list, the North Carolina Division of Water Quality (DWQ) has
identified a 3.6-mile segment (16-17) of Town Branch in the Cape Fear Basin as impaired by fecal
coliform bacteria. The impaired segment extends from the stream’s source to its confluence with
the Haw River. This section of the stream is located in subbasin 03-06-02. Town Branch is
designated as a class C water. Class C waters are freshwaters that are protected for secondary
recreation, fishing, and propagation and survival of aquatic life.
Section 303(d) of the Clean Water Act (CWA) requires states to develop a list of waters not meeting
water quality standards or which have impaired uses. This list, referred to as the 303(d) list, is
submitted biennially to the U.S. Environmental Protection Agency (EPA) for review. The 303(d)
process requires that a Total Maximum Daily Load (TMDL) be developed for each of the waters
appearing on Part I of the 303(d) list. A TMDL is the maximum amount of a pollutant (e.g., fecal
coliform) that a waterbody can receive and still meet water quality standards, and an allocation of
that load among point and nonpoint sources. The objective of a TMDL is to estimate allowable
pollutant loads and allocate to known sources so that actions may be taken to restore the water to its
intended uses (USEPA, 1991). Generally, the primary components of a TMDL, as identified by
EPA (1991, 2000a) and the Federal Advisory Committee are as follows:
Target identification or selection of pollutant(s) and endpoint(s) for consideration. An endpoint is an
instream numeric target. The pollutant and endpoint are generally associated with
measurable water quality related characteristics that indicate compliance with water quality
standards. North Carolina indicates known problem pollutants on the 303(d) list.
Source assessment. Sources that contribute to the impairment should be identified and loads
quantified, to the extent that that is possible.
Reduction target. Estimation or level of pollutant reduction needed to achieve water quality goal. The
level of pollution should be characterized for the waterbody, highlighting how current
conditions deviate from the target endpoint. Generally, this component is identified through
water quality modeling.
Final version submitted to EPA
3
Margin of safety. The margin of safety addresses uncertainties associated with pollutant loads,
modeling techniques, and data collection. Per EPA (2000a), the margin of safety may be
expressed explicitly as unallocated assimilative capacity (portion of TMDL) or implicitly
through conservative assumptions. The margin of safety should be included in the reduction
target.
Allocation of pollutant loads. Allocating available pollutant load (TMDL), and hence pollutant control
responsibility, to the sources of impairment. The wasteload allocation portion of the TMDL
accounts for the loads associated with existing and future point sources. The load allocation
portion of the TMDL accounts for the loads associated with existing and future nonpoint
sources. Any future nonpoint source loading should remain within the TMDL that is
calculated in this assessment; in other words, this TMDL does not leave allocation for future
sources.
Seasonal variation. The TMDL should consider seasonal variation in the pollutant loads and endpoint.
Variability can arise due to streamflows, temperatures, and exceptional events (e.g., droughts
and hurricanes).
Critical conditions. Critical conditions occur when fecal coliform levels exceed the standard by the
largest amount. If the modeled load reduction is able to meet the standard during critical
conditions, then it should meet the standard at all, or nearly all, times.
Section 303(d) of the CWA and the Water Quality Planning and Management regulation (USEPA,
2000a) require EPA to review all TMDLs for approval or disapproval. Once EPA approves a
TMDL, then the waterbody may be moved to Part III of the 303(d) list. Waterbodies remain on
Part III of the list until compliance with water quality standards is achieved. Where conditions are
not appropriate for the development of a TMDL, management strategies may still result in the
restoration of water quality.
The goal of the TMDL program is to restore designated uses to water bodies. Thus, the
implementation of bacteria controls will be necessary to restore designated uses in Town Branch.
Although an implementation plan is not included as part of this TMDL, reduction strategies are
needed. The involvement of local governments and agencies will be critical in developing an
implementation plan and reduction strategies. DWQ will seek to begin development of the
implementation plan during public review of the TMDL.
Final version submitted to EPA
4
1.1 Watershed Description
Town Branch, located in the upper Cape Fear River basin, drains into the Haw River about three
miles northeast of the City of Graham (see Figure 1). The creek’s watershed lies entirely within
Alamance County and is slightly less than 4 square miles in area. The City of Graham (2000
population of 12,833) covers more than three-quarters of the watershed. DWQ has an ambient
water quality monitoring site (Storet number B1260000) near the creek’s confluence with the Haw.
Figure 1.
#
Haw River
Town Branch
City of Graham Interstates 40 & 85
Graham municipal boundary
Major Roads
Reach File, V3 (03030002)
#DWQ ambient station
Town Branch watershed
1 0 1 2 Miles
Town Branch
N
Town Branch Watershed
Vicinity Map of Town Branch Watershed
The land use/land cover characteristics of the watershed were determined using 1996 land cover
data that were developed from 1993-94 LANDSAT satellite imagery. The North Carolina Center
for Geographic Information and Analysis, in cooperation with the NC Department of
Transportation and the United States Environmental Protection Agency Region IV Wetlands
Division, contracted Earth Satellite Corporation of Rockville, Maryland to generate comprehensive
land cover data for the entire state of North Carolina. Tabulated land cover/land use data for the
Final version submitted to EPA
5
Town Branch watershed are shown in Table 1. During the formation of this geographic dataset,
developed land was identified using the proportion of synthetic cover present; low density
developed was 50-80% synthetic cover, and high density developed was 80-100% synthetic cover
(Earth Satellite Corporation, 1997). Assuming that synthetic cover is impervious, and that all non-
developed land cover classes have 1% impervious cover, the Town Branch watershed is estimated to
have 29-42% impervious surface.
Table 1. Land use/land cover in Town Branch watershed.
Land Use/Land Cover Town Branch Watershed Acres
High Density Developed 275 (10.8%)
Low Density Developed 991 (38.8%)
Cultivated 5 (0.0%)
Managed Herbaceous 395 (15.5%)
Forest 886 (34.7%)
Total 2552
The USGS 14-digit hydrologic unit code (HUC) for Town Branch is 030300020080.
1.2 Water Quality Monitoring Program
There are two sources of fecal coliform data for this project: 1) ambient monitoring data; 2) special
study data collected by the City of Graham. More information on each is provided below and a
more detailed report is included in Appendix I.
Town Branch was listed as impaired based on data from the ambient monitoring station, which is
located at SR 2109 (Cooper Rd.) or about one-quarter mile upstream from the stream’s confluence
with the Haw River. Fecal coliform samples are collected at this site on a monthly basis.
The second source of Town Branch fecal coliform data is a special study conducted by the City of
Graham in July and August of 2001. Ten samples were collected at three different sites, including:
Town Branch at Cooper Road (ambient site at SR 2109); Town Branch at the wastewater treatment
plant (just upstream of major tributary, County Home Branch); County Home Branch at Gilbreath
Street. The purposes of this study were to evaluate whether the creek was complying with the state
fecal coliform standard, and to provide information on bacteria source areas in watershed.
Final version submitted to EPA
6
The ambient monitoring data were used in model calibration. Both sets of monitoring data may be
seen in Appendix I.
1.3 Water Quality Target
The North Carolina fresh water quality standard for fecal coliform in Class C waters (T15A:
02B.0211) states:
Organisms of the coliform group: fecal coliforms shall not exceed a geometric mean of
200/100ml (membrane filter count) based upon at least five consecutive samples examined
during any 30 day period, nor exceed 400/100 ml in more than 20 percent of the samples
examined during such period; violations of the fecal coliform standard are expected during
rainfall events and, in some cases, this violation is expected to be caused by uncontrollable
nonpoint source pollution; all coliform concentrations are to be analyzed using the
membrane filter technique unless high turbidity or other adverse conditions necessitate the
tube dilution method; in case of controversy over results, the MPN 5-tube dilution technique
will be used as the reference method.
The instream numeric target, or endpoint, is the restoration objective expected to be reached by
implementing the specified load reductions in the TMDL. The target allows for the evaluation of
progress towards the goal of reaching water quality standards for the impaired stream by comparing
the instream data to the target. For this TMDL the water quality target is the geometric mean
concentration of 200cfu/100ml over a 30-day period. A geometric mean is obtained by calculating
the average of the log values of the individual samples. Basically, the geometric mean will discount
higher values so that it should be lower than the arithmetic mean (average of measurements, no log
taken). Cfu stands for colony-forming units; it may also be referred to as simply ‘counts’ in this
assessment. In this TMDL, DWQ will consider the entire model period to address the portion of
the standard that limits the percentage of instantaneous excursions over 400cfu/100ml to twenty
percent.
In order to evaluate the fecal coliform model, monitor water quality conditions and assess progress
of the TMDL, an evaluation location was established for the Town Branch watershed. The
evaluation location of this watershed is Town Branch at SR 2109, which is the location of the
ambient monitoring station.
Final version submitted to EPA
7
2.0 SOURCE ASSESSMENT
A source assessment is used to identify and characterize the known and suspected sources of fecal
coliform bacteria in the watershed. DWQ completed a source assessment and used it to develop the
water quality model for the TMDL calculation.
2.1 Point Source Assessment
General sources of fecal coliform bacteria are divided between point and nonpoint sources.
Currently, there are no facilities in the watershed that discharge waste through the National Pollutant
Discharge Elimination System (NPDES), which is considered to be the regulatory approach for all
but the smallest of point sources. The City of Graham has a municipal treatment facility (NPDES
NC0021211), however that discharges into the Haw River. North Carolina also has general
wastewater permits for package plants (small sand filter operations), but there is very limited
information about those; DWQ cannot identify any that discharge to Town Branch.
2.2 Nonpoint Source Assessment
Nonpoint sources of fecal coliform bacteria include those sources that can not be identified as
entering the waterbody at a specific location (e.g., a pipe). Nonpoint source pollution includes
urban, agricultural and background sources. For this TMDL, background loading is considered to
be that which originates from wildlife; this is primarily from forestland, but wildlife are considered to
exist on cropland as well (see 2.2.5 Wildlife below). Fecal coliform bacteria may originate from
human and non-human sources. Table 2 lists the potential human and animal nonpoint sources of
fecal coliform bacteria (Center for Watershed Protection, 1999). The nonpoint sources of fecal
coliform bacteria in Town Branch include runoff from urban development (stormwater), sewer line
systems (leaky sewer lines and sewer system overflows), wildlife, failing septic systems, and probably
illicit connections in unknown locations.
2.2.1 Livestock
DWQ conferred with an Alamance County extension agent, Paul Walker, to derive estimates for
livestock populations in the Town Branch watershed. From this conversation, DWQ decided that
no livestock reside in the watershed, and that none would be accounted for in the TMDL model.
Final version submitted to EPA
8
Table 2. Potential sources of fecal coliform bacteria in urban and rural watersheds (Center for
Watershed Protection, 1999).
Source Type Source
Human Sources Sewered watershed Combined sewer overflows
Sanitary sewer overflows
Illegal sanitary connections to
storm drains
Illegal disposal to storm drains
Non-sewered watershed Failing septic systems
Poorly operated package plant
Landfills
Marinas
Non-human Sources Domestic animals and urban wildlife Dogs, cats
Rats, raccoons
Pigeons, gulls, ducks, geese
Livestock and rural wildlife Cattle, horse, poultry
Beaver, muskrats, deer, waterfowl
2.2.2 Miscellaneous Sources
Illicit discharges (e.g., straight pipes) and any loading that occurs during baseflow (no runoff) are
called ‘miscellaneous sources’ in the TMDL allocation. In the model, miscellaneous sources are
treated as a constant, instream source of bacteria. It is necessary to separate these instream
sources from land based ones, because they are defined as one instream source through
modeling. That is, it is difficult to determine individual estimates for the fecal coliform that
originates from illicit discharges, but it is possible to estimate them cumulatively in the model
during periods of low streamflow. During such periods, contributions from nonpoint sources of
fecal coliform, which rainfall transport, are assumed to be negligible. What is being fit to the
model, through calibration, is a constant instream source of fecal coliform.
2.2.3 Failed Septic Systems
Failing septic systems have been cited as a potential source of fecal coliform bacteria to water bodies
(USEPA, 2000). For the most part, household waste from the City of Graham and its surroundings
is treated at the municipal treatment plant, which discharges to the Haw River. Based on
investigation by the City of Graham, it appears that there are at least 13-16 legacy septic tanks (no
longer in active use) and just a few currently used septic tanks in the watershed (Quick, 2002).
DWQ estimates that there are 4 failing legacy septic systems in the Town Branch watershed. During
the public comment period on the draft Town Branch TMDL, the City of Graham alerted DWQ to
Final version submitted to EPA
9
additional failing septic systems in a neighborhood immediately north and south of Cheek Lane
(Sullivan, 2002). After consulting the map and other information provided by the City of Graham,
DWQ decided to add 20 more failing septic systems to the TMDL model. Negotiations are in
progress to annex this neighborhood to the City of Graham; this may lead to connecting the
neighborhood to Graham’s sewage collection and treatment systems.
Additionally, DWQ assumed that, on average, there are 3 people per system. Assuming the average
concentration of septic waste reaching the stream is 1.0 x 104 counts/100 ml and that the septic
overcharge flow rate is 70 gallons/day/person (Horsely & Whitten, 1996), the contribution from
failing septic systems is 4.77 x 107 counts/hour. DWQ also assumed that 60% of the septic
overcharge reached the stream channel; this estimate is not scientifically based and was selected as a
seemingly moderate to high number for transport from a failing septic system to the stream
network. The loading rate from septic systems using these assumptions was 1.61 x 108 counts/30
days.
2.2.4 Urban Development/Sanitary Sewer Overflows
Fecal coliform bacteria can originate from various urban sources. These sources include pet waste,
runoff through stormwater sewers, illicit discharges/connections of sanitary waste, leaky sewer
systems and sanitary sewer overflows.
Fecal coliform accumulation rates on urban land cover were derived using the following:
1) the proportion of the watershed that is covered by high and low density developed land cover;
2) the types of urban land use that occurs in the watershed (e.g., residential, and heavy and light
commercial);
3) the fecal coliform build-up (accumulation) rates for each land-use in 2), as calculated from
instream stormwater samples collected by the United States Geological Survey (USGS) from
December 1993 to September 1997 in Mecklenburg County (Bales et al., 1999). In the USGS
study, each of the urban land uses was paired with a sample site. The land use descriptions and
calculated accumulation rates for fecal coliform may be seen in Table 3.
Final version submitted to EPA
10
Table 3. Rate of accumulation and maximum storage of fecal coliform by land use (from Bales et
al., 1999).
Land Use Rate of Accumulation
(count per acre per day)
Maximum Storage
(count per acre)
Residential 6.86 x 109 1.44 x 1010
Heavy Commercial 2.68 x 109 5.63 x 109
Light Commercial/Light Industrial 3.20 x 1010 6.72 x 1010
Woods/Brush 5.48 x 109 1.15 x 1010
A step-by-step description of the approach to determine initial urban fecal coliform accumulation
rates follows. DWQ first calculated the proportion of low and high-density land cover in the Town
Branch watershed. Using local knowledge of the watershed, the NC GIS land cover data were
converted into the four land use classes referenced by the USGS study (see Table 4). By combining
the proportions of the four land classes with the accumulation rates, DWQ assigned a
comprehensive urban accumulation rate of fecal coliform. The accumulation rate is an important
model parameter that describes how much fecal coliform is generated on each land use; actual fecal
coliform loading (delivery) to the stream network is determined through subsequent modeling.
Essentially, the model tracks fecal coliform build-up through the accumulation rate and simulates
fecal coliform wash-off as precipitation falls. More description of the model appears later in this
document.
Table 4. Estimated conversion from NC GIS land cover to land use in USGS study. Note that this
is for urban (developed) land cover only*
Watershed Urban Land cover classification
(from GIS database)
Land use classification
(estimated)
Town Branch 21.7% high density developed,
78.3% low density developed
16% light commercial/industrial
4% heavy commercial
51% light residential
29% woods/brush
By combining the information in Tables 3 and 4, DWQ calculated initial estimates of urban
accumulation and maximum storage. The results are shown in Table 5.
Final version submitted to EPA
11
Table 5. Initial (pre-calibration) estimates of accumulation and storage.
Watershed Rate of Accumulation
(count per acre per day)
Maximum Storage
(count per acre)
Town Branch 1.03 x 1010 1.85 x 1010
Since these numbers are based on studies in different, somewhat distant watersheds (Mecklenburg
Co.), they were subject to calibration in the model.
The city of Graham owns and operates a wastewater treatment plant and sewage collection system.
From 1997-2001, Graham reported seventeen sanitary sewer overflows (SSOs) of greater than 1000
gallons, including five SSOs of greater than 50,000 gallons. DWQ did not explicitly account for
SSOs in the modeling; rather, DWQ used a relatively high (compared to other land uses), constant
value for urban interflow fecal coliform concentration in the calibration model, which, along with a
calibrated urban accumulation rate may account for leaky sewers and infrequent SSOs.
2.2.5 Wildlife
Wildlife can be a source of fecal coliform bacteria in forest, wetland, pasture and cropland areas.
Wildlife deposit fecal material in these areas, which can be transported to a stream in a rain event.
Wildlife in Town Branch watershed is expected to include deer, raccoons, squirrels, and birds
(including waterfowl). DWQ derived population density estimates for all but squirrels and non-
waterfowl birds; consequently, these animals were not included in the model.
DWQ obtained estimates for deer population of 15-30 per square mile from the North Carolina
Wildlife Resources Commission (WRC, 2001). The lower end of the range (15) was applied to
cropland and pastureland, and the higher end of the range (30) was applied to forestland. DWQ
assumed that no wildlife live in areas with urban land cover.
There was very little basis for estimating populations of raccoon, duck and geese density – this is
one of many areas of uncertainty in the model. DWQ considered that there are 5 geese and 10
ducks per square mile of forestland and managed herbaceous land, and that there are 10 raccoons
Final version submitted to EPA
12
per square mile on such land. The numbers for geese, duck and raccoon are not scientifically based
and are intended as a rough, moderate estimate.
2.3 Source Assessment Conclusion
All of the aforementioned source assessment data were entered into a spreadsheet called Fecal Tool,
which calculates accumulation rates on the different land covers, and loading from direct sources
such as leaking septic systems and miscellaneous instream sources. TetraTech, Inc. developed Fecal
Tool. Output from this spreadsheet was used in the initial estimates for the corresponding
parameters in the water quality model. Some of the model parameter estimates calculated in the
spreadsheet were later altered through calibration (e.g., urban coliform accumulation rates).
3.0 MODELING APPROACH
An important component of the TMDL is to establish the relationship between instream water
quality and sources of fecal coliform. A model that simulates or statistically characterizes hydrology
and water quality is a helpful tool for this purpose. Models provide the relative contribution of the
sources, as well as the predictions of water quality resulting from changes in these source
contributions; these are the basic elements of Total Maximum Daily Load determination.
3.1 Model Framework
The model selected for this TMDL needed to meet several objectives:
1) To simulate watershed loading and instream transport of fecal coliform bacteria, and to capture
some of the temporal and spatial variation that those processes demonstrate.
2) To simulate instream fecal coliform concentrations over several years, so that critical conditions
(definition on page 2) may be identified. Critical conditions will be the basis for this TMDL.
3) To evaluate seasonal effects on the production and fate of fecal coliform bacteria.
EPA’s BASINS software includes a model, Nonpoint Source Model (NPSM), that is suited for
TMDL development. NPSM is based on another model, the Hydrologic Simulation Program –
FORTRAN (HSPF). Because it meets the objectives stated above, DWQ chose NPSM as the
model for this TMDL.
Final version submitted to EPA
13
NPSM (HSPF) is a dynamic watershed model capable of simulating nonpoint source runoff and
associated pollutant loads. It does this by tracking water and fecal coliform in the watershed.
Specifically, modules named PWATER and IWATER are used to calculate the components of the
water budget, and to predict the runoff from pervious and impervious areas, respectively (EPA,
1993). The model considers the following hydrologic processes: precipitation, interception, surface
runoff, interflow, groundwater, evaporation and evapotranspiration. Fluxes or storages within
subroutines of the model simulate these processes. Fecal coliform is simulated in the PQUAL and
IQUAL modules (from pervious and impervious land segment) using simple relationships with
water. Fecal coliform occurs in both the surface and subsurface outflow, though the former is
considered to be more complex in the model. On the surface, fecal coliform can be affected by
adhesion to the soil, and by light, wind, temperature and direct human influence. The approach is to
simulate fecal coliform using basic accumulation (build-up) and depletion rates, in concert with
depletion by wash-off; in other words, fecal coliform outflow from the surface is a function of the
water flow and the amount of fecal coliform in storage (EPA, 1993). Constant rates are assumed for
subsurface loading from the different land use categories.
NPSM (HSPF) performs flow routing and pollutant decay in stream reaches through the RCHRES
module. Flow is assumed to be unidirectional and decay is assumed to be first-order in nature (see
section 3.2.1 below). Also, NPSM allows discrete simulation of the required components of the
TMDL (e.g., WLA and LA components).
3.2 Model Setup
Town Branch was modeled as a single watershed (no subwatershed delineation) based on Reach File
3 (RF3) stream coverage and a digital elevation model of the area. The farthest downstream point of
the delineation was the DWQ ambient water quality sampling station, B1260000, or Town Branch at
SR 2109 near Graham. Dewberry and Davis Consultants compiled hourly meteorological data from
a weather station near Greensboro (see Figure 3 for specific location and its proximity to Town
Branch watershed). The meteorological data begin on 7/1/1996 and end on 8/11/2001. DWQ
measured the stream channel cross section near the watershed outlet (Appendix IV), and
incorporated this into the model’s f-table (table that describes relationship between streamflow
depth and streamflow volume).
Final version submitted to EPA
14
3.2.1 Instream Decay Rate
Once fecal coliform bacteria reach a waterbody, environmental factors influence the extent of their
growth and decay. Physical factors that influence the bacteria populations include photo-oxidation,
adsorption, flocculation, coagulation, sedimentation and temperature (USEPA, 1985). Chemical
toxicity, pH, nutrient levels, algae and the presence of fecal matter may also influence the fecal
coliform population. The water quality model utilizes a first order decay rate to calculate instream
decay of fecal coliform bacteria.
Ct = Coe-kt
Ct= coliform concentration at time t (cfu/100ml)
Co= initial coliform concentration (cfu/100ml)
k= decay rate constant (day-1)
t = exposure time (days)
Bacterial die-off has been modeled as a first-order decay equation, using a decay rate (k) between
0.7/day and 1.5/day (Center for Watershed Protection, 1999). Another study found that the
median decay rate for fecal coliform was 1.15/day (Lombardo, 1972); that value was used in the
Town Branch model for the existing condition and allocation runs.
3.3 Hydrologic Calibration
Because NPSM is driven by precipitation and by the subsequent treatment of the water budget, it is
important to calibrate the hydrologic parameters prior to calibrating the water quality parameters. In
the hydrologic calibration, simulated streamflows were compared to the historic streamflow data
recorded at a continuous stream gage. There is not a continuous gage in Town Branch, so instead
DWQ used one at Reedy Fork near Oak Ridge (USGS 02093800), which is about 30 miles away (see
Figure 2 below). The Reedy Fork watershed is relatively similar to the Town Branch watershed in
terms of shape, however it has proportionally less urban land cover and is considerably larger in area
(about 20 square miles to 4 square miles). To calibrate the model, hydrologic parameters, including
infiltration, upper and lower zone storage, groundwater storage and recession, interflow, and
evapotranspiration, were adjusted within a recommended range until the simulated and observed
hydrographs were as close as possible. DWQ determined the best match by assessing statistical fit;
more description is provided below.
Final version submitted to EPA
15
Figure 2.
Haw River
Greensboro
Weather Station
$T
Reedy Fork stream discharge gage
Haw River
Town Branch
City of Graham Interstates 40 & 85
Town Branch Watershed
Municipal boundaries
WDM Weather Data StationsMajor RoadsReach File, V1
Town Branch watershed
10 0 10 20 Miles
N
Regional View of Important Model Sites
A four-year period from 1/1/97 to 12/31/00 was used as the calibration period for the hydrologic
parameters. Relative fit of the modeled flow compared to the recorded flow is shown in Figure 3
below. The hydrologic parameters used to calibrate the model developed at the Reedy Fork gage
were assumed to apply to the Town Branch watershed, and were used to develop the water quality
model for Town Branch watershed.
Final version submitted to EPA
16
Figure 3. Simulated and observed flows recorded at USGS 02093800 Reedy Fork, 1997-2001.
Four conventional statistics for assessing model fit are correlation coefficient (r), the root mean
squared error (RMSE), mean error (ME) and mean absolute error (MAE). DWQ calculated these to
calibrate the model. Further description on each of the statistics is provided below, and the
formulas for these statistics appear in Appendix II; much of this follows Stow et al. (in review). A
log transformation was used because, as with most environmental data, fecal coliform has a log
normal distribution. The log transformation more equally weights low and high values when
calculating fit statistics (it gives less leverage to the relatively few high values).
Correlation coefficient, r, is a measure of the variability in the observed data that is explained by the
model. This was chosen instead of r2, because r2 hides some negative correlation. The closer it is to
1, the better. If the correlation between observed and predicted values is close to 1, the values don’t
necessarily track each other, they just tend to vary similarly (Stow et al., in review). For the Reedy
Fork hydrologic calibration, using log base 10 values, the four-year correlation coefficient is 0.79;
this is a typical level of prediction for hydrologic models (Stow, 2002).
Reedy Fork Hydrologic Calibration
1
10
100
1000
1/
1
/
9
7
3/
1
/
9
7
5/
1
/
9
7
7/
1
/
9
7
9/
1
/
9
7
11
/
1
/
9
7
1/
1
/
9
8
3/
1
/
9
8
5/
1
/
9
8
7/
1
/
9
8
9/
1
/
9
8
11
/
1
/
9
8
1/
1
/
9
9
3/
1
/
9
9
5/
1
/
9
9
7/
1
/
9
9
9/
1
/
9
9
11
/
1
/
9
9
1/
1
/
0
0
3/
1
/
0
0
5/
1
/
0
0
7/
1
/
0
0
9/
1
/
0
0
11
/
1
/
0
0
Fl
o
w

(
c
u
b
i
c

f
e
e
t

p
e
r

s
e
c
o
n
d
)
Observed Flow (cfs)NPSM Flow (cfs)
Final version submitted to EPA
17
The root mean squared error (RMSE) is the standard deviation of the model residuals, which are the
difference between the model predictions and observed data. A lower RMSE is better (0 is best),
though its value is relative to what the model attempts to predict; the mean of the observed data is a
good measure to compare. In this application, the RMSE is 0.19 and the observed mean is 1.20 (log
base 10 values), which indicates decent precision.
The average error (AE) and average absolute error (AAE) are measures of aggregate model bias.
Again, closer to 0 is better, though for the mean error, values near zero can be misleading since
negative and positive differences can offset each other (Stow et al., in review). For the Reedy Fork
hydrologic calibration, the AE and AAE were 0.01 and 0.13, respectively. This indicates that the
model does not over- or under-predict consistently, and again, that the precision is good.
Table 6. Hydrologic calibration fit statistics
Statistic Value (using log base 10)
r 0.79
RMSE 0.19
AE 0.01
AAE 0.13
Observations (n)1,461
Obs. mean 1.20
DWQ tested the Reedy Fork calibration on North Buffalo Cr. and found it to be a reasonable fit
there as well. North Buffalo Cr. is in between Reedy Fork and Town Branch, and has a proportion
of urban land cover that is more similar to the Town Branch watershed.
3.4 Water Quality Calibration
Once the hydrologic calibration is complete the model is calibrated for water quality by adjusting
parameters until simulated and observed fecal coliform concentrations achieve acceptable
agreement. To calibrate the model, several parameters were adjusted including the accumulation
rates of fecal coliform bacteria, wash-off rates, maximum storage of fecal coliform bacteria and
contributions from direct sources. By matching the trends in simulated and observed
Final version submitted to EPA
18
concentrations resulting from a variety of streamflows, the model may be a reasonable predictor of
instream water quality. Selected values from the water quality calibration may be seen in Appendix
III.
Through model calibration, DWQ estimated that the constant instream source (see miscellaneous
sources, Section 2.2.2) is 7.2 x 106 counts/30 days. Illicit discharges are assumed to be included in
this estimate.
DWQ focused on calibrating urban coliform accumulation and maximum storage rates in the Town
Branch watershed, as that was identified as the primary source through the source assessment.
Through calibration, these values decreased by approximately 80% from the values shown in Table
5 (estimates from Mecklinburg Co. USGS study). It seems rather evident that the Charlotte values
are much too high for the Town Branch watershed. DWQ also calibrated urban interflow
concentration, which was considered to be a constant value in the model.
The calibration period for the water quality model spanned October, 1996 into August, 2001. The
beginning time was limited by the meteorological file, which began in July, 1996. DWQ allowed
three months for the model to stabilize before comparing predicted and observed data. The
meteorological file also limited the simulation end time.
3.4.1 Calibration Results
Fecal coliform samples collected at B1260000 (the ambient monitoring station) between October,
1996 and August, 2001 were compared to simulated concentrations and rainfall collected at the
meteorological stations. The calibration objective was to obtain the best fit to the observed data, as
determined by the same statistics used in the hydrologic calibration (r, RMSE, AAE, AE; see Section
3.3). Graphical fit of the model to observed data is shown in Figures 4 and 5. These figures indicate
that the model does a fair job at simulating the response of fecal coliform bacteria over time, with
variations in flow. The model calibration statistics for fecal coliform are not nearly as good as those
for the hydrologic calibration. Additionally, the model results indicate that instream fecal coliform
concentrations following rainfall events frequently exceed the fecal coliform standard, while
instream concentrations during dry periods typically do not exceed the standard. Finally, a plot of
modeled flow and fecal coliform concentration appears in Appendix V. One important concept that
Final version submitted to EPA
19
this figure demonstrates is higher fecal coliform concentrations occur in Town Branch following
rain events.
Table 7. Water quality calibration fit statistics
Statistic Value (using log base 10)
r 0.47
RMSE 0.79
AE -0.38
AAE 0.72
Observations (n)50
Obs. mean 2.58
Figure 4. Simulated versus observed fecal coliform concentrations from 9/1996 to 3/1999
10
100
1000
10000
9/22/96 12/31/96 4/10/97 7/19/97 10/27/97 2/4/98 5/15/98 8/23/98 12/1/98 3/11/99
DATE
FE
C
A
L

C
O
L
I
F
O
R
M

(
#
/
1
0
0

m
L
)
0
1
2
3
4
5
6
RA
I
N
F
A
L
L

(
i
n
/
d
a
y
)
RAINFALL (IN/DAY)MODEL OUTPUT OBSERVED DATA

Final version submitted to EPA
20
Figure 5. Simulated versus observed fecal coliform concentrations from 3/1999 to 8/2001
The low correlation coefficient indicates a lack of predictive power. On the other hand, the
relatively small RMSE, AE and AAE suggest that the model may predict the mean fairly well, but
there may be a lot of individual scatter about the mean.
To resolve this discrepancy, DWQ examined modeling efficiency, which is calculated using the
following formula:
å
åå
=
==
-
÷÷
ø
ö
çç
è
æ ---
n
i
n
i
n
i
OOi
OiPiOOi
1
2
2
11
2
)(
)()(
where:
O_ = mean of the observations
Oi = ith of n observations
Pi = ith of n predictions
10
100
1000
10000
3/31/99 7/9/99 10/17/99 1/25/00 5/4/00 8/12/00 11/20/00 2/28/01 6/8/01 9/16/01
DATE
FE
C
A
L

C
O
L
I
F
O
R
M

(
#
/
1
0
0

m
L
)
0
1
2
3
4
5
6
RA
I
N
F
A
L
L

(
i
n
/
d
a
y
)
RAINFALL (IN/DAY)MODEL OUTPUT OBSERVED DATA

Final version submitted to EPA
21
The modeling efficiency measures how well the model predicts relative to the average of the
observations. In terms of the its value, the following guidelines apply:
· The closer to 1, the better. A value near 1 indicates a close match between model predictions
and observations.
· If > 0, then the model predicts better than the average of the observations.
· If < 0, then the mean of the observations predict better than the model.
· If = 0, then the model predicts individual observations no better than the average of the
observations.
The modeling efficiency for the Town Branch coliform model is –0.40. This confirms that the
model predictive precision is low.
3.4.2 Prediction Uncertainty
The inability to accurately simulate specific observed data points can sometimes be attributed to
more specific aspects of a model, such as differences in rainfall at the meteorological gage and in the
watershed, or illicit point discharges. More often though, the lack of agreement between modeled
and observed fecal coliform is due to the general high degree of uncertainty associated with
predicting any water quality variable, especially fecal coliform. Prediction uncertainty comes from a
number of sources, including (from Reckhow and Chapra, 1983 and Reckhow, 1995):
· Gaps in our scientific knowledge.
· Natural variability – spatial and temporal variability in chemistry, hydrology and ecology is great.
Model predictions are on a much coarser scale than what occurs in nature.
· Measurement error – measurement of fecal coliform in the field and laboratory has error.
· Aggregation error – with increased endpoint specificity (space and time), the uncertainty
associated with the prediction increases.
· Model error:
· Mis-specification – model expressions that characterize processes may be wrong.
· Error in parameters – reaction rates may be inappropriate.
· Error in model inputs – e.g., loading terms such as accumulation rate of bacteria have error.
Final version submitted to EPA
22
Unfortunately, many water quality models employed for TMDL analysis, including NPSM, are not
adept at characterizing prediction uncertainty. With these models, all we know is that the
uncertainty is certain to be large. Emphasizing an adaptive management approach is one way to
address this. Specifically, the model may guide initial decision making, but continued observation of
the watershed and creek, as fecal coliform controls are implemented (e.g., exclusion fencing, leaky
sanitary sewer repair), is expected to be our best approach for determining the appropriate level of
management. This is especially true, considering the results of the water quality model calibration,
for the Town Branch TMDL.
4.0 Total Maximum Daily Load
A Total Maximum Daily Load is the maximum amount of a pollutant that a water body can receive
and still meet water quality standards, and an allocation of that amount among point and nonpoint
sources. A TMDL comprises the sum of wasteload allocations (WLA) for point sources, load
allocations (LA) for nonpoint sources, and a margin of safety. This definition is expressed by the
equation:
TMDL = S WLA + S LA + MOS
The objectives of the TMDL are to estimate allowable pollutant loads, and to allocate to the known
pollutant sources in the watershed, so the appropriate control measures can be implemented and the
water quality standard can be achieved.
The TMDL will be expressed in units of counts/30 days, as this is the period over which the water
quality target/standard is evaluated.
The two main components of a TMDL, the reduction target, including a margin of safety, and the
allocation strategy will be presented in the following sections.
4.1 Reduction Target
To determine the amount of fecal coliform load reduction necessary to comply with the water
quality standard, the period of critical conditions must be established. Critical conditions, and
loading that represents current conditions and TMDL conditions were determined in the following
manner:
Final version submitted to EPA
23
1) The calibrated model was rerun for the entire, nearly 5-year period.
2) Simulated fecal coliform concentrations for the nearly 5-year period were plotted as rolling 30-
day geometric mean concentrations. A 30-day geometric mean is determined by calculating the
geometric mean of an individual day’s fecal coliform prediction and the daily predictions for the
29 days that precede it. The rolling aspect is achieved by moving to the next day and performing
the same calculation (the earliest date from the previous 30-day geometric mean calculation
would be dropped).
3) DWQ applied load reductions in the model until all of the 30-day geometric means were below
the standard. The date on which the last 30-day geometric mean peak falls below the criterion
determines when critical conditions occur. The 30-days prior to and including this date form the
critical conditions period, during which the TMDL is determined.
4) For the 30-day critical period, DWQ calculated current loading from the calibrated model,
before any load reductions were taken.
5) Next, again using the 30-day critical period, DWQ calculated loading during TMDL conditions
(after load reductions were taken), when the predicted instream fecal coliform concentrations
indicated no exceedances of the water quality standard. With no predicted exceedances of the
standard, the model fulfills the TMDL criterion of allowing the maximum load while still
achieving water quality standards.
6) For both the current condition case and the TMDL condition case, the simulated daily fecal
coliform loads from sources such as runoff from all lands, leaking septic systems and
miscellaneous sources were summed for the 30-day critical period.
These six steps will be further explained in the following sections.
4.1.1 Critical Conditions
The Town Branch fecal coliform monitoring data indicate that elevated fecal coliform levels occur
throughout the year, primarily during wet weather conditions. To determine when critical
conditions occur, DWQ applied load reductions in the model until all of the 30-day geometric
means were below the standard. The date on which the last 30-day geometric mean peak falls below
the criterion determines when critical conditions occur. By definition, the 30-days prior to and
Final version submitted to EPA
24
including this date form the critical conditions period. See Figure 6 below. For the Town Branch
fecal coliform model, this occurred on February 5, 1998. The critical conditions occur between
January 7, 1998 and February 5, 1998.
Figure 6. Reduction of rolling 30-day geometric mean of fecal coliform from existing loading to
TMDL allocation.
Rain was recorded in Greensboro on 17 days during the 30-day critical period; at least 0.10 inch fell
on 11 of those days. To wit, November and December, 1997 had 2.19 and 2.66 inches of
precipitation, respectively, while the 30-day critical period had 7.62 inches of precipitation. Include
the day before the critical period (January 6) and the total increases to 8.35 inches.
Relatively high fecal coliform accumulation rates resulted from the model calibration (see Section
3.4), and consequently, rainy periods such as the critical conditions period produce high instream
0
200
400
10/12/96 4/30/97 11/16/97 6/4/98 12/21/98 7/9/99 1/25/00 8/12/00 2/28/01 9/16/01
DATE
PREDICTED (EXISTING)TMDL ALLOCATION WATER QUALITY CRITERION

critical conditions
2/5/1998
Final version submitted to EPA
25
fecal coliform concentrations, as even small (>0.10 inch precipitation) events wash recently
accumulated bacteria into the stream network.
4.1.2 Current Conditions
Current loading conditions were calculated by summing the loading from the critical 30-day period
before any reductions were taken (calibrated model).
DWQ separated the principal coliform source categories, as used in NPSM, in Table 8; these include
runoff from all lands, leaking septic systems and miscellaneous sources. Runoff from all lands
includes estimated fecal coliform load from deposits by wildlife, as well as an estimate of loading
from urban areas. Leaking septic systems estimates loading related to septic systems.
Miscellaneous sources is an estimate of loading from unknown, or illicit, instream sources.
According to the model, storm-driven runoff from all land provides the largest load of fecal
coliform bacteria to the stream. Loads from miscellaneous sources are constant loads that are
applied directly to the stream; these sources will have the greatest impact on instream water quality
during periods of low flow.
Table 8. Summary of predicted existing coliform loads in the Town Branch watershed.
Runoff from all lands1
(counts/30 days)
Leaking septic systems
(counts/30 days)
Miscellaneous sources2
(counts/30 days)
Instream conc.3
(counts/100 ml)
1.03 x 1012 1.61 x 108 7.20 x 106 350
1 Includes urban runoff and wildlife.
2 Includes illicit discharges.
3 Maximum simulated concentration during the critical period (geometric mean).
4.1.3 Margin of Safety
TMDL conditions were calculated by summing the loading from the critical 30-day period after
reductions were taken to bring all of the 30-day geometric means below the criterion. Before this is
done, however, a margin of safety must be included.
A TMDL requirement is that a margin of safety must be included to provide further insurance that
the impaired waterbody will meet its designated uses once load reductions are realized. The margin
of safety may be accounted for implicitly, through conservative (more protective of water quality)
model assumptions, or explicitly, by reserving a portion of the allocated load. The Town Branch
Final version submitted to EPA
26
TMDL includes explicit and implicit margins of safety; more explanation on the margin of safety
follows below.
In Figure 7 below, observe that the target for the rolling 30-day geometric mean of fecal coliform is
170 counts/100 ml, instead of the standard of 200 counts/100 ml. By using this lower target, DWQ
provides an explicit margin of safety for the Town Branch TMDL. This explicit margin of safety
may be interpreted to account for 15% greater assurance of achieving the instream water quality
target. [(200-170)/200]*100 = 15%
Also, an implicit margin of safety is included because the model assumes that bacteria delivered
from the land surface do not decay as it travels from its source to the stream network; in reality,
some decay would occur, though it is difficult to estimate or model.
Figure 7. Reduction of rolling 30-day geometric mean of fecal coliform to include explicit margin of
safety.
0
200
400
10/12/96 4/30/97 11/16/97 6/4/98 12/21/98 7/9/99 1/25/00 8/12/00 2/28/01 9/16/01
DATE
PREDICTED (EXISTING)TMDL ALLOCATION WATER QUALITY CRITERION

Final version submitted to EPA
27
4.1.4 TMDL Allocation
To calculate the TMDL, DWQ summed the loads from the critical 30-day period after reductions
were taken to bring all of the 30-day geometric means below the criterion, plus the explicit margin of
safety (170 counts/100 ml).
The allocation strategy for the Town Branch fecal coliform TMDL is limited to nonpoint sources, as
there are no permitted point sources in the watershed. An allocation scenario that predicts
compliance with the instream water quality criterion and the required reductions from the individual
categories may be seen in Table 9. A nonlinear relationship between fecal coliform inputs and
instream fecal coliform concentration produces an inequitable reduction in load and concentration
in percent terms (51% reduction in concentration requires 60 or more % reduction in load).
Table 9. Allocation strategy by major nonpoint sources for TMDL conditions
Runoff from all lands
(counts/30 days)
Leaking septic systems
(counts/30 days)
Miscellaneous Sources2
(counts/30 days)
Instream f.c. concentration1
(counts/100 ml)
3.25 x 1011 6.46 x 107 2.88 x 106 170
68% reduction 60% reduction 60% reduction 51% reduction
1 Maximum simulated instream concentration during critical period. Percent reduction represents
the difference in simulated instream concentration between the existing loads (Table 8., 350
counts/100 ml) and TMDL allocation scenario (this table, 170 counts/100 ml).
2 Should not include illicit discharges.
The nonpoint sources are summed to produce a load allocation (LA), which is displayed in Tables
10 and 11 below.
Tables 10 and 11. Allocation strategy by TMDL component for Town Branch.
In terms of load:
Wasteload allocation (WLA)
(counts/30 days)
Load allocation (LA)
(counts/30 days)
Explicit Margin of safety
(MOS)1
(counts/30 days)
TMDL
(counts/30 days)
0 3.25 x 1011 0.90 x 1011 4.15 x 1011
1 Explicit margin of safety is equal to 22% since the load allocation is reduced by that much from
the TMDL allocation (e.g., [(4.15 x 1011 - 3.25 x 1011)/4.15 x 1011] * 100 = 22%).
Final version submitted to EPA
28
In terms of concentration:
Wasteload allocation (WLA)
(counts/100 ml)
Load allocation (LA)
(counts/100 ml)
Explicit Margin of safety
(MOS)1
(counts/100 ml)
TMDL
(counts/100 ml)
0 170 30 200
1 Explicit margin of safety is equal to 15.0% since the instream water quality target is reduced to 170
counts/100 ml from 200 counts/100 ml (e.g., [(200-170)/200] = 15%).
The explicit margin of safety differs depending on whether it is viewed in terms of concentration or
load (15% vs. 22%, respectively). This is because of a nonlinear relationship between load and
concentration.
The implicit margin of safety, from the assumption that fecal coliform bacteria do not decay as they
are transported from the land surface to the stream network, is not quantified nor included in the
tables above. Basically though, by not including this decay, the listed load allocation is higher than if
the decay were included. Consequently, the actual (expected) load allocation will be lower than what
is shown in Tables 10 and 11. Because less fecal coliform should actually be delivered to the
streams, therein lies the implicit margin of safety.
4.2 Instantaneous portion of standard
To assess the instantaneous portion of the fecal coliform standard, DWQ considered observed data
and predictions from the full modeling period. The observed data from 1995 to 2001 show that
44% of the samples were over 400 counts/100 ml. The calibrated model has 28% of the daily
predictions over 400 counts/100 ml. The TMDL model that meets the geometric mean part of the
fecal coliform standard (overall about a 70% reduction) drops the percent of daily predictions above
400 counts/100 ml to less than 19% over the nearly 5 year model period.
4.3 Seasonal variation
DWQ used a nearly 5-year simulation period to assess the TMDL. This longer period allows for
consideration of seasonal variation.
Final version submitted to EPA
29
5.0 SUMMARY AND FUTURE CONSIDERATIONS
The sources of fecal coliform in the Town Branch watershed include urban sources in the Graham
area, and wildlife in the forested areas of the watershed. The Nonpoint Source Model in EPA’s
Basins software was used to simulate instream fecal coliform concentrations and to allocate the fecal
coliform loads to the various sources. In order for the water quality target to be met, the final
allocation of the fecal coliform requires the major sources (not wildlife as that is considered part of
background) to reduce loading by approximately 70%. DWQ considers the major source to be
runoff from urban areas, possibly including leaking sewer lines. In fact, based on the special study
monitoring which is shown in Appendix I, and based on the model results, it appears that if Town
Branch is to meet the fecal coliform standard, the City of Graham must considerably reduce loading
from runoff of urban areas, as well as address SSOs, possibly through infrastructure repair. As
evidenced by Appendix V, the fecal coliform problem in Town Branch occurs following rain events.
Additional monitoring is required to identify specific sources of fecal coliform, in terms of both area
and animal. Finally, it would be beneficial to the fecal coliform impairment of Town Branch if the
City of Graham would annex, and connect to the sewer system, the neighborhood in the southern
portion of the watershed where septic system failure rate is high (see Section 2.2.3 Failing Septic
Systems).
5.1 Monitoring
Fecal coliform monitoring will continue on a monthly interval at the ambient monitoring site
(SR2109). The continued monitoring of fecal coliform concentrations will allow for the evaluation
of progress towards the goal of reaching water quality standards. In addition to this data collection,
further fecal coliform monitoring may be considered. Additional monitoring could focus on fecal
coliform source assessment in the watershed; this would further aid in the evaluation of the progress
towards meeting the water quality standard. Also, a synoptic survey of instream fecal coliform
during storm events may improve model calibration during those important loading events.
To comply with EPA guidance, North Carolina may adopt new bacteria standards in the near future
using Escherichia coli (E. coli) and enterococci. Thus, future monitoring efforts to measure
compliance with this TMDL should include E. coli and enterococci. Per EPA recommendations
(EPA, 2000b), if future monitoring for E. coli/enterococci indicates the standard has not been
Final version submitted to EPA
30
exceeded, these monitoring data may be used to support delisting the water body from the 303(d)
list. If a continuing problem is identified using E. coli/enterococci, the TMDL may be revised.
5.2 Implementation
An implementation plan is not included in this TMDL. The involvement of local governments and
agencies will be needed in order to develop the implementation plan. Graham was very helpful in
the development of this TMDL; they conducted a monitoring program in 2001 to assess compliance
with the water quality criterion, and to better identify source areas (see Appendix I). Additionally, a
local planning agency (Piedmont Triad Council of Governments), with collaboration from DWQ,
developed a proposal for Section 319 funds to implement the Town Branch fecal coliform TMDL.
In June, that proposal was awarded and work is scheduled to begin in 2003.
6.0 PUBLIC PARTICIPATION
The City of Graham was notified of DWQ’s intention to develop the Town Branch fecal coliform
TMDL. The county extension service supplied information on wildlife and livestock to aid in the
source assessment portion of the TMDL.
To publicly notice the Town Branch fecal coliform TMDL, DWQ submitted a legal advertisement
to the daily newspaper for nearby Burlington. The advertisement appeared in this newspaper on
June 19, 2002. Additionally, also on June 19, 2002, DWQ electronically distributed a draft of the
TMDL and public comment information to the known interested parties. Finally, DWQ held a
public meeting in Graham on June 24, 2002 to present the TMDL and offer opportunity for
questions and comments by the public. Eleven people, mostly employees of the City of Graham,
attended this meeting.
Final version submitted to EPA
31
References Cited
Alabama Department of Environmental Management. 2001. Draft Big Nance Creek TMDL
Development for Fecal Coliform. Prepared by the Water Quality Branch and Tetra Tech, Inc.
American Society of Agricultural Engineers (ASAE). 1998. ASAE Standards, 45th Edition, Standard
Engineering Practices Data.
Bales, J.D., Weaver, J.C., and Robinson, J.B. 1999. Relation of Land Use to Streamflow and Water
Quality at Selected Sites in the City of Charlotte and Mecklenburg County, North Carolina, 1993-
1998. U.S. Geological Survey, Raleigh NC. Water Resources Investigations Report 99-4180.
Canale, R.P., Auer, M.T., Owens, E.M., Heidtke, T.M., and S.W. Effler. 1993. Modeling Fecal
Coliform Bacteria-II. Model Development and Application. Wat. Res. 27(4):703-714.
Center for Watershed Protection. 1999. Microbes and Urban Watersheds: Concentrations, Sources
and Pathways. Watershed Protection Techniques 3(1): 554:565.
Davis, M. USEPA, Region IV. 2002. Personal communication.
Division of Environmental Health (DEH). NCDENR. 2000. Report on the Proper Maintenance of
Septic Tank Systems in Accordance with Section 13.5 of HB 1160 (Clean Water Act of 1999). March
15, 2000.
Horner, 1992. Water Quality Criteria/Pollutant Loading Estimation/Treatment Effectiveness
Estimation. In R.W. Beck and Associates, Covington Master Drainage Plan, King County Surface
Water Management Division, Seattle, Washington.
Horsley and Witten. 1996. Identification and evaluation of nutrient and bacterial loadings to
Maquiot Bay, New Brunswick and Freeport, Maine. Final Report. Casco Bay Estuary Project,
Portland, ME.
Lombardo, P.S. 1972. Mathematical Model of Water Quality in Rivers and Impoundments.
Hydrocomp, Inc., Mountain View, California.
McDaniel, A.H. Dewberry & Davis Consultants, Inc. 2001. Personal communication.
Mecklenburg County Department of Environmental Protection and NC Department of
Environment and Natural Resources. 2001. Fecal Coliform Total Maximum Daily Load for the
Irwin, McAlpine, Little Sugar and Sugar Creek Watersheds, Mecklenburg County, North Carolina.
North Carolina Department of Agriculture. 2001. Agricultural Statistics Division-County Statistics.
www.ncagr.com/stats/cntysumm/ and http://govinfo.library.orst.edu/cgi-bin/ag-
state?North+Carolina.
Final version submitted to EPA
32
North Carolina State University (NCSU). 1994. Livestock Manure Production and Characterization
in North Carolina. North Carolina Cooperative Extension Service. College of Agriculture and Life
Sciences, Raleigh, NC.
NC Wildlife Resources Commission. 2001. Estimates for deer and turkey populations in
Rockingham Co. NC Department of Environment and Natural Resources, Raleigh, NC.
Quick, V. City of Graham Treatment Plants Superintendent. 2002. Personal communication.
Sullivan, J. City of Graham. 2002. Public comment and personal communication.
Reckhow, K.H. and S.C. Chapra. 1983. Engineering Approaches for Lake Management. Volume I:
Data Analysis and Empirical Modeling, Ann Arbor Science Pub. Inc.
Reckhow, K.H. 1995. Duke University Water Quality Modeling course notes.
Stow, C.A. Duke University. 2002. Personal communication.
Stow, C.A., Roessler, C., Borsuk, M.E., and Reckhow, K.H. A Comparison of Estuarine Water
Quality Models for TMDL Development in the Neuse River Estuary. In review.
Tetra Tech, Inc. 2001. Fecal Tool User’s Guide. Draft report submitted to USEPA. Fairfax,
Virginia.
U.S. Environmental Protection Agency (USEPA). 1991. Guidance for Water Quality-Based
Decisions: The TMDL Process. Assessment and Watershed Protection Division, Washington, DC.
U.S. Environmental Protection Agency, Federal Advisory Committee (FACA). Draft final TMDL
Federal Advisory Committee Report. 4/28/98.
U.S. Environmental Protection Agency (USEPA). 2000a. Revisions to the Water Quality Planning
and Management Regulation and Revisions to the National Pollutant Discharge Elimination System
Program in Support of Revisions to the Water Quality Planning and management Regulation; Final
Rule. Fed. Reg. 65:43586-43670 (July 13, 2000).
U.S.Environmental Protection Agency (USEPA). 2000b. Implementation Guidance for
Ambient Water Quality Criteria for Bacteria – 1986. DRAFT. Office of Water. EPA-823-D-
00-001.
U.S. Environmental Protection Agency (USEPA). 1985. Rates, constants, and kinetics
formulations in surface water quality modeling (II ed.). Athens, GA: EPA-600-3-85-040.
U.S. Environmental Protection Agency (USEPA). 1998. Better Assessment Science Integrating
Point and Nonpoint Sources (BASINS), Version 2.0 User’s Manual. Office of Water, Washington
DC. EPA-823-B-98-006.
Final version submitted to EPA
33
U.S. Environmental Protection Agency (USEPA). 1993. Hydrologic Simulation Program –
FORTRAN. User’s Manual for Release 10. Office of Research and Development, Washington DC.
EPA-600-R-93-174.
U.S.Environmental Protection Agency (USEPA). 2001. Protocol for Developing Pathogen TMDLs.
Office of Water, Washington DC. EPA 841-R-00-001.
U.S. Environmental Protection Agency (USEPA). 2001. Total Maximum Daily Load Development
for Fecal Coliform in the Crooked Creek Watershed, Cullman County, Alabama. USEPA Region
IV, Atlanta, Georgia.
U.S.Environmental Protection Agency (USEPA). 2000. BASINS Technical Note 6. Estimating
Hydrology and Hydraulic Parameters for HSPF. Office of Water, Washington DC. EPA 823-R-00-
012.
http://www.epa.gov/waterscience/basins/tecnote6.html
United States Geological Survey (USGS). 2001. Water Resources of the United States. NWIS web
online hydrologic data: http://water.usgs.gov.
Walker, P. Alamance County Agriculture Extension. 2002. Personal communication.
Final version submitted to EPA
34
APPENDIX I. OBSERVED DATA
Table A. Data from DWQ Ambient Station – Town Branch at SR 2109
DATE Fecal Coliform (#/100 ml)DATE Fecal Coliform (#/100 ml)
1/19/95 110 10/28/98 10
2/23/95 260 11/19/98 120
3/22/95 680 1/21/99 18
4/12/95 640 4/22/99 140
5/18/95 620 7/19/99 690
6/21/95 9100 7/12/99 230
7/24/95 4400 8/16/99 830
9/26/95 220 9/29/99 2100
10/23/95 380 11/18/99 20
12/20/95 230 2/22/00 10
1/25/96 270 3/21/00 1700
2/26/96 27 4/19/00 250
3/26/96 320 5/31/00 150
4/17/96 320 6/26/00 2500
5/29/96 4000 7/25/00 3700
6/24/96 1600 8/24/00 740
7/24/96 400 9/20/00 540
8/27/96 200 10/18/00 390
9/24/96 310 11/30/00 4300
10/23/96 310 12/14/00 210
11/19/96 2400 1/31/01 450
12/17/96 18 4/25/01 3500
1/30/97 210 6/19/01 2200
2/26/97 19 8/30/01 110
3/26/97 870 9/26/01 640
4/28/97 1200 10/23/01 120
5/27/97 5300 11/28/01 48
6/25/97 100
7/17/97 4900
8/26/97 240
9/16/97 450
10/22/97 910
11/24/97 2400
12/18/97 45
1/27/98 1200
2/18/98 6400
3/24/98 140
4/29/98 350
5/26/98 4600
6/25/98 63
7/27/98 220
8/24/98 100
9/29/98 230
Final version submitted to EPA
35
Special Study by City of Graham during July and August, 2001. The purposes of this study were
to evaluate whether the creek was complying with the state fecal coliform standard, and to provide
information on bacteria source areas in watershed.
#
#
#
#
#
#
Town Branch
County Home Br.
Graham Haw River
I-85
Hwy 54
12
3
N
Sampling Sites for Town Branch
Site 1 - SR2109, Town Branch at Cooper Rd.
Site 2 - Upper Town Branch at Graham WWTP entrance
Site 3 - County Home Branch at Gilbreath St.
Heavy rains occurred on 7/4/2001, which explains the extremely high fecal coliform concentrations
on 7/5/2001.
As will be evidenced below, Town Branch had fecal coliform levels that exceed the water quality
criterion, and consequently, that pollutant impairs one of the stream’s designated uses. County Home
Branch, which is characterized by low density residential development, typically had the highest fecal
coliform levels.
Final version submitted to EPA
36
Individual Sample Results
Site One is at SR2109 - Town Branch at Cooper Rd.
Date Counts/100 ml
7/5/01 10,571
7/11/01 210
7/16/01 192
7/18/01 156
7/25/01 724
8/1/01 620
8/8/01 1645
8/13/01 417
8/15/01 438
8/22/01 317
8/29/01 276
Graham Special Study of Fecal Coliform on Town Branch
and County Home Branch
100
1,000
10,000
100,000
Date
Site 1
Site 2
Site 3
Final version submitted to EPA
37
Site Two - upper Town Branch at Graham WWTP entrance
Date Counts/100 ml
7/5/01 17,560
7/11/01 230
7/16/01 191
7/18/01 156
7/25/01 279
8/1/01 460
8/8/01 249
8/13/01 852
8/15/01 440
8/22/01 597
8/29/01 192
Site Three - County Home Branch at Gilbreath St.
Date Counts/100 ml
7/5/01 61,000 >
7/11/01 260
7/16/01 245
7/18/01 303
7/25/01 938
8/1/01 800
8/8/01 800 >
8/13/01 12,300 >
8/15/01 233
8/22/01 218
8/29/01 92
Final version submitted to EPA
38
Summary of Town Branch Fecal Coliform Data from City of Graham Special Study
(for station locations see map on next page)
Station Description Dates Number of Days Observations Geometric Mean
Town Branch at SR 2109 7/5/2001 to 8/3/2001 30 6 557
Town Branch at SR 2109 7/11/2001 to 8/9/2001 30 6 408
Town Branch at SR 2109 7/16/2001 to 8/14/2001 30 5 457
Town Branch at SR 2109 7/18/2001 to 8/16/2001 30 6 525
Town Branch at SR 2109 7/25/2001 to 8/22/2001 30 6 591
Town Branch at SR 2109 8/1/2001 to 8/29/2001 29 6 504
Station Description Dates Number of Days Observations Geometric Mean
Town Branch above WWTP 7/5/2001 to 8/3/2001 30 6 499
Town Branch above WWTP 7/11/2001 to 8/9/2001 30 6 246
Town Branch above WWTP 7/16/2001 to 8/14/2001 30 5 305
Town Branch above WWTP 7/18/2001 to 8/16/2001 30 6 351
Town Branch above WWTP 7/25/2001 to 8/22/2001 30 6 439
Town Branch above WWTP 8/1/2001 to 8/29/2001 29 6 412
Station Description Dates Number of Days Observations Geometric Mean
County Home Branch at Gilbreath 7/5/2001 to 8/3/2001 30 6 980
County Home Branch at Gilbreath 7/11/2001 to 8/9/2001 30 6 476
County Home Branch at Gilbreath 7/16/2001 to 8/14/2001 30 5 905
County Home Branch at Gilbreath 7/18/2001 to 8/16/2001 30 6 897
County Home Branch at Gilbreath 7/25/2001 to 8/22/2001 30 6 849
County Home Branch at Gilbreath 8/1/2001 to 8/29/2001 29 6 577
Station Description Dates Number of Days Observations Geometric Mean
All Stations 7/5/2001 to 8/3/2001 30 18 648
All Stations 7/11/2001 to 8/9/2001 30 18 363
All Stations 7/16/2001 to 8/14/2001 30 15 502
All Stations 7/18/2001 to 8/16/2001 30 18 549
All Stations 7/25/2001 to 8/22/2001 30 18 604
All Stations 8/1/2001 to 8/29/2001 29 18 493
Final version submitted to EPA
39
APPENDIX II. FIT STATISTICS FOR MODEL CALIBRATION
1) r – the correlation coefficient of the model predictions and observations:
r =
å å
å
==
=
--
--
n
i
n
i
ii
n
i
ii
PPOO
PPOO
1 1
22
1
)()(
))((
,
2) RMSE - the root mean squared error:
RMSE = n
OPn
i
iiå
=
-
1
2)(
,
3) RI – the reliability index
RI = å=÷÷ø
ö
ççè
æn
i i
i
P
O
n 1
2
log1exp ,
4) AE – the average error
AE = OP
OPn
i
ii -=
-å=1
)(
,
5) AAE – the average absolute error
AAE = n
OPn
i
iiå
=
-
1 , and
Final version submitted to EPA
40
6) MEF – the modeling efficiency
MEF = å
åå
=
==
-
÷ø
öçè
æ ---
n
i
i
n
i
ii
n
i
i
OO
OPOO
1
2
1
2
1
2
)(
)()(
,
where n = the number of observations, Oi = the ith of n observations, Pi = the ith of n predictions,
and O and P are the observation and prediction averages, respectively. All observations and
predictions were log-transformed before calculating the fit statistics so that differences between
predicted and observed values would not be highly skewed and dominated by a small proportion
of high values.
Final version submitted to EPA
41
APPENDIX III. MODEL CALIBRATION INFORMATION
Calibrated Hydraulic Parameters for HSPF application to Reedy Fork
(later applied to Town Branch)
Parameter Description/Units Calibration value Typical range*
LZSN Lower zone nominal storage (inches)7.0 3 - 8
INFILT Soil infiltration rate (in./hr.)0.10 0.01 – 0.25
LSUR Length of assumed overland plane
(ft.)
300 200 - 500
SLSUR Slope of assumed overland plane 0.035 0.01 – 0.15
AGWRC Groundwater recession rate ( /day)0.99-forest,
0.96-nonforest
0.92 – 0.99
UZSN Upper zone nominal storage (inches)0.56 0.10 – 1.0
INTFW Interflow inflow (no units)0.60 Default is 0.75
IRC Interflow recession coefficient 0.40 0.5 – 0.7
LZETP Lower zone evapotranspiration 0.20 0.2 – 0.7
* From Basins Technical Note 6 – Estimating Hydrology and Hydraulic Parameters for HSPF
Calibrated Water Quality Parameter for HSPF application to Town Branch
Land use/PARAMETER ACQOP SQOLIM WSQOP IOQC
Urban 2 x 109 3.6 x 109 0.7 6.7 x 105
Managed Herbaceous 1.95 x 108 3.52 x 108 1.5 11,300
Cropland 1.95 x 106 3.52 x 106 1.5 11,300
Forest 1.95 x 106 3.52 x 106 2.8 5,670
ACQOP is the accumulation rate (in count/acre/day )
SQOLIM is the maximum storage rate (in count/acre)
WSQOP is the surface runoff rate removing 90% of pollutant (in inches)
IOQC is the interflow concentration (in count/acre/day)
These values were calculated in Fecal Tool, except for managed herbaceous, which was assumed to have an
accumulation rate that approaches urban land (it is usually associated with urban land in this watershed).
Final version submitted to EPA
42
APPENDIX IV. CHANNEL CROSS SECTION.
This cross section was used to develop the model F-table (table that describes relationship between
streamflow depth and streamflow volume).
Town Branch Cross Section at SR 1209
0
2
4
6
8
10
12
2 10 18 26 32 40 47 5663.5 70 76 84 92100 108 116
Horizontal distance (ft.)
Bank Full
APPENDIX V. MODELED FLOW and MODEL FECAL COLIFORM PREDICTIONS
(Calibration run, also shown in Figures 4 and 5)
This figure demonstrates that the higher fecal coliform concentrations in Town Branch occur following rainfall events.
During dry stretches the concentration is below the criterion.
10
100
1000
10000
6/19/97 1/5/98 7/24/98 2/9/99 8/28/99 3/15/00 10/1/00 4/19/01 11/5/01
DATE
FE
C
A
L

C
O
L
I
F
O
R
M

(
#
/
1
0
0

m
L
)
0.0
20.0
40.0
60.0
80.0
100.0
120.0
140.0
FL
O
W

(
c
f
s
)
MODELED FLOW MODEL OUTPUT