Influence of Climate Change, Tidal Mixing, and ... - ACS Publications

Oct 20, 2005 - Figure 1 Map of Newport Bay indicating the location of monitoring stations, gauge stations, and major creeks draining into the Bay. Acr...
0 downloads 0 Views 698KB Size
Environ. Sci. Technol. 2005, 39, 9071-9082

Influence of Climate Change, Tidal Mixing, and Watershed Urbanization on Historical Water Quality in Newport Bay, a Saltwater Wetland and Tidal Embayment in Southern California ABHISHEK M. PEDNEKAR,† S T A N L E Y B . G R A N T , * ,† YOUNGSUL JEONG,† YING POON,‡ AND CARMEN OANCEA§ Department of Chemical Engineering and Materials Science, Henry Samueli School of Engineering, University of California, Irvine, Irvine, California 92697, Everest International Consultants, Inc., 444 West Ocean Boulevard, Long Beach, California 90802, and Geomatics/ Land Information System Division, County of Orange, Santa Ana, California 92702

Historical coliform measurements (n ) 67 269; 32 years) in Newport Bay, a regionally important saltwater wetland and tidal embayment in southern California, have been compiled and analyzed. Coliform concentrations in Newport Bay decrease along an inland-to-ocean gradient, consistent with the hypothesis that this tidal embayment attenuates fecal pollution from inland sources. Nearly 70% of the variability in the coliform record can be attributed to seasonal and interannual variability in local rainfall, implying that stormwater runoff from the surrounding watershed is a primary source of coliform in Newport Bay. The storm loading rate of coliform from the San Diego Creek watershedsthe largest watershed draining into Newport Baysappears to be unaffected by the dramatic shift away from agricultural land-use that occurred in the watershed over the study period. Further, the peak loading of coliform during storms is larger than can be reasonably attributed to sources of human sewage, suggesting that nonhuman fecal pollution and/or bacterial regrowth contribute to the coliform load. Summer time measurements of coliform exhibit interannual trends, but these trends are site specific, apparently due to within-Bay variability in landuse, inputs of dry-weather runoff, and tidal mixing rates. Overall, these results suggest that efforts to improve water quality in Newport Bay will likely have greater efficacy during dry weather summer periods. Water quality during winter storms, on the other hand, appears to be dominated by factors outside of local management control; namely, virtually unlimited nonhuman sources of coliform in the watershed and global climate patterns, such as the El Nino Southern Oscillation, that modulate rainfall and stormwater runoff in southern California. * Corresponding author tel: (949) 824-8277; fax: (949) 824-2541; e-mail: [email protected]. † University of California, Irvine. ‡ Everest International Consultants, Inc. § County of Orange. 10.1021/es0504789 CCC: $30.25 Published on Web 10/20/2005

 2005 American Chemical Society

Introduction Urbanization of the southern California coastline has nearly eliminated what was historically an extensive network of tidal saltwater wetlands (1). Of the 29 tidal saltwater wetlands that remain, nearly all have been altered in some way by human activity and many are completely reconstructed systems (2). As human activities continue to transform tidal saltwater wetlands and their surrounding watersheds in southern California, coastal managers are faced with developing strategies to minimize, or perhaps even reverse, the negative effects of urbanization on these key ecosystems. However, forecasting the response of a tidal saltwater wetland to specific management decisions (e.g., the building of a housing development in the surrounding watershed) is made difficult by the many physical and biological feedback loops that characterize such systems (3, 4). Furthermore, factors completely outside of local control can dominate a wetland’s ecological status. For example, global climate change alters the intensity and frequency of rainfall and forest fires in the western United States (5, 6). In turn, changes in the frequency and intensity of rainfall and forest fires can alter terrestrial inputs of organic material and sediments into coastal wetlands, with consequent impacts on wetland health and function (3). In this study we compile and analyze 32 years of water quality datasspecifically coliform datasmeasured in Newport Bay, a regionally important embayment and tidal saltwater wetland in southern California. These historical coliform data are interesting for a number of reasons. First, coliform are present at high concentrations in sewage, urban runoff, and agricultural runoff, and hence their presence in coastal wetlands may be an indication of human stress (1). Second, while much is known about the utility of freshwater wetlands for removing fecal indicator bacteria and pathogens in urban runoff (7-13), to our knowledge there is only one published study that addresses this issue in tidal saltwater wetlands (14). The historical data described in this study helps to fill these data gaps. Finally, many wetlands in southern Californiasincluding the one we focus on heres are listed by the U.S. EPA as coliform impaired. Hence, insights into the nature of, and causative agents for, temporal and spatial coliform trends are directly relevant to ongoing water quality management programs (e.g., the development of total maximum daily load management plans) (15). A companion paper (16) focused on the contribution of marinas to fecal indicator pollution in Newport bay, utilizing data from special studies in which samples were collected at a few marina sites over time scales ranging from hours to months. The focus of the present article, on the other hand, is at the scale of the entire wetland, and utilizes historical monitoring data collected at sites scattered throughout Newport Bay and the adjacent surf zone over time scales ranging from months to decades.

Field Site Description Newport Bay (hereafter referred to as the Bay) is the second largest embayment in southern California (Figure 1). The Bay can be divided into regions inland and seaward of the Pacific Coast highway bridge (denoted PCH in Figure 1). The lower portion of the Bay (Lower Bay) is a regionally important recreational area, and one of the largest pleasure craft harbors in the United States. The upper region (Upper Bay) is a state ecological reserve and provides refuge, foraging areas, and breeding grounds for a number of threatened and endangered VOL. 39, NO. 23, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9071

FIGURE 1. Map of Newport Bay indicating the location of monitoring stations, gauge stations, and major creeks draining into the Bay. Acronyms include San Diego creek (SDC), Santa Ana Delhi channel (SAD), and Pacific coast highway (PCH); the alphanumeric code assigned to each sampling station by the Orange County Health Care Agency is indicated next to each station marker. species, including the Clapper Rail (Rallus longirostris levipes), the Belding’s Savannah Sparrow (Ammodramus sandwichensis beldingi), and the Salt Marsh Bird’s Beak (Cordylanthus maritimus spp. maritimus). Within Upper Bay there are seagrass beds (Zostera marina and Ruppia maritima), unvegetated tidal flats, and a salt marsh dominated by Spartina foliosa and Salicornia virginica. The Bay receives episodic freshwater input linked to infrequent rain events (average of 29 cm of rainfall per year) and significant inputs of energy, nutrients, and organisms from the sea and the surrounding watersheds. Similar lagoon morphologies and climate occur at other locations along the southern California coastline (2) and along the coasts of South Africa (17, 18) and Australia (19). The Bay is listed by the U.S. Environmental Protection Agency as impaired for fecal coliform, nutrients, sediment, and toxic pollutants (15). A significant fraction of the Bay’s pollution problem is thought to originate from dryweather and stormwater runoff that flows into the Bay from numerous creeks, channels, and drains (16, 20, 21). The San Diego Creek (SDC) and Santa Ana Delhi Channel (SAD)s which drain into Upper Bay (see Figure 1)saccount for the largest runoff volume, and collectively drain approximately 352 km2 of highly developed land in the cities of Newport Beach, Orange, Costa Mesa, Santa Ana, Laguna Hills, Laguna Woods, Lake Forest, Irvine, and Tustin. 9072

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 23, 2005

Materials and Methods Land-Use Changes. Land-use in the SDC and SAD watersheds was characterized at three times (1968, 1993, and 2003). At each time point land-use was grouped into six different categories (agricultural, commercial, industrial, residential, public/recreational, and undeveloped/unaccounted) based on historical land-use maps prepared by the County of Orange (Figure S1 in the Supporting Information). Historic land use for 1968 was digitized and interpolated from an archived Orange County land use map prepared by the planning department. Land-use data from 1993 and 2003 were derived from detailed parcel-based information stored in Orange County’s land information system database. Historical Water Quality Data. Weekly (and in some cases, monthly) coliform data were obtained from the Orange County Health Care Agency (OCHCA). Written records for total coliform (TC) were available for the period 1969-1985; electronic records were available after 1985. Electronic records were available for fecal coliform (FC) from 1985 to 1999. The number of sites at which testing was carried out in Newport Bay, and the measurements reported for each site, varied over the 32-year time frame. To carry out the time-series analyses described below, we utilized only those data, and sites, for which a more-or-less unbroken data record was available. Accordingly, 25 sampling sites (excluding SAD

and SDC) were selected and TC was adopted as a long-term water quality index. Because the FC data has more gaps, these data were used only for comparison to TC; i.e., trend analyses were not performed on FC data. TC measurements were carried out by OCHCA, using multiple tube fermentation (MTF) (APHA 9221 B) from 1969 to 2001. FC measurements were carried out by OCHCA using MTF (APHA 9221 E.1) from 1985 to 2001. These standard methods yield the concentration of TC and FC in units of most probable number (MPN) per 100 mL of test sample. After 2001 OCHCA changed the analysis method from MTF to membrane filtration (MF). Because results from the MTF and MF methods may not be comparable, we opted not to include the more recent MF data in the analysis reported here. The MTF records of TC are nearly continuous, with the exception of the period 19951997 during which sampling at some stations in the Bay was curtailed due to funding constraints (data gaps at each station are summarized in Table S1). Written data records were transcribed into the computer by two students, one of whom called out the test result and observed as the other student entered the number into the computer. TC was also measured in the surf zone along the ocean side of the spit that separates Newport Bay from the ocean (blue stations, Figure 1). These data, which were available for the entire 32-year study period, were obtained either directly from the Orange County Sanitation District, or from data records compiled for a separate study (22). As explained in the Results section, the within-Bay and surf zone data were grouped into four different station categories as follows: Upper Bay, Lower Bay, Western Bay, and Surf Zone. These station categories are designated by different colored symbols in Figure 1. Stream Gauge and Rainfall Data. Daily measurements of surface water flowing into Upper Bay from the SDC and SAD were available for most of the 32-year study period; these data were recorded at the two stream gauge stations indicated by large yellow rectangles in Figure 1 (see Table S1 for data gaps). From 1971 to 1992, stage was measured at SAD using a Stevens Type A-35 analogue recorder driven by a float inside an 18-in. corrugated metal pipe; after 1992 stage was measured using a Stevens A-71 analogue recorder driven by a Fluid Data Balance beam manometer. The latter stage recorder was installed a short distance (approximately 0.4 km) upstream of the older recorder. The gauge station at SDC is equipped with a continuous water stage recorder (Stevens A-71) and ALERT (automated local evaluation in real time) transmitter/data logger. Stream discharge is computed from stage using a Windows Software Hydrologic software program (XStream Measures, Auberry, CA); the Western Hydrological Systems’ Surface Water Program (XStream Measures, Auberry, CA) was also used early in the study period. Surface water discharge measurements were carried out bimonthly at SDC using a current meter or a Marsh-McBirney Flo-Mate (Ashtead Technologies, Rochester, NY). Historical rainfall data were measured at a station in Newport Bay (data are available since 1922, designated by NOAA as RNST 88, see blue cross in circle marker in Figure 1). For those periods of time when discharge measurements were not carried out at SDC and/or SAD, discharge values were estimated from rainfall records using the Rational formula (23): Qs ) kAR(t) + b, where k is the dimensionless runoff coefficient, A is the drainage area, R(t) is the rainfall measurement (in meters/second) at a particular time, and b is the base flow rate in m3/s. These constants were estimated by regressing measured daily discharge against daily rainfall measurements: kA ) 80 × 106m2 (SDC) and 16.1 × 106m2 (SAD), and b ) 0.55m3/s (SDC) and 0.11m3/s (SAD). Over the period of time covered by this study (32 years since 1969), rainfall was measured at RNST 88 using a tipping bucket located under a funnel assembly on top of a standpipe structure maintained by County of Orange personnel.

Modeling of Flushing Time Scales for Newport Bay. Flushing times in Newport Bay were estimated by simulating flow in the Bay with a two-dimensional hydrodynamic model, RMA2 (24), and simulating mass transport in the Bay with a two-dimensional water quality model, RMA4 (25). RMA2 is a two-dimensional depth-averaged finite element hydrodynamic numerical model that computes a finite element solution of the Reynolds form of the Navier-Stokes equations for turbulent flows. RMA4 model takes the velocity field from the RMA2 model and solves the depth-integrated transport equations. Figure S2 (Supporting Information) shows the model grid for the Bay. Bay bathymetry was constructed from multiple data sources. Bathymetry for the Lower Bay and the region upstream of the PCH Bridge (see Figure 1) was based on a bathymetric survey conducted by the U.S. Army Corps of Engineers in 1999 and 2001, respectively (26). Bathymetry in Western Bay was based on a dredge plan prepared by the City of Newport Beach in 1976 (27). Bathymetry in Upper Bay was estimated from a 1999 NOAA Chart (No. 18754). The RMA2 model was run using a mean tide consisting of two daily highs (MHHW and MHW), and two daily lows (MLW and MLLW). The tidal elevations and tidal currents simulated by the RMA2 were then used to drive the water quality model RMA4. To calculate flushing times for the Bay, a numerical experiment was carried out as follows. At time t ) 0, each within-Bay node was assigned the same initial pollutant concentration, C ) C0. Grid nodes on the ocean side of the tidal outlet were assumed to be pollutant free, C ) 0. The concentration of the pollutant at each within-Bay grid node was then followed over time as contaminated Bay water was replaced with clean ocean water by tidal flushing. An example of one of the resulting concentration-time curves is shown in Figure S3 (Supporting Information). This particular curve was calculated at the numerical gauge shown in Figure S2. Referring to Figure S3, the subtidal simulated concentration decays roughly exponentially with time. Accordingly, the flushing time was taken as the time required for the concentration at a particular site to drop to e-1 of the initial concentration (i.e., the time when C/C0 ) e-1 or 0.368). Frequency Domain Analysis. To identify periodic patterns in the coliform time series, periodograms were calculated using a fast Fourier transform (FFT) routine numerically implemented in the computer program Matlab (MathWorks, Natick, MA). Cumulative Residual Analysis. Cumulative residuals analysis is a procedure for revealing long-period trends in time series measurements, and involves summing departures of a measurement from the measurement’s overall mean. Let Xj represent the value of the measurement in question at the jth time point, and let X h represent the arithmetic mean N of Xj over the entire N points in the time series, X h ) ∑i)1 Xi/N. The cumulative residual up to the jth time point is calculated by summing residual values from the start of the time series to the jth time point: j

Cj )

∑R

(1)

i

i)1

where the residual values are departures from the mean at the ith time point, Ri ) Xi - X h . Cumulative residuals were computed for the following quantities: (1) monthly cumulative rainfall recorded by the rain gauge in Newport Bay; (2) monthly mean volumetric discharge from SDC and SAD; and (3) monthly average log-mean TC concentration measured at four different station categories (Upper Bay, Lower Bay, Western Bay, and Surf Zone). Empirical Orthogonal Function (EOF) Analysis. EOF analysis is an approach for identifying dominant spatial and temporal patterns (or “modes”) in time series measurements VOL. 39, NO. 23, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9073

TABLE 1. Land Use Details for San Diego Creek and Santa Ana Delhi Watershed San Diego Creek watershed (km2)a

Santa Ana Delhi watershed (km2)a

land use

1968

1993

2003

1968

1993

2003

1. agricultural 2. commercial 3. industrialb 4. residential 5. other (public, recreational) 6. undeveloped & unaccounted 7. developed (1+2+3+4+5) total (6+7)

104 (34%) 2 (1%) 5 (2%) 33 (11%) 34 (11%) 129 (42%) 178 (58%) 307

19 (6%) 26 (9%) 65 (21%) 47 (15%) 2 (1%) 148 (48%) 159 (52%) 307

5 (2%) 30 (10%) 61 (20%) 76 (25%) 13 (4%) 123 (40%) 184 (60%) 307

6 (14%) 1 (3%) 9 (19%) 22 (49%) 3 (7%) 4 (8%) 41 (92%) 45

0 (0%) 7 (17%) 16 (35%) 15 (34%) 1 (3%) 5 (11%) 40 (89%) 45

0 (0%) 7 (15%) 15 (33%) 16 (35%) 2 (5%) 5 (12%) 40 (88%) 45

a

The numbers in parentheses indicate percentage land use based on total area. b Industrial land use includes transportation.

that are collected at a number of different spatial locations (28). The application of EOF to the historical water quality data in the Bay involved the following steps: (1) grouping of stations into the four categories described above (Upper Bay, Lower Bay, Western Bay, and Surf Zone); (2) computation of monthly log-mean TC concentrations for each station category; (3) organization of the monthly log-mean TC concentration data into a data matrix, Cij; rows and columns of the matrix correspond to, respectively, station category and time (i.e., month and year); (4) preparation of a demeaned data matrix, D ) [dij] ) (Cij - C h i)/ri, where C h i and ri() σi) represents the 32-year mean and standard deviation, respectively, of all log-transformed TC concentrations measured in a specific station category (28); and (5) decomposition of the de-meaned data matrix into EOF modes, where each mode consists of a set of spatial and temporal eigenvectors and associated eigenvalues. A detailed description of how the spatial and temporal eigenvectors were computed can be found in a companion paper (16). The temporal and spatial eigenvectors associated with a particular mode represent the dominant temporal and spatial variability patterns captured by that mode. The modes are ordered such that the first mode captures the most data variance, the second captures the next most data variance, and so on. The fraction of variance captured by the ith mode is numerically equal to the magnitude of the corresponding eigenvalue, or “loading” λi.

Results Historical Land-Use Change. Table 1 summarizes the evolution of land-use patterns in the SDC and SAD watersheds from 1968 to 2003. Despite rapid population growth in the two watersheds (from 105 671 residents in 1968 to 576 038 residents in 2000), the percentage of land with some form of human development (i.e., agriculture, commercial, industrial, residential, and public/recreational) remained relatively constant at 52-60% and 88-92% in the SDC and SAD watersheds, respectively. While the developed area remained relatively constant over the 35 years, land-use within the developed area changed dramatically, from substantial agricultural land-use in 1968 (34% and 14% of total land in the SDC and SAD, respectively) to virtually no agricultural land-use in 2003 (2% and 0% of total land). In the SDC watershed, agricultural land-use was replaced with commercial land-use (from 1% in 1968 to 10% in 2003), industrial land-use (from 2% of total land in 1968 to 20% in 2003) and residential land-use (from 11% in 1968 to 25% in 2003). In the SAD watershed, agriculture and residential landuse was replaced with commercial and industrial land-use (from 3% and 19% of total land in 1968 to 15% and 33% in 2003). Comparing the SDC and SAD entries in Table 1, it is striking how different the two watersheds are with respect to their overall size (307 km2 vs 45 km2), percentage of the different land use categories, and the degree and nature of land-use change over the 35-year time frame. 9074

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 23, 2005

Correlation of TC and FC Data. The Bay is designated by the U.S. EPA as impaired for FC (15). Because the historical record for TC is more complete than the historical record for FC (see Table S1 in the Supporting Information), ideally the former could be used as a proxy for the latter. All within-Bay samples tested for FC were also tested for TC (n ) 6811) and hence the comparability of these two indicator bacteria can be assessed from the historical data. Measurements of TC and FC in the Bay are significantly correlated (p < 0.01, Figure 2). The correlation strengthens with distance inland in the order Lower Bay (Spearman Rank correlation, Sp ) 0.5), Western Bay (Sp ) 0.72), and Upper Bay (Sp ) 0.87). TC therefore appears to be a reasonable proxy for FC, particularly in portions of the Bay (i.e., Upper Bay and Western Bay) that exhibit the most severe coliform contamination (see next section). Spatial Distribution of Total Coliform in Newport Bay. As a first step toward analyzing this very large data set, logmean concentrations of TC were computed from all measurements conducted over the 32-year study period (top left panel in Figure 3). Log-means were used, instead of arithmetic means, because coastal bacterial monitoring data typically vary over many orders of magnitude (22, 30). The 32-year log-mean TC concentrations exhibit an inland-to-ocean gradient, with the highest concentrations in Upper Bay, intermediate concentrations in Lower Bay, and lowest concentrations in the surf zone outside of the Bay. Standard deviations computed from these data exhibit a similar trend, with the highest variability recorded in Upper Bay, intermediate variability in Lower Bay, and the lowest variability in the surf zone (top right panel, Figure 3). Previous studies (22, 30-33) have noted that coastal water quality in this region of southern California tends to be worse during storms, which in this semi-arid climate occur primarily in the January through March time frame (34). To determine if the same result applies to the Bay, log-means were calculated from all TC measurements conducted during the winter months of January, February, and March (JFM); logmeans were also calculated from all TC measurements conducted during the traditionally dry weather months of June, July, and August (JJA). JFM and JJA anomaly plots were then calculated by subtracting the 32-year log-means from the JFM and JJA logmeans. When calculated in this way, positive (negative) anomalies indicate that the concentration at a particular site is, on average, higher (lower) compared to the 32-year mean at that site. All within-Bay and surf zone sites exhibit positive JFM anomalies (ranging from 0 to 0.5 log units) and weakly positive to strongly negative JJA anomalies (ranging from 0.1 to -0.4 log units) (Figure 3, bottom two panels). An examination of the plots in Figure 3 reveals that sampling sites can be divided into four groups based on the following patterns: (1) stations located north of the Pacific Coast Highway bridge (PCH in Figures 1 and 3) have large positive JFM anomalies; (2) stations located to the south of

FIGURE 2. Cross plot of log transformed total coliform (TC) and fecal coliform (FC) concentrations. The Spearman rank correlations between log TC and log FC are indicated on each panel and the slope of the best-fit line is provided in the key. the Pacific Coast Highway have smaller positive JFM anomalies; (3) Stations located in the western portion of Lower Bay have weakly positive to weakly negative JJA anomalies; and (4) finally, surf zone stations have lower log-mean, and smaller (in absolute value) JFM and JJA anomalies. In the analysis presented in this paper, all TC data collected in the Bay and nearby surf zone were therefore grouped into these four station categories, hereafter referred to as Upper Bay, Lower Bay, Western Bay, and Surf Zone stations. Grouping TC data into these four categories facilitated presentation of the time series (by reducing the number of separate time series from 25 to 4), and virtually eliminated data gaps present in the raw water quality data sets. The station category assigned to each sampling station is indicated in Table S1 (Supporting Information) and by the color of sampling stations in Figure 1. Spatial Distribution of Flushing Time Scales in Newport Bay. In tidal embayments such as Newport Bay, the

rate at which pollutants locally dissipate due to tidal mixing can vary significantly by location (35, 36). This phenomenon can be characterized by plotting the spatial distribution of flushing times, where the latter represent an estimate of the time it takes a conservative pollutant to locally dissipate due to tidal mixing alone (see Materials and Methods Section). Flushing times calculated for the Bay vary from 0 to >30 days (Figure 4). The shortest flushing times occur in regions of the Bay located near the ocean outlet (blue regions in Figure 4). The longest flushing times (ca. 30 days) occur in dead-end regions of the Bay where tidal flow is muted, such as Western Bay and Upper Bay (red regions in Figure 4). Comparing Figures 3 and 4, it is evident that regions of the Bay with the highest 32-year log-mean TC concentrations (and highest standard deviations in the log-transformed TC concentrations) are, in many cases, the same areas with the longest flushing times. However, flushing times alone are not a predictor of the magnitude of local contamination. For example, both Western Bay and Upper Bay have very long tidal flushing time scales (>28 days), yet the 32-year logmean TC concentration in Upper Bay is approximately 0.6 log units higher than the 32-year log-mean TC concentration in Western Bay. Time series of Total Coliform, Rainfall, and Runoff. Figure 5 presents a 32-year time series of rainfall (blue bars), stream discharge from SAD and SDC (black bars), and monthly log-mean TC concentrations in SAD, SDC, Upper Bay, Western Bay, Lower Bay, and the Surf Zone (red bars). As expected, monthly cumulative rainfall exhibits significant seasonal variability, with most of the rain falling during the winter (JFM) season and little rain falling during the summer (JJA) season. Monthly average discharge records for SAD and SDC also exhibit significant seasonal variability, with peak flow occurring during JFM periods. Spearman rank correlation between monthly cumulative rainfall and average monthly discharge in SDC and SAD is large (Sp > 0.8) and significant (p < 0.05, Figure S4). TC concentrations measured in SDC and SAD exhibit a weak positive correlation with stream discharge (Sp ) 0.34, 0.33, p < 0.01), implying that TC in these two creeks is not diluted during periods of high flow (see fourth and fifth panels, Figure 5). Correlations between rainfall and TC concentrations are strong (Sp > 0.6) and significant (p < 0.05) in Upper Bay, Lower Bay, and the Surf Zone, but lower for Western Bay (Sp ) 0.4, p < 0.05) (bottom four panels in Figure 5; see also Table S2 and Figure S4 in Supporting Information). In Upper Bay and SDC, TC concentrations during the summer appear to exhibit interannual (i.e., multi-year) trends, particularly during the second half of the study. For example, summer (JJA) TC concentrations measured in SDC decline during the second half of the study, from a geometric mean of ca. 10 000 MPN/100 mL in 1979 to ca. 1000 MPN/100 mL in 2000 (gray curve in the panel labeled SDC, Figure 5). By way of comparison, the geometric mean standard for TC in California coastal recreational waters is 1000 MPN/100 mL. TC concentrations measured during the summer in Upper Bay exhibit roughly the same interannual trend, peaking in 1979 and declining thereafter (gray curve in panel labeled “Upper Bay stations”, Figure 5). Interannual trends at the other sites are more complex and analyzed later in the context of the cumulative residual analysis. Frequency Domain Analysis of TC Data. Periodograms calculated from the 32-year TC data reveal strong annual return periods (12-month periods) in Upper Bay, Lower Bay, and the Surf Zone (Figure 6). These annual return periodss which imply that periods of poor (and good) water quality tend to repeat every 12 monthsspresumably reflect the role that winter storms play in mobilizing fecal indicator bacteria pollution into the Bay (see discussion above). Periodograms calculated from the Western Bay data, on the other hand, VOL. 39, NO. 23, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9075

FIGURE 3. Color contour plots showing the 32-year mean of the log transformed TC concentrations (top left panel), standard deviation of the log transformed TC concentrations (top right panel), JFM anomaly plot (bottom left panel), and JJA anomaly plot (bottom right panel).

FIGURE 4. Flushing time scales calculated from numerical models of mass and momentum transport in the Bay. exhibit a weaker annual peak, and a stronger semiannual peak, suggesting that periods of poor (or good) water quality tend to repeat every 6 months at this location. It was noted earlier that the JJA and JFM anomalies for Western Bay were weakly positive to weakly negative, in contrast to the other station categories (Upper Bay, Lower Bay, and Surf Zone) which all had strongly negative JJA anomalies. The fact that 9076

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 23, 2005

Western Bay has near or above-average TC concentrations during both JFM and JJA periods is consistent with the semiannual return period revealed by the frequency domain analysis. Empirical Orthogonal Function (EOF) Analysis of TC Data. The goal of the EOF analysis is to derive a set of empirical spatial and temporal modes that, together, capture

FIGURE 5. Time series of monthly rainfall totals (blue bars) at Newport Harbor gauge station, volumetric discharge from SAD and SDC (black bars), and monthly TC log means at SAD (Santa Ana Delhi), SDC (San Diego creek), Upper Bay, Western Bay, Lower Bay, and Surf Zone (red bars). Yellow boxes in the discharge time series indicate periods of time when data gaps in the discharge record were filled using the Rational method (see Materials and Methods section). The green and gray solid curves represent the JFM and JJA signals. a large percentage of the variance in the de-meaned and normalized TC data set (see Materials and Methods section). The first modeswhich accounts for 69% of all the variance in the de-meaned and normalized TC data, as shown in Table S3s is strongly correlated (Sp ) 0.7, p < 0.01) with total monthly rainfall recorded in the Bay (compare rainfall measurements with Mode 1 curve in Figure 7). In other words, nearly 70% of the variance associated with the de-meaned and normalized TC measurements in the Bay is apparently caused by stormwater runoff. The second mode accounts for approximately 50% of the remaining variance, and is most pronounced in Western Bay (see spatial eigenvalue entries for Western Bay, Table S3). The second mode does not exhibit obvious seasonal patterns before 1985, but after 1985 the second mode is frequently higher in the summer and lower in the winter, precisely the opposite seasonal pattern exhibited by the first mode (third panel in Figure 7). Cumulative Residual Analysis of Rainfall, Runoff and Total Coliform. Cumulative residual analysis is a sensitive method for identifying long-period trends in a time series record (37, 38). Cumulative residuals exhibit an increasing (decreasing) trend when measurements are consistently higher (lower) than average over a sustained period of time

(see Materials and Methods section). Cumulative residuals calculated from rainfall records at Newport Beach (top panel, Figure 8) reveal interannual periods of dryer-than-average weather (from the start of the time series in 1969 to 1977, from 1983 to 1991, and from 1998 through the end of the time series in 2001) and interannual periods of wetter-thanaverage weather (from 1977 to 1983 and from 1991 to 1998). These 5-10 year wet/dry cycles are well-documented for southern California, and correlate with the return period of El Nino Southern Oscillation (ENSO) anomalies which bring wetter than usual winters to this part of the world (38). Long period trends in runoff discharged from the SDC and SAD closely parallel the rainfall trends described above (compare top two panels in Figure 8). Some of this similarity is an artifact of using rainfall records to fill data gaps in the SDC and SAD discharge records (see Materials and Methods and yellow boxes in the second and third panels of Figure 5). Even if these data gaps are excluded, however, the correlation between cumulative residual curves calculated from rainfall and discharge is strong and significant (Sp ) 0.96 and 0.61 for SDC and SAD, p < 0.01). It is also striking that the cumulative residual curves calculated from the SAD and SDC discharge data are very similar, despite quite VOL. 39, NO. 23, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9077

average thereafter. In Lower Bay and in the Surf Zone, summer time TC concentrations are near average for the duration of the 32-year study period. The site-specific nature of the interannual trends in the within-Bay summer coliform data suggests that local features of the wetland and surrounding landscape (e.g., changes in land-use and civil infrastructure in the local subdrainage, tidal mixing rates, etc.) dominate long-period changes in dry weather water quality in the Bay.

Discussion

FIGURE 6. Frequency domain analysis of log mean TC data measured in Upper Bay, Lower Bay, Western Bay, and Surf Zone. different land-use, and land-use changes, in the respective watersheds (see earlier description of Table 1). Cumulative residuals calculated from the TC data reveal different long-period trends for winter (JFM) and summer (JJA) periods (see bottom two panels, Figure 8). For data collected during JFM, long-period trends in the TC concentrations are very similar at all within-Bay stations (compare red, green, and black curves, third panel, Figure 8). Furthermore, the within-Bay winter TC trends are similar to the rainfall and runoff trends described above, particularly during the first half of the study period from 1969 to 1985. The winter TC trends in the surfzone are relatively featureless over the 32-year study (pink curve in third panel). For TC data collected during JJA, long-period changes are site specific, and do not follow the patterns described above for rainfall and runoff. In Western Bay, JJA TC concentrations are below average until 1983, and then above or near average thereafter (green curve in the fourth panel, Figure 8). In Upper Bay, JJA TC concentrations are above average until 1987, and then below 9078

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 23, 2005

The results presented in this paper indicate that fecal indicator bacteria concentrations in the Bay are strongly forced by stormwater runoff, over time scales ranging from months to decades. This conclusion is based on the following evidence: (1) the 32-year log-mean TC concentration is highest in Upper Bay where the largest volume of stormwater runoffsfrom SDC and SAD channelssdischarges to the Bay; (2) TC concentrations within the Bay are higher than average during JFM when most rain falls in southern California; (3) TC concentrations within the Bay are significantly correlated with rainfall; (4) TC concentrations within the Bay exhibit an annual return period, consistent with a seasonal source for these contaminants; (5) 69% of the variability in the demeaned and normalized within-Bay TC data is apparently caused by rainfall and stormwater runoff; (6) winter time TC concentrations within the Bay exhibit the same interannual variability as rainfall, particularly during the first half of the study period, from 1969 through 1985. Once in the Bay, the fate and transport of fecal indicator bacteria is influenced by tidal mixing, sunlight-induced dieoff, and possibly other nonconservative processes that act to increase (e.g., regrowth in the sediments) or reduce bacteria (e.g., sedimentation) concentrations in the wetland (39,40). Winter time TC concentrations are consistently highest in Upper Bay, which is likely due to this region’s combination of high coliform loading rate (from SDC and SAD channel) and long tidal flushing times. This hypothesissthat the combination of high loading rates and long flushing times leads to water quality impairmentswould also explain why Western Bay is less contaminated than Upper Bay, despite the fact that both regions have similar flushing times. A simple scaling analysis presented in the supporting information for this paper predicts that the loading of coliform from storm runoff should be 550 times greater in Upper Bay, compared to Western Bay, consistent with the observed TC concentrations in these two regions of the Bay. Studies are currently underway to determine if observed concentrations in the different regions of the Bay can be predicted using simple box models from known loading rates, flushing time scales, and bacterial die-off rates. In interpreting the flushing times reported in Figure 4, several caveats should be kept in mind. First, the bathymetry data used as input to the hydrodynamic model were pieced together from different data sets collected over a 25-year period, from 1976 to 2001. Over that period of time, the bathymetry of the Bay was likely quite dynamic due to ongoing sediment loading from the surrounding watersheds (21), together with periodic modification of the bottom topography by dredging. The highly dynamic and incompletely documented nature of the Bay’s bathymetry raises questions about both the accuracy of the flushing time calculations and the degree to which flushing time scales for a particular region of the Bay would remain constant over the 32-year time frame considered in this article. Second, the numerical method employed here is likely to yield an upper bound on flushing times, because it estimates the time it would take a local region of the Bay to clean up under the assumption that the entire Bay was initially contaminated (see Materials and Methods section). As will be described in

FIGURE 7. Empirical orthogonal function analysis of log transformed TC data. The top panel displays the raw data. Mode 1 and 2 (gray solid curves) represents temporal Eigenvector 1 and 2. Rainfall is shown in the bottom panel (blue bars). a future study by our group, depending on the approach adopted, significantly smaller flushing times (e.g., days instead of months in Upper Bay) can also be obtained (see ref 35 for a good review of flushing time scale calculations). Taken together, these considerations suggest that the absolute magnitude of the flushing time scales plotted in Figure 4 are subject to significant uncertainty, although, on the whole, the within-Bay trends illustrated are likely to be accurate. It is noteworthy that both the 32-year TC log-mean and JFM TC anomaly decline approaching the Bay’s ocean outlet (see Figure 3). This last result suggests that the Bay acts as a natural treatment system for stormwater runoff from the surrounding urban landscape, by tidal dilution, die-off, filtration, sedimentation, or some combination thereof. In an earlier study of a much smaller tidal saltwater wetland called Talbert Marsh (0.1 km2 for Talbert Marsh compared to 8 km2 for Newport Bay), Grant et al. (14) concluded that the Talbert marsh is a net source of fecal indicator bacteria to the surrounding surf zone, most likely as a result of bacterial regrowth in the sediment and/or bird droppings deposited on the tidal mudflats from the relatively large number of gulls that visit the marsh on a daily basis. The very different conclusion offered here for Newport Baysthat it reduces rather than increases bacterial concentrationssis likely the result of the very different residence times of water in these two systems. The residence time in Talbert Marsh (ca., 20

min.) is less than the duration of a single flood or ebb tide, implying that any contaminants mobilized into the water column (e.g., from sediment resuspension) (41) will quickly flush into the ocean over the course of a single ebb tide. In contrast, the relatively long residence times in the upper reaches of Newport Bay (ca. 30 days according to the flushing time scales indicated in Figure 4) implies that contaminants entering this region of the Bay have much more time to undergo nonconservative transformationsssuch as die-off, filtration, and sedimentationsbefore being discharged to the ocean. Several lines of evidence indicate that human fecal contamination is probably not the sole source of TC pollution in stormwater runoff flowing into the Bay. First, TC concentrations in the SDC creek exhibit a weak positive correlation with volumetric discharge, implying that the concentrations of these fecal indicator bacteria increase with increasing flow. Given that there are a finite number of sources of human sewage in the watershed, markers for human sewage should dilute with increasing stream discharge, contrary to observations. Second, the peak loading rate of FC in SDCscalculated by taking the product of volumetric discharge and FC concentrationsis ca. 2 × 1014 MPN/day. If all FC in SDC originated from human sewage, this peak-loading rate would correspond to the combined fecal excretions of 100 000 people. This sewage-loading rate VOL. 39, NO. 23, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9079

FIGURE 8. Cumulative residuals of log transformed TC monthly means (seasonally separated into JFM and JJA, bottom two panels), rainfall (top panel), and volumetric discharge from SDC and SAD. is unrealistically high, given that the storm and sanitary sewer systems are separate in this watershed and hence combined sewer overflows are not a source of fecal indicator bacteria in SDC. For this last calculation, a human excretion rate of 2 × 109 FC/human/day was used (42). Other potential sources of coliform in the SDC and SAD watersheds, besides human sewage, include fecal matter deposited on the watershed by birds and (nonhuman) mammals, and regrowth of the bacteria in the water column and sediments (39, 40). The idea that not all coliform bacteria in this system originate from sources of human sewage may explain the variable correlation between TC and FC by region of the Bay (see Figure 2). In particular, if the sources of coliform were different in different regions of the Bay, the distribution of total and fecal coliform might vary by region as well. Finally, from the data presented in this paper we can provide some insight into the relative importance of climate change, on one hand, and local scale factors that might be under the control of coastal managers, on the other hand. Winter TC concentrations appear to be forced primarily by global processes (e.g., ENSO events) that influence the 9080

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 23, 2005

amount of rain that falls in southern California. This conclusion is based on the following lines of evidence. First, cumulative residual analysis reveals that winter (JFM) TC concentrations increase and decrease in unison across all within-Bay station categories (i.e., Upper Bay, Lower Bay, and Western Bay), and these water quality trends roughly parallel interannual trends in rainfall and stormwater runoff. Second, Mode 1 calculated from the EOF analysis is strongly correlated with rainfall, and the spatial eigenvalues associated with Mode 1 (see Table S3) indicate that this mode applies more-or-less equally to all within-Bay station categories. Third, winter time TC concentrations in SDC and in Upper Baysthe region of the Bay directly impacted by discharges from SDCsappear to be relatively insensitive to the rather dramatic shift away from agricultural land-use that occurred in the watershed over the 32-year study period. Interannual trends in the summer TC concentrations, on the other hand, are highly location dependent, suggesting that management actionsse.g., in the volume of dry-weather runoff discharged to different regions of the Baysmight have a significant impact on local water quality. The local nature

of summer time water quality in the Bay is evidenced in several ways. First, cumulative residual analysis reveals that summer (JJA) TC concentrations at the different within-Bay station categories (i.e., Upper Bay, Lower Bay, and Western Bay) exhibit very different interannual trends, and these trends do not parallel interannual trends in rainfall and stormwater runoff. Second, Mode 2 calculated from the EOF analysis is inversely correlated with rainfall, particularly after 1985, and the spatial eigenvalues associated with Mode 2 (see Table S3) indicate that this mode is more pronounced in some regions of the Bay (e.g., Western Bay) and much less pronounced in other regions (e.g., Upper Bay).

Acknowledgments We acknowledge the feedback and guidance provided by Clinton Winant, Lisa Levin, Richard Ambrose, Brett Sanders, and Robert Stein. Special thanks to Monica Mazur of the Orange County Health Care Agency for making available the historical data utilized in this study and for effectively addressing our many questions. We also thank Lane Waldner and Bryan Pastor of the Orange County Resources and Development Management Department for providing stream gauge and rainfall data. We gratefully acknowledge the many people involved in compiling the historical database utilized in this paper. Special thanks to two anonymous reviewers for their useful suggestions.

Supporting Information Available A brief scaling analysis, land use maps, additional figures, and additional tables. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Larson E. J. Coastal wetlands- Emergent marshes, Californias living marine resources: A status report; California Department of Fish and Game: Sacramento, CA, 2001; pp 483-486. (2) Southern California Wetlands Recovery Project Information Station; California Coastal Conservancy: Oakland, CA; www.wrpinfo.scc.ca.gov/index.html. (3) Levin, L. A.; D. F.; Boesch, A.; Covich, C.; Dahm, C.; Erseus, C.; Ewel, C. K.; Kneib, T. R.; Moldenke, A.; Palmer, A. M.; Snelgrove, P.; Strayer, D.; Weslawski, J. M. The function of marine critical transition zones and the importance of sediment biodiversity. Ecosystems 2001, 4, 430-451. (4) Smith, C. R.; Austen, M. C.; Boucher, G.; Heip, C.; Hutchings, P. A.; King, G. M.; Koike, I.; Lambshead, D. J.; Snelgrove, P. Global change and biodiversity linkages across the sediment water interface. Bioscience 2000, 50 (12), 1108-1120. (5) Cook, E. R.; Woodhouse, C. A.; Eakin, C. M.; Meko, D. M.; Stahle, D. W. Long-term Aridity Changes in the Western United States. Science 2004, 306, 1015-1018. (6) Pierce, J. L.; Meyer, G. A.; Jull, A. J. T. Fire Induced erosion and millennial-scale climate change in northern Ponderosa pine forests. Nature 2004, 432, 87-91. (7) Hill, V. R.; Sobsey, M. D. Removal of Salmonella and microbial indicators in constructed wetlands treating swine wastewater. Water Sci. Technol. 2001, 44, 215-222. (8) Gonzalez, J. M.; Ansola, G.; Luis, E. Experimental results on constructed wetland pilot system. Water Sci. Technol. 2001, 44, 387-392. (9) Pekins, J.; Hunter, C. Removal of enteric bacteria in a surface flow constructed wetland in Yorkshire, England. Water Res. 2000, 34 (6), 1941-1947. (10) Quinonez-Diaz, M.; Karpiscak, M. M.; Ellman, E. D.; Gerba, C. P. Removal of pathogenic and indicator microorganisms by a constructed wetland receiving untreated domestic wastewater. J. Environ. Sci. Health, Part A: Toxic/Hazard. Subst. Environ. Eng. 2001, A36, 1311-1320. (11) Thurston, J. A.; Gerba, C. P.; Foster, K. E.; Karpiscak, M. M. Fate of indicator microorganisms, Giardia and Crytosporidium in subsurface flow constructed wetlands. Water Res. 2001, 35 (6), 1547-1551 (12) Stenstrom, T. A.; Carlander A. Occurrence and die-off of indicator organisms in the sediment in two constructed wetlands. Water Sci. Technol. 2001, 44 (11-12), 223-230.

(13) Gerba, C. P.; Thurston, J. A.; Falabi, J. A.; Watt, P. M.; Karpiscak, M. M. Optimization of artificial wetland design for removal of indicator microorganisms and pathogenic protozoa. Water Sci. Technol. 1999, 40, 363-368. (14) Grant, S. B.; Sanders, B. F.; Boehm, A. B.; Redman, J. A.; Kim, J. H.; Mrse, R. D.; Chu, A. K.; Gouldin, M.; McGee, C. D.; Gardiner, N. A.; Jones, B. H.; Svejkovsky, J.; Leipzig, G. V.; Brown, A. Generation of Enterococci bacteria in coastal saltwater marsh and its impact on surf zone water quality. Environ. Sci. Technol. 2001, 35 (12), 2407-2416. (15) U.S. Environmental Protection Agency, Region 9. Total maximum daily loads for toxic pollutants San Diego creek and Newport bay, California; Public Review draft, 2002. (16) Jeong, Y.; Grant, S. B.; Ritter, S.; Pednekar, A.; Candelaria, L.; Winant, C. Identifying Pollutant sources in Tidally mixed systems: Case Study of fecal Indicator bacteria in Newport bay, southern California. Environ. Sci. Technol. 2005, 39, 9083-9093. (17) Teske, P. R.; Wooldridge, T. A comparison of the macrobenthic faunas of permanently open and temporarily open/closed South African estuaries. Hydrobiologica 2001, 464, 227-243. (18) Allanson, B., Baird, D., Eds. Estuaries of South Africa; Cambridge University Press: Cambridge, 1999. (19) Arthington, A. H.; Hadwen, W. L. Effects of catchment management on the food web ecology of intermittently closed and open lakes and lagoons (ICOLLs) in Coastal New South Wales, Australia; Abstract NABS meeting, Athens, Georgia, 2003. (20) Masters, P. M.; Inman, D. L. Transport and fate of organochlorines discharged to the salt marsh at upper Newport bay, California, USA. Environ. Toxicol. Chem. 2000, 19 (8), 20762084. (21) Trimble, S. W. The contribution of stream channel erosion to sediment yield from an urbanizing watershed. Science 1997, 278, 1442-1444 (22) 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 and shorter period variability of surf zone water quality at Huntington Beach, California. Environ. Sci. Technol. 2002, 36 (18), 38853892. (23) Mays, L. W. Water Resources and Engineering; John Wiley and Sons: New York, 2001. (24) U.S. Army Corps of Engineers Waterways Experiment Station. Users Guide to RMA2 WES Version 4.3; U.S. Army Corps of Engineers: May 19, 1997. (25) U.S. Army Corps of Engineers Waterways Experiment Station. Users Guide to RMA4 WES Version 4.5; U.S. Army Corps of Engineers: August 1, 2000. (26) U.S. Army Corps of Engineers, Los Angeles District Navigation Home Page. http//www.spl.usace.army.mil/co/navigation/ Newport/newport1.html (Accessed 2002). (27) City of Newport Beach (California), Public Works Department. Drawing title: Newport Island Channel Dredging - Rivo Alto and Portion of North Channel; Drawing No. H-5071-S; September 1976. (28) Emery, W. J.; Thomson, R. E. Data Analysis Methods in Physical Oceanography; Elsevier Sciences B. V.: Amsterdam, 2000. (29) Winant, C. D.; Inman, D. L.; Nordstorm, C. E. Description of seasonal beach changes using empirical eigenfunctions. J. Geophys. Res. 1975, 80 (15), 1979-1986. (30) Kim, J. H.; Grant, S. B.; McGee, C. D.; Sanders, B. F.; Largier, J. L. Locating sources of surf zone pollution: A mass budget analysis of fecal indicator bacteria at Huntington Beach, California. Environ. Sci. Technol. 2004, 38 (9), 2626-2636. (31) Reeves, R. L.; Grant, S. B.; Mrse, R. D.; Copil Oancea, C. M.; Sanders, B. F.; Boehm, A. B. Scaling and management of fecal indicator bacteria in runoff from a coastal urban water shed in southern California. Environ. Sci. Technol. 2004, 38 (9), 26372648. (32) Dwight, R. H.; Baker, D. B.; Semenza, J. C.; Olson, B. H. Health effects associated with recreational coastal water use: Urban versus rural California. Am. J. Public Health 2004, 94 (4), 565567 (33) Dwight, R. H.; Semenza, J. C.; Baker, D. B.; Olson, B. H. Association of urban runoff with coastal water quality in Orange County, California. Water Environ. Res. 2002, 74 (1), 82-90. (34) Water Control Manual, Prado Dam & Reservoir, Santa Ana River, California; U.S. Army Corps of Engineers: September, 1994. VOL. 39, NO. 23, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9081

(35) Monsen, N. E.; Cloern, J. E.; Lucas, L. V.; Monismith, S. G. A comment on the use of flushing time, residence time, and age as transport time scales. Limnol. Oceanogr. 2002, 47 (5), 15431553. (36) Fischer, H. B.; List J. E.; Koh, R. C. Y.; Imberger, J.; Brooks, N. H. Mixing in Inland and Coastal Waters; Academic Press: San Diego, CA, 1979. (37) Hurst, H. E. Long-term storage capacity of reservoirs. Am. Soc. Civil Eng. Trans. 1951, 116, 770-799. (38) Inman, D. L.; Jenkins, S. A. Climate change and the episodicity of sediment flux of small California rivers. J. Geol. 1999, 107, 251-270. (39) Solo-Gabriele, H. M.; Wolfert, M. A.; Desmarais, T. R.; Palmer, C. J. Sources of Escherichia coli in a coastal subtropical environment. Appl. Environ. Microbiol. 2000, 66 (1), 230-237.

9082

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 23, 2005

(40) Steets, B. M.; Holden, P. A. A mechanistic model of runoffassociated fecal coliform fate and transport through a coastal lagoon. Water Res. 2003, 37, 589-608. (41) Sanders, B. F.; Arega, F.; Sutula, M. Loading of enteric bacteria to surface waters of an intertidal wetland by urban runoff, bird feces and resuspended sediments. Water Res. Submitted for publication. (42) Tchobanoglous, G.; Schroeder, E. D. Water Quality; AddisonWesley Publishing Company: Reading, MA, 1987.

Received for review March 10, 2005. Revised manuscript received July 27, 2005. Accepted August 4, 2005. ES0504789