Environ. Sci. Technol. 2008, 42, 3609–3614
Treatment of Dry Weather Urban Runoff in Tidal Saltwater Marshes: A Longitudinal Study of the Talbert Marsh in Southern California YOUNGSUL JEONG, BRETT F. SANDERS, KAREN MCLAUGHLIN, AND STANLEY B. GRANT* Interdisciplinary Environmental Engineering Program, Henry Samueli School of Engineering University of California, Irvine, California 92697
Received October 23, 2007. Revised manuscript received January 28, 2008. Accepted February 6, 2008.
The scientific literature presents conflicting assessments of whether tidal saltwater wetlands reduce or increase fecal indicator bacteria (FIB) impairment of marine bathing waters. In this paper we describe the use of a two end-member salinitymixing model to calculate FIB treatment efficiencies for the Talbert Marsh, a tidal saltwater wetland in Orange County, California. The mixing model utilized FIB and salinity measurements (n ) 10 716) collected during a three-year longitudinal study of the Talbert Marsh. Over the course of the study the marsh received progressively less dry weather surface water runoff from the surrounding urban landscape due to the implementation of a runoff interception and treatment program. As the volume of dry-weather runoff entering the marsh declined, the Talbert Marsh more efficiently removed one FIB group (total coliform) and became a significantly smaller source of two other FIB groups (Escherichia coli and enterococci bacteria). Hence, there may be a maximum volume of dry weather urban runoff (in this case 1 (or < -1) the concentration. Consequently, when N measured FIB concentration is significantly greater (or lower) than the highest (or lowest) 10% of predicted FIB concentrations. Two-End-Member Salinity Mixing Model. Hourly predictions for [FIB]pred and σ pred were calculated from hourly salinity measurements at the marsh outlet using the twoend member salinity mixing model for tidal saltwater wetlands described in McLaughlin et al. (35), but modified here to account for the variability of salinity and FIB concentrations in the runoff and ocean end-members. In this model, the predicted concentration of FIB at the marsh outlet, [FIB]pred, is calculated from eq 2, where f is the fraction of runoff water present at the outlet (as determined by the measured salinity, see below), and [FIB]runoff and [FIB]ocean represent the concentration of FIB in the urban runoff and ocean end-members: [FIB]pred ) f × [FIB]runoff + (1 - f) × [FIB]ocean
(2)
During dry weather periods, FIB concentrations in ocean water are generally below our lower-detection limit of 10 MPN/100 mL (unpublished data), and hence, for these calculations we set [FIB]ocean ) 0. The fraction f of outlet water that is urban runoff is calculated from the salinity measured at the outlet [S]outlet and the slope m and intercept b of the linear mixing line: f ) m × [S]outlet + b
(3)
The mixing line parameters are estimated by applying the slope and intercept equations to the (f, [S]) pairs {(1, [S]runoff),(0, [S]ocean)}: m ) -1 ⁄ ([S]ocean - [S]runoff) b ) (1 - [S]runoff ⁄ [S]ocean)
-1
(4a) (4b)
For each hourly measurement of salinity at the marsh outlet, the salinity-mixing model was used to generate a probability distribution of predicted FIB concentrations. This was accomplished in four steps as illustrated in SI Figure S3. In the first step, cumulative density functions (CDFs) were created from all measurements of salinity in samples of runoff and ocean end-members. These CDFs were randomly sampled 1000 times to generate CDFs for the slope and intercept parameters m and b (eq 4a and 4b). In the second step, each hourly measurement of salinity at the marsh outlet was converted into a CDF for the fraction f of runoff at the marsh outlet by randomly sampling 1000 times CDFs for m and b (eq 3). In the third step, a CDF was constructed from all measurements of FIB in samples of runoff, and CDFs for f and [FIB]runoff were randomly sampled 1000 times to generate a CDF for the concentration of FIB at the marsh outlet (eq 2). From the latter CDF, 10th, 50th, and 90th percentiles of
FIGURE 1. Cumulative distribution functions of salinity and fecal indicator bacteria concentrations measured at the Talbert Marsh outlet (left column) and in nuisance runoff collected from the forebays (right column). California marine bathing water standards (single-sample standards and geometric-mean standard) are marked with dashed lines and dotted lines, respectively. predicted FIB concentrations were extracted, and used to calculate the wetland treatment efficiency parameter (eq 1a). Steps 2 through 4 were repeated for every hourly measurement of salinity at the marsh outlet. In carrying out the simulations above, we found that results were unchanged if CDFs were sampled 1000 or 2000 times; hence, the 1000 realizations used here appears adequate to characterize the variability of all dependent variables.
Results and Discussion Forebay and Outlet Measurements. Figure 1 displays CDFs of salinity and FIB concentrations measured at the Talbert Marsh outlet (left column of panels) and in urban runoff collected from the forebays (right column of panels); time series plots of these same data can be found in SI Figure S4-S6. The different curves in Figure 1 represent the three different sampling events (TM1999, TM2000, and TM2001) that collectively makeup the longitudinal study. Comparing the left and right columns in Figure 1, it is evident that samples collected from the Talbert Marsh outlet had much higher salinity and lower FIB concentrations, compared to samples of dry weather runoff collected from the forebays. A large percentage of samples collected from the forebays had FIB concentrations above California’s marine bathing water standards (compare colored curves to vertical dotted and dashed lines). In contrast, most samples collected from the Talbert Marsh outlet had concentrations below marine bathing water standards. At the marsh outlet, ENT most frequently exceeded the California single-sample and geometric mean recreational marine bathing water standards (104 and 35 MPN/100 mL, respectively). Salinity measured at the Talbert Marsh outlet became progressively less brackish (i.e., more like ocean water) over the three-year longitudinal study. The fraction of water samples collected from the marsh outlet with low salinity VOL. 42, NO. 10, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3611
FIGURE 2. Monte Carlo simulation results of marsh treatment for fecal pollution. A. Time series of the measured and the predicted ˆ value for three Talbert Marsh studies. total coliform (TC) concentration at the marsh outlet. B. CDF plots of N (i.e., salinity TM2000 > TM1999) (Figure 1). Reeves et al. (5) suggest that changing environmental conditions may account for the increasing concentration of FIB in the runoff. Ambient air temperature increased progressively over the three studies (14 °C during TM1999, 18 °C during TM2000, and 22 °C during TM2001), as did the total rain that fell 6 months prior to each study (23 mm during TM1999, 154 mm during TM2000, and 277 mm during TM2001) (5). Warmer temperatures have been implicated in the survival and regrowth of coliform in nonhost environments, notably in sediments and soils (36). Of the three types of FIB measured, TC was the only group that had comparable longitudinal trends in the forebays and at the marsh outlet. Wetland Treatment Efficiency. The ability of the marsh to treat FIB cannot be assessed solely from the concentration of FIB at the marsh outlet because, as noted above, the concentration of FIB in the forebays increased progressively over the three-year longitudinal study, while the flow of runoff into the marsh decreased. Highly variable mixing of ocean water with runoff in the marsh (caused by constantly changing tidal conditions) further complicates interpretation of concentration measurements at the marsh outlet (see Figures S4-S6 in SI). Consequently, we tested the possibility of using salinity measurements to estimate the contribution of dry weather 3612
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 10, 2008
runoff to FIB measured at the outlet of the marsh. Figure 2A compares the measured and salinity-based predictions of TC concentrations at the marsh outlet for the TM2000 study. Similar results for different FIB groups, and different study years, are presented in SI Figures S7-S9. For TM2000, TC concentrations measured at the marsh outlet are lower, often by orders of magnitude, than TC concentrations predicted by the two-end-member salinity-mixing model (Figure 2A). This observation is consistent with the idea that, during the TM2000 field study, TC was removed from the water column by nonconservative processes (e.g., die-off, sedimentation, filtration) within the marsh. The removal of TC in the marsh is likely to be even greater than is evident in Figure 2A, because the Colilert-18 test used in this study can have a high falsepositive rate for TC in marine waters (see Field Site and Study Design section). The difference between predicted and measured FIB concentrations at the marsh outlet can be used to calculate ˆ (see Field what we called a wetland treatment parameter, N Site and Study Design section). ˆ are presented for the three Probability distributions for N studies (TM1999, TM2000, and TM2001) and for the three ˆ values become progressively less FIB groups in Figure 2B. N positive (or progressively more negative, depending on the ˆ FIB group) over the longitudinal study. In particular, N values for TC went from weakly negative during TM1999 ˆ value -0.2) to strongly negative during TM2001 (median N ˆ value -0.7). N ˆ values for EC went from strongly (median N ˆ ) 0.5) to weakly positive positive during TM1999 (median N ˆ ) 0.2) to near zero during during TM2000 (median N ˆ ) 0). Similarly, N ˆ values for ENT went TM2001 (median N ˆ ) 4.5), to from strongly positive during TM1999 (median N ˆ ) 0.1), to near weakly positive during TM2000 (median N ˆ ) 0). While FIB concentrazero during TM2001 (median N tions in the marsh and forebay runoff exhibit complex and conflicting longitudinal trends (see above), when these data are recast in terms of a wetland treatment efficiency, we find ˆ steadily declines over the that for all three FIB groups N three year longitudinal study. ˆ progressively declined over the three-year The fact that N longitudinal study can be interpreted to mean that the marsh was more effective at removing (or at least not contributing) FIB from the water column, as progressively more dry weather
runoff was captured and diverted to the sanitary sewer system. The diversion of dry weather runoff flowing into the wetland could have improved the wetland’s performance as a FIB treatment system, by (1) reducing the load of FIB coming into the marsh environment, perhaps bringing it to within the treatment capacity of this low-residence-time system; (2) reducing the flow of nutrients into the marsh that might serve as a substrate for regrowth (36); (3) increasing the salinity of the marsh, thereby making it a less hospitable environment for the persistence and regrowth of environmentally adapted FIB strains; and/or (4) increasing the salinity of the marsh, thereby making the marsh a less desirable destination for birds to congregate and forage, thereby reducing the input of FIB on the mudflats from bird droppings (12). In the dry weather study of the Talbert Marsh reported by Grant et al. (12), a reduction in runoff flowing into the marsh from the surrounding forebays did not improve water quality in the marsh (or the adjacent surf zone) over an 8 day trial period. Hence, if the improved FIB treatment efficiency reported here is indeed related to the coincident reduction in dry weather runoff flowing into the marsh, the impacts of runoff on marsh water quality are felt over long (months to years) not short (days to weeks) time periods. Although the precise mechanism by which reduction in dry weather runoff volume improved the marsh’s treatment efficiency cannot be ascertained from the present study, the data and analysis presented here support the hypothesis that tidal saltwater marshes in southern California, like the Talbert Marsh, may have a maximum carrying capacity for dry weather runoff from the surrounding landscape, as has been suggested by others (2). Under this hypothesis, when the volume of dry weather runoff flowing into a marsh exceeds its carrying capacity, the result is poor treatment efficiency, and amplification of the FIB from a variety of within-marsh sources (e.g., regrowth in sediments, and tidal washing of mudflats contaminated with bird droppings). Because all tidal saltwater marshes in southern California have been modified to some degree, and most are engineered systems, it is instructive to compare the volume of dry weather runoff flowing into the Talbert Marsh in the last year of the longitudinal study (when treatment efficiencies were at their best) with the marsh’s tidal prism—the volume of water in the marsh that is exchanged over a tidal cycle. For the three studies reported here, the daily average tidal prisms were similar over the three sampling events (2.30 × 105 m3/day, 2.57 × 105 m3/day, and 2.17 × 105 m3/day for the TM1999, TM2000, and TM2001 studies, respectively). Over those same three studies, the daily average flow of runoff into the marsh from the surrounding urban watershed is estimated to be 6100 m3/day (during TM1999), 3300 m3/day(during TM2000), and 1700 m3/day (during TM2001), respectively (37). Thus, the marsh’s treatment efficiency improved markedly when the ratio of urban runoff volume to tidal prism volume fell below 0.0078 (
