Chemosphere 100 (2014) 175–181

Contents lists available at ScienceDirect

Chemosphere

journal homepage: www.elsevier .com/locate /chemosphere

Bridging the gap between life cycle inventory and impact assessmentfor toxicological assessments of pesticides used in crop production

0045-6535/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.chemosphere.2013.11.037

⇑ Corresponding author. Tel.: +31 243652923.E-mail address: r.vanzelm@science.ru.nl (R. van Zelm).

Rosalie van Zelm a,⇑, Pyrène Larrey-Lassalle b, Philippe Roux b

a Department of Environmental Science, Institute for Water and Wetland Research, Radboud University Nijmegen, P.O. Box 9010, 6500 GL Nijmegen, The Netherlandsb Irstea, Research Unit: Information & Technologies for Agro-processes, 361 rue JF Breton, 34196 Montpellier, France

h i g h l i g h t s

� A framework is provided to link life cycle inventory and impact assessment for pesticides.� The framework prevents overlaps and gaps between LCI and LCIA modeling.� Efficient and inefficient management practices can be distinguished.

a r t i c l e i n f o

Article history:Received 30 July 2013Received in revised form 7 November 2013Accepted 13 November 2013Available online 7 December 2013

Keywords:EcotoxicityHuman toxicityTechnosphereLife cycle assessmentFateExposure pathway

a b s t r a c t

In Life Cycle Assessment (LCA), the Life Cycle Inventory (LCI) provides emission data to the variousenvironmental compartments and Life Cycle Impact Assessment (LCIA) determines the final distribution,fate and effects. Due to the overlap between the Technosphere (anthropogenic system) and Ecosphere(environment) in agricultural case studies, it is, however, complicated to establish what LCI needs to cap-ture and where LCIA takes over. This paper aims to provide guidance and improvements of LCI/LCIAboundary definitions, in the dimensions of space and time. For this, a literature review was conductedto provide a clear overview of available methods and models for both LCI and LCIA regarding toxicologicalassessments of pesticides used in crop production. Guidelines are provided to overcome the gapsbetween LCI and LCIA modeling, and prevent the overlaps in their respective operational spheres.

The proposed framework provides a starting point for LCA practitioners to gather the right data and usethe proper models to include all relevant emission and exposure routes where possible. It is also able topredict a clear distinction between efficient and inefficient management practices (e.g. using differentapplication rates, washing and rinsing management, etc.). By applying this framework for toxicologicalassessments of pesticides, LCI and LCIA can be directly linked, removing any overlaps or gaps in betweenthe two distinct LCA steps.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Over the last years, a significant number of Life Cycle Assess-ment (LCA) studies were conducted on agricultural products. Sev-eral of them are reported in scientific journals (e.g. Basset-Mensand Van der Werf, 2005; Mouron et al., 2006; Torrellas et al.,2012), but most of them are in the form of reports commissionedby stakeholders and oftentimes in the national language (e.g. Blonket al., 2007; Schmidt, 2008; Weidema et al., 2008; Ademe, 2010).To perform such LCAs, a detailed inventory of all resources usedand all emitted chemicals needs to be conducted. This includesan inventory of agricultural inputs and outputs, e.g. fertilizersand pesticides, that are of importance. Termed the Life CycleInventory (LCI), this phase supplies the amount of inputs used

per functional unit, generally per kg of agricultural product orper hectare, but also assesses the amount of them emitted to eachenvironmental compartment. Subsequently, in the Life Cycle Im-pact Assessment (LCIA) step, the chemical emissions are convertedinto impact scores for ecotoxicity and human toxicity.

For classic industrial applications, the flux exchanges betweenTechnosphere, the studied anthropogenic system, and Ecosphere,the natural environment, can often be easily assessed becausethe system boundaries, e.g. factory walls boundaries, dischargepipes & chimneys are clearly defined. For agricultural applications,the boundaries between Technosphere and Ecosphere are not soeasy to define. An agricultural field and its soil can be set to belongto the Ecosphere, as is done by e.g. Ecoinvent (Frischknecht et al.,2007), or the Technosphere, as is done by e.g. PestLCI (Dijkmanet al., 2012). Furthermore, substance transfers between environ-mental compartments are complex. Given the unclear boundariesbetween Technosphere and Ecosphere in agricultural case studies

176 R. van Zelm et al. / Chemosphere 100 (2014) 175–181

it is complicated to establish what LCI needs to characterizeregarding degradation and partitioning of the pesticides in air,water, and soil at the local scale, and where LCIA takes over, mov-ing onto a larger temporal and spatial scale such as long-rangetransmission of air pollutants at regional, continental, and globalscale. Up to now, LCA practitioners, who are often not even awareof the boundaries problem, have been using different hypothesesto build agricultural inventories. They apply, for example, a regio-nal or global scale model of substances transfers in the LCI phase(e.g. EMEP/EEA, 2009), or they apply a simplified approach assum-ing that pesticides are entirely emitted to the soil compartment(Nemecek and Kägi, 2007; Panichelli et al., 2009), or that 85% isemitted to soil, 5% to crops, and 10% to air (Margni et al., 2002;Audsley et al., 2003). Previous framing and conceptualization ofagricultural LCA focused on combining either LCI or LCIA inputsfrom various methods or applicable to various countries (e.g.Bentrup et al., 2004; Kah and Brown, 2007). To date, neither cleardistinction nor guidance is provided on how to apply LCI and LCIAmodels so they link up with respect to toxicological assessments ofpesticides applied in agriculture.

This paper aims to provide guidance to better define the bound-aries between what should be included in LCI and where LCIA takesover. For this, we provide an overview of the possibilities of bothLCI and LCIA methods regarding toxicological assessments of in-puts used in crop production. Guidelines are provided to overcomethe gaps between LCI and LCIA modeling and remove the overlaps,as well as to harmonize the comparisons of agricultural LCAresults.

2. Current practice and available methodologies

2.1. From spraying to environmental damage: biophysical phenomenaand farmers practices

When a pesticide is released into the environment duringspraying, not the entire applied chemical reaches the target crop– for fungicides and insecticides, or weed – for herbicides). Differ-ent fractions are directly emitted to air (drift), soil, and, in somecases, surface water (Fig. 1), where non-target organisms can be af-fected. The amount of pollution depends on both the substance ap-plied and the pesticide application practice. The latter is a

Fig. 1. Examples of transfer mechanisms from pestic

combination of the technology used (type of sprayer and nozzles)and human practices, such as the type of equipment used andadjustment or timing of pesticide application depending on weath-er condition and the evolution of crop diseases. If this practice isefficient, i.e. applied at the correct dose directly where and whenit is needed, damage is expected to be small. These drift water con-taminations can be drastically reduced directly by the use of lowdrift equipment or indirectly by buffer zones, i.e. zones with grassor specific crops around the field, where spraying is prohibited (seeFig. 1) (De Snoo and De Wit, 1998). The buffer zones increase thedistance that runoff water has to cover in order to reach water-course. Furthermore, in the case of aerial spraying, buffer zoneswill catch over-sprays, leading to reduced environmental damage.Spraying by airplanes can cause pollution in areas outside the cropfield via spray drift. This leads on the one hand to environmentaldamage in a larger area and on the other hand, if the phytosanitaryefficacy is low, this increases the need for more pesticide treat-ments on the crop field. In addition to emissions during pesticideapplication, contaminations may occur before, during the prepara-tion of the pesticide formulation while filling the sprayer tank, andafter the treatment, by managing the mixture remaining in thesprayer tank and during washing and rinsing operations (Ramwellet al., 2007). The fate of pesticides packaging residues could also bea contributor to those impacts. These emissions can be drasticallyreduced by the use of integrated technologies, such as filling/wash-ing areas associated with purification systems using a biological orphysicochemical treatment (see e.g. De Wilde, 2009). In LCA, theaim is to identify average impacts, and not a worst case scenario.Therefore, it should be noted that contaminations due to inefficientbehavior should only be included when appropriate data are avail-able, i.e. actual data regarding these inefficient practices, or whenthe ultimate goal of the LCA is to compare agricultural practices.

Commercial pesticide products rarely consist only of their pureactive substance, but are usually formulated with several otheringredients. Pesticide formulations have been developed to im-prove the efficiency of the active substance against pests, fungaldiseases, and weeds. Surfactants are among the most importantcomponents used and can improve the biological activity by mod-ifying spray droplet size, drift phenomenon, retention, and spread-ing on leaf surfaces or by enhancing immediate uptake andpenetration of the active ingredient into the crops. Formulationmay also improve the properties of a chemical for handling,

ide spraying to emissions in air, soil, and water.

R. van Zelm et al. / Chemosphere 100 (2014) 175–181 177

storage, and application, and may substantially influence treat-ment effectiveness and operators safety. This formulation may af-fect their distribution among environmental compartments, andenvironmental effects (Benachour and Seralini, 2009; Lee et al.,2009). LCA practitioners often only have access to the main activesubstances used and not to the detailed composition of full formu-lations and additives.

Multimedia transport phenomena, i.e. evaporation, volatiliza-tion, spray drift, runoff, leaching, absorption, adsorption, and washoff cause pesticides to end up in other environmental compart-ments than initially emitted (see Fig. 1). Extensive field tests havebeen carried out on some of these phenomena, such as spray driftand subsequent deposition (Ganzelmeier, 2000; Gil and Sinfort,2005; Wang and Rautmann, 2008). Wang and Rautmann (2008)concluded that wind speed, agricultural equipment (nozzle type,spray pressure), and relative humidity have the strongest effectson spray drift. For some other phenomena, the knowledge is poorerbut research is on-going. For instance, few experimental data setsof pesticide volatilization from plants at the field scale are available(Bedos et al., 2010).

The time-integrated amounts of pesticides that remain in eachcompartment depend on degradation and transformation, immobi-lization, and multimedia transport: the three main aspects of fate.Some phenomena such as dispersion in the atmosphere or leachingmay occur a few minutes after pesticide application, while groundtransfers and associated microbial and chemical breakdown mayresult in pollution transfers in water bodies after several weeksor months.

Pesticide residues in air, water, soil, groundwater, and crops canbe inhaled or taken up by humans, plants, and animals where theyharm the species by causing diseases and death. Chemicals enterthe human body via oral ingestion, dermal contact, or inhalation,where they can cause respiratory effects, cancer, or other diseases.Intake of pesticides through food ingestion has been shown to bethe dominant exposure route of these chemicals in humans (Luet al., 2008; Juraske et al., 2009; Fantke et al., 2012). Chemical up-take can accumulate along the food chain, leading to impacts onwarm-blooded predators as well (Golsteijn et al., 2012). Further-more, chemical transformation products in the environment ormetabolites in the human body can on their turn be harmful aswell (Van Zelm et al., 2010). The fate, distribution and effects ofpesticides will vary depending on location, e.g. habitat characteris-tics, local stressors, background concentrations, weather, and time,e.g. season characteristics and life stage sensitivity.

2.2. From spraying to environmental damage: models

2.2.1. Models for LCIVarious sources provide specific factors for volatilization,

groundwater leaching, and drift based on experiments (see Sup-porting information). More broader overviews are provided by anumber of models. Pesticide loss to the air (drift) was studied usingfuzzy inference models (Gil et al., 2008). PEARL (pesticide emissionassessment at regional and local scales, Leistra et al. (2001)) is aone-dimensional numerical model of pesticide behavior in thesoil–plant system. It includes simplified approaches for the predic-tion of volatilization from soil and, partially, from plants. To betterdescribe pesticide volatilization, Wolters et al. (2004) used a mech-anistic approach based on a concept of laminar air-boundary. Theirmodel was calibrated on wind tunnel experiments using three dif-ferent pesticides and they showed that 6% to 29% of the pesticidecan be volatilized after 10 d. Audsley et al. (2003) developed sim-plified methods to calculate emissions from pesticide applicationto three environmental compartments (air, surface/groundwater,and soil), and estimated pesticide residues in food based on theirtolerable values. However, their method is generic to climate and

sprayer type, and they do not justify all emission flows. Birkvedand Hauschild (2006) proposed a model named ‘‘PestLCI’’ to esti-mate field emissions of pesticides in agricultural LCA. This model,updated recently (Dijkman et al., 2012), assumes the Technosphereboundary to be the crop field borders (horizontal field area) andfrom 1 m soil depth up to 100 m high in the air. Biophysical phe-nomena such as volatilization, sorption, leave uptake, runoff, andbiodegradation are integrated in the model. As the agriculturalfield in PestLCI is considered a part of the Technosphere, pesticidesapplied to it are not considered as emissions to the environment,but merely intra Technosphere (re-)distribution. In order to reachthe Ecosphere soil (i.e. the environment surrounding the crop field)the pesticides have to migrate outside it, which can only occur viaother compartments (i.e. mainly advective transport via air andwater). The philosophy behind this reasoning is that arable landrepresents highly ‘‘manipulated ecosystems’’, very different fromthe environment that is expected to be protected. This, however,might not be the case for a less intensive practice such as organicfarming or no tillage farming. The PestLCI model structure distin-guishes between primary distribution (leaf and soil deposition aswell as air drift), and secondary distribution (fate processes afterapplication). The fate processes are modeled to end at the startof the first precipitation event after pesticide application. All pesti-cide that reached the top soil is assumed to start leaching into thesubsoil towards the groundwater. In PestLCI, the fraction reaching1 m below the surface is calculated taking into account tillage,macropore transport, soil column properties, meteorological con-ditions, degradation, sorption, etc. and is probably its most com-plex module. PestLCI provides emissions from the Technosphereto surface water, groundwater, and air. As the agricultural soil be-longs to the Technosphere (up to 1 m of depth) there are no emis-sions to agricultural soil. The updated PestLCI version 2.0.4.contains a pesticide database of 94 active ingredients, 25 Europeanclimate profiles, and 7 European soil profiles with differentcompositions.

2.2.2. Models for LCIAIn the LCIA phase, the cause-effect pathway of chemicals caus-

ing toxicity impacts is modeled to come to final impact scores. Thewhole ecosystem is affected by toxic compounds, which, in termsof biodiversity, can be expressed as the potentially affected fraction(PAF) or potentially disappeared fraction (PDF) of species. Humantoxicity can be assessed up to life years disabled (YLD) or life yearslost (YLL). This is called the endpoint, and all points earlier on thecause-effect pathway are referred to as midpoints. At the moment,various models are available for the LCIA of toxicity (see Support-ing information). Most of these models were originally developedfor regional risk assessment purposes. USEtox (Rosenbaum et al.,2008), the so-called scientific consensus model, was built by meth-od developers of the various existing LCIA toxicity models to estab-lish recommended practice in chemical characterization for LCIA.

As in almost all current life cycle inventories, for all LCIA mod-els, emissions are summed up per pollutant regardless of their geo-graphical place of occurrence. USEtox for example, is a genericcontinental-scale model. Inventory outcomes are attributed to ascale that lacks any retrievable relation with a particular region(Rosenbaum et al., 2008). The continental scale in USEtox capturessix environmental compartments, i.e. urban air, rural air, agricul-tural soil, industrial soil, freshwater and coastal marine water.The global scale accounts for impacts outside the continental scale.The emission scenarios are continental emission to urban air, ruralair, freshwater, and agricultural soil. Human exposure throughinhalation of (rural and urban) air and ingestion of drinking water(untreated surface freshwater), leaf crops (exposed produce), rootcrops (unexposed produce), meat, milk, and fish from freshwaterand marine aquatic compartments, are included. For ecotoxicity,

178 R. van Zelm et al. / Chemosphere 100 (2014) 175–181

only effects on the freshwater environment are included(Henderson et al., 2011). The model uses two soil types to accountfor the fraction of agricultural soil relative to the total soil surfaceso as to allow for specific (e.g. pesticide) emissions occurring onagricultural soil only. The soil compartment is modeled as a homo-geneous single-layer compartment to a depth of 10 cm.

USES–LCA has several extra emission compartments, i.e. strato-sphere, continental air (general), sewage treatment plant, seawa-ter, and natural soil. The model also includes marine andterrestrial ecotoxicity (Van Zelm et al., 2009). USES–LCA and IM-PACT 2002 take into account intake of fish from marine waterand groundwater contamination leading to contaminant intakewith drinking water (Huijbregts et al., 2000; Pennington et al.,2005). GLOBOX (Wegener Sleeswijk, 2006) includes region-specificinformation and includes groundwater as a stand-alone compart-ment. Spatial differentiation was investigated with IMPACT 2002(Manneh et al., 2010), but has not been standardized yet.

In the plant uptake, USEtox uses a simplified one-compartmentapproach suitable to account for chemical exposure limiting theroot concentration factor (RCF) for high Kow (>105) compoundsto 200 and distinguishing leaf surfaces from overall above-groundplant tissues when calculating the plant–air partition coefficient.No degradation processes in the vegetation were included. Thiswill overestimate exposures of humans via agricultural produceand meat/milk (Rosenbaum et al., 2008). Exposure by pesticide res-idues in food is excluded (for now) in USEtox due to lack of scien-tific consensus indicating further research needs (Rosenbaum et al.,2011). Other LCIA models, e.g. IMPACT2002 do include vegetationcompartments. More recently, Fantke et al. (2011b) proposedDynamiCROP, a dynamic plant uptake model for the quantificationand valuation of human health impacts caused by the directapplication of pesticides onto food crops and subsequent ingestion.The modeling framework is based on a crop-specific multi-

Fig. 2. Proposed framework to include pesticide app

compartment system, i.e. environmental compartments (atmo-spheric ground layer, root-zone soil layer) and vegetation compart-ments (leaf and fruit surface deposit, leaf, fruit, stem, and thickroot) (Fantke et al., 2011a). DynamiCROP analyzes the mass evolu-tion in the different compartments over time. Important dynamiccontributions are the date of application and harvest times.

Margni et al. (2002) specifically focused on pesticides and in-cluded transfer to groundwater, showing that very few substancesreach this environmental compartment. They assumed that thepesticide concentrations in food correspond to 5% of their respec-tive tolerance values.

Many pesticides can ionize in the environment. Until recent,LCIA models did not properly take into account ionization pro-cesses in fate modeling, while PestLCI includes pH dependency inthe topsoil only. The updated USEtox model now better addressthe dissociation of chemicals and consequently their altered fateand effects, based on the work of Van Zelm et al. (2013).

3. Proposed framework for pesticide inventory and impactassessment

To enable harmonized comparisons of agricultural LCAs results,a consensual flowchart is required to describe system boundariesas well as the biophysical phenomena that have to be taken intoaccount during the LCI phase. This harmonized framework shouldthen enable the use of generic models of pesticide transfers as wellas experimental results when needed, in particular if the availablemodels do not cover worldwide all specific crops, practices, andpedoclimatic conditions. Furthermore, overlaps or gaps betweenLCI and LCIA modeling need to be prevented. With this in mind,Fig. 2 provides a proposal flowchart for pesticide LCI and LCIA,respectively, where both LCA phases link up to each other. It

lication mass flow in LCI and LCIA, respectively.

R. van Zelm et al. / Chemosphere 100 (2014) 175–181 179

describes the main potential contamination as well as the systemboundary in space and time, resulting in an inventory of emissionsin the four LCIA environmental compartments off-site topsoil,groundwater, freshwater (surface), and air. This flowchart also al-lows for improving the description of contamination pathways tobetter differentiate between efficient and inefficient agriculturalpractices, when known to occur in a specific case study. To do so,it includes emissions that occur during the filling of sprayer tankand the washing and rinsing of sprayer and pesticide packaging.These flows are summarized as a box ‘flows during pre-post appli-cation’ in Fig. 2 with more detail given in the Supportinginformation.

We propose that the system boundary, i.e. the Technosphere,for pesticide LCI is the crop field, including buffer zones. Basedon that, water emission factors need to be adjusted depending onwhether buffer zones are present, and on the corresponding aver-age distance to surface water, or on the use of low drift technolo-gies, reducing direct water contamination. That is for instancedone in PestLCI v2, by corrections of emissions done for the fieldslope and the pesticide-specific buffer zone, increasing the distancethat runoff water has to cover in order to cross the Technosphereborders (Dijkman et al., 2012).

Vertically, a 25 m high boundary in the air (above the top of thecrops, including fruit trees and aerial application of the pesticide) isproposed. This height is compatible with the order of magnitude ofdrift field test measurements that were conducted for a number ofsituations and technologies (see e.g. Ganzelmeier, 2000; Gil andSinfort, 2005). At depth, the proposition is to set the LCI systemboundary to the surface of the topsoil and the surface of the crops.In such a simplified approach it is a major issue to limit the com-plexity of modeling the transfers in soils and crops and the re-quired data. This can be handled by LCIA models which areindeed designed to assess chemical fate for long durations, suchas those that occur in soil(s), and will avoid potential double count-ing between LCI and LCIA.

Regarding time scale, this cannot be seen separate from themodeled processes in space. It does not depend on the pesticideused or environmental compartment considered, but only on thephenomena taking place, as presented in Fig. 2. Therefore, it is rec-ommended that durations are adapted to the Technosphere andthe time chemicals stay here. E.g., drift and volatilization withinthe Technosphere will take up to several minutes, while local shortterm water and soil transfer processes can take up to a few daysafter application. This allows LCA practitioners the possibility totake into account the probability of local events such as rain occur-ring soon after pesticide application inducing a potential effect onleaching and runoff. LCI modeling stops when the chemical enterscrops, agricultural topsoil, freshwater, and air outside the Techno-sphere, respectively. Subsequently, the LCIA models properly ac-count for fate issues that occur in the Ecosphere. The LCIA willthen consist of a plant compartment as well as soil, groundwater,freshwater, marine water, and air on regional and larger scales(see Fig. 2).

4. Discussion and conclusion

This paper presents a framework to ‘bridge the gap’, i.e. set theboundaries between LCI and LCIA modeling in toxicological assess-ments. This general framework is meant to help agricultural LCApractitioners to better define the needed models, data, and toolsfor a better assessment of pesticides in agriculture. This sectiondiscusses the perspective for improving the LCI(A) models and sug-gests areas requiring further research.

Our framework includes all processes occurring in the environ-ment and captures them consistently in either LCI or LCIA. To build

this framework, no model was chosen as a basis, and, unfortu-nately, current models cannot be applied as such. PestLCI, forexample, the currently most advanced LCI model, does not includeemissions to soil, while ILCD recommended LCIA model USEtoxcannot handle groundwater emissions. To be able to apply our pro-posed framework in a modeling context, PestLCI can be used with-out the groundwater modeling and a groundwater compartmentneeds to be included in USEtox. Furthermore, a vegetation model,such as dynamiCROP needs to be applied in addition to USEtoxto fully cover all environmental compartments.

We recommend to look for experimental data in LCI as far aspossible, before coupling it to an inventory model. PestLCI forexample, assumes that the farmer follows recommended use of apesticide (Birkved and Hauschild, 2006). A case study incorporat-ing known insufficient management practices, or the effects offorthcoming practices and technologies (reducing emissions) cantherefore not be handled by LCA practitioners using this modelonly.

Coverage of the complete framework scheme in a full LCA is notyet possible. Methods to address effects to terrestrial and marineecosystems are preliminary and no single method is currently rec-ommended by ILCD to address these ecological groups (Hauschildet al., 2011). Damage to agricultural soil as well as to other soilshould be included in the terrestrial ecosystem category, as soilspecies belong to the Ecosphere. We recommended to include onlyagricultural topsoil in the Technosphere, because we consider thesoil as part of the Ecosphere since it contains a lot of organic matterand microbial biomass. Currently, ILCD recommended model USE-tox only includes freshwater toxicity.

Additionally, several issues are currently not included in ourframework as they are very data intensive and uncertain. One ofthem is related to adjuvants and fillers that can add a tenfold effi-ciency but also toxicity of pesticides by, for instance, improving thepenetration in vegetation. A second issue is the disposal or reuse ofcrop residues. When performing a consequential LCA, the secondlife of crop residues needs to be included as well.

Several structural uncertainties remain in the fate and effectmodeling of pesticides as is currently done in LCI(A) models. Start-ing from plants and cold-blooded organisms, chemicals can beaccumulated along the food chain. Golsteijn et al. (2012) show thatthe influence on cold-blooded species is different than theinfluence on warm-blooded species. Up to now, only damage tocold-blooded species has been included in LCIA, but to obtain abest estimate on ecosystem damage, warm-blooded organismsneed to be included as well. Moreover, pesticide degradation itselfis assessed, but generally the formation of pesticide transformationproducts and metabolites are ignored in the LCIA pathway,although they can be more toxic than the primary active substance.For instance, the characterization factor for freshwater toxicity ofglyphosate can be 10 times larger when including its transforma-tion products in the environment (Van Zelm et al., 2010). It shouldalso be noticed that the adjunction of nano-particles in pesticideformulation is more and more common. That may increase theirefficiency, but often also their toxicity and current knowledge isvery poor on that topic.

Based on our framework, the key forthcoming issues include (i)the need for a classification of ‘‘pesticide application’’ systems (i.e.combining crop types, vegetative stages and shape with availabletechnologies and practices) for different worldwide conditions,(ii) the need to review all available models or data that can be use-ful to conduct the LCI for all identified typologies (as defined in (i))and to adapt it to build a suitable LCI model, and (iii) to build anLCIA model (i.e. combine and adapt existing LCIA models) that in-cludes all relevant environmental compartments. This will allow,depending on the goal and scope of an LCA, to develop differentLCI approaches from the simplest ones, for global product LCAs,

180 R. van Zelm et al. / Chemosphere 100 (2014) 175–181

to more complex ones for comparing agricultural systems moreaccurately.

The framework provides a starting point for LCA practitioners togather the right data and include all relevant emission and expo-sure routes where possible. By applying our framework for toxico-logical assessments of pesticides, LCI and LCIA are directly linkedup to each other and overlap in LCA steps is avoided.

Acknowledgements

PLL and PR are members of the ELSA research group and theythank Claudine Basset-Mens, Mitchell Burns and Carole Sinfortfrom this group for their relevant advice. (Environmental Life Cycle& Sustainability Assessment, www.elsa-lca.org). In addition, wethank Mitchell Burns for English proofreading and Mark Huijbregtsfor reviewing an earlier version of the manuscript.

Appendix A. Supplementary material

Supplementary data associated with this article can befound, in the online version, at http://dx.doi.org/10.1016/j.chemosphere.2013.11.037.

References

Ademe, 2010. Analyses de Cycle de Vie appliquées aux biocarburants de premièregénération consommés en France. French Environment and EnergyManagement Agency (Ademe), DEPD (Direction Production et EnergiesDurables), p. 236 (in French).

Audsley, E., Alber, S., Clift, R., Cowell, S., Crettaz, P., Gaillard, G., Hausheer, J., Jolliet,O., Kleijn, R., Mortensen, B., Pearce, D., Roger, E., Teulon, H., Weidema, B., VanZeijts, H., 2003. Harmonisation of Environmental Life Cycle Assessment forAgriculture. Final Report Concerted Action AIR3-CT94-2028. EuropeanCommission, Silsoe, United Kingdom, p. 70.

Basset-Mens, C., Van der Werf, H.M.G., 2005. Scenario-based environmentalassessment of farming systems: the case of pig production in France. Agric.Ecosyst. Environ. 105, 127–144.

Bedos, C., Rousseau-Djabri, M.F., Loubet, B., Durand, B., Flura, D., Briand, O., Barriuso,E., 2010. Fungicide volatilization measurements: inverse modeling, role ofvapor pressure, and state of foliar residue. Environ. Sci. Technol. 44, 2522–2528.

Benachour, N., Seralini, G.E., 2009. Glyphosate formulations induce apoptosis andnecrosis in human umbilical, embryonic, and placental cells. Chem. Res. Toxicol.22, 97–105.

Bentrup, F., Küsters, J., Kuhlmann, H., Lammel, J., 2004. Environmental impactassessment of agricultural production systems using the life cycle assessmentmethodology I. Theoretical concept of a LCA method tailored to cropproduction. Eur. J. Agron. 20, 247–264.

Birkved, M., Hauschild, M.Z., 2006. PestLCI – a model for estimating field emissionsof pesticides in agricultural LCA. Ecol. Model. 198, 433–451.

Blonk, H., Alvarado, C., De Schryver, A., 2007. Milieuanalyse vleesproducten. PRéConsultants B.V. & Blonk Milieu Advies, Amersfoort and Gouda, TheNetherlands, p. 37 (in Dutch).

De Snoo, G.R., De Wit, P.J., 1998. Buffer zones for reducing pesticide drift to ditchesand risks to aquatic organisms. Ecotoxicol. Environ. Saf. 41, 112–118.

De Wilde, T., 2009. Sorption and Degradation of Pesticides in BiopurificationSystems. Ghent University, Ghent, Belgium.

Dijkman, T.J., Birkved, M., Hauschild, M.Z., 2012. PestLCI 2.0: a second generationmodel for estimating emissions of pesticides from arable land in LCA. Int. J. LCA17, 973–986.

EMEP/EEA, 2009. Air Pollutant Emission Inventory Guidebook. Technical Guidanceto Prepare National Emission Inventories. European Environment Agency,Copenhagen, Denmark.

Fantke, P., Charles, R., de Alencastro, L.F., Friedrich, R., Jolliet, O., 2011a. Plant uptakeof pesticides and human health: dynamic modeling of residues in wheat andingestion intake. Chemosphere 85, 1639–1647.

Fantke, P., Juraske, R., Antón, A., Friedrich, R., Jolliet, O., 2011b. Dynamic multicropmodel to characterize impacts of pesticides in food. Environ. Sci. Technol. 45,8842–8849.

Fantke, P., Friedrich, R., Jolliet, O., 2012. Health impact and damage cost assessmentof pesticides in Europe. Environ. Int. 49, 9–17.

Frischknecht, R., Jungbluth, N., Althaus, H.-J., Doka, G., Heck, T., Hellweg, S., Hischier,R., Nemecek, T., Rebitzer, G., Spielmann, M., Wernet, G., 2007. Overview andMethodology. Ecoinvent Report No. 1. Swiss Centre for Life Cycle Inventories,Dübendorf, Switzerland.

Ganzelmeier, H., 2000. Drift Studies and Drift Reducing Sprayers – A GermanApproach. 2000 ASAE Annual International Meeting, Technical Papers:Engineering Solutions for a New Century, pp. 1809–1830.

Gil, Y., Sinfort, C., 2005. Emission of pesticides to the air during sprayer application:a bibliographic review. Atmos. Environ. 39, 5183–5193.

Gil, Y., Sinfort, C., Guillaume, S., Brunet, Y., Palagos, B., 2008. Influence ofmicrometeorological factors on pesticide loss to the air during vine spraying:data analysis with statistical and fuzzy inference models. Biosyst. Eng. 100,184–197.

Golsteijn, L., Van Zelm, R., Veltman, K., Musters, G., Hendriks, A.J., Huijbregts, M.A.J.,2012. Including ecotoxic impacts on warm-blooded predators in life cycleimpact assessment. Integr. Environ. Assess Manage. 8, 372–378.

Hauschild, M.Z., Goedkoop, M., Guinée, J., Heijungs, R., Huijbregts, M., Jolliet, O.,Margni, M., De Schryver, A., Humbert, S., Laurent, A., Sala, S., Pant, R., 2011.Identifying best existing practice for characterization modeling in life cycleimpact assessment. Int. J. LCA 18, 683–697.

Henderson, A.D., Hauschild, M.Z., Van de Meent, D., Huijbregts, M.A.J., Larsen,H.F., Margni, M., McKone, T.E., Payet, J., Rosenbaum, R.K., Jolliet, O., 2011.USEtox fate and ecotoxicity factors for comparative assessment of toxicemissions in life cycle analysis: sensitivity to key chemical properties. Int. J.LCA 16, 701–709.

Huijbregts, M.A.J., Thissen, U., Guinée, J.B., Jager, T., Kalf, D., Van de Meent, D., Ragas,A.M.J., Wegener Sleeswijk, A., Reijnders, L., 2000. Priority assessment of toxicsubstances in life cycle assessment. Part I: Calculation of toxicity potentials for181 substances with the nested multi-media fate, exposure and effects modelUSES–LCA. Chemosphere 41, 541–573.

Juraske, R., Mutel, C.L., Stoessel, F., Hellweg, S., 2009. Life cycle human toxicityassessment of pesticides: comparing fruit and vegetable diets in Switzerlandand the United States. Chemosphere 77, 939–945.

Kah, M., Brown, C., 2007. Prediction of the adsorption of ionizable pesticides in soils.J. Agric. Food Chem. 55, 2312–2322.

Lee, H.L., Kan, C.D., Tsai, C.L., Liou, M.J., Guo, H.R., 2009. Comparative effects of theformulation of glyphosate-surfactant herbicides on hemodynamics in swine.Clin. Toxicol. 47, 651–658.

Leistra, M., Van der Linden, A.M.A., Boesten, J.J.T.I., Tiktak, A., Van den Berg, F., 2001.PEARL Model for Pesticide Behaviour and Emissions in Soil–Plant Systems;Descriptions of the Processes in FOCUS PEARL v 1.1.1, Alterra-Rapport 013.Alterra Green World Research, Wageningen, The Netherlands.

Lu, C.S., Barr, D.B., Pearson, M.A., Waller, L.A., 2008. Dietary intake and itscontribution to longitudinal organophosphorus pesticide exposure in urban/suburban children. Environ. Health Perspect. 116, 537–542.

Manneh, R., Margni, M., Deschenes, L., 2010. Spatial variability of intake fractions forCanadian emission scenarios: a comparison between three resolution scales.Environ. Sci. Technol. 44, 4217–4224.

Margni, M., Rossier, D., Crettaz, P., Jolliet, O., 2002. Life cycle impact assessment ofpesticides on human health and ecosystems. Agric. Ecosyst. Environ. 93, 379–392.

Mouron, P., Nemecek, T., Scholz, R.W., Weber, O., 2006. Management influence onenvironmental impacts in an apple production system on Swiss fruit farms:combining life cycle assessment with statistical risk assessment. Agric. Ecosyst.Environ. 114, 311–322.

Nemecek, T., Kägi, T., 2007. Life Cycle Inventories of Agricultural ProductionSystems. Data v2.0. Agroscope Reckenholz-Tänikon Research Station ART,Zürich and Dübendorf, Switzerland.

Panichelli, L., Dauriat, A., Gnansounou, E., 2009. Life cycle assessment of soybean-based biodiesel in Argentina for export. Int. J. LCA 14, 144–159.

Pennington, D.W., Margni, M., Amman, C., Jolliet, O., 2005. Multi-media fate andhuman intake modeling: spatial versus nonspatial insights for chemicalemissions in Western Europe. Environ. Sci. Technol. 39, 1119–1128.

Ramwell, C.T., Leak, J., Cooper, S.E., Taylor, W.A., 2007. The potential environmentimpact of pesticides removed from sprayers during cleaning. Pest Manage. Sci.63, 1146–1152.

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., Van de Meent, D., Hauschild, M.Z., 2008. USEtox—theUNEP–SETAC toxicity model: recommended characterisation factors for humantoxicity and freshwater ecotoxicity in Life Cycle Impact Assessment. Int. J. LCA13, 532–546.

Rosenbaum, R.K., Huijbregts, M.A.J., Henderson, A.D., Margni, M., McKone, T.E., Vande Meent, D., Hauschild, M.Z., Shaked, S., Li, D.S., Gold, L.S., Jolliet, O., 2011.USEtox human exposure and toxicity factors for comparative assessment oftoxic emissions in life cycle analysis: sensitivity to key chemical properties. Int.J. LCA 16, 710–727.

Schmidt, J.H., 2008. Life Cycle Assessment of Palm Oil at United Plantations Berhad.Aalborg University, United Plantations Berhad, Perak Darul Ridzuan, Malaysia.

Torrellas, M., Anton, A., Lopez, J.C., Baeza, E.J., Parra, J.P., Munoz, P., Montero, J.I.,2012. LCA of a tomato crop in a multi-tunnel greenhouse in Almeria. Int. J. LCA17, 863–875.

Van Zelm, R., Huijbregts, M.A.J., Van de Meent, D., 2009. USES–LCA 2.0: a globalnested multi-media fate, exposure and effects model. Int. J. LCA 14, 282–284.

Van Zelm, R., Huijbregts, M.A.J., Van de Meent, D., 2010. Transformation products inthe life cycle impact assessment of pesticides. Environ. Sci. Technol. 44, 1004–1009.

Van Zelm, R., Stam, G., Huijbregts, M.A.J., Van de Meent, D., 2013. Making fate andexposure models for freshwater ecotoxicity in life cycle assessment suitable fororganic acids and bases. Chemosphere 90, 312–317.

Wang, M., Rautmann, D., 2008. A simple probabilistic estimation of spray drift –factors determining spray drift and development of a model. Environ. Toxicol.Chem. 27, 2617–2626.

R. van Zelm et al. / Chemosphere 100 (2014) 175–181 181

Wegener Sleeswijk, A., 2006. GLOBOX – a spatially differentiated multimedia fateand exposure model. Int. J. LCA 11, 141.

Weidema, B.P., Wesnæs, M., Hermansen, J., Kristensen, T., Halberg, N., 2008.Environmental improvement potentials of meat and dairy products. In: Eder, P.,Delgrado, L. (Eds.), JRC/IPTS. Seville, Spain.

Wolters, A., Leistra, M., Linnemann, V., Klein, M., Schaffer, A., Vereecken, H., 2004.Pesticide volatilization from plants: improvement of the PEC model PELMObased on a boundary-layer concept. Environ. Sci. Technol. 38, 2885–2893.