Environ. Sci. Technol. 2007, 41, 2445-2451
Indicators for the Exposure Assessment of Transformation Products of Organic Micropollutants
and the atmosphere. Several instances have been reported where single transformation products are present in higher concentrations or detected more frequently than their parent compound. Important examples include pesticides such as organochlorine compounds, triazines or chloroacetanilides (1-3), surfactants (4), and chlorinated solvents (5).
L U K A S G A S S E R , †,‡ K A T H R I N F E N N E R , * ,† A N D MARTIN SCHERINGER‡ Eawag (Swiss Federal Institute for Aquatic Science and Technology), CH-8600 Du ¨ bendorf, Switzerland, Institute for Biogeochemistry and Pollutant Dynamics, Swiss Federal Institute of Technology (ETH), CH-8092 Zu ¨ rich, and Institute for Chemical and Bioengineering, Swiss Federal Institute of Technology (ETH), CH-8093 Zu ¨ rich, Switzerland
Since transformation products can be similarly or even more mobile, persistent, or toxic than their parent compounds (6), the assessment of their exposure potential and potential for toxic effects should be included in chemical risk assessment procedures and in the assessment of soil and water quality (7). However, producing high quality substance property data and carrying out a full-fledged risk assessment for each possible transformation product is not feasible because of limited time and resources. Instead, what is needed are screening approaches that efficiently prioritize transformation products for further investigations according to their exposure potential and their toxicological and ecotoxicological effects. Screening approaches for pesticide transformation products have been described (8, 9). However, none of the methods suggested so far accounts for the dynamics of transformation products formation and transport because they are based on qualitative or semiquantitative scores only. Therefore, the goal of this study is to develop a process-based model and corresponding indicators that integrate the available information on transformation products and quantitatively reflect the environmental fate of parent compounds and transformation products. Rather than simulating a particular region, we use a model representing an evaluative or generic region in which all transformation products can be consistently assessed and whose level of detail matches the available substance data.
Environmental transformation products of organic micropollutants have the potential to be similarly or even more mobile, persistent, or toxic than their parent compounds. They should, therefore, be included in chemical hazard and risk assessment procedures as well as in the assessment of soil and water quality. To fulfill this requirement most efficiently, screening approaches that select relevant transformation products for detailed assessment are needed. This paper presents two process-based multimedia, multispecies models that allow us to quantitatively estimate the environmental fate of transformation products. The resulting exposure patterns are assessed with two indicators: joint persistence (JP), which describes the temporal extent of environmental exposure to a parent compound and its transformation products, and the predicted relative aquatic concentrations (RAC), which estimate the relative concentrations of parent compounds and their transformation products in surface water bodies. As a case study, JP and RAC are calculated for 16 pesticides and their relevant transformation products. The results for the JP indicator confirm the importance of considering transformation products in the assessment of overall persistence; for example, in the context of PBT assessments. Comparison of RAC results with monitoring data on herbicides and their transformation products shows the suitability of our approach for estimating relative concentrations in surface water, and as a consequence, its usefulness in identifying transformation products for future water quality monitoring programs. Transformation products of triketones and other highly used acidic herbicides are specifically identified as targets.
Introduction Environmental contamination can be caused not only by synthetic organic compounds but also by their recalcitrant transformation products. Such products of biological and chemical transformation have been detected in all environmental media, including groundwater, surface waters, soils, * Corresponding author phone: +41-44-823 50 85; fax: +41-44823 54 71; e-mail:
[email protected]. † Eawag and Institute for Biogeochemistry and Pollutant Dynamics, (ETH). ‡ Institute for Chemical and Bioengineering (ETH). 10.1021/es062805y CCC: $37.00 Published on Web 03/06/2007
2007 American Chemical Society
In previous studies (10-12), we laid the basis for including transformation products in exposure assessment models and introduced joint persistence (JP) as a measure that describes the temporal extent of environmental exposure to a parent compound and its transformation products and is, therefore, useful in hazard-based assessment. In this article, we extend this approach by introducing an additional indicator, the predicted relative aquatic concentrations (RAC) of parent compounds and transformation products. This indicator specifically focuses on the protection of surface water resources and provides a basis for comparative risk assessment of parent compounds and their transformation products and for targeted monitoring of transformation products. It reflects differences in the extent of formation of transformation products in soil and water, in their transport from soil to water, and in their fate in surface waters. We calculate both indicators for 16 pesticides from 14 different chemical classes and their major transformation products (69 compounds in total). The models and indicators are intended to be generally applicable, but here we use pesticides and their transformation products as case studies. This has the following advantages: First, chemical property data needed for fate modeling are scarce for most transformation products. For pesticides, however, transformation schemes and substance properties are available for most parent compounds and even some transformation products. Using these data as a basis for comparison allows us to discuss the applicability of various substance property estimation tools, which will provide crucial knowledge for extending the approach to other substance classes. Second, for some pesticide transformation products, extensive monitoring information is available, which makes an evaluation of the predictive power of our modeling approach possible. VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2445
FIGURE 1. Scheme of three-box model, river model, and the interconnecting flux (fsw) from soil of the three-box model to first box of the river model. The two models are used to calculate the indicators PP, JP, and RAC. The river model consists of flowing water (shaded boxes), stagnant water (white boxes) and an underlying river sediment (not shown).
Methods Indicators and Models. Joint Persistence and Contribution to Joint Persistence. Overall persistence (Pov), i.e., a chemical’s residence time in a system of all environmental compartments of concern, has repeatedly been suggested as a hazardbased indicator for screening and prioritization purposes in chemical risk assessment (13-15). We have extended the concept of Pov to include transformation products by defining two Pov indicators (10): primary persistence (PP), which refers to the parent compound alone (eq 1), and joint persistence (JP), which represents the persistence of the entire substance family, i.e., the parent compound (PC) in combination with all relevant transformation products (TPs) (eq 2).
PP )
JP )
Mss PC +
∑
n i)1
SPC
Mss PC SPC
Mss i
) PP +
(1)
∑
n i)1
CTPi
(2)
with Mss i being the steady-state mass (mol) of each comss pound summed over all compartments j (Mss i ) ∑j)s,a,wmi,j), n the number of TPs, and SPC the release rate (mol/d) of the parent compound. The contribution to the joint persistence (CTPi) indicates the importance of single transformation products, i. As shown in eq 2, the CTPi values of all substances within a substance family add up to yield the JP. To calculate these persistence indicators, we use a generic three-box multimedia multispecies model with the compartments soil, water, and air, and averaged global properties (see Figure 1). The dimensions of each compartment and the quantification of transfer processes between compartments and of the distribution between phases within compartments are the same as used in ref 10. For pesticides, which are the focus of this work, we assume emission of the parent compound only, with 90% emitted to surface soil and 10% to air to reflect a certain amount of spray drift. Relative Aquatic Concentrations. The predicted relative aquatic concentrations (RAC) reflect the surface water concentrations of a set of TPs and their PC for unit emission rates of the PCs. They indicate the intrinsic potential of the different compounds for aquatic exposure, independent of actual quantities used. For comparison within or across substance families, RACs are further normalized either to the RAC of each substance family’s PC or to the RAC of atrazine as a well-known reference compound, respectively. We determine the RAC values with a river model and define them as the calculated concentrations in the last box of the river model. We use the evaluative river model from Scheringer et al. (16) for this purpose; the length of the river is 700 km, represented by 70 boxes of 10 km length and 2446
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007
increasing volumes representing the inflow of tributaries (see Figure 1). Each box contains flowing water, stagnant water, and sediment; the dimensions of the compartments and the quantification of the phase-transfer processes are described in ref 16. With a water residence time of about 8 days and the dimensions chosen, this model is roughly representative of the river Rhine. All compounds are continuously released into the first river box. Their release rates are set equal to their fluxes from soil to water in the three-box model. Transformation Schemes and Substance Properties. The 16 case study pesticides (13 herbicides, 2 fungicides, 1 insecticide) were selected according to the following criteria: total amount used in four Swiss catchments (Greifensee, Baldeggersee, Murtensee, and Genfersee), data availability, diversity of the substance classes, and ecotoxicological hazard potential. They are listed in Table S1 in the Supporting Information (SI). For each pesticide, transformation schemes for the compartments soil, water, and sediment were assembled. No information on transformation products formed through atmospheric oxidation could be found. However, since only small fractions of the pesticides and their TPs are expected to reside in air, we neglected the formation of TPs in air. For each pesticide, only those TPs were included in the schemes that were either referred to as major metabolites or reported in several of the data sources used (i.e., handbooks (17, 18) and registration information from the EU (19), the U.S. (20), and the UK (21)). This approach focuses on current knowledge on recalcitrant TPs and assumes that possible other TPs are considerably less stable. As a consequence, it leads to different numbers of TPs considered for each pesticide. Since in the river model the sediment layer is assumed to be a thin, aerobic layer, the sediment transformation scheme is assumed to be the same as the water transformation scheme. All transformation schemes, including the nomenclature for PCs and TPs used throughout the paper are given in the SI. To calculate JP and RAC values, chemical fate properties for all PCs and TPs need to be entered into the models, namely degradation rate constants and, for the TPs, fractions of formation in all compartments, as well as partition coefficients between the environmental phases considered in the model. From the re-registration procedure in both the U.S. and the EU, chemical property data for pesticides and their major TPs are, at least to some extent, publicly available. The main sources for chemical properties used here were the same as those used to establish the transformation schemes. In addition, the Herbicide Handbook (22) and the Pesticide Manual (23) were consulted. In some cases, additional data were obtained from original publications (see references in SI).
Whereas for all PCs enough reliable data exist, property data for most TPs are incomplete. When no experimental data on fate properties were available, BIOWIN from EPI (Estimation Programs Interface) Suite (24) was used for the estimation of biodegradation half-lives, PCKOCWIN for the organic carbon-water partition coefficient (Koc), and AOPWIN for second-order rate constants for reaction with OH radicals in the air. The outputs of these estimation methods were further processed to yield input values for the two fate models. This treatment of the estimated property data and the final values used are described in Tables S2 and S3 in the SI. In some cases fractions of TP formation in soil and water (ffs and ffw) could be derived from experimental data from degradation studies. If the half-life of the precursor in such a study is known, the fraction of formation and the half-life of the TP can be estimated from its maximal amount formed and the time elapsed between the start of the study and the time of its maximal concentration (see SI). In those cases where no information from degradation studies was available, we assigned generic fractions of formation: 1 in the case of a single transformation pathway, 0.5 for two parallel pathways, and 0.33 for three parallel pathways. These generic values were further multiplied by an estimated correction factor of 0.8 to account for the fact that all fractions of formation derived from experimental data were smaller than these generic values.
Results and Discussion Joint Persistence and Contribution to Persistence. Figure 2a gives an overview of calculated PP and JP values for all pesticides. More detailed results on the contributions to persistence (CTPi) of the single TPs are given in Table S4 in the SI. The results confirm the importance of including transformation products in persistence assessments. For 10 of 16 pesticides, the JP more than doubles compared to the PP. The ratio Q between JP and PP indicates how much the transformation products contribute to the persistence of a substance family. For the 16 pesticides investigated, Q varies between values close to 1, which means that the TPs hardly contribute to persistence (2,4-D, mecoprop-P), and values around 50 (kresoxim-me). High values of Q are typically observed for pesticide families where the parent compound is a pro-pesticide (bromoxynil-oct, fluoroglycofen-et, and kresoxim-me) that is hydrolyzed rapidly to the active substance. For all other pesticides, the finding of Fenner et al. (12) is confirmed that Q does usually not exceed a value of 4 (with the exception of alachlor, Q ) 12.5). Predicted Relative Aquatic Concentrations. Figure 2b shows the relative aquatic concentrations (RAC) of the parent pesticides compared to the sum of RACs of the PC and its TPs (joint RAC). For intercomparison across substance families, all values are normalized to the RAC of atrazine. The sequence of pesticide families is different from the JP ranking, with families including more mobile compounds exhibiting higher ranks (e.g., dicamba). Most notably, for some pesticide families, the ratio Q between primary and joint RAC, now ranging from 1.02 to 105, is considerably higher than the ratio between PP and JP. The RAC indicator focuses exclusively on the water compartment; therefore, differences in mobility between PCs and transformation products as determined by their Koc values are more pronounced than in the JP indicator, which integrates steady-state masses in soil and water. Also, there is a higher degree of differentiation between families, i.e., the highest joint RAC (for the mesotrione family) is more than 100-fold higher than the lowest joint RAC (for the glyphosate family). To gain further insight into the RAC indicator, we compare relative values of the chemicals’ steady-state masses in soil
FIGURE 2. (a) Primary (PP) and joint persistence (JP) sorted according to JP; (b) primary and joint relative aquatic concentrations normalized to RAC of atrazine and sorted according to joint RAC for 16 pesticides. If no primary RAC is depicted, values are too small for graphical representation. (msss), of their fluxes from soil to water (fsw), and of their final RAC values within each pesticide family by normalizing them to the values for the PCs. This leads to three groups of substance families: (i) all three quantities show similar patterns within a family, (ii) Koc causes differences in fsw patterns compared to msss patterns; (iii) the RAC pattern is different from the fsw pattern and this again differs from the msss pattern. (i) Similar msss, fsw and RAC Patterns. Transformation products are less important than the parent compounds regarding both fluxes and RAC. In this group (2,4-D, mecoprop-P, amidosulfuron, dicamba, diuron, atrazine, orbencarb), the TPs have similar Koc values, and hence, similar mobility as the parent compounds. Additionally, both PC and TPs degrade slowly in water (kw < 0.1 d-1). As a consequence, the TPs remain mainly in the soil compartment where they are formed, their flux from soil to water relative to their mass in soil is similar or even lower than for the PC, and transformation reactions in the river water are unimportant (see Figure 3a, diuron). Within this group, fsw and RAC values of transformation products can be slightly higher or lower than the msss values. VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2447
FIGURE 3. Comparison of msss (steady-state mass in soil), fsw (flux from soil to water), and RAC (relative aquatic concentration) within substance families (a: diuron, b: mesotrione, c: fluoroglycofenet). All values normalized to values of PCs. This variation reflects the difference in polarity between PC and TPs. In the case of orbencarb, e.g., TPs are somewhat more polar than the PC and, therefore, show slightly increased fsw and RAC values in comparison to their masses in soil relative to the PC. However, since the soil half-lives of most pesticide PC and TPs lie in a similar range (10-100 d) but the shares formed of the TPs are always smaller than 1, the effect of small changes in polarity in this group is, in general, not strong enough to make RAC values of TPs higher than those of the PC (only exception: orbencarb sulfonic acid). (ii) fsw Patterns Different from msss Patterns. Similar fsw and RAC Patterns. Fluxes and RAC values of transformation products exceed values for parent compounds. In this group, TPs have a markedly higher polarity than the parent compounds either because the PC is neutral and TPs are acidic compounds (alachlor), or because a major structural change takes place such as the loss of a phosphate group that sorbs strongly in soils (glyphosate) or the cleavage of the 2448
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007
PC into two smaller molecules (mesotrione, sulcotrione). As in group (i), PC and TPs degrade slowly in water (kw < 0.1 d-1). The higher polarity of the TPs leads to higher fluxes compared to the PC, whereas, because of the low reactivity in water of all compounds, the RAC pattern reflects the flux pattern (see Figure 3b, mesotrione). Generally, it is observed that if the Koc values of the TPs are, by a factor of 50 or more, lower than that of the PC, the TPs are more prevalent in surface waters than the PC. (iii) fsw Pattern Different from msss Patterns; RAC Patterns Different from fsw Patterns. Fluxes and RAC values of transformation products exceed values for parent compounds. In this group, too, the TPs have a markedly higher polarity than the PCs and, in addition, the PCs degrade rather rapidly in soil and water (kw > 0.1 d-1). This group contains mostly PCs that are pro-pesticides (kresoxim-me, bromoxynil-oct, fluoroglycofen-et, chlorpyrifos-me) as well as TCPN that is biodegraded exceptionally fast in surface waters (kw ) 0.63 d-1). The pro-pesticides are neutral esters of the acidic active compounds and hydrolyze rapidly into the highly mobile acids. As a consequence, the TPs exhibit not only higher fluxes than the PC, but their RAC values also exceed their fsw values (note the logarithmic scale in Figure 3c, fluoroglycofenet). Differences between PCs and TPs within single substance families as described above are important to gain mechanistic insights into the factors driving environmental exposure to TPs. For practical purposes such as monitoring or prioritization of contaminants, however, it is also of interest to know how all these factors determine relative aquatic concentrations across substance families. For this purpose, we ranked the 69 PCs and TPs of all 16 substance families according to their RAC values. Table 1 lists the top 20 RAC compounds with all RAC values normalized to that of atrazine (chosen as a reference compound because of its well-known environmental behavior); for a list of all 69 compounds, see SI. Table 1 also contains scores of data quality for the relevant chemical properties. 12 compounds in the top 20 RAC list are TPs, which confirms the importance of transformation products as water pollutants. These predictions agree well with findings of the U.S. Geological Survey (USGS), who run a long-term monitoring program for herbicides (mainly triazines and chloroacetanilides) and their TPs in U.S. Midwestern streams and groundwaters. They report total TP concentrations to be about 20-fold higher than total PC concentrations and frequencies of detection for TPs to be systematically higher than for PCs (25-27). Our calculations clearly indicate that the acidic TPs of chloroacetanilide compounds (here alachlor OXA, ESA, and sulfinyl acetic acid) have a high potential for entry into surface waters. In refs 25 and 27, median concentration ratios between total chloroacetanilide TPs and PC of 2:1-9:1 (25), and 1.25:1-8.75:1 (27) in different streams and under different flow conditions are reported. Our calculated RAC ratio, which aims to represent average conditions, is 11:1 for alachlor and lies at the upper limit of those ranges. Our measurements in the lake Greifensee (unpublished data) show average concentration ratios of metolachlor ESA: metolachlor of roughly 8:1 and metolachlor OXA:metolachlor of 3:1. For alachlor, corresponding calculated RAC ratios are 7:1 for ESA and 4:1 for OXA and thus agree well if we assume that transformation patterns are roughly equal across different chloroacetanilides. The TPs of the triazine herbicide atrazine are not part of the top 20 list because they are not particularly mobile compared to TPs of groups (ii) and (iii). However, because triazine herbicides are used in high amounts, their TPs are ubiquitous in surface and groundwater samples. In ref 28,
TABLE 1. RAC and RAC Greifensee Ranks and Values for the Top 20 Compounds (RAC: Relative Aquatic Concentration; RAC Greifensee: Relative Aquatic Concentration for the Greifensee Catchment)a compound
RAC rank
RAC relative to atrazine
RAC Greifensee rank
RAC Greifensee relative to atrazine
CMBA dicamba amidosulfuron kresoxim-me-acetic acid CHD MNBA AMBA alachlor ESA 3,6-di-Cl-salicylic acid B-benzoic acid alachlor sulfinyl acetic acid mecoprop-P atrazine acifluorfen alachlor OXA B-benzamide mesotrione sulcotrione ADHP amino-acifluorfen DEA DIA HA 6-Cl-salicylic acid 2,5-OH-di-Cl-salicylic acid orbencarb sulfonic acid alachlor
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
3.30 2.63 2.33 2.23 2.10 2.01 1.46 1.43 1.35 1.26 1.08 1.07 1.00 0.80 0.78 0.78 0.72 0.62 0.57 0.51
1 5 >20 19 3 >20 >20 11 8 13 12 6 2 >20 14 17 >20 7 >20 >20 4 9 10 15 16 18 20
1.08 0.32 0.015 0.69 0.084 0.16 0.048 0.064 0.24 1.00 0.046 0.030 0.20 0.41 0.14 0.08 0.03 0.03 0.03 0.01
ks
kw
** *** *** *** * * * ** ** * ** *** *** *** ** * ** *** ** * *** *** *** * * * ***
* * ** ** * * * ** * * * ** *** * ** * * * * * *** *** *** * * * ***
data quality Koc + +++ +++ +++ + + + +++ + + ++ +++ +++ +++ +++ + +++ +++ + ++ +++ +++ +++ + + + +++
ffs
ffw
° ° ° ° ° °°° °°° ° °°° °° °°° °° °° ° °°° °°° °°° ° ° ° -
° ° ° ° ° °°° °° °° °° ° °°° °° ° ° °°° °°° °°° ° ° ° -
a Compounds sorted according to RAC rank. All values normalized to atrazine. Compounds that are part of the top 20 RAC Greifensee list but not of the top 20 RAC list are added at the bottom. Quality of input data indicated for degradation rate constants ks and kw (*** is >2 empirical values, ** is 1-2 empirical values or estimated from degradation study, * is estimated using BIOWIN), for organic carbon-water partition coefficient Koc (+++ is >2 empirical values, ++ is 1-2 empirical values or analogy reasoning, + is estimated using PCKOCWIN), and for fractions of formation, ffs and ffw (°°° is deduced from >1 empirical degradation studies, °° is deduced from 1 empirical degradation study, ° is generic, - is no ff because compound is PC).
median DEA-to-atrazine ratios (DAR) of 0.2-0.4 in all streams under preapplication, application, and postapplication conditions are reported, whereas our RACs yield a DAR of 0.4. Reported total TP:PC ratios for atrazine vary between 0.1 and 0.5 (25, 27, 29). Our RAC ratio of 0.7 of total atrazine TP:PC is somewhat higher but still close to this value. The USGS studies also show that under baseflow conditions, where a higher contribution of groundwater infiltration is expected, TP:PC ratios can be several times higher than the ones used for comparison here. Four more substance families are well represented in the top 20 RAC list. These are the triketones sulcotrione and mesotrione and the acidic herbicides dicamba and bromoxynil-oct. Hardly any monitoring information is available on these compounds and their TPs. In addition, measurements of the compound properties for these substance families are few. Especially the generic fractions of formation used in the calculations may thus be overestimated. Nevertheless, our results indicate that it would be important to include triketones, highly used acidic herbicides, and their TPs in future monitoring programs. The RAC indicator reflects the relative aquatic concentrations of a range of PCs and TPs for unit emission rates of the PCs. However, it can also be transformed to more closely reflect the situation in a given catchment by multiplying the RAC values of each substance family by the actual amounts of parent pesticide used. As an example, we adjusted our RAC values to reflect actual pesticide usage in the Greifensee catchment in Switzerland. Usage data were obtained from a stratified survey during the period 1997-2003 (see SI). The 20 compounds with highest RAC values for the Greifensee catchment in Switzerland are also listed in Table 1 (again relative to atrazine). Compared to the RAC values for unit
emissions, atrazine TPs now appear on the list since atrazine is the most-used herbicide in the catchment. Whereas alachlor and sulcotrione and their TPs remain important, mesotrione is not used in the catchment. Since metolachlor is the chloroacetanilide used in higher amounts than alachlor in the catchment, metolachlor TPs can also be expected to be present at high concentrations. Our measurements of metolachlor OXA, metolachlor ESA, and the atrazine transformation product DEA in the epilimnion of lake Greifensee during or shortly after the application period confirm the presence of all three TPs at concentrations in the same order of magnitude (i.e., 10-100 ng/L, unpublished data). For a more detailed, quantitative comparison of modeled RAC values with measured concentrations, which is not the purpose of this study, accounting for actual emissions may have to be accompanied by an adjustment of residence time and geometry of the modeled water body.
Uncertainties Induced by Parameter Estimation Methods All the above-mentioned rankings and absolute values are uncertain due to uncertainty and variability in the input parameters. Whereas for most parent pesticides, several experimental half-lives and Koc values can be found, property data for TPs, except for the most well-known ones (alachlor OXA and ESA; DEA and HA), are scarce. An in-depth uncertainty analysis of the JP indicator (12) showed that the most influential and uncertain substance parameters for JP and CTPi are, in order of decreasing importance, (i) halflives in soil, (ii) half-lives in water, (iii) fraction of formation in the main residence compartment of the precursor, and (iv) Koc. For the RAC indicator, Koc gains importance since it governs the flux from soil to water. It is, therefore, of interest VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2449
to know how uncertain the estimates of these parameters are and how this influences the RAC results. If no experimental half-lives are available, estimation with BIOWIN and conversion into half-life categories as proposed in BIOWIN are used in this study (see SI). Others have suggested that linear correlations established between BIOWIN output and experimental half-lives could be used to derive unknown half-lives from BIOWIN results (30). In an accompanying study, we have compared the validity of these two methods for 20 pesticides and 18 pesticide TPs with good experimental data (31). The results showed that none of the methods to convert BIOWIN output into half-lives is superior and that both methods produce mean errors in halflife predictions of a factor of around 5 and maximal errors of a factor of around 30. For our 69 case study compounds, 28 soil half-lives and 48 water half-lives had to be estimated with BIOWIN. The related uncertainties propagate through the model and influence the results for TPs with estimated half-lives. In the top 20 list, RAC values for the TPs of the two substance families of bromoxynil-oct and fluoroglycofen-et can be expected to exhibit highest uncertainties since all half-lives of the second and higher generation TPs had to be estimated. To illustrate the consequences of these uncertainties, we recalculated the RAC values for these two substance families under the extreme assumption that all estimated soil and water half-lives are too high by a factor of 5. B-benzamide showed the highest sensitivity to these changes, resulting in a 17-fold decreased RAC value, relegating it to position 45 of the RAC ranking. For all other TPs of these two substance families the effect of the factor of 5 in the half-lives is smaller (reduction in RAC values by a factor of 8 or less). In summary, uncertainty in half-life estimates affects calculated RAC values by a maximal factor of about 20. Since this factor is small compared to the entire range of 6 orders of magnitude spanned by all RAC values, the discriminatory power of the model calculations remains intact despite uncertainty in substance parameters. Further uncertainty may arise from Koc values, which had to be estimated for 38 out of 69 compounds as described in the SI. Comparison of the PCKOCWIN predictions for 17 neutral PCs and TPs with experimental Koc values indicated a mean prediction error of a factor of 6.5. Since we use analogy considerations in the derivation of unknown Koc values for neutral compounds (see SI), our error should be somewhat smaller. Comparison of Koc values for 15 anionic PCs and TPs with PCKOCWIN predictions yields a factor of 2.5 between the predicted neutral Koc values and the measured Koc value for the anionic compound. Thus, our assumed extrapolation factor of 10 between PCKOCWIN predictions and predicted Koc values for anionic compounds might be too high. All in all, uncertainties in predicted values can be assumed to be around a factor of 5 for neutral compounds and a factor of 20 for anionic compounds. In the relevant Koc range for pesticides and their TPs, i.e., 1-1000 L/kg, this translates into a maximum uncertainty factor of 5 in the estimation of the dissolved fraction for neutral compounds, and a corresponding uncertainty factor of 10 for anionic compounds, and hence, their RAC values. In conclusion, the current rankings for compounds where all parameters are estimated (see indications of data quality in Table 1) are subject to considerable uncertainty. Nevertheless, the results and rankings obtained in this study can serve as a starting point to target screening of environmental water samples for TPs. Although high-resolution mass spectrometers have recently become available, allowing for the identification of substances without the need for reference standards, these methods are still not suitable for nontarget screening of trace contaminants such as TPs. Therefore, lists of possible target contaminants as in Table 1 are indispensable for the application of these techniques to gain a better 2450
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007
understanding of the prevalence of TPs in water bodies. The outcome of such targeted monitoring studies, in turn, will allow for a further refinement of the input data and the modeling approach for TP prioritization.
Environmental Significance The models and indicators developed in this study confirm that TPs of environmental contaminants such as pesticides significantly contribute to the total environmental exposure. While the JP is an important indicator to reflect this aspect in the context of PBT assessments or hazard-based assessments in general, the RAC is useful in the context of water quality assessment, risk-based assessment, and for targeting monitoring strategies. We have used pesticides and their TPs as case study compounds because chemical property data and monitoring information are available. This enabled a comparative evaluation of model predictions with field data for the betterknown substance families such as the chloroacetanilides and triazines. An important next step will be to corroborate RAC rankings of some of the lesser-studied pesticides, such as acidic herbicides and their TPs by further, targeted monitoring. The models developed here can be used directly for other water-relevant pollutants being emitted to soil, which is the case for compounds contained in sludge or manure used as fertilizer on agricultural fields (e.g., veterinary medicines). For compounds being released to the environment through point sources such as sewage treatment plants, the models need to be extended by a module representing the sewage treatment plant. Also, for compounds forming cationic and zwitterionic species, Koc will not be sufficient as a descriptor of mobility in soil. However, a more critical factor currently limiting the evaluation of the importance of TPs for other substance classes is the lack of information on transformation schemes. Current tools for the prediction of biodegradation schemes are still in their infancy (32, 33). However, this situation can be expected to improve in the future, since research to enhance the predictive power of such tools is ongoing. At the same time, the experimental database will grow with the implementation of REACH, the new chemicals legislation in Europe, which requests identification of relevant TPs for all compounds produced in amounts exceeding 100 t/year. Improved biodegradation estimation tools and experimental information about TPs will make it possible to extend the approach proposed here to other chemical classes.
Acknowledgments Funding by the Swiss Federal Office for the Environment is gratefully acknowledged. We thank Heinz Singer, Beate Escher, Juliane Hollender, Konrad Hungerbu ¨ hler, and Rene´ Schwarzenbach for helpful comments.
Supporting Information Available List of case study pesticides and transformation schemes; information on how unknown substance properties were derived and a complete list of half-lives and Koc values used in modeling; method how fractions of formation were derived from degradation study data; complete listing of all calculated persistence and RAC values. This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) Coats, J. R. What happens to degradable pesticides. Chemtech 1993, 23, 25-29. (2) Kalkhoff, S. J.; Kolpin, D. W.; Thurman, E. M.; Ferrer, I.; Barcelo, D. Degradation of chloroacetanilide herbicides: The prevalence of sulfonic and oxanilic acid metabolites in Iowa groundwaters and surface waters. Environ. Sci. Technol. 1998, 32, 1738-1740.
(3) Scribner, E. A.; Thurman, E. M.; Zimmermann, L. R. Analysis of selected herbicide metabolites in surface and ground water of the United States. Sci. Total Environ. 2000, 248, 157-167. (4) Ahel, M.; Giger, W.; Schaffner, C. Behavior of alkylphenol polyethoxylate surfactants in the aquatic environment. 2. Occurrence and transformation in rivers. Water Res. 1994, 28, 1143-1152. (5) Yeh, H. C.; Kastenberg, W. E. Health risk assessment of biodegradable volatile organic-chemicalssa case-study of PCE, TCE, DCE and VC. J. Hazard. Mater. 1991, 27, 111-126. (6) Tixier, C.; Sancelme, M.; Bonnemoy, F.; Cuer, A.; Veschambre, H. Degradation products of a phenylurea herbicide, diuron: Synthesis, ecotoxicity, and biotransformation. Environ. Toxicol. Chem. 2001, 20, 1381-1389. (7) Boxall, A. B. A.; Sinclair, C. J.; Fenner, K.; Kolpin, D.; Maud, S. J. When synthetic chemicals degrade in the environment. Environ. Sci. Technol. 2004, 38, 368A-375A. (8) Belfroid, A. C.; van Drunen, M.; Beek, M. A.; Schrap, S. M.; van Gestel, C. A. M.; van Hattum, B. Relative risks of transformation products of pesticides for aquatic ecosystems. Sci. Total Environ. 1998, 222, 167-183. (9) Sinclair, C. J.; Boxall, A. B. A.; Parsons, S. A.; Thomas, M. R. Prioritization of pesticide environmental transformation products in drinking water supplies. Environ. Sci. Technol. 2006, 40, 7283-7289. (10) Fenner, K.; Scheringer, M.; Hungerbu ¨ hler, K. Persistence of parent compounds and transformation products in a level IV multimedia model. Environ. Sci. Technol. 2000, 34, 3809-3817. (11) Fenner, K.; Kooijman, C.; Scheringer, M.; Hungerbu ¨ hler, K. Including transformation products into the risk assessment for chemicals: The case of nonylphenol ethoxylate usage in Switzerland. Environ. Sci. Technol. 2002, 36, 1147-1154. (12) Fenner, K.; Scheringer, M.; Hungerbu ¨ hler, K. Joint persistence of transformation products in chemicals assessment: Case studies and uncertainty analysis. Risk Anal. 2003, 23, 35-53. (13) Klasmeier, J.; Matthies, M.; Macleod, M.; Fenner, K.; Scheringer, M.; Stroebe, M.; Le Gall, A. C.; McKone, T.; Van De Meent, D.; Wania, F. Application of multimedia models for screening assessment of long-range transport potential and overall persistence. Environ. Sci. Technol. 2006, 40, 53-60. (14) Kloepffer, W. Environmental hazard - assessment of chemicals and products, part II: Persistence and degradability of organic chemicals. Environ. Sci. Pollut. Res. 1994, 1, 108-116. (15) Scheringer, M. Persistence and spatial range as endpoints of an exposure-based assessment of organic chemicals. Environ. Sci. Technol. 1996, 30, 1652-1659. (16) Scheringer, M.; Halder, D.; Hungerbu ¨ hler, K. Comparing the environmental performance of fluorescent whitening agents with peroxide bleaching of mechanical pulp. J. Ind. Ecol. 2000, 3, 77-95. (17) Roberts, T. R. Ed. Herbicides and Plant Growth Regulators; The Royal Society of Chemistry: Cambridge, 1998; Vol. Part 1. (18) Roberts, T. R. Ed. Insecticides and Fungicides; The Royal Society of Chemistry: Cambridge, 1999; Vol. Part 2.
(19) European Commission, Plant Protection Products, http:// europa.eu.int/comm/food/plant/protection/evaluation/new_subs_rep_en.htm, 2006. (20) U.S. EPA, Pesticides: Reregistration, http://cfpub.epa.gov/ oppref/rereg/status.cfm?show)rereg, 2006. (21) U.K. PSD, http://www.pesticides.gov.uk/psd_evaluation_all.asp, 2006. (22) Ahrens, W. H., Ed. Herbicide Handbook, 7 ed.; Weed Science Society of America: Champaign, Illinois, 1994. (23) Tomlin, C. D. S. Ed. The Pesticide Manual; 13 ed.; BCPC: Alton, UK, 2003. (24) U.S. EPA EPI (Estimation Programs Interface) Suite, version 3.12; November 30, 2004. http://www.epa.gov/opptintr/exposure/pubs/episuitedl.htm. (25) Battaglin, W. A.; Thurman, E. M.; Kalkhoff, S. J.; Porter, S. D. Herbicides and transformation products in surface waters of the Midwestern United States. J. Am. Water Resour. Assoc. 2003, 39, 743-756. (26) Kalkhoff, S. J.; Lee, K. E.; Porter, S. D.; Terrio, P. J.; Thurman, E. M. Herbicides and herbicide degradation products in upper Midwest agricultural streams during August base-flow conditions. J. Environ. Qual. 2003, 32, 1025-1035. (27) Rebich, R. A.; Coupe, R. H.; Thurman, E. M. Herbicide concentrations in the Mississippi River Basinsthe importance of chloroacetanilide herbicide degradates. Sci. Total Environ. 2004, 321, 189-199. (28) Gilliom, R. J.; Barbash, J. E.; Crawford, C. G.; Hamilton, P. A.; Martin, J. D.; Nakagaki, N.; Nowell, L. H.; Scott, J. C.; Stackelberg, P. E.; Thelin, G. P.; Wolock, D. M. The quality of our nation’s watersspesticides in the nation’s streams and ground water, 1992-2001. U. S. Geol. Surv. Circ. 1291 2006. (29) Steen, R. J. C. A.; Van Hattum, B.; Brinkman, U. A. T. A study on the behaviour of pesticides and their transformation products in the Scheldt estuary using liquid chromatography-electrospray tandem mass spectrometry. J. Environ. Monit. 2000, 2, 597602. (30) Arnot, J.; Gouin, T.; Mackay, D. Practical Methods for Estimating Environmental Biodegradation Rates; Canadian Environmental Modelling Network: Peterborough, ON, 2005. (31) Fenner, K.; Canonica, S.; Escher, B. I.; Gasser, L.; Spycher, S.; Tuelp, H. C. Developing quantitative structure-activity relationships (QSARs) for REACH. Chimia 2006, 60, 683-690. (32) Hou, B. K.; Ellis, L. B. M.; Wackett, L. P. Encoding microbial metabolic logic: predicting biodegradation. J. Ind. Microbiol. Biotechnol. 2004, 31, 261-272. (33) Jaworska, J.; Dimitrov, S.; Nikolova, N.; Mekenyan, O. Probabilistic assessment of biodegradability based on metabolic pathways: Catabol system. SAR QSAR Environ. Res. 2002, 13, 307-323.
Received for review November 27, 2006. Revised manuscript received January 25, 2007. Accepted January 30, 2007. ES062805Y
VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2451
