J Env Eng Paper2004

7/31/2019 J Env Eng Paper2004

1/14

419

Modeling coliforms in storm water plumes1

J. Alex McCorquodale, Ioannis Georgiou, Susanne Carnelos, andAndrew J. Englande

Abstract: The recreational waters near many large cities in the United States and Canada are severely impaired bypathogens that are present in the storm water runoff. In separated sewers the pathogen sources may be cross-flows betweenthe sanitary and storm water systems. This paper presents the methodology that was used in developing a forecastingmodel for pathogen indicators for recreational sites in the receiving waters of multiple storm water outfalls. The objectiveof the model is to give a timelier indicator of beach water quality than conventional beach monitoring, which takes about2 d for laboratory results. The model used for the study was based on the Princeton Ocean Model. The forecasting systemconsists of nested hydrodynamic models and a bacteria fatetransport submodel. Calibration and validation is based on6 years of field studies, laboratory analyses, and experiments. The methodology is illustrated by a case study of the impactof storm water flows on the south shore of Lake Pontchartrain, Louisiana, which has been banned for swimming since1985. The water quality data included: pathogen indicators (fecal coliform, Enterococci, and E. Coli), water chemistryparameters, turbidity, and nutrients.

Key words: modeling, water quality, pathogens, fecal coliform, stormwater runoff.

Rsum : Leau des cours deau destins aux loisirs prs des grandes villes des tats-Unis et du Canada est gravementdtriore par les pathognes se trouvant dans le ruissellement des eaux pluviales. Dans des conduites dgout spares,les sources de pathognes peuvent provenir dune contamination croise entre les systmes dgouts pluviaux et dgoutssanitaires. Cet article prsente la mthode utilise pour dvelopper un modle prdictif des indicateurs pathognes pourles sites rcratifs dans les eaux recevant plusieurs exutoires dgouts pluviaux. Lobjectif de ce modle est de fournir unindicateur plus rapide de la qualit des eaux la plage que celui offert par la surveillance conventionnelle des plages,laquelle ncessite environ deux jours pour avoir les rsultats du laboratoire. Le modle utilis pour cette tude est bassur le modle ocanique de Princeton. Le systme de prdiction consiste en des modles hydrodynamiques embots etdun sous-modle de transportdestin des bactries. Ltalonnage et la validation sont bass sur des tudes de terrain dunedure de six ans, sur des analyses en laboratoire et des expriences. La mthode est illustre par une tude de cas delimpact du ruissellement des eaux pluviales sur la rive sud du lac Pontchartrain, en Louisiane, dans lequel la baignadeest interdite depuis 1985. Les donnes sur la qualit de leau comprennent : les indicateurs pathognes (coliformes fcaux,

Enterococci et E. coli), les paramtres de chimie de leau, la turbidit et les nutriments.

Mots cls : modlisation, qualit de leau, pathognes, coliformes fcaux, ruissellement des eaux pluviales.

[Traduit par la Rdaction]

Introduction

During the past two decades engineers have been imple-menting abatement schemes to address the impacts of urbanstorm water pollutants on recreational waters, water supply,and aquatic biota. Storm and waste waters discharging from ur-ban areas are often polluted by pathogens that impose a risk ofdisease due to recreational use of the water. In spite of the sig-

nificant efforts to reduce bacteria counts in storm water, thereare often unacceptable levels of such indicators as E. Coli andfecal coliform (FC) in storm drainage. In separated systems,pathogens may enter the storm water via cross-flows betweenthe sanitary and storm water systems. The pathogens in thecross-flows represent a pollutant on the receiving waters. Pub-lic health officiers monitor beach areas by analyzing samplesfor pathogen indicators such as fecal coliforms and E. Coli;

Received 3 July 2003. Revision accepted 3 December 2003. Published on the NRC Research Press Web site at http://jees.nrc.ca/ on 6 October2004.

J.A. McCorquodale, I. Georgiou,2, 3 and S. Carnelos. Department of Civil and Environmental Engineering, 2000 Lakeshore Drive, 817 En-gineering Bldg., University of New Orleans, New Orleans, LA 70148, USA.A.J. Englande. Tulane University, Medical Center, School of Public Health and Tropical, Medicine, New Orleans, LA 70118, USA.

Written discussion of this article is welcomed and will be received by the Editor until 31 January 2005.

1This article is one of a selection of papers published in this Special Issue on Environmental Hydraulics.2 Corresponding author (e-mail: [email protected]).3 Current address: FMI Center for Environmental Modeling, 2000 Lakeshore Drive, 349 CERM Bldg., University of New Orleans, NewOrleans, LA 70148, USA.

J. Environ. Eng. Sci. 3: 419431 (2004) doi: 10.1139/S03-055 2004 NRC Canada


2/14

420 J. Environ. Eng. Sci. Vol. 3, 2004

Fig. 1. Field location map (inset) and field monitoring site map. Single headed arrow indicate storm outfall channel; double arrowindicates tidal boundary condition.

-90.12 -90.08 -90.04

30.00

30.04

30.08

NCDC Lakefront Airport Weather StationUNO Shoreline Monitoring Sites

LondonCanal

Bayou

St. JohnOrleans

Canal

InnerHarborNavigationCanal

Old Pontchartrain Beach LakefrontAirport

LP#5

LP#4

LP#3LP#2 LP#1

NCDC MSY International Airport Weather Station

17th Street

Canal

NORTH

LakePontchartrain

however, the results of these test take 2 d to complete. Duringthe waiting period, the loading, hydrodynamics, and pathogendecay may result in a completely different water quality indi-cator concentration. The objective of this study was to developa tool to assist in the near real-time assessment of the pathogen

indicators in the receiving waters of urban storm water outfalls.This is achieved by the development of a fate and transportmodel for these pathogens, in the receiving waters, during andfollowing a storm water runoff event. A case study involvingstorm discharges into Lake Pontchartrain, Louisiana, is used toillustrate the modeling procedure. The research was supportedby a grant from the US Environmental Protection Agency (Mc-Corquodale et al. 2003).

Lake Pontchartrain is located in the southeast portion ofLouisiana and forms the northern boundary of the City of NewOrleans (Fig. 1). The lake is approximately 1660 km2 in areawith an average depth of less than 4 m. The lake drains a wa-tershed of approximately 12 200 km2. Salt water from the Gulf

of Mexico enters Lake Pontchartrain through the MississippiSound and Lake Borgne at the east as well as through theInner Harbour Navigation Canal (IHNC) to the south, whichconnects the lake to the Gulf of Mexico through the Missis-sippi River Gulf Outlet (MRGO). Typical salinities in the lakesystem vary from 1 ppt in the west to 6 ppt in the east. To-gether, Lakes Maurepas, Borgne, and Pontchartrain form oneof the largest estuaries in the United States. Historically, LakePontchartrain has provided recreational opportunities for overone million residents in the metropolitan area of New Orleans.In the 1980s, the concentrations of FC bacteria regularly ex-

ceeded State water quality standards for primary contact recre-ation of 200 MPN/100 ml (MPN, most probable number)(Louisiana Department of Health and Hospitals (LDHH 1982).The advisory issued in 1985 to discourage swimming and otherprimary contact recreation in Lake Pontchartrain is still in ef-

fect for the south shore in New Orleans. One question that thisresearch addressed was whether swimming could be resumedwith an improved and timely advisory, i.e., when is a particularsite likely to be unsafe for swimming?

Urban storm water runoff from the New Orleans metropoli-tan area is pumped to the Lake since the average elevation inNew Orleans is below sea level and it is surrounded by floodprotection levees. An extensive drainage network consisting ofsubsurface and open surface canals exists to collect the runoff,which is then discharged to the lake at various points along theshoreline via a system of pump stations and drainage canals.The drainage system, operated and maintained by the Sewer-age and Water Board of New Orleans (SWB), has a pumping

capacity of over 120 106 m3/d. The system is designed tohandle 25 mm of rain in the first hour and 12 mm per hourthereafter (SWB 1990). The urban storm water at the point ofdischarge contains high counts of FC bacteria that are believedto be the result of sanitary sewer cross-flows. The storm waterdischarge points are at LP#3, LP#5, and the 17th Street Canalas show in Fig. 1. An intermittent source is located near LP#1.Figure 2 shows a typical storm water plume on the south shoreof Lake Pontchartrain.

The general approach was to collect field data on the stormwater flows and loads and the corresponding receiving water

2004 NRC Canada


3/14

McCorquodale et al. 421

Fig. 2. Aerial photograph showing typical outfall plume in Lake Pontchartrain. (Courtesy of Jefferson Parish).

responses. The widely used Princeton Ocean Model (POM)was adapted for this study. This is a three-dimensional modelthat includes density currents due to temperature and salinity(Blumberg and Mellor 1987). The forecasting model consistedof a hydrodynamic model for the whole lake that was driven bytidalandwindconditionsandahigh-resolutionnearshoremodelwith lake-side boundary conditions transferred from the whole-lake model. The resolution of this model was based on the sizeof theoutfallcanals. Themodel wascalibrated andverified withthe drifter trajectory and associated bacteria and conventional

water quality data. The modeling methodology is illustratedby application to the south shore of Lake Pontchartrain that isimpacted by storm runoff from Greater New Orleans.

Field studies methodology

There were three aspects of thefield studies: (1) theshorelinewater quality study, (2) the plume tracking study, and (3) thesedimentstudy. In addition,there werelaboratory investigationsassociated with the field work. A detailed description of thelaboratory and field studies is given by McCorquodale et al.(2003).

Monitoring of the water quality at five sampling locationsalong thesouthshoreof Lake Pontchartrain wascompleted overa 3-year period. Both wet and dry weather data were obtainedduring a period of drought-like conditions as well as duringa period characterized by more normal precipitation amounts.The stations (shown in Fig. 1) are as follows: LP #1 is a fishingboat launch and beach area near the outlet of the IHNC; LP#2is the old Pontchartrain Beach near the University of New Or-leans (UNO) Research and Technology Park; LP#3 is the Lon-don Canal outlet; LP#4 is the Old Beach recreational park nearBayou St. John; and LP#5 is the Orleans Canal outlet. Mon-

itoring was performed 5 times per 30-d period to analyze thestate of compliance with State regulations for safe recreationaluse of these waters. The parameters included: salinity, conduc-tivity, water temperature, FC, total nitrogen (N) as nitrite and(or) nitrate (NO2 and NO

3 ), ammonia (NH3), total Kjeldahl

nitrogen (TKN), ortho-phosphate (ortho-PO34 ), total phospho-rus (P), total suspended solids (TSS), volatile suspended solids(VSS), and pH. Daily rainfall data were obtained from the Na-tional Climatic DataCenter (NCDC) Lakefront Airport weatherstation.

The pathogen indicators at LP#3 in the London drainagecanaleffluentwerealsomonitoredduringsixstormwaterrunoffevents. Salinity, temperature, fecal coliforms, E. Coli, and En-trococci were determined at intervals of 10 to 30 min. The samelaboarory procedures were used for these smples as for theshoreline samples. The records of pumping stations operatedby the SWB were used to obtain the storm water hydrographsfor each event.

The pathogen indicator samples were analyzed on the dayof collection by the LDHH Water Laboratory. The nutrient andsuspended solids grab samples were analyzed at the UNO En-vironmental Engineering Laboratory within 2 h of collection.

All laboratory analyses, sample container preparation, samplecollection, and transportation were performed by the methodsoutlined foreach parameter in theStandard Methods for theEx-amination of Water and Wastewater (American Public HealthAssociation 1998). Salinity, conductivity, and water tempera-ture were measured in situ using aYSI 85 S-C-DO-T meter. Aspart of the UNO external QAQC program, the Lake Pontchar-train nutrient samples were compared and found to be in goodagreement with those reported for the same time period by theLouisiana Department of Environmental Quality (LA DEQ);the difference in the means were within the standard error in

2004 NRC Canada


4/14


the measurements. In addition, the UNO laboratory analyticalprocedures were verified with standard solutions.

Drifter studies were conducted to provide information onplume migration during actual storm water runoff pump events.The drifter study field methods included deployment of a buoy-ant GPS drifter (manufactured by METOCEAN Limited with

ARGOS satellitesupport)at themouth of a drainage canal uponnotification of a pump event by SWB pump station operators.The drifter was carried by the storm water runoff plume andreported, via the ARGOS Satellite, the path coordinates (lati-tude and longitude) at hourly intervals based on satellite avail-ability. Water quality monitoring in the area of the plumes wasperformed concurrently with thedrifter studies. The water sam-pling field methods included generating a 25-station grid with150-m sample intervals centered on the most recent drifter po-sition. The parameters measured at each location on the gridwere:E. Coli,Enterococci,FC,and Salmonella as thepathogenindicators, as well as salinity, conductivity, temperature, anddissolved oxygen. Sampling was repeated once a day for three

consecutive days following a storm event. These samples wereanalyzed to determinethe netdilution die-off along theplume.Background data were obtained through bi-monthly monitor-ing of all parameters during dry weather periods using a similar25-station grid positioned at the mouth of the canal.

Laboratory studies were performed to determine the kineticsof the bacterial indicators. Some of the factors influencing thedie-off rate of FC, Enterococci, and E. Coli are temperature,salinity, pH, turbidity, dissolved oxygen, and solar radiation.A laboratory evaluation was made under aerobic conditions,simulated sunlight, and fixed pH typical of Lake Pontchartrainwaters. Temperature, salinity, and light intensity were varied sothat water column die-off rates could be correlated to these pa-rameters. The decay rates and their relationships to these envi-ronmental factors as determined through these studies are usedin the decay sub-model incorporated into the three-dimensionalhydrodynamic fate and transport model.A sedimentation studywas conducted to determine the rate of reduction of bacteriafrom the water column due to the settling of suspended solidswith attached bacteria.

Field studies results

A detailed description of the laboratory and field results isgiven by McCorquodale et al. (2003). Data collected through

the field studies provided information on the urban stormwa-ter discharges to the study area. The shoreline FC data wereclassified as: wet weather or dry weather based on whetherthe samples were collected during or within 24 h of a rainfall>12 mm. Figure 3 shows the average water quality parametersfor all the shoreline data for wet and dry conditions. Table 1shows a statistically significant dependency (>95%) of FC onwet weather conditions found at 4 of the 5 stations. Similarresults were found for dissolved inorganic nitrogen (DIN) andTSS. The FC levels at the LP#1 site appear to be independent ofwet weather. This is a tidal navigation channel that appears to

Fig. 3. Comparison of water quality indicators under wet and dryweather conditions for shoreline study sites.

0

1

2

3

4

5

6

7

8

9

Beach near

INHC (LP#1)

Old

Pontchartain

Beach (LP#2)

London Canal

outlet (LP#3)

Old Beach at

Bayou St. John

(LP#4)

Orleans canal

outlet (LP#5)

Wet weather log FC (MPN/100 mL)

Dry weather log FC (MPN/100 mL)

Wet weather DIN (10-1 mg/L)

Dry weather DIN (10-1 mg/L)

Wet weather TSS (102 mg/L)

Dry weather TSS (102 mg/L)

Table 1. Students t test results for FC dependency onsignificant rainfall events and percent non-compliancewith Louisiana surface water quality regulations.

Site% Probability of wetweather dependency

% Non-compliance(LDEQ Regs.)

LP#1 66.36 47LP#2 98.43 53LP#3 99.99 59LP#4 99.46 28LP#5 99.99 72

have an intermittent source of FC that is not strongly dependenton storm water runoff.

The effluent at the outfall of the London Canal (LP#3) wassampled periodically during selected storm water pumpingevents. Figure 5 shows a typical pathogen and salinity responseat the outfall of the London Canal for a pumped discharge ofapproximately 75 m3/s.

Figure 4 shows the drifter trajectories for two runoff eventsat Orleans Canal. It is noted that north winds produced an east-ward movement while a more easterly initial wind resulted ina westward trajectory.

Table 2 summarizes the results of the kinetic studies that

were conducted on Enterococci. Similar results were obtainedfor E. Coli. (General note: pH for all samples was 7.7 (LakePontchartrain Average); standard temperature was 20 C; noradiation was used in the laboratory reference tests; all refer-ence tests were performed in the dark; the reference salinitywas 3 ppt.) The environmental kinetic coefficients that weredeveloped in the laboratory were kS and kP, which are the inac-tivation rateconstants incorporating salinityandsettlingeffects,respectively; kL is the inactivation rate constant incorporatinglight intensity. It was assumed that the decay effects were addi-tive. Enterococci was used as a surrogate for FC; this ensures a

2004 NRC Canada


5/14


Fig. 4. Examples of storm water plumes tracked using the buoyant GPS drifter.

-90.15

-90.15

-90.10

-90.10

30.02 30.02

30.04 30.04

30.06 30.06

30.08 30.08

OrleansCanal

BayouSt. John

LondonCanal

Plume tracking period :5:12 pm on 10/6/2000 to5:38 am on 10/7/2000.

17th StreetCanal

Plume tracking period :5:25 pm on 11/18/2000 to8:00 am on 11/20/2000.

OldBeach

NORTH

Open and closed circles aredrifter positions at hourly intervals

Wind Pattern over 11/18/2000Tracking Period :

Windspeed Range: 6 to 24 knots

Wind Pattern over 10/06/2000Tracking Period :

Windspeed Range: 14 to 19 knots

Fig. 5. Typical pathogen loading at London Canal outfall.(Sample date 26 July 2001. Mean outfall velocity was 0.2 m/s ata canal cross-sectional area of 150 m2.)

0

1

2

3

4

5

6

0 20 40 60 80 100

Time of sample in minutes

Measuredvariable

Salinity [ppt]

E Coli log(MPN/100 mL)

Enterococci log(MPN/100 mL)

FC log(MPN/100 mL)

conservative approach.The tremendous variabilityin the organ-isms that make up FC makes it difficult to duplicate laboratoryresults from sample to sample. The final inactivation kineticrates for FC in the forecasting model were calibrated to thefield observations.

Modeling approach

The POM is a free surface three-dimensional sigma coor-dinate primitive variable model. The model solves the conti-nuity and momentum equations and the transport equation for

Table 2. Laboratory analysis data for effects of environmentalfactors and sedimentation loss on overall inactivation rateconstant, k, for Enterococci.

Salinity (ppt) ks (h1)Light intensity(lux) kL (h1)

3 0.039 0 0.039

6 0.057 10 000 0.079 0.078 20 000 0.0912 0.09 50 000 0.116

Loss due to sedimentationFraction of bacteria attached to suspendedsolids, FP 0.09Sedimentation velocity (weighted avg. forlarge and small particles), VS 0.4 m h1

Constant for particle shape effect, 1Inactivation rate due to settling, kP 0.04/H h1

Note: ks, inactivation rate constant incorporating salinity; kp, inactivationrate constant incorporating settling effects; kL is the inactivation rateconstant incorporating light intensity; H, is the water depth (m).

temperature and salinity. The temperature and salinity trans-port is coupled to velocity through the equation of state rela-tionship. The model also incorporates a 2.5-level turbulenceclosure scheme to provide vertical mixing coefficients (Mellorand Yamada 1982) and horizontal mixing is accomplished bytheSmagorinsky formulation.Thegeneric form of thetransportequation is given as eq. [1].

[1] D

t+

UD

x+

V D

y+

W

=

KV

D

+F

2004 NRC Canada


6/14


where is the transportablevariable; U, V, Warethehorizontaland vertical velocity components; x, y, , t are the independentvariables representing horizontal and vertical space and time;D is the depth of the water column; KV is the vertical diffusioncoefficient; F is the horizontal viscositydiffusion term. Thehorizontal viscosity and diffusion terms are defined as

[2] F

x(H qx) +

y

H qy

where,

[3] qx AH

x, qy AH

y

where AH is the horizontal diffusivity. The Smagorinsky for-mulation was used to represent horizontal mixing and varieswith velocity gradients and grid spacing according to

[4] AM = Exy 12u

x

2+ v

x

+ u

y

2

2+

vy

212

where x and y is the grid spacing and Eis a dimensionlessparameter. Values ofEvary for each application and are gener-ally between 0.1 and 0.2 and can be zero if sufficient resolutionexists; a value of 0.05 was used for all the runs in this reportbecause of the fine grid spacing. Finally, the ratio ofAH/AM isused in the model to relate the diffusivity and momentum trans-fer. This is referred to as the inverse turbulent Prandtl numberand was set to 0.1 in the model.

The model formulation uses the finite control volume prin-ciple. The model has a two time step solution scheme. The hor-izontal (external) free surface mode solves the depth-averagesurface wave equation using a small time step. The internalmode solves the three-dimensional part using a much largertime step of the order of 30 times the external time step. It isa three time level model and time stepping is accomplished bythe leap frog scheme.

Model development

Two computational grids were used; the first grid includedthewholelake(Fig.6)andthesecondnear-fieldgridwaslocallyrefined in the area of study (Fig. 7). Since large scale circula-

tion could notbe properly simulated with a near-field model,thefar-field whole-lake grid was used to generate boundary condi-tions for the near-field domain. On the other hand, the far-fieldmodel could not resolve the outlet geometry of the drainagecanals. The far-field model had a horizontal resolution varyingfrom 150 m to 950 m, with high resolution near the study area,and 15 vertical layers. The near-field model had an alongshorehorizontal resolution of 50 m except near the eastern boundary,cross-shore horizontal resolution varying from 50 m to 500 m,and 10 vertical layers. Both models had logarithmic refinementat the surface and bottom.

Wind and tidal induced hydrodynamics and general circula-tion were predicted using the far-field model. Boundary condi-tions for use in the near-field model were saved every hour forparameters such as depth-averaged velocity and water surfaceelevation. Far-field model simulations were started from restand weretypicallyrun for 7 d toprovideenoughstartup time for

boundary effects to dissipate. The last 3 d of the solution werethen used to drive the near-field model. The near-field modelwas then run for 3 d resolving local hydrodynamics and trans-ported concentrations for salinity, temperature, and pathogens.The near-field model was modified to accept hourly forcing fordepth-averaged velocity and water surface elevation at the east-ern open boundary and at the IHNC, depth-averaged velocityat the western open boundary and discharge hydrographs andtime-dependent pathogen concentrations for all drainage canalswithastochasticpathogenloadingattheIHNC.Theseboundaryconditions were used after testing several alternatives includingthe transfer of elevation and velocity boundary conditions ontheeast, north,and west boundaries from the whole lake model;

the instability problems with this boundary condition appearsto be related to the dominance of the tidal boundary conditionat the IHNC. Pathogen distributions were simulated using themodel transport equation given as eq. [1]. To avoid over- andunder-shootingofthesolutionduetohighgradientsbetweenthepathogen source and the background concentrations, a positivedefinite upwind advection scheme was adopted (Smolarkewitz1984, 1986).

Following the transport, pathogen concentrations were thencorrected for decay. Field data suggested that the bacteria die-off follows first-order kinetics. Thus, a first-order kinetics equa-tion was applied as a correction to the concentration obtainedby the transport equation in the model. Laboratory data alsosuggested that the overall inactivation kinetic rate, k, is a func-tion of salinity, temperature, light, turbidity, and sedimentation.Quadratic relationships obtained by curve fitting of the fielddata were used to achieve the overall pathogen decay. The tem-perature correction was adapted from the literature (Thomannand Mueller 1987). Equations [5], [6], and [7] show the effectsof salinity, loss due to sedimentation, and temperature, respec-tively, on the inactivation constant. The combined effects wereincorporated into one coefficient and eq. [8] was used for thedecay calculation at each time step.

[5] kS = 0.00014S2 + 0.0024S+ 0.0253

[6] kP = FPVS

H= 0.04

H

[7] k = 1.07(T20) (kS + kP + kL)

[8] C = (Co Cb) (1 t k) + Cb

where kS and kP are the inactivation rateconstants incorporatingsalinity and settling effects (h1), respectively; kL is the inacti-vation rate constant incorporating light intensity (= 0.06 h1);S is salinity (ppt); H is the water column depth (m); FP is thefraction of pathogens attached to suspended sediment; VS is

2004 NRC Canada


7/14


Fig. 6. Fine-grid version of computational domain for the whole-lake model showing lake bathymetry in metres.

4.03

.0

5.0

2.0

3.0

2.0

3.0 3

.0

2.0

5.0

3.0

4.0

3.02.0

-90.60

-90.60

-90.40

-90.40

-90.20

-90.20

-90.00

-90.00

-89.80

-89.80

30.00 30.00

30.10 30.10

30.20 30.20

30.30 30.30

30.40 30.40

Fig. 7. Computational domain for the near-field model.

-90.12

-90.12

-90.09

-90.09

-90.06

-90.06

-90.03

-90.03

30.02 30.02

30.04 30.04

30.06 30.06

30.08 30.08

30.10 30.10

the combined particle settling velocity (m h1); T is tempera-

ture (Celsius); kis the overall inactivation rate constant (h1

);Co is the transported pathogen concentration prior to decay(MPN/100 mL); Cis the corrected concentration including de-cay (MPN/100 mL); Cb is the background pathogen concentra-tion for the lake (MPN/100 mL); t is the model time step (h).

Calibration and validation

The whole-lake model was calibrated for wind and tidaldriven circulation. The calibration parameter for the hydrody-namic model was the bottom roughness that was varied from

0.5 to3 cm with1.5 cmgiving the bestreproduction ofstageand

currents. The drifter data from 5 out of a total of 6 deploymentswere used to estimate velocities along the path of the plume.Table 3 compares these data with the model predictions for thesame deployments. The model currents were extracted for thetop0.9-mthatcorrespondedto thesaildepthof thedrifter. Themodel predicted the mean and root-mean-square (rms) currentsquite well;however, themaximum velocities weregenerallyun-derestimated.A fewdrifter velocities wereunreasonably highand may have been moved by pleasure boaters; these data werenot used in the calibration. The POM model stage predictionwas compared with mid-lake free surface stage data; Georgiou

2004 NRC Canada


8/14


Table 3. Comparison of currents for drifter data and modeledcurrents for five drifter deployments.

Event date

Meanvelocity(cm/s)

Minvelocity(cm/s)

Maxvelocity(cm/s

RMSvelocity

(cm/s)

7 Oct 2000 6 1.5 15.7 4.1Model 6.2 1.7 16.5 3.919 July 2001 7.9 1.4 27.5 3.1Model 7.4 1.7 14 326 July 2001 6.4 1.2 1.319 Sept 2001 5.3 0.9 17.8 1.4Model 5.9 1.5 13.5 1.37 Aug 2001 12.2 1.3 3.4Model 7.8 1.4 12.5 1.7

(2002) found that the rms difference between the predicted andobserved stage was approximately 4 cm for stage variations of

up to 45 cm.The model results were also compared with other Lake

Pontchartrain models (Georgiou 2002; Haralampides 2000;Signell and List 1997); all of the models predicted similar dou-ble gyre patterns in the depth averaged currents with nearshorecurrents in the range of 5 to 15 cm/s and interior velocities of2 to 6 cm/s. Figure 8 shows a typical double gyre circulationobtained from depth averaged currents with a north wind assimulated by POM. Figure 9 shows an almost identical gyrepattern as predicted by the RMA2 model developed by Har-alampides (2000). The double gyre pattern was observed forall wind directions. Figure 10 is an example of a south windinduced circulation predicted by POM.

Thenear-field model used thesame bed roughnessparameteras thewhole lake model.The near-field model wasdriven by thevelocity and elevation time series produced by the whole-lakemodel for the event of interest. These velocities and elevationswereinterpolatedintimeandspaceandusedasboundarycondi-tionsattheeastandwestboundariesofthenear-fieldgridshownin Fig. 7. The north boundary of the grid was modeled as anopen(radiation) boundary. This was necessary to avoid instabil-ity problems that were encountered when the north boundarywas represented by elevations and velocities that were trans-ferred from the whole-lake model. This instability may havebeen due to the over specification of the boundary conditionsin the presence of the locally strong tidal signal from the IHNCon the south boundary. The open boundary provided the neces-sarydegrees of freedom to satisfy continuity without unrealisticspikes in the water surface elevation. Internal disturbances inthe near-field model were allowed to radiate out of the domainfrom all water boundaries using a radiation condition for ele-vation.

Figure 11 shows shoreline contaminant dilutiondecay val-ues for dissolved inorganic nitrogen (DIN) and FC as deter-mined from the 3-year shoreline monitoring study. Also shownare minimum dilution data obtained from a physical model of

a thermal plume (McCorquodale et al. 2000). The apparent di-lution values, S, were estimated by

[9] S= {Co Cb} /

C Cb

where Co is the source concentration; C is the concentrationat the dimensionless distance, /w, along the shoreline ( isthe distance along the shoreline in the direction of the plumeand w is the width of the outfall canal); Cb is the backgroundconcentration.

A conservative tracer was simulated in the model and com-pared with the estimated dilutions based on observed DIN; itis assumed that DIN is nearly conservative over a period of24 h after discharge from the pumping station into the drainagecanal. Minimum dilutions determined from a thermal plume ina physical model are slightly lower than the DIN based valuesand the model predictions. The outlet aspect ratio in the physi-cal model was varied in the range of 6:1 to 12:1, while the fieldoutfalls were approximately 16:1.

In the case of FC, the C value includes the effect of decay;the standard error in the mean values estimated from the 3-yearmonitoring data is shown. The model was run with the decayand solids removal constants developed in the laboratory. Thecalibration was accomplished by varying the vertical diffusionKv, the inverse Prandtl Number (AH/AM), and the coefficientEin the transport equation (eqs. [1][4]) until the predicted andobserved average DIN dilutions were in agreement for a rangeof wind conditions. The model calibration captures the dilutionof the nearly conservative DIN very well. It should be notedthat the shoreline dilutions, as modeled and as estimated from

the field samples, were not necessarily the minimum dilution.The laboratory determined bacterial kinetics, a fate and trans-port model of the storm water plumes were calibrated for fieldobservation along the drifter path; in particular the light inac-tivation parameter of 0.06 h1 gave the best fit to the averageof the field observations. The shoreline FC data were used formodel verification. Figure 11 also shows the typical range ofmodeled FC obtained under various wind directions using theAH/AM and the coefficientEfrom the DIN calibration. The FCdilutiondecay values are well represented by the model; how-ever, the uncertainty is very high in the observed mean. At leastpart of the variability in the field estimates is explained by pos-sible sediment-derived FC introduced to the water column bywave action, the intermittent dry-weather source in the IHNC,differences in FC near the shore andoffshore, phase differencesin the modeled and actual tide at the sampling locations, erraticvariations in the source concentrations or flows, longitudinaldispersion, variation in solar radiation effects and experimentalerror in estimating the counts. The FC dilutiondie-off predic-tions of the model are subject to high uncertainties due to theinput and the field determinations of FC. The model resultsfor dilutiondecay are within the range of uncertainties in themodel results and the field data.

2004 NRC Canada


9/14


Fig. 8. Lake Pontchartrain circulation for a north wind by POM with a coarse grid at slack tide (Georgiou 2002).

-90.40

-90.40

-90.20

-90.20

-90.00

-90.00

-89.80

-89.80

30.00 30.00

30.20 30.20

30.40 30.40

25

1912

6

0

150 hours

Windat 5 ms

-1

(9.8 knots)

IHNC

Chef Menteur

Rigolets

VelocityMagnitude

(cm/s)

NORTH

Fig. 9. Lake Pontchartrain circulation for a north wind by RMA2 at slack tide (Haralampides 2000).

Model results

Figures 8 and10, respectively, show thegeneral lake currentsfor north and south winds. The predominant summer windshave a southerly component, while the winter winds tend tobe from the north. These patterns are shown at slack tide toemphasize the wind driven circulation. Figure 12 shows a near-field surface velocity pattern for a south-east wind as predictedby the near-field model.This figure illustrates a significant tidaleffect from the IHNC and FC plumes from the IHNC and theLondon Canal. During the 3-d period the plume responded to

the wind and tidal conditions. In some cases, the plume movedin one direction along the shore with a small tidal signal; inother cases, it oscillated in response to tidal effects and (or)shifts in the wind direction.

Figure 13 shows a model simulation for the FC plumes re-sulting from pumped discharges from the London, Orleans, and17th Street Canals for an easterly wind scenario. The resultsshow the FC plume from the end of pumping up to 3 d afterthe event. The simulated pump event is typical of those re-sulting from fairly significant storms in terms of duration and

2004 NRC Canada


10/14


Fig. 10. Lake Pontchartrain circulation for a south wind by POM with a coarse grid (Georgiou 2002).

IHNC

-90.40

-90.40

-90.20

-90.20

-90.00

-90.00

-89.80

-89.80

30.00 30.00

30.20 30.20

30.40 30.40

25

19

12

6

0

162 hours

Windat 5 ms

-1

(9.8 knots)

VelocityMagnitude

(cm/s)

NORTH

Fig. 11. Results of model calibration and validation (w is the canal width = 50 m and is the distance along the shoreline; ranges formodeled results represent the range of predicted values at a selected location; ranges for field data are means plus or minus the standarddeviations.)

flow rate. Figure 13 shows the results for 8 h of pumping at75 m3/s from each outfall (discharge velocity 0.2 m/s) witheast winds of 5 m/s. The east wind causes the plume to becomeshore-attached and move westward, which is common duringthe summer. A northwest wind also causes shore-attachmentbut moves the plume eastward, which is more common duringthe spring and fall. Figure 14 shows the plume under calm con-ditions. In this case the plume is affected mainly by tides andbuoyant spreadingso that thedilutions near theoutfall aresmallbut the beach areas are not strongly impacted because it takeslonger for the plume to spread to those areas. It is, however,possible that winds capable of forcing the contaminated plume

into the nearshore waters of the beach areas may occur duringthe 3 d following the runoff event.

Discussion

Since the runoff has a freshwater density and the lake istypically a brackish water body of greater density, the efflu-ent usually behaves as a surface-buoyant plume. This meansthat the movement and spreading of the plume water is initiallyconfined to the surface layers of the water column. The dif-ference in buoyancy between the effluent and receiving watersincreases the spreading action of the outfall plumes and mod-

2004 NRC Canada


11/14


Fig. 12. Simulated nearshore currents during a pumping event.

-90.08

-90.08

-90.04

-90.04

30.03 30.03

30.06 30.06

V el oc it y (cm/s ): 0 2 4 6 8 1 0

LondonCanal INHC

BayouSt. John

OrleansCanal

Universityof New OrleansReseach and Techology Park

Fig. 13. Fecal coliform plume simulation including decay for an 8-h pump event with an easterly wind of 5 m/s (FC sourceconcentration of 60 000 MPN/100 mL; standard for safe swimming is less than 200 MPN/100 mL.)

200

30

10000100030

20030

100010000

1000

0

-90.12 -90.09 -90.06 -90.03

30.03

30.06

30.09

constant pumping at 0.2 m/s (approx. 1100 cfs)

Diurnal tidecycle w/amplitude= 15 cm

17th StreetCanal

OrleansCanal

LondonCanalBayou

St. John

IHNCOld

PontchartrainBeachOld

Beach

Easterly wind

5 m/s

13 hours

FC source= 1445 MPN/100 mL

(a)

3020

0

1000

100020030

1000

30

200

200

-90.12 -90.09 -90.06 -90.03

30.03

30.06

30.09



17th StreetCanal

OrleansCanal

LondonCanalBayou

St. John

IHNCOld


Beach

Easterly wind

5 m/s

25 hours


(b)

30

30

200

30

30

200

30

-90.12 -90.09 -90.06 -90.03

30.03

30.06

30.09



17th StreetCanal

OrleansCanal

LondonCanalBayou

St. John

IHNCOld


Beach

Easterly wind

5 m/s

48 hours


(c)

30200

1000

-90.12 -90.09 -90.06 -90.03

30.03

30.06

30.09



17th StreetCanal

OrleansCanal

LondonCanalBayou

St. John

IHNCOld


Beach

Easterly wind

5 m/s

73 hours


(d)

2004 NRC Canada


12/14


Fig. 14. Fecal coliform plume simulation including decay for an 8-h pump event for calm conditions (FC source concentration of 60 000MPN/100 mL; standard for safe swimming is less than 200 MPN/100 mL.)

3020

0

1000

0

30 200

1000

302001000

30

200

1000

-90.12 -90.09 -90.06 -90.03

30.03

30.06

30.09

8 hours of pumping at 0.2 m/s (approx. 1100 cfs)


17th StreetCanal

OrleansCanal

LondonCanal

BayouSt. John

IHNC

OldPontchartrain

BeachOldBeach

No Wind

13 hours


(a)

30 200200

30

30

200

-90.12 -90.09 -90.06 -90.03

30.03

30.06

30.09



17th StreetCanal

OrleansCanal

LondonCanal

BayouSt. John

IHNC

OldPontchartrain

BeachOldBeach

No Wind

48 hours


(c)

30

200

10002001000

30200 1000

30

200

1000

30

-90.12 -90.09 -90.06 -90.03

30.03

30.06

30.09



17th StreetCanal

OrleansCanal

LondonCanal

BayouSt. John

IHNC

OldPontchartrain

BeachOldBeach

No Wind

25 hours


(b)

30200

100030

-90.12 -90.09 -90.06 -90.03

30.03

30.06

30.09



17th StreetCanal

OrleansCanal

LondonCanal

BayouSt. John

IHNC

OldPontchartrain

BeachOldBeach

No Wind

73 hours


(d)

ifies the mixing and dilution processes. Ideally, as the stormwater runoff is discharged from the drainage canals along theshoreline it should behave like a jet propelling the discharge outof the nearshore waters and diluting thecontaminated runoff byentraining the surrounding lake water thereby reducing the ex-tent of the impact. Unfortunately, many of the outfalls alongthe south shore of the lake have discharge velocities that areinsufficient to provide the momentum required to move the ef-fluent away from the nearshore region and to provide adequatemomentum mixing. The presence of ambient alongshore cur-rents often produces narrow outfall plumes that travel along theshoreline resulting in inhibited spreading and dilution. Figure 2isanaerialphotographofastormwaterrunoffdischargetoLakePontchartrain that is representative of the plume behaviour ob-served in the south shore of the lake. Evident from this figureis the significant impact that these discharges may have on thenearshore waters of adjacent recreational areas. The model andfield observation indicated that the denser lake flows caused atwo-layer flow to develop in the drainage canals; this providedsome in-canal mixing and increased salinity in the canals thatapparently has contributed to increased settling out of contam-

inated particles in the canals. The stored water in the canalsat the end of the runoff event provided an extended source ofbacteria that was available for mixing with the lake water. Themodel showed to be an extended source of bacteria and this wasconfirmed by the persistence of higher FC counts near the canalmouths even during so-called dry weather conditions.

The modelconsistentlyindicated that thecombined effectsofdilution and decay required 2 to 3 d for the ambient conditionsto return to safe levels (FC less than 200 MPN/100 mL) at thebeach areas after a storm water event. In addition to the decay

of organisms, the actual safety of the beaches depends stronglyon the pumped volumes and wind conditions. Other factorsthat affected the reduction of FC are salinity, temperature, andturbidity. A typical range for kinthisstudywas0.08to0.12h1

depending mainlyon thesalinity compared with a typical rangeof 0.1 to 0.2 h1 in the literature (Thomann and Mueller 1987).

Since the execution time for the model is of the order of afew hours (with current desktop computers), it could be usedas a tool to obtain daily updates for swimming advisories thatcan be posted on a public website.

2004 NRC Canada


13/14


Conclusions

Research on urban storm water runoff discharges to LakePontchartrainprovidedstatistical evidencethat urban storm wa-ter discharges are, indeed, a significant source of pathogens.The effect of storm water runoff discharges was found to last

for 2 to 3 d after a significant rainfall event (>12 mm) in theareas where FC levels have been linked to urban runoff dis-charges. Using laboratory determined bacterial kinetics, a fateand transport model of the storm water plumes was developedand confirmed from field observation along the drifter path.The mixing parameters in the model were calibrated to the fielddata for currents and a tracer (dissolved inorganic nitrogen).The model was verified using measured bacteria counts alongthe shoreline. The numerical model showed the dynamic natureof the effluent plume and also reproduced the 3-d wet weatherimpact period.

The concentration of microbes in the storm water effluentis highly variable, with observed values for FC in the range

of 6000 to more than 160 000 counts/100 mL. Variability ofthis order of magnitude makes it difficult to calibrate a modelusing field bacteria data. Furthermore, the uncertainty in theprediction of the model can be very large. For this reason, arisk-based prediction is recommended.

This study revealed that up to 30% of the bacteria can be at-tached to fine sediment. The removal of this fraction due to set-tling is significant; however, the laboratory study also showedthat these microbes had a very slow die-off rate in the sediment.The possibility of resuspension of these bacteria due to waves,currents,or swimming activityis a concern that is not addressedin the model.

The modeling methodology presentedhereprovides a frame-

work for a management tool for updating swimming advisorieson a timely basis. It is anticipated that regulatory pathogen in-dicator sampling would also be used in conjunction with thesimulation model.

Acknowledgments

This research was supported by a grant to the UNO SchliederEnvironmental Systems Center by the US EPA. Support wasalso provided by theSewerage andWater Board of New Orleansand the Lake Pontchartrain Basin Foundation.

ReferencesAmerican Public HealthAssociation,American Water Works Associa-

tion, Water Pollution Control Federation. 1998. Standard Methods

for the Examination of Water and Wastewater (20th ed.). AmericanPublic Health Association, New York.

Blumberg, A.F., and Mellor, G.L. 1987. A description of a three-dimensional coastal ocean circulation model. In Three dimensionalcoastaloceanmodels.Edited by N. S. Heaps.American GeophysicalUnion, Washington, D.C. pp. 116.

Georgiou, I. 2002. Three dimension hydrodynamic modeling of salt-water intrusion and circulation in Lake Pontchartrain. Ph.D. disser-tation, Department of Civil and Environmental Engineering, Uni-versity of New Orleans, New Orleans, La.

Haralampides, K. 2000. A study of the hydrodynamics and salinityregimes of the Lake Pontchartrain system. Ph.D. dissertation, De-partment of Civil and Environmental Engineering, University ofNew Orleans, New Orleans, La.

Louisiana Department of Health and Hospitals. 1982. Hu-man health protection through fish consumption and swim-ming advisories in Louisiana, February 18 [online]. Availablefrom http://www.deq.state.la.us/owr/mercury/fishadvi.html [cited14 January 2000].

McCorquodale, J.A., Hannoura, A.A., and Carnelos, S. 2000. Urbanstorm and waste water outfall modeling, Urban Waste ManagementResearchCenter, Final Project Report to the US Environmental Pro-tectionAgency, Collegeof Engineering, University of NewOrleans.New Orleans, La.

McCorquodale, J.A., Englande, A.J., Carnelos, S., Georgiou, I., andWang, Y. 2003. Fate of pathogen indicators in storm water runoff,Final Project Report to US Environmental Protection Agency, Ur-ban Waste Management Research Center, College of Engineering,University of New Orleans, New Orleans, La.

Mellor, G. L., and Yamada T. 1982. Development of a turbulence clo-sure model for geophysical fluid problems. Rev. Geophys. SpacePhys. 20: 851875.

Sewerage and Water Board of New Orleans (SWB). 1990. How itbegan, the problems it faces, the way it works, the job it does (pam-

phlet). Sewerage and Water Board of New Orleans, New Orleans,La.Signell, R.P., and List. J.H. 1997. Modeling waves and circulation in

Lake Pontchartrain. Gulf Coast Assoc. Geol. Soc., Trans. 47: 529532.

Smolarkiewicz, P.K. 1984. A fully multidimensional positive definiteadvection transport algorithmwithsmall implicitdiffusion. J. Com-put. Phys. 54: 325362.

Smolarkiewicz, P.K.,and Clark, T.L.1986. Themultidimensionalposi-tive definite advection transport algorithm: further development andapplications. J. Comput. Phys. 67: 396438.

Thomann, R.V., and Mueller, J.A. 1987. Principles of surface waterquality modeling and control, Harper & Row Publishers, NewYork.

2004 NRC Canada


14/14

J Env Eng Paper2004

Documents