Determination of Fecal Coliform Loading and its Impact on River Water Quality for TMDL Development

DETERMINATION OF FECAL COLIFORM LOADING AND ITS IMPACT ON RIVER WATER QUALITY FOR TMDL DEVELOPMENT

Brajesh Gautam, Murthy Kasi, and Wei Lin

Department of Civil Engineering North Dakota State University

Fargo, ND 58105

ABSTRACT The reach of Red River of the North flowing through the Cities of Fargo, ND and Moorhead, MN has been impaired due to the fecal coliform. Analyses of water quality samples and rainfall data showed that high coliform concentration in the river was related to rain events, and runoffs from the urban areas are major sources of fecal coliform contamination. To determine the fecal coliform total maximum daily load (TMDL) and to develop an implementation plan for water quality improvement, an extensive water quality sampling and model simulation program was carried out to calculate storm runoff flow rates, fecal coliform loads from the runoffs, and their impact on Red River water quality. The Storm Water Management Model (SWMM) was used to simulate the runoff flow rates and fecal coliform concentrations from different drainage areas under various rainfall conditions. Hydrographs and pollutographs generated by the model were used to determine the average daily fecal coliform load from each drainage area and the total load to the Red River. Steady state water quality model, QUAL2E, was employed for determining the impact of storm water runoff on Red River fecal coliform concentration. Event mean concentration (EMC) was used as a tool to link dynamic SWMM model to steady state QUAL2E model. The runoff sampling and modeling results were used to analyze the effect of rainfall properties on fecal coliform EMC. The study also identified sanitary sewer bypass as the potential source of high fecal coliform concentration in the River. KEYWORDS TMDL, storm runoff, SWMM, EMC, QUAL2E, and fecal coliform INTRODUCTION The Red River of the North, located in the north-central plains of the United States, plays an important rule in the regional development and is used for water supply, irrigation, industry, livestock, and recreation. The reach of the Red River main stem that flows through the Fargo, ND and Moorhead, MN areas was identified as impaired for its designated use due to high fecal coliform concentrations. Under the Clean Water Act, a pollutant-specific total maximum daily load (TMDL) needs to be developed. Led by Minnesota Pollution Control Agency (MPCA), a project team was assembled with participation of North Dakota Department of Health, Cities of Fargo and Moorhead, River Keepers, a voluntary group, and North Dakota State University

3851

WEFTEC®.06

Copyright 2006 Water Environment Foundation. All Rights Reserved©

(NDSU). Researchers from the NDSU Civil Engineering Department were responsible for determining fecal coliform loads from various sources and simulate its impact on the Red River water quality. The results of this study are presented in this paper. The Red River formed at Wahpeton, ND and Breckenridge, MN by the confluence of the Ottertail and Bois de Sioux rivers flows northward through the Red River Valley and drains into Lake Winnipeg in Canada. It flows across the flat, former bottom of the ancient glacial Lake Agassiz with average relief of only 2-3 feet per mile towards the river and 1-2 feet per mile northwards. The Red River forms the boundary between North Dakota and Minnesota as it meanders for 394 river miles (mi) to the Canadian border, a path that is nearly double the straight-line distance. The river flow rate varies greatly through out a year. Snow melting in the spring often causes flood conditions and rainfalls in early summer result in high river flows (Stoner et al., 1993). During the dry months of a year, the river flow rate can be very low, especially in the upper portion of the river, causing water shortage concerns. The area is renowned for its harsh and long winters with an annual average temperature of only 41.0◦F. A map of the Red River Basin is shown in Figure 1, with the impaired reach area highlighted. Figure 1. Red River Basin and the study area (Source: USGS, 1998)

Study Area

3852

WEFTEC®.06


The Cities of Fargo and Moorhead lie in the central part of the Red River basin and are the major urban areas in the region. Fargo-Moorhead is a growing metro area with a population of 174,367 (2000) with an increase of 13.7 percent from 1990 to 2000. The Red River main stem from the mid portion of Fargo-Moorhead area to the confluence of Sheyenne River is identified as impaired for exceeding fecal coliform standard, which specifies fecal coliform geometric mean less 200 CFU/100 mL and no more than 10% sample has greater than 2000 CFU/100 mL from April 1 till October 31. See highlighted reach in Figure 2. Figure 2. Impaired reach and the Study Area

3853

WEFTEC®.06


To identify the sources of fecal coliform contamination, an extensive water quality sampling was carried out along the Red Rive from its confluence with the Wild Rice River to the confluence of the Sheyenne River was carried out in the summer of 2002. From the analyses of water quality sampling data and rainfall records, it was concluded that fecal contamination in the Red River was a wet season problem and was related to storm runoff. After the sources were identified, next steps in TMDL development are to estimate source loadings and to assess the impact of the loads on river water quality. To address possible contributions of fecal coliform load from the entire urban areas and surrounding agricultural land uses, the study area was extended to include all the drainage areas upstream of the reach until the confluence of the Wild Rice River, as shown in Figure 2. The objectives of the project reported in this paper included:

• Determination of storm runoff flow rates and fecal coliform concentrations from representative drainage areas through field sampling.

• Calibration of the Storm Water Management Model (SWMM) • Estimation of average storm runoff flow rates and fecal coliform loads from all drainage

areas within the study area under various rainfall conditions; and • Assessment of the impact of storm runoff on Red River fecal coliform concentration

through model simulations. METHODOLOGY The methodology developed for this study consist different components of a hydrologic study. The two main components were application of time based flow and concentration results to calculate the amount of fecal coliform loading; and assessment of the impact of the runoff fecal coliform load on the River water quality. Storm Runoff Monitoring Both storm runoff hydrograph and water quality are highly dependent on land use. Degree of urbanization and storm water drainage system designs affects runoff duration, volume, peak flow rate and peaking time (Meyer, et al., 2005; and Walsh et al., 2005). Link between fecal coliform concentration and subcatchment land use has been reported based on field studies (Mallin et al., 2000; Petersen et al., 2005; and Young and Thackston, 1999). For these reasons landuse division is commonly used in runoff water quality studies (Choi and Ball, 2002; Moyer and Hyer, 2003; and Tong and Chen, 2002). In this study, storm water drainage areas were grouped based on their major land uses and drainage systems. Storm runoff monitoring was carried out at river outfalls or in manholes near outfalls at representative land use locations. Land use grouping and sampling point selection. Landuse map for the study area was obtained from the Fargo-Moorhead Metropolitan Council of Governments (Metro COG). Major land use in each storm water drainage area was identified by overlaying the drainage map on the land use map using GIS. As a result of this analysis, drainage areas in the study area were grouped into the five categories: residential, commercial, agricultural, parks and mixed, as shown in Figure 2. There are also two types of drainage system designs: storm water sewers discharging directly into the Red River, and storm water discharging to the Red River through

3854

WEFTEC®.06


drainage ditches. Direct discharging storm sewer systems are common in the areas next to the river and are usually smaller in size. The ditch systems are found in agricultural land use areas and used to collect storm sewer discharges from communities that are not adjacent to the river. Sampling locations were selected with the following considerations:

1. All major land use types like residential, commercial, mix and agricultural areas within Fargo and Moorhead should be covered.

2. Areas with various sizes were selected so that different hydrologic responses could be observed.

3. Outfalls from both direct sewer discharge and ditch collection systems should be sampled.

4. The outfalls should be accessible and safe for taking samples. Based on these selection criteria, five different drainage areas were selected for sampling. These drainage areas were sampled for runoff flowrate and concentration during the summer of 2004. The sampled drainage areas are shown in Figure 3 and the physical properties of the drainage areas are given in Table 1. Table 1. Storm Runoff quantity and quality monitoring locations

Landuse (%) Sampling

Site ID

Agricultural (AG)

Residential (RES)

Commercial (COM)

Open/Park (PA)

Water (WA)

MajorLand Use

Area (acres)

MHD L14 0 20 80 0 0 COM 80 FAR W11 0 40 28 30 2 MIX 815 MHD GF 0 50 28 22 0 RES 365

RC 67 3 11 18 1 AG 22,067 MHD D 100 0 0 0 0 AG 2,300

Storm water runoff measurements. Flow measurement was used to generate runoff hydrograph for the sampled rainfall event. Pressure transducers were used to measure the flow. These flow sensors measure the water depth in conduits. Then Manning’s equation was used to calculate the flow rate from the water depth and bottom slope of the conduit. When a conduit is a circular pipe, Manning’s roughness coefficient was corrected with depth of flow (Camp, 1946). For most of the sampling results using pressure transducers, one-minute interval flow hydrographs were obtained. In the circumstances when pressure transducers reading were not available, manual measurement of the depth was carried out and fifteen-minute hydrograph was calculated. Precipitation data for this study was collected using an onsite tipping bucket rain gauge. When the onsite rain gauge data were not available, data from Moorhead rain gauge and North Dakota Agricultural Weather Network (NDAWN) were used. The onsite rain gauge data was of one

3855

WEFTEC®.06


minute interval while the other source data had 10-minute and 15-minute interval. The Moorhead rain gauge was located in northern side of Moorhead near Highway 75 while the NDAWN site was located at Hector International Airport in North Fargo. The rain gauge locations are shown in Figure 3. Forty-eight hours of time difference was allowed between the runoff sampling events to account for the proper accumulation of the pollutant on the drainage area. Figure 3. Locations of point sources and sanitary sewer bypass discharge points from cities of Fargo and Morhead, and runoff sampling locations.

Fecal coliform sampling and testing. Time based concentration measurement was required to develop a pollutograph for a given runoff event. Manual grab sampling was the chosen method of sampling. Grabbed samples were taken at every fifteen minutes from the chosen manhole or outfall with the help of a bucket. The sampling was carried out during the whole duration of runoff for dry conduits or until the flow has receded to base flow level in the case of wet

3856

WEFTEC®.06


conduits. The bucket was rinsed between each sample. Fecal coliform sample were collected in 100mL pre-sterilized bottles. The samples were put on ice during the entire sampling event and storage period. Proper QA/QC measures were followed during the sampling which consisted of taking both field blank and duplicated samples. The procedures for the QA/QC from the Standard Operating Procedures for Water Quality Monitoring in the Red River Watershed created by Red Lake Watershed District (2003) were followed. The samples were analyzed in the Fargo Water Treatment Plant Lab, Moorhead Wastewater Treatment Plant Lab and NDSU Environmental Engineering Lab. The fecal coliform analysis was carried out using membrane filtration method (Standard Method, 9222D). A test was also carried out to check the consistency of the fecal coliform analysis among different labs that gave satisfactory result. With the result of concentration sampling, fifteen-minute interval fecal coliform pollutograph were obtained for the runoff events. Runoff Simulation and Model Calibration Storm Water Management Model (SWMM) is a dynamic rainfall-runoff simulation model used for single event or continuous simulation of runoff quantity and quality. SWMM has been a very popular model for the estimation of urban runoff hydrographs, and the model is also increasingly being used to generate pollutographs. Within the model, various buildup and washoff formulations can be used to calculate pollutant concentrations. With a consideration of natural die off in the environment, non-linear buildup functions with upper limits or saturation buildup models are commonly used for estimation of accumulation of fecal coliform during dry days (Moyer and Hyer, 2003; Zarriello et al., 2002). Saturation buildup model was selected for this study. Exponential washoff model is typically used for simulating fecal coliform concentration in runoff and was employed in this project. Using the hydrograph and pollutograph generated by SWMM, loadograph, which expresses coliform load as a function of time, was generated using spreadsheet software. The SWMM model given saturation buildup formulation is based on the buildup rate and the number of dry days however there is an upper ceiling on the maximum buildup that can occur.

tCtCB+

=2

1 (1)

where,

B = buildup rate (mass per unit area) C1 = maximum buildup (mass per unit area) C2 = half saturation constant (days to reach half of the maximum buildup) t = number of dry days preceding the rain event

The washoff occurs on the surface of catchment during a runoff period. The exponent formulation allows the washoff at every time step to be calculated based on runoff rate and remaining pollutant buildup amount.

3857

WEFTEC®.06


BqCW C21= (2)

where,

C1 = washoff coefficient C2 = washoff exponent q = runoff rate per unit area (inches/hour or mm/hour) B = pollutant buildup in mass (lbs or kg) per unit area or curb length

The SWMM was calibrated with field hydrograph and loadograph data. Because the main goal of this study was to determine the pollutant loads, it was decided that reliable loadograph estimation was more important. The model calibration was carried out separately for the runoff flow rate and fecal coliform load. The fecal coliform load was determined based on the runoff flow rate in the SWMM model hence it is important that the model is properly calibrated for runoff flow rate prior to the calibration for the load. The parameters adjusted during hydrograph calibrations include: percent imperviousness, drainage area characteristic width, and depression storage because these were identified as the most sensitive parameters (Jewell et al, 1978; Tsihrintzis and Hamid, 1998; and Warwick and Tadepalli, 1991). When multiple rain events data were available, all the data were used for model calibration simultaneously to provide more reliable calibration. A calibration objective function was developed using the concept of root mean square errors (RMSE). Matching the total runoff volumes and peak flow rates were determined as primary targets for the hydrograph calibration. Difference between measured and calculated runoff volume and peak flow rates were normalized by dividing measured volume and measured peak flow rates, respectively, to make them comparable in the objective function. To account for the impact of rainfall volume on the simulation error, rainfall volume of different events was also included. The objective function for hydrograph calibration (RMSEH) is given as:

n

PM

MS

Wn

PM

MS

WRMSEP

PP

p

V

VV

vH

∑∑⎥⎥⎦

⎤

⎢⎢⎣

⎡×⎟⎟

⎠

⎞⎜⎜⎝

⎛ −

+⎥⎥⎦

⎤

⎢⎢⎣

⎡×⎟⎟

⎠

⎞⎜⎜⎝

⎛ −

=

22

(3)

where, SV = Simulated total volume SP = Simulated peak flow rate MV = Measured total volume MP = Measured peak flow rate n = Number of runoff events Wv, Wp = Weight for runoff volume and peak flow rate, Wv=Wp=1 was used in this study P = Event total rainfall volume (inch) The model calibration for runoff flow rate was achieved by means of an iterative process of trial and error, by adjusting the parameters and minimizing the objective function and the shape of the hydrograph as well as runoff duration was compared graphically.

3858

WEFTEC®.06


The fecal coliform load calibration was carried out after the model was calibrated for the runoff flowrate. Similar to the runoff hydrograph calibration, both numerical and visual comparison were used in loadograph calibration. Total load was compared numerically through objective function RMSEL minimization while loadograph was compared graphically. The RMSEL function is given below.

( )n

MSRMSE LL

L∑ −

=2

(4)

where, SL = Simulated total event load

ML = Measured total event load n = Number of runoff events Loadograph calibrations were carried out by adjusting all four parameters in the selected buildup and washoff equations: maximum limiting buildup (C1), saturation buildup rate (C2), washoff coefficient (C1) and washoff exponent (C2). SWMM Model calibration for rural drainage areas The agricultural landuse in the surrounding areas of Fargo and Moorhead within the study area discharge the runoff into Red River through ditches. The flow routing through ditch differs from the urban storm sewer system. The higher percentage of impervious areas and storm drainage network in urban areas cause higher total runoff volume and larger peak flowrates than the rural catchments of equivalent areas and slopes. Hence, in the present study, a different approach for rural runoff flow estimation was adopted. The areas of rural catchments in the study area are greater than 1000 acres. For flow routing purposes, these large rural catchments were thus conceptually sub divided into smaller catchments and assumed to discharge through a common ditch. Similar to the model calibration for urban areas, the model was first calibrated for flowrate and then for fecal coliform load. The calibration parameters used for model calibration for flowrate were, percent impervious area and width. Model calibration for runoff quality was achieved through the RMSE minimization between predicted and observed fecal loads and the visual comparison of loadographs for predicted and observed. All four parameters of the saturation buildup and exponential washoff were used for the calibration of model calibration for quality. Simulation of Storm Runoff Impact on River Water Quality Estimation of the impact of urban runoff and other non point sources on the quality of receiving water body requires both flow volume and pollutant concentration in runoff. For the purpose of TMDL development load calculations are often carried out on a daily basis. Daily average concentrations for runoff loads are generally calculated using event mean concentration (EMC) approach (US EPA, 1983; Lee et al., 2002; Khan et al. 2006). EMC is the flow-weighted average

3859

WEFTEC®.06


concentration collected over the entire storm event, i.e. total load generated in a runoff event divided by total runoff volume. EMCs are also calculated for a time less than the full runoff duration, also known as partial event mean concentrations (PEMC). PEMCs are also useful in determining the initial or first flush impact (Lee et al., 2002; Kim et al. 2005). In the present study, PEMC approach is used and average concentrations and average flow rates were calculated in 24-hour storm duration to obtain the daily loads for TMDL development.

∑

∑

=

=

Δ×

Δ×== n

iii

n

iiii

tQ

tCQ

fVolumeTotalRunofTotalLoadEMC

0

0)(

(5)

24240∑

=

Δ×==

n

iii

avg

tQfVolumeTotalRunofQ (6)

where,

Qavg = Average flow rate, ft3/s Qi = Flow rate at each sampling interval, ft3/s Ci = Concentration at each sampling interval, CFU/100mL n = Number of samples ∆ti = Sampling interval, hours

Enhanced Stream Water Quality model (QUAL2E) is chosen to simulate the steady-state conditions in the river. QUAL2E is the most commonly used water-quality modeling program in the world (Chaudhury et al. 1998; Ning et al. 2001; Drolc and Končan 1996). It is one of the recommended models for TMDL studies by US EPA and has been widely used in TMDL studies. QUAL2E simulates fecal coliform in the river using one dimensional advection-dispersion mass transport equation, with a first order decay reaction (Brown and Barnwell, 1987). The river is modeled as series of completely mixed reactors, defined as computational elements, which are grouped into sub reaches with 20 or less computational elements in each sub reach. Sub reaches will have same geometric properties such as river bed slope, channel cross section, and Manning’s roughness, hydrological properties such as dispersion, and biological properties such as decay rate. The computational elements are used to input the sources such as wastewater treatment plant discharges, runoff outfall discharge points and sinks such as water treatment plant withdrawals. To define the geometric, hydrological, and biological properties of the Red River, 42.6 mile long study reach was divided into 15 sub reaches (unequal lengths) with a computational element length of 0.2miles. The data for geometric properties of sub reaches were obtained from a past study done by USGS (Wesolowski, 1994) and from Houston Engineering Inc., Fargo. The point sources from Fargo and Moorhead wastewater treatment plants, and American Crystal Sugar Plant, were input to QUAL2E as sources in the computational elements. The water treatment plant withdrawals from Fargo and Moorhead were the sinks in the computational elements.

3860

WEFTEC®.06


Average flowrates and concentrations for point sources and average intake rates for withdrawals are presented in Table 2. Table 2. Average values of flowrates and concentrations for point sources and withdrawals

Source/Sink

Average Flowrate (ft3/sec)

Average fecal coliform (CFU/100mL)

Fargo wastewater treatment plant 18.26 24

Moorhead wastewater treatment plant 6.82 5 American Crystal Sugar Plant 7.74 0

Fargo water treatment plant 17 Not applicable

Moorhead water treatment plant 6.2 Not applicable The QUAL2E model was first calibrated using measured river cross section areas, dry season (without the impact of runoff) river flow rates and fecal coliform concentrations. The calibration parameters for flowrate and concentration were identified from a sensitivity analysis (results are not presented here). The measured channel cross sections are adjusted to obtain equivalent areas of trapezoidal sections. Manning’s roughness constant and cross sectional areas were then adjusted during the model calibration process for flowrate. Though in reality, the channel geometry varies with river flowrates; it was assumed that the calibrated channel geometry did not change with varying river flow in the present study. The model calibration for fecal coliform was performed with dry weather data collected during 2000 and 2002. Fecal coliform die-off rate and headwater river base flow concentration were adjusted during model calibration for fecal coliform. Considering headwater concentration as one of the calibrating parameter is to specify it as a constant river base flow concentration (dry weather concentration without runoff impact). EMCs and average flowrates were calculated from the SWMM simulated dynamic runoff flowrates and concentrations (hydrographs and pollutographs). The EMCs and average flowrates were then used in the calibrated QUAL2E model as sources. There are 77 storm runoff outfalls and ditches from the Fargo and Moorhead discharging into the river. Hence, the sources closer than 0.2 miles (length of each computation element) were summed together using dilution method and used as one source. The QUAL2E was then simulated for the response of river fecal coliform concentrations to runoff discharges for rainfall events between 2000 and 2002. RESULTS AND DISCUSSION Storm Runoff and Fecal Coliform Load Simulation The SWMM model was calibrated with the field data obtained from the runoff sampling. Runoff flow rate and concentration was measured during the five rainfall events from May 2004 to

3861

WEFTEC®.06


August 2004 covering the five selected drainage areas. The rainfall characteristics of these 5 events are shown in Table 3. Table 3. Characteristics of rainfall data used for model calibration

Rainfall Event

Rainfall Volume (in)

Rainfall Duration (hr)

Peak Intensity (in/hr)

Antecedent Dry Period (days)

5/19/04 0.24 0.75 1.14 7 5/29/04 0.11 0.25 0.44 5 6/5/04 0.18 2.0 0.16 6 7/3/04 0.46 1.25 0.68 17 8/2/04 0.38 1.25 0.60 4 8/6/04 1.55 3 0.36 3

The SWMM was calibrated for hydrographs first by adjusting percent imperviousness and drainage area characteristic width until minimum values of RMSEs were obtained. Comparison of measured and simulated total runoff volumes and peak flow rates are shown in Table 5. For majority cases, close matches, less than ±15, of the field values and model simulation results were achieved, except for two events. The discrepancies between the model results and field data could be caused by several reasons including uneven distribution of intensity across the drainage area of interest. When multiple sets of field data were available, all the data sets were used simultaneously in the model calibration. Simulation results for two rain events, May 29 and August 3, for the drainage area MHD L14 are shown in Figures 4 and 5, respectively. Reasonably good match between the field data and calibrated hydrograph was shown in Figure 4 for the May 29 event. However, difficulties were experienced during the calibration for the July 3 rain event. Although percentage errors for the total volume and peak flow rate calibrations are relatively low, the field hydrograph has only one peak, but the model simulation showed two peaks in response to two separated higher rainfall intensity periods as shown in Figure 5. It is believed that this difference is caused by uneven distribution of rainfall during the event, and the rainfall data recorded by the rain gauge were not representative for the entire drainage area. Table 4. Hydrograph calibration results for directly discharging (through outfalls) drainage areas

Total Volume (ft3) Peak Flowrate (ft3/sec)

Drainage Area

Rainfall Event Measured Simulated

Error (%) Measured Simulated

Error (%)

MHD L14 5/19/2004 20412 20470 0.3 14.63 13.7 -6.8 5/29/2004 2831 2438 -16.2 1.43 1.46 2.1 6/5/2004 6575 9509 30.8 2.63 2.99 12.1 7/3/2004 47874 52639 9 25.72 28.05 8.3

FAR W11 7/3/2004 313933 281804 -11.4 66 65.3 -1.1

MHD GF 8/2/2004 37496 90465 58.6 8.9 26.8 66.8

3862

WEFTEC®.06


Figure 4. Calibration result for runoff flowrate of MHD L14 area (May 29, 2004)

0

0.5

1

1.5

2

2.5

3

17:30 18:00 18:30 19:00 19:30 20:00

Flow

(cfs

)

0

0.5

1

1.5

2

Rai

nfal

l Int

ensit

y (in

/hr)

Rainfall FieldSimulated

Figure 5. Calibration result for runoff flowrate of MHD L14 area (July 3, 2004)

0

10

20

30

40

18:00 18:30 19:00 19:30 20:00 20:30 21:00

Flow

(cfs

)

0

1

2

3

4

5

Rai

nfal

l Int

ensit

y (in

/hr)

Rainfall FieldSimulated

To estimate the hydrographs for all the directly discharging drainage areas, the SWMM parameters for each drainage area have to be determined. Three parameters were used during the calibration, depression storage, percent imperviousness, and characteristic width. Among the three calibrated parameters for hydrograph, depression storage is considered to be constant value for all the drainage areas. The %imperviousness is then determined for each landuse in the three calibrated drainage areas. This was done by writing the %imperviousness for each drainage area as an equation in terms of weighted landuse (Equation 5) and solving them as simultaneous linear equations. The final imperviousness for each landuse division in directly discharging drainage areas is given in Table 5. Using the percentages of landuse in each of the remaining directly discharging drainage areas, % imperviousness values were calculated.

3863

WEFTEC®.06


Bbababa nn =×++×+× K2211 (5) where, a1 = Percent of landuse 1 in the drainage area b1 = Imperviousness of the landuse 1 B = Calibrated imperviousness of the drainage area The characteristic width is calculated using the following equation.

AkW ×= (6) where,

W = subcatchment width k = constant based on the size of area A = subcatchment area

From the calibration results, drainage areas were divided into three sets, the drainage areas with less than 100 acres, the drainage areas between 100 acres to 500 acres and the drainage areas with more than 500 acres. The k values for these three groups were calculated as 4.82, 1.38 and 1.34 respectively. Table 5. Calibrated values of percent imperviousness for directly discharging drainage areas

Type of land use Calibrated % impervious area

Residential 30

Commercial 55

Park /Open 5

Agricultural 0

Water 0 After the hydrograph calibration was performed, the model was calibrated against the field loadographs. The field loadographs were obtained from runoff flowrates and concentrations using spreadsheet calculations. The model simulated hydrographs and pollutographs for four different rain events (Table 6) were used to calculate loadographs. Similar to hydrograph calibration, the multiple storm approach was applied for loadograph calibration too. Two rain events were used to minimize the RMSE values between measured and simulated total loads for MHD L14. The calibration results for these two rain events for MHD L14 are shown in Figures 4 and 5. During the RMSE minimization, the maximum limiting buildup and washoff exponent were found to be the controlling parameters. For the other two drainage areas (FAR W11 and MHD GF), the loadograph calibration was performed for one rain event each.

3864

WEFTEC®.06


Table 6. Fecal coliform load calibration results for directly discharging drainage areas

Total Load (CFU) Drainage Area Rainfall Event Measured Simulated Error (%)

MHD L14 5/29/2004 9.18×109 1.15×1010 25.3

7/3/2004 9.77×1011 1.01×1012 3.4

FAR W11 7/3/2004 1.56×1012 1.54×1012 -1.1

MHD GF 8/2/2004 2.70×1011 2.81×1011 4.1

Figure 6. Calibration result for runoff fecal coliform load of MHD L14 (May 29, 2004)

1.0E+05

1.0E+06

1.0E+07

1.0E+08

17:30 18:00 18:30 19:00 19:30 20:00

Feca

l Loa

d (C

FU/s)

FieldSimulated

Figure 7. Calibration result for runoff fecal coliform load of MHD L14 (July 3, 2004)

1.0E+06

1.0E+07

1.0E+08

1.0E+09

18:00 18:30 19:00 19:30 20:00 20:30 21:00

Feca

l Loa

d (C

fu/s)

FieldSimulated

3865

WEFTEC®.06


The quality parameters for the three calibrated drainage areas were used to determine the model parameters for the remaining directly discharging drainage areas. Since quality parameters are functions of landuse, a set of simultaneous linear equations with landuse percentage were built for each calibrated drainage area as shown in Equation 5. The equations were then solved for quality parameters for different types of landuse. The results are shown in Table 7. Using the percentages of landuse in each drainage area, weighted values of quality parameters were then calculated. Table 7. Calibrated runoff quality parameters

Maximum limiting Buildup Washoff Type of land use C1 (CFU/acre) C2 (day-1) C1 (in-1) C2

Commercial 5.5×1010 2 4.6 1.6 Mix 3.1×1010 4 4.6 1.5

Residential 2.9×1010 5 3.8 1.6 Agricultural/Open 2.1×1010 3 6.0 1.1

Calibration of SWMM model for rural drainage areas SWMM model was calibrated for the runoff data collected for rural drainage area RC for rainfall on August 6, 2004. Due to large size of the sampled drainage area (22,067 acres), the complete data for hydrograph and loadograph for this rainfall were not available. The calibration was performed by matching the simulation results with the available field data points through the trial and error process. The calibration results for hydrograph and loadograph are shown in Figures 8 and 9. The calibrated parameters are summarized in Table 7. Figure 8. Calibration result for runoff flowrate of RC area (August 6, 2004)

6567697173757779818385

13:00 13:30 14:00 14:30 15:00 15:30 16:00

Flow

(cfs

)

0.00

0.30

0.60

0.90

1.20

Rai

nfal

l (in

RainfallFieldSimulated

3866

WEFTEC®.06


Figure 9. Calibration result for runoff fecal coliform load of RC area (August 6, 2004)

1.00E+06

1.00E+07

1.00E+08

1.00E+09

13:00 13:30 14:00 14:30 15:00 15:30 16:00

Feca

l Loa

d (C

FU/s)

FieldSimulated

Table 7. SWMM calibrated parameters for rural drainage areas

EMC and its Dependence on Rainfall Intensity and Duration EMCs were used as 24 hour average concentrations in the present study to determine the total loads in the runoffs. They were also used to link the dynamic model with steady state model. EMCs were calculated from the loadograph and hydrograph which were generated using SWMM and spreadsheet calculations. To calculate EMC, total loads and total runoff volumes in a 24 hour window from loadographs and hydrographs were calculated, and EMC was then calculated as total load divided by total runoff volume. The 24 hour window was chosen to consider only daily loads for TMDL development. EMCs are affected by rainfall characteristics which include rainfall total volume, rainfall intensity, and rainfall duration (USEPA 1983). To better understand the effect of rainfall characteristics on pollutant washoff rate and its impact on river quality, an analysis on effects of rainfall intensities and durations on runoff EMCs was performed. The calibrated SWMM model was simulated for different uniform hypothetical rainfall events. The rain events were varied in intensity and duration. The intensities were varied from 0.1 in/hr

Drainage Area Hydrograph Calibration

Fecal Coliform loadograph calibration

Percent

Impervious Width

(ft) Depression Storage (in)

C1 (CFU/acre)

C2 (day-1)

C1 (in-1)

C2

RC-1 1.9 5267 0.1 1.7×1010 3 4.6 1.55

3867

WEFTEC®.06


to 3 in/hr, and the durations were varied from 0.5 hour to 12 hours. Runoff fecal coliform EMCs for a drainage area FAR W11 (a mixed landuse) were calculated from SWMM simulated hydrographs and loadographs. A summary of the drainage area characteristics are presented in Table 8. Table 8. Characteristics of subcatchment (FAR W11) used for EMC analysis

Subcatchment Property Value

Area, acres 815.4 Agricultural (%) 0 Commercial (%) 28.4

Open (%) 30.3 Residential (%) 39.5

Water (%) 1.7 The total runoff volumes and total fecal coliform loads were initially analyzed for different rainfall intensities and durations. As expected, the runoff volume increased for increasing intensities and with the increasing rainfall durations (Figure 10). The total runoff load has also increased with increase in rainfall intensity; however, the total load reached its maximum value quickly at shorter durations and higher intensities (Figure 11). At higher runoff durations and higher intensities the total load reached a constant value, because most of the fecal coliform built-up on the surfaces is washed off at the initial time periods. Figure 10. Impact of rainfall intensity on runoff volume

RainfallIntensity, in/hr

0.0E+00

2.5E+07

5.0E+07

7.5E+07

1.0E+08

0 2 4 6 8 10 12Duration, hr

Run

off

Vol

ume,

ft3

0.10.20.30.40.51.01.52.02.53

Figure 11. Impact of rainfall intensity and duration on runoff fecal coliform load

3868

WEFTEC®.06


Rainfall Intensity, in/hr

0.0E+00

2.0E+11

4.0E+11

6.0E+11

8.0E+11

0 2 4 6 8 10 12

Duration, hr

Feca

l col

iform

load

CFU

/day

0.10.20.30.40.51.01.522.53.0

Figure 12. Impact of rainfall intensity and duration on runoff fecal coliform EMC

Rainfall Intenstiy, in/hr

0

10000

20000

30000

40000

0 2 4 6 8 10 12

Duration, hr

Feca

l col

iform

EM

C, C

FU/1

00m

L

0.1 0.2 0.3 0.4 0.5

1 1.5 2 2.5 3

3869

WEFTEC®.06


Figure 13. Impact of rainfall volume on runoff fecal coliform EMC.

0

10000

20000

30000

40000

0 0.5 1 1.5 2 2.5 3 3.5

Rainfall total volume (inches)

Feca

l col

iform

EM

CC

FU/1

00m

L0.51.02.0

Duration, hr

The results were then analyzed for the EMC values versus rainfall intensity and duration. Higher EMCs were resulted for higher intensity rainfall with shorter durations (Figure 12). This is an important observation because; the higher EMC values in runoff will have adverse impact on river fecal coliform concentrations. The longer durations and higher intensities have resulted in decreased EMCs and hence will have reduced impact on river water quality. From the above analysis of results, it was identified that storms with higher intensities and shorter durations produced higher EMCs. The results were verified using the approach described in Critical Flow-Storm (CFS) concept by Zhang et al. (2001). According to CFS concept, a small storm combined with a low river flow can cause worst pollutant concentration in the river. In the present study, the instead of time varying concentrations, EMCs were analyzed with different rainfall volumes (Figure 13). Smaller rainfall volume with smallest duration (0.5 hr) had the maximum EMC in runoff. At the same rainfall volume, smaller durations (higher intensities) have caused higher EMCs in the runoff, which supports the earlier statement that higher intensities at lower durations cause higher EMCs. However, as the total volume increased the differences in EMC for different rainfall duration decreased. This is due to the total load at higher durations and higher intensities approaching a constant value (Figure 11). The results from this analysis have identified the characteristics of rain events that cause high EMCs in the storm runoff. The analysis results were then combined with river flow conditions to determine the critical conditions that could cause worst river fecal coliform concentrations, as part of the Red River fecal coliform TMDL development.

3870

WEFTEC®.06


Impact of Storm Runoff on River Water Quality The QUAL2E model was calibrated for flow with the data collected during 2004 and for fecal coliform with the data collected during 2002. From the model calibrations, fecal coliform die-off rate of 0.05 day-1 and river dry season concentration of 40 CFU/100mL were obtained. The calibrated die-off rate for Red River is in the lower range of 0.05 to 4.0 day-1 as specified by Brown and Barnwell (1987). The calibrated QUAL2E model was simulated for rainfall events between 2000 and 2002 and the simulation results were compared with field measured concentrations. The simulation results have closely matched with observed river fecal coliform data for all the rain events except for two major rain events. The models have underestimated the river fecal coliform concentrations significantly. This indicated that either the models were improperly calibrated or some major fecal coliform sources were not considered. A frequent communication with local stakeholders, and State and City governments was considered vital in each step of the TMDL study to avail their knowledge and understanding of the local settings. This type of communication has been found important in other bacterial TMDL studies too (Benham et al., 2005). After a discussion with representatives from state agencies and Cities of Fargo and Moorhead, it was found that both of the cities had experienced sanitary sewer overloading to their wastewater treatment plants during these two major rainfall events and allowed part of the raw sewage discharge directly into the Red River through sanitary sewer bypasses. The rainfall events were 6.9 inches of total rainfall volume in 9 hours on June 19, 2000 and 3.4 inches in 3 hours on June 9, 2002. Both the rainfall events have return periods with more than 100 years. Based on the information provided from the cities, two sanitary sewer lift stations from each city have bypassed part of the raw sewage during these rain events. The bypass discharge points are shown in Figure 3 (LS#1 and LS#2 from Fargo, and SD002 and SD003 from Moorhead). Average flows, EMCs, and total fecal coliform loads from these bypasses were calculated based on cities’ bypass records and were included in the QUAL2E simulations. The bypass loads were summed up with near by outfall loads which are closer than 0.2 miles (which is the length of each computation element) to input in QUAL2E. The QUAL2E model was simulated for with and without sanitary sewer bypass flows and results for one rain event are presented in Figure 12. The model simulation results gave a better match with the inclusion of bypass flows than without bypass flows. The results indicated that bypass flows had significant impact on river fecal coliform concentrations, which indicates that diluted raw sewage had caused a significant increase in river fecal coliform concentrations. As shown in Figure 12 at river mile 451.6, an increase of 1401 CFU/100mL was observed by the first bypass discharge (SD003). The total increase in river fecal coliform concentrations due to all the sanitary sewer bypasses was 4291 CFU/100mL. It is also can be seen from the Figure 12 that the maximum fecal coliform concentration in the river (12,125 CFU/100mL) occurred at river mile 449.4, which is immediately after the sanitary sewer bypass discharges. The initial large increase at river mile 470 was observed for both simulations (with and without sanitary sewer bypass flows); and this increase was observed because, the initial river concentration was very low (40 CFU/100mL) as compared to high concentration (20113 CFU/100mL) coming in the runoff at this river mile.

3871

WEFTEC®.06


Figure 14. QUAL2E simulation results of Red River fecal coliform concentrations on June 19, 2000 with consideration of sanitary sewer bypass from Cities of Fargo and Moorhead

0

2000

4000

6000

8000

10000

12000

14000

425.0435.0445.0455.0465.0475.0Red River Miles

(According to US Army Corps of Engineers)

Red

Riv

er fe

cal c

olifo

rm c

once

ntra

tion,

C

FU/1

00m

L

Simulated - withBypassSimulated - withoutBypassBypass discharge locationsMeasured

Wild Rice River Confluence

Sheyenne River Confluence

River flow direction

CONCLUSIONS To support the Red River fecal coliform TMDL development, a comprehensive project was carried out to estimate loadings from urban runoffs, to predict the impact of fecal coliform discharge on river quality, and to determine the load capacity of the river under various conditions. The project team worked closely with state agencies, city officials and other stakeholders to meet different expectations and requirements. The results of this study will benefit the decision makers in developing the goals and plans for water quality improvement. During urban runoff load estimation, grouping drainage areas based on landuse has helped in selecting representative sampling locations. Landuse also played an important role in modeling efforts to determine runoff quantity and quality parameters for different drainage areas. Model calibrations with RMSE minimization approach provided reasonable calibration parameters for both runoff flowrates and loads by reducing the errors between measured and predicted values. Using EMCs and average flowrates to determine runoff impact on river water quality reduced the data requirements and intensive modeling efforts. EMCs were also used in understanding the rainfall impacts on pollutant concentrations in the runoff. Based on the analysis performed in the present study, rainfall volume, intensity and duration, were found to have significant impact on runoff EMC. Small rainfall events have caused large EMCs generated in the runoff. This analysis results were used in identifying critical rainfall conditions that could cause adverse impact on river fecal coliform concentrations.

3872

WEFTEC®.06


Model simulations results for runoff impact on river water quality indicated that sanitary sewer bypasses had significant impact on river fecal coliform concentrations during large rain events. The necessary information about the bypass discharges provided by officials from state and city governments, and other local stakeholders had helped to quantify the bypass impact on river water quality. ACKNOWLEDGEMENTS The work was supported by Minnesota Pollution Control Agency. Authors appreciate North Dakota Department of Health, Red River Basin Commission, City of Fargo, North Dakota, City of Moorhead, Minnesota and River Keepers at Fargo that have provided their valuable suggestions during the project. The research was conducted at the Department of Civil Engineering, North Dakota State University. REFERENCES American Public Health Association (APHA), American Water Works Association (AWWA),

Water Environment Federation (WEF). (1998) Standard Methods for the Examination of Water and Wastewater; 20th Edition, American Public Health Association, American Water Works Association and Water Environment Federation, Washington DC.

Benham, B.L., Brannan, K.M., Yagow, G., Zeckoski, R.W., Dillaha, T.A., Mostaghimi, S., Wynn, J.W. (2005) Development of bacteria and benthic total maximum daily loads: A case study, Linville Creek, Virginia. Journal of Environmental Quality, 34 (5), 1860.

Brown, L.C., Barnwell T.O., Jr. (1987) Documentation and User Manual for the Enhanced Stream Water Quality Models QUAL2E and QUAL2E-UNCAS; USEPA Environmental Research Laboratory, Athens, GA, EPA/600/3-87/007.

Camp, T.R. (1946) Design of sewers to facilitate flow. Sewage Works Journal, 18, 3 Chaudhury, R.R., Sobrinho, J.A.H., Wright, R.M., Sreenivas, M. (1998) Dissolved oxygen

modeling of the Blackstone River (Northeastern United States). Water Research 32 (8), 2400

Choi, K., Ball, J.E. (2002) Parameter Estimation for Urban Runoff Modeling. Urban Water 4 (1), 31

Drolc, A., Končan, J.Z. (1996) Water quality modelling of the river Sava, Slovenia. Water Research 30 (11), 2587

Jewell T.K., Nunno T.J., and Adrian D.D., 1978. Methodology for Calibrating Stormwater Models. ASCE Journal of Environmental Engineering Division 104 (3), 485

Khan, S., Lau, S.-L., Kayhanian, M., Stenstrom, M.K. (2006) Oil and grease measurement in highway runoff - Sampling time and event mean concentrations. Journal of Environmental Engineering 132 (3), 415

Kim, L.-H., Kayhanian, M., Zoh, K.-D., Stenstrom, M.K. (2005) Modeling of highway stormwater runoff. Science of the Total Environment, 348 (1-3), 1

Lee, J.H., Bang, K.W., Ketchum, L.H., Choe, J.S., Yu, M.J. (2002) First flush analysis of urban storm runoff. Science of the Total Environment, 293 (1-3), 163

3873

WEFTEC®.06


Meyer, J.L., Paul, M.J., Taulbee, W.K. (2005) Stream ecosystem function in urbanizing landscapes. Journal of the North American Benthological Society, 24 (3), pp. 602-612

Mallin, M.A., Williams, K.E., Esham E.C., Lowe, P.R. (2000) Effect of human development on bacteriological water quality in coastal watersheds. Ecological Applications, 10 (4), 1047

Moyer, D.L., Hyer, K.E. (2003) Use of the HSPF and Bacterial Source Tracking for Development of the Fecal Coliform TMDL for Christians Creek, Augusta County, VA; Water Resources Investigations Report 03-4162, USGS

Ning, S.K., Chang, N.-B., Yang, L., Chen, H.W., Hsu, H.Y. (2001) Assessing pollution prevention program by QUAL2E simulation analysis for the Kao-Ping River Basin, Taiwan. Journal of Environmental Management, 61 (1), 61

Petersen, T.M., Rifai H.S., Suarez M.P., and Stein A.R. (2005) Bacterial loads from point and nonpoint sources in an urban watershed. Journal of Environmental Engineering, 131 (10), 1414

Red Lake Watershed District, 2003. Standard operating procedures for water quality monitoring in the Red River watershed. Revision 6.

Stoner, J.D., Lorenz, D.L., Wiche, G.J., and Goldstein, R.M. (1993) Red River of the North Basin, Minnesota, North Dakota, and South Dakota. American Water Resources Association Monograph Series No. 19 and Water Resources Bulletin, 29 (4), 575.

Tsihrintzis V.A and Hamid R. (1998) Runoff quality prediction from small urban catchments using SWMM, Hydrological Processes, 12 (2), 311.

U.S. Environmental Protection Agency (1983) Results of the Nationwide Urban Runoff Program, Volume I. Final Report; NTIS PB84-185552, U.S. EPA: Washington, DC.

U.S. Geological Survey (1998) Water Quality in the Red River of the North Basin, U.S. Geological Survey Circular, 1169.

Warwick J.J and Tadepalli P. (1991), Efficacy of SWMM Application, Journal of Water Resources Planning and Management, 117 (3), 352.

Wesolowski, E.A., 1994, Calibration, verification, and use of a water-quality model to simulate effects of discharging treated wastewater to the Red River of the North at Fargo, North Dakota. U.S. Geological Survey Water-Resources Investigations Report 94-4058, 143pp.

Young K.D., Thackston, E.L. (1999) Housing density and bacteriological loading in urban streams. Journal of Environmental Engineering 125 (12), 1177.

Zarriello, P.J., Breault, R.F., and Weiskel, P.K. (2002) Potential Effects of Structural Controls and Street Sweeping on Stormwater Loads to the Lower Charles River, MA; Water-Resources Investigations Report 02-4220, USGS.

Zhang H.X. (2001) The Critical Flow-Storm Approach for Nitrate TMDL Development in the Muddy Creek Watershed, Virginia. Conference proceedings of WEFTEC 2001.

3874

WEFTEC®.06


Determination of Fecal Coliform Loading and its Impact on River Water Quality for TMDL Development

Documents