Coastal Loading and Transport of Escherichia coli at an Embayed Beach in Lake Michigan

Coastal Loading and Transport ofEscherichia coli at an EmbayedBeach in Lake MichiganZ H O N G F U G E , * , † M E R E D I T H B . N E V E R S , †

D A V I D J . S C H W A B , ‡ A N DR I C H A R D L . W H I T M A N †

United States Geological Survey, Great Lakes Science Center,Lake Michigan Ecological Research Station, 1100 NorthMineral Springs Road, Porter, Indiana 46304, and NationalOceanic and Atmospheric Administration, Great LakesEnvironmental Research Laboratory, 4840 S. State Road, AnnArbor, Michigan 48108

Received March 11, 2010. Revised manuscript received July14, 2010. Accepted July 16, 2010.

A Chicago beach in southwest Lake Michigan was revisitedto determine the influence of nearshore hydrodynamic effectson the variability of Escherichia coli (E. coli) concentrationin both knee-deep and offshore waters. Explanatory variablesthat could be used for identifying potential bacteria loadingmechanisms, such as bed shear stress due to a combined wave-current boundary layer and wave runup on the beachsurface, were derived from an existing wave and currentdatabase. The derived hydrodynamic variables, along with theactual observed E. coli concentrations in the submergedand foreshore sands, were expected to reveal bacteria loadingthrough nearshore sediment resuspension and swash on thebeachsurface,respectively.Basedontheobservationthatonshorewaves tend to result in a more active hydrodynamic systemat this embayed beach, multiple linear regression analysis ofonshore-wavecasesfurther indicatedthesignificanceofsedimentresuspension and the interaction of swash with gull-droppings in explaining the variability of E. coli concentrationin the knee-deep water. For cases with longshore currents,numerical simulations using the Princeton Ocean Model revealedcurrent circulation patterns inside the embayment, whichcan effectively entrain bacteria from the swash zone into thecentral area of the embayed beach water and eventually releasethem out of the embayment. The embayed circulation patternsare consistent with the statistical results that identified that1) the submerged sediment was an additional net source of E.coli to the offshore water and 2) variability of E. coliconcentration in the knee-deep water contributed adverselyto that in the offshore water for longshore-current cases. Theembayed beach setting and the statistical and numericalmethods used in the present study have wide applicability foranalyzing recreational water quality at similar marine andfreshwater sites.

IntroductionHuman health is often threatened by contact with orswimming in contaminated water at recreational beaches.

Traditional microbiological techniques for detecting fecalindicator bacteria (FIB) levels in beach water, such as thosebased on bacteria culturing and incubation, usually take18-24 h, considerably longer than the time scale of variationsof FIB concentrations (1). Modeling FIB variability based onobservations of hydrometeorological and biological variablesusing statistical or process-based mathematical models is apromising alternative. Statistical models in the exploratorystage of recreational water quality modeling mostly have beenmultiple linear regression models constructed using easilyobserved hydrometeorological variables, including near-shore wave height, wind speed, turbidity, precipitation, water/air temperature, and dew point (2–4). Although these modelshave proven to be effective for managing particular beaches,the explanatory variables seldom provide direct evidence forbacteria importation, transport, and fate. Wind speed, forexample, is often used to identify wind storms that in turngenerate high waves on the ocean/lake surface. It has beenestablished, however, that the propagation direction of wind-generated waves is not necessarily the same as the winddirection (5). Onshore/offshore wind speed, as a useful meansfor partitioning data sets (3), may not be effective for allbeaches (6) as wind speed has only an indirect influence onbacteria transport and fate in the nearshore hydrodynamicsystem. Similarly, antecedent precipitation sometimes hasa loose connection with FIB counts at a recreational beachespecially when there is a complex watershed and ecologicalsystem or various nonpoint sources of contamination (7).

Previous studies attempting to identify the mechanismsthat may directly explain the variability of beach FIBconcentration have found that foreshore and submergedsands are important bacteria sources (8–10). These studiesconfirmed that beach water can actively receive FIB fromforeshore sands through swash, suspension, or tidal move-ments (10). Bacteria can also be deposited from water intothe foreshore sand, which makes the water-sand interactionbidirectional (9). The significant bacteria sources in foreshoreand submerged sands underscore the importance of hy-drodynamic mechanisms in the swash and the surf zones.Besides strong point sources and solar radiation (11),nearshore hydrodynamics is possibly another rapid mech-anism responsible for bacteria importation, transport, anddeposition at a beach. Investigating hydrodynamic phe-nomena such as sediment suspension and swash, therefore,will offer more direct evidence than meteorological param-eters of the mechanism involved in water-sand interactionand bacteria transport in the beach water.

The Chicago 63rd Street Beach is a popular recreationalresource in southern Lake Michigan. It suffers from highclosures each swimming season (9), partly because theChicago area is one of the regions under the heaviest influenceof sediment resuspension throughout Lake Michigan, with,for example, a bed shear stress over 0.1 Pa for 20% of thetime during 1994-1995 (12). Extensive studies have beenconducted at this beach to understand the causes of FIBcontamination. Prediction-oriented regression models de-termined that the hydrometeorological variables includingwind speed/direction, solar insolation, and rainfall couldexplain a portion of the variation in E. coli concentration(13). Additional research identified the importance of water-sand interaction at this beach and its associated ecologicaland microbiological effects on water quality (9), whilegroundwater discharging to the lake was eliminated as asignificant E. coli source for the beach water (7). In the presentstudy, we focused on bacteria loading and transport in directresponse to nearshore hydrodynamic events. The embayment

* Corresponding author phone: (219)926-8336 ext. 430; fax:(219)929-5792; e-mail: [email protected].

† United States Geological Survey.‡ National Oceanic and Atmospheric Administration.

Environ. Sci. Technol. 2010, 44, 6731–6737

10.1021/es100797r 2010 American Chemical Society VOL. 44, NO. 17, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 6731

Published on Web 08/05/2010

setting, with beach water bounded by breakwaters or jetties,is typical in coastal areas. Methods and findings in the presentwork have widespread significance for other sites with similarinfrastructure and environmental conditions.

Materials and MethodsField Observations. Detailed description of the Chicago 63rdStreet Beach has been given elsewhere (9, 13). Briefly, thestudy area is located on the southwest shore of Lake Michigan.The beach area is in an embayment that is open to thenortheast and bounded by two breakwaters in the north andsouth (Figure 1). Five transects were established 100 m apartfrom one another. Water samples were taken at approximately07:00 h on three consecutive days per week, usually fromTuesday through Thursday, in 45 cm deep (knee-deep) waterat each transect from April to September 2000. At the samelocations, sediment samples were collected from submergedsand. Sand samples were simultaneously collected from theforeshore sand about 1 m landward from the farthest extentof wave actions at each transect. Additional water sampleswere obtained from the offshore extremity of the southernbreakwater (water depth approximately 4 m). After samplecollection, water and sediment samples were kept at 4 °Cand later analyzed for E. coli concentration in the laboratorywithin 3 h. Water samples were analyzed by membranefiltration onto mTEC agar as outlined in EPA/600/4-85 076(14). Besides the same analytical methods, sediment samplesrequired additional preparation including the estimation ofthe total sample volume with the core liner, the dilution ofthe test sediment, and the shaking of the sample bottles for5 min at 210 rpm on an Eberbach platform shaker. Resultsare reported as colony forming units per 100 mL of water(CFU/100 mL). E. coli concentration averaged over the fivetransects for the foreshore sand, submerged sediment, and45-cm water, and of offshore water were further log10-transformed and are denoted as ECforesh, ECsed, EC, andECoffsh, respectively.

Onsite weather and water monitoring stations wereestablished on the southern breakwater and near transect#1, respectively. Meteorological parameters observed fromthe onsite weather station included wind speed, direction,gust (denoted as GT), and air temperature (Ta). The numberof gulls on the beach (GL) was observed over the entire beacharea. The density of gull-droppings (GD) was counted in threerandomly placed 1-m2 quadrats 1-6 m from the furthest

extent of waves at each transect followed by averaging. Waveand current conditions were obtained from a nearby offshorelocation (the origin of the coordinate system in Figure 1)based on an existing numerical simulated database and willbe described in next section.

The definitions of onshore, offshore, and longshoredirections are particularly important for isolating cases fromthe entire data set. According to Figure 1, exact onshoredirection is toward the southwest, and hence onshore windand current velocity components are defined as the projec-tions of wind and current velocity vectors in this direction,denoted as WDon and CTon, respectively. The northwest-southeast line is the exact longshore direction in which theprojection of current velocity is denoted as CTlong. Fur-thermore, any vector (wind, wave, or current) that is directedfrom 135° to 315° (a 180° range) clockwise from the due northis defined as “onshore”. A 90° subrange of these angles, from180° to 270° clockwise from the due north, is defined as“strictly onshore” directions. Similarly, vectors directed nomore than 22.5° from the exact longshore direction wereconsidered to be “longshore”. Categorical variables indicatingwhether wave propagated onshore (1 for onshore and 0otherwise) and whether current velocity was longshore (1for longshore and 0 otherwise) were also used in the followinganalyses and denoted as IfWVon and IfCTlong, respectively.

New Hydrodynamic Variables. Bed shear stress (BSS) andtotal runup height (R) were derived from numericallysimulated wave and current data to reflect the potential ofsediment resuspension and bacteria input from the foreshoresand, respectively.

Bed Shear Stress and Sediment Resuspension. Theoccurrence of sediment resuspension is determined by bedshear stress as well as other parameters such as the bed formand sediment mobility (15). Bed shear stress can be estimatedfrom the fluid flow that overlies the sea/lake bed, which oftenis a combined wave-current boundary layer (WCBL) flow(16). The combined WCBL is not a simple addition of waveand current flow fields but a nonlinear superposition of them.A widely used method for estimating bed shear stress causedby a WCBL was outlined in ref 16. Detailed assumptions andequations are given in the Supporting Information.

When bed shear stress exceeds a critical value, empirically0.05-0.1 Pa, sediment suspension is considered to occur(12). While all suspended FIB are not necessarily sediment-attached (17), elevation of bacteria concentration is usually

FIGURE 1. 63rd Street Beach of Chicago. Sampling locations for E. coli concentration in the foreshore sand (open triangle), in theknee-deep water (open circle), in the submerged sediment (open circle), and in the offshore water (filled circle) are marked. Opensquare: weather station; dark square: water monitoring point; crossed circle: the location where numerically simulated wave andcurrent data are available; directions within 90° and 45° from the exact onshore direction (SW) are defined as onshore and strictlyonshore, respectively; exact longshore direction is the NW-SE direction indicated by the wide dashed-line. Box-plot of thedistribution of E. coli concentrations (log10 CFU/100 mL) measured in the foreshore sand (ECforesh), in the submerged sand (ECsed),in the knee-deep water (EC), and in the offshore water (ECoffsh) are inserted on the left.

6732 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 17, 2010

observed immediately close to the location of sedimentdisturbance, as seen, for example, in dredging (18). Therefore,it was assumed that, provided bacteria are initially presentin the sediment, sediment resuspension closely coincideswith bacteria importation into the water column.

In the absence of actually observed current and wavedata for the Chicago 63rd Street Beach in 2000, numericallysimulated current and wave parameters were obtained fromthe Great Lakes Environmental Research Laboratory of theNational Oceanic and Atmospheric Administration (NOAA).Their numerical simulations of current circulation andsurface wave parameters in Lake Michigan were based on athree-dimensional Princeton Ocean Model and wave modelsadapted for the Great Lakes, which have been extensivelyvalidated and applied to many advanced studies (19–21).For the present case, the location for both current and waveparameters is close to the midpoint of the opening of theembayment (Figure 1) with a nominal water depth of 5 m.Current and wave data at 06:00 h were used in order to leadwater sampling time, 07:00 h, by an hour, allowing foradaptation of beach water to hydrodynamic conditions. Thecurrent velocity used for estimating the bed shear stress inthe combined WCBL was obtained at the lowest verticalcomputational cell in the numerical model, namely 0.12 mabove the lake bed.

Potential Bacteria Input in the Swash Zone. An additionalvariable that is potentially important for explaining thebacteria loading at the study beach is wave runup on thebeach surface. The total wave runup R is defined as the sumof the maximum wave setup and the standard deviation ofthe fluctuating swash height (22). The total runup on a beachis empirically proportional to the surf similarity parameter

where � denotes the beach slope, and H0 and L0 are the deep-water wave height (or significant wave height Hs for randomwaves) and wavelength, respectively. The empirical propor-tionality leads to a convenient estimate of total runup on abeach

where Tp is the peak wave period, and C is a coefficient ofproportionality that is at least dependent on the beach slope� and can be calibrated for specific beaches. For many studies,deep-water wave parameters such as Hs and Tp are availablefrom publicly accessible sources.

Physically, shear stress on the beach surface during waverunup-backwash cycles is hypothesized to be a direct agentfor bacteria input from the foreshore sands to the swashzone. Due to the lack of reliable models for directly estimatingswash shear stress (23) and the complexity of the wave-sandinteractions (24), we simply assumed that the total runupreflects the potential of bacteria loading from the foreshoresand.

In estimating the total runup for the present work, eq 2was used, but the coefficient C was omitted without affectingany analysis in the following sections. Beach width (Bw), thedistance from a fixed inland point to the upper tip of swashlenses, was actually observed during the field study and is(inversely) comparable to the total runup.

A summary of explanatory variables used in the followingstatistical analyses are listed in Table 1. In the Results section,parsimonious models under particular hydrodynamic con-ditions are identified by a backward elimination processbased on the probability of F-to-remove larger than 100.Backward elimination started with 10-14 variables (i.e., thefull model) introduced in the present section. The number

of variables included in the full models was limited by theissue of no multicollinearity. No serious multicollinearitywas found between E. coli concentrations at differentlocations. It is important to note that linear regression modelsused here are not for prediction purposes but for identifyingrelationships among variables. In practical predictive models,for example, E. coli concentrations in the submerged andforeshore sands are usually unavailable, and more waterchemistry variables should be considered.

ResultsGeneral Results. Distributions of the four E. coli concentra-tion variables, ECforesh, ECsed, EC, and ECoffsh, in the entirestudy period (N) 75) are shown in the insert of Figure 1. Theforeshore sand had the highest E. coli concentration with amedian slightly over 104 CFU/100 mL. The median of E. coliconcentration in the submerged sand was approximately103 CFU/100 mL, while those in the nearshore and offshorewaters were about 102 and 10 CFU/100 mL, respectively.There thus was an approximately exponential decay of E.coli concentration from the sand farthest onshore to the wateroffshore (9).

The simulated wave height and current speed as well astheir directions are shown in Figure 2, with onshore-wavecases highlighted by symbols. The 75-d study period can benearly evenly divided into two categories: onshore wave cases(38 d) and offshore-wave cases (37 d). Moreover, a majorityof onshore-wave cases, 25 out of 38 d (the filled circles thatfall in the horizontal belt bounded by solid lines in Figure2c), are strictly onshore. Longshore currents were prevailingnearshore outside the embayment. According to the defini-tion given in Figure 1, there were 64 d (out of 75 d) withlongshore currents.

Figure 3a shows the estimated bed shear stress due to acombined WCBL in the nearshore water. On the days whenthe wave action was negligible (wave height <0.18 m) a purecurrent boundary layer flow was assumed instead of acombined WCBL (Supporting Information). If 0.08 Pa ischosen as the critical bed shear stress for sediment suspension(12), 11 occurrences of sediment suspension are identifiedin Figure 3a. Since the numerically simulated current and

�0 ) �

√H0/L0

(1)

R ) C√HsTp (2)

TABLE 1. Explanatory Variables Used in Statistical Analysis

variable description

ECforesh E. coli concentration in the foreshore sand(log10 CFU/100 mL)

EC E. coli concentration in the knee-deep beachwater (log10 CFU/100 mL)

ECsed E. coli concentration in the submerged sediment(log10 CFU/100 mL)

ECoffsh E. coli concentration in the offshore deep water(log10 CFU/100 mL)

BSS bed shear stress (Pa)Hs significant wave height (m)R estimated total runup (arbitrary unit)Bw beach width (m)CTon onshore component of current velocity (ms-1)CTlong longshore component of current velocity (ms-1)WDon onshore component of wind speed (ms-1)GT gust (ms-1)Ta air temperature (°C)GL number of gulls

GD gull dropping density on the beach(number m-2)

BwGD product of Bw and GDIfWVon 1 if wave direction is onshore; 0 if otherwiseIfCTlong 1 if current velocity is longshore; 0 if othereise

IfDnCoast 1 if current velocity is longshore and down-coast; 0 if longshore but up-coast

VOL. 44, NO. 17, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 6733

wave data as well as estimated beach slope and grain sizemight have been affected by diverse sources of error, nocritical shear stress value was adopted. The original shearstress values, rather than those above a threshold, were usedfor further statistical analysis.

Figure 3b shows the estimated total runup based onsimulated offshore wave parameters (Figure 2). Actuallyobserved beach width, Bw, was inversed and shifted (i.e.,40-20Bw) to have the same trend as the total runup (notethat a larger runup is corresponding to a smaller beach width)and clarity in the figure.

Effect of Onshore Waves. Nine out of fourteen environ-mental variables had significant differences in the mean valuefor onshore and offshore waves and are highlighted in Table1 of the Supporting Information. Results based on inde-pendent-samples t tests revealed that the variables ECoffsh,BSS, R, Hs, and GT were significantly higher (P < 0.05) foronshore waves than for offshore waves, while EC, Ta, WDon,and GL were significantly lower (P < 0.05). It appears thatonshore waves enhanced hydrodynamic activities near the

beach, including increased significant wave height (Hs), bedshear stress (BSS), and total runup (R). Onshore waves alsocoincided with increased gusts (GT) and decreased onshorewind speed (WDon), consistent with established theories thatturbulence in winds is more important than the mean windvelocity in generating surface waves (5). It might not beintuitive that onshore waves were consistent with decreasedEC (significantly). A possible reason is that most of theonshore-wave cases (34 out of 38 d) also had longshorecurrents, which considerably facilitated bacteria exchangebetween the inside and outside of the embayment. Signifi-cantly more E. coli imported into the beach water does notnecessarily mean more retained. Further analysis of thephenomenon can be found in the Current Pattern in theEmbayment and the Discussion sections.

We further isolated a subsample with onshore waves (N) 38) and developed a multiple linear regression model,which is shown as model A in Table 2. A model selectionprocess through backward elimination identified a bestparsimonious model with four explanatory variables: ECsed,ECforesh, BSS, and BwGD (product of Bw and GD). Theinterpretation of the model is straightforward: the pair ofvariables ECsed and BSS reflect the supply and mechanism(sediment resuspension) of bacteria loading from the sub-merged sand, and the other variables indicate the strengthsof two different types of bacteria sources, the foreshore sandand gull-droppings, and their common mechanism, swash.The partial correlation coefficient in a multiple linearregression analysis can provide additional information ofthe net contribution (and, in the present context, the netdirection of bacteria loading and transport) of individualexplanatory variables to the variability of the dependentvariable (25). In this four-variable parsimonious model, thepartial correlation coefficients of all variables are positive.This implies positive net contributions of all variables to E.coli concentration in the knee-deep water for onshore-wavecases.

Offshore E. coli Concentration. Using the entire data set(N ) 75), we identified an optimal model (model B in Table2) for estimating offshore E. coli concentration. The variabilityof offshore E. coli concentration was significantly dependenton the variability of ECforesh, R, and IfCTlong. While it wasnot unexpected that ECforesh and R signify a bacteria loadingmechanism from foreshore sands, the significance of thevariable IfCTlong, indicating whether the currents movedlongshore, was not as clear. A further model (model C inTable 2) that includes only cases with longshore currents (N) 64) reveals the significance of EC and ECsed in additionto ECforesh and R. Considering models B and C together, itappears that current direction facilitates a meaningfulpartitioning of the data set. On the days with longshorecurrents, bacteria loading from the foreshore sand (ECforeshand R), from the sediment (ECsed), and a bacteria receiver(i.e., embayed beach water represented by EC) contributedto the variability of E. coli concentration right outside theembayment. More specifically, bacteria imported from theforeshore sand through swash (a net source with positivepartial correlations between ECoffsh and ECforesh and R inboth models B and C) moved offshore, during which processE. coli could be either entrained from or deposited to thelake bed in the embayment. The positive partial correlationbetween ECoffsh and ECsed (0.227) further indicates that,with the influence of the foreshore sand as a strong bacteriasource eliminated, the variability of ECsed is still a positivenet contribution to that of ECoffsh. The submerged sandtherefore would have behaved as a second bacteria source(suspension exceeding deposition) for the offshore water evenif there had been no primary bacteria supply from the swashzone (ECforesh and R). The negative partial correlationbetween ECoffsh and EC (-0.251) implies a division between

FIGURE 2. Simulated wave and current vectors during the studyperiod. (a) wave height; (b) current speed; (c) wave propagationdirection (red curve) and current direction (blue curve); filledcircle in (a) and (c): onshore-wave case. In (c), directionsbetween the two horizontal dashed lines (i.e., larger than -45°and smaller than 135°) are offshore; the range bounded by theupper and lower solid lines (i.e., between -180° and -90°) arestrictly onshore directions; the middle solid line indicates theexact onshore direction (-135°).

FIGURE 3. Hydrodynamic variables derived from wave andcurrent data: (a) bed shear stress due to a combined WCBLflow and (b) total runup R compared to inversed beach width40-20Bw.

6734 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 17, 2010

the nearshore and offshore waters of E. coli supply so that,with all other factors held constant, more E. coli retained inthe embayed beach water was statistically associated withless E. coli that could reach the offshore water.

Current Pattern in the Embayment. A bacteria inputand transport path (from the foreshore sand passing the knee-deep water to the deep water outside the embayment) wasinferred from the statistical model in the previous section.A numerical simulation was further conducted to reveal thedetailed current pattern both inside and outside the em-bayment. The Princeton Ocean Model, a full three-dimen-sional finite-difference program using a second-momentturbulence closure scheme (26), was applied to a rectangularcomputation domain that is sufficiently large and surroundsthe study beach. The domain was divided into 150 by 75square grids with a grid size of 13.92 m. The bathymetry atChicago 63rd Street Beach was obtained by the U.S. Geolog-ical Survey in July 2009. For this relatively small domain,heat exchange at the air-water interface and the effect ofwind stress were ignored so that the current pattern in thecomputation domain was only driven by prescribed currentvelocities at open-water boundaries.

Figure 4 shows the distribution of current velocities aroundthe embayment driven by an inflow from the top (A-B) and

right (B-C) boundaries (i.e., a roughly up-coast inflow) andan outflow at the left (A-D) boundary. The inflow velocitywas in the exact longshore direction (Figure 1). The outflowvelocity was given to approximately hold a mass balance(continuity of flow) in the domain. It is observed that theinflow is disturbed by the southern breakwater (the one onthe right), forms a shear layer from the offshore end of thebreakwater, and establishes a large anticlockwise gyre insidethe embayment, which appears to influence the entireembayment. It is also important to note that the flow patternshown by arrows in Figure 4 is only the mean pattern of theturbulent current flow. The trajectory of any particle in theflow field is stochastic due to the random fluctuating part ofthe turbulent flow. Bacteria released from the swash zonecan either be entrained into the gyre, remaining inside theembayment for a longer period of time or travel toward thesouthern breakwater, exit the embayment near the end ofthe breakwater, and eventually reach the offshore samplinglocation. With a constant supply of bacteria from theforeshore sand, more bacteria entrained into the gyre resultsin less exiting the embayment. This is consistent with theinference made from model C (Table 2).

Current pattern in the embayment resulting from a down-coast longshore current is shown in Figure 5. External currententers the computation domain through the A-D boundaryand exits it through B-C. The v-component of the currentvelocity was set at zero at all open-water boundaries. A

TABLE 2. Multiple Linear Regression Models under Various Hydrodynamic Conditionse

model N variable coefficient Pd partial correlation R2(adj. R2)

A: EC for onshore wavea 38

Const -1.424 0.057

0.496(0.435)ECsed 0.575 0.010 0.427BSS 0.212 0.095 0.286ECforesh 0.335 0.038 0.352BwGD 1.587 0.030 0.368

B: ECoffshb 75

Const -1.379 0.001

0.461(0.438)ECforesh 0.597 0.000 0.612R 0.338 0.009 0.304IfCTlong 0.337 0.081 0.205

C: ECoffsh for longshore currentc 64

Const -1.324 0.014

0.451(0.413)EC -0.238 0.051 -0.251ECsed 0.311 0.079 0.227ECforesh 0.540 0.000 0.494R 0.359 0.009 0.332

a Dependent variable: EC; subsample with onshore waves. b Dependent variable: ECoffsh; full sample. c Dependentvariable: ECoffsh; subsample with longshore currents. d P-value for t tests on regression coefficients. e Parsimoniousmodels were obtained by a model selection process with a backward elimination based on the probability of F-to-removelarger than 100.

FIGURE 4. Current flow pattern around Chicago 63rd StreetBeach driven by an external current entering the computationdomain through boundaries A-B, B-C, and exits at A-D.Boundary conditions at A-B: u ) -0.15 ms-1, v ) -0.05 ms-1;at B-C: current velocity decreases linearly from u ) -0.15ms-1, v ) -0.05 ms-1 at B to u ) v ) 0 ms-1 at C; along A-D:uniform u component determined by mass balance of thecomputation domain and a linearly decreasing v componentfrom -0.05 to 0 ms-1. Arrow: current velocity vector.

FIGURE 5. Current flow pattern around Chicago 63rd StreetBeach driven by an external current entering the computationdomain through boundary A-D and exits at B-C. Boundaryconditions at A-D: u ) 0.15 ms-1, v ) 0 ms-1; at B-C: uniformu and zero v determined by mass balance in the computationdomain; along A-B: u ) 0.08 ms-1, v ) 0 ms-1; red curve:streamline of the current flow.

VOL. 44, NO. 17, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 6735

double-gyre pattern is observed inside the embayment, whichplays an important role in entraining contaminants close tothe beach into the central area of the embayment. Bacteria,for example, can either be temporarily retained inside theembayment or be released out of the embayment due to theeffect of turbulent fluctuations on their actual trajectories.Once out of the embayment, bacteria are transported down-coast following the streamlines and observed from theoffshore sampling location. Although the transport patterndriven by a down-coast longshore current (Figure 5) appearsmore complex than that driven by an up-coast current (Figure4), the transport path of bacteria originating from the swashzone is consistent with the statistical results from model Cin Table 2.

DiscussionIn previous sections it has been shown that nearshorehydrodynamic effects, including currents, waves, the com-bined WCBL, and swash, can have an enormous impact onthe variability of E. coli concentration at an embayed beach.For simplicity, we studied the characteristics of bacterialoading and transport under particular conditions: cases withonshore waves and with longshore currents.

Onshore waves significantly (P < 0.05) increased bed shearstresses, responsible for sediment resuspension nearshore,and total runup on the beach face. Cases with onshore waveswere thus major occasions when the beach water receivedbacteria (E. coli) loading from foreshore sand and submergedsediment. Although a combined WCBL also formed on thedays with offshore waves, the resulting bed shear stress turnedout to be significantly lower (Supporting Information). Theeffect of sediment resuspension was thus surpassed bymeteorological effects such as air temperature.

Once in the beach water, bacteria were efficientlyentrained into current flows inside the embayment andreleased out of the embayment when the external currentswere favorable (i.e., longshore). The mean flow patternaround the study beach has been numerically simulated usingthe Princeton Ocean Model with prescribed open-waterboundary conditions. Our simulations (Figures 4 and 5) arestrongly supported by previous observations. In their experi-ments on rectangular model harbors (similar to embayment)with aspect ratios 1:1 and 1:2 (ratio of cross-shore to longshorelengths), Yin et al. (27) showed current circulation patternsinside the harbors that were similar to the single-gyre patternin Figure 4. The circulations in the harbors were found toenhance the mass and momentum exchange between theharbor and the external main flow. The double-gyre patternin Figure 5 was clearly suggested in both field and laboratoryobservations in a test embayment with an aspect ratio 1:3(28). Since the embayment at our study beach is not strictlyrectangular, whether the circulation pattern inside theembayment is a single- or a double-gyre seems to bedependent on the ratio of the length of the breakwater in theupstream to that of the shoreline. When, for example, anexternal current travels down-coast, the shorter northernbreakwater makes the circulation inside the embayment moresimilar to a double-gyre pattern (Figure 5) than a single-gyreone (Figure 4). With either pattern, bacteria can be releasedout of the embayment, which is subsequently observed fromthe offshore sampling location.

The internal circulation generated by external longshorecurrents established a strong connection between the bacteriain the swash zone and those that appeared offshore, by whichthe two variables, ECforesh and R, have positive contribution(partial correlations) to ECoffsh in model C (Table 2). Thepositive partial correlation between ECsed and ECoffshindicates that the variability of E. coli concentration in thesubmerged sediment also influenced that of E. coli concen-

tration offshore positively through sediment suspension andsubsequently transport following the same circulation pat-tern. The associated bed shear stress, nevertheless, wasgenerated very locally (i.e., inside the embayment) by themean internal circulation pattern and complicated by theturbulent shearing. This explains why the variable BSS, simplybased on offshore wave and current data, was not foundsignificant in model C (Table 2). The inference that ECsedwas a net bacteria source to ECoffsh does not eliminate thepossibility of sediment deposition inside the embayment. Asa matter of fact, sediment deposition and suspension canboth happen inside the embayment, but on average thenearshore submerged sediment behaved as a source, ratherthan a sink, for E. coli concentration offshore when the currentwas longshore.

Thupaki et al. (29) investigated budget of E. coli at adifferent Lake Michigan beach by numerical simulation. Theyfound that dilution due to advection and diffusion accountedfor a large portion of the total E. coli budget in the nearshorearea and the net E. coli loss (settling, bacterial base mortality,and solar inactivation) rate within the water column was anorder of magnitude smaller compared to the horizontal andvertical transport rates. Although for a significantly differentbeach setting, their results are consistent with our statisticalinferences that hydrodynamic mechanisms (variables) suchas wave actions and current circulations can connect E. coliconcentrations observed from various locations withoutbeing completely obscured by bacterial mortality andinactivation. That the role of bacteria settling/resuspensionin the present study was not as secondary as that in ref 29is because hydrodynamic transport by currents is typicallymuch weaker in an embayment than in open beach water.

The methods used in the present investigation have wideapplicability; embayed beaches are common in both fresh-water and marine coastal areas, and the models employedhere for estimating or simulating wave runup, bed shearstress, and current circulation patterns are generic enoughto be transferred to other locations. Taking advantage ofnumerical results with high fidelity, accuracy, and resolution,beach monitoring programs can easily estimate pertinenthydrodynamic variables for empirical predictive models, i.e.,a “hybrid” modeling technique as the authors envisioned.The methods presented here are both promising andeconomical for numerous recreational sites in the Great Lakesand marine coastal regions as nowcast/forecast hydrody-namic databases, such as those maintained by NOAA, areextensively available there. Future work at the 63rd StreetBeach has also been planned, including detailed analyses ofsediment textures and sizes and bacteria transport from beachsand to the swash zone through infiltration and exfiltration.

AcknowledgmentsThe authors would like to thank Gregory Lang of the NOAAand Steve Corsi of the USGS for their help with data. Helpfuldiscussions with Rachael Jones and Pramod Thupaki are alsoappreciated. This article was funded by USGS Great LakesOcean Research Priorities Plan (ORPP) and NOAA’s Oceanand Human Health Initiative. This article is Contribution1601 of the USGS Great Lakes Science Center and Contribu-tion No. 1569 of the NOAA GLERL.

Supporting Information AvailableDetails about the estimation of bed shear stress in a wave-current boundary layer flow and the paired t test comparingthe variable mean values between onshore- and offshore-wave cases. This material is available free of charge via theInternet at http://pubs.acs.org.

6736 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 17, 2010

Literature Cited(1) Boehm, A. B.; Grant, S. B.; Kim, J. H.; Mowbray, S. L.; McGee,

C. D.; Clark, C. D.; Foley, D. M.; Wellman, D. E. Decadal andshorter period variability of surf zone water quality at HuntingtonBeach, California. Environ. Sci. Technol. 2002, 36 (18), 3885–3892.

(2) Francy, D. S.; Gifford, A. M.; Darner, R. A. Escherichia coli atOhio bathing beaches - Distribution, sources, waste-waterindicators, and predictive modeling; Water Resources Investiga-tions Report 02-4285; U.S. Geographical Survey: Columbus, OH,2003.

(3) Nevers, M. B.; Whitman, R. L. Nowcast modeling of Escherichiacoli concentrations at multiple urban beaches of southern LakeMichigan. Water Res. 2005, 39, 5250–5260.

(4) Frick, W. E.; Ge, Z.; Zepp, R. G. Nowcasting and forecastingconcentrations of biological contaminants at beaches: Afeasibility and case study. Environ. Sci. Technol. 2008, 42, 4818–4824.

(5) Phillips, O. M. On the generation of waves by turbulent wind.J. Fluid Mech. 1957, 2, 417–445.

(6) Ge, Z.; Frick, W. E. Time-frequency analysis of beach bacteriavariations and its implication for recreational water qualitymodeling. Environ. Sci. Technol. 2009, 43, 1128–1133.

(7) Whitman, R. L.; Nevers, M. B.; Byappanahalli, M. N. Watershed-wide distribution of Escherichia coli along southern LakeMichigan: An integrated approach. Appl. Environ. Microbiol.2006, 72, 7301–7310.

(8) Alm, E. W.; Burke, J.; Spain, A. Fecal indicator bacteria areabundant in wet sand at freshwater beaches. Water Res. 2003,37, 3978–3982.

(9) Whitman, R. L.; Nevers, M. B. Foreshore sand as a source ofEscherichia coli in nearshore water of a Lake Michigan beach.Appl. Environ. Microbiol. 2003, 69, 5555–5562.

(10) Yamahara, K. M.; Layton, B. A.; Santoro, A. E.; Boehm, A. B.Beach sands along the California coast are diffuse sources offecal bacteria to coastal waters. Environ. Sci. Technol. 2007, 41,4515–4521.

(11) Sinton, L. W.; Finlay, R. K.; Lynch, P. A. Sunlight inactivationof fecal bacteriophages and bacteria in sewage-polluted sea-water. Appl. Environ. Microbiol. 1999, 65, 3605–3613.

(12) Lou, J.; Schwab, D. J.; Beletsky, D.; Hawley, N. A model ofsediment resuspension and transport dynamics in southernLake Michigan. J. Geophys. Res. 2000, 105, 6591–6610.

(13) Olyphant, G. A.; Whitman, R. L. Elements of a predictive modelfor determining beach closures on a real time basis: The case

of 63rd Street Beach Chicago. Environ. Monit. Assess. 2004, 98,175–190.

(14) U.S. Environmental Protection Agency. E. coli and enterococciin water - Membrane filter; EPA/600/4-85 076; Office of Scienceand Technology: Washington, DC, 1985.

(15) Nielsen, P. Coastal Bottom Boundary Layers and SedimentTransport; World Scientific: 1992.

(16) Grant, W. D.; Madsen, O. S. The continental-shelf bottomboundary layer. Annu. Rev. Fluid Mech. 1986, 18, 265–305.

(17) Bai, S.; Lung, W.-S. Modeling sediment impact on the transportof fecal bacteria. Water Res. 2005, 39, 5232–5240.

(18) Grimes, D. J. Release of sediment-bound fecal coliforms bydredging. Appl. Microbiol. 1975, 29, 109–111.

(19) Schwab, D. J.; Bennett, J. R.; Liu, P. C.; Donelan, M. A. Applicationof a simple numerical wave prediction model to Lake Erie. J.Geophys. Res. 1984, 89, 3586–3589.

(20) Beletsky, D.; Schwab, D. J. Modeling circulation and thermalstructure in Lake Michigan: Annual cycle and interannualvariability. J. Geophys. Res. 2001, 106, 19745–19771.

(21) Lee, C.; Schwab, D. J.; Beletsky, D.; Stroud, J.; Lesht, B. Numericalmodeling of mixed sediment resuspension, transport, anddeposition during the March 1998 episodic events in southernLake Michigan. J. Geophys. Res. 2007, 112, C02018.

(22) Holman, R. A.; Sallenger Jr., A. H. Setup and swash on a naturalbeach. J. Geophys. Res. 1985, 90, 945–953.

(23) Conley, D. C.; Griffin, J. G., Jr. Direct measurements of bedstress under swash in the field. J. Geophys. Res. 2004, 109, C03050.

(24) Puleo, J. A.; Beach, R. A.; Holman, R. A.; Allen, J. S. Swash zonesediment suspension and transport and the importance of bore-generated turbulence. J. Geophys. Res. 2000, 105, 17021–17044.

(25) Tabachnick, B. G.; Fidell, L. S. Using Multivariate Statistics, 5thed.; Pearson Education: 2007.

(26) Mellor, G. L. Users Guide for A Three-Dimensional, PrimitiveEquation, Numerical Ocean Model; Princeton University: Prin-ceton, NJ, 2004.

(27) Yin, J.; Falconer, R. A.; Chen, Y.; Probert, S. D. Water and sedimentmovements in harbours. Appl. Energy 2000, 67, 341–352.

(28) Muto, Y.; Baba, Y.; Aya, S. Velocity measurements in openchannel flow with rectangular embayments formed by spurdykes. Ann. Disas. Prev. Res. Inst. Kyoto Univ. 2002, 45B–2.

(29) Thupaki, P.; Phanikumar, M. S.; Beletsky, D.; Schwab, D. J.;Nevers, M. B.; Whitman, R. L. Budget analysis of Escherichiacoli at a southern Lake Michigan beach. Environ. Sci. Technol.2010, 44, 1010–1016.

ES100797R

VOL. 44, NO. 17, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 6737