ARTICLE IN PRESS

Water Research 38 (2004) 15201529

Occurrence and fate modeling of MTBE and BTEX compounds in a Swiss Lake used as drinking water supply
Torsten C. Schmidta,*, Stefan B. Haderleina,1, Rolf Psterb, Richard Forsterb
a

Swiss Federal Institute for Environmental Science and Technology (EAWAG) and Swiss Federal Institute of Technology, Zurich (ETHZ), P.O. Box 611, Ueberlandstr. 133, CH-8600 Duebendorf, Switzerland b Wasserversorgung Zurich, Hardhof 9, CH-8023 Zurich, Switzerland Received 28 April 2003; received in revised form 8 December 2003; accepted 17 December 2003

Abstract Emissions of fuel components from boating use on multiple-use lakes and reservoirs are of high concern with regard to the drinking water supply from such water bodies. We report results of a detailed study on the occurrence, sources and fate of aromatic hydrocarbons and methyl tert-butyl ether (MTBE) in a typical holomictic lake, Lake Zurich, that supplies drinking water for the largest Swiss city. Emphasis of the investigation was on the fuel oxygenate MTBE, which was found in concentrations up to 1.4 mg/L in the epilimnion and up to 0.05 mg/L in the hypolimnion of the lake. The concentration difference was due to the stratication of the lake during the boating season with very limited water exchange across the thermocline. MTBE and BTEX nearly completely volatilized before vertical lake mixing occurred in winter. Spatial and temporal variations of MTBE concentrations in the lake were observed and successfully predicted using two complementary box models (MASAS Light and Aquasim). The drinking water supply from holomictic lakes is not at risk for the scenarios studied if water is extracted from well below the thermocline. Since emissions of unburned gasoline into such water bodies are caused predominantly by boating activities, restrictions of highly emitting twostroke engines could substantially reduce the MTBE and BTEX load of the epilimnion during the boating season. r 2004 Elsevier Ltd. All rights reserved.
Keywords: MTBE; BTEX; Gasoline; Drinking water; Boating; Lake

1. Introduction The concentration of fuel-derived hydrocarbons in lakes, rivers, and coastal waters has been intensively studied in the past [1]. The major emission source was found to be boating trafc, in particular emissions of unburned fuel from two-stroke-engines mainly used in small motor and sailing boats [24]. More recently,
*Corresponding author. Present address: Center for Applied Geoscience, Eberhard-Karls-University Tuebingen, Wilhelmstr. 56, D-72074 Tuebingen, Germany. Tel.: +49-7071-29-73147; fax: +49-7071-29-5139. E-mail address: torsten.schmidt@uni-tuebingen.de (T.C. Schmidt). 1 Center for Applied Geoscience, Eberhard-Karls-University Tuebingen, Wilhelmstr. 56, D-72074 Tuebingen, Germany.

oxygenates added to gasoline such as methyl tert-butyl ether (MTBE) have been the focus of investigations in the US. Reported peak concentrations of MTBE in the epilimnion of (shallow) freshwater lakes were as high as 31 mg/L [5,6], which is well above the odor threshold of about 15 mg/L. As for other hydrocarbons, MTBE concentrations were strongly correlated with the number of boats in use. Consequently, various communities have restricted the use of personal watercraft on multiple-use lakes, in particular in California [7]. Pankow et al. [8] modeled volatilization rates for MTBE and benzene in surface water as a function of the three parameters depth, wind speed and water ow velocity. The calculated half-lifes ranged from 0.1 to 85 days for MTBE and from 0.005 to 42 days for benzene. Thus, a fairly rapid air/water exchange is generally expected for MTBE in shallow surface waters. In

0043-1354/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2003.12.027

ARTICLE IN PRESS
T.C. Schmidt et al. / Water Research 38 (2004) 15201529 1521

conclusion, elevated concentrations of MTBE, primarily due to boating activities, may only be expected during periods of high input, e.g., during the boating season at marinas and resorts. Reuter et al. [9] studied in detail the potential sources of MTBE at Lake Donner and have shown that neither road runoff nor precipitation contributed much to the MTBE load. They found MTBE concentrations rapidly decreasing after the end of the boating season and attributed the MTBE loss mainly to volatilization, which was corroborated by comparing model results based on MTBE loss via volatilization with measured data in two lakes [10]. Recently, Bradley et al. [11] suggested microbial degradation of MTBE in lake sediments as an additional sink for MTBE in stratied lakes. In Europe, much less information on the occurrence of MTBE in freshwater lakes is available than in the US. The few reports, together with data for other surface water bodies and US lakes, are summarized in Table 1. Due to the lack of MTBE concentration data for European lakes, data from US lakes is often used to
Table 1 Reported concentrations of MTBE in surface water bodies Country Denmark Finland Germany Germany Germany Germany Switzerland Switzerland Switzerland Switzerland US US US US US US Water body (if specied) Various lakes Creeks and rivers Helsinki area Various rivers and ponds Various rivers Various rivers Various rivers Various rivers, Kanton Zurich, year round Lake Zurich, epilimnion data, boating seasond Lake Zurich, epilimnion data, off-seasone Lake Zurich, hypolimnion data, year roundf Lake Donner, year round Lake Tahoe, boating season New Jersey lakes, boating season California lakes, year round Lake Texoma, year round Cranberry Lake, New Jersey, year roundh No. of samples 4 5 16 118 430 6 132 172 59 200 na 14 40 na na 54

estimate if boating use may lead to disturbance of wildlife, human activities or poses a risk for drinking water supplies [13]. However, there are major differences between the US and a typical mid-European situation, regarding boating use, climate and lake hydraulics. Of particular concern in the US are personal water crafts for high-speed boating with carbureted or nondirect fuel injection two-stroke engines (essentially all jet skis and wave runners manufactured before 1999). Such carbureted two-stroke engines emit up to 2530% of the fuel unburned into the water or the atmosphere [7]. In Europe, such watercrafts are hardly seen yet on freshwater lakes and streams, and only might become a future source of fuel (and MTBE) input in lakes. Regarding lake hydraulics, many of the lakes in the southwestern US studied over the last years are nearly stagnant in summer, i.e., they have hardly any outow. Lake Zurich is a typical mid-European multiple-use lake with intense boating activities during summer. At the same time, it serves as the main drinking water supply for the city of Zurich. About 45 million m3 per

% Positivesa na 5 81 97 87 67 45 100 100 100 na 100 98 na 20c 100

Concentration average and range (in mg/L) na; oLOD0.59 na; 0.190.62 0.06; o0.010.2 0.21; o0.059.8b 0.17; o0.012.4 0.11; o0.010.16 0.15; o0.052.53g 0.29; 0.101.4 0.059; 0.0360.10 0.038; 0.0300.047 na; o0.112 0.90; 0.184.2 14; o0.228.1 na; 120c 2.4c; o110.2 13.5; 0.430.9

Reference [12] [13] [14] [15] [16] [17] [18] This study This study This study [9] [19] [5] [10] [20] [6]

na: not available, LOD: limit of detection. a Note that % positives depends heavily on the LOD, which ranges from 0.01 to 1 mg/L. b Outlier of 9.8 mg/L not included in average as suggested by the authors. c Estimated value. d Measurements 19992003 from mid-May to September at depth 05 m. e Measurements from October 2002 to April 2003 and October to November 2003 at all depths (lower LOD of 0.02 mg/L). f Measurements from April 2002 to November 2003 at depth >20 m (lower LOD of 0.02 mg/L). g Outlier of 2.53 mg/L not included in average. h Data set partially overlapping with [5].

ARTICLE IN PRESS
1522 T.C. Schmidt et al. / Water Research 38 (2004) 15201529

year are extracted by two drinking water treatment plants from 30 m below surface. Due to the reports from the US, Zurich water works (WVZ) were concerned about impacts of boating on the quality of the extracted water. The aims of this study were therefore: (i) to determine the content of MTBE and other fuel-derived aromatic hydrocarbons in Lake Zurich depending on spatial and temporal distribution, (ii) to identify the major immission and elimination pathways for such compounds in a holomictic lake and, (iii) to model the behavior of these compounds over time and space. Thus, we provide data for a risk assessment of MTBE and aromatic hydrocarbons in a holomictic lake used as drinking water supply.

Fig. 1. Lake Zurich with location of sampling points (1=Riesbach, 2=Thalwil) and lling stations ().

2. Experimental section 2.1. Lake Zurich Lake Zurich is a prealpine, glacio-uvial lake of wellknown hydraulics and morphology (Table 2). Its length from Rapperswil to Zurich is about 30 km in north west direction (Fig. 1). At Rapperswil, Lake Zurich is separated by a dam from Lake Obersee, which is the main inow to the lake. The outow of Lake Zurich is the river Limmat in the city of Zurich. Lake Zurich is holomictic with regular deep-mixing over winter (from January to April) and a (theoretical) mean water residence time of 430 days. In summer, however, water exchange by inow and outow is almost completely conned to the epilimnion, which results in a residence time of less than 3 months. Since inow during the deepmixing phase is about 50% of the summer inow, more than 6 years are actually needed for the total renewal of the lake water [22]. 2.2. Study site and sampling points Preliminary investigations were carried out in 1999 at various sampling points near Zurich harbor and in summer and winter 2000 during two regular sampling campaigns. From 2001 on, depth proles for MTBE were measured in Thalwil and Zollikon in February, May, August and November (sampling always in the morning). In order to study in more detail daily and spatial variations of MTBE concentrations in the lake, two sites were selected for further measurements: Riesbach, close to Zurich harbor, and Thalwil, at the deepest point of the lake (Fig. 1). Riesbach is located near the outow of Lake Zurich with intense boating trafc arising from the various marinas in or near the city. Thalwil is about 8 km further up the lake, where boats cross only occasionally. At these sites, depth proles from 0 to 30 m were measured on two sunny Sundays in June and July with intense boating trafc (one sampling round in the morning, one in the afternoon). 2.3. Chemicals MTBE, benzene, toluene, ethylbenzene, o-xylene, mxylene, and p-xylene were purchased as aromatics- and ether mix 30 from Dr. Ehrenstorfer (Augsburg, Germany). This mixture contains all seven analytes at a concentration of 1 g/L in methanol (certied maximum concentration error 70.5%). Water and methanol used for standard preparations were HPLC-grade from Baker (Stehelin, Basel, Switzerland). 1,4-dichlorobenzene

Table 2 Characteristic data of Lake Zurich Lake morphology Volume (m3) Surface area (m2) Length (m) Max/mean width (m) Max/mean depth (m) Lake hydraulics Theoretical mean residence time of water (d) Mean, max and min throughow of water (m3/d) Mean water exchange rate epilimnion/hypolimnion (summer) (m3/d) Mean epilimnion depth (summer) (m) Mean epilimnion volume (summer) (m3) Mean temperature in epilimnion (summer) ( C) 3.3 109 66.6 106 29 103 4/2.3 103 136/51

430 7.7 106, 23 106, 2.5 106 0.49 106 [21]

5 3.3 108 16.3

ARTICLE IN PRESS
T.C. Schmidt et al. / Water Research 38 (2004) 15201529 1523

(purum, Fluka, Buchs, Switzerland) was used as internal standard. 2.4. Analytical methods All analytes (MTBE and BTEX) were quantied using gas chromatography with split-/splitless injector 1177 and ame ionization detection (Varian GC 3800, Varian, Zug, Switzerland). For enrichment and injection, headspace SPME with a Combi-PAL autosampler (CTC, Zwingen, Switzerland) was used. Ten milliliter of a sample was lled in a 20-mL headspace vial and 1 mL of an internal standard solution was added. Subsequent SPME enrichments were done with a 75 mm Carboxen/ PDMS ber (Supelco, Bellefonte, PA). Equilibration time was 15 min at room temperature, desorption time was 20 min at an injector temperature of 280 C (split open after 60 s). Separation of analytes was carried out with a GC column CP-Select 624, 30 m 0.32 mm, df 1.8 mm (Chrompack/Varian, Middleburg, Netherlands). Carrier gas was helium at a constant ow of 2 mL/min. The following temperature program was used: 5 min at 40 C, to 150 C at 5 C/min, to 250 C at 40 C/min, 5 min at 250 C. For the FID the following settings were used: temperature 300 C, make-up gas (nitrogen) at 25 mL/min, hydrogen at 30 mL/min, and synthetic air at 300 mL/min. Quantitation was done with an external calibration curve after correction by the internal standard. Important parameters of the analytical method are summarized in Table 3. The reliability of the method was validated in two round-robin tests for the determination of MTBE at low concentrations. Selected gasoline samples from boat lling stations were analyzed by the Swiss Federal Institute for Testing and Materials (EMPA) according to the European regulation EN 228. The MTBE content in these gasoline samples was 2.5 and 9.8 vol% for a regular and premium grade gasoline, respectively. 2.5. Modeling approach For all model calculations physico-chemical parameters of MTBE and BTEX compounds published by
Table 3 Analytical method parameters Substance MTBE Benzene Toluene Ethylbenzene m/p-Xylene o-Xylene Limit of quantitation (LOQ) (mg/L) 0.10 0.10 0.50 0.10 0.20 0.10 Working range (mg/L) 0.1010 0.1010 0.5010 0.1010 0.2010 0.1010

Squillace et al. [23] were used. The Henry constants KH of the compounds were extrapolated to 16 C where possible. In the case of MTBE, the following equation was derived for the temperature dependency of KH from measurements in a temperature range from 5 C to 35 C [24]: ln KH=5041 1/T+9.83, yielding KH=0.50 atm L mol1 at 16 C. The relevant elimination processes and overall concentrations in epi- and hypolimnion of Lake Zurich were modeled with MASAS Light, a spreadsheet-based program developed at EAWAG [25]. This very userfriendly program was used to model situations where horizontal concentration gradients were negligible. A time frame with the following four phases was implemented to predict contaminant concentrations: (I) mixed lake, no input (105 days), (II) stratied lake, no input (20 days), (III) stratied lake, input (153 days), (IV) stratied lake, no input (87 days). In order to model the spatial distribution of MTBE in Lake Zurich, the more sophisticated program Aquasim [26] had to be used. The conceptual model of the lake used for Aquasim simulations is shown in Fig. 2. Due to the limited availability of experimental data, two wellmixed reactors representing two sections of the epilimnion of the lake during the input phase (boating season) were coupled with an advective and a diffusive link. The two reactors comprised only half of the total lake volume because at the upper end of the lake another area of intense boating exists at the cities of Rapperswil . fkon. Therefore, at the outer boundary of and Pfa reactor 2, the concentration gradient was assumed to be zero. Consequently, net losses of MTBE from reactor 2 occurred neither by advection nor by diffusion across the outer boundary. Since water exchange with the hypolimnion during this phase is very small and does not inuence epilimnion concentrations, it was neglected. Both reactors were in contact with an air compartment with a constant MTBE concentration of 0.5 mg/m3 assumed as average value for Switzerland [27]. The atmospheric background concentration could only be considered in Aquasim. Input parameters for both models are summarized in Table 4. Major uncertainties regarding MTBE input comprise the average MTBE content of gasoline used for boats (2.510 vol.%) and

Measurement uncertainty (%) 715 710 710 710 710 710

Recoveries (%) 110 110 105 105 105 108

Retention time (min) 5.49 9.33 13.85 17.74 18.08 19.13

ARTICLE IN PRESS
1524 T.C. Schmidt et al. / Water Research 38 (2004) 15201529

Fig. 2. Conceptual model of Lake Zurich epilimnion for modeling spatial distribution of MTBE and BTEX with Aquasim. vtot: mass transfer coefcient for airwater exchange. Khor: horizontal diffusivity. Water exchange between epi- and hypolimnion is small, hence the hypolimnion was not considered as an additional compartment. Note that the model comprises only half of the total epilimnion volume (see text for details).

Table 4 Model input parameters for calculations with MASAS Light and Aquasim Parameters for estimating MTBE input into the lake Number of boats with two-stroke enginesa Gasoline consumption of all twostroke engines on Lake Zurichb Average MTBE content of gasoline used in boats Emission rate of unburned gasoline from two-stroke engines Dissolution rate of unburned gasoline emitted by two-stroke engines Boating season from May to September Resulting MTBE input in Lake Zurich during boating season Other model input parameters Wind speed Epilimnion depth Average epilimnion temperature during summer Average ow velocity in epilimnion Fraction of MTBE released in source and non-source zone (Aquasim only)c
a b

the share of fuel emitted unburned from two-strokeengines (450%). Emissions from four-stroke engines were neglected. 3. Results and discussion

Value 4600

3.1. Seasonal and daily uctuations

193000 L/a 4.6 vol% 20% 80% 153 d 6.9 kg/d

1.6 m/s 5m 16 C 5 cm/s 80%/20%

Estimated from [28] and regional boat registry. Estimated from [28]. c These fractions are estimates based on observations of boat numbers during sample campaigns.

Since preliminary investigations in 1999 and 2000 indicated that concentrations of MTBE in Lake Zurich ranged from 0.1 to 1.4 mg/L during summer the uctuations of MTBE concentrations over time and space were subsequently studied in more detail. A summary of all MTBE measurements obtained from 19992003 is included in Table 1. During the boating season, MTBE was always detected in the epilimnion at concentrations >0.1 mg/L. Because off-season and hypolimnion concentrations of MTBE always were below the limit of quantitation (0.1 mg/L), the analytical method was changed from FID to MS detection, which provided a ve-fold improvement in sensitivity. For hypolimnion and off-season epilimnion measurements included in Table 1, only measurements from May 2002 on are therefore included. The concentrations in the hypolimnion are remarkably constant at both sampling points and over time, i.e., they do not depend on the MTBE input in the epilimnion. The average concentration in 200 samples taken from the lake hypolimnion since April 2002 is 0.038 7 0.004 mg/L. This is nearly 10 times lower than the average epilimnion concentrations during the boating season. The average off-season

ARTICLE IN PRESS
T.C. Schmidt et al. / Water Research 38 (2004) 15201529 1525

epilimnion concentrations are only slightly elevated compared with hypolimnion concentrations (less than a factor of 2), which substantiates the rapid decrease of the MTBE load in the lake after the end of the boating season. To investigate if MTBE input over the day resulted in signicant diurnal concentration patterns, morning and afternoon sampling campaigns were carried out on two days in June and July 2001 with intense boating trafc. The depth proles for MTBE obtained during these four sampling events at both sites are shown in Fig. 3a (Riesbach) and b (Thalwil). Very similar depth proles for MTBE were measured on other sampling dates during the boating season and at other sampling sites of the WVZ water quality monitoring at Lake Zurich. Thus, a similar correlation of MTBE concentrations with boating use as already described for multiplepurpose lakes in the US [9,10] was found. However, maximum epilimnion concentrations were about 10 times lower than reported from US lakes, which might be explained by the lower MTBE input due to a lower average content of MTBE in gasoline and less emitting boating types. At another lake nearby, Lake Walen, with much less intense boating use, summer concentra-

tions of MTBE did not exceed 0.14 mg/L and BTEX could never be detected. Diurnal uctuations of MTBE concentrations at Lake Zurich during the two sampling campaigns on June 24 and July 29 were signicant. Whereas MTBE concentrations in the epilimnion were rather constant during the morning measurements, over the day a steep concentration gradient developed in the upper few meters that concurred with the temperature gradient. Thus, diurnal epiliminic stratication with reduced vertical mixing might explain the higher MTBE concentrations beneath the lake surface. Vertical mixing within the epilimnion during the night presumably destroyed the epilimnic concentration gradient. 3.2. Spatial distribution Depth proles measured at Riesbach and Thalwil (Fig. 3a and b) revealed that MTBE concentration proles very well matched the corresponding temperature proles (data not shown), i.e., concentrations were nearly constant in the epilimnion (see above for exemption) and decreased rapidly with depth. This

MTBE conc. in g/L 0.00 0 5 Depth in m 10 15 20 25 24.06.01 AM 24.06.01 PM 29.07.01 AM 29.07.01 PM LOQ 0.20 0.40 0.60 0.80 1.00 1.20 1.40

(a)
MTBE conc. in g/L 0 0 5 Depth in m 10 15 20 25 0.1 0.2 0.3 0.4 0.5 0.6

24.06.01 AM 24.06.01 PM 29.07.01 AM 29.07.01 PM 08.08.01 AM LOQ

(b)
Fig. 3. Depth proles of MTBE at the two sampling points Riesbach (a) and Thalwil (b). Temperature proles are not shown to enhance legibility. LOQ: limit of quantitation (see Table 3).

ARTICLE IN PRESS
1526 T.C. Schmidt et al. / Water Research 38 (2004) 15201529

concentration prole is consistent with previously published data from Lake Donner [9]. The vertical distribution of BTEX, however, differed from the pattern seen for MTBE. BTEX concentrations decreased more rapidly with depth than MTBE concentrations. Frequently, 2.5 m below the surface BTEX concentrations were below the limit of quantitation (0.1 mg/l). Though the reason for this behavior is not exactly clear, our nding is consistent with results by Wick et al. [29] who attributed the loss of benzene in the metalimnion of a lake to a stable community of benzene degrading bacteria in this region. Comparison of MTBE concentrations in the epilimnion of Lake Zurich at the two sampling sites Riesbach and Thalwil shows clearly a correlation of MTBE concentrations with boating trafc intensity. In the morning sampling campaigns, the concentration difference between the two sites (1.1 and 1.4 times higher at the highly frequented site Riesbach in June and July, respectively) was lower due to horizontal mixing during the night when no MTBE input occurred. During day time, the MTBE concentration at Riesbach was 1.8 times higher on both sampling dates. For BTEX compounds, a similar but less-pronounced trend was observed. This indicates that the assumption of horizontally homogenous concentrations in the epilimnion is an oversimplication of the system which tends to underestimate peak diurnal concentrations in boating areas. In 2001/2002, in addition to the monitoring of MTBE and BTEX in Lake Zurich, MTBE was measured several times in the outow of Lake Zurich, river Limmat. Concentrations in May 2001, i.e., shortly after the boating season commenced, were 0.23 mg/L, and rose to

0.45 mg/l in August 2001. In agreement with the modeling results (see below), after cessation of the boating season, no MTBE was detected in river Limmat in November 2001 and February 2002. This result is consistent with monthly monitoring data of the regional authorities in 2000 that gave maximum MTBE concentrations of 0.66 mg/l in August [18]. In that study, Purge & Trap-GC/MS was used with a lower limit of detection (LOD) of 0.05 mg/L. MTBE was detected year round but from October to May concentrations were below 0.1 mg/L. In contrast, BTEX were only detected between May and August above 0.05 mg/L. Peak concentrations in August for benzene, toluene, ethylbenzene and xylenes were 0.47, 1.02, 0.39, and 2.05 (sum of all isomers) mg/L, respectively. 3.3. Modeling results The modeling with MASAS Light and Aquasim served several purposes: rst, the major elimination processes for MTBE and BTEX were identied with the use of lake and substance-specic data (using MASAS Light). The results were then used to predict average epiand hypolimnion concentrations in the whole lake. Second, a simulation over several years was run to evaluate long-term trends of MTBE in the lake. The Aquasim model was used to analyze the differences in concentration found at the two sampling sites Thalwil (non-source zone) and Riesbach (source zone). Aquasim also allowed to include a non-zero air concentration. The major elimination processes from the lake epilimnion were airwater exchange (approx. 2 3 of total 1 elimination rate) and ushing 3 (Table 5). Elimination

Table 5 Average rst-order rate constants for elimination of MTBE and BTEX in the epilimnion of Lake Zurich during summera Rate constant Symbol Units Compound MTBE Flushing through outow Water exchange with hypolimnionb Abiotic hydrolysis Gas exchanged Sedimentation Direct photochemical transformation Indirect photochemical transformation Biological transformatione Total elimination rate constant
a b

Benzene 0.023 0.0015 0.055 o0.001 o0.001 o0.001 o0.001 0.080



Toluene 0.023 0.0015 0.051 o0.001 o0.001 o0.001 o0.001 0.076

o-Xylene 0.023 0.0015 0.047 o0.001 o0.001 0.001 o0.001 0.073

kw kex khyd kgas ksed kphotod kphotoi kbio k

d1 d1 d1 d1 d1 d1 d1 d1 d1

0.023 0.0015 o0.001c 0.050 o0.001 o0.001 o0.001 o0.001 0.075

Estimated from model calculations with MASAS Light using physico-chemical parameters at 25 C (see text). The small hypolimnion concentrations were neglected. c No hydrolysis at lake pH (7.88.6) [30]. d Using Henry constants extrapolated to 16 C (where data for extrapolation was available); however, this does not change the result more than 3%. e Microbial degradation was neglected for all compounds since no rate constants appropriate for a lake epilimnion were available. In particular for the BTEX compounds this is a worst-case scenario since degradation might considerably contribute to overall loss [29].

ARTICLE IN PRESS
T.C. Schmidt et al. / Water Research 38 (2004) 15201529 1527

rates were surprisingly similar for all compounds, considering that the Henry constants differ by more than one order of magnitude between MTBE and the three BTEX compounds. Water exchange with the hypolimnion contributed only about 2% to the total removal from the epilimnion. Other elimination processes were not of importance (o1%), except for indirect photochemical transformation via reaction with OH-radicals in the case of o-xylene (approx. 1.4%). In contrast to studies on US lakes that involved a calculation of MTBE loss [9,10], ushing via the outow needs to be considered for Lake Zurich as for most other mid-European lakes. This is evident by the rather high concentrations of MTBE in the lake outow, river Limmat, during summer (see above). Bradley et al. [11] recently suggested microbial degradation of MTBE in lake sediments as a potential sink for MTBE in stratied lakes. However, for MTBE and BTEX, sedimentation is not an important elimination process. Since hardly any MTBE reaches the lake sediment, microbial degradation in lake bottom sediments may not contribute substantially to the overall elimination of MTBE. Fig. 4 shows MASAS Light predictions of MTBE concentrations in the epilimnion and hypolimnion of the lake (assumed to be horizontally well mixed) in comparison with measured concentrations. Predicted epilimnion concentrations for the whole lake are well within the range of experimental results at the Thalwil site, which represents the average lake concentrations better than the source-zone site (Riesbach). The equilibrium concentration of 0.28 mg/L is reached 40 days after the boating season (and thus the MTBE input) commences (75%). Over 4 years of constant MTBE input during Phase 3 (P3), there is hardly a rise in hypolimnic concentrations due to the small vertical water exchange during stratication (see Table 2) and

since most of the MTBE has already been eliminated from the lake before complete mixing occurs. A further model prediction shows that even after 15 years (MTBE has been used in Switzerland in substantial amounts since about 1985), hypolimnion concentrations remain o0.015 mg/L, under the given assumptions. The measured hypolimnic concentrations of 2002 and 2003 are about 2.5 times higher than the predicted ones. The reason is mainly that a zero concentration for air had to be assumed due to constraints of the MASAS Light program. The upper limit of predicted concentrations is obtained by simply assuming equilibrium between the mixed lake and the atmosphere and no net changes in the hypolimnion after stratication. For this case (using KH=0.304 atm L mol1 at 8 C from our own measurements) the predicted equilibrium concentration in water assuming a constant air concentration of 0.5 mg/m3 is about 0.038 mg/L. This perfectly matches the average measured hypolimnion concentration of 0.038 mg/L. The Aquasim prediction of horizontal segmentation in the lake is shown in Fig. 5. With this model it is possible to match the measured concentrations at the Riesbach (source zone) and Thalwil (diffusive-linked zone) site. The effective horizontal diffusivity Khor was used as a tting parameter. The obtained value of 5.18 106 m2/d (or 60 m2/s) is at the upper end of turbulent horizontal diffusivities found in lakes [31] but can be explained by the fact that in addition to horizontal diffusivity other horizontal exchange processes are included in the effective parameter. The predicted epilimnion concentrations at Riesbach are at the higher end of the measured data. This might be due to the fact that MTBE concentrations were measured in the morning, and no MTBE was emitted in

Fig. 4. MASAS Light prediction of epi- and hypolimnion concentrations of MTBE in Lake Zurich over 4 years compared with measurement results at Thalwil in August 2001. P1, P2, P3, P4: phases 14, respectively (details see text). Note that air concentrations could not be considered in this model and are therefore assumed to be 0.

Fig. 5. Aquasim prediction of epilimnion concentrations of MTBE at the two sampling points Riesbach (within source zone) and Thalwil (within non-source zone). For parameters not listed in Table 4 the following values were used: vtot=0.249 m/d, Khor=5.18 106 m2/d, l=6000 m, w=2000 m. For comparison, the range of measured MTBE concentrations in the epilimnion on July 29, 2001 are given (data from the morning sampling campaign were used).

ARTICLE IN PRESS
1528 T.C. Schmidt et al. / Water Research 38 (2004) 15201529

the approx. twelve hours before sampling whereas the model assumes a constant input over the whole boating season. Since there are a number of model input parameters that could only be estimated, a sensitivity analysis was carried out with Aquasim to identify which of these parameters dominate uncertainties in the model result. MTBE input and water throughow inuence concentrations in both lake compartments in the same direction whereas the fraction of MTBE released in the source and non-source zone, and the effective horizontal diffusivity Khor determine the concentration difference between both compartments. Overall, for both lake compartments, the uncertainty in MTBE input in the lake was most important. MTBE input is governed by several factors: gasoline consumption of boats, average MTBE content of gasoline used in these boats, average emission rate of unburned gasoline and dissolution rate of MTBE in water from unburned gasoline. For all of these parameters estimated values had to be used, either based on local data or on literature values. 3.4. Environmental signicance Due to the density stratication of holomictic lakes there is hardly any water exchange in summer and thus transport of dissolved contaminants between epi- and hypolimnion. If contaminants are almost completely eliminated during the stratication period, their concentration in the hypolimnion will remain very low even over long time periods. Drinking water is typically extracted well below the thermocline, therefore no risk is expected for the drinking water supply in such lakes. The frequently used assumption of horizontal homogeneity in lakes is not always valid on a diurnal time scale as shown for the spatial MTBE distribution in Lake Zurich. Although elimination rates of VOC in running water are higher than in lakes, river water below multiple-use lakes may be contaminated with MTBE at the sub-mg/L level as is the case in river Limmat. Furthermore, it is advisable to include benzene and other BTEX when studying boating emissions since substantial amounts of these toxicologically more relevant compounds are emitted along with MTBE and total elimination rates are similar. In order to further reduce emissions of unburned fuel into surface water, limitations of the use of high-emitting two-stroke engine types in motorboats should be considered.

Light. Rowena Crockett is acknowledged for the analysis of gasoline samples from one of the boat lling stations. Several students have provided valuable input during a case study about MTBE in Lake Zurich using MASAS Light. We thank Markus Ulrich, Frank Peeters and Heinz Singer for helpful comments on the manuscript. Financial support has been provided by Com! ne ! rale des Eaux (CGE). pagnie Ge

References
[1] Mantoura RFC, Gschwend PM, Zariou OC, Clarke KR. Volatile organic compounds at a coastal site. 2. Short-term variations. Environ Sci Technol 1982;16:3845. [2] Gabele PA, Pyle SM. Emissions from two outboard engines operating on reformulated gasoline containing MTBE. Environ Sci Technol 2000;34(3):36872. [3] Juttner . F, et al. Emissions of two- and four-stroke outboard enginesI. Quantication of gases and VOC. Water Res 1995;29(8):197682. [4] Juttner . F, et al. Emissions of two- and four-stroke outboard enginesII. Impact on water quality. Water Res 1995;29(8):19837. [5] Baehr AL, Zapecza OS. Methyl tert-butyl ether (MTBE) and other volatile organic compounds in lakes in Byram township, Sussex County, New Jersey, Summer 1998. 984264, US Geological Survey, West Trenton, 1998. [6] Toran L, Lipka C, Baehr A, Reilly T, Baker R. Seasonal and daily variations in concentrations of methyltertiarybutyl ether (MTBE) at Cranberry Lake, New Jersey. Water Res 2003;37(15):375666. [7] Anonymous. Two-stroke and MTBE issues. California Department of Boating and Waterways, 2003. Available at http://www.dbw.ca.gov/mtbe.htm. [8] Pankow JF, Rathbun RE, Zogorski JS. Calculated volatilization rates of fuel oxygenate compounds and other gasoline-related compounds from rivers and streams. Chemosphere 1996;33(5):92137. [9] Reuter JE, et al. Concentrations, sources, and fate of the gasoline oxygenate methyl tert-butyl ether (MTBE) in a multiple-use lake. Environ Sci Technol 1998;32(23): 366672. [10] Stocking AJ, Kavanaugh MC. Modeling volatilization of MTBE from standing surface waters. J Environ Eng ASCE 2000;126(12):11316. [11] Bradley PM, Landmeyer JE, Chapelle FH. Widespread potential for microbial MTBE degradation in surfacewater sediments. Environ Sci Technol 2001;35(4):65862. [12] Anonymous. Handlingsplan for MTBE, Miljo- og Energiministeriet Miljostyrelsen, Kobenhavn, 1998. [13] Hansen BG, et al. editors. EUR 20417 ENEuropean Union Risk Assessment Report on Tert-butyl methyl ether. Environment and quality of life, vol. 19. Luxembourg: Ofce for Ofcial Publications of the European Communities; 2002. 282pp. [14] Effenberger M, Weiss H, Popp P, Schirmer M. Untersu. r-butylether chungen zum Benzininhaltstoff Methyltertia . chenwasser in Deutsch(MTBE) in Grund- und Obera land. Grundwasser 2001;2:5160.

Acknowledgements We thank Frank Peeters for his help in the conceptual model and its implementation in AQUASIM for the modeling of spatial distribution. Markus Ulrich helped with all features of modeling with his program MASAS

ARTICLE IN PRESS
T.C. Schmidt et al. / Water Research 38 (2004) 15201529 . r[15] Klinger J, Sacher F, Brauch H-J. MTBE (methyl-tertia butylether)Ein Problemstoff fur . die Wasserversorgung? gwf-Wasser/Abwasser 2002;143(3):16671. [16] Achten C, Kolb A, Puttmann W, Seel P, Gihr R. Methyl tert-butyl ether (MTBE) in river and wastewater in Germany1. Environ Sci Technol 2002;36(17):365261. [17] Achten C, Puttmann . W. Determination of methyl tertbutyl ether in surface water by use of solid-phase microextraction. Environ Sci Technol 2000;34(7):135964. . ssern [18] Pster H, Balsiger C. Untersuchung von Fliessgewa auf VOC im Kanton Zurich. . Baudirektion Kanton Zurich, . . sserschutz, Zurich, AWEL, Abteilung Gewa . 2001. [19] Boughton CJ, Lico MS. Volatile organic compounds in Lake Tahoe, Nevada and California, JulySeptember 1997. USGS Fact Sheet FS-055-98, US Geological Survey, Carson City, NV, 1998. [20] An YJ, Kampbell DH, Sewell GW. Water quality at ve marinas in Lake Texoma as related to methyl tert-butyl ether (MTBE). Environ Poll 2002;118(3):3316. [21] Dzambas Z, Ulrich M. MASAS library les for Swiss lakes, EAWAG/ETH, Duebendorf, 1994. [22] Gammeter S, Forster R, Zimmermann U. Limnologische Untersuchung des Zurichsees . 19721996, Wasserversorgung Zurich (WVZ), Zurich, 1997. [23] Squillace PJ, Pankow JF, Korte NE, Zogorski JS. Review of the environmental behavior and fate of methyl tertbutyl ether. Environ Toxicol Chem 1997;16(9):183644. 1529 [24] Arp HP. Master thesis. Eberhard-Karls-University, Tuebingen, 2004. [25] Ulrich MM. MASAS light. Swiss Federal Institute for Environmental Science and Technology (EAWAG), Duebendorf, 1998. [26] Reichert P. AQUASIMa tool for simulation and data analysis of aquatic systems. Water Sci Technol 1994;30(2):2130. . tzung der Altlastenrele[27] Lothenbach B, Munz C. Abscha vanz von Methyl-tert-butylether (MTBE). Umweltmaterialien, Nr. 151, Bundesamt fur . Umwelt, Wald and Landschaft, Bern, 2002. [28] Schurmann . P. Reduktion der VOC-Emissionen bei der Bootsbetankung. Schriftenreihe Umwelt Nr. 254, Bundesamt fur . Umwelt, Wald und Landschaft, Bern, 1996. [29] Wick LY, McNeill K, Rojo M, Medilanski E, Gschwend PM. Fate of benzene in a stratied Lake receiving contaminated groundwater discharges from a Superfund site. Environ Sci Technol 2000;34(20): 435462. [30] OReilly KT, Moir ME, Taylor CD, Smith CA, Hyman MR. Hydrolysis of tert-butyl methyl ether the (MTBE) in dilute aqueous acid. Environ Sci Technol 2001;35(19): 395461. [31] Schwarzenbach RP, Gschwend PM, Imboden DM. Environmental organic chemistry. New York: Wiley; 2003.