Examination of the Watershed-Wide Distribution of Escherichia coli along Southern Lake Michigan: an Integrated Approach

Recent research has highlighted the occurrence of Escherichiacoli in natural habitats not directly influenced by sewage inputs.Most studies on E. coli in recreational water typically focuson discernible sources (e.g., effluent discharge and runoff)and fall short of integrating riparian, nearshore, onshore,and outfall sources. An integrated "beachshed" approach thatlinks E. coli inputs and interactions would be helpful to understandthe difference between background loading and sewage pollution;to develop more accurate predictive models; and to understandthe differences between potential, net, and apparent culturableE. coli. The objective of this study was to examine the interrelatednessof E. coli occurrence from various coastal watershed componentsalong southern Lake Michigan. The study shows that once establishedin forest soil, E. coli can persist throughout the year, potentiallyacting as a continuous non-point source of E. coli to nearbystreams. Year-round background stream loading of E. coli caninfluence beach water quality. E. coli is present in highlyvariable counts in beach sand to depths just below the watertable and to distances at least 5 m inland from the shore, providinga large potential area of input to beach water. In summary,E. coli in the fluvial-lacustrine system may be stored in forestsoils, sediments surrounding springs, bank seeps, stream marginsand pools, foreshore sand, and surface groundwater. While rainfallevents may increase E. coli counts in the foreshore sand andlake water, concentrations quickly decline to prerain concentrations.Onshore winds cause an increase in E. coli in shallow nearshorewater, likely resulting from resuspension of E. coli-laden beachsand. When examining indicator bacteria source, flux, and context,the entire "beachshed" as a dynamic interacting system shouldbe considered.

INTRODUCTION

Top

Introduction

The use of Escherichia coli as an indicator of sewage contaminationhas in recent years become increasingly controversial. It hasbeen argued that the reliability of this bacterium as an indicatorof recent sewage contamination is compromised by its persistenceand common occurrence in nature (14, 42); its capacity to resuscitateor grow under selected conditions (31, 33); the length of analysistime required, leading to inadequate predictability (38); anddifferential survival rates with variations in light, substrate,salinity, and predatory community (4, 15, 24, 25). Lack of hostor geographic specificity (17) may explain problems in identifyingan E. coli contamination source. While many scientists agreethat its presence in high densities suggests recent human oranimal pollution, moderate to low densities of E. coli are moreproblematic for ascertaining pollution, and substantial effortand capital are spent to search for causes (of beach closures)when solutions are typically unclear.

For indicator bacteria, point source contributions are generallymore straightforward than non-point sources, natural inputs,or seasonal effects. Beach closures are a familiar example ofundesirable effects of elevated indicator bacteria; thousandsof potential swimming days are lost every year due to excessiveE. coli (26). In many cases, especially those where bacterialdensities are near the regulatory limit (<235 CFU/100 ml)for safe swimming (35), the sources of the indicator bacteriaare unknown. Even though the solution to these problems shouldideally integrate both source and flux of E. coli, few studiesinclude both aspects. Source studies are generally concentratedon animal or human origins, i.e., source tracking (32, 34);however, ambient contributions (soil and sediments) and environmentalinfluences (bacterial regrowth, inactivation, and die off) arenot adequately included in such exercises. Process, transport,and flux inquiries tend to be largely mathematical and groupedinto mechanistic and empirical modeling approaches.

The ability to generalize from findings has been further limitedby the focus of many investigations. Studies have typicallybeen devoted to a particular type of environment, geographiclocation, or contamination source and have largely concentratedon effluent discharges or stream or beach water. Synoptic studiesacross shores and water are limited but suggest an interactionbetween land and surface water E. coli densities (12, 39). Anintegrated approach that links habitats within watersheds (e.g.,riparian soils with creek and sediments, creeks/streams withreceiving waters, and beach sand with swimming water) is importantfor understanding the distribution, proximate sources and sinks,transient storage, and transportation of microbial pollutantsand the effects on apparent water quality. Eventually, an E.coli "budget" would then be possible, given a better understandingof local bacteria partitioning and flux. Further, by examiningfluvial and lacustrine systems, perhaps a more integrated modelof how the watershed influences E. coli concentrations in beacheswill emerge. A better understanding of the groundwater, physicochemical,and hydrometeorological influences is needed, and knowledgeof bacterial spatial distribution along and within the shoremay provide clues to land-water interactions.

In this paper, we present data from a series of studies conductedover 10 years in the temperate coastal lacustrine and fluvialwatersheds of Lake Michigan to show that E. coli from seeminglydifferent local habitats can be related. Using the results ofthese studies and related published studies, we suggest thata more integrated model of land-water interaction emerges thatmay further our understanding of the environmental occurrenceof this enteric bacterium. By studying numerous interrelatedhabitats, the continuum of potential contamination, from riparianforest soils to beach shoreline, is described along with theimplications for recreational water quality. The derived conceptsmay influence how we approach contaminant tracking and haveimplications for beach monitoring and assessment, regulatoryand public health, and remediation approaches.

MATERIALS AND METHODS

Top

Materials and Methods

Site descriptions.
The described studies occurred between 1991 and 2002 mostlywithin the coastal watersheds of southern Lake Michigan, focusingon three habitats: (i) riparian forest soils, (ii) creek basins,and (iii) beach shore and water (Fig. 1). Stream and foreststudies were done in Derby Ditch and the Dunes Creek watershed,small perennial streams of Lake Michigan.

View larger version (16K):
[in this window]
[in a new window]
FIG. 1. Sampling areas included in the study.

Lake studies were conducted in southern and northwest Lake Michiganin Illinois (63rd Street Beach), Indiana (Lakeview, West, CentralAvenue, and Mount Baldy beaches), and Michigan (Warren Dunes,Good Harbor Bay, and Little Glen Lake). Most were wide, sandybeaches bounded by foredunes and forested uplands within publicparks; 63rd Street Beach is an urban, partially embayed beachwithin the City of Chicago without foredunes, as described elsewhere(39). Little Glen Lake is a small inland lake in Sleeping BearDunes National Lakeshore.

Microbiological analysis.
Substrate samples (beach sand and soil) were analyzed for E.coli by the Colilert-18 system; water was analyzed by the mTECmethod (2) with positive (E. coli ATCC 25922) and negative blanks.Substrate collection and extraction are described elsewhere(5). At least 10% of the presumptive E. coli was confirmed byappropriate methods (3).

Field sampling: stream and watershed. (i) Stream reach.
Water was sampled for E. coli from outfall to upper reachesof Dunes Creek and Derby Ditch at sites previously identified(37). Twelve previously established fixed locations were sampledweekly from 2 August to 20 September 1991 at Dunes Creek, and42 randomly selected sites were sampled over a 1-week periodat Derby Ditch. On 6 August 2001, creek water and sediment from15 randomly chosen sites along 800 m of the main branch of DunesCreek were sampled (5). The stream was only a few centimetersdeep. Sediment was collected by hand to a depth of 2 to 3 cmand placed in a polyethylene bag; a bag was dipped below thestream's surface for water collection.

(ii) Population recovery.
A 0.09-m² plot 5 m above the bank of Dunes Creek was treatedwith 8 liters of 88°C water on 7 June 2001. Leaf litterwas removed prior to treatment. Monitored temperature showedthat it was less than 25°C below 6 cm deep. Composite soilsamples were collected with a sterile, 15-ml liquid medicinedispenser pre- and posttreatment. The plot was then coveredwith sterilized leaf litter. Soil samples were collected atvarious time intervals from June 2001 to November 2002; disturbanceor waste in the area was documented. Limited animal disturbance(digging) was observed only on the day following treatment.An intensive sampling for E. coli distribution in the affectedarea was made 88 days posttreatment. Thirteen samples were takenwithin the plot along primary cardinal directions at the center(within 10 cm) and at intervals of 20, 40, and 60 cm from thecenter. Resultant densities were estimated and mapped usingSurfer 8.0 (Golden Software).

(iii) In vitro growth.
Fresh surface soil (0 to 6 cm depth) was collected from theDunes Creek riparian forest. After thoroughly homogenizing thesoil, aliquots (100 g) were weighed into five plastic containers.The soil in the containers was unamended (control) or amendedwith one of the following treatments: (i) bile salts, 0.15%(wt/wt) (Difco Laboratories, Sparks, MD); (ii) glucose, 1.0g; (iii) liquid fertilizer, 1 ml (Expert Gardner Rose Food,Schultz Company, Bridgeton, MO), containing nitrogen (0.1 g),phosphorus (0.12 g), and potash (0.12 g); or (iv) bile salts,0.15%, plus glucose, 1.0 g. The chemicals were dissolved separatelyin 5 ml of water and added to the corresponding containers.For the control, 5 ml of sterile water was added. The soil ineach container was thoroughly mixed, and aliquots (20 g) wereweighed into five Whirl-Pak bags. The bags were then incubatedat 35°C. At 0, 24, 48, and 72 h of incubation, a bag fromeach treatment was randomly selected and analyzed for E. coli.

(iv) Riparian distribution.
Dunes Creek water, adjacent bank seep, and artesian spring waterand surrounding sediment were sampled along with interveningsoils on 16 June 2001. Creek, seep, and artesian well waterswere collected directly with polyethylene bags, and soil, bank,and creek sediment samples were taken by compositing 5-ml subsamples.

Field sampling: lake and shore. (i) Foreshore pore water.
Weekly pore water samples were taken in triplicate at Lakeviewand West Beach at 1-m intervals from the shoreline to 5 m inlandduring June to September 1993. In 1994, the sampling was repeated;additionally, monthly pore water samples were taken from GoodHarbor Bay. Pore water samples were occasionally taken during1993 and 1994 from Central Avenue (n = 58), Mount Baldy (n =12), and Warren Dunes (n = 15) for reference. A post hole digger(12-cm diameter) that had been cleaned with ethanol and allowedto air dry was used to reach the groundwater; exposed waterwas sampled with a sterile pipette.

(ii) Pore water chemistry.
Water chemistry was monitored weekly during July to August 1994at Lakeview Beach. Three wells were buried 5 m inland from shore,1 m apart, and up to 10 cm below the water table. Sample tubingwas flushed 1 week prior to sampling. Using a 50-ml syringe,one full volume was discarded and another 100 ml of pore waterwas retrieved for analysis. Samples were analyzed for ammonianitrogen, nitrate nitrogen, total phosphorus, total hardness,in situ-dissolved oxygen, pH, specific conductance, and temperature(2).

(iii) Foreshore hydrology.
At West Beach in 1994, five water table wells were positionedat 1-m intervals from the shoreline to 5 m inland. Deep piezometersof the same construction were installed at the 1- and 2-m locations.Water level data were collected continuously using pressuretransducers and data loggers. Seepage meters, to measure theamount of water being discharged or recharged below the lake,were used at 63rd Street Beach in 2000; five such setups wereestablished in about 10 to 20 cm of water on two occasions.On one occasion, three seepage meters were used at West Beachin about 30 cm of water. Water was retrieved from shallow wells(piezometers) at various distances inland along the foreshoreat both locations for E. coli analysis.

(iv) Vertical distribution.
At Lakeview in 1999, sand was excavated to the water table usingthe method described above. Three holes 1 m apart and parallelto the shoreline were excavated at 1, 2, and 3 m from the shore.In each hole, 10 g of sand was retrieved from the exposed verticalwall surface at 5-cm intervals from the water table. The watertable was reached at 10, 25, and 30 cm from the sand surfacefor distances 1, 2, and 3 m inland from the shore, respectively.

Deep sand vibracoring was used at 63rd Street Beach in 2000at two foreshore locations that were 10 m inland and 10 m apart.Ten-centimeter pipes were forced into the sand using a vibratingdevice (10), with resulting core lengths of 1.26 and 1.92 m,including sand up to 1.76 m below the water table. Subsamplesfrom the center of each core were taken at 5-cm intervals abovethe water table; at 2-cm intervals to 20 cm below the watertable (–20 cm); at 10-cm intervals to –60 cm; andthen at –80, –100, –150, and –176 cm.Intervals for the shorter core ended at –110 cm.

(v) Sand-water interaction.
E. coli from sand, pore water, and lake water was monitoredintensively from 23 September to 2 October 1999 at LakeviewBeach. Sand was excavated using methods described above. Threeholes were dug in the foreshore sand 2 m from the shoreline.A 10-ml water sample was collected from pore water using a sterilepipette. Twenty-five milliliters of sand was collected fromthe wall of the hole at 5-cm intervals. A lake water samplewas collected simultaneously with a polyethylene bag. Rainfallbegan the third day of sampling, occurring periodically for30 h and totaling 2.5 cm.

The coincidence of wind direction shifts on E. coli counts wasexamined at Little Glen Lake on 16 to 17 July 1997, where anearly rainless front passed through the area during a 4-h period.Water samples were collected hourly at distances of 7.5, 15,75, 150, and 225 m from the shore.

E. coli from sand and water were monitored at 63rd Street Beachfrom April through September 2000. Sampling and enumerationare detailed elsewhere (39). In short, five transects were sampledthree times a week at the foreshore and submerged sands andat 45-, 90-, and 240-cm water depths.

Statistical path analysis was done using the AMOS version 5software package (Smallwater Corp., Chicago, IL). Path analysis,also called structural equation modeling, is a diagrammaticuse of the general linear model and is a powerful tool for testingpotential relationships between complex and interacting factors.It considers both additive and multiplicative relations andcan be used to test time series relationships. Its major featureis that it allows researchers to structure hypothetical factorsin a diagrammatic format that can then be tested with routineparametric statistics. In this case the overall model was testedby chi-square, and relationships (using maximum likelihood)were assessed by correlation moments and regression weights.Statistical significance was set at P = 0.05.

All data were log₁₀ transformed prior to quantitative assays.

RESULTS AND DISCUSSION

Top

Results and Discussion

Numerous internal and external sources and sinks of E. coliin swimming water complicate the accuracy of water samples collectedfor beach monitoring. With inputs from point and non-point sourcesand the internal processes of deposition, inactivation, andexportation, E. coli counts can change within moments of collectinga water sample. In order to assess the factors influencing E.coli levels in the water, an integrated approach incorporatingthe upland as well as the nearshore processes needs to be used.In this series of studies, we examined upland areas, beach areas,and interactions between sand and water to evaluate the interconnectednessof the system; based on the results and on previous studies,we then developed an illustrative diagram to highlight the complexityof the numerous interacting processes (Fig. 2). While this figureis itself oversimplified, it provides a conceptual frameworkthat helps structure the general research questions put forthin this paper.

View larger version (14K):
[in this window]
[in a new window]
FIG. 2. Conceptual diagram of E. coli within and between stream and beach watersheds (left and right triangles, respectively). For streams, the model partitions stream inputs as human or nonhuman; the latter might include bacterial inputs originating from growth within riparian soils or long-term storage. Within the stream, major processes include (i) inactivation and reactivation, (ii) deposition (temporary or permanent loss) and resuspension, and (iii) death and direct input (e.g., animal defecation and multiplication). Culturable bacteria are delivered from the stream to the beachshed as surface water or groundwater. Internal inputs of E. coli into the lake include foreshore bacteria resuspension (from storage, defecates, and growth) and wastewater releases. Within the lake, there are dynamic interactions between inactivation/reactivation, deposition/resuspension, and offshore importation/exportation of bacteria. The whole beachshed system eventually yields the net culturable E. coli counts (enumerated) monitored by managers at a targeted beach location.

Stream and watershed.
The relationship between a stream and its basin has been longemphasized by both physical scientists and biologists (19, 22,23). Studies have demonstrated the relationships of watershedsto contaminant, nutrient, and sediment loadings and consequentialrelationships to eukaryotic ecology, especially fish and invertebrates.Far less is known about interactions of watersheds, non-pointpollution, and microbial response. Most of the research thathas been done has focused on wet weather events during warmseasons or anthropogenic disturbance (e.g., deforestation, ditching,erosion, farm animals, and sludge applications). We posit thatE. coli levels within Dunes Creek are strongly influenced bybackground levels of E. coli within soils, springs, and submergedsediments and that these sources help explain background bacteriallevels even in the absence of human contamination.

E. coli recovery.
Previous research has revealed the high and variable countsof E. coli in Dunes Creek forest soils (5), but original sourcesremain unknown. After hot water treatment, surface, 4-, and6-cm-deep soil temperature was 72, 48, and 41°C, respectively.Thus, only the top few centimeters was hot enough to kill thebacteria. Pretreatment E. coli count in the experimental plotwas 150 most probable numbers (MPN) g^–1, and no culturableE. coli was found in samples taken 10 min after treatment toa depth of 5 cm. After this, surface populations quickly recoveredto maintain pretreatment densities or higher for over 8 months:1,112, 95, 134, >2,400, 166, 171, and 140 MPN g^–1 atrespective days 5, 20, 28, 81, 160, 216, and 250 (mean, >603MPN g^–1).

The rapid recovery of the surface E. coli was likely due tointroduction of organisms from adjacent or underlying soil.Regrowth remains a possibility (20); see below for additionalevidence. Regardless of the recolonization source, the experimentshows that once established, E. coli can persist for months,through winter and into the next year. The hot water may havealso reduced the densities of potential competitors and predators,allowing the E. coli to flourish. E. coli population pulseshave been witnessed in soil when existing predators and/or competitorsare removed by desiccation (33) or inhibitory substances areadded (6). Also, hot water undoubtedly released a surplus oforganic material that otherwise would have been trapped in thesoil and litter. Neither mechanism explains why E. coli wasable to persist through the following year.

Mapping the posttreatment E. coli population (Fig. 3) showsthe variable density in a given area (<0.1 m²) and emphasizesthe patchiness of microbial distribution in soil (28): E. colipopulations on June 15 varied from 3 x 10² to 8 x 10⁴ MPN g^–1.This highlights the need for multiple samples to approximateE. coli concentration and also shows that high E. coli levelsin the soil can be independent of stream concentrations andpresent even without apparent contamination (7). The experimentalplot was within a few meters of Dunes Creek but was locatedwell above the creek's normal flood plain. We surmise that soilborneE. coli, like that described here, could easily enter the creekby runoff or erosion events, resulting in elevated E. coli levelsin the creek. In the collective watershed, the overall backgroundload in the stream could be substantial.

View larger version (22K):
[in this window]
[in a new window]
FIG. 3. Spatial distribution of E. coli 5 months after treatment in a plot treated with hot (88°C) water. The plot was located in Dunes Creek forest soil, approximately 5 m upland from the creek.

In vitro growth.
With the establishment of naturally occurring E. coli in soil,creeks, and beach sand, the question of the potential for growthmust be considered. In a laboratory experiment, E. coli countsincreased by 2 logs after 72 h in unamended (control) soil;its increase in the amended soils was much more rapid: 2 to5 logs in 24 h (Fig. 4). E. coli growth patterns in glucoseand fertilizer-amended soils were generally similar, as werethe patterns in soils amended with bile salts and bile saltsplus glucose, although the growth rate and relative counts werehigher in the soil amended with bile salts plus glucose. Whenthe experiment was terminated after 72 h, E. coli counts insoil amended with bile salts as well as that amended with bilesalts plus glucose were at least 10 times higher those in thecontrol, fertilizer-, or glucose-amended soils.

View larger version (16K):
[in this window]
[in a new window]
FIG. 4. In vitro growth potential of E. coli in unamended (control) and amended (bile salts, fertilizer, glucose, or bile salts plus glucose) forest soils at 35°C.

As E. coli growth requirements in soil are relatively simple(8), we hypothesized that E. coli can grow in the organicallyrich Dunes Creek forest soils at least during summer months.The delayed E. coli growth in the unamended soil is an indicationthat zymogenous (opportunistic) bacteria (e.g., culturable heterotrophicbacteria), which are more abundant (for every E. coli bacterium,there were 1.46 x 10⁵ to 2.58 x 10⁶ heterotrophic bacterialcounts g^–1) and perhaps better adapted to soil conditions,may have been competing for limited nutrients. On the otherhand, by reducing competition through the addition of bile salts(0.15%), a known inhibitor of nonfecal bacteria, E. coli countsincreased by approximately 4.0 logs in 24 h. These results suggestthat in natural soils, E. coli may grow under certain conditions:in the presence of metabolizable substrates and reduced or minimalcompetition from other microflora (6). Two potential factorssupport this hypothesis: (i) warm soil conditions during summermonths (June to September) and (ii) nutrient availability—itis believed that the hot-water application experiment generateda nutrient pool (simple carbohydrates and other essential nutrients)from dead microbiota and macrobiota; these nutrients were perhapssufficient to promote E. colirecolonization and growth in thesoil. The sudden increase in nutrients can explain short-termresponses; it does not, however, explain the long-term persistenceof E. coli in the treated plot.

Springs.
Soil moisture limits E. coli in riparian areas (5), which isinfluenced by precipitation and groundwater in the form of seepsalong the banks and forest springs. While there was no E. coliin the emanating spring some 20 m from the stream, surroundingspring sediments contained 6, 2, and 1 MPN g^–1 E. coliat the pool center, pool margin, and in the forest soil betweenpool and creek (Fig. 5). This pattern of decreasing counts alongthe margins of spring pools where coarser textured sedimentsgrade to soils and litter has been seen elsewhere (5). E. colimay do well in substrates with moderately low levels of organiccontents but not as well in finer organically rich soils dueto habitat restrictions and negative interspecific relationships.E. coli counts were lower in the creek water than in submergedsands; surrounding bank seeps had high counts similar to thoseof submerged sands. It appears that E. coli accumulates in adownstream direction, largely due to bacterial accrual fromsources, such as soils (18, 33), seeps, and bank sediments,and not necessarily from apparent pollution.

View larger version (27K):
[in this window]
[in a new window]
FIG. 5. Diagram of sampling locations and E. coli counts in creek, bank, soil, and seep in the Dunes Creek watershed. E. coli counts are in most probable number/gram of dry weight for soils and most probable number/100 ml for water.

Stream water.
Earlier studies have shown that E. coli, as well as nutrients,typically increase over stream order, with the highest concentrationsoften found at the outfall and the lowest concentrations nearthe stream origin or below wetlands (5, 37). E. coli densitiesin the Dunes Creek main branch ranged from 10² to 10³ CFU/100ml, but just below wetlands or large pools E. coli was nearlyabsent. The negative correlation with distance from the outfall(P = 0.04; R = –0.535) indicates the trend of E. coliincreasing from headwaters to outfall. E. coli levels were higherin stream sediments than in overlying water, usually by 2 logs,and they were significantly correlated (P < 0.05; R = 0.516).Levels of E. coli in the secondary branches of Derby Ditch andDunes Creek were significantly lower than those in the mainbranch of either, but the creeks were not significantly differentfrom one another (P = 0.05).

In an experiment conducted previously in Dunes Creek, the potentialfor watershed and stream interaction was explored by comparingE. coli counts in creek water, bottom sediment, margin sediment,and sediment at 1 and 4 m from the creek (5). Hierarchical clusteringby squared Euclidean distance of those data reveals similaritiesbetween stream margins and the bank 1 m from the stream edgeand also stream bottom sediment and overlying water (Fig. 6).The lack of similarity between upland and basin E. coli concentrationdoes not necessarily negate a relationship. Perhaps more extensivesampling along a longer stretch of stream would yield strongerbasin-riparian interaction. This approach shows more clearlythe similarity and potential interaction between the water-streambottom and margin-bank components of the watershed.

View larger version (6K):
[in this window]
[in a new window]
FIG. 6. Hierarchical cluster for E. coli counts in Dunes Creek water and upland sites. Media sampled included the creek water (at the center), creek sediment (at the center), creek margin (0.25 m from the creek center), creek bank (1 m from the creek center), and forest (4 m from the creek center).

Past studies show that Dunes Creek no longer has any apparentmajor inputs of point source contamination (37, 40), but ithas been impacted by heavy historical ditching and loss of wetlands.E. coli isolates (n = 188) from Dunes Creek water, soil, andsediments were tested for multiple antibiotic resistance, andover 95% of the isolates tested were susceptible to the antibiotics,suggesting that the source of E. coli within the watershed wasless likely from sewage contamination, as sewage (point source)-associatedstrains exhibit greater resistance to many commonly used antibioticsin humans and animals (30). E. coli densities increased as thesestreams neared the outfalls and recreational beaches and remainedrelatively high throughout the warm season. The transport ofbacteria from creek sediment to the water can be high even duringbase flow (21), which highlights a large potential source ofindicator bacteria for downstream areas. During weekly monitoringin 2001, there was a significant correlation between streamand beach water E. coli concentrations at the Dunes Creek outfall(R = 0.67; P < 0.005; 16 df).

Lake and shore.
The literature on E. coli in large lakes, such as Lake Michigan,is heavily weighted toward point source effects on recreationalwater quality, with special emphasis on water quality managementand health effects. Studies on non-point source contaminationare more sparse, and these tend to lack the integrative synthesisof observations necessary to develop a broad perspective ofthe beach-water system and the influences of environmental factorson water quality.

Foreshore.
Foreshore sands studied were generally similar in characteristics,being more than 95% sand and less than 1% organics, and theorganic component was mostly detritus. Of the physical parametersmonitored, only dissolved oxygen (positively) and ammonia (negatively)were significantly correlated with E. coli (<0.05). Thesefindings suggest that E. coli does better in pore water thathas moderate amounts of organic matter and aerobic conditions.E. coli may lack a competitive advantage in reducing, organicallyrich environments, or possibly the abundant fine material thatclogs interstitial spaces excludes microbial productivity andinhibits pore water circulation (36).

Groundwater influences.
At 63rd Street Beach and West Beach, E. coli was not presentin the groundwater discharging to the lake, nor was it normallyretrieved in well water, similar to findings reported elsewhere(11). E. coli already present is likely held within the sandand cannot readily move horizontally or vertically. These resultshelp explain why most well tests in beach sands rarely showsubstantial E. coli content unless sediments are saturated withcontaminants or testing is in karst environments (15). Verticalfluctuations in the foreshore water table tracked lake levelfluctuations, although they were much attenuated. Mean amplitudewas 6 cm and 15 min apart 2 m inland, and flow reversals occurredon short time scales. The reoccurring displacement of watersuggests there is adequate water exchange for the nutrient andwaste fluxes required to promote microflora in the shallow groundwater.

Horizontal distribution.
There was no significant difference in log mean E. coli concentrationin pore water taken from West Beach (2.45 ± 0.336 standarddeviations) and Lakeview Beach (2.37 ± 0.679) over thesampling period. E. coli counts in pore water were higher thanin lake water, but they did not vary with distance from shore.While both beaches have streams nearby, their respective watershedsare quite different, and E. coli concentrations in lake waterat West Beach are lower in mean and variance. West Beach, however,is more subject to direct sewage contamination, being near theoutfall of the Little Calumet River (27). Regardless of beachlocation or type, E. coli in sand and pore water is persistent,similar to previous findings (1, 39).

Vertical distribution.
Vertical patterns of distribution in the sand above the watertable showed no significant difference (P = 0.536) with distancefrom shore (1 to 3 m). While sand E. coli concentrations atLakeview were lower than those at 63rd Street, results fromboth locations again showed that E. coli on a volume comparisonwas higher in sand than in lake water.

Results from deep sand coring at 63rd Street Beach showed thatE. coli in sand was somewhat variable in depth distributionbut not relative pattern (Fig. 7). The depth range for the firstcore sample was 6 cm above (+6 cm) to 100 cm below (–100cm) the water table. Maximum E. coli counts from this core wereat –8 cm (6,500 MPN/100 ml), although concentrations of1,000 or more MPN/100 ml were consistently found to depths of–18 cm. Below this depth E. coli concentrations exponentiallydecreased; no E. coli was detected by –60 cm. In the secondcore, maximum E. coli concentrations occurred at the water table,but bacterial concentrations were very similar from +15 to –2cm. E. coli levels fell precipitously below –4 cm, withonly occasional occurrence of cells at a depth of –20cm and none thereafter.

View larger version (15K):
[in this window]
[in a new window]
FIG. 7. Depth distribution of E. coli in two deep sand cores collected at 63rd Street Beach.

The high bacterial concentration within the first 30 cm of surfacesand means that swash and wave action can resuspend these bacteria,affecting beach water E. coli content, especially in shallowwater. Pore water was, on average, about 25% by volume of thesand content. Assuming that the entire 30-cm depth of sand canbe resuspended by wave action and wind setup or seiches, itfollows that there is more than an adequate quantity of E. colistored within the foreshore sand ( $~$ 5 to 6 log/100 ml) to exceedthe current U.S. Environmental Protection Agency criterion (2.38log E. coli/100 ml) for closing beaches. This estimation assumesno lakeward dilution and a 22% bottom gradient.

Interaction of water and sand.
Fig. 8 shows the results of structural equation modeling ofmorning E. coli in sand and water at 63rd Street Beach duringthe swimming season of 2000. The overall model, which uses maximumlikelihood to optimize potential regression relationships, failedto be rejected (n = 66; chi-square = 6.6; 5 df; P = 0.251).It should be noted that the model represents the "average" conditionsand that relationships between components would likely changeduring varying ambient conditions (e.g., wind direction, rainfall,and wave heights). The strongest relationships were betweenwaters at 240-, 90-, and 45-cm depths. This is understandable,since waters are contiguous and exchange readily. The relationshipbetween shallow water and shore sand is less but is still highlysignificant. We would have guessed that the net flow would befrom the sand to the beach water, but the model suggests a depositionaltrend; likewise, there was a relationship between foreshoreand submerged sand. Perhaps this is because under summer conditionswinds are predominately offshore, which encourages depositionas opposed to onshore winds, which tend to resuspend and exportE. coli-laden foreshore sands (37). Overall, the model accountsfor a cumulative correlation coefficient of 0.66 in knee-deepwater, where routine sampling normally occurs. The model emphasizesand reinforces findings that foreshore sands are an importantcomponent of sediment-water interactions and both deeper waterand submerged and exposed sediments can act as potential sourcesof E. coli under certain conditions (39). These sources shouldbe identified as proximate sources, since they may not be theoriginal source but only sinks and avenues of contamination.

View larger version (9K):
[in this window]
[in a new window]
FIG. 8. Structural equation model of E. coli in water and beach sediments at 63rd Street Beach from May through September 2000. Cumulative R² values are above the box, and regression weights are above the arrows. Regression weight significance is as follows: *, 0.05; **, <0.01; ***, <0.001. E1 to E4 represent residual error.

Seasonal distribution.
E. coli counts generally show a seasonal pattern in recreationalwaters, with higher counts associated with late summer, a patternnot examined for pore water. In pore water (West and Lakeview)and sand (63rd Street) studies, E. coli counts were significantlydifferent between sampling years, which may be due to a combinationof evolving sampling techniques and real differences betweenthe Chicago, Indiana, and Michigan beaches. These data werestandardized for differences in sampling techniques for yearsand locations by converting yearly concentrations to z scores,which indicates the number of standard deviations a value differsfrom the mean of the population. While E. coli levels in sandduring April and early May were lower than during summer months,there were no particular trends during summer. E. coli levelsin lake water, in contrast, increased during the summer. Yearlymonitoring of sand E. coli strongly suggests that these populationsare in equilibrium (9), but it is not known if this is due toimportation (gulls and lake deposition) or from the bacteriaitself (multiplication and persistence). The founding populationin the latter case remains unknown, but past research demonstratesthat origins are varied (39).

Factor interaction.
Background E. coli counts at Lakeview Beach prior to a rainstormwere log mean 2.55 (±0.11 standard errors) for sand,2.13 (±0.18) for pore water, and 2.28 CFU/100 ml forlake water (Fig. 9). Counts in all three media responded tothe 1-cm, 8-h rain event, and the counts increased more withan additional 0.6 cm of rainfall. By the third day after allrainfall had subsided, E. coli counts had fallen to levels belowthose prerainfall for sand (2.36 ± 0.17) and lake water(2.00 ± 0.15). Counts in pore water were still slightlyelevated over initial counts (2.83 ± 0.39). The immediateincrease in E. coli counts in the lake water indicates thatit was not a runoff-driven event. Rather, the resuspension ofE. coli in the sand and possibly the physical turbulence causedan increase in E. coli in sand, pore water, and lake water.This phenomenon indicates the potential for E. coli to movelakeward from sand, a process that may be greatly elevated duringrain events.

View larger version (20K):
[in this window]
[in a new window]
FIG. 9. Response of E. coli in beach sand, pore water, and lake water to a rain event. SE, standard errors.

At Little Glen Lake, background E. coli counts in the lake werehigher closest to shore (163 CFU at 7.5 m from shore) and decreasedwith increasing depth and distance from the shore: the E. colicount was 0 at 150 m offshore, a phenomenon seen in larger lakes(39). After the wind shifted, counts fell to zero at all locationsexcept 7.5 m from shore. Two hours later, the wind was calmand E. coli was detected in one sample (75 m) at very low counts(21 CFU).

The movement of E. coli between beach sand and swimming wateris the direct link between background E. coli and recreationalwater quality. Because the flux is influenced by numerous factors,including rainfall, wind direction, concentration of E. coli,and beach orientation, E. coli concentration in the water canbe directly impacted by naturally occurring E. coli populationsin the beach sand (9, 39). Once E. coli cells are adsorbed ontoparticulate matter (sand and silt) in moist, nearshore areas,they are protected from environmental stresses, such as desiccationand solar radiation (41). Bacteria that are introduced can establishand perhaps grow in the warm, moist sand and in the presenceof nutrients, e.g., decomposing vegetation or other availablesources, and suitable temperature (5). E. coli cells suspendedin the open water are not only stressed by biotic and abioticfactors but soon settle or become diluted such that concentrationdecreases exponentially with depth or distance from the shore.

The ubiquity of E. coli in beach sand from the shoreline 5 minland and to a depth of up to 70 cm below the surface in thebackshore indicates this is a persistent or even a naturallyoccurring population. There are obvious potential onsite sources(birds, visitors, etc.) of E. coli, but the expanse over whichthe E. coli population extends indicates sources must eitherbe widespread or autochthonous in nature. Also, the common occurrenceof E. coli in beach sand indicates it may be a recurring sourceof E. coli for the nearshore water, accounting for some of thebeach closures during increased wave conditions.

The watershed-wide occurrence and documented persistence ofE. coli throughout a coastal watershed has significant consequencesfor downstream recreational waters. Regardless of original source,E. coli is abundant in forest soil, natural streams, deep backshoresand, pore water, and foreshore sand, all of which may eventuallytravel into receiving waters (Fig. 2). The direct input of streamoutfalls has already been established as a cause of higher countsof E. coli delivered to the monitored swimming water (5), andbeach sand has also been implicated as a potential source (39).

While the contribution of sewage and other sources (e.g., runoffand streams) to increased E. coli levels in water is periodicallya problem for many recreational beaches (26), the continuouspresence of endogenous sources within watershed soils and sandsand throughout the year is likely to have an effect on beachwater quality. Several studies have shown the common occurrenceof indicator bacteria in beach sand (1, 16, 29, 39) and soil(13), but few have connected indicator bacteria densities throughoutthe watershed with adjacent beach water quality. Original sourcesof these bacteria are likely wide ranging and dependent on location,but persistence throughout the watershed indicates a certaininterconnectedness that would affect downstream water quality.

The description of the distribution of E. coli from forest floorto creek bank, creek, outfall, beach water, and foreshore andbackshore sand should not necessarily imply a unidirectionalmovement from upland to beach sinks. There are several transitorypathways in this regard, most notably cell inactivation, death,burial, dilution, and deep-water deposition. Likewise, whileriparian areas may be an important source, sewage, streams,wetlands, and coastal wildlife may significantly contributeto E. coli loadings. All of the contaminants represent the collectivenet E. coli concentrations cultured from beach water duringsampling (i.e., apparent concentration), which makes the partitioningof these various sources difficult even with the aid of modernmolecular techniques (34) or sanitary surveys. Determining thecomponents, fluxes, and interactions of a dynamic conceptualmodel or simply an E. coli "budget" will help define the anticipatedpathogen concentration at a given beach and the real healthconsequences (Fig. 2). Understanding the relationships betweenthe factors that ultimately lead to the net E. coli concentrationsat bathing beaches will help optimize monitoring approachesand define remediation strategies for using indicators thatreflect actual, not apparent, fecal pollution.

Conclusions.
The interaction of E. coli between the watershed and beach showsthere is a complex system with many interacting factors. Onceestablished in forest soil, E. coli can persist throughout theyear and for many months, potentially acting as a continuousnon-point source of E. coli for nearby streams. Storage of E.coli in the fluvial-lacustrine system may include forest soils,sediments surrounding springs, bank seeps, stream margins andpools, foreshore sand, and surface groundwater, and year-roundbackground loading from these components can influence beachwater quality. Further, along the shore E. coli is present inhighly variable counts in beach sand to depths just below thewater table and distances of at least 5 m inland from the shore,which provides a large potential area of input to beach water.All of these potential E. coli inputs may be released into thelake when there are interacting factors, such as wind and rainfall.In these studies, rainfall events increased E. coli counts inthe foreshore sand and lake water, but concentrations quicklydeclined to prerain concentrations. Onshore winds caused anincrease in E. coli in shallow nearshore water, likely resultingfrom resuspension of E. coli-laden beach sand. Overall, whenexamining indicator bacteria source, flux, and context, theentire "beachshed" as a dynamic, interacting system should beconsidered.

ACKNOWLEDGMENTS

This paper is dedicated to Roger Fujioka, whose work on ambientindicator bacteria inspired our initial studies. We also acknowledgethe assistance of Laurel Last, Dawn Shively, Paul Gerovac, MelanieFowler, Maria Goodrich, Tim Fisher, and Paul Murphy.

These studies were funded by the City of Chicago, Chicago ParkDistrict, National Park Service, and Indiana Department of NaturalResources.

FOOTNOTES

* Corresponding author. Mailing address: U.S. Geological Survey, Great Lakes Science Center, 1100 N. Mineral Springs Road, Porter, IN 46304. Phone: (219) 926-8336. Fax: (219) 929-5792. E-mail: rwhitman{at}usgs.gov.

This article is contribution 1395 of the USGS Great Lakes ScienceCenter.

Published ahead of print on 15 September 2006.

REFERENCES

Top