Environ. Sci. Technol. 2002, 36, 3482-3489
Phototransformation of Triclosan in Surface Waters: A Relevant Elimination Process for This Widely Used BiocidesLaboratory Studies, Field Measurements, and Modeling C EÄ L I N E T I X I E R , H E I N Z P . S I N G E R , SILVIO CANONICA, AND STEPHAN R. MU ¨ LLER* Swiss Federal Institute for Environmental Science and Technology (EAWAG), CH-8600 Du ¨ bendorf, Switzerland
The phototransformation of the widely used biocide triclosan (5-chloro-2-(2,4-dichlorophenoxy)phenol) was quantified for surface waters using artificial UV light and sunlight irradiation. The pH of surface waters, commonly ranging from 7 to 9, determines the speciation of triclosan (pKa ) 8.1) and therefore its absorption of sunlight. Direct phototransformation of the anionic form with a quantum yield of 0.31 (laboratory conditions at 313 nm) was identified as the dominant photochemical degradation pathway of triclosan. Combining the photochemical parameters with actual meteorological data and field measurements allowed us to validate a model describing the behavior of triclosan in the water column of a Swiss lake (Lake Greifensee). From August to October 1999, direct phototransformation accounted for 80% of the observed total elimination of triclosan from the lake. The remaining major sink for triclosan was the loss in the outflow. Thus, during the summer season, direct phototransformation appears to be a major elimination pathway of triclosan in this lake. Based on absorption spectra and quantum yield data, the phototransformation half-lives of triclosan were calculated under various environmental conditions typical for surface waters. Daily averaged half-lives were found to vary from about 2 to 2000 days, depending on latitude and time of year.
Introduction Triclosan, 5-chloro-2-(2,4-dichlorophenoxy)phenol (Figure 1), commercially known as Irgasan DP 300 or Irgacare MP, is a broad-spectrum antibacterial agent widely used in personal care products (e.g. toothpaste, soap, deodorant, cosmetics). Triclosan is also suitable for incorporation in polymers and fibers to give these materials antibacterial properties. It is used in mattress pads, food cutting boards, shoes, and sportswear. Triclosan, previously found in the aquatic environment (water and sediments) as the result of wastewater emission from a specialty chemicals manufacturing plant (1, 2), was recently detected in urban wastewaters (3-5). The latter studies showed that it was eliminated from wastewater with an efficiency of 85-95% in sewage treatment plants, but the residual amount of triclosan in these wastewaters led to surface water triclosan concentrations in the range of about 5-90 ng L-1. * Corresponding author phone: +41-1-823-5460; fax: +41-1-8235471; e-mail:
[email protected]. 3482
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 16, 2002
FIGURE 1. Chemical structure of triclosan. The toxicity of triclosan has been tested on various aquatic organisms and the major concern of triclosan in surface water is its toxicity to certain algae species, e.g., Scenedesmus subspicatus (6). The no-observed effect concentration (NOEC) of this species is 500 ng/L (7), which leads to a predicted no effect concentration (PNEC) of about 50 ng/L (assuming the commonly used safety factor of 10). Furthermore, some recent findings indicate a possible bacterial resistance to triclosan, which may result from its property to block lipid biosynthesis by specifically inhibiting the enzyme enoyl-acyl carrier protein reductase (8, 9). However, clinical studies with longterm exposure to triclosan-containing products such as deodorants or toothpastes have not resulted in development of bacterial resistance on skin or mucous membrane (1012). Further research is currently being undertaken to evaluate these findings. For proper exposure estimates and for an improved risk assessment of triclosan in the aquatic environment, a better knowledge of the transformation processes of triclosan in surface water is needed. According to preliminary work (3), the phototransformation process should be the main elimination process for triclosan from a lake water column. As a chlorinated phenol derivative, triclosan, present in surface waters both in its molecular and anionic form (pKa ) 8.1 (13)), is expected to undergo direct phototransformation (1416) and probably also photochemical transformations sensitized by dissolved natural organic matter (DOM), i.e., reaction with singlet molecular oxygen (17, 18), excited triplet states (19), or other oxidizing radicals (20, 21). As shown by Kawaguchi in the case of 2-chlorophenol (22, 23), both direct and indirect phototransformation pathways may be relevant for phenolic compounds in surface waters. In the present paper, we report results of photochemical kinetic studies using UV light produced in the laboratory and sunlight irradiation. These photochemical parameters determined under controlled conditions were then combined with field data to simulate vertical concentration profiles of triclosan in Lake Greifensee. The model used is similar to the one recently validated to quantify the transformation of phenylurea herbicides in the same lake (24). With the goal of improving risk assessment procedures for triclosan, we also discuss how such photochemical parameters can be employed, using the computer program GCSOLAR (25), to calculate phototransformation rates in various aquatic environments. Moreover, we present a short review of all the parameters influencing the occurrence and fate of triclosan in surface water.
Experimental Section Chemicals. Triclosan was obtained from Ciba SC, Basel, Switzerland. Valerophenone (98%) was supplied by EGAChemie, Steinheim/Albuch, Germany. All solvents (HPLC grade) were purchased from Scharlau (Barcelona, Spain) and were used as received. Suwannee River fulvic and humic acid (type standard) were purchased from the International Humic Substances Society. Deionized water was further purified with a Nanopure water purification device (NANOpure 4, Skan, Basel, Swit10.1021/es025647t CCC: $22.00
2002 American Chemical Society Published on Web 07/10/2002
zerland). Lake water for photochemical experiments was collected at 1 m depth in the middle of Lake Greifensee and exhibited the following chemical properties: pH ) 8.6, [DOC] ) 4.0 mg C L-1, and [NO3-] ) 1.6 mg N L-1. Analyses. During the photochemical experiments, the depletion of the parent compound was followed by HPLC on a Hewlett-Packard System 1050 using a reverse-phase column (Nucleosil C18-5 µ, 125*4 mm - Macherey-Nagel, Oensingen, Switzerland) at room temperature. A mobile phase of acetonitrile/water/acetic acid 1% vol/vol (55:35:10) was used, at a flow rate of 1 mL min-1. Detection was performed with a Hewlett-Packard 1050 UV detector set at 235 nm. UV-spectra were recorded on a Kontron Instruments Uvikon 930 spectrophotometer, using quartz cuvettes of 1-10 cm optical path length. Irradiation Setup. The setup for photoirradiation experiments in the laboratory or for sunlight irradiation is described in detail elsewhere (26). Briefly, photoirradiations in the laboratory were carried out in quartz tubes (internal diameter 15 mm, external diameter 18 mm), using a merry-go-round photoreactor (MGRR) DEMA (Hans Mangels GmbH, Bornheim-Roisdorf, Germany) model 125, equipped with a Hanau TQ718 medium-pressure mercury lamp, which was operated at 500 W power throughout the experiments. Irradiations were performed using a Duran 50 borosilicate glass cooling jacket (total thickness of glass 4 mm) and the following filter solutions: λ > 290 nm, deionized water; λ ) 313 nm, 0.206 g L-1 potassium chromate with 1.00 g L-1 sodium carbonate solution; λ g 334 nm, 12.75 g L-1 sodium nitrate solution. Alternatively, irradiations were carried out using a photoirradiation system from Applied Photophysics (London, UK), consisting of a 900 W xenon source and a f/3.4 grating monochromator set at 279 nm or at 292 nm, with an optical bandwidth of 10 nm. Chemical actinometry with valerophenone was used to determine the photon fluence rate during monochromatic irradiation experiments (27). Sunlight irradiations were performed in quartz tubes mounted in a linear rack placed across a bath of deionized water or Greifensee water. A stair-shaped support was used to place tubes at various depths (every 5 cm) in the water bath. The temperature in all irradiation experiments was kept at 25.0 ( 0.5 °C. Triclosan solutions in Nanopure water were prepared at pH 9.0 (5 mM phosphate buffer) by dilution of a 80 µM aqueous stock solution, and the pH was subsequently adjusted to the desired value by adding HCl or NaOH. For sunlight irradiation, triclosan was dissolved in 90% Greifensee water which was previously filtered (0.45 µm cellulose acetate). Field Site. Lake Greifensee is a small eutrophic lake (surface area: 8.46 km2 - max. depth: 32 m) located 10 km east of Zu ¨rich, Switzerland (47°21′N/8°41′E). Lake Greifensee is a holomictic lake with regular deep-mixing in winter (December/March) and a mean water residence time of 408 days. The catchment area of Greifensee (detailed description in ref 3) is of about 130 km2 with approximately 89 000 inhabitants. We can assume that no industrial discharge of triclosan takes place in this area. The effluents of seven municipal wastewater treatment plants (WWTP) are discharged into the epilimnion either directly or indirectly via the two main tributaries, Aa Uster and Aabach Moenchaltorf. The river Glatt is the only outflow of the lake. Sampling Program and Analytical Procedure. Between August 16, 1999 and October 22, 1999, an intensive sampling program was conducted (for details see ref 3). Lake water samples at various depths above the deepest point were regularly collected using a Niskin bottle, from which the water was transferred to 1-L glass bottles. The two main tributaries, Aa Uster and Aabach Moenchaltorf, as well as the effluents of three WWTPs discharging directly into the lake were
sampled using volume-proportional sampling devices. The sampler was emptied twice a week, and samples from WWTPs were filtrated (0.45 µm, regenerated cellulose filter, Schleicher and Schuell, Dassel, Germany) immediately after collection to reduce biological activity. Samples were stored overnight at 4 °C in the dark and analyzed the following day. The analytical procedure used is described in detail elsewhere (3). In brief, filtered samples (1 L), spiked with 13 C6-triclosan as an internal standard, were enriched by solidphase extraction. After a derivatization step with diazomethane, the samples were analyzed on a GC/MS system, which consisted of a HRGC 8000, a MD800 mass spectrometer, and an autosampler A200S, all Fisons Instruments (Beverly, MD). The detection limit with this method was 1 ng L-1 for triclosan in surface water. Lake Model. Vertical concentration profiles were simulated using the computer software AQUASIM (28), which allows to construct mathematical models describing the dynamic behavior of chemicals in lakes. Lake Greifensee was represented by a vertical stack of 64 horizontal boxes each of 50 cm thickness. Complete horizontal homogeneity was assumed for each layer. In contrast, vertical mixing was explicitly described by turbulent diffusion with time- and depth-dependent diffusivities, which were calibrated by means of vertical temperature profiles. In the epilimnion, this heat-budget approach is, in this case, not adequate, since there, the heat balance is dominated by radiation and airwater exchange. In the top 1 m diffusivity was estimated by adopting the classical boundary layer (Law-of-the-Wall) assumption (29). Turbulence measurements in other lakes (30) have demonstrated that short-term turbulent mixing can reliably be quantified based on wind speed and stratification alone (31). Thus hourly measured temperature profiles (providing the water column stability) were combined with hourly measured wind speed (providing the turbulence level) to determine the surface boundary diffusivity on a 1-h time step. Phototransformation Process. Under optically thin conditions, the pseudo-first-order rate constant for direct phototransformation of a compound of interest in a given volume element, kp,direct (s-1) can be described by eq 1, where λ is the wavelength of light (nm), (λ) is the decadic molar absorption coefficient of the compound of interest (m2 mol-1), Φ(λ) is the reaction quantum yield (mol einstein-1), and dI(λ)/dλ is the irradiation spectrum (einstein m-2 s-1 nm-1).
kp,direct ) 2.303‚
∫ (λ)‚Φ(λ)‚ dλ ‚dλ dI(λ)
(1)
λ
Triclosan is a phenol derivative and as such is present in aqueous solution as two distinct chemical species in rapid dynamic equilibrium within each other. Considering only direct phototransformation, its apparent first-order rate app constant kp,direct may be expressed as app AH Akp,direct ) kp,direct ‚(1 - fA-) + kp,direct ‚f A-
(2)
where the superscripts AH and A- indicate that the quantity is related to the molecular and anionic form, respectively, and fA- is the fraction of triclosan present in the anionic form and depends on pH according to eq 3.
f A- )
1 1 + 10pKa-pH
(3)
We recall here that the kinetics of environmental contaminants is often discussed in terms of half-lives, which are inversely proportional to first-order rate constants. The direct phototransformation half-life of triclosan is defined as t1/2p,direct app ) ln2/kp,direct . Calculations of rate constants for photoVOL. 36, NO. 16, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3483
chemical transformations under terrestrial sunlight were performed with the computer program GCSOLAR (25) which implements eq 1 as a function of season, latitude, time-ofday, depth in the water body and ozone layer thickness, assuming a perfectly clear sky. Considering the pH of lake Greifensee during the field study, triclosan is mainly present in its anionic form. As the direct phototransformation rate constant of the molecular form is much lower than the one of the anionic form (see results), the following approximation was used for the model calculations: Aapp kp,direct ≈ kp,direct ‚f A-
(4)
Daily ephemeris values for Du¨bendorf, Switzerland (47°24′ N/8°37′ E) were calculated (32), and ozone layer thickness values were obtained from satellite data (33). Rate constants were calculated either for surface conditions for verification under controlled conditions or for a well-mixed water layer of 50 cm thickness for implementation into the lake model. To calculate effective triclosan phototransformation rate constants, a further correction accounting for reduced sunlight transmission through the atmosphere was considered, eq 5.
Wmeasured Wtheory
app Akp,direct (calc) ) kp,direct (GCSOLAR)‚f A-‚
(5)
Wmeasured is the hourly measured global radiation at Du ¨ bendorf and Wtheory is the hourly theoretical global radiation, calculated as described by Gerecke et al. (24). Sedimentation Process. Hydrophobic partitioning was assumed. The river particles already contaminated with triclosan were assumed to have no influence on this partitioning as they settled down near the lake entrance. In the middle of the lake, the adsorption to particles was assumed to take place only on newly built particles, whose vertical concentration profiles were deduced from the monthly measured vertical concentration profiles of particulate organic carbon (POCepilimnion = 1.8 mg L-1). An equilibrium process was then defined, corresponding to the adsorption/desorption equilibrium of the nondissociated form of triclosan to particles. The fraction in particulate form (F) was calculated as follows: F ) 1-(1/(1+Koc‚POCnew)) where Koc is the natural organic matter/water partition constant of triclosan (47 000 L kgoc-1) and POCnew is the concentration in new particulate organic carbon. The settling of these particles was assumed to occur with a depth dependent sedimentation velocity: 0.5 m d-1 for depths between 0 and 5 m, 1 m d-1 for depths below 10 m and a linearly interpolated value for depths between 5 and 10 m (38).
Results and Discussion pH Dependence of Direct Phototransformation. The pKa of triclosan (pKa ) 8.1) lies in the pH range commonly encountered in surface freshwaters, which means that photochemical reactions of both the molecular and anionic form have to be investigated. The mathematical formulation of the apparent direct phototransformation rate constant is given in the Experimental Section (eqs 1-4). Figure 2 illustrates how the pH affects the absorption spectrum of a triclosan solution and, consequently, the light absorption rate of the solution. This rate is directly proportional to the overlap between the absorption spectrum of the solution and the spectrum of terrestrial sunlight. Such an overlap increases with increasing pH, i.e., with increasing fraction of anionic form of triclosan. Thus, at equal phototransformation quantum yields of both forms, the transformation of the anionic form under sunlight is expected to be much faster than the transformation of the molecular form. This was 3484
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 16, 2002
FIGURE 2. Upper part: Variation in the absorption spectrum of triclosan solution according to pH. Lower part: Overlap area of the terrestrial sunlight intensity spectrum and triclosan absorption spectrum at various pH: ‚‚‚ sunlight; - - pH ) 5.9; - - - pH ) 9; s pH ) 11.5 verified in the laboratory using the MGRR (irradiation at λ > 290 nm) as a sunlight surrogate. The pseudo-first-order rate constants for direct phototransformation of triclosan (Table 1) increased by a factor of about 30 by increasing the pH from 5.9 (99% molecular form) to 11.0 (99.9% anionic form). Quantum Yields of Direct Phototransformation. The quantum yield for depletion of triclosan was determined at the absorption maxima of the molecular (279 nm) and anionic (292 nm) forms using the photoirradiation system. Using the MGRR, quantum yield determinations were also attempted with monochromatic light at 313 nm, which is close to the wavelength of maximum spectral overlap between sunlight and triclosan absorption spectrum (Figure 2). Owing to the very small absorption of the molecular form at this wavelength, only the quantum yield of the anionic form could be determined. The results are summarized in Table 2. Similar and relatively high quantum yield values at the absorption maximum were found for both the molecular and anionic forms. These values are essentially higher than those known for the related compound 3-chlorophenol (0.13-0.15 for the anionic form, 0.09 for the molecular form (15)) probably due to the presence of the o-phenoxy substituant in triclosan. Influence of Dissolved Organic Matter. DOM affects the photochemical transformation kinetics of a trace pollutant by acting as a light absorber and, possibly, as a photosensitizer or a scavenger of reaction intermediates (34). To determine the role played by DOM in triclosan phototransformation, humic or fulvic acid (DOC: 2 mg L-1) was added to buffered triclosan solution in Nanopure water (pH 9.0). By irradiating these solutions in the MGRR at wavelengths above 334 nm (for which photoabsorption and therefore direct phototransformation of triclosan is negligible), the observed transformation rate was very low: indirect phototransformation seems to be negligible for triclosan. The same solutions and buffered triclosan solutions without organic matter were then exposed to natural sunlight: the presence of organic matter led to a decrease of around 20% in the phototransformation rate of triclosan. According to the absorption spectrum of the organic matter used, this decrease was mainly due to the light absorption of dissolved organic matter. Another series of experiments was carried out under natural sunlight with
TABLE 1. Pseudo-First-Order Rate Constants for Direct Phototransformation of Triclosan at Various pH Valuesa pH ) 5.9 (1% anionic form)
pH ) 8.0 (48% anionic form)
pH ) 9.1 (91% anionic form)
(3.8 ( 0.6) × 10-4
(6.9 ( 0.5) × 10-3
(1.1 ( 0.2) × 10-2
kbexp (s-1)
Experimental conditions: initial triclosan concentration: 2 µM in phosphate buffer solution - λ > 290 nm. linear regression of ln(concentration) versus time plots. a
pH ) 11.0 (99.9% anionic form) (1.2 ( 0.3) × 10-2 b
Standard deviation obtained from
TABLE 2. Quantum Yield at 279, 292, and 313 nm for Triclosan Direct Phototransformationa
quantum yield
pH ) 5.9 λ ) 279 nm
pH ) 11.5 λ ) 292 nm
pH ) 11.0 λ ) 313 nm
0.40 ( 0.04
0.50 ( 0.04
0.31 ( 0.03
a
Experimental conditions: initial triclosan concentration: 2 µM in phosphate buffer solution. Standard deviation from four independent measurements.
TABLE 3. Experimental (exp) and Calculated (calc) Triclosan Phototransformation Kinetic Parameters under Sunlight Irradiation at Du1 bendorf, Switzerland, for Water Surface Conditions at pH 8.6 date
halflifeexp (min)
exel k p,direct (s-1)a
calc k p,direct (eq 5) (s-1)b
08.11.00 14.6 (7.9 ( 0.7) × 10-4 (7.5 ( 1.7) × 10-4 08.15.00 16.3 (7.1 ( 0.8) × 10-4 (7.0 ( 1.4) × 10-4 09.12.00 25.7 (4.5 ( 0.5) × 10-4 (4.2 ( 0.8) × 10-4
FIGURE 3. Influence of depth (z) on triclosan phototransformation rate constant (k) under sunlight: laboratory conditions (exp)/ GCSOLAR calculations (GCSOLAR). Experimental conditions: initial triclosan concentration: 2.0 µM in lake water; pH ) 8.6.
exel k p,direct / calc k p,direct (-)
0.95 0.98 0.94
a Standard deviation obtained from linear regression of ln(concentration) versus time plots. b Estimated relative standard deviation of 15% from the individual errors of different parameters (error propagation rule).
FIGURE 4. Time course of triclosan input in the Lake Greifensee. triclosan solutions (2 µM) prepared either in Nanopure water, buffered using 5 mM phosphate, or in natural water collected in the Lake Greifensee (DOC ) 4.0 mg L-1). The pH of the solutions was adjusted to be identical in all samples (pH ) 8.6). The transformation rates (data not shown) of triclosan under both conditions differ by only 10% and can be considered identical within experimental error. Thus, within the scope of the present study, DOM acts only as an optical filter for triclosan phototransformation. Verification of Laboratory Data under Natural Sunlight. To verify that, on the basis of quantum yield and spectral data determined in the laboratory, GCSOLAR and eq 5 give correct results, several rate constants were determined experimentally under sunlight irradiation and compared with calculated values. In the calculations, we used the quantum yield value of 0.31, corresponding to the wavelength of 313 nm, which is very close to the maximum spectral overlap of anionic form absorption and terrestrial sunlight. Calculation using a linearly decreasing quantum yield (determined using the quantum yield values at 292 and 313 nm) yielded basically identical results (less than 4% of difference). Table 3 gives the measured and calculated values of direct phototransformation rate constants and half-lives for triclosan, obtained in 90% Greifensee water on different sunny days under surface conditions (zero depth in the water body). During the late summer period under consideration, half-lives of the order of about 20 min were observed. They increased as the available sunlight intensity decreased following the seasonal change. The ratio of experimental to calculated rate constants was always close to unity, suggesting that our calculation method is a good predictor of the phototranformation rate of triclosan at the surface. The average of the three values of this ratio, 0.96, was used as a factor to further correct calculated rate constants obtained by eq 5. Experiments conducted at different pH values showed that the transfor-
mation rate constant of the molecular form is about 19 times lower than that of the anionic form. Sunlight irradiations at different depths in Greifensee water were also performed under controlled conditions as described in the Experimental Section. Very good agreement was obtained between experimental and calculated rate constants (Figure 3). According to these experiments, the rate constant at 50 cm depth corresponds to only 5% of the surface rate constant, i.e., 95% of the photochemically active sunlight is absorbed in the upmost 50 cm Greifensee water layer. This detailed photochemical study clearly shows that triclosan can be quickly eliminated from surface waters by direct phototransformation and that indirect phototransformation does not play an important role under the conditions investigated. Quantification of the Different Elimination Processes of Triclosan in Lake Greifensee. To evaluate and quantify the different elimination processes of triclosan in the real world, a field study with input measurements (Figure 4) and vertical concentration profiles (Figure 5) was conducted in Lake Greifensee (August to October 1999). An average input of 10 g d-1 was determined over this time period, with two maxima of 21 and 27 g d-1 at the end of August and at the beginning of October, respectively (for more details see ref 3). The temperature and vertical concentration profiles illustrate that the lake was stratified. The concentration in the hypolimnion remained constant, around 11 ng L-1 during the whole sampling period, whereas triclosan concentration in the epilimnion showed pronounced fluctuations. Despite the high input in August, triclosan concentration decreased from around 9 ng L-1 in August to 5 ng L-1 in September. Then the concentration slowly increased to reach 8 ng L-1 at the end of October. These fluctuations clearly indicate an elimination process of triclosan in the epilimnion. VOL. 36, NO. 16, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3485
FIGURE 5. Vertical concentration profiles of triclosan in Lake Greifensee. ( measured concentrations; ‚‚‚ temperature profile; simulation results considering as triclosan elimination process only the flushing (scenario I, -) or the flushing, the sedimentation and the phototransformation (scenario III, s); - - - error bounds for scenario III (( standard deviation). Based on a continuous one-dimensional lake model (see Experimental Section) using the input measurements and the profile of the 16th of August as the starting point, three scenarios were simulated (35). The following processes were considered: (i) elimination by flushing, (ii) sorption and sedimentation/resuspension, and (iii) photochemical transformation and the simulation results were compared to measured data. Scenario I. The thin lines (Figure 5) correspond to simulation results assuming flushing as the sole elimination process. The simulation results revealed a continuous increase of the concentration over time up to 15 ng/L at the end of October. Obviously, compared to the measured value of 8 ng/L the triclosan concentrations were overestimated in the epilimnion. Therefore, additional removal processes must be considered in the epilimnion. Scenario II. The second introduced elimination process was adsorption of the molecular form of triclosan to particles, followed by the sedimentation of these particles. The sorption process was defined as a simple hydrophobic partitioning on the newly formed particles in the lake. This process corresponded to a six times lower elimination rate than flushing and was unimportant for the time period considered, because of the low percentage of the nondissociated form of triclosan in the epilimnion (from 20 to 30%, pH ranging from 8.5 to 8.8). Scenario III. The simulated profiles (solid line in Figure 5) including a third elimination process, i.e., the direct photolysis of the anionic form of triclosan, matched the measured values very well. This elimination process was 5 times faster than flushing and corresponded for 80% of the elimination of triclosan. Therefore phototransformation was the most important elimination process during the study period. Note that the direct phototransformation of the anionic form of triclosan was assumed to take place only in the upmost 50 cm layer of the lake, where 95% of the sunlight available for triclosan phototransformation was absorbed. As the phototransformation rate of triclosan was very high in the first centimeters of the lake, a high time resolution (hourly averaged diffusion coefficient) had to be considered to describe the mixing and photransformation processes in the epilimnion (see Experimental Section). Reliability of the Results. A sensitivity analysis showed that the measured start profile and the phototransformation rate were the most important parameters. Thus, for the uncertainty analysis, relative or absolute standard deviations were defined for the most important model parameters (Table 4). For the start profile, an error of 1 ng L-1 coming from the uncertainty in the analysis method was assumed. An average 3486
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 16, 2002
TABLE 4. Standard Deviations Considered for Error Contribution in Model Calculations standard deviations parameter concentration of start profile in lake GCSOLAR phototransformation rate concentration effluent WWTP concentration river water inflow of WWTP water inflow of Moenchaltorf River water inflow of Uster River measured solar intensity
relative
absolute (ng L-1) 1
15% 6 3 10% 10% or 15% (rain events) 15% or 25% (rain events) 15%
error of 15% was assumed for the calculated phototransformation rate. For the input, the uncertainty in the measured concentration took into account the possible degradation during storage (3). Two different standard deviation values were taken for water inflow of rivers according to the water level: the uncertainty on the water flow measurements was assumed to be higher during rain events. In the AQUASIM software, a simple error propagation method is implemented: the linearized propagation of standard deviations of uncorrelated parameters. This error contribution was then combined with the error corresponding to a variation in pH of 0.2. This pH variation accounts for the error in pH measurements and for a possible daily variation in lake water pH. Figure 5 presents the results of such an error analysis. The dashed line corresponds to error bounds limiting the range of results values to minus or plus one standard deviation. Over the time period considered, the total standard deviation for the simulated concentrations ranged from 1 to 2 ng L-1. These results show that the direct phototransformation of the anionic form is the main elimination process for triclosan in the Lake Greifensee from August to October 1999, accounting for 80% of the overall elimination. Elimination by flushing is also important, around 17%. In contrast, the sedimentation and the biological degradation were found to be negligible over the time period considered. Assessing Phototransformation Half-Lives of Triclosan in the Aquatic Environment. In the previous sections we determined the photochemical parameters necessary to describe the transformation kinetics of triclosan in sunlit
FIGURE 6. Dependence of the direct phototransformation half-life of triclosan on latitude and season. Values calculated for the 15th day of each month and at intervals of 10° in latitude for a well mixed Greifensee water layer of 1.0 m depth, pH ) 8.0, and r (310 nm) ) 2.5 m-1. freshwaters and validated them with measured data in the Lake Greifensee. It is the aim of the present section to extend predictions of triclosan phototransformation rates to different environmental situations. Together with the rates for other elimination processes of triclosan in sewage treatment plants and surface waters, a good estimate of average phototransformation rates would allow improved exposure assessment from triclosan input data into the aquatic environment. Environmental factors affect the direct phototransformation of triclosan through (i) the fluence rate of available sunlight in the volume element, where the photoreaction takes place, (dI(λ)/dλ in eq 1), and (ii) the speciation of triclosan, which is controlled by pH (eqs 2 and 3). The effect of pH, already discussed above, results in triclosan half-lives at low pH (10). The fluence rate of available sunlight depends, itself, on various environmental factors, including latitude, season, time of day, climatic conditions affecting the transmission of the atmosphere, average ozone layer thickness, and water absorption. For risk assessment purposes, we are mainly interested in parameters that vary over a time period of several days or weeks. Consequently, we first studied the effect of latitude and season on the phototransformation half-life of triclosan (Figure 6). In our calculations performed AAH using GCSOLAR and eq 2 (assuming kp,direct /kp,direct ) 19), we used “standard” conditions corresponding to a water layer thickness of 1.0 m, a pH of 8.0, and an absorption coefficient of the water of 2.5 m-1 for a wavelength of 310 nm (the absorption coefficient of Greifensee water used in this study). We also averaged out the effect of time of day, using 24-haveraged values of phototransformation rate constants, and atmosphere transmission to UV-B radiation, using an ozone layer thickness of 300 D.u. and an average broadband transmission of 70% to account for the effect of clouds and aerosols (value valid for central Europe (36) and recently confirmed for the Greifensee area (37)). Figure 6 shows the expected seasonal variation of direct photolysis half-lives for different latitudes (25). The seasonal change in triclosan halflives is restricted within a factor of at most 2 from 0° to 20° latitude but increases significantly at higher latitudes. For example, at 40° latitude there is a 9-fold increase in half-life from summer to winter, and at 60° latitude this increase is 160-fold. The absorption coefficient of the water and the thickness of the (mixed) water layer exposed to sunlight are also important parameters for determining phototransformation half-lives in surface waters (25, 38). If the water layer is thick enough to allow total absorption of light in the significant wavelength range (condition which should hold for triclosan in many surface waters), then we may use eq 7 to describe
FIGURE 7. Seasonal variation of triclosan direct phototransformation half-life for the four examples detailed in Table 5.
TABLE 5. Parameters Used for Direct Phototransformation Half-Life Calculations of Typical Water Bodies name of water body River Neckar (Germany) River Rhine at Weil (France/Germany) Orinoco River outflow (Venezuela) hypothetical shallow lake (southern Finland)
short name latitude (see (rounded Figure 7) to 10°)
pH
r(310 zmix nm) (m) (m-1)
Neckar
50° N
8.0a
2b
5.7b
Rhine
50° N
8.0c
8d
2.5e
Orinoco
10° N
8.3f
10f 10f
Finland
60° N
6.1g
1h 10i
a Estimated value. b From ref 42. c Long-time average from ref 43. Reference 44. e From refs 45 and 46. f From ref 47. g Median for about 800 lakes in central and southern Finland (48, 49). h Working hypothesis. i Estimated assuming direct proportionality between R(310 nm) and organic carbon concentration. d
the dependence of direct phototransformation half-life of triclosan on these parameters
R(310 nm) (7) 2.5
tot t1/2p,direct = t1/2p,direct(1 m, Greifensee)‚zmix‚
where the superscript “tot” stands for total absorption, zmix (m) is the thickness of the mixed water layer, and R(310 nm) (m-1) is the decadic absorption coefficient (or beam attenuation coefficient (38)) of the water at 310 nm (the wavelength of maximum spectral overlap between terrestrial sunlight and triclosan absorption, see Figure 2). In Figure 7 and Table 5 we exemplify the use of eq 7 (dependence on zmix and R(310 nm)) and eq 2 (dependence on pH) to calculate direct phototransformation half-lives of triclosan for specific water bodies, employing the half-life data represented in Figure 6. The two example rivers in central Europe exhibit similar direct phototransformation half-lives, which are lower than 10 days for more than half of the year and reach a maximum of about 90 days in winter time. The calculated half-lives in the River Rhine are about twice as high as in the River Neckar due to the greater depth of the Rhine. For smaller rivers and streams, we estimate that direct phototransformation half-lives of triclosan will be of the same order of magnitude as for the above two example rivers. In these smaller streams, the effect of a smaller depth (zmix) may be compensated by shading by vegetation, as was reported for the River Glatt, Switzerland (zmix = 0.7 m) (37, VOL. 36, NO. 16, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3487
39). Large rivers in tropical regions are well-known to be rich in natural organic matter and therefore have a reduced transmission to UV-B. However, the availability of a high sunlight intensity throughout the year may still allow relatively short triclosan phototransformation half-lives, as shown in Figure 7 for the River Orinoco (near its estuary, located close to the Caribbean Sea). Figure 7 also displays an example of slow direct photodegradation of triclosan for a hypothetical lake in southern Finland. Owing to low pH, the half-life here is about an order of magnitude higher than for the two European rivers and is remaining lower than 100 days only for half of the year. For such a situation, alternative transformation or removal mechanisms of triclosan, i.e., biodegradation, sedimentation, and indirect phototransformation, have to be evaluated in more detail for accurate exposure assessment. Moreover these comparisons are based on the influence on the phototransformation rate of only three parameters: the pH, the thickness, and the decadic absorption coefficient of the mixed water layer. For a detailed risk assessment in any of these river systems, other factors which may affect the phototransformation rate like, the flow rate, the turbulence rate, the cross-sectional profile, etc. have to be considered.
Outlook to an Improved Risk Assessment The main motivation for this work and a previous study (3) was the fairly low predicted no effect concentration (PNEC) of triclosan of about 50 ng/L. This value should, from an ecological point of view, not be exceeded in surface waters. However, a few studies have reported triclosan concentrations in surface waters higher than 50 ng/L (3-5). The present work demonstrated the relevance of phototransformation as an elimination process of triclosan in surface waters. However we showed that the efficiency of such a process was highly dependent on seasonal and geographic parameters. For an improved risk assessment, information has thus to be collected about the different factors affecting the occurrence and the fate of triclosan in surface waters. The most important parameters determining triclosan concentration in surface waters are (i) use of triclosan in products which may reach surface waters via the sewer network, e.g., toothpaste, soaps, washing powder, etc.; (ii) the performance of the wastewater treatment plant which can vary between zero elimination (where there is no wastewater treatment plant) and up ca. 95% in modern plants; (iii) the dilution factor of the effluent with the receiving water; and (iv) parameters specific to the location or to the water system considered, i.e. parameters affecting the degradation processes of triclosan in surface waters such as pH, available light intensity, the absorption coefficient of the water body, the concentration of suspended particles, and the prevailing mixing processes. Thus, in addition to triclosan measurements, data have to be collected on several environmental parameters that affect the fate of triclosan in surface water. This study provides important information for triclosan risk assessment in surface water. However, in the case of low direct phototransformation rate, more information needs to be collected on alternative removal process for triclosan (biodegradation, sedimentation, and indirect phototransformation). Moreover, a previous study (3) on the fate of triclosan in wastewater treatment plants revealed significant adsorption of triclosan to sludge. Other studies described the formation of dichlorodibenzo-p-dioxins after exposure of triclosan to artificial sunlight in the solid state (40) or after combustion (41). However, to date, no information is available about the fate of triclosan on sludge. Another point of interest for triclosan risk assessment is its transport into lake sediment, where it seems to be rather persistent (3). To our knowledge, only these data from the 3488
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 16, 2002
Greifensee sediment are available. The risk assessment for triclosan should also be extended to sediments as well.
Acknowledgments The authors are grateful to P. Reichert and A. Wu ¨ est for their help with AQUASIM modeling. They also thank P. Douben, R. van Egmond, A. Hauk, J. Hoigne´, J. Inauen, H.P. Kohler, E. Mu ¨ ller, D. Sabaliunas, R. Schwarzenbach, and M. Whelan for reviewing this manuscript. This work was partially funded by Ciba Specialty Chemicals Inc. (Switzerland), Unilever (United Kingdom), Procter & Gamble (United Kingdom), and the Swiss Federal Office of Environment, Forest and Landscape (BUWAL).
Literature Cited (1) Jungclaus, G. A.; Lopez-Avila, V.; Hites, R. A. Environ. Sci. Technol. 1978, 12, 88-96. (2) Lopez-Avila, V.; Hites, R. A. Environ. Sci. Technol. 1980, 14, 13821390. (3) Singer, H. P.; Mu ¨ ller, S. R.; Tixier, C.; Pillonel, L. Environ. Sci. Technol. (submitted for publication). (4) Sabaliunas, D.; Webb, S. F.; Hauk, A.; Eckhoff, W. S. In 11th SETAC Europe Annual Meeting; Madrid, Spain, 2001. (5) McAvoy, D. C.; Schatowitz, B.; Jacob, M.; Hauk, A.; Eckhoff, W. S. Environ. Toxicol. Chem. (accepted for publication). (6) Orvos, D. R.; Versteeg, D.; Inauen, J.; Capdevielle, M.; Rothenstein, A.; Cunningham, V. Environ. Toxicol. Chem. 2001, accepted for publication. (7) Ciba Specialty Chemical Holding Inc., Irgasan, Irgacare Toxicological and ecological data; Official registration/2521, 1998. (8) McMurry, L. M.; Oethinger, M.; Levy, S. B. Nature 1998, 394, 531-532. (9) Levy, C. W.; Roujeinikova, A.; Sedelnikova, S.; Baker, P. J.; Stuitje, A. R.; Slabas, A. R.; Rice, D. W.; Rafferty, J. B. Nature 1999, 398, 383-384. (10) Cox, A. R. J. Soc. Cosmet. Chem. 1987, 38, 223-231. (11) Zambon, J. J.; Reynolds, H. S.; Dunford, R. G.; DeVizio, W.; Volpe, A. R.; Berta, R.; Tempro, J. P.; Bonta, Y. Oral Microbiol. Immunol. 1995, 10, 247-255. (12) Jones, C. L.; Ritchie, J. A.; Marsh, P. D.; Van der Ouderaa, F. J. Dent. Res. 1988, 67, 46-50. (13) Pillonel, L. Diploma Thesis; Swiss Federal Institute of Technology: Zu ¨ rich, 1999. (14) Audureau, J.; Filiol, C.; Boule, P.; Lemaire, J. J. Chim. Phys. 1976, 73, 613-620. (15) Boule, P.; Guyon, C.; Lemaire, J. Chemosphere 1982, 11, 11791188. (16) Boule, P.; Guyon, C.; Lemaire, J. Chemosphere 1984, 13, 603612. (17) Scully, F. E., jr.; Hoigne´, J. Chemosphere 1987, 16, 681694. (18) Tratnyek, P. G.; Hoigne´, J. Environ. Sci. Technol. 1991, 25, 15961604. (19) Canonica, S.; Jans, U.; Stemmler, K.; Hoigne´, J. Environ. Sci. Technol. 1995, 29, 1822-1831. (20) Mill, T.; Hendry, D. G.; Richardson, H. Science 1980, 207, 886887. (21) Faust, B. C.; Hoigne´, J. Environ. Sci. Technol. 1987, 21, 957-964. (22) Kawaguchi, H. Chemosphere 1992, 25, 635-641. (23) Kawaguchi, H. J. Contam. Hydrol. 1992, 9, 105-114. (24) Gerecke, A. C.; Canonica, S.; Mu ¨ ller, S. R.; Scha¨rer, M.; Schwarzenbach, R. P. Environ. Sci. Technol. 2001, 35, 39153923. (25) Zepp, R. G.; Cline, D. M. Environ. Sci. Technol. 1977, 11, 359366. (26) Wegelin, M.; Canonica, S.; Mechsner, K.; Fleischmann, T.; Pesaro, F.; Metzler, A. J. Water SRT-Aqua 1994, 43, 154-169. (27) Zepp, R. G.; Gumz, M. M.; Miller, W. L.; Gao, H. J. Phys. Chem. A 1998, 102, 5716-5723. (28) Reichert, P. Water Sci. Technol. 1994, 30, 21-30. (29) Imboden, D. M.; Wu ¨ est, A. In Physics and Chemistry of Lakes; Lerman, A., Imboden, D. M., Gat, J. R., Eds.; Springer: New York, U.S.A., 1995; pp 83-138. (30) Kocsis, O.; Prandke, H.; Stips, A.; Simon, A.; Wu ¨ est, A. J. Mar. Syst. 1999, 21, 67-84.
(31) Wu ¨est, A.; Piepke, G.; Van Senden, D. C. Limnol. Oceanogr. 2000, 45, 1388-1400. (32) Meeus, J. Astronomical formulae for calculators; 4th, enlarged and revised ed.; Willmann-Bell, Inc.: Richmond, VA, U.S.A., 1988. (33) http://jwocky.gsfc.nasa.gov/. (34) Kramer, J. B.; Canonica, S.; Hoigne´, J.; Kaschig, J. Environ. Sci. Technol. 1996, 30, 2227-2234. (35) Ulrich, M. M.; Mu ¨ ller, S. R.; Singer, H. P.; Imboden, D. M.; Schwarzenbach, R. P. Environ. Sci. Technol. 1994, 28, 16741685. (36) Frank, R.; Klo¨pffer, W. Chemosphere 1988, 17, 985-994. (37) Poiger, T.; Kari, F. G.; Giger, W. Environ. Sci. Technol. 1999, 33, 533-539. (38) Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. Environmental Organic Chemistry; Wiley: New York, U.S.A., 1993. (39) Kari, F. G.; Giger, W. Environ. Sci. Technol. 1995, 29, 28142827. (40) Kanetoshi, A.; Ogawa, H.; Katsura, E.; Kaneshima, H.; Miura, T. J. Chromatogr. 1988, 454, 145-155. (41) Kanetoshi, A.; Ogawa, H.; Katsura, E.; Kaneshima, H.; Miura, T. J. Chromatogr. 1988, 442, 289-299. (42) Frank, R.; Klo¨pffer, W. Ecotox. Environ. Saf. 1989, 17, 323-332.
(43) NADUF. Messresultate 1977-1998; Bundesamt fu ¨r Umwelt, Wald und Landschaft, Bundesamt fu ¨ r Wasser und Geologie, Eidg. Anstalt fu ¨r Wasserversorgung, Abwasserreinigung und Gewa¨sserschutz: Bern, Switzerland, 2000. (44) Streit, D., Personal communication. (45) Haag, W. R.; Hoigne´, J. Environ. Sci. Technol. 1986, 20, 341348. (46) Freiburghaus, M. Diploma Thesis; Swiss Federal Insitute of Technology, Zu ¨ rich, 1996. (47) Blough, N. V.; Zafiriou, O. C.; Bonilla, J. J. Geophys. Res. 1993, 98, 2271-2278. (48) Forsius, M. In Publications of the Water and Environment Research Institute; National Board of Waters and the Environment: Helsinki, Finland, 1992; Vol. 10, pp 1-37. (49) Forsius, M.; Ka¨ma¨ri, J.; Kortelainen, P.; Mannio, J.; Verta, M.; Kinnunen, K. In Acidification in Finland; Kauppi, P., Anttila, P., Kentta¨mies, K., Ed.; Springer: Berlin, Germany, 1990; pp 759780.
Received for review March 14, 2002. Revised manuscript received May 30, 2002. Accepted June 4, 2002. ES025647T
VOL. 36, NO. 16, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3489
