Transformation Products in the LifeCycle Impact Assessment ofChemicalsR O S A L I E V A N Z E L M , * , †

M A R K A . J . H U I J B R E G T S , † A N D D I K V A ND E M E E N T † , ‡

Department of Environmental Science, Institute for Water andWetland Research, Radboud University Nijmegen,P.O. Box 9010, 6500 GL, Nijmegen, The Netherlands, andLaboratory for Ecological Risk Assessment, National Instituteof Public Health and the Environment, P.O. Box 1, 3720 BA,Bilthoven, The Netherlands

Received July 14, 2009. Revised manuscript receivedDecember 17, 2009. Accepted December 18, 2009.

The current life cycle impact assessment (LCIA) of chemicalsfocuses only on the fate and effects of the parent compound,neglecting the potential impact of transformation products. Here,we assess the importance of including the potential impactof transformation products in the calculation of characterizationfactors (CF). The developed method is applied to freshwaterecotoxicity for 15 pesticides and perchloroethylene, which areall known to have potentially persistent transformationproducts. The inclusion of transformation products resulted ina median increase in CF that varied from negligible to morethan 5 orders of magnitude. This increase, however, can be highlyuncertain, particularly due to a lack of toxicity data fortransformation products and a lack of mode of action-specificdata. We show in a case study that replacement of atrazinewith other pesticides for application on corn results in a medianimpact score of 2 orders of magnitude lower when the fateand effects of only the parent compounds are included. Whentransformation products are included, the reduction inmedian impact score would likely be lower (less than 1 orderof magnitude). An uncertainty analysis showed that thedifference in impact scores of atrazine and the atrazinereplacements was not statistically significant when only theparent chemical was considered. When transformation productswere included, the uncertainty in impact scores was evengreater.

Introduction

Characterization factors (CFs) are used in the life cycle impactassessment (LCIA) of products to determine the impact thata stressor causes to humans and ecosystems. For ecotoxicityin LCIA, the fate and effects of the parent compound onlyare taken into account (1, 2). In the fate step, multimediafate models are commonly used to calculate the persistenceof a chemical, accounting for intermedia transport processes,intramedia partitioning, and degradation in the environment.The effect step represents the calculation of the average

toxicity of a chemical on the basis of a set of toxicity data forvarious species.

Degradation in the environment, however, may yieldtransformation products (TPs) that can pollute the environ-ment as well. Inclusion of the impacts of these TPs in thecharacterization factor of a chemical can therefore beimportant in LCIA. This is particularly relevant when adegradation product is more toxic, more persistent, moremobile, or more bioaccumulative than its parent compound(3). Although information is available on the risk of TPs (3-8),to our knowledge, the inclusion of TPs has not previouslybeen addressed in LCIA.

The goal of this paper is to assess the relevancy of theinclusion of ecotoxicological impacts due to transformationproducts in LCIA. We provide a method that includes thefate and effects of TPs in the characterization factor of achemical and calculate characterization factors for 15pesticides and perchloroethylene. Each chemical has one ormore TPs that are thought to be potentially persistent. Withthis work, persistence, mobility, and toxicity of transforma-tion products are addressed together with their parentcompounds. An uncertainty analysis is carried out to quantifythe uncertainty in characterization factors with and withoutthe inclusion of transformation products. Moreover, in apractical example, we assess the relevancy of includingtransformation products for addressing the impacts ofatrazine application on corn compared to the application ofsubstituting pesticides.

MethodologyCharacterization Factor. Up to now, the characterizationfactor of a chemical x (CFx) for freshwater ecotoxicity hasbeen obtained by multiplying the fate factor (FFx in yr/m3),which expresses the environmental persistence, with theeffect factor (EFx m3/kg), which expresses the toxicity of asubstance (9, 10) as

with

where ∆M is a small emission change (kg/yr), ∆Cx is thecorresponding change in freshwater concentration (kg/m3),and ∆E is the corresponding change in the effect (dimen-sionless) of chemical x. S is the toxic mode of action (TMoA)-specific slope factor of the dose-response curve that showsthe effect in the environment due to a toxic unit. Thegeometric mean toxicity of chemical x (kg/m3) is representedas10µx.

In the current research we took into account n transfor-mation products t of a chemical x in the calculation of itscharacterization factor. CFx now becomes

where FFxfta is the fate factor of transformation product tdue to transformation from the parent chemical x, whichwas calculated as follows:

* Corresponding author phone: +31-24-3652923; e-mail:r.vanzelm@science.ru.nl.

† Radboud University Nijmegen.‡ National Institute of Public Health and the Environment.

CFx ) FFx × EFx (1)

FFx )∆Cx

∆Mx(2)

and EFx )∆Ex

∆Cx) S × 1

10µx(3)

CFx,tot ) FFx × EFx + ∑a)1

n

(FFxfta× EFta

) (4)

Environ. Sci. Technol. 2010, 44, 1004–1009

1004 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 3, 2010 10.1021/es9021014 2010 American Chemical SocietyPublished on Web 01/05/2010

EFta is the effect factor of a transformation product and wascalculated in the same way as EFx (eq 3).

Fate Factor. The multispecies chemical version of Sim-pleBox 3.0 (11) was adapted here for Life Cycle Assessment(LCA) purposes. SimpleBox is the underlying fate model ofthe European Union System for the Evaluation of Substances(12) and is also part of the multimedia fate exposure andeffects model USES-LCA (10, 13). With this multispeciesmodel, we calculated fate factors for each parent compoundand up to four generations of its TPs. The fate model includedan algorithm that approximated the effect of intermittentrain events, following the method of Jolliet and Hauschild(14).

Effect Factor. Effect factors were calculated according tothe nonlinear msPAF (multisubstance potentially affectedfraction of species) method, which explicitly accounts forthe nonlinearity in concentration-response relationships(15, 16). This method is based on toxic pressure assessments,and its representative is the potentially affected fraction ofspecies (PAF). A PAF-value quantifies the toxic pressure puton ecosystems by a single chemical or a mixture of chemicals,and it reflects the fraction of all species that are expected tobe exposed above a certain effect-related benchmark, suchas the effect concentration for 50% of species (EC50) or theno observed effect concentration (NOEC) (17). For mixtures,the estimated fraction is called the msPAF. The nonlinearmsPAF method provides the possibility to explicitly quantifyuncertainty in the effect factor.

Uncertainty Assessment. To investigate the influence ofuncertainty in the chemical-specific input parameters forthe characterization factors, an uncertainty estimation wasobtained by means of Monte Carlo simulation using LatinHypercube Sampling with Crystal Ball version 7.1.2 (18). ThisMonte Carlo simulation consisted of 10,000 iterations. Foreach iteration, the ratio (CFtotal)/(CFparent) was calculated,where CFtotal is the characterization factor including TPs, andCFparent is the characterization factor of the parent compoundonly.

A log-normal uncertainty distribution was applied for thechemical-specific parameters vapor pressure, solubility, Kow,Koc, and degradation half-lives. Some advantages of the log-normal distribution are that it avoids negative numbers,captures a large value range, and the uncertainty in manyprocesses and parameters follows a skewed distribution (19).Uncertainties in fractions of formation were treated astriangular distributions (0, likeliest, 1) because the likeliestvalues were available, and fractions of formation cannot beless than 0 or more than 1 (4). The maximum fraction of eachTP obtained in a degradation step was 1. For compoundsthat yielded more than one transformation product (like DDT)10,000 iterations, with realistic fraction of formation data(i.e., which led to a total fraction of 1 or less) were obtainedwith the Monte Carlo simulation. To calculate effect factors,data on the slope factors (S) and the geometric mean toxicityof chemical x (10µx) were required (see eq 3). Uncertaintydistributions were attributed to S according to the work ofVan Zelm et al. (15, 16). Uncertainty in µx was treated as aStudent’s t-distribution, following Aldenberg and Jaworska(20). In the case of EC50 data derived with QSAR estimations,we added an uncertainty measure for the EC50 itself basedon the work of Reuschenbach et al. (21). The SupportingInformation provides more details on the applied uncertaintydistributions of S and µx.

Data Set. We calculated characterization factors for thechemicals 2,4-D, alachlor, atrazine, bromoxynil-octanoate,chlorothalonil, chlorpyrifos, DDT, dicamba, diuron, glypho-sate,R-HCH, heptachlor, malathion, mecoprop-p, orbencarb,

and perchloroethylene in the freshwater environment andemitted to air, freshwater, and agricultural soil. These 16chemicals were chosen based on available information forthe transformation pathways, availability of fraction offormation and degradation data, and availability of experi-mental toxicity data for the parent compound. For eachparent compound and transformation product, toxicity datahad to be available for at least three test species, as Van Zelmet al. (16) showed that uncertainty drastically decreases whenthree instead of two test species are available.

If available, TMoA-specific slope factors were applied inthe calculations (15, 16). The TMoA-specific slope factor wasnot available for 13 out of 64 parent compounds and TPsincluded in this analysis. For these substances, the weightedaverage slope factor of 0.55, as calculated from Van Zelm etal. (15, 16), was taken. We calculated geometric meantoxicities with acute freshwater toxicity data (EC50). EC50 datawere preferably based on experiments (23-26). Experimentalecotoxicity data were, however, scarce for the TPs. TheECOSAR (ecological structure activity relationship) modelprovided by the U.S. Environmental Protection Agency (27)was used to estimate toxicity data when measured data werenot available. Effect factors for 33 out of 48 transformationproducts were based on ECOSAR only.

Uncertainty distribution ranges were based on experi-mental data (see literature sources in Table S1 in theSupporting Information). If no chemical-specific uncertaintydata for the above-mentioned parameters were available,uncertainty distributions specific for pesticides and their TPswere taken from Rikken et al. (28), and distributions for theTPs of perchloroethylene were taken from MacLeod et al.(29). The transformation schemes, degradation rates, frac-tions of formation, physicochemical properties, TMoAs, andall ecotoxicity data, including their (literature) sources, forthe 16 parent compounds and their TPs are listed in theSupporting Information.

Case Study. A case study that addresses the applicationof atrazine to corn in the United States was performed toshow the application of the obtained characterization factors.We assessed whether the ratio of the impact scores for theecotoxicity of atrazine application and the application ofsubstitute pesticides changed with and without the inclusionof transformation products after a possible atrazine ban. TheImpact Score (ISecotox in yr per kg corn) was calculated asfollows:

where ARx is the application rate of pesticide x (kg per kgcorn), and Mx,a and Mx,s are the emissions of pesticide x toair and agricultural soil (in fraction of total emission),respectively. CFx,a and CFx,s are the characterization factorsfor emissions to air and soil, respectively.

We followed the work of Tesfamichael and Kaluarachchi(30), which stated that 63% of the corn treated with atrazinein the United States in 2002 could be replaced with 2,4-D(12%), bromoxynil (11%), dicamba (27%), and nicosulfuron(13%). Given that another 10% of atrazine is replaced by anunknown mix of pesticides, we left this 10% out of thecomparison. Application rates were obtained from the ratesof pesticides per hectare (30) and corresponding corn yieldsper hectare (31). Standard air emission factors for pesticidespresented by the U.S. EPA (32) were used to estimate theamount of pesticides applied on agricultural soil that isreleased to air. New characterization factors were calculatedfor this case study for bromoxynil (first generation trans-formation product of bromoxynil-octanoate) and its TPs, andfor nicosulfuron (see Table S5). For nicosulfuron, no infor-mation on possible TPs was available and therefore only the

FFxfta)

∆Cta

∆Mx(5)

ISecotox ) ∑x

[ARx × (Mx,a × CFx,a + Mx,s × CFx,s)] (6)

VOL. 44, NO. 3, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 1005

parent compound was taken into account. All pesticide-specific input data can be found in the Supporting Informa-tion.

ResultsCharacterization Factors. Figure 1 shows the increase incharacterization factors when transformation products aretaken into account for the 16 chemicals included for emissionsto air, freshwater, and agricultural soil (CFtotal/CFparent).Relatively high CFtotal/CFparent ratios (>10) were found foralachlor, bromoxynil-octanoate, chlorothalonil, DDT, hep-tachlor, and orbencarb. The largest ratio was found for thecharacterization factors of bromoxynil-octanoate emitted toagricultural soil, with a typical increase of more than 5 ordersof magnitude. Taking into consideration uncertainty, therewas a 5% chance that this increase was greater than 14 ordersof magnitude. For emissions to air and freshwater, theincrease in characterization factors of bromoxynil-octanoatewas most prominent, with a typical increase of around 4orders of magnitude, but which could be up to a 12-13 ordersof magnitude increase (95% confidence interval). For 2,4-D,atrazine, chlorpyrifos, diuron, malathion, and perchloroet-hylene, the median increase in CF was less than a factor of

2 for each emission compartment, but could be up to 6 ordersof magnitude (95%).

In terms of the likelihood that taking TPs into accountresults in a significant increase in CF, these results can beexpressed as follows: For four chemicals (bromoxynil-octanoate, chlorothalonil, DDT, and heptachlor) it is likely(at least a 50% chance) that the CF increases by more thana factor of 10 for emissions to every compartment. For fivechemicals (alachlor, bromoxynil-octanoate, chlorothalonil,heptachlor, and orbencarb) there is a slight chance (at least25%) that the CF increases by more than a factor of 100.

The uncertainty in the characterization factor of the TPsis systematically greater than the uncertainty in the char-acterization factor of the parent compound, except forheptachlor. Figure S2 shows the uncertainty in the charac-terization factors of each parent compound, its TPs, and thetotal CF. Each median individual fate, effect, and charac-terization factor, and the fifth and 95th percentiles are inTable S5.

Case Study. Figure 2 displays the impact scores for theapplication of atrazine and substitute pesticides on corn.The median impact score shows a decline of 2 orders ofmagnitude upon substitution when the fate and effects of

FIGURE 1. Ratio of CFtotal/CFparent for 16 chemicals emitted to air (A), freshwater (B), and agricultural soil (C). The center of each boxequals the median ratio, the edges of each box represent the 25th and 75th percentiles, and the whiskers represent the 5th and 95thpercentiles of uncertainty in the ratio CFtotal/CFparent.

1006 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 3, 2010

only the parent compounds are included. When TPs are takeninto account, however, the median impact score for atrazineis only 3.6 times greater than the score of its replacingpesticides. The decrease in impact score caused by an atrazineban is not statistically significant (R ) 0.1, two sidedconfidence interval). A standard z-test on the log-transformedimpact scores revealed that this was the case both when TPswere not taken into account (p ) 0.17), and when TPs wereconsidered (p ) 0.86).

Characterization factors of bromoxynil caused the greatestincrease in impact scores when including transformationproducts, and uncertainty in the CFs of bromoxynil alsocontributed most to the increase in uncertainty in impactscores. Considering each replacement pesticide, 2,4-Dseemed the best alternative to replace atrazine for use oncorn. The impact scores were lower than the atrazine impactscores when TPs were included in the analysis (p ) 0.03),just as when only the parent compound was taken intoaccount (p ) 0.01).

DiscussionContribution of Transformation Products. In this study,characterization factors for freshwater ecotoxicity, whichinclude the fate and effects of transformation products, werecalculated for 16 chemicals. Fate as well as effect could bethe decisive factor for the increase in characterization factorswhen TPs are taken into account. CFs for bromoxynil-octanoate increased the most, which was mainly caused bythe large toxicity of the TPs compared with the toxicity of theparent compound (23, 33) and the small TMoA-specific slopefactor S of bromoxynil-octanoate (15). The effect factor ofbromoxynil was 3 orders of magnitude greater than that ofits parent compound, and four of the other TPs were alsoestimated to be more toxic than bromoxynil-octanoate.Moreover, the parent compound degrades relatively fast inthe environment, while some of the TPs, such as 4-OH-bezonitrile in soil, do not. For alachlor, chlorothalonil, DDT,heptachlor, and orbencarb, the characterization factorstypically increased by more than a factor of 10 when TPswere taken into account. Each transformation product ofchlorothalonil and heptachlor was more persistent in theenvironment than its parent compound (5, 34, 35). DDT wasalmost completely degraded to DDE and DDD. DDDcontributed the greatest part of the total characterizationfactor, which was caused by the relatively high persistenceof DDD in water (36). The transformation products of alachlorand orbencarb have approximately the same physicochem-ical properties as their parent compounds, which resultedin a significant contribution to the total characterizationfactors.

For 5 out of the 16 chemicals, the contribution of theparent compound to the total characterization factorsdominated in every compartment. Atrazine, chlorpyrifos,diuron, and malathion were much more toxic than their TPs(23, 24, 26, 27). Malaoxon is formed from malathion in aironly (7), which explains why there was only a small increasein CF for emissions to freshwater and agricultural soil.Furthermore, malaoxon is less persistent in air and has alower toxicity than malathion (7, 23, 25). Perchloroethylenewas more persistent and had a greater toxicity than its TPs(4, 6, 23, 27).

In Figure 1, the characterization factors are compared ina relative way. As the chemicals are mostly pesticides,environmental impacts of these chemicals are relatively large.Perchloroethylene was the only nonpesticide and had arelatively low characterization factor compared to the otherchemicals (see Table S5). Bromoxynil-octanoate also had arelatively low CF; inclusion of TPs, however, led to a significantincrease in the CF of this chemical. The greatest median CFsin this research were found for diuron, atrazine, andchlorpyrifos; inclusion of TPs for these chemicals hardly ledto an increase in their CFs.

Figure 3 shows how our results compare to those of Fenneret al. (4), Gasser et al. (5), and Schenker et al. (6), who tookinto account persistence of transformation products. Ourcalculation routine for the characterization factors concep-tually differs from the persistency concept in two ways. First,our characterization factors assess the fate of an emission inrelation to the freshwater environment, while the concept ofpersistency addresses the overall residence time of a chemicalover all compartments involved. Second, our characterizationfactors include the toxicity of a chemical, while this is excludedfrom the persistency calculations. For several chemicals, theincrease in CF was greater than the increase in persistence,which was mainly caused by the inclusion of toxicity in ourwork. The transformation products of heptachlor and bro-moxynil-octanoate were much more toxic than their parentcompounds, while the TPs of dicamba, R-HCH, and meco-prop-p were equally or slightly more toxic when comparedto their parent compounds. The increase in CF was less thanthe increase in persistence for atrazine, chlorpyrifos, diuron,glyphosate, and perchloroethylene, because these five pes-ticides were considerably more toxic than their transforma-tion products. For the other chemicals, the differences weremainly caused by differences in persistence calculationsversus fate factor calculations. For example, for alachlor andchlorothalonil, the increase in CFs was much greater thanthe increase in persistence because the Kocs for the parentcompounds were orders of magnitude greater than the Kocsof their TPs. The transfer from the soil (emission compart-ment) to the water (receiving compartment) was much greater

FIGURE 2. Impact scores for the fate and effects of pesticideapplication on corn with only the parent compound (PC) andincluding the transformation products (TPs), comparing twoapplication scenarios. The center of each box equals themedian score, the edges of each box represent the 25th and75th percentiles, and the whiskers represent the 5th and 95thpercentiles of the impact score.

FIGURE 3. Median increase in characterization factors from ourresearch compared to the increase in persistence fromprevious research, when including transformation products, foremissions to air or soil. Persistence scores of Gasser et al. (5)were calculated for emissions of 90% to soil and 10% to air.

VOL. 44, NO. 3, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 1007

for the TPs than for the parent compounds, causing a majorincrease in CFtotal. Persistence calculations, on the contrary,also included soil as a receiving compartment.

Uncertainty Analysis. Uncertainty in the effect factor wasdominant in the overall uncertainty for the total character-ization factors. The primary cause was the large uncertaintyin one or more effect factors of the transformation products.The effect factor uncertainty (95% confidence interval) wasat least 6 orders of magnitude for chemicals where no TMoA-specific slope factor was available. Estimation of µx withECOSAR caused an uncertainty of up to 8 orders of mag-nitude. Models to more accurately predict ecotoxicity of(transformation) chemicals are underway (37-39). Ap-propriate assignment of the toxic mode of action is, however,still a weak point in the assessment (38). The acute toxicitymodeling component of ASTER, for example, assesses thestructural characteristics of chemicals and evaluates whetheror not a compound contains specific functional moietiesthat are associated with a specific TMoA. If the structuralcharacteristics of a chemical do not suggest that a specificTMoA may be involved, nonpolar narcosis is assigned. Thisresults in a large variety of chemicals within the nonpolarnarcosis category, which creates large uncertainties in theTMoA-specific slope factor S. More information and betterdefinitions of toxic modes of action are required.

We quantified the added chemical-specific parameteruncertainty for the inclusion of transformation products inthe calculation of characterization factors for ecotoxicity. Adecrease in uncertainty was found for the characterizationfactors of heptachlor when including its transformationproduct. Uncertainty in the effect factors of heptachlorep-oxide, the transformation product of heptachlor, was lowcompared with the uncertainty in the effect factor of its parentcompound. Fenner et al. (4) found similar trends for thejoint persistence (including TPs) of nonylphenol polyethoxy-lates and atrazine compared with the primary persistence(without TPs). The most influential input parameters of thejoint persistence showed lower uncertainties than the mostinfluential parameters of the primary persistence. We foundthe opposite for the uncertainty in the CF for atrazine, becausein our study the effect portion was found to be the most in-fluential factor for the uncertainty in the CF of this chemical.

The complete degradation with all intermediate sub-stances was not included for the compounds in our analysis;we only took into account known TPs. For atrazine, a moreextensive transformation pathway was suggested by Fenneret al. (4), but they showed that the first degradation steps aremost important, as the chemicals that are formed in thesesteps are much more persistent than the subsequent TPs.For each transformation pathway, the complete degradationscheme may not be known, therefore, more research isneeded to find out whether additional transformations occurin the environment (6).

Relevancy of the Study. Inclusion of the fate and effectsof transformation products inevitably leads to greatercharacterization factors. This increase varies from negligibleto an increase of more than 5 orders of magnitude. Forbromoxynil-octanoate, chlorothalonil, DDT, and heptachlor,there is more than a 50% chance for emissions to everycompartment that the CF will increase by more than a factorof 10. We also found that the CFs of the chemicals consideredcan be highly uncertain. Particularly, the reliability of thetoxicity data for the TPs and the toxic mode of action-specificdata need to be improved.

The case study of pesticide application on corn showedthat, without the inclusion of the effects of TPs, it is likelythat pesticide impact scores will be reduced by 2 orders ofmagnitude when atrazine is banned (Figure 2). When TPsare included, however, this reduction would likely be lower(less than 1 order of magnitude). Uncertainty estimates

showed that before the inclusion of TPs a decrease in impactscore was not significant, while after inclusion of TPsuncertainty in the impact scores was even greater. Therefore,it is not certain that replacement of atrazine will cause adecrease in freshwater ecotoxicity effects. We found that itwas only beneficial to replace atrazine application on cornwith 2,4-D and no other pesticides investigated in the casestudy in this paper.

Although the results are highly uncertain, we show forseveral chemicals that the exclusion of transformationproducts can lead to unjustified conclusions concerning thelife cycle impacts for freshwater ecotoxicity of these chemi-cals. We should note, however, that the chemicals in thisresearch were selected because they were known to havepersistent transformation products; therefore, they are nota random selection providing results that can be directlyapplied to other chemicals. If reliable chemical-specific dataare available, CFs can be substantially greater when trans-formation products are included, while uncertainty does notneed to increase.

AcknowledgmentsThis research was financially supported by FP7 EU ProjectCADASTER (Contract 212668). We are grateful to KathrinFenner for providing data on fractions of formation and forfruitful discussions. Furthermore, we thank Arjen Wintersenfor making the RIVM e-toxBase available and to AnitaPomplun for providing data from U.S. EPA’s ASTER. We thankthree anonymous reviewers for their valuable comments.

Supporting Information AvailableTransformation schemes, fractions of formation, chemical-specific and ecotoxicity input data, characterization factors(CFs), CF comparison tables for emissions to freshwater andagricultural soil, and input data for the case study. Thismaterial is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited(1) Rosenbaum, R. K.; Bachmann, T. M.; Swirsky Gold, L.; Huijbregts,

M. A. J.; Jolliet, O.; Juraske, R.; Koehler, A.; Larsen, H. F.; Macleod,M.; Margni, M.; McKone, T. E.; Payet, J.; Schuhmacher, M.; Vande Meent, D.; Hauschild, M. Z. USEtoxsThe UNEP-SETACtoxicity model: recommended characterisation factors forhuman toxicity and freshwater ecotoxicity in Life Cycle ImpactAssessment. Int. J. LCA 2008, 13, 532–546.

(2) Schulze, C.; Jodicke, A.; Scheringer, M.; Margni, M.; Jolliet, O.;Hungerbuhler, K.; Matthies, M. Comparison of different life-cycle impact assessment methods for aquatic toxicity. Environ.Toxicol. Chem. 2001, 20, 2122–2132.

(3) Fenner, K.; Scheringer, M.; Hungerbuhler, K. Persistence ofparent compounds and transformation products in a level IVmultimedia model. Environ. Sci. Technol. 2000, 34, 3809–3817.

(4) Fenner, K.; Scheringer, M.; Hungerbuhler, K. Joint persistenceof transformation products in chemicals assessment: casestudies and uncertainty analysis. Risk Anal. 2003, 23, 35–53.

(5) Gasser, L.; Fenner, K.; Scheringer, M. Indicators for the exposureassessment of transformation products of organic micropol-lutants. Environ. Sci. Technol. 2007, 41, 2445–2451.

(6) Schenker, U.; Scheringer, M.; Hungerbuhler, K. Including de-gradation products of persistent organic pollutants in a globalmulti-media box model. Environ. Sci. Pollut. Res. 2007, 14, 145–152.

(7) Cahill, T. M.; Mackay, D. A high-resolution model for estimatingthe environmental fate of multi-species chemicals: applicationto malathion and pentachlorophenol. Chemosphere 2003, 53,571–581.

(8) Sinclair, C. J.; Boxall, A. B. A. Assessing the ecotoxicity of pesticidetransformation products. Environ. Sci. Technol. 2003, 37, 4617–4625.

(9) Pennington, D. W.; Payet, J.; Hauschild, M. Aquatic ecotoxi-cological indicators in life-cycle assessment. Environ. Toxicol.Chem. 2004, 23, 1796–1807.

1008 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 3, 2010

(10) Huijbregts, M. A. J.; Struijs, J.; Goedkoop, M.; Heijungs, R.;Hendriks, A. J.; Van de Meent, D. Human population intakefractions and environmental fate factors of toxic pollutantsin life cycle impact assessment. Chemosphere 2005, 61, 1495–1504.

(11) Van Zelm, R.; Huijbregts, M. A. J.; Russell, M. H.; Jager, T.; Vande Meent, D. Modeling the environmental fate of perfluorooc-tanoate and its precursors from global fluorotelomer acrylatepolymer use. Environ. Toxicol. Chem. 2008, 27, 2216–2223.

(12) Vermeire, T.; Rikken, M.; Attias, L.; Boccardi, P.; Boeije, G.;Brooke, D.; De Bruijn, J.; Comber, M.; Dolan, B.; Fischer, S.;Heinemeyer, G.; Koch, V.; Lijzen, J.; Muller, B.; Murray-Smith,R.; Tadeo, J. European union system for the evaluation ofsubstances: the second version. Chemosphere 2005, 59, 473–485.

(13) Van Zelm, R.; Huijbregts, M. A. J.; Van de Meent, D. USES-LCA2.0: a global nested multi-media fate, exposure and effectsmodel. Int. J. LCA 2009, 14, 282–284.

(14) Jolliet, O.; Hauschild, M. Modeling the influence of intermittentrain events on long-term fate and transport of organic airpollutants. Environ. Sci. Technol. 2005, 39, 4513–4522.

(15) Van Zelm, R.; Huijbregts, M. A. J.; Harbers, J. V.; Wintersen, A.;Struijs, J.; Posthuma, L.; Van de Meent, D. Uncertainty in msPAF-based ecotoxicological effect factors for freshwater ecosystemsin life cycle impact assessment. Integr. Environ. Assess. Manage.2007, 3, 203–210.

(16) Van Zelm, R.; Huijbregts, M. A. J.; Posthuma, L.; Wintersen, A.;Van de Meent, D. Pesticide ecotoxicological effect factors andtheir uncertainties for freshwater ecosystems. Int. J. LCA 2009,14, 43–51.

(17) De Zwart, D.; Posthuma, L. Complex mixture toxicity for singleand multiple species: proposed methodologies. Environ. Toxicol.Chem. 2005, 24, 2665–2676.

(18) Decisioneering Inc. Crystal Ball, version 7.1.2. Forecasting andrisk analysis for spreadsheet users; Decisioneering; Denver, CO,2005.

(19) Slob, W. Uncertainty analysis in multiplicative models. RiskAnal. 1994, 14, 571–576.

(20) Aldenberg, T.; Jaworska, J. S. Uncertainty of the hazardousconcentration and fraction affected for normal species sensitivi-ty distributions. Ecotoxicol. Environ. Saf. 2000, 46, 1–18.

(21) Reuschenbach, P.; Silvani, M.; Dammann, M.; Warnecke, D.;Knacker, T. ECOSAR model performance with a large test setof industrial chemicals. Chemosphere 2008, 71, 1986–1995.

(22) Cahill, T. M.; Cousins, I.; Mackay, D. General fugacity-basedmodel to predict the environmental fate of multiple chemicalspecies. Environ. Toxicol. Chem. 2003, 22, 483–493.

(23) RIVM; National Institute for Public Health and the Environment,2008.

(24) Stratton, G. W. Effects of the herbicide atrazine and itsdegradation products, alone and in combination, on pho-totrophic microorganisms. Arch. Environ. Contam. Toxicol. 1984,13, 35–42.

(25) Aker, W. G.; Hu, X.; Wang, P.; Hwang, H. Comparing the relativetoxicity of malathion and malaoxon in blue catfish Ictalurusfurcatus. Environ. Toxicol. 2008, 23, 548–554.

(26) Kotrikla, A.; Gatidou, G.; Lekkas, T. D. Toxic effects of atrazine,deethyl-atrazine, deisopropyl-atrazine and metolachlor onchlorella fusca var-fusca. Global Nest: The Int. J. 1999, 1, 39–45.

(27) USEPA. Ecological structure activity relationships (ECOSAR) v.1.00a; U.S. EPA: Washington, DC, 2009.

(28) Rikken, M. G. J. ; Van Wijnen, H. J.; Linders, J. B. H. J.; Jager, D. T.Uncertainty analysis of USES 3.0. Improving risk managementthrough probabilistic risk assessment of agricultural pesticides;RIVM 601450011/2002; National Institute of Public Health andthe Environment: Bilthoven, The Netherlands, 2002.

(29) MacLeod, M.; Fraser, A. J.; Mackay, D. Evaluating and expressingthe propagation of uncertainty in chemical fate and bioaccu-mulation models. Environ. Toxicol. Chem. 2002, 21, 700–709.

(30) Tesfamichael, A. A.; Kaluarachchi, J. J. A methodology to assessthe risk of an existing pesticide and potential future pesticidesfor regulatory decision-making. Environ. Sci. Policy 2006, 9,275–290.

(31) USDA. Corn: yield by year, US; 2009; available from http://www.nass.usda.gov/Charts_and_Maps/Field_Crops/cornyld.asp.

(32) USEPA. AP42. In Compilation of Air Pollutant Emission Factors,Volume 1: Stationary Point and Area Sources, 5th edition; U.S.EPA: Washington, DC, 1995.

(33) USEPA. Assessment Tools for the Evaluation of Risk, NationalHealth and Environment; U.S. EPA, Effects Research Laboratory,Mid-Continent Ecology Division: Duluth, MN; 2004.

(34) Mackay, D.; Shiu, W. Y.; Ma, K. C. Illustrated Handbook ofPhysical-Chemical Properties and Environmental Fate for Or-ganic Chemicals. Volume V: Pesticide Chemicals; Lewis: BocaRaton, FL, 1997.

(35) Howard, P. H.; Boethling, R. S.; Jarvis, W. F.; Meylan, W. M.;Michalenko, E. M. Handbook of Environmental DegradationRates; Lewis: Chelsea, MI, 1991.

(36) Hertwich, E. G.; McKone, T. E.; Pease, W. S. Parameteruncertainty and variability in evaluative fate and exposuremodels. Risk Anal. 1999, 19, 1193–1204.

(37) Escher, B. I.; Baumgarter, R.; Lienert, J.; Fenner, K. Predictingthe ecotoxicological effects of transformation products. In Boxall,A. B. A. Handbook of Environmental Chemistry - Degradationof Synthetic Chemicals in the Environment; Springer: Heidelberg,Germany, 2008; pp 1-33.

(38) Jager, T.; Posthuma, L.; De Zwart, D.; Van de Meent, D. Novelview on predicting acute toxicity: Decomposing toxicity data inspecies vulnerability and chemical potency. Ecotoxicol. Environ.Saf. 2007, 67, 311–322.

(39) Porcelli, C.; Boriani, E.; Roncaglioni, A.; Chana, A.; Benfenati,E. Regulatory perspectives in the use and validation of QSAR.A case study: DEMETRA model for daphnia toxicity. Environ.Sci. Technol. 2008, 42, 491–496.

ES9021014

VOL. 44, NO. 3, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 1009