Combined Sewer Overflows to Surface Waters Detected by the

(Agroscope), CH-8820 Wädenswil, Switzerland. Continuous progress in wastewater treatment technology and the growing number of households connected to...
0 downloads 0 Views 441KB Size
Environ. Sci. Technol. 2006, 40, 4096-4102

Combined Sewer Overflows to Surface Waters Detected by the Anthropogenic Marker Caffeine IGNAZ J. BUERGE,* THOMAS POIGER, MARKUS D. MU ¨ LLER, AND HANS-RUDOLF BUSER Plant Protection Chemistry, Swiss Federal Research Station (Agroscope), CH-8820 Wa¨denswil, Switzerland

Continuous progress in wastewater treatment technology and the growing number of households connected to wastewater treatment plants (WWTPs) have generally resulted in decreased environmental loading of many pollutants. Nonetheless, further reduction of pollutant inputs is required to improve the quality of surface waters in densely populated areas. In this context, the relative contribution of combined sewer overflows as sources of wastewater-derived contaminants has attracted more and more attention, but the quantitative importance of these overflows has barely been investigated. In this study, caffeine was successfully used as a chemical marker to estimate the fraction of sewer overflows in the catchment area of lake Greifensee, Switzerland. Caffeine is a ubiquitous compound in raw, domestic wastewater with typical per capita loads of ≈16 mg person-1 d-1. In WWTPs of the Greifensee region, caffeine is largely eliminated (>99%), resulting in much smaller loads of e0.15 mg person-1 d-1 in treated wastewater. However, in receiving streams as in the inflows to Greifensee, caffeine loads (0.1-1.6 mg person-1 d-1) were higher than those in WWTP effluents, indicating additional sources. As the loads in the streams correlated with precipitation during sampling, it was concluded that combined sewer overflows were the most likely source of caffeine. Using a mass balance approach, it was possible to determine the fraction of wastewater (in dry weather equivalents) discharged untreated to the receiving streams (up to 10%, annual mean, ≈2-3%). The concept of caffeine as a marker for combined sewer overflows was then applied to estimate phosphorus inputs to Greifensee with untreated and treated wastewater (≈1.5 and 2.0 t P y-1, respectively), which corresponded well with P inputs determined in a separate study based on hydraulic considerations. For compounds with high elimination in WWTPs such as phosphorus (96-98% in the Greifensee area), inputs from combined sewer overflows are thus of similar magnitude as inputs from treated wastewater. The study demonstrated that caffeine is a suitable marker for untreated wastewater (from combined sewer overflows, direct discharges, etc.), but its sensitivity depends on regional conditions and decreases with decreasing elimination efficiency in WWTPs.

* Corresponding author phone: ++41 44 783 63 83; fax: ++41 44 780 63 41; e-mail: [email protected]. 4096

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 13, 2006

Introduction With the construction of wastewater treatment plants in the 1970s, the ban of phosphates in detergents, and the ecological measures taken in agriculture, the concentration of phosphorus, the growth-limiting nutrient for primary production, and thus the degree of eutrophication, has been decreasing in most lakes of the Swiss Midland region (1). For example, in Greifensee, total phosphorus decreased by almost 1 order of magnitude from 0.50 in 1969 to 0.07 mg L-1 in 2003 (2). Nonetheless, during summer, respiration in the deeper water layers of the lake is still so high that oxygen levels become critically low (2), with adverse effects on fish and other organisms (3). Authorities thus aim at an even lower target value of 0.025 mg Ptotal L-1. To take appropriate measures, knowledge on the magnitude of phosphorus inputs from different sources is required, such as from discharge of wastewater, agricultural runoff, or geogenic sources. Domestic wastewater is certainly a major source of phosphorus. In the densely populated catchment area of Greifensee (population, 108000), virtually all households are connected to WWTPs (Figure 1, (4)). However, in combined sewer systems, the capacities of WWTPs and retention basins may be exceeded during or after rain events with the consequence that untreated wastewater is discharged to the receiving waters (combined sewer overflows) (5). Whereas phosphorus inputs from treated wastewater can adequately be estimated from measurements in WWTP effluents, the contribution of combined sewer overflows is hardly known. Quantification of these wastewater discharges is difficult as there are usually numerous retention basins with overflows situated in different locations throughout the catchment area (e.g., 32 in the region of Greifensee). Alternatively, chemical markers may be helpful for detecting wastewater inputs from combined sewer overflows (6-8). Such a marker should indicate untreated wastewater specifically, which is, in principle, the case for compounds that are largely eliminated in WWTPs. Caffeine has been shown to be a suitable marker for domestic wastewater and is degraded in Swiss WWTPs with efficiencies of typically >99% (9). Caffeine concentrations found in Swiss Midland lakes (up to 164 ng L-1) could not solely be explained by the discharge of treated wastewater; hence, inputs of untreated wastewater were assumed to be responsible for the comparatively high concentrations. In a more detailed investigation on caffeine in Zu ¨ richsee, input loads seemed to be linked to rain events and it was concluded that caffeine associated with untreated wastewater had been discharged to the lake or its inflows because the capacities of WWTPs had been exceeded (9). As there were far too many WWTPs and retention basins located around Zu ¨richsee, these input loads could not directly be determined, but were estimated indirectly by fitting regularly measured vertical concentration profiles in the lake with a simulation program (AQUASIM, (10)), considering all relevant elimination processes and the hydrology of the lake. A factor of uncertainty, thereby, was the rate of biodegradation of caffeine in the lake, which was simply estimated from batch incubation experiments under laboratory conditions. The aim of the present study was (i) to generally assess the suitability of caffeine as a marker specific for untreated domestic wastewater, (ii) to quantify the importance of combined sewer overflows in the catchment area of Greifensee by analyses of caffeine in the relevant inflows to the lake under different weather conditions, and (iii) to estimate the rate of biodegradation of caffeine in Greifensee with a 10.1021/es052553l CCC: $33.50

 2006 American Chemical Society Published on Web 05/20/2006

FIGURE 1. Map of the catchment area of lake Greifensee, Switzerland, showing the location of WWTPs (triangles, numbers indicate population served by the plants) and the sampling sites (arrows) in the two major inflows (Mo1 nchaltdorfer Aa and Ustermer Aa), which cover inputs from treated wastewater of ≈23200 and 42600 residents, respectively, and inputs from combined sewer overflows of ≈25000 and 58700 residents, respectively. Note that, in Mo1 nchaltdorf and Uster, WWTPs discharge downstream, whereas about half of the combined sewer overflows are discharged upstream of the sampling sites. mass balance approach. (iv) The concept of caffeine as a marker for combined sewer overflows was finally applied to estimate the relative magnitude of phosphorus inputs to Greifensee from treated and untreated wastewater.

Experimental Section Chemicals. Caffeine anhydrous (purity, >99%) was from Fluka, Buchs, Switzerland, and 13C3-labeled caffeine, used as an internal standard, was from Cambridge Isotope Laboratories, Cambridge, MA (courtesy of C. Schaffner, Eawag, Switzerland). Water Samples. Water samples for vertical concentration profiles of caffeine in Greifensee (depths, 1-30 m) were taken monthly between June and September 2002 at the location above the deepest point of the lake with a 10-L Niskin bottle. From two major inflows to Greifensee (Mo¨nchaltdorfer Aa and Ustermer Aa; average discharge, 1.1 and 1.7 m3 s-1, respectively (11), Figure 1), flow-proportional 24-h composite samples were taken between June 17 and July 22, 2002, and between August 26 and September 23, 2002. In the laboratory, consecutive 24-h samples were combined to flow-proportional 7-d composite samples, which were then analyzed for caffeine. All surface water samples were refrigerated and protected from light after sampling. Extraction was performed within a few days. On November 13, 2002, and August 12, 2003, the temporal course of the caffeine concentration was determined in wastewater from WWTP Wetzikon (population served, 19600). The installation operates with a mechanical, biological (activated sludge, with nitrification and denitrification), and chemical treatment (phosphate precipitation by iron salts, no chlorination), and subsequent sand filtration. Flowproportional samples were collected every 2 h, stored at ≈4

°C, and extracted within the next few hours. Influent samples were taken before (August sampling) or after the primary sedimentation basin (November, time delay, ≈1 h) and treated effluent samples after sand filtration (both sampling dates). Note that caffeine is not removed to a significant extent during primary clarification (9). Analytical Procedure. Caffeine was analyzed with gas chromatography-mass spectrometry (GC-MS) in selected ionmonitoring mode after addition of 13C3-caffeine as the internal standard, solid-phase extraction using a macroporous polystyrene divinylbenzene adsorbent, and a cleanup over silica. The procedure was described in detail in Buerge et al. (9).

Results and Discussion WWTP Effluents Are Not the Main Source of Caffeine in Receiving Waters. In the catchment area of Greifensee, most treated wastewater from WWTPs, as well as untreated wastewater from combined sewer overflows, is discharged to two major inflows to the lake (Mo¨nchaltdorfer Aa and Ustermer Aa) or to small tributaries of these streams (Figure 1). Only 2 of the 9 installations within the catchment area release wastewater directly to the lake (WWTP Uster and Maur). This geographical situation thus allows covering of a significant part of the wastewater inputs finally ending up in Greifensee by sampling water from the two streams shortly before their inflow to the lake (Figure 1). During several weeks between June and September 2002, the marker caffeine was analyzed in flow-proportional, 7-d composite samples of Mo¨nchaltdorfer Aa and Ustermer Aa. Concentrations ranged from 60 to 410 ng L-1 (Table 1), astonishingly, reached those found in effluents of WWTPs typical for the Greifensee region (30-360 ng L-1, dry and wet weather situations, Table 1). Considering that the percentage VOL. 40, NO. 13, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4097

TABLE 1. Concentrations and Loads of Caffeine in WWTPs, Lake Greifensee, and Its Two Major Inflows (Mo1 nchaltdorfer Aa, Ustermer Aa), Switzerland

WWTPs, influent

wastewatera

WWTPs, effluent wastewatera,b Mo¨ nchaltdorfer Aac Ustermer Aac Greifenseed

concentration [ng L-1]

load per capita [mg person-1d-1]

no. of samples

7000-73000e

10-23e

13

28-355e 62-410 103-400 36-114

15.8 ( 3.8e,f 0.02-0.15e 0.06 ( 0.03e,f 0.13-1.61 0.24-1.13 0.22-0.32g

11 9 9 28

a 24-h composite samples. b WWTPs with sludge age >5 d. c 7-d composite samples, June-September 2002. d Depths, 1-30 m, June-September 2002. e Buerge et al. (9). f Mean ( standard deviation. g Load in the outflow of Greifensee, river Glatt, estimated from the concentration in the lake in 1 m depth and the mean discharge between sampling.

of treated wastewater in the two streams is not exceptionally high (annual means: 13% in Mo¨nchaltdorfer Aa and 17% in Ustermer Aa (11, 12)), treated wastewater would be expected to be diluted by a factor of ≈6-8 in the streams, resulting in estimated concentrations of only ≈4-60 ng L-1. Comparison of predicted and measured concentrations thus suggested that treated wastewater was apparently not the main source of caffeine in these streams. Caffeine Indicates Discharge of Untreated Wastewater to Receiving Waters. Untreated domestic wastewater from combined sewer overflows and households not connected to the sewer system as well as wastewater from the beverage and food industry may be additional potential sources of caffeine. In the catchment area of Greifensee, however, essentially all households are connected to WWTPs (>99.6%, (13)) or discharge their wastewater to manure tanks (remote farmhouses), and industrial sources are not considered to be relevant. Therefore, untreated wastewater from sewer overflows was assumed to be the predominant additional source of caffeine. Concentrations of caffeine in untreated wastewater are, in fact, orders of magnitude higher (typical range, 7-70 µg L-1, Table 1) than in treated wastewater, reflecting the high elimination efficiencies of typically >99% in WWTPs. The difference also becomes evident when comparing corresponding loads. On a per capita basis, average loads in untreated and treated wastewater are Iuntreated ) 15.8 ( 3.8 mg person-1 d-1 and Itreated ) 0.06 ( 0.03 mg person-1 d-1, respectively (Table 1). The total loads of caffeine in the receiving rivers (Lriver [g d-1]) thus result from direct discharges of untreated wastewater (Luntreated) as well as from inputs of treated wastewater (Ltreated) and can be described with the following mass balance:

Lriver ) Luntreated + Ltreated

(1)

with

Lriver ) criver × Qriver ) P × Iriver Luntreated ) P × f × Iuntreated Ltreated ) P × (1 - f) × Itreated where criver is the concentration of caffeine measured in the river [g m-3] (Table 1), Qriver is the average discharge during sampling [m3 d-1] (11), P is the population upstream of the sampling site in the river [person] (according to Figure 1), Iriver is the per capita load of caffeine in the river (0.13-1.61 mg person-1 d-1, Table 1), which is calculated from criver, Qriver, and P, and f is the fraction of direct discharges of untreated wastewater [dimensionless], the only unknown parameter in these equations. Note that f refers to wastewater as dry weather equivalents. 4098

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 13, 2006

FIGURE 2. Loads of caffeine in 7-d composite samples (Lriver) from the two major inflows to lake Greifensee, (a) Mo1 nchaltdorfer Aa and (b) Ustermer Aa, plotted vs the mean precipitation during sampling (measured in Mo1 nchaltdorf). The second ordinate denotes the calculated fraction of combined sewer overflows (f). Error bars refer to f; for details see text. According to this mass balance approach, the fraction of untreated wastewater discharged to Mo¨nchaltdorfer Aa and Ustermer Aa was up to ≈10% during the observation period of 9 weeks in Summer 2002. Even a small percentage of untreated wastewater does considerably increase the load of caffeine in receiving waters; e.g., discharge of 1% of untreated wastewater delivers more than half, 10% of untreated wastewater > 90% of the total caffeine inputs. Input of Untreated Wastewater to Receiving Waters Correlates with Precipitation: Indication for Combined Sewer Overflows. Assuming that these inputs of untreated wastewater were linked to combined sewer overflows, their importance would have to increase during rainy weather. Caffeine loads in the streams Mo¨nchaltdorfer Aa and Ustermer Aa, in fact, showed a positive correlation with the mean precipitation in the region during sampling (Figure 2). Consequently, a positive correlation was also found between the calculated fraction of supposed sewer overflows and precipitation (second ordinate in Figure 2). For Mo¨nchaltdorfer Aa, the correlation was more or less linear (Figure 2a), suggesting that, with increasing rainfall, capacities of more and more retention basins were exceeded. The “hockey stick”-like trendline indicates, in a qualitative way, that combined sewer overflows become relevant only from a certain threshold of precipitation. For Ustermer Aa, data points were somewhat more scattered (Figure 2b). The data also suggested a small percentage of sewer overflows even in situations with virtually no rainfall, which is, of course, not reasonable. The likely explanation for the higher base load in this stream is that wastewater from two WWTPs first passes through Pfa¨ffikersee (Figure 1), which delivers a more or less constant input of caffeine to Ustermer Aa. Error bars for f in Figure 2 were determined by error propagation, considering the standard deviations for Itreated

FIGURE 4. Influence of wastewater throughput on elimination efficiency of caffeine in WWTPs with sludge ages >5 d (data from this study and from Buerge et al. (9)).

FIGURE 3. Temporal course of caffeine loads in influent and effluent wastewater of WWTP Wetzikon (population served, 19600) on (a) November 13, 2002 (throughput, 14925 m3 d-1; residence time in activated sludge basin, ≈10 h; water temperature, 13.6 °C) and (b) August 12, 2003 (6226 m3 d-1, 24 h, 21.2 °C). Note the different scaling factors for effluent loads. Elimination efficiencies were ≈99.0 and 99.95%, respectively. and Iuntreated listed in Table 1 and assuming a relative error of 10% for Iriver. The estimate of low fractions of combined sewer overflows has a higher relative error than that of high fractions of overflows due to the relatively high error of Itreated. The higher loads of caffeine in receiving waters during rainy weather are certainly a strong indication for combined sewer overflows. From the annual mean precipitation (3.07 mm d-1) and the tentative linear correlations shown in Figure 2, the annual mean fractions of sewer overflows to Mo¨nchaltdorfer Aa and Ustermer Aa were estimated at ≈1.9 and 2.5%, respectively. Negligible Influence of Wastewater Throughput on Elimination Efficiency of Caffeine in WWTPs. Elevated loads of caffeine in receiving waters during heavy rainfall may, in principle, also be the result of lower elimination efficiencies in WWTPs at higher throughput of (diluted) wastewater and thus shorter residence time in the plant. In WWTP Wetzikon, the temporal course of the caffeine loads was followed at low and high throughput in influent and effluent wastewater (flow-proportional 2-h composite samples). Caffeine loads in influent wastewater showed distinct maxima, however, at different times on the two sampling days (Figure 3). The time delay between input of caffeine to the sewer system (e.g., from discarded caffeinated beverages and renal excretion) and occurrence in WWTP influents depends on many factors such as the construction of the sewer system (distances, hydraulic gradients, etc.) and the actual throughput of wastewater. The mean residence time of wastewater in the sewer system is decreasing with increasing throughput. On November 13, 2002 (throughput, 14925 m3 d-1, rainy weather on the previous days), caffeine loads reached quite a clear (narrow) maximum at ≈1800-

2000 h, whereas on August 12, 2003 (6226 m3 d-1, dry weather period), the maximum was already observed at ≈1000-1200 h and was less pronounced. It was therefore concluded that the earlier maximum at low throughput had been from inputs to the sewer system on the previous day, provided that the pattern of caffeine consumption was similar in November and August. Effluent loads also showed distinct diurnal variations with maxima and minima at a time similar to the corresponding influent loads. The amplitude, however, was much higher in November than in August, which is reasonable for a highdischarge compared to a low-discharge situation, respectively. Removal rates of caffeine were high on both sampling days, but higher in August (99.95%), at low throughput and consequently high residence time in the activated sludge basin (≈24 h) and high water temperature (21.2 °C), than in November (99.0%, 10 h, 13.6 °C, respectively, Figure 3). A compilation of elimination efficiencies determined in other WWTPs representative for the Greifensee area also showed a slight negative trend with increasing throughput (Figure 4), but altogether, removal of caffeine was still g99%. Therefore, the higher loads of caffeine in the streams Mo¨nchaltdorfer Aa and Ustermer Aa during rainy weather (suggesting inputs of up to 10% untreated wastewater, see above) cannot be rationalized with a less efficient elimination in WWTPs, so that caffeine inputs with combined sewer overflows remain the most likely cause. Biodegradation of Caffeine in Greifensee. Concentrations of caffeine in Greifensee, analyzed between June and September 2002, ranged from 36 to 114 ng L-1 (Figure 5) and were generally lower than those found in the inflowing streams Mo¨nchaltdorfer Aa and Ustermer Aa (62-410 ng L-1, Table 1). As dilution by additional water inflows to the lake and by precipitation is relatively small (≈17% between June and September, 2002), caffeine must have been eliminated to some degree in the lake. Caffeine was shown to undergo slow photochemical degradation by reaction with HO• radicals (9). For Greifensee, however, the corresponding estimated half-life of ≈10 years (integrated value over the whole depth of the lake) exceeds the mean water residence time (420 d, (14)) by an order of magnitude; i.e., photodegradation is not a relevant removal process in this lake. Furthermore, dark-chemical reactions and partitioning (sorption/sedimentation and volatilization) are negligible for the fate of caffeine in surface waters (9). Hence, biodegradation is probably the dominant elimination process in Greifensee. Biodegradation rates for caffeine may be obtained from batch incubation experiments in the laboratory, but the VOL. 40, NO. 13, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4099

FIGURE 5. Vertical concentration profiles of caffeine (circles) and temperature profiles (dashed line) in lake Greifensee, 2002.

FIGURE 6. Mass balance of caffeine in the catchment area of lake Greifensee in August/September 2002 (observation period, 28 d; population, 108000). extrapolation to natural conditions in a lake is difficult. Alternatively, the importance of biodegradation may be estimated from a mass balance approach for the lake:

input ) export + biodegradation + storage

(2)

Input. As mentioned above, a major fraction of caffeine inputs to Greifensee was covered by analyses in the two inflows, Mo¨nchaltdorfer Aa and Ustermer Aa. These inputs amounted to 1.2 kg between June 25 and July 22, 2002 (observation period, 27 d), and 1.8 kg between August 26 and September 23 (28 d). Further inputs from discharges of treated wastewater from the WWTPs of Uster and Maur as well as from combined sewer overflows in Uster and Mo¨nchaltdorf, downstream the sampling sites in the corresponding streams (Figure 1), were estimated at 0.3 and 0.5 kg for June/July and August/September, respectively. It was assumed that the fraction of sewer overflows downstream of the sampling sites was the same as that upstream. Export. In the same way, caffeine export loads may be estimated from analyses in flow-proportional samples taken in the outflow of the lake, the river Glatt, considering the actual discharge of the river. Alternatively, the concentration of caffeine in the surface water layer of the lake may also represent the concentration in the outflowing river since, in Greifensee, the discharging water is primarily from this layer and horizontal mixing is relatively fast (15). The mean concentrations of caffeine in 1 m depth were ≈105 and 94 4100

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 13, 2006

ng L-1 during June/July and August/ September 2002, respectively. Multiplication with the mean discharge of the river Glatt between sampling (6.2 × 106 and 8.8 × 106 m3, respectively (11)) yields export loads of 0.7 and 0.8 kg for the two observation periods, respectively. Storage. The total amount of caffeine in Greifensee, calculated from vertical concentration profiles (Figure 5) and bathometric data, decreased from ≈11.8 kg in June to 10.2 kg in July and 9.5 kg in August, followed by a slight increase to 9.8 kg in September. The “storage” during the two observation periods thus amounted to -1.6 kg in June/July and 0.3 kg in August/September. Biodegradation. With the above mass balance, the contribution of biodegradation was then estimated at 2.4 kg for June/July and 1.2 kg for August/ September. Of course, such a mass balance approach is subject to errors. Assuming relative errors of 10% for input and export loads and 50% for storage, error propagation results in estimates of 2.4 ( 0.8 and 1.2 ( 0.3 kg for biodegradation, respectively (standard deviations). Figure 6 summarizes the whole mass balance of caffeine in Greifensee for the 28-d period in August/September. More than 90% (2.1 kg) of the caffeine inputs to the lake were from combined sewer overflows, corresponding to a fraction of 4.4% wastewater (dry weather equivalents) discharged untreated, whereas inputs with treated wastewater (0.2 kg) were less important. In this season, the total amount of caffeine

in the lake was about 10 kg. Biodegradation (1.2 kg) and export (0.8 kg) were the dominant elimination processes, leading to a replacement of the inventory of caffeine in the lake within ≈5 months. Rate Constants for Biodegradation of Caffeine in Greifensee and in WWTPs. It was assumed that the biodegradation of caffeine in Greifensee basically followed first-order kinetics, and corresponding mean rate constants, kbio [d-1], may be derived from the following equation:

Lbio kbio ) kw Lexport

(3)

where kw is the mean flushing rate coefficient (0.0015 and 0.0021 d-1 in June/July and August/September, 2002, respectively, derived from daily discharge values at the outflow of the lake and its volume, (11)) and Lexport and Lbio are the above-calculated loads for export and biodegradation, respectively [kg per observation period]. In this way, rate constants of kbio ≈ 0.006 and 0.003 d-1 were calculated for June/July and August/September, which is equivalent to halflives of ≈120 and 240 d, respectively. The higher rate constant in June/July may partly be rationalized by the somewhat higher temperature in the lake during this time (Figure 5). Another approach to estimate rate constants for biodegradation is to follow the concentration decrease of caffeine in deeper water layers during summer. In this season, Greifensee is stratified in a warmer surface water layer (epilimnion), a thermocline at ≈5-15 m, and a colder deep water layer (hypolimnion) (Figure 5). Due to the different density of warm and cold water, vertical mixing between epi- and hypolimnion is minimal in this season. Water inflow and outflow are basically to and from the epilimnion, respectively. The hypolimnion can therefore be considered as an isolated water layer without significant inputs or outputs. Between June and September, the concentration of caffeine in 15 and 30 m depth decreased with rate constants of ≈0.003 and 0.007 d-1, respectively (assuming first-order kinetics). These in situ degradation rate constants can be compared to rate constants obtained from batch incubation experiments under laboratory conditions (dark, 20 °C) using water from Greifensee, where caffeine degraded with a rate constant of ≈0.006 d-1 (9). To sum up, all these estimates for biodegradation are in the range of ≈0.003-0.007 d-1, corresponding to half-lives of several months. Caffeine thus shows, at first sight, an astonishingly high persistence in Greifensee, given the rapid elimination in WWTPs. Rate constants for biodegradation in WWTPs can be estimated from elimination efficiencies (typically >99%) and residence times in the activated sludge basins (again assuming first-order kinetics). Rate constants calculated in this way range between ≈3 and 20 d-1 (half-lives, ≈0.8-5.0 h, data not detailed here) and are thus about 3 orders of magnitude higher than those estimated for the lake. Nevertheless, this difference is reasonable considering the high microbial activity in activated sludge. For comparison, rate constants for biodegradation of the pharmaceutical compound ibuprofen were also about 2-3 orders of magnitude higher in activated sludge than in lake water (16). Phosphorus Inputs to Greifensee from WWTP Effluents and Combined Sewer Overflows. As outlined in the Introduction, knowledge of the relative magnitude of phosphorus inputs from WWTP effluents and combined sewer overflows to Greifensee is important to take appropriate measures for further input reduction. WWTPs in the region of Greifensee all operate with phosphate precipitation by iron salts and subsequent sand filtration. Elimination rates for phosphorus are therefore high, typically in the range of ≈96-98% (17). Mean per capita loads in influent and effluent wastewater

are Iuntreated ≈ 0.67 kg P person-1 y-1 and Itreated ≈ 0.02 kg P person-1 y-1, respectively (17). Wastewater-derived phosphorus inputs to Greifensee could now be estimated from Iuntreated, Itreated, P, and f using eq 1. As described above, the marker caffeine suggested annual mean fractions of combined sewer overflows of ≈1.9% to Mo¨nchaltdorfer Aa and ≈2.5% to Ustermer Aa. With use of these values, an annual input of ≈1.5 t P to Greifensee via combined sewer overflows was calculated. Combined sewer overflows are thus equally important sources of phosphorus as WWTP effluents (≈2.0 t P y-1). A recently published report (17) provided an alternative estimate for the fraction of combined sewer overflows in the region of Greifensee, basically from GIS data (geographical information system), considering precipitation, fraction of sealed areas, dimensions and capacities of sewer systems and WWTPs, population, etc. The study anticipated phosphorus inputs of ≈1.6 t P y-1 from combined sewer overflows (17), which is remarkably similar to our results and an excellent validation of the concept of using caffeine as a chemical marker for combined sewer overflows. The contribution of further major phosphorus sources to Greifensee was estimated in other reports (cited in (18)) and is listed here for comparison: agricultural runoff, 4.6 t P y-1; geogenic sources, 2.9 t P y-1; treated wastewater, 2.0 t P y-1; stormwater, 1.6 t P y-1 (total input: 13.2 t P y-1). Suitable measures for reduction of phosphorus inputs to Greifensee are suggested and discussed elsewhere (18). The consistency between the caffeine-based and the GISbased estimate of phosphorus inputs from combined sewer overflows was nonetheless somewhat astonishing as caffeine is a highly hydrophilic compound, whereas phosphates tend to sorb to inorganic particles such as iron and aluminum oxides (19). The transport of particulate-bound compounds in sewer systems, WWTPs, and overflows may be different from that of dissolved compounds. Therefore, caffeine is certainly best suited for extrapolation to other hydrophilic, wastewater-derived compounds. Application of Caffeine as a Marker for Combined Sewer Overflows to Other Surface Waters. The correlation between caffeine loads in receiving waters and precipitation (Figure 2), and the fact that caffeine inputs from treated wastewater were not significantly increased at high throughput in WWTPs (Figure 4), basically demonstrated the suitability of caffeine as a marker for combined sewer overflows in the catchment area of Greifensee. Its application to other surface waters, however, requires some region-specific knowledge, in particular, on per capita loads of caffeine in untreated and treated wastewater representative of the respective catchment area (Iuntreated and Itreated). The fraction of combined sewer overflows can only be determined with adequate accuracy if elimination efficiencies in WWTPs are high; i.e., the percentage of remaining caffeine in WWTP effluents should be lower than the expected fraction of overflows. Furthermore, the percentage of residents that are not connected to a WWTP has to be considered and may contribute to the overall load of caffeine in receiving waters. In rivers and streams, the loads of caffeine show pronounced temporal variations (Table 1). Therefore, flowproportional sampling is important to quantitatively assess the burden of rivers by domestic wastewater. The fraction of overflows may then be estimated with eq 1. In lakes, temporal variations are less pronounced and representative water samples are easier to obtain. However, in lakes with water residence times in the order of months to years, the contribution of biodegradation to the overall removal of caffeine cannot be neglected. In this case, the fraction of combined sewer overflows can be estimated with the following mass balance, which is analogous to eq 1: VOL. 40, NO. 13, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4101

Llake ) P × f × Iuntreated + P × (1 - f) × Itreated

(4)

with

Llake ) clakeV(kw + kbio) where Llake is the total influent load of caffeine to the lake, calculated from the steady-state concentration in the lake (clake), the lake volume (V), the water exchange rate coefficient (kw), and the biodegradation rate constant (kbio), which is, as discussed above, dependent on the local situation (temperature, biological activity, etc.) and difficult to determine. In summary, if some region-specific conditions are considered, caffeine can successfully be used as a marker for combined sewer overflows to surface waters. Applying a combination of several markers is, of course, advisable, but the choices of markers that specifically indicate combined sewer overflows are limited (6-8).

Acknowledgments This research project was sponsored by AWEL (Office for Waste, Water, Energy, and Air of the Canton of Zu ¨ rich, Switzerland). Interesting discussions were held with M. Koch, W. Meier, P. Spohn, and C. Balsiger (AWEL) and are kindly acknowledged. We thank the personnel of AWEL and Labor Veritas, Zurich, for providing samples from Greifensee and from streams, M. Sobaszkiewicz from WWTP Wetzikon for wastewater samples, A. Kelly from Ernst Winkler und Partner AG for information on the local situation in Uster, and our colleagues A. Ba¨chli, M. E. Balmer, and A. Hauser for their help in the lab.

Literature Cited (1) Liechti, P. Der Zustand der Seen in der Schweiz; Swiss Agency for the Environment, Forests and Landscape (BUWAL): Bern, Switzerland, 1994; 159 pp. (2) Office for Waste, Water, Energy, and Air of the Canton of Zurich (AWEL), Switzerland. Gewa¨sserschutz. www.gewaesserschutz.zh.ch (accessed December 2005). (3) Dodds, W. K. Freshwater Ecology; Academic Press: San Diego, CA, 2002; 569 pp. (4) Swiss Agency for the Environment, Forests and Landscape (BUWAL), Bern, Switzerland. Anschlussgrad an Kla¨ranlagen. www.umwelt-schweiz.ch/buwal/de/fachgebiete/gewaessers chutz/abwasser/kommunal/anschlussgrad (accessed December 2005). (5) Moffa, P. E. Control and Treatment of Combined Sewer Overflows, 2nd ed.; John Wiley & Sons: New York, 1997; 292 pp.

4102

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 13, 2006

(6) Eganhouse, R. P.; Sherblom, P. M. Anthropogenic organic contaminants in the effluent of a combined sewer overflow: impact on Boston Harbor. Marine Environ. Res. 2001, 51, 5174. (7) Fono, L. J.; Sedlak, D. L. Use of the chiral pharmaceutical propranolol to identify sewage discharges into surface waters. Environ. Sci. Technol. 2005, 39, 9244-9252. (8) Marvin, C.; Coakley, J.; Mayer, T.; Brown, M.; Thiessen, L. Application of faecal sterol ratios in sediments and effluents as source tracers. Water Qual. Res. J. Canada 2001, 36, 781-792. (9) Buerge, I. J.; Poiger, T.; Mu ¨ ller, M. D.; Buser, H. R. Caffeine, an anthropogenic marker for wastewater contamination of surface waters. Environ. Sci. Technol. 2003, 37, 691-700. (10) Reichert, P. AQUASIM - a tool for simulation and data-analysis of aquatic systems. Water Sci. Technol. 1994, 30, 21-30. (11) Office for Waste, Water, Energy, and Air of the Canton of Zurich (AWEL), Switzerland. Hochwasserinformationsstelle des Kantons Zu ¨rich. www.hochwasser.zh.ch (accessed December 2005). (12) Office for Waste, Water, Energy, and Air of the Canton of Zurich (AWEL), Switzerland. Abwasser- und Kla¨rschlammanfall im Kanton Zu ¨ rich 2002 (2003). (13) Office for Waste, Water, Energy, and Air of the Canton of Zurich (AWEL), Switzerland. Personal communication, P. Spohn (November 2005). (14) Kupper, U.; Koch, M.; Meier, W.; Niederhauser, P. Oberfla¨chengewa¨sser und Abwasserreinigungsanlagen; Office for Waste, Water, Energy, and Air of the Canton of Zurich (AWEL): Zurich, Switzerland, 1998; 67 pp. (15) Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. Environmental Organic Chemistry, 2nd ed.; John Wiley & Sons: Hoboken, NJ, 2003; 1313 pp. (16) Buser, H. R.; Poiger, T.; Mu ¨ ller, M. D. Occurrence and environmental behavior of the chiral pharmaceutical drug ibuprofen in surface waters and in wastewater. Environ. Sci. Technol. 1999, 33, 2529-2535. (17) Morgenthaler Ingenieure AG. Gesamtphosphorbelastungen aus Abwasserreinigungsanlagen im Einzugsgebiet des Greifensees; Office for Waste, Water, Energy, and Air of the Canton of Zurich (AWEL): Zurich, Switzerland, 2003; 76 pp. (18) Meier, W. Phosphorbelastung des Greifensees: Aktuelle Belastungen, Reduktionspotenziale und deren Kosten, Auswirkungen von Sanierungsmassnahmen auf den Seezustand; Office for Waste, Water, Energy, and Air of the Canton of Zurich (AWEL): Zurich, Switzerland 2003; 14 pp. (19) Stumm, W.; Morgan, J. J. Aquatic Chemistry: Chemical Equilibria and Rates in Natural Waters, 3rd ed.; John Wiley & Sons: New York, 1996; 1022 pp.

Received for review December 21, 2005. Revised manuscript received March 21, 2006. Accepted April 4, 2006. ES052553L