JEQ Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 17 July 2007
Published in J Environ Qual 36:1338-1345 (2007)
DOI: 10.2134/jeq2007.0025
© 2007 American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Nevers, M. B.
Right arrow Articles by Ge, Z.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Nevers, M. B.
Right arrow Articles by Ge, Z.
Agricola
Right arrow Articles by Nevers, M. B.
Right arrow Articles by Ge, Z.
Related Collections
Right arrow Water Management
Right arrow Ecosystem Management
Right arrow Water Pollution

TECHNICAL REPORTS

Surface Water Quality

Interaction and Influence of Two Creeks on Escherichia coli Concentrations of Nearby Beaches: Exploration of Predictability and Mechanisms

Meredith B. Neversa,*, Richard L. Whitmana, Walter E. Frickb and Zhongfu Geb

a Great Lakes Science Center, Lake Michigan Ecological Research Station, 1100 N. Mineral Springs Road, Porter, IN 46304 USA
b Ecosystems Research Division, National Exposure Research Laboratory, 960 College Station Road, Athens, GA 30605 USA

* Corresponding author (mnevers{at}usgs.gov).

Received for publication January 12, 2007.

   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 Study Sites
 Materials and Methods
 Results
 Discussion
 REFERENCES
 
The impact of river outfalls on beach water quality depends on numerous interacting factors. The delivery of contaminants by multiple creeks greatly complicates understanding of the source contributions, especially when pollution might originate up- or down-coast of beaches. We studied two beaches along Lake Michigan that are located between two creek outfalls to determine the hydrometeorologic factors influencing near-shore microbiologic water quality and the relative impact of the creeks. The creeks continuously delivered water with high concentrations of Escherichia coli to Lake Michigan, and the direction of transport of these bacteria was affected by current direction. Current direction reversals were associated with elevated E. coli concentrations at Central Avenue beach. Rainfall, barometric pressure, wave height, wave period, and creek specific conductance were significantly related to E. coli concentration at the beaches and were the parameters used in predictive models that best described E. coli variation at the two beaches. Multiple inputs to numerous beaches complicates the analysis and understanding of the relative relationship of sources but affords opportunities for showing how these complex creek inputs might interact to yield collective or individual effects on beach water quality.

Abbreviations: CFU, colony-forming units • MPN, most probable number


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
Study Sites
 Materials and Methods
 Results
 Discussion
 REFERENCES
 
BEACHES INFLUENCED by creek or river outfalls are often subject to periodic or continuous inputs of water with high concentrations of microbiologic contaminants (e.g., Escherichia coli). During dry periods, outfalls provide a continuous source of water that often has high levels of background E. coli. The amount of E. coli present may increase by several orders of magnitude during rain events because the large discharge associated with rainfall carries with it contaminants washed from areas throughout the watershed (Olyphant et al., 2003; Wyer et al., 1995). Further, rain events are often associated with direct sewage contamination resulting from combined sewer overflows (Crowther et al., 2001). Beaches that are situated near multiple river mouth outfalls are subject to a highly complex point-source system, in addition to the numerous possible nonpoint sources, where creeks may act singularly or in association with one another to affect microbiologic water quality.

Determining the magnitude of influence of any factor associated with high E. coli concentrations can be complicated. Although a point-source input, such as a creek or river, would seem to be the primary source of E. coli to any nearby beach, numerous ambient parameters interact to affect the amount of E. coli that reaches the beaches (Kim et al., 2004; Molloy et al., 2005). Similarly, E. coli present on or near the beach are influenced by water conditions. Many weather and water parameters have been directly associated with E. coli concentrations, including wave height (Francy et al., 2003; Nevers and Whitman, 2005), wind direction (Haack et al., 2003), near-shore currents (Boehm et al., 2002b), and sunlight (Davies-Colley et al., 1994; Fujioka et al., 1981; Whitman et al., 2004). Understanding the interaction between the river outfall and ambient factors is necessary for determining the extent of influence on the beaches and potential for remediation.

Using predictive modeling, it is often possible to estimate the relative influence or association of ambient conditions in the creek and along the beach with E. coli concentrations in the beach water. Empirical modeling studies have been conducted specifically at river outfalls that have linked E. coli concentrations in the river with wind direction, prevailing current (Francy et al., 2003), and wave height (Ahn et al., 2005; Hou et al., 2006; Nevers and Whitman, 2005). The influence of E. coli concentration in outfalls has been related to E. coli concentration on the beaches (Ahn et al., 2005). Although the idea of modeling multiple beaches is not new (Crowther et al., 2001; Nevers and Whitman, 2005), we do not know of other studies that analytically examine effects of two creeks on the bacteria content of multiple beach waters simultaneously. This understanding is important not only in determining the collective content of bacteria at a given beach but also in analyzing individual point source contributions to help understand along-shore dynamics and relative contributions from nearby outfalls.

Predictive modeling has been examined as an alternate approach to traditional beach monitoring, which requires a time-intensive culturing of water samples to assess E. coli concentrations. Because E. coli variation is so high and concentrations can change quickly, sometimes within minutes, hours, or days (Boehm et al., 2002a; Whitman and Nevers, 2004), predictive modeling may be superior to the traditional approach because it provides a real-time estimate of bacterial water quality. Further, characterizing the point-source inputs influencing near-shore water quality can help in managing the beaches relative to periodic increased discharge events.

Along Indiana's Lake Michigan coast, bacteria loading from several creeks has been examined, with most creeks designated as potential sources of high concentrations of fecal indicator bacteria to the lake, particularly during rainfall events (Olyphant et al., 2003; Whitman et al., 1995). The impact of these creeks on E. coli concentrations at the beaches has been long-suspected, such that routine beach monitoring over the past 28 yr at coastal beaches has included sampling creek outfalls. Among these beaches are Mount Baldy and Central Avenue, and interspersed with these two beaches are two creek mouths, Trail Creek and Kintzele Ditch. The separate and combined influence of these creeks on E. coli concentration at these two popular beaches has not been explored, and modeling E. coli concentration in relation to the creeks and ambient conditions may help improve monitoring effectiveness at similar freshwater beaches impacted by multiple point-source inputs.


   Study Sites
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Study Sites
Materials and Methods
 Results
 Discussion
 REFERENCES
 
Central Avenue and Mount Baldy beaches are located within the Indiana Dunes National Lakeshore and are separated by Kintzele Ditch (Fig. 1). Both are wide, sandy beaches bounded landward by natural foredunes. The beaches are popular recreational areas, particularly Mount Baldy, that are periodically closed to swimming due to elevated E. coli concentrations, an indicator of sewage pollution. Kintzele Ditch drains a watershed that includes natural, residential, and urban areas of Michigan City. Immediately to the east of Mount Baldy is Trail Creek, a much larger watershed that drains urban Michigan City and upstream agricultural and residential areas. Escherichia coli concentrations in the two creeks typically have been elevated. Recent improvements to the wastewater treatment facility located on Trail Creek have virtually eliminated combined sewer overflows. There are no known point-source sewage inputs to Kintzele Ditch, but both creeks presumably have animal and human nonpoint sources (Triad Engineering Incorporated, 2003).


Figure 1
View larger version (21K):
[in this window]
[in a new window]

  		Fig. 1. Study locations, including water sampling sites and sites of hydrometeorologic instruments. Water samples for E. coli were collected daily at Central and Mount Baldy Beaches and at the mouths of Kintzele Ditch and Trail Creek. Trail Creek instruments included a gauging station and weather station; Kintzele Ditch instruments included two pressure transducers (one offshore) and a multiprobe monitoring sonde. An acoustic Doppler current profiler was located offshore of Mount Baldy, and predicted measurements for lake conditions were developed by the National Oceanic and Atmospheric Administration a location offshore of Mount Baldy/Kintzele Ditch.

 

   Materials and Methods
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Study Sites
 Materials and Methods
Results
 Discussion
 REFERENCES
 
Water Samples
Water samples were collected at 0800 h 7 d wk–1 for 8 wk in the summer of 2004. Site locations included two sites at Central Avenue Beach, three sites at Mount Baldy Beach, and one site each at the mouths of Kintzele Ditch and Trail Creek. Two water samples were collected at each location: one for E. coli analysis and one for water chemistry. Water samples were collected by immersing a sterile bottle (for E. coli analysis) or a polyethylene bag (for water chemistry) below the surface in approximately 45-cm-deep water. Samples for E. coli analysis were placed in a closed cooler and held at 4°C until analysis. Samples for water chemistry analysis were stored at ambient temperature out of direct light. Field observations made at each sampling site included air temperature, water temperature, wind speed and direction, and wave height.

Instrumentation
A multiprobe sonde (YSI 6920; YSI Inc., Yellow Springs, OH) was deployed in Kintzele Ditch 20 m upstream of the outfall. The instrument recorded water temperature, specific conductance, turbidity, pH, dissolved oxygen, and water depth every 15 min. Data from the instrument were downloaded every 2 wk when the probes were cleaned and recalibrated. Data from the 4-h period preceding E. coli sample collection were averaged for the analysis.

Data for weather and water conditions were collected from several sources. Weather data were collected from an onsite station (Onset Computer Corporation, Pocasset, MA) located near the Trail Creek outfall (N41.72240° 86.90577° W). Wave height and water depth were recorded near the outfall of Kintzele Ditch in Lake Michigan using a freestanding pressure transducer (In Situ Inc., Laramie, WY). An additional pressure transducer was placed upstream in Kintzele Ditch near Beverly Drive, where discharge was also calculated. Discharge in Trail Creek was collected from a United States Geological Survey gauging station (USGS 04095380) located at Michigan City Harbor.

Physical water data were generated by the Great Lakes Environmental Research Laboratory using a model that incorporated hydrometeorologic data from the Gary Regional Airport and included air temperature, dew point, wave height, wind speed, cloud cover, current direction, and current speed.

Near-shore currents were measured using an RDI Acoustic Doppler Current Profiler (ADCP) (Poway, CA) at 16T 0505781 UTM 4618380 (N41° 43.032 W86° 55.830). The 600 kHz ADCP was deployed on 29 June 2004 (Julian Day 181) at 1630CDT at a depth of approximately 7 m and was recovered on 8 Aug. 2004 (Julian Day 221) at 1130CDT. Data from five pings were recorded into 1.0-m bins every 15 min, including current direction, wave direction, water temperature, water level, and wave period.

Laboratory Analyses
In the laboratory, water samples were analyzed for E. coli using the Colilert-18 method (American Public Health Association, 1998), which provides results as most probable number (MPN)/100 mL. The additional water samples were analyzed with laboratory instruments for specific conductance (Acumet meter; Fisher Scientific, Pittsburgh, PA), color spectrophotometry (Hach, Loveland, CO), and chlorophyll and turbidity (Aquafluor; Turner Instruments, Sunnyvale, CA).

Statistical Analyses
Escherichia coli concentrations were log-transformed to achieve normality, so results are presented as log (E. coli MPN/100). Pearson correlation and ANOVA were used in initial comparisons of E. coli data. All exploratory statistics were calculated using SPSS software (SPSS, 2003). Type I error, rejecting a null hypothesis that is true, is defined as predicting that a beach will have an E. coli concentration greater than 235 colony-forming units (CFU)/100 mL when the actual result is less than 235 (235 CFU/100 mL is the United States Environmental Protection Agency's single-sample limit for issuing a swimming advisory [USEPA, 1986]). A type II error, accepting a null hypothesis that is not true, is defined as predicting that an E. coli concentration will be less than 235 CFU/100 mL when the actual E. coli concentration is greater than 235 CFU/100 mL. Laboratory comparisons between Colilert (MPN) and membrane filtration (CFU) results have shown the two are comparable (Eckner, 1998).

Parameters were considered for inclusion in the predictive model using several criteria. Redundant parameters were reduced to a single representative. Number of days for inclusion in the model was maximized by eliminating parameters that were available only for a small portion of the sampling period. From the remaining candidate variables, models were selected using the Akaike's Information Criteria (AIC) method in SAS software (SAS Institute, 2003):

Formula
where L is the maximum likelihood estimation, and k is the number of free parameters (Burnham and Anderson, 1998). AIC method helps to limit overfitting by maximizing fit with the fewest number of parameters. It also limits bias and exaggerated Type I probabilities related to multiple testing (Whittingham et al., 2006). The resulting model was tested for collinearity using the variance inflation factor and the Durbin-Watson test for the residuals. Root mean square error was calculated to compare modeling results.

Correlation analysis was used to determine the potential source of increased E. coli concentrations relative to current direction. The correlation between a vectorial time series (current velocity) and a scalar one (E. coli concentrations) was examined, with an extra projection process implemented to make vectors into associated scalars. Current speeds were projected on axes from 0°, true easterly, to 180°, true westerly, with an increment of 3°. (Scientific and surveying conventions are used to describe directions. In scientific notation, 0° is due east, and angles increase in the counter-clockwise direction. In surveying notation, 0° is due north, and angles increase in the clockwise direction. For the latter, the term "bearing" is used.) For each projection direction, a time series of current-speed components was obtained to yield a correlation coefficient with the constant bacteria concentration time series.


   Results
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Study Sites
 Materials and Methods
 Results
Discussion
 REFERENCES
 
Mean E. coli concentration was significantly higher (P = 0.05) at Mount Baldy (log mean, 1.54 MPN/100 mL ± 0.68 SD) than Central Avenue (log mean, 1.33 ± 0.59). The E. coli concentration at the mouth of each creek was significantly higher than at the beaches (P < 0.01), with Kintzele Ditch having the highest concentration (log mean, 2.91 ± 0.37), followed by Trail Creek (log mean, 2.1 ± 0.68). The sampling site at the Trail Creek outfall was very close to the lake, and it was determined that there was significant mixing with Lake Michigan water, which likely diluted the E. coli concentration. The E. coli concentration exceeded 235 MPN/100 mL—the USEPA single-sample advisory limit (USEPA, 1986)—on 11 occasions at Mount Baldy (N = 71), three occasions at Central Avenue (N = 70), 55 occasions at the Kintzele Ditch mouth (N = 56), and 15 occasions at the Trail Creek mouth (N = 56).

Escherichia coli concentrations for both beaches and both creeks were significantly correlated with one another, with concentrations at Mount Baldy and Central Avenue more highly correlated with the concentration in Trail Creek (Pearson R = 0.545 and P < 0.001 for Mount Baldy; R = 0.391 and P = 0.004 for Central Avenue) than those in Kintzele Ditch (R = 0.378 and P = 0.005 for Mount Baldy; R = 0.285 and P = 0.039 for Central Avenue).

Over the course of the sampling season, Kintzele Ditch had E. coli concentrations consistently higher than both beaches and almost always higher than Trail Creek. Every sample collected at the mouth of Kintzele Ditch except for one had an E. coli concentration of 235 MPN/100 mL or higher (N = 56), with the highest E. coli concentration of log 4.15 (1.4 x 104 MPN/100 mL).

Hydrometeorologic Parameters
Comparisons between E. coli concentrations on each beach and water and weather parameters sampled revealed similar results for both beaches (Table 1). Rain events that had totals over 1 cm were followed by an increase in E. coli concentrations in both creeks (Fig. 2). These increases in the creeks typically lasted for 1 d, after which E. coli concentrations fell to baseline levels. There was often a corresponding increase in E. coli concentrations at the beaches, but not all high E. coli concentrations at the beaches were associated with rain events. Of the 14 occasions when beach E. coli concentration exceeded log mean 2.38 (235 MPN/100 mL), there were rain data available for nine. For those nine events, seven had rainfall within the previous 48 h, and four had near or greater than 2 cm of rain.


View this table:
[in this window]
[in a new window]

  		Table 1. Significant Pearson correlations between Escherichia coli concentrations at Mount Baldy and Central Avenue beaches and the measured hydrometeorologic parameters.

 

Figure 2
View larger version (35K):
[in this window]
[in a new window]

  		Fig. 2. Daily rainfall amounts and Escherichia coli concentrations over the course of the study period (A) at the beaches and (B) in the creeks.

 
Escherichia coli concentrations collected under different prevailing wind directions were considered separately because, historically, onshore winds have been associated with high concentrations of E. coli. Also, wind direction alters where the longshore currents force the creek outfalls to shore. Concentrations were significantly higher during onshore winds (P < 0.001) at Central Avenue (log MPN/100 mL: 1.59 ± 0.48 onshore vs. 1.14 ± 0.60 offshore), but there was no difference for Mount Baldy (log MPN/100 mL: 1.72 ± 0.63 onshore vs. 1.42 ± 0.69 offshore). When wind direction was considered by east and west winds, there was no difference in E. coli concentration between wind directions at Central Avenue (1.39 ± 0.53 east, 1.28 ± 0.62 west), and E. coli concentration was significantly higher (P = 0.017) at Mount Baldy during west winds (1.70 ± 0.67 west, 1.32 ± 0.63 east), a condition that might force Kintzele Ditch water toward the swimming beach.

Model Development
In modeling E. coli concentrations at these two beaches, they were examined separately to assess how the creeks might differently affect the beaches. For Mount Baldy, several variables contributed to the best model, including wave height, specific conductance in Kintzele Ditch, barometric pressure, and wave period (Table 2). There was no autocorrelation according to Durbin-Watson test (P < 0.01), and ANOVA showed the model was significantly different from zero (N = 43; df = 4; p < 0.001). The model had an R2 of 0.722 and an adjusted R2 of 0.694 (Fig. 3):

Formula
where waveht is mean wave height, KDspcond is mean specific conductance in Kintzele Ditch, baropress is barometric pressure, and wvperiod is mean wave period. Means were calculated for the 4 h before E. coli sample collection.


View this table:
[in this window]
[in a new window]

  		Table 2. Results for best models developed using linear regression.

 

Figure 3
View larger version (14K):
[in this window]
[in a new window]

  		Fig. 3. Predicted Escherichia coli concentration versus actual E. coli concentration measured at Central Avenue and Mount Baldy Beaches.

 
When these results were examined for beach closure errors, there were one type I error and three type II errors, so that even though much of the variation was explained, the number of errors in determining beach closures was relatively high (9.3% of the samples).

The best model developed for Central Avenue Beach included three parameters: turbidity in Kintzele Ditch, wave height, and north/south wind component (Table 2). There was no autocorrelation according to Durbin-Watson test (p < 0.01), and ANOVA showed the model was significantly different from zero (N = 58; df = 3; p < 0.0001). The model had an R2 of 0.504, with an adjusted R2 of 0.477 (Fig. 3):

Formula
where KDturb is mean turbidity in Kintzele Ditch, waveht is mean wave height, and NSwind is a binary code for north or south wind direction. Means were calculated for the 4 h before E. coli sample collection.

There were fewer beach closure errors with the Central Avenue model (2.9%): no type I errors and two type II errors. However, over the course of the season, there were only three occasions on which E. coli concentration was 235 MPN/100 mL or higher.

Comparison of Empirical Models to Traditional Bacteria Culturing
The traditional monitoring approach for freshwater involves collecting a water sample, filtering and selectively culturing any present E. coli, and counting the resulting organisms 18 to 24 h later. Beach advisories are posted the day the results are available, at least 24 h after sample collection. An examination of the model on which this approach is based—day 1 E. coli = day 2 E. coli—for the data collected as part of this study resulted in an R2 of 0.01 (adjusted R2 = –0.006) for Mount Baldy, a relationship similarly low as previous modeling studies (Hou et al., 2006). For Central Avenue, the R2 was 0.05 (adjusted R2 = 0.035).

The two models (the predictive model developed here vs. the traditional model) can be compared by examining the RMSE. This value can be used more accurately to compare two models because it depends on the spread of data points rather than the overall fit of a regression line. For Mount Baldy, the predictive model had an RMSE of 0.371, and the traditional model had an RMSE of 0.673, suggesting far more variation in the traditional method of monitoring. Similarly, for Central Avenue, the predictive model had an RMSE of 0.419, and the traditional model had an RMSE of 0.573. Thus, Mount Baldy and Central Avenue models performed similarly when respective RMSE values were compared.

Effects of Current Direction
Current direction influenced the flow of both creeks, changing with prevailing hydrometeorologic conditions. Escherichia coli delivery to the beach is in part a function of distance from the outfall, which relates to time of travel, dilution, and original loading. Given the amount of E. coli measured in the two creeks, it was determined that average loading, according to these data, would be 4.65 x 1010 MPN/h (±2.09 x 1010) for Trail Creek and 1.95 x 1010 MPN/h (±7.29 x 109) for Kintzele Ditch. Distances from Kintzele Ditch and Trail Creek to Central Avenue are 0.6 and 3.9 km, and distances to Mount Baldy are 1.0 and 2.4 km, respectively. Given these distances, lake current velocities of 0.16, 0.28, 0.66, and 1.08 m/s would transport creek water to the beaches in 1 h. Measured currents range 0.05 to 0.1 m s–1, so time of travel is 2 to 10 h. These lag times are influenced by momentum of the outflow jet for the creeks themselves.

Correlation coefficients between the current speed and E. coli concentration at Mount Baldy for all degrees were below the significance level (i.e., none of the correlation coefficients was significantly different from zero) (Fig. 4). In contrast, the E. coli concentration at Central Avenue had a consistently significant correlation with the current speed for most projection angles, especially for those between 20 and 80° (the shoreline is 23°). This implies a strong source to the northeast of Central Avenue Beach that may supply E. coli and associated pathogens to the beach through alongshore current movements.


Figure 4
View larger version (17K):
[in this window]
[in a new window]

  		Fig. 4. Correlation coefficients between current speed components and bacteria concentrations at Mount Baldy and Central Avenue Beaches during the period of 29 June 2004 to 8 Aug. 2004.

 
To take the transport time of the contaminants into consideration (2–10 h as previously estimated), the current data measured at 0600 h (i.e., 2 h before the E. coli data) were used to investigate further the relationship between the occurrence of current shift and the increase of E. coli concentrations at both beaches. A current shift occurs when the alongshore component of the current speed changes sign from the previous day. During the 40 d that current data were available, the alongshore current shifted 15 times. In the same period, E. coli concentration exceeded 150 MPN/100ml on 5 d at Central Avenue Beach; a current reversal occurred on four of these days (Fig. 5). Escherichia coli concentration exceeded the same threshold on 7 d at Mount Baldy Beach, but only three of these days coincided with a current reversal. A Z test for odds ratios (Ramsey and Schafer, 2002) supported the finding that E. coli concentrations greater than 150 MPN/100 mL were associated with current reversals at Central Avenue (p = 0.01) but not at Mount Baldy (p > 0.05). Although the 150 MPN/100 mL threshold was chosen only for a suitable sample size, different thresholds from 140 to 190 MPN/100 mL led to the same inference.


Figure 5
View larger version (20K):
[in this window]
[in a new window]

  		Fig. 5. (A) Escherichia coli concentrations from Days 182 through 221, indicating incidents of a current shift. White bars indicate a current reversal from the previous day; black bars indicate days when E. coli count exceeded 150 MPN/100 ml at Central Avenue Beach; gray bars indicated days when E. coli exceeded 150 MPN/100 mL at Mount Baldy beach. (B) Alongshore component of the current velocity at 0600 h each day; positive component corresponds to the upshore direction, toward northeast.

 

   Discussion
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Study Sites
 Materials and Methods
 Results
 Discussion
REFERENCES
 
Two creek outfalls carry bacteria-laden water from throughout their watersheds and convey it directly to Lake Michigan, where prevailing winds and currents influence its transport to the swimming beaches at Central Avenue and Mount Baldy. The impact and interaction of these outfalls had not been examined before the current study. Even after dispersion, the elevated concentrations of E. coli found at nearby Lake Michigan recreational swimming areas may be partially explained by the much higher E. coli concentrations found in both creeks. Empirical models that interpret this relationship may result in improved predictions of E. coli concentrations for recreational beach management.

After we partitioned the effects of the two creeks, it appeared that Kintzele Ditch likely affected Central Avenue Beach, primarily due to the inclusion of Kintzele Ditch–related parameters in the model and the influence of current shifts on high E. coli concentrations. Trail Creek could have an impact on Mount Baldy but not necessarily on Central Avenue, likely as a result of the great distance between Trail Creek and Central Avenue. Results from correlation analyses, modeling, and the impacts of current direction support these notions. In the case of the Santa Ana River in California, a river of comparable discharge, the predicted river-borne bacteria impact was limited to less than 5 km of the adjacent marine shore due to dilution and dispersal (Ahn et al., 2005). Given the shallower water around Mount Baldy, with the potential to trap fecal bacteria, and the lower momentum of the Trail Creek outflow compared with the Santa Ana River, it is reasonable to conclude that bacteria originating in Trail Creek might not reach Central Avenue in concentrations high enough to affect the bacterial water quality. However, the correlations between E. coli at both beaches and E. coli in both creeks indicate that some larger-scale factors may simultaneously affect all sampling locations. Rainfall has been suggested in this analysis and has been designated in previous analyses (Whitman et al., 1995; Wyer et al., 1995) as a potential factor that can influence or act as a surrogate factor for increasing and decreasing E. coli concentrations at all sampling sites.

The negative correlation coefficients between E. coli at Central Avenue and currents (approximately –0.5) indicate that an increase in E. coli concentration was accompanied by increasing current-speed components in the opposite direction of 20 to 80°; E. coli concentrations were significantly correlated with current components in direction of 200 to 260°. The effect was enhanced when the current speed increased in the southwest direction so that the beach came into the path of the creek source plume, indicating that Kintzele Ditch contributes to E. coli concentrations at Central Avenue. The maximum negative correlation of 0.5 at about 65° also implicates Kintzele Ditch. Given that the shoreline is oriented at 23°, the alongshore component of the currents was very active in bringing the influence of Kintzele Ditch to Central Avenue. Similar significant correlations were not found for Mount Baldy; E. coli concentrations at Mount Baldy beach do not seem to respond noticeably to currents from one direction or the other, according to our analysis. In a previous examination of transport and inactivation of E. coli, it was determined that both creeks influenced E. coli concentrations at Mount Baldy; however, the impact could not be explained only by dispersion, and inactivation was a primary explanatory factor (Liu et al., 2006).

The impact of the creeks on E. coli concentration at the beaches seems to be strongest during events when there is a shift in current direction, a phenomenon that is more perceptible at Central Avenue Beach. When current direction pushes the Kintzele Ditch outfall away from Central Avenue and toward Mount Baldy, there may be a storing of E. coli to the east of the ditch. Beach sand at Mount Baldy is annually replenished, and significant shoaling has occurred (Garza and Whitman, 2004). This may result in facilitated deposition of the Kintzele Ditch load when it is moving to the east. When the current shift occurs, Kintzele Ditch water and residual stored E. coli in surface sediments and water to the east is redirected toward Central Avenue beach, resulting in a spike in E. coli concentrations. Current direction greatly influences the tracking of river outfalls, with waters forced periodically on- or offshore (Boehm et al., 2005), which can result in the release of E. coli stored in the sand. Most events of shifting current occurred simultaneously or shortly after rainfall, so the two factors likely were interacting.

The majority of river plume studies have been conducted in marine environments, but a mechanistic model of southern Lake Michigan currents and associated E. coli inputs from river outfalls was developed that included Trail Creek (Liu et al., 2006). In that examination, it was determined that currents were generally weak, averaging 1 to 3 cm s–1 along the shoreline in summer, and direction was generally from east to west along the study area beaches (Liu et al., 2006). Some of the E. coli present in river plumes are immediately forced toward the beach, depending on wave direction (Ahn et al., 2005). In Trail Creek, that amount of E. coli is influenced by the breakwater that forces flow toward the west—and the study beaches—thereby preventing it from tracking along the prevailing westerly current.

The influence of these outflows on beach bacteria was considered in selecting lake and creek parameters for inclusion in a predictive model that could simultaneously describe the physical characteristics of the system and assess whether the E. coli count in beach water would exceed the limit designated by the USEPA for issuing a swimming advisory. Wave height most often correlated with E. coli concentrations. High waves may increase E. coli due to turbulence and resuspension of bacteria in the sand. When high waves are associated with rainfall, resuspension of sand-borne bacteria and high influx from creek outfalls likely interact to affect E. coli concentration in the nearshore water (Alm et al., 2003; Whitman and Nevers, 2003). Additional factors included in the model that helped account for the variation in E. coli at Mount Baldy and Central Avenue included barometric pressure, wave period, specific conductance, Kintzele Ditch turbidity and wind direction, many of which may be surrogate indicators for rain events. Noticeable among the parameters not included in the model was rainfall; this likely was the result of its sporadic nature, and barometric pressure, a more normally distributed parameter, could substitute for rainfall. The rainfall phenomenon has been evaluated in numerous studies (Ackerman and Weisberg, 2003; Hou et al., 2006; Noble et al., 2003). Further, the incidents of current shift and associated high E. coli concentration often coincided with rain events. In a model developed for currents and fecal pollution at these beaches (Molloy et al., 2005), a human marker indicating sewage input was found in lake water on three occasions, all of which were associated with significant rainfall events and all of which were on days of a noticeable shift in current direction. The strength of rainfall-related factors in determining E. coli concentration at the beaches indicates that point-source fecal contamination is easier to model than nonpoint-source contamination.

The developed predictive models were superior to traditional models (E. coli concentration on testing day = E. coli concentration on day of results determination) in assessing actual E. coli concentrations in the beach waters. Although variation in E. coli at Mount Baldy was better predicted, models for both beaches had lower error than traditional models. Wind direction, which has been identified as a key influence on surface water concentration of E. coli (Nevers and Whitman, 2005), factored into the model for Central Avenue, which may have been similarly related to current direction. Given the dominance of the two creek outfalls in affecting these beaches, these may be reasonable locations for implementing predictive models for monitoring purposes.

Trail Creek and Kintzele Ditch contribute large amounts of E. coli to Lake Michigan. Higher loading rates from Trail Creek and the correlation between concentrations in Trail Creek and Mount Baldy may translate to higher rates of beach closures at Mount Baldy than Central Avenue Beach. Current direction can be a very strong indicator when the main sources are only to one side of the beach. Further, current direction directly affects the forcing of river outfalls toward the beaches, and sudden shifts in current associated with rain events or after prolonged periods in one direction should be monitored for influences on E. coli concentration at the beaches. In combination with the outfall, lake conditions can be incorporated into an empirical model that may be a better approach to monitoring recreational water quality because they resulted in fewer errors. Defining the factors influencing E. coli concentration on a beach from a single source can be complicated, but explaining the added complexity in a system with two outfalls may be possible when individual beaches and creeks are used as references to partition influences.


   ACKNOWLEDGMENTS
 
We acknowledge the contributions of those who assisted in this research, including David Schwab, NOAA; Dawn Shively, Muruleedhara Byappanahalli, Alyssa Bishel, Stacey Byers, Jessica Hardesty, and Cassie Peller, USGS; and the Michigan City Port Authority. We also thank three anonymous reviewers for their constructive comments that helped improve this manuscript. This research was funded by the National Park Service. This article is Contribution 1421 of the USGS Great Lakes Science Center.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 Study Sites
 Materials and Methods
 Results
 Discussion
 REFERENCES
 
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Study Sites
 Materials and Methods
 Results
 Discussion
 REFERENCES
 




This article has been cited by other articles:


Home page
 J. Environ. Qual.Home page
M. B. Nevers, D. A. Shively, G. T. Kleinheinz, C. M. McDermott, W. Schuster, V. Chomeau, and R. L. Whitman
Geographic Relatedness and Predictability of Escherichia coli along a Peninsular Beach Complex of Lake Michigan
J. Environ. Qual., October 29, 2009; 38(6): 2357 - 2364.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Nevers, M. B.
Right arrow Articles by Ge, Z.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Nevers, M. B.
Right arrow Articles by Ge, Z.
Agricola
Right arrow Articles by Nevers, M. B.
Right arrow Articles by Ge, Z.
Related Collections
Right arrow Water Management
Right arrow Ecosystem Management
Right arrow Water Pollution

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Vadose Zone Journal
Soil Science Society of America Journal Journal of Plant Registrations The Plant Genome