- Email: [email protected]
Aquatic fate assessment of the polycyclic musk fragrance HHCB
Chemosphere 41 (2000) 671±679
Aquatic fate assessment of the polycyclic musk fragrance HHCB Scenario and variability analysis in accordance with the EU risk assessment guidelines Stefan Schwartz *, Volker Berding, Michael Matthies Institute of Environmental Systems Research, University of Osnabr uck, 49069 Osnabr uck, Germany Received 16 July 1999; accepted 30 September 1999
Abstract By means of the environmental fate and distribution models laid down in the Technical Guidance Documents (TGD) and implemented in the European Union System for the Evaluation of Substances (EUSES) environmental concentrations of the polycyclic musk fragrance HHCB (1,3,4,6,7,8-hexahydro-4,6,6,7,8,8-hexamethyl-cyclopenta-[g]-2-benzopyrane; trade name: e.g. Galaxolideâ ) were calculated for the aquatic environment under consideration of various scenarios. The results were then compared to monitoring data from the region of North Rhine-Westphalia (River Ruhr). An uncertainty analysis was carried out to determine sensitive parameters, to integrate environmental variability and to con®rm the model's calculations. The standard scenario of EUSES overestimates the measured concentrations, which con®rms the conservative nature of the calculations. The regional-speci®c scenarios lead to lower deviations from the measured values than the standard scenario. Deviations range from one to two orders of magnitude in the euent of sewage treatment plants; they amount to one order of magnitude for surface water concentrations on a local scale and conform to monitoring data on a regional scale. The use of measured bioconcentration factors for ®sh instead of estimated ones reduces deviations remarkably. The investigation reveals that unrealistic worst-case calculations of HHCB can at best be ameliorated by the application of more realistic emission rates and measured bioconcentration factors. The use of regional-speci®c parameters also diminishes the deviations of the calculations from the measured concentrations. Ó 2000 Elsevier Science Ltd. All rights reserved. Keywords: Fate assessment; Polycyclic musk fragrances; HHCB; Risk assessment guidelines; TGD; EUSES
1. Introduction Synthetic musk fragrances are an essential ingredient in numerous perfumes, cosmetics and cosmetic care products, soaps, detergents, and other cleaning agents (Ohlo, 1990). Since the early 1990s, not only nitro
*
Corresponding author. Tel.: +49-0541-969-2573; fax: +490541-969-2599. E-mail address: [email protected] (S. Schwartz).
musk compounds (Rimkus and Brunn, 1996), but also polycyclic musk fragrances have been detected in rivers and the sea, ®sh, human adipose tissue and human milk (Eschke et al., 1994, 1995a,b; Rimkus and Wolf, 1996; Bester et al., 1998; Winkler et al., 1998). Some of the polycyclic compounds are the prevailing musk fragrances in Western Europe with concentrations in aquatic biota that exceed those of the nitro musks by up to three orders of magnitude (Gatermann et al., 1999). Seven single compounds are involved, of which HHCB (1,3,4,6,7,8-hexahydro-4,6,6,7,8,8-hexamethyl-
0045-6535/00/$ - see front matter Ó 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 5 - 6 5 3 5 ( 9 9 ) 0 0 5 1 0 - X
672
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
cyclopenta-[g]-2-benzopyrane; CAS-No. 1222-05-5; trade name: e.g. Galaxolideâ ) occurs with the highest concentrations in the environment. As early as the 1970s, scientists became aware of this substance class because of the neurotoxic eects of the representative compound Versalideâ (Spencer et al., 1979). Due to its ubiquitous occurrence in the aquatic environment and its bioaccumulation potential, it has recently returned to the public attention (Mrasek, 1998). Despite its widespread usage, a consumption rate of up to 2400 t/a in Europe, and environmental concentrations of up to 1.2 lg/l in rivers, 63 mg/kg (dry weight) in sewage sludge and 63.6 mg/kg in fat tissue of ®sh (Plassche and Balk, 1997), these chemicals have not yet been investigated suciently. This situation hampers the comprehensive assessment of ecological risks posed by these fragrances. Further investigations are required to obtain a comprehensive evaluation (Eschke et al., 1995a,b). In addition to data concerning the eects of these substances, information on their environmental fate and distribution is also necessary. In 1996 the European Commission published the Technical Guidance Documents (TGD) (EC, 1996a), which enable the assessment of the risk posed by new noti®ed and existing chemicals to humans and the environment. The models and procedures laid down in the TGD have been made available in software form by the European Union System for the Evaluation of Substances (EUSES) (EC, 1996b). EUSES makes default values for most of the parameters available; just four physicochemical substance properties and the tonnage are required as input parameters. The default values claim to represent a hypothetical ``European average region'', the so-called standard region (EC, 1996a). Within the scope of a risk assessment of HHCB by Plassche and Balk (1997), environmental concentrations were assessed according to the TGD. However, their investigation is based only on the standard scenario for the calculation of environmental concentrations. Schwartz et al. (1999) carried out a HHCB fate assessment for dierent scenarios and recommended further assessments. In particular, the need for considering environmental variability was emphasised. Using the example of HHCB, our purpose is to determine how accurately the models, parameters and procedures laid down in the TGD can be applied to an environmental risk assessment for this substance. The deviations of predicted environmental concentrations (PECs) from measured values obtained by monitoring studies are of particular interest. A further objective is to determine, by consideration of various scenarios, the impact of dierent model assumptions on the results and to point out possibilities to reduce over- and underestimations. The state of North Rhine-Westphalia (Germany) was chosen for the comparison of measured with predicted concentrations. This con®rms the strength
of the fate models and contributes to an extensive evaluation of HHCB. A stepwise approach is pursued: Firstly, a scenario analysis identi®es the most realistic scenario, then an uncertainty analysis is carried out for it. 2. Database 2.1. Regional emissions Only a few production and consumption rates of HHCB have been published to date: Ohlo (1990) states an annual production volume of about 1000 t worldwide, whereas Plassche and Balk (1997) quote a consumption of 2400 t (1992) and 1482 t (1995) in Europe, based on data from the producers. The substance is emitted into wastewater during use, minus a negligible loss due to volatilisation (fragrance!). Thereafter, it reaches the surface waters and the aquatic food chain via municipal sewage treatment plants, or agricultural ®elds via sludge application. Dermal uptake is viewed as a major exposure pathway to humans (Rimkus and Brunn, 1996). EUSES estimates the emissions from a given tonnage using emission tables for a certain use category. In this investigation, the tonnage for a region is estimated in two dierent ways: Firstly, it is estimated by EUSES, whereby 10% of the European consumption rate is taken as a regional tonnage, i.e. a rate of 240 t/a is given for 1992. Secondly, the tonnage is calculated on the basis of the average per-capita consumption of 18 mg/d (1992) and 11 mg/d (1995), respectively (Plassche and Balk, 1997), and the number of inhabitants in the region. Thus, consumption rates of 117 t/a (1992) and 72.3 t/a (1995) are obtained for the region of North RhineWestphalia and 14.5 t/a (1992) for the River Ruhr catchment, respectively. Both scenarios assume a continuous emission into surface water. Regional parameters, which dier from the standard scenario, and estimated tonnage for each region are summarised in Table 1. 2.2. Substance-speci®c parameters A minimum of four physico-chemical parameters are necessary to run a calculation (Table 2). Estimated partition coecients and bioconcentration factors may be replaced by more reliable (e.g. measured) values on demand. The physico-chemical data applied in this work are experimentally determined and listed in Table 2. Reasonable data on aquatic degradation have not yet been published, thus, for the purpose of a worst-case estimation, no degradation is assumed. However, Plassche and Balk (1997) allude to a possible aquatic biodegradation. To allow for this eventuality, HHCB is classi®ed as readily biodegradable in one scenario, and a
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
673
Table 1 Characteristics of the investigated regions and estimated tonnage Parameter
Standard (EC, 1996b)
North Rhine-Westphalia (Berding et al., 1999a)
River Ruhr catchment (AWWR, 1997)
Area of region (km2 ) Average annual precipitation (mm/a) Sewage treatment: BOD50 (g/d) Sewage treatment: mode of aeration No. of inhabitants discharging into one STP Fraction connected to STP ()) Depth of water (m) Fraction of water area in region ()) Fraction of water ¯ow from continental scale to the region ()) No. of inhabitants in region Estimated regional tonnage (t/a) 1992 (based on 10% rule) 1992 (based on per-capita cons. 18 mg/d) 1995 (based on per-capita cons. 11 mg/d)
40,000 700 54 Surface 10,000 0.70 3 0.030 0.034
34,400 679 60 Bubble 17,226 0.92 3 0.018 0.029
4488 1056 60 Bubble 22,804 0.95 2 0.018 0.002
2:00 107
1:78 107
2:21 106
117 72.3
14.5
240
Table 2 Physico-chemical data (Plassche and Balk, 1997)a Parameter Molar mass Vapour pressure HenryÕs law constant log KOW log KOC Water solubility BCFfish a
Value 258.4 0.0727 11.3 5.9 4.86 1.75 1584
Unit g/mol Pa, 25°C Pa m3 /mol ) l/kg mg/l, 25°C l/kg
The HHCB content in the Ruhr is comparatively high, with a mean of 0.37 lg/l. In the river Elbe and the river Rhine average concentrations of 0.12 lg/l (Winkler et al., 1998) and 0.07 lg/l (Plassche and Balk, 1997), respectively, were measured. Since EUSES only dierentiates between a local, regional and continental scale, the average monitoring data from the river Ruhr are viewed as regional concentrations from the region of North Rhine-Westphalia.
Parameters marked with * may be estimated by EUSES.
3. Methods half-life of 15 days in bulk surface water and 0.7 h in sewage treatment plants, respectively, was assumed. 2.3. Monitoring data Measured concentrations from the river Ruhr by Eschke et al. (1994, 1995a) are used to compare the model's results with monitoring data. The river Ruhr, located in North Rhine-Westphalia (Germany), is a relatively large river with a length of 217 km, a catchment area of 4488 km2 and an average annual mouth out¯ow of more than two billion m3 (AWWR, 1997). It is of major importance to the drinking water supply for the surrounding conurbation. The measurement of 30 samples at the beginning of 1994 along a stretch of 160 km was carried out downstream from three sewage treatment plant discharges as well as at locations without the direct in¯uence of emitters. Sampling was carried out daily over the period of a week. In the same period, the contamination contained in nine ®sh from the river Ruhr was investigated. The given concentrations refer to the wet weight.
3.1. Models Model calculations were carried out by EUSES according to the equations laid down in the TGD. Various models are linked (Fig. 1). The emission in the vicinity of a point source (here the euent of a sewage treatment plant) is calculated by multiplying the release into water by the fraction of the main local source. The fraction of the main local source provides the fraction of the total volume released that can be assumed to be released through a single point source. The sewage treatment model SimpleTreat (STP) estimates indirect emissions into the environment and the
Fig. 1. Model structure of EUSES.
674
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
concentration in the euent of sewage treatment plants. The PEC of surface waters is then ascertained on a regional and a local scale. The former is calculated by the Level III multimedia model SimpleBox and represents a background concentration. The local concentration is equivalent to the concentration in the euent corrected by a dilution factor plus the regional background concentration. By means of a regression relationship the concentration in ®sh is estimated from the average of local and regional concentrations. We refer to EC (1996a,b) and Berding et al. (1999b) for a detailed presentation of the model structure and equations.
data specifying the region (e.g. area of region, number of inhabitants, fraction of population connected to sewage treatment plants) were aligned where possible. As a comparison the results from the study by Plassche and Balk (1997) are also shown. The dierences in the resulting concentrations and their comparison to measured values enable us to determine to what extent 1. an adjustment of the standard environment towards a selected region and 2. an alteration of model assumptions in¯uence the results, i.e. the concentrations in sewage treatment euent, water and ®sh.
3.2. Scenario analysis
3.3. Sensitivity and variability analysis
In order for EUSES to carry out release estimations it is essential to de®ne the use category. Due to its main area of application as a cleaning/washing additive, HHCB is categorised as such. Alternatives, e.g. cosmetics or odour agents, only have a negligible impact on the results. Besides the standard scenario according to EUSES, four dierent emission and distribution scenarios were calculated for North Rhine-Westphalia (Table 3), whereby in Scenarios NRW1 and NRW2 the amount produced in 1992 is taken to be 2400 t. In Scenario NRW3 the value of 1995 (1482 t) is taken into consideration, whilst in NRW4 a ready biological degradability is assumed. Except for the standard scenario and NRW1 the partition coecients and bioconcentration factors calculated by EUSES are replaced by the measured values in Table 2. The scenario Ruhr reduces the spatial scale to that of the River Ruhr catchment, i.e.
When predicting environmental concentrations the question of their inherent uncertainties arises. Due to the large amount of parameters used in EUSES ± almost 500 parameters within the overall system (Berding et al., 1999b) ± it is advisable to identify those with the largest impact on the model results. For this reason a sensitivity analysis was ®rst carried out. Since EUSES is not capable of performing probabilistic calculations, a spreadsheet version was implemented. Various scenarios for dierent chemicals were investigated to verify that EUSES and the newly developed spreadsheet led to the same results. Furthermore, a sensitivity analysis was performed for both implementations. The same results additionally con®rm that both implementations are equivalent. The sensitivity analysis was carried out for the PECs in the surface water by varying input parameters by 10% and correlating them with the output. Thus, the
Table 3 Survey of the scenariosa
a
Scenario name
Region
EU production volume (t/a)
Regional tonnage (t/a)
Measured partition coecients
Biodegradable
Description/characteristics
Standard NRW1
Standard NRW
2400 2400
240 240
No No
No No
NRW2
NRW
2400
117
Yes
No
NRW3
NRW
1482
72.3
Yes
No
NRW4
NRW
2400
117
Yes
Readily
Ruhr
Ruhr
2400
14.5
Yes
No
P&B
Standard
1482
58.6
Yes
No
Input of the minimal parameter set. Regional-speci®c parameters for the North Rhine-Westphalian environment are used. Regional tonnage is based on per-capita consumption and number of inhabitants. As NRW2, but production volume of 1995 is assumed. As NRW2, but biodegradability is assumed. As NRW2, but regional-speci®c parameters for the river Ruhr catchment are used. Environmental fate assessment carried out by Plassche and Balk (1997).
Regional tonnage is calculated according to Table 1.
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
675
sensitivity S x of a parameter x and model f is expressed by S x f 0 x
x f x10% ÿ f xÿ10% x : f x x10% ÿ xÿ10% f x
In the context of an uncertainty analysis a distinction between uncertainty and variability is commonly claimed. Uncertainty represents a lack of knowledge about factors aecting exposure or risk, whereas variability arises from true heterogeneity across people, places or time (EPA, 1997). This work focuses on variability, which can be successfully dealt with by MonteCarlo simulations. A variability analysis was carried out with special attention to the most sensitive parameters. The analysis was performed by Monte-Carlo simulations using the spreadsheet version and the risk analysis tool Crystal Ball Proâ 4.0 (Decisioneering, 1996). Etienne et al. (1997) investigated the uncertainty of SimpleBox, identi®ed variable parameters, and collated probability distributions. These distributions were used, whereby some parameters were adjusted to the North Rhine-Westphalian environment. When in doubt, a distribution was assigned a rather broad standard deviation, so that the whole spectrum of heterogeneity is covered. The Ruhrverband (1999) provided the results of 35 further samples that were taken from one sampling site from March to December 1994. Based on this total of 65 samples, two log-normal distributions were ®tted to represent the spatial and temporal variability. Finally, this distribution is compared to those obtained from the Monte-Carlo simulation.
4. Results and discussion 4.1. Scenario analysis Sewage treatment euent (Fig. 2): With sewage treatment euent the calculated concentrations are considerably higher than the measured values. The deviations range between 1 and 2 orders of magnitude. Marked overestimations even arise when HHCB is classi®ed as readily biodegradable. Reasons for this could be that emissions have been overestimated, eventual absorption and degradation mechanisms, or the sewage treatment model itself. However, the reasons can only be determined once more detailed information on the behaviour of HHCB in sewage treatment plants is available. The deviations between the scenarios dier by up to two orders of magnitude. A relatively low predicted concentration in the Ruhr scenario is attributable to the reduced consumption rate for the catchment, while the fraction of the main local source is kept con-
Fig. 2. Comparison of calculated concentrations [mg/l] with monitoring data in the euent of sewage treatment plants.
stant by EUSES. Accordingly, the local concentrations in water and ®sh (see below) are also reduced. However, the local concentration for the Ruhr scenario should correspond to that of the NRW2 scenario, since just regional parameters have been changed. To obtain the same concentration, the fraction of the main local source has to be increased by the same ratio by which the consumption rate is decreased. Surface water (Fig. 3): Calculated concentrations in surface water show relatively small deviations. Concentrations calculated on the local scale overestimate the median of the measured values by up to 1.5 orders of magnitude, whilst the dierent results on the regional scale correspond to the range of the measured values. Depending on the scenario, the median of the measured values is either over- or underestimated on a regional scale, while the local model generally overestimates the median in each of the seven scenarios. The median of the concentrations measured in the Ruhr represents a mixture of local and regional concentrations, and can
Fig. 3. Comparison of calculated surface water concentrations (mg/l) with monitoring data (PEC water (local), PEC water (regional)).
676
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
Fig. 4. Comparison of calculated concentrations in ®sh (mg/kg) (wet weight) with monitoring data.
therefore be viewed as an increased regional contamination. This is also con®rmed by the fact that lower contamination was measured in rivers such as the Elbe and the Rhine. On this basis EUSES delivers conservative estimations on a local scale, but realistic estimations on a regional one. However, in the sense of the TGD, the measured values are rather to be classi®ed as regional, since local concentrations are always de®ned by the direct vicinity of an emitter. The variation in the regional scale results of the individual scenarios does not exceed one order of magnitude. It is remarkable that the concentration of the Ruhr scenario equals that of the NRW2 scenario and hence lies between those of the NRW1 and NRW4 scenario. This shows that scaling down the environment from the North Rhine-Westphalian region to the Ruhr catchment only in¯uences the resulting regional concentrations negligibly. The predicted concentrations of the Ruhr scenarios on both scales correspond to the median of the measured values very well. Fish (Fig. 4): If EUSES is left to estimate the BCF, the deviations range up to four orders of magnitude. The cause of this discrepancy is explained by an unrealistically high calculated bioconcentration factor (20,700 l/ kg), which probably arises from disregarding metabolism in the model. But the concentration in ®sh from the Ruhr is even considerably overestimated when a bioconcentration factor gained from experiments is used: with a measured BCF of 1584 l/kg the bioconcentration model for ®sh overestimates the measured values by around two orders of magnitude. Recent works con®rm the inaccuracy of lipophilicity/ bioconcentration potential regression equations regarding HHCB: a measurement reveals a BCF of 624 l/kg, while a comparison of calculated bioconcentration factors leads to values ranging from 2000 to 220,000 l/kg (Gatermann, 1999; Butte and Ewald, 1999). Furthermore, a species-dependent metabolism is assumed which
additionally points out the inexactness of the regression equation laid down in the TGD. With regard to all of the submodels, a slight improvement of the model results can be gained by replacing the standard region by the NRW-speci®c regional data. More precise emission rates and the use of measured partition coecients and bioconcentration factors lead to a considerable improvement of the model results. However, the values used are still aicted by uncertainty: The consumption quantities are based on production ®gures provided by the fragrance industry. These ®gures do not contain information on import and export trade outside the EU. Nor do they include all manufacturers. Furthermore, it is not known whether the reduction in consumption represents a general trend or ¯uctuations over a number of years (Plassche and Balk, 1997). This could be a signi®cant source of uncertainty regarding statements on the comparison of measured vs. predicted data. But the ®gures can be taken as a best estimate, and it is expected that qualitative statements (i.e. over- or underestimations) do not change, because in our scenarios the emission rate shows a linear impact on the predictions and it is expected that variations in emissions do not exceed one order of magnitude. As can be seen in Fig. 3 (regional scale), changing per-capitabased emission rates (NRW2 vs. NRW3) does not in¯uence the results more than any other changes, e.g. replacing emissions calculated by EUSES by per-capitabased emissions (NRW1 vs. NRW2). The classi®cation of the substance as readily degradable leads, of course, to noticeably lower environmental concentrations, but this does not alter the basic statement either. Suspended matter can in¯uence the distribution behaviour of substances considerably. Winkler et al. (1998) ascertained a dissolved fraction of 92% for HHCB in the Elbe with a mean partition coecient between water and suspended matter of 4500 l/kg. The values generated by EUSES are 88% and 1900 l/kg, respectively. According to this, adsorption to suspended matter has a negligible in¯uence on the result. 4.2. Sensitivity and variability analysis An investigation of the interdependency of the EUSES parameters on the basis of the scenarios used in this study reveals that only 15% of all EUSES parameters in¯uence the regional and local water concentrations. These parameters are classi®ed into three categories concerning their sensitivity. One third of the parameters show a minor impact, nearly half with a sensitivity of S x > 0:01 show a moderate impact and the remainder, with a sensitivity of S x > 0:5, have a strong impact on the predicted surface water concentrations. The parameters of the latter category are depicted in Table 4 for two scenarios at the local and regional scale.
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
677
Table 4 Results of the sensitivity analysis for concentrations in surface water for two scenariosa
a
PEC water
Scenario NRW 1 and substance assumed to be not biodegradable
Scenario NRW 1 and substance assumed to be readily biodegradable
Regional
Fraction connected to STP (Fconnect) Area of region (AreaReg) Reg. emission to waste water (ERegWater) Vapour pressure (VP) Water solubility (Sol) Area fraction of water in region (FWaterReg)
Fraction connected to STP (Fconnect) Area of region (AreaReg) Area fraction of water in region (FWaterReg) Reg. emission to surface water (ERegDirectWater)
Local
Dilution factor (Dilution) Local emission to waste water (ELocalWater) No. of inhabitants discharging into one STP (NLocal) Octanol water partitioning (KOW )
Dilution factor (Dilution) Local emission to waste water (ELocalWater) No. of inhabitants discharging into one STP (NLocal) Fraction connected to STP (Fconnect) Total degradation rate in STP (kdegSTP)
The most sensitive parameters S x > 0:5 are listed, nomenclature of EUSES is given in brackets.
The model's results are always sensitive to the emission rates. In addition, the fraction of people connected to sewage treatment plants and the regional area are sensitive parameters for the regional scale. For the local scale the dilution factor and the number of inhabitants discharging into one plant are sensitive parameters. Comparing both scenarios, the impact of some substance properties (vapour pressure, water solubility,
Fig. 5. Variability of monitoring data and the model's results for surface water.
Table 5 Parameters (lg/l) of the ®tted log-normal probability distributions for concentrations in the surface water and the NRW 1 scenario Monitoring data (temporal variability) Monitoring data (spatial variability) PEC regional PEC local
Mean
10%-tile
90%-tile
0.23
0.11
0.38
0.45
0.18
0.81
0.91 1.98
0.62 1.81
1.23 2.16
KOW ) is emphasised, while results for substances classi®ed as readily biodegradable are sensitive to the degradation rate. Special attention was paid to these parameters in the subsequent variability analysis: a broad database is available for a fraction of households connected to STP, so that we were able to construct a custom distribution. The variability in the parameter area is already included in the de®nition of the scenarios. To observe the eect of a more realistic dilution factor the constant default factor of 10 was replaced by an estimation (depending on the ¯ow rate of the river, the number of inhabitants discharging into one STP and the wastewater production rate) as proposed in the TGD. The estimation yields a dilution factor of 2000. The ensuing concentrations are given in Fig. 5. Also the probabilistic view shows an overestimation of the measured values, even if 90-percentiles are compared (Table 5). Thus, the conservative character of the models is con®rmed. The higher mean of the spatial variability is explained by the time of sampling, since concentrations were one of the highest over the year. The large range of the monitoring data distribution is noticeable, especially for the spatial variability. The data also include samples taken downstream from three emitters, but this fact hardly in¯uences the probability distribution (standard deviations are 0.29 and 0.23, respectively). A similar standard deviation for the calculated concentrations may be obtained by elaborating the uncertainty of parameters, which, with respect to the restricted database, was not feasible. However, it can be concluded that the estimated variability in the model cannot account for the discrepancy between observed and modelled concentrations. To rationalise the discrepancy, an uncertainty analysis would be required. The calculated cumulative probability distribution represents a mixture of spatial and temporal variability, since in multimedia models most parameters are aver-
678
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
aged over space and time. An exact statement on the spatial and temporal variability in the model's outcome can only be made by considering geographic-referenced fate models (Matthies et al., 1997). Additionally, the high in¯uence of the dilution factor on the local concentration can be derived from Fig. 5, since the median of the predicted concentrations (0.002 mg/l) is half a log unit lower than the value (0.005 mg/l) obtained in the scenario analysis (Fig. 3, scenario NRW1, local scale), where the default dilution factor was applied. Variable concentrations may arise from the river's varying ¯ow rates. Taking the percentile values of the measured data into account, it can be justi®ed that the variability will not exceed one log unit.
5. Conclusions In principle for the aquatic environment, EUSES can be employed for the investigated fragrance HHCB. Predicted concentrations are sensitive to only a few parameters. The individual EUSES submodels deliver dierent deviations from the measured values and contain numerous sources of uncertainty. The consequence for EUSES users is that the program does not necessarily deliver conservative results, although estimates on a regional scale are usually conservative. Concentrations may arise in the waters of one region that do not represent an extremely high local contamination, but which are still underestimated by the model. However, the uncertainties related to this are low for concentrations in the surface water, as long as emissions based on percapita consumption are used. The uncertainty lies within one order of magnitude. Deviations of up to three orders of magnitude have to be reckoned with for sewage treatment euent and biota in comparison to measured values. It is recommended not to use the standard environment for the investigated region, but it should be taken into consideration that more realistic emission rates and bioconcentration factors in¯uence the model results more noticeably. One may suppose that the models describing the regional distribution as laid down in the TGD are invariant regarding the scaling of the underlying region from large-scale to low-scale and vice versa for fragrances such as HHCB. The use of the programÕs standard estimation of the regional tonnage leads to the greatest deviations and is not recommended. In view of these results, the application of EUSES to other polycyclic musk fragrances seems to be possible, since AHTN (e.g. Tonalideâ ) or ADBI (e.g. Celostolideâ ) show similar environmental behaviour. Furthermore, correlations between these substances are reported (Winkler et al., 1998), due to their use in mixtures. Individual statements for the other substances will
either be more or less appropriate, depending on the respective sorption or accumulation behaviour. For concentrations in water on a certain scale the variations between the individual scenarios are smaller than the dierence between the minimum and maximum of the measured values. From the comparison of the concentrations of the three investigated regions it can be concluded that an exact selection of representative environmental segments, including an adjustment of the emission rates to the selected region, leads to the most realistic modelling of the contamination of the rivers. This requires closer scrutiny of the monitoring data and their spatial classi®cation, which could be carried out more easily using a geographical referenced information system. In this paper, we dealt with potential uncertainties in an aquatic fate assessment using dierent scenarios and by investigating the variability. An analysis of the uncertainties of substance parameters or a detailed investigation into the sewage treatment process of HHCB was omitted; the necessary database for such research is not yet available. Acknowledgements Funding by the German Federal Environmental Agency, Berlin is gratefully acknowledged (R+D project FKZ 206 01 075). We also thank H.D. Eschke from the Ruhrverband, Essen for his support and the anonymous reviewers for their helpful comments. References AWWR, 1997. Ruhrwasserg utebericht. The study group of the waterworks at the Ruhr (AWWR) and Ruhrverband. Essen. Berding, V., Schwartz, S., Matthies, M., 1999a. Scenario analysis of a Level III multimedia model using generic and regional data. Environ. Sci. & Pollut. Res., accepted. Berding, V., Schwartz, S., Matthies, M., 1999b. Visualisation of the Complexity of EUSES. Environ. Sci. & Pollut. Res. 6, 37±43. Bester, K., H uhnerfuss, H., Lange, W., Rimkus, G., Theobald, N., 1998. Results of nontarget screening of lipophilic organic pollutants in the German Bight II: polycyclic musk fragrances. Water Res. 32, 1857±1863. Butte, Ewald, 1999. Kinetics of accumulation and clearance of the polycyclic musk compounds galaxolide and tonalide. Abstract book of the Ninth Annual Meeting of SETACEurope, 2o/P16. Society of Environmental Toxicology and Chemistry, Brussels. Decisioneering, 1996. Crystal Ballâ Version 4.0 User Manual. Decisioneering, Inc. Denver, Colorado. EC, 1996a. Technical Guidance Document in Support of the Commission Directive 93/67/EEC on Risk Assessment for New Noti®ed Substances and the Commission Regulation (EC) 1488/94 on Risk Assessment for Existing Substances,
S. Schwartz et al. / Chemosphere 41 (2000) 671±679 Parts I±IV. Oce for Ocial Publications of the European Communities, Luxembourg. EC, 1996b. EUSES ± the European Union System for the Evaluation of Substances. Institute of Public Health and the Environment (RIVM), The Netherlands. Available from European Chemicals Bureau, Ispra. EPA, 1997. Exposure Factors Handbook, Vol. I-III. EPA/600/ P-95/002Fa, Oce of Research and Development, US Environmental Protection Agency, Washington. Eschke, H.D., Traud, J., Dibowski, H.J., 1994. Untersuchungen zum Vorkommen polycyclischer Moschus-Duftstoe in verschiedenen Umweltkompartimenten ± Nachweis und Analytik mit GC/MS in Ober¯ achen-, Abw assern und Fischen (1. Mitteilung). Z. Umweltchem. Okotox. 6, 183± 189. Eschke, H.D., Traud, J., Dibowski, H.J., 1995a. Untersuchungen zum Vorkommen polycyclischer Moschus-Duftstoe in verschiedenen Umweltkompartimenten ± Befunde in Ober¯achen-, Abwassern und Fischen sowie in Waschmitteln und Kosmetika (2. Mitteilung). Z. Umweltchem. Okotox. 7, 131±138. Eschke, H.D., Traud, J., Dibowski, H.J., 1995b. Nachweis und Quanti®zierung von polycyclischen Moschus-Duftstoen mittels Ion-Trap GC/MS/MS in Humanfett und Muttermilch. Dt. Lebensmittel-Rundschau 12, 375±379. Etienne, R.S., Ragas, A., vandeMeent, D., 1997. Operational Uncertainties in SimpleBox. National Institute of Public Health and the Environment (RIVM), Report 719101 031, Bilthoven. Gatermann, R., 1999. Bioaccumulation and enantioselective degradation of synthetic musks in dierent aquatic species. Abstract book of the Ninth Annual Meeting of SETACEurope, 2o/P07, Society of Environmental Toxicology and Chemistry, Brussels. Gatermann, R., Hellou, J., H uhnerfuss, H., Rimkus, G., Zitko, V., 1999. Polycyclic and nitro musks in the environment: a comparison between canadian and european aquatic biota. Chemosphere 38, 3431±3441.
679
Matthies, M., Wagner, J., Koormann, F., 1997. Combination of regional exposure models for european rivers with GIS information. In: Hrsg. Alef et al. K. (Ed.), Proceedings of ECO-INFORMAÕ97, Eco-Informa Press, Bayreuth, vol. 12, pp. 523±529. Mrasek, V., 1998. Duftstoe, die unter die Haut gehen. S uddeutsche Zeitung, 22, January 1998. Ohlo, G., 1990. Riechstoe und Geruchssinn: Die molekulare Welt der D ufte. Springer, Heidelberg. Plassche, E.J.van de, Balk, F., 1997. Environmental risk assessment of the polycyclic musks AHTN and HHCB according to the EU-TGD. National Institute of Public Health and the Environment (RIVM), Report 601503008, Bilthoven. Rimkus, G., Brunn, H., 1996. Synthetische Moschusduftstoe ± Anwendung, Anreicherung in der Umwelt und Toxikologie; Teil 1: Herstellung, Anwendung, Vorkommen in Lebensmitteln, Aufnahme durch den Menschen. Ern ahrungsUmschau 43, 442±449. Rimkus, G., Wolf, M., 1996. Polycyclic musk fragrances in human adipose tissue and milk. Chemosphere 33, 2033± 2043. Ruhrverband, 1999. Results of Monitoring of Polycyclic Musk Fragrances in River Ruhr ± Personal communication. Ruhrverband (Institution for River Ruhr water management), Essen. Schwartz, S., Berding, V., Matthies, M., 1999. Umweltexpositionsabsch atzung des polycyclischen MoschusDuftstoes HHCB ± Szenarienanalyse mit EUSES. Umweltmed. Forsch. Prax. 4, 7±11. Spencer, P.S., Sterman, A.B., Horopian, D.S., Foulds, M.M., 1979. Neurotoxic fragrance produces ceroid and Myelin Disease. Science 20, 633±635. Winkler, M., Kopf, G., Hauptvogel, C., Neu, T., 1998. Fate of arti®cial musk fragrances associated with suspended particulate matter spm from the river elbe germany in comparison to other organic contaminants. Chemosphere 37, 1139±1158.
Aquatic fate assessment of the polycyclic musk fragrance HHCB Scenario and variability analysis in accordance with the EU risk assessment guidelines Stefan Schwartz *, Volker Berding, Michael Matthies Institute of Environmental Systems Research, University of Osnabr uck, 49069 Osnabr uck, Germany Received 16 July 1999; accepted 30 September 1999
Abstract By means of the environmental fate and distribution models laid down in the Technical Guidance Documents (TGD) and implemented in the European Union System for the Evaluation of Substances (EUSES) environmental concentrations of the polycyclic musk fragrance HHCB (1,3,4,6,7,8-hexahydro-4,6,6,7,8,8-hexamethyl-cyclopenta-[g]-2-benzopyrane; trade name: e.g. Galaxolideâ ) were calculated for the aquatic environment under consideration of various scenarios. The results were then compared to monitoring data from the region of North Rhine-Westphalia (River Ruhr). An uncertainty analysis was carried out to determine sensitive parameters, to integrate environmental variability and to con®rm the model's calculations. The standard scenario of EUSES overestimates the measured concentrations, which con®rms the conservative nature of the calculations. The regional-speci®c scenarios lead to lower deviations from the measured values than the standard scenario. Deviations range from one to two orders of magnitude in the euent of sewage treatment plants; they amount to one order of magnitude for surface water concentrations on a local scale and conform to monitoring data on a regional scale. The use of measured bioconcentration factors for ®sh instead of estimated ones reduces deviations remarkably. The investigation reveals that unrealistic worst-case calculations of HHCB can at best be ameliorated by the application of more realistic emission rates and measured bioconcentration factors. The use of regional-speci®c parameters also diminishes the deviations of the calculations from the measured concentrations. Ó 2000 Elsevier Science Ltd. All rights reserved. Keywords: Fate assessment; Polycyclic musk fragrances; HHCB; Risk assessment guidelines; TGD; EUSES
1. Introduction Synthetic musk fragrances are an essential ingredient in numerous perfumes, cosmetics and cosmetic care products, soaps, detergents, and other cleaning agents (Ohlo, 1990). Since the early 1990s, not only nitro
*
Corresponding author. Tel.: +49-0541-969-2573; fax: +490541-969-2599. E-mail address: [email protected] (S. Schwartz).
musk compounds (Rimkus and Brunn, 1996), but also polycyclic musk fragrances have been detected in rivers and the sea, ®sh, human adipose tissue and human milk (Eschke et al., 1994, 1995a,b; Rimkus and Wolf, 1996; Bester et al., 1998; Winkler et al., 1998). Some of the polycyclic compounds are the prevailing musk fragrances in Western Europe with concentrations in aquatic biota that exceed those of the nitro musks by up to three orders of magnitude (Gatermann et al., 1999). Seven single compounds are involved, of which HHCB (1,3,4,6,7,8-hexahydro-4,6,6,7,8,8-hexamethyl-
0045-6535/00/$ - see front matter Ó 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 5 - 6 5 3 5 ( 9 9 ) 0 0 5 1 0 - X
672
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
cyclopenta-[g]-2-benzopyrane; CAS-No. 1222-05-5; trade name: e.g. Galaxolideâ ) occurs with the highest concentrations in the environment. As early as the 1970s, scientists became aware of this substance class because of the neurotoxic eects of the representative compound Versalideâ (Spencer et al., 1979). Due to its ubiquitous occurrence in the aquatic environment and its bioaccumulation potential, it has recently returned to the public attention (Mrasek, 1998). Despite its widespread usage, a consumption rate of up to 2400 t/a in Europe, and environmental concentrations of up to 1.2 lg/l in rivers, 63 mg/kg (dry weight) in sewage sludge and 63.6 mg/kg in fat tissue of ®sh (Plassche and Balk, 1997), these chemicals have not yet been investigated suciently. This situation hampers the comprehensive assessment of ecological risks posed by these fragrances. Further investigations are required to obtain a comprehensive evaluation (Eschke et al., 1995a,b). In addition to data concerning the eects of these substances, information on their environmental fate and distribution is also necessary. In 1996 the European Commission published the Technical Guidance Documents (TGD) (EC, 1996a), which enable the assessment of the risk posed by new noti®ed and existing chemicals to humans and the environment. The models and procedures laid down in the TGD have been made available in software form by the European Union System for the Evaluation of Substances (EUSES) (EC, 1996b). EUSES makes default values for most of the parameters available; just four physicochemical substance properties and the tonnage are required as input parameters. The default values claim to represent a hypothetical ``European average region'', the so-called standard region (EC, 1996a). Within the scope of a risk assessment of HHCB by Plassche and Balk (1997), environmental concentrations were assessed according to the TGD. However, their investigation is based only on the standard scenario for the calculation of environmental concentrations. Schwartz et al. (1999) carried out a HHCB fate assessment for dierent scenarios and recommended further assessments. In particular, the need for considering environmental variability was emphasised. Using the example of HHCB, our purpose is to determine how accurately the models, parameters and procedures laid down in the TGD can be applied to an environmental risk assessment for this substance. The deviations of predicted environmental concentrations (PECs) from measured values obtained by monitoring studies are of particular interest. A further objective is to determine, by consideration of various scenarios, the impact of dierent model assumptions on the results and to point out possibilities to reduce over- and underestimations. The state of North Rhine-Westphalia (Germany) was chosen for the comparison of measured with predicted concentrations. This con®rms the strength
of the fate models and contributes to an extensive evaluation of HHCB. A stepwise approach is pursued: Firstly, a scenario analysis identi®es the most realistic scenario, then an uncertainty analysis is carried out for it. 2. Database 2.1. Regional emissions Only a few production and consumption rates of HHCB have been published to date: Ohlo (1990) states an annual production volume of about 1000 t worldwide, whereas Plassche and Balk (1997) quote a consumption of 2400 t (1992) and 1482 t (1995) in Europe, based on data from the producers. The substance is emitted into wastewater during use, minus a negligible loss due to volatilisation (fragrance!). Thereafter, it reaches the surface waters and the aquatic food chain via municipal sewage treatment plants, or agricultural ®elds via sludge application. Dermal uptake is viewed as a major exposure pathway to humans (Rimkus and Brunn, 1996). EUSES estimates the emissions from a given tonnage using emission tables for a certain use category. In this investigation, the tonnage for a region is estimated in two dierent ways: Firstly, it is estimated by EUSES, whereby 10% of the European consumption rate is taken as a regional tonnage, i.e. a rate of 240 t/a is given for 1992. Secondly, the tonnage is calculated on the basis of the average per-capita consumption of 18 mg/d (1992) and 11 mg/d (1995), respectively (Plassche and Balk, 1997), and the number of inhabitants in the region. Thus, consumption rates of 117 t/a (1992) and 72.3 t/a (1995) are obtained for the region of North RhineWestphalia and 14.5 t/a (1992) for the River Ruhr catchment, respectively. Both scenarios assume a continuous emission into surface water. Regional parameters, which dier from the standard scenario, and estimated tonnage for each region are summarised in Table 1. 2.2. Substance-speci®c parameters A minimum of four physico-chemical parameters are necessary to run a calculation (Table 2). Estimated partition coecients and bioconcentration factors may be replaced by more reliable (e.g. measured) values on demand. The physico-chemical data applied in this work are experimentally determined and listed in Table 2. Reasonable data on aquatic degradation have not yet been published, thus, for the purpose of a worst-case estimation, no degradation is assumed. However, Plassche and Balk (1997) allude to a possible aquatic biodegradation. To allow for this eventuality, HHCB is classi®ed as readily biodegradable in one scenario, and a
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
673
Table 1 Characteristics of the investigated regions and estimated tonnage Parameter
Standard (EC, 1996b)
North Rhine-Westphalia (Berding et al., 1999a)
River Ruhr catchment (AWWR, 1997)
Area of region (km2 ) Average annual precipitation (mm/a) Sewage treatment: BOD50 (g/d) Sewage treatment: mode of aeration No. of inhabitants discharging into one STP Fraction connected to STP ()) Depth of water (m) Fraction of water area in region ()) Fraction of water ¯ow from continental scale to the region ()) No. of inhabitants in region Estimated regional tonnage (t/a) 1992 (based on 10% rule) 1992 (based on per-capita cons. 18 mg/d) 1995 (based on per-capita cons. 11 mg/d)
40,000 700 54 Surface 10,000 0.70 3 0.030 0.034
34,400 679 60 Bubble 17,226 0.92 3 0.018 0.029
4488 1056 60 Bubble 22,804 0.95 2 0.018 0.002
2:00 107
1:78 107
2:21 106
117 72.3
14.5
240
Table 2 Physico-chemical data (Plassche and Balk, 1997)a Parameter Molar mass Vapour pressure HenryÕs law constant log KOW log KOC Water solubility BCFfish a
Value 258.4 0.0727 11.3 5.9 4.86 1.75 1584
Unit g/mol Pa, 25°C Pa m3 /mol ) l/kg mg/l, 25°C l/kg
The HHCB content in the Ruhr is comparatively high, with a mean of 0.37 lg/l. In the river Elbe and the river Rhine average concentrations of 0.12 lg/l (Winkler et al., 1998) and 0.07 lg/l (Plassche and Balk, 1997), respectively, were measured. Since EUSES only dierentiates between a local, regional and continental scale, the average monitoring data from the river Ruhr are viewed as regional concentrations from the region of North Rhine-Westphalia.
Parameters marked with * may be estimated by EUSES.
3. Methods half-life of 15 days in bulk surface water and 0.7 h in sewage treatment plants, respectively, was assumed. 2.3. Monitoring data Measured concentrations from the river Ruhr by Eschke et al. (1994, 1995a) are used to compare the model's results with monitoring data. The river Ruhr, located in North Rhine-Westphalia (Germany), is a relatively large river with a length of 217 km, a catchment area of 4488 km2 and an average annual mouth out¯ow of more than two billion m3 (AWWR, 1997). It is of major importance to the drinking water supply for the surrounding conurbation. The measurement of 30 samples at the beginning of 1994 along a stretch of 160 km was carried out downstream from three sewage treatment plant discharges as well as at locations without the direct in¯uence of emitters. Sampling was carried out daily over the period of a week. In the same period, the contamination contained in nine ®sh from the river Ruhr was investigated. The given concentrations refer to the wet weight.
3.1. Models Model calculations were carried out by EUSES according to the equations laid down in the TGD. Various models are linked (Fig. 1). The emission in the vicinity of a point source (here the euent of a sewage treatment plant) is calculated by multiplying the release into water by the fraction of the main local source. The fraction of the main local source provides the fraction of the total volume released that can be assumed to be released through a single point source. The sewage treatment model SimpleTreat (STP) estimates indirect emissions into the environment and the
Fig. 1. Model structure of EUSES.
674
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
concentration in the euent of sewage treatment plants. The PEC of surface waters is then ascertained on a regional and a local scale. The former is calculated by the Level III multimedia model SimpleBox and represents a background concentration. The local concentration is equivalent to the concentration in the euent corrected by a dilution factor plus the regional background concentration. By means of a regression relationship the concentration in ®sh is estimated from the average of local and regional concentrations. We refer to EC (1996a,b) and Berding et al. (1999b) for a detailed presentation of the model structure and equations.
data specifying the region (e.g. area of region, number of inhabitants, fraction of population connected to sewage treatment plants) were aligned where possible. As a comparison the results from the study by Plassche and Balk (1997) are also shown. The dierences in the resulting concentrations and their comparison to measured values enable us to determine to what extent 1. an adjustment of the standard environment towards a selected region and 2. an alteration of model assumptions in¯uence the results, i.e. the concentrations in sewage treatment euent, water and ®sh.
3.2. Scenario analysis
3.3. Sensitivity and variability analysis
In order for EUSES to carry out release estimations it is essential to de®ne the use category. Due to its main area of application as a cleaning/washing additive, HHCB is categorised as such. Alternatives, e.g. cosmetics or odour agents, only have a negligible impact on the results. Besides the standard scenario according to EUSES, four dierent emission and distribution scenarios were calculated for North Rhine-Westphalia (Table 3), whereby in Scenarios NRW1 and NRW2 the amount produced in 1992 is taken to be 2400 t. In Scenario NRW3 the value of 1995 (1482 t) is taken into consideration, whilst in NRW4 a ready biological degradability is assumed. Except for the standard scenario and NRW1 the partition coecients and bioconcentration factors calculated by EUSES are replaced by the measured values in Table 2. The scenario Ruhr reduces the spatial scale to that of the River Ruhr catchment, i.e.
When predicting environmental concentrations the question of their inherent uncertainties arises. Due to the large amount of parameters used in EUSES ± almost 500 parameters within the overall system (Berding et al., 1999b) ± it is advisable to identify those with the largest impact on the model results. For this reason a sensitivity analysis was ®rst carried out. Since EUSES is not capable of performing probabilistic calculations, a spreadsheet version was implemented. Various scenarios for dierent chemicals were investigated to verify that EUSES and the newly developed spreadsheet led to the same results. Furthermore, a sensitivity analysis was performed for both implementations. The same results additionally con®rm that both implementations are equivalent. The sensitivity analysis was carried out for the PECs in the surface water by varying input parameters by 10% and correlating them with the output. Thus, the
Table 3 Survey of the scenariosa
a
Scenario name
Region
EU production volume (t/a)
Regional tonnage (t/a)
Measured partition coecients
Biodegradable
Description/characteristics
Standard NRW1
Standard NRW
2400 2400
240 240
No No
No No
NRW2
NRW
2400
117
Yes
No
NRW3
NRW
1482
72.3
Yes
No
NRW4
NRW
2400
117
Yes
Readily
Ruhr
Ruhr
2400
14.5
Yes
No
P&B
Standard
1482
58.6
Yes
No
Input of the minimal parameter set. Regional-speci®c parameters for the North Rhine-Westphalian environment are used. Regional tonnage is based on per-capita consumption and number of inhabitants. As NRW2, but production volume of 1995 is assumed. As NRW2, but biodegradability is assumed. As NRW2, but regional-speci®c parameters for the river Ruhr catchment are used. Environmental fate assessment carried out by Plassche and Balk (1997).
Regional tonnage is calculated according to Table 1.
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
675
sensitivity S x of a parameter x and model f is expressed by S x f 0 x
x f x10% ÿ f xÿ10% x : f x x10% ÿ xÿ10% f x
In the context of an uncertainty analysis a distinction between uncertainty and variability is commonly claimed. Uncertainty represents a lack of knowledge about factors aecting exposure or risk, whereas variability arises from true heterogeneity across people, places or time (EPA, 1997). This work focuses on variability, which can be successfully dealt with by MonteCarlo simulations. A variability analysis was carried out with special attention to the most sensitive parameters. The analysis was performed by Monte-Carlo simulations using the spreadsheet version and the risk analysis tool Crystal Ball Proâ 4.0 (Decisioneering, 1996). Etienne et al. (1997) investigated the uncertainty of SimpleBox, identi®ed variable parameters, and collated probability distributions. These distributions were used, whereby some parameters were adjusted to the North Rhine-Westphalian environment. When in doubt, a distribution was assigned a rather broad standard deviation, so that the whole spectrum of heterogeneity is covered. The Ruhrverband (1999) provided the results of 35 further samples that were taken from one sampling site from March to December 1994. Based on this total of 65 samples, two log-normal distributions were ®tted to represent the spatial and temporal variability. Finally, this distribution is compared to those obtained from the Monte-Carlo simulation.
4. Results and discussion 4.1. Scenario analysis Sewage treatment euent (Fig. 2): With sewage treatment euent the calculated concentrations are considerably higher than the measured values. The deviations range between 1 and 2 orders of magnitude. Marked overestimations even arise when HHCB is classi®ed as readily biodegradable. Reasons for this could be that emissions have been overestimated, eventual absorption and degradation mechanisms, or the sewage treatment model itself. However, the reasons can only be determined once more detailed information on the behaviour of HHCB in sewage treatment plants is available. The deviations between the scenarios dier by up to two orders of magnitude. A relatively low predicted concentration in the Ruhr scenario is attributable to the reduced consumption rate for the catchment, while the fraction of the main local source is kept con-
Fig. 2. Comparison of calculated concentrations [mg/l] with monitoring data in the euent of sewage treatment plants.
stant by EUSES. Accordingly, the local concentrations in water and ®sh (see below) are also reduced. However, the local concentration for the Ruhr scenario should correspond to that of the NRW2 scenario, since just regional parameters have been changed. To obtain the same concentration, the fraction of the main local source has to be increased by the same ratio by which the consumption rate is decreased. Surface water (Fig. 3): Calculated concentrations in surface water show relatively small deviations. Concentrations calculated on the local scale overestimate the median of the measured values by up to 1.5 orders of magnitude, whilst the dierent results on the regional scale correspond to the range of the measured values. Depending on the scenario, the median of the measured values is either over- or underestimated on a regional scale, while the local model generally overestimates the median in each of the seven scenarios. The median of the concentrations measured in the Ruhr represents a mixture of local and regional concentrations, and can
Fig. 3. Comparison of calculated surface water concentrations (mg/l) with monitoring data (PEC water (local), PEC water (regional)).
676
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
Fig. 4. Comparison of calculated concentrations in ®sh (mg/kg) (wet weight) with monitoring data.
therefore be viewed as an increased regional contamination. This is also con®rmed by the fact that lower contamination was measured in rivers such as the Elbe and the Rhine. On this basis EUSES delivers conservative estimations on a local scale, but realistic estimations on a regional one. However, in the sense of the TGD, the measured values are rather to be classi®ed as regional, since local concentrations are always de®ned by the direct vicinity of an emitter. The variation in the regional scale results of the individual scenarios does not exceed one order of magnitude. It is remarkable that the concentration of the Ruhr scenario equals that of the NRW2 scenario and hence lies between those of the NRW1 and NRW4 scenario. This shows that scaling down the environment from the North Rhine-Westphalian region to the Ruhr catchment only in¯uences the resulting regional concentrations negligibly. The predicted concentrations of the Ruhr scenarios on both scales correspond to the median of the measured values very well. Fish (Fig. 4): If EUSES is left to estimate the BCF, the deviations range up to four orders of magnitude. The cause of this discrepancy is explained by an unrealistically high calculated bioconcentration factor (20,700 l/ kg), which probably arises from disregarding metabolism in the model. But the concentration in ®sh from the Ruhr is even considerably overestimated when a bioconcentration factor gained from experiments is used: with a measured BCF of 1584 l/kg the bioconcentration model for ®sh overestimates the measured values by around two orders of magnitude. Recent works con®rm the inaccuracy of lipophilicity/ bioconcentration potential regression equations regarding HHCB: a measurement reveals a BCF of 624 l/kg, while a comparison of calculated bioconcentration factors leads to values ranging from 2000 to 220,000 l/kg (Gatermann, 1999; Butte and Ewald, 1999). Furthermore, a species-dependent metabolism is assumed which
additionally points out the inexactness of the regression equation laid down in the TGD. With regard to all of the submodels, a slight improvement of the model results can be gained by replacing the standard region by the NRW-speci®c regional data. More precise emission rates and the use of measured partition coecients and bioconcentration factors lead to a considerable improvement of the model results. However, the values used are still aicted by uncertainty: The consumption quantities are based on production ®gures provided by the fragrance industry. These ®gures do not contain information on import and export trade outside the EU. Nor do they include all manufacturers. Furthermore, it is not known whether the reduction in consumption represents a general trend or ¯uctuations over a number of years (Plassche and Balk, 1997). This could be a signi®cant source of uncertainty regarding statements on the comparison of measured vs. predicted data. But the ®gures can be taken as a best estimate, and it is expected that qualitative statements (i.e. over- or underestimations) do not change, because in our scenarios the emission rate shows a linear impact on the predictions and it is expected that variations in emissions do not exceed one order of magnitude. As can be seen in Fig. 3 (regional scale), changing per-capitabased emission rates (NRW2 vs. NRW3) does not in¯uence the results more than any other changes, e.g. replacing emissions calculated by EUSES by per-capitabased emissions (NRW1 vs. NRW2). The classi®cation of the substance as readily degradable leads, of course, to noticeably lower environmental concentrations, but this does not alter the basic statement either. Suspended matter can in¯uence the distribution behaviour of substances considerably. Winkler et al. (1998) ascertained a dissolved fraction of 92% for HHCB in the Elbe with a mean partition coecient between water and suspended matter of 4500 l/kg. The values generated by EUSES are 88% and 1900 l/kg, respectively. According to this, adsorption to suspended matter has a negligible in¯uence on the result. 4.2. Sensitivity and variability analysis An investigation of the interdependency of the EUSES parameters on the basis of the scenarios used in this study reveals that only 15% of all EUSES parameters in¯uence the regional and local water concentrations. These parameters are classi®ed into three categories concerning their sensitivity. One third of the parameters show a minor impact, nearly half with a sensitivity of S x > 0:01 show a moderate impact and the remainder, with a sensitivity of S x > 0:5, have a strong impact on the predicted surface water concentrations. The parameters of the latter category are depicted in Table 4 for two scenarios at the local and regional scale.
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
677
Table 4 Results of the sensitivity analysis for concentrations in surface water for two scenariosa
a
PEC water
Scenario NRW 1 and substance assumed to be not biodegradable
Scenario NRW 1 and substance assumed to be readily biodegradable
Regional
Fraction connected to STP (Fconnect) Area of region (AreaReg) Reg. emission to waste water (ERegWater) Vapour pressure (VP) Water solubility (Sol) Area fraction of water in region (FWaterReg)
Fraction connected to STP (Fconnect) Area of region (AreaReg) Area fraction of water in region (FWaterReg) Reg. emission to surface water (ERegDirectWater)
Local
Dilution factor (Dilution) Local emission to waste water (ELocalWater) No. of inhabitants discharging into one STP (NLocal) Octanol water partitioning (KOW )
Dilution factor (Dilution) Local emission to waste water (ELocalWater) No. of inhabitants discharging into one STP (NLocal) Fraction connected to STP (Fconnect) Total degradation rate in STP (kdegSTP)
The most sensitive parameters S x > 0:5 are listed, nomenclature of EUSES is given in brackets.
The model's results are always sensitive to the emission rates. In addition, the fraction of people connected to sewage treatment plants and the regional area are sensitive parameters for the regional scale. For the local scale the dilution factor and the number of inhabitants discharging into one plant are sensitive parameters. Comparing both scenarios, the impact of some substance properties (vapour pressure, water solubility,
Fig. 5. Variability of monitoring data and the model's results for surface water.
Table 5 Parameters (lg/l) of the ®tted log-normal probability distributions for concentrations in the surface water and the NRW 1 scenario Monitoring data (temporal variability) Monitoring data (spatial variability) PEC regional PEC local
Mean
10%-tile
90%-tile
0.23
0.11
0.38
0.45
0.18
0.81
0.91 1.98
0.62 1.81
1.23 2.16
KOW ) is emphasised, while results for substances classi®ed as readily biodegradable are sensitive to the degradation rate. Special attention was paid to these parameters in the subsequent variability analysis: a broad database is available for a fraction of households connected to STP, so that we were able to construct a custom distribution. The variability in the parameter area is already included in the de®nition of the scenarios. To observe the eect of a more realistic dilution factor the constant default factor of 10 was replaced by an estimation (depending on the ¯ow rate of the river, the number of inhabitants discharging into one STP and the wastewater production rate) as proposed in the TGD. The estimation yields a dilution factor of 2000. The ensuing concentrations are given in Fig. 5. Also the probabilistic view shows an overestimation of the measured values, even if 90-percentiles are compared (Table 5). Thus, the conservative character of the models is con®rmed. The higher mean of the spatial variability is explained by the time of sampling, since concentrations were one of the highest over the year. The large range of the monitoring data distribution is noticeable, especially for the spatial variability. The data also include samples taken downstream from three emitters, but this fact hardly in¯uences the probability distribution (standard deviations are 0.29 and 0.23, respectively). A similar standard deviation for the calculated concentrations may be obtained by elaborating the uncertainty of parameters, which, with respect to the restricted database, was not feasible. However, it can be concluded that the estimated variability in the model cannot account for the discrepancy between observed and modelled concentrations. To rationalise the discrepancy, an uncertainty analysis would be required. The calculated cumulative probability distribution represents a mixture of spatial and temporal variability, since in multimedia models most parameters are aver-
678
S. Schwartz et al. / Chemosphere 41 (2000) 671±679
aged over space and time. An exact statement on the spatial and temporal variability in the model's outcome can only be made by considering geographic-referenced fate models (Matthies et al., 1997). Additionally, the high in¯uence of the dilution factor on the local concentration can be derived from Fig. 5, since the median of the predicted concentrations (0.002 mg/l) is half a log unit lower than the value (0.005 mg/l) obtained in the scenario analysis (Fig. 3, scenario NRW1, local scale), where the default dilution factor was applied. Variable concentrations may arise from the river's varying ¯ow rates. Taking the percentile values of the measured data into account, it can be justi®ed that the variability will not exceed one log unit.
5. Conclusions In principle for the aquatic environment, EUSES can be employed for the investigated fragrance HHCB. Predicted concentrations are sensitive to only a few parameters. The individual EUSES submodels deliver dierent deviations from the measured values and contain numerous sources of uncertainty. The consequence for EUSES users is that the program does not necessarily deliver conservative results, although estimates on a regional scale are usually conservative. Concentrations may arise in the waters of one region that do not represent an extremely high local contamination, but which are still underestimated by the model. However, the uncertainties related to this are low for concentrations in the surface water, as long as emissions based on percapita consumption are used. The uncertainty lies within one order of magnitude. Deviations of up to three orders of magnitude have to be reckoned with for sewage treatment euent and biota in comparison to measured values. It is recommended not to use the standard environment for the investigated region, but it should be taken into consideration that more realistic emission rates and bioconcentration factors in¯uence the model results more noticeably. One may suppose that the models describing the regional distribution as laid down in the TGD are invariant regarding the scaling of the underlying region from large-scale to low-scale and vice versa for fragrances such as HHCB. The use of the programÕs standard estimation of the regional tonnage leads to the greatest deviations and is not recommended. In view of these results, the application of EUSES to other polycyclic musk fragrances seems to be possible, since AHTN (e.g. Tonalideâ ) or ADBI (e.g. Celostolideâ ) show similar environmental behaviour. Furthermore, correlations between these substances are reported (Winkler et al., 1998), due to their use in mixtures. Individual statements for the other substances will
either be more or less appropriate, depending on the respective sorption or accumulation behaviour. For concentrations in water on a certain scale the variations between the individual scenarios are smaller than the dierence between the minimum and maximum of the measured values. From the comparison of the concentrations of the three investigated regions it can be concluded that an exact selection of representative environmental segments, including an adjustment of the emission rates to the selected region, leads to the most realistic modelling of the contamination of the rivers. This requires closer scrutiny of the monitoring data and their spatial classi®cation, which could be carried out more easily using a geographical referenced information system. In this paper, we dealt with potential uncertainties in an aquatic fate assessment using dierent scenarios and by investigating the variability. An analysis of the uncertainties of substance parameters or a detailed investigation into the sewage treatment process of HHCB was omitted; the necessary database for such research is not yet available. Acknowledgements Funding by the German Federal Environmental Agency, Berlin is gratefully acknowledged (R+D project FKZ 206 01 075). We also thank H.D. Eschke from the Ruhrverband, Essen for his support and the anonymous reviewers for their helpful comments. References AWWR, 1997. Ruhrwasserg utebericht. The study group of the waterworks at the Ruhr (AWWR) and Ruhrverband. Essen. Berding, V., Schwartz, S., Matthies, M., 1999a. Scenario analysis of a Level III multimedia model using generic and regional data. Environ. Sci. & Pollut. Res., accepted. Berding, V., Schwartz, S., Matthies, M., 1999b. Visualisation of the Complexity of EUSES. Environ. Sci. & Pollut. Res. 6, 37±43. Bester, K., H uhnerfuss, H., Lange, W., Rimkus, G., Theobald, N., 1998. Results of nontarget screening of lipophilic organic pollutants in the German Bight II: polycyclic musk fragrances. Water Res. 32, 1857±1863. Butte, Ewald, 1999. Kinetics of accumulation and clearance of the polycyclic musk compounds galaxolide and tonalide. Abstract book of the Ninth Annual Meeting of SETACEurope, 2o/P16. Society of Environmental Toxicology and Chemistry, Brussels. Decisioneering, 1996. Crystal Ballâ Version 4.0 User Manual. Decisioneering, Inc. Denver, Colorado. EC, 1996a. Technical Guidance Document in Support of the Commission Directive 93/67/EEC on Risk Assessment for New Noti®ed Substances and the Commission Regulation (EC) 1488/94 on Risk Assessment for Existing Substances,
S. Schwartz et al. / Chemosphere 41 (2000) 671±679 Parts I±IV. Oce for Ocial Publications of the European Communities, Luxembourg. EC, 1996b. EUSES ± the European Union System for the Evaluation of Substances. Institute of Public Health and the Environment (RIVM), The Netherlands. Available from European Chemicals Bureau, Ispra. EPA, 1997. Exposure Factors Handbook, Vol. I-III. EPA/600/ P-95/002Fa, Oce of Research and Development, US Environmental Protection Agency, Washington. Eschke, H.D., Traud, J., Dibowski, H.J., 1994. Untersuchungen zum Vorkommen polycyclischer Moschus-Duftstoe in verschiedenen Umweltkompartimenten ± Nachweis und Analytik mit GC/MS in Ober¯ achen-, Abw assern und Fischen (1. Mitteilung). Z. Umweltchem. Okotox. 6, 183± 189. Eschke, H.D., Traud, J., Dibowski, H.J., 1995a. Untersuchungen zum Vorkommen polycyclischer Moschus-Duftstoe in verschiedenen Umweltkompartimenten ± Befunde in Ober¯achen-, Abwassern und Fischen sowie in Waschmitteln und Kosmetika (2. Mitteilung). Z. Umweltchem. Okotox. 7, 131±138. Eschke, H.D., Traud, J., Dibowski, H.J., 1995b. Nachweis und Quanti®zierung von polycyclischen Moschus-Duftstoen mittels Ion-Trap GC/MS/MS in Humanfett und Muttermilch. Dt. Lebensmittel-Rundschau 12, 375±379. Etienne, R.S., Ragas, A., vandeMeent, D., 1997. Operational Uncertainties in SimpleBox. National Institute of Public Health and the Environment (RIVM), Report 719101 031, Bilthoven. Gatermann, R., 1999. Bioaccumulation and enantioselective degradation of synthetic musks in dierent aquatic species. Abstract book of the Ninth Annual Meeting of SETACEurope, 2o/P07, Society of Environmental Toxicology and Chemistry, Brussels. Gatermann, R., Hellou, J., H uhnerfuss, H., Rimkus, G., Zitko, V., 1999. Polycyclic and nitro musks in the environment: a comparison between canadian and european aquatic biota. Chemosphere 38, 3431±3441.
679
Matthies, M., Wagner, J., Koormann, F., 1997. Combination of regional exposure models for european rivers with GIS information. In: Hrsg. Alef et al. K. (Ed.), Proceedings of ECO-INFORMAÕ97, Eco-Informa Press, Bayreuth, vol. 12, pp. 523±529. Mrasek, V., 1998. Duftstoe, die unter die Haut gehen. S uddeutsche Zeitung, 22, January 1998. Ohlo, G., 1990. Riechstoe und Geruchssinn: Die molekulare Welt der D ufte. Springer, Heidelberg. Plassche, E.J.van de, Balk, F., 1997. Environmental risk assessment of the polycyclic musks AHTN and HHCB according to the EU-TGD. National Institute of Public Health and the Environment (RIVM), Report 601503008, Bilthoven. Rimkus, G., Brunn, H., 1996. Synthetische Moschusduftstoe ± Anwendung, Anreicherung in der Umwelt und Toxikologie; Teil 1: Herstellung, Anwendung, Vorkommen in Lebensmitteln, Aufnahme durch den Menschen. Ern ahrungsUmschau 43, 442±449. Rimkus, G., Wolf, M., 1996. Polycyclic musk fragrances in human adipose tissue and milk. Chemosphere 33, 2033± 2043. Ruhrverband, 1999. Results of Monitoring of Polycyclic Musk Fragrances in River Ruhr ± Personal communication. Ruhrverband (Institution for River Ruhr water management), Essen. Schwartz, S., Berding, V., Matthies, M., 1999. Umweltexpositionsabsch atzung des polycyclischen MoschusDuftstoes HHCB ± Szenarienanalyse mit EUSES. Umweltmed. Forsch. Prax. 4, 7±11. Spencer, P.S., Sterman, A.B., Horopian, D.S., Foulds, M.M., 1979. Neurotoxic fragrance produces ceroid and Myelin Disease. Science 20, 633±635. Winkler, M., Kopf, G., Hauptvogel, C., Neu, T., 1998. Fate of arti®cial musk fragrances associated with suspended particulate matter spm from the river elbe germany in comparison to other organic contaminants. Chemosphere 37, 1139±1158.