Life-Cycle Assessment in Pesticide Product Development: Methods

This procedure provides inventory data for a best and worst case scenario of production .... If low environmental efficiency is assumed for supply cha...
0 downloads 0 Views 582KB Size
Download PDF
Environ. Sci. Technol. 2005, 39, 2406-2413

Life-Cycle Assessment in Pesticide Product Development: Methods and Case Study on Two Plant-Growth Regulators from Different Product Generations G E O R G G E I S L E R , † S T E F A N I E H E L L W E G , * ,† THOMAS B. HOFSTETTER,‡ AND KONRAD HUNGERBUEHLER† Swiss Federal Institute of Technology (ETH), Safety and Environmental Technology Group, Ho¨nggerberg, CH-8093 Zu ¨ rich, Switzerland, and Department of Water Resources and Drinking Water, Swiss Federal Institute for Environmental Science and Technology (EAWAG), CH-8600 Du ¨ bendorf, Switzerland

Environmental assessments in pesticide product development are generally restricted to plant uptake and emissions of active ingredients. Life-cycle assessment (LCA) enables a more comprehensive evaluation by additionally assessing the impacts of pesticide production and application (e.g. tractor operations). The use of LCA in the product development of pesticides, in addition to the methods commonly applied, is therefore advisable. In this paper a procedure for conducting LCA in early phases of product development is proposed. In a case study, two plant-growth regulators from different product generations were compared regarding their application in intensive production of winter wheat. The results showed that the reduced emissions from active ingredients of the newer pesticide were compensated by higher impacts from the production process. The authors draw the conclusion that it is important to consider environmental objectives in the procurement of precursors, in addition to the classical goals of increasing the efficacy and reducing the nontarget effects of pesticides. Moreover, the case study showed that decisions based on uncertain results in early stages of product development may need to be revised in later stages, e.g. based on investigations of pesticides’ effects on crop yield.

Introduction Producers of pesticides strive to proceed toward environmental sustainability (1-3). To this end and to comply with legal requirements, producers apply environmental assessment methods in decision-making. However, such methods generally focus on isolated phases of the life cycle of pesticides. For instance, Risk assessment (RA) only evaluates pesticide use, because the goal of risk assessment is to ensure the safe use of chemicals (4). Such isolated analyses may however lead to suboptimizations of the overall system. * Corresponding author phone: +41-1-6334337; fax: +41-16321189; e-mail: [email protected]. † Swiss Federal Institute of Technology, Safety and Environmental Technology Group. ‡ Swiss Federal Institute for Environmental Science and Technology (EAWAG). 2406

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 7, 2005

Therefore, a comprehensive method is needed, in addition to conventional risk assessment, to evaluate all relevant environmental impacts associated with the service provided by pesticides. Life-cycle assessment (LCA) (5, 6) offers such a comprehensive scope. The use of LCA in early stages of product development is desirable, because an application in early stages offers the greatest leverage for the reduction of environmental impacts of new products (7). Moreover, LCA can be used to assess a variety of issues in product development, such as the environmental need of the product as such, the environmental advantages and drawbacks of the replacement of existing pesticides with new ones, or the technical optimization of production processes and product formulation. However, significant gaps in knowledge make an application of LCA in early stages of product development difficult (7, 8). Therefore, one key question is, at what stages in product development is LCA best implemented, considering the tradeoff between data availability and degrees of freedom in decision-making. Figure 1 displays the different stages of pesticide product development and the potential use of LCA. Different objectives are applied in each stage in order to ensure a safe use and high profitability. Knowledge on efficacy, substance properties, and toxicity of active substances and products increases continuously during product development, due to increasingly more realistic and comprehensive testing methods. Likewise, the quality of information on production processes and supply chains improves. In the high-throughput screening, the structure of the active substances is often not even known. The lack of knowledge and data impedes the use of LCA in this stage. During the evaluation stage, primary data on substance properties, synthesis routes, and applications of active substances are obtained through laboratory/field tests and pilot studies. Data gaps may be closed using extrapolation procedures and default values. Variability is, however, high; e.g. which synthesis route for an active substance will ultimately be implemented may be unknown in early stages (9). The most important decision during pesticide development concerns the passing of an active substance from the evaluation to the development stage, as substantial monetary investments are dependent on this decision. Therefore, the application of LCA in making this decision would have the greatest leverage for obtaining products with inherently low environmental impacts. In later stages, degrees of freedom decrease sharply, as only one active substance is usually developed for a given application. Still, choices between various formulations or production processes for one active substance may be supported by LCA. LCA may further be useful in communicating progress in environmental objectives to authorities, farmers, or consumers. LCA methods have been developed for use in the product development of chemicals (7) and consumer goods (8). However, these methods are not applicable to pesticides. The use of LCA in pesticide product development thus requires an adaptation of the methodology to a number of specific needs: First, many chemical production steps are necessary for the manufacture of active substances and formulation ingredients. The gathering of life-cycle inventory (LCI) data for these processes is impeded by considerable data gaps that can, however, be subverted by systematic estimation procedures (10). Second, pesticide use is always part of a farming system in which farmers use several agricultural inputs (fertilizers, machine work, pesticides) to achieve the full yield potential of a crop under the given environmental conditions (11). The use of agricultural inputs 10.1021/es049145m CCC: $30.25

 2005 American Chemical Society Published on Web 03/04/2005

FIGURE 1. Objectives and activities in different stages of the economic lifespan of a pesticide, and possible applications of LCA. A substance in the development stage has a 95% probability of reaching the sales stage. therefore depends on such environmental factors, as well as the farming system and the crop variety chosen (11, 12). Appropriate functional units (5) need to be defined in order to include these important interdependencies in LCA. Finally, uncertainty in LCA is generally high, particularly concerning toxic effects (13), which are important for an assessment of pesticide use. Therefore, the significance of results should be checked with methods for uncertainty analysis (13, 14). The goal of this paper is to propose a method for the LCA of pesticides that is applicable in early stages of product development. To this end, we evaluate the method’s applicability during different stages of the economic lifespan of pesticides. The method is applied to a case study comparing two pesticides with similar functions from two product generations. LCA-based objectives are derived from this case study to help design pesticides with inherently lower environmental impacts. These objectives are finally compared to environmental goals commonly taken into consideration during the development of pesticides.

Methods Functional Units. During the registration process for pesticides, a recommended dose is defined, the application of which is to guarantee the desired effect of the product (4). Pesticides with similar functions may therefore be compared using their recommended doses for a specific crop application as a functional unit. We calculated quotients (Q) of impact scores for this comparison (14)

Q ) Ic,A/Ic,B

(1)

where I is an environmental impact potential (unit of the impact category c). Indices A and B designate products compared. Using the recommended dose as the functional unit might not always include all aspects of pesticide efficacy, as the original purpose of recommended doses is determined

FIGURE 2. Scheme of agricultural inputs in the base-case and the efficiency-increase scenario. Input In(m) depends on the output mass of agricultural produce and input In(A) on the area under cultivation. independent of any consistent comparison among pesticides but is rather determined on the basis of ensuring efficacy and safe application of individual products (4). Effects on crop yield may differ considerably between pesticides if recommended doses are applied. Such differences should be taken into account in a product comparison: If a pesticide increases crop yield, impacts otherwise necessary to produce this extra yield should be credited to the impact score of the pesticide. Hence, the yield increase necessary to reach the break-even point is of particular interest (15), i.e., the point where impacts due to the production and application of a pesticide (the environmental investment) are equal to the impacts avoided due to the yield increase achieved by using this product (the environmental revenue). The calculation of environmental break-even points is based on the comparison of two scenarios (base case and efficiency increase) of a farming system for a given crop (Figure 2). In this case, the functional unit is the output of a defined mass of agricultural produce (mfu, eq S1, Supporting Information). The pesticide under study is used only in the efficiency-increase scenario, while all other agricultural inputs VOL. 39, NO. 7, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2407

FIGURE 3. Step-by-step assessment of pesticides. Explicit schemes in shaded boxes apply in analogy to the shaded boxes carrying only the corresponding title. are needed in both scenarios. These other agricultural inputs are assumed to be a linear function of the mass of agricultural produce (In(m) in Figure 2) or the area under cultivation (In(A)). Agricultural inputs that depend on the output mass are equal in both scenarios, as the output mass of agricultural produce is the same for both scenarios (mfu). Agricultural inputs that depend on the area under cultivation are higher in the base case than in the efficiency-increase scenario. This is due to the smaller yield per area in the base-case scenario relative to the efficiency-increase scenario. The break-even point is depicted here as a function of the relative yield increase (for detailed derivation, see the Supporting Information)

∆mrel,bep(p) ) Ippp(p)/

∑I

i,incr(A)

(2)

i

where ∆mrel,bep is the relative yield increase (dimensionless) at the break-even point, Ippp is a percentile of the impact score of the environmental investment (impact from pesticide production and application in the efficiency increase scenario), p is the probability of this percentile (dimensionless), and Ii,incr(A) is the impact score for area-dependent agricultural inputs i in the efficiency increase scenario. Impact scores carry the unit of the respective impact category. Percentiles of the impact score of the environmental investment are derived by an uncertainty assessment (14). 2408

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 7, 2005

Relative yield increases necessary for reaching break-even points (eq 2) are compared to ranges of possible yield increases found in field studies. Break-even points are considered to be reached under average agricultural conditions, if the relative yield increase at the break-even points is lower than the median of possible yield increases. The robustness factor (RF) quantifies this relation

RF ) median(∆mrel)/∆mrel,bep(0.95)

(3)

where RF is the robustness factor (dimensionless), median(∆mrel) is the median of the possible yield increases (dimensionless) determined in field studies, and ∆mrel,bep(0.95) the 95th percentile of the relative yield increase at break-even point (eq 2). If the robustness factor is larger than one, it can be assumed that the break-even point is reached under average agricultural conditions. Application of the pesticide is then considered environmentally preferable to not using it. In addition to this application of the robustness factor, RF can also be applied in the comparison of different pesticides used in the same crop and farming system. Calculation of Pesticide LCAs. In Figure 3, a method is proposed for calculating LCAs in pesticide product development (Figure 1). The assessment is carried out separately for each field of use of the pesticides considered. Prior to the assessment, comparable products are determined. These may

TABLE 1. Characteristics of Trinexapac-Ethyl and Chlorocholine Chloride, the Two Active Substances in the Growth Regulator Products Used as Case Study

TABLE 2. Application Schemes for Moddus and Stuntan in Intensive Cereal Management of Winter Wheat (19) as Field of Use and Corresponding Break-Even Point Scenarios (Figure 2) application scheme number

efficiency increase scenario

base-case scenario

1 2

combined application of Moddus and Stuntan application of Stuntan alone

no use of growth regulators no use of growth regulators

be products under development or already established benchmark products. In case comparable products exist, a comparison is initially carried out using the recommended dose as the functional unit. Data acquisition to this end comprises LCIs for the production of active substances and formulation ingredients as well as substance data for the life-cycle impact assessment (LCIA) of these compounds. Some of the data are routinely generated by pesticide producers (Figure 1). Public LCI and substance property databases are preferable for filling data gaps. Otherwise, estimation routines are applied. To estimate LCIs of chemical production, a procedure for the systematic estimation of LCIs for fine and speciality chemical production processes is used (10). This procedure provides inventory data for a best and worst case scenario of production efficiency in chemical production processes, thus depicting important sources of uncertainty for such LCIs. In the best case scenario, high reaction yields, low fresh solvent and utility demands, and highly efficient emission abatement are assumed. Concerning the impact assessment, missing substance data for the calculation of characterization factors of active substances are estimated by applying the Epiwin (16) suite of QSARs. Once the data acquisition is completed, the LCA based on the recommended dose is calculated for different scopes. The first scope (a) comprises only the environmental fate and effects of active substances. Leaching to groundwater was assessed for average European conditions with the method of Geisler et al. (17). As input data for substance properties, we used the values in Table S1 and assumed a generic value of 80% for the coefficient of variation of KOC and DT50,soil, as recommended in ref 17. All other pesticide emissions from the field were calculated with the method of Hauschild (18). This scope allows examining whether impact pathways of relatively high concern are the same in LCIA and RA. If this is not the case, reasons should be found to explain divergent results from these two methods to decision makers. Next, products are compared by applying a lifecycle scope (b), using the recommended dose as functional unit (see Functional Units). After that, environmental breakeven points are calculated, provided that data on crop-yield effects are available. Uncertainty of the results is assessed and evaluated according to Geisler et al. (14): If results show insignificant

differences between alternative products, uncertainty should be reduced, if possible, for instance, by acquiring more precise values for LCI data. By contrast, uncertainty attributable to LCIA methods cannot be reduced during routine use of LCA (e.g. uncertainty of characterization factors). Finally, if no comparable product is available, but yield data are known, break-even points (eq 2) and the corresponding robustness factors (eq 3) provide the important information for determining whether using a pesticide is environmentally preferable to not using it. Case Study. Two growth regulators are compared using the procedure in Figure 3. The two products stem from different generations. Moddus contains the relatively new active substance trinexapac-ethyl, while Stuntan (a fictive name representing a number of similar products) contains chlorocholine chloride, which has been used for a long time. We assess the use of these products for growth regulation in intensive cereal management (12) of winter wheat. In such a farming system, both products were found to reduce stem height and increase stem thickness of wheat plants, thus preventing lodging, which is the bending of stems during storms (19). Moddus was additionally found to increase yield by affecting plant hormones (19). The step 1 product comparison is based on recommended doses (Figure 3). However, in agricultural practice Moddus and Stuntan are usually used in combination (19). Fieldstudy data were available on the effects of the combined application of Moddus and Stuntan on wheat yield, as well as for the application of Stuntan alone (19). Accordingly, two application schemes were assessed for their environmental break-even points (Table 2). The system boundaries included the production of the plant-growth regulator (active ingredient and formulation ingredients), transport processes, tractor operations, and the impacts from the emissions of the product on the agricultural field (Figure S1 and Section 2 of the Supporting Information). In addition, impacts from those agricultural inputs that changed as a function of the plantgrowth regulator applied were considered (Table S8). For a detailed description of case-study LCIs and data used to calculate the environmental revenue, see the Supporting Information. Concerning LCIA, relevant impact categories of the CML-baseline method (6) are used, as is the primary energy demand (22). VOL. 39, NO. 7, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2409

FIGURE 4. Characterized impact scores (CML-baseline method (6)) of active substance emissions from the field for the two growth regulators and exposure pathways (functional unit: 1 ha treated with the recommended dose; step 1a in Figure 3).

FIGURE 5. Percentiles of the quotient of impact scores for the comparison of Moddus (index A in eq 1) and Stuntan (index B) based on the recommended dose as functional unit (steps 1b and 3 in Figure 3). Distributions obtained in the best (left) and the worst case scenario (right) of chemical production efficiency (10) are juxtaposed. Bold, italic figures: Probability (%) of the quotient of impact scores to be larger than 1, with asterisks designating significant differences between the pesticides. Uncertainty was modeled according to Geisler et al. (14).

Results Concerning the isolated assessment of active ingredient emissions (step 1a in Figure 3), trinexapac-ethyl exhibits lower human and ecotoxicity impact-scores than chlorocholine chloride (Figure 4). This can be accounted for by the lower dose (Table 1), shorter environmental half-lives, and higher no-effect concentrations of trinexapac-ethyl as compared to chlorocholine chloride (Supporting Information). The main exposure pathway is via air, concerning toxicity to humans and ecosystems. In contrast to trinexapac-ethyl, some chlorocholine chloride leaches to the groundwater (mean leached fraction, 0.15% of applied dose, calculated according to ref 17) (Figure 4), because it has a higher soil half-life and a lower organic carbon-water partitioning-coefficient than trinexapac-ethyl (Table S1). 2410

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 7, 2005

Next, quotients of impact scores (eq 1) for the full life cycles of the growth regulators are compared in Figure 5 (step 1b, Figure 3). Results are shown for two scenarios, depicting uncertainty in the environmental efficiency of chemical production processes (10). Parameter uncertainty is propagated into the quotient of impact scores (eq 1) according to Geisler et al. (14). In most instances, no significant difference is observed between the two products. If low environmental efficiency is assumed for supply chains of active substances and formulation ingredients (worst case scenario), the impact score is lower for Stuntan than for Moddus regarding acidification, photooxidant creation, and human toxicity. Only with regard to freshwater ecotoxicity in the best case scenario are the impact scores for Stuntan higher than for Moddus, with a probability of 71% (Figure

TABLE 3. Evaluation of Break-Even Points for Two Application Schemes (Table 2) of Moddus and Stuntan (steps 2 and 3 in Figure 3)e Combined Application of Moddus and Stuntan

primary energy demand, MJ

global warming potential, kg CO2 equiv

median 95th percentile 5th percentile

280 340 230

19 22 16

median 95th percentile 5th percentile

2.4 4.0 1.6 2.5

Environmental Investment photooxidant human toxicity creation potential, potential, kg ethene equiv kg 1,4-DCB equiv 5.3 × 10-3 1.3 × 10-2 3.0 × 10-3

2.5 10 0.80

freshwater ecotoxicity potential, kg 1,4-DCB equiv 0.70 2.3 0.27

Relative Yield Increasea Needed To Reach Break-Even Point (wt %) 2.8 1.9 1.2 0.082b 4.4 15c 46c 2.7 2.0 0.69b 0.12b 0.010b 2.3

Robustness Factor (dimensionless) 0.67 0.22

3.7

terrestrial ecotoxicity potential, kg 1,4-DCB equiv 5.3 × 10-4 2.0 × 10-3 1.6 × 10-4 1.2 × 10-3 b 5.1 × 10-2 b 8.9 × 10-5 b 200

Application of Stuntan Alone

primary energy demand, MJ

global warming potential, kg CO2 equiv

median 95th percentile 5th percentile

140 170 110

9.1 11 7.8

median 95th percentile 5th percentile

1.2 2.0 0.79 2.0

Environmental Investment photooxidant human toxicity creation potential, potential, kg ethene equiv kg 1,4-DCB equiv 2.9 × 10-3 8.1 × 10-3 1.5 × 10-3

1.2 5.2 0.39

freshwater ecotoxicity potential, kg 1,4-DCB equiv 0.54 1.6 0.21

Relative Yield Increased Needed To Reach Break-Even Point (wt %) 1.3 1.0 0.61 0.063 2.1 10c 23c 1.3 0.98 0.31 0.059 7.6 × 10-3 1.9

Robustness Factor (dimensionless) 0.40 0.17

3.1

terrestrial ecotoxicity potential, kg 1,4-DCB equiv 2.2 × 10-4 9.0 × 10-4 6.4 × 10-5 5.0 × 10-4 0.024 3.2 × 10-5 170

a

Possible yield increases in field studies range from 1 to 14 wt % (Supporting Information). b Yield increase at break-even point is smaller than the minimum yield increase in the field studies. c Yield increase at break-even point is larger than the maximum yield increase in the field studies. d Possible yield increases in field studies range from 0 to 7 wt % (Supporting Information) e Impact score distributions of the environmental investment are displayed, as well as relative yield increases necessary to reach break-even points (eq 2) for selected percentiles of the environmental investment and robustness factors (eq 3).

5). This is attributable to the emissions of chlorocholine chloride (Figure 4). Concerning the other toxicity impact categories, active-substance emissions from the field did not influence quotients of life-cycle impact scores substantially. In the second assessment step (Figure 3), whether breakeven points are reached in farming practice, is examined. To this end, we regarded realistic growth-regulator application schemes (Table 2) and used field data on yield effects of these schemes in intensive cereal production (Supporting Information). The results of Figure 5 represent the environmental investment (Ippp in eq 2) for each of the growth regulators. Only the best case scenario of chemical production efficiency is used any longer, as this scenario was found to be more realistic than the worst case scenario concerning Western European standards in the chemical industry (10). With respect to primary energy, global warming, and ecotoxicity impact scores, the robustness factors (eq 3) are larger than one for both application schemes (Tables 2 and 3), indicating that the break-even point is reached in farming practice. The combined use of the growth regulators as well as the use of Stuntan alone is therefore environmentally preferable to not applying growth regulators at all, with respect to these impacts. The robustness factors of Stuntan are always lower than those of the combined application, as yield increases are smaller for the sole application of Stuntan (Supporting Information). This indicates that break-even points are reached with a higher probability for the combined application of growth regulators than with Stuntan alone.

Regarding photooxidant creation and human toxicity, relative yield increases at break-even points span the whole range of possible yield increases found in field studies. Due to this considerable uncertainty, it cannot be claimed that the two application patterns of the growth regulators are preferable to no growth regulator use, with respect to photooxidant creation and human toxicity.

Discussion Case Study. The goal of this work was to illustrate the application of LCA in decision-making during pesticide product development (Figure 1). Therefore, we use Moddus as a hypothetical example of a new product in the evaluation stage. If only active substance emissions from the field are considered (step 1a in Figure 3, Figure 4), the application of trinexapac-ethyl is environmentally preferable to using chlorocholine chloride. These results agree with the common findings of risk assessment, according to which reduced active substance doses and environmental half-lives lead to a lower risk of impacts. However, considering a life-cycle scope (steps 1b and 3 in Figure 3), no significant progress of Moddus is apparent in the product comparison based on recommended dose (Figure 5). This can, in part, be accounted for by the difference in the complexity of molecular structures of the active substances: Trinexapac-ethyl shows a more complex molecular structure than chlorocholine chloride and requires more chemical production steps than Stuntan, leading to higher production impacts. This is apparent in that quotients of impact scores in the worst case scenario of chemical VOL. 39, NO. 7, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2411

TABLE 4. Objectives for an Increased Environmental Efficiency of Pesticides Derived from the Case Study, Importance from the Point of View of LCA, and Current Application in Routine Decision-making during Pesticide Development

objective maximize yield increase minimize applied dose procurement for maximum production efficiency, including commodity precursors and formulation ingredients minimize environmental exposure and effects due to the use of active substances efficiency of chemical production carried out by pesticide producers themselves reduce impacts of application technique a

importance according to LCA results

applied in routine decision-support?

very important very important very important

yes yes no

important

yes

important

yes

important

not explicitlya

Impacts, for example, of tractor operation are not considered explicitly, but reduction of tractor use is an agronomic objective.

production are always higher than those in the best case (Figure 5): If the production efficiency for precursors is low (worst case), the impacts from the production of trinexapacethyl increase more strongly than for chlorocholine chloride. In addition to the active ingredients, impacts from formulation ingredients also contribute considerably to the overall impact of both growth regulators. The comparison based on recommended doses neglects specific yield effects and realistic application schemes of the growth regulators (Table 2). In the Evaluation stage, neither data on the combined application nor on yield effects would have been available for Moddus. Preliminary yield data may have been accessible by expert judgment from field plot studies, but uncertainty of such estimations would have been high. Therefore, any assessment in the evaluation stage would have neglected important aspects in the comparison of Moddus and Stuntan. It is thus advisable to revise the LCA results in each stage of product development when more information becomes available. We are considering, as a second hypothetical example, Moddus as an established product, and we question whether this product should remain in the portfolio of a pesticide producer. One objective may be that the use of Moddus should show reduced environmental impacts as compared with other established products. This is evaluated by considering the break-even points of realistic application schemes of the growth regulators (steps 2 and 3 in Figure 3). The LCA results show that robustness factors are higher for the combined application than for the application of Stuntan alone (Table 3). The use of Moddus in combination with Stuntan is therefore environmentally preferable to relying only on Stuntan. Hence, Moddus should remain in the portfolio of the company, as far as environmental objectives are concerned. In the case study we focused on the use of plant-growth regulators in intensive cereal management, comparing three alternatives, i.e., the application of Stuntan, the joint use of Moddus and Stuntan, and no plant-growth regulator use. Assessing completely different systems of agricultural management, e.g. organic farming, was not in the scope of this work. Such alternative systems may be environmentally superior to the use of plant-growth regulators. Also, the breakeven point depends strongly on the yield increase data determined in field trials, which are specific to the use of growth regulators in intensive agriculture. Pesticide Product Development. Some general recommendations for the reduction of environmental impacts in the life cycle of pesticides may be derived from the case study (Table 4). Some important objectives in reducing LCA 2412

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 7, 2005

impact scores are already routinely applied in decisionmaking during pesticide development. These are the maximization of yield increases and the minimization of applied doses. Due to regulatory requirements, minimizing environmental exposure and effects from the active substances also has a high priority in the development of pesticides. However, in the LCA such impacts may be dominated by impacts from pesticide production and from machine use during application. The little relevance of active substance emissions from the field in our case study is to be expected, providing RA really ensures the safe use of pesticides. Therefore, both assessments, RA and LCA, are important in their own fields, according to differing goals and scopes (23). An additional objective for reduced environmental impacts emerges from LCA, namely the production efficiency of active substances, formulation ingredients, and their precursors (see worst case scenario of production efficiency, Figure 5). Pesticide producers routinely optimize the production efficiency of their processes. By contrast, they rarely analyze the production efficiency along the whole supply chain including precursors and formulation ingredients. The case study showed that considerable reductions in impact scores would be achieved if pesticide producers applied environmental criteria in the procurement of chemicals. The method presented here enables the implementation of LCA as early as the evaluation stage (Figure 1) of pesticide product development. Considerable uncertainty is attached to LCA at this stage. For instance, potential long-term effects of the developed pesticide on human and ecosystem health as well as plant resistance are unknown early in product development and therefore cannot be assessed within LCA. However, decisions still need to be made and should therefore be based on the information made available with methods as the one discussed here. Some types of variability, due to several possible formulations, for example, can be depicted in LCA by analyzing all alternatives. Such analyses would help in finding optimum realizations of a product. Product development of pesticides moves toward increasing complexity of molecular structures, which leads to desired reductions in doses and environmental half-lives of active substances. This progress may however be at a tradeoff with increased environmental impacts from the production of ever more complex molecular structures. Substantial increases in crop yield may again compensate increased production impacts, as shown in the present case study. No general rules can be established in this work to quantify these counteracting trends. Thus, a case-based assessment of different types of pesticides is needed.

Method. A valuable link between agronomy and LCA is offered by expressing avoided impacts due to pesticide use in terms of crop yield, as in the break-even point. However, the assumption of direct linear relations between agricultural inputs and yield or cultivated area (Figure 2) is a simplification. For instance, Brentrup (24) found that nitrate leaching may increase more strongly than at a linear rate with rising nitrate fertilizer doses. In this case, nonlinear shares of nitrate leaching would have to be added to the environmental investment, as well as corresponding fertilizer inputs (eq 2). Further, we considered all agricultural produce to meet the desired quality criteria. In reality, there is a risk of not meeting such criteria (12), the relevance of which still needs to be analyzed for LCA. Also, potential synergism and antagonism of mixtures, such as that of Moddus and Stuntan, have not been taken into consideration in this study. Uncertainty in production data was described by a best case and a worst case scenario. As a consequence of the high uncertainties of such estimated data, significant differences between the two product alternatives were only apparent with respect to few impact categories. It would therefore be desirable to decrease these uncertainties, e.g. by basing the calculations more strongly on measured than on estimated production data. If LCA were fully integrated in the product development of pesticides, information routinely acquired by pesticide producers would certainly improve LCI data quality. However, data on the production of precursors would probably remain inaccessible. To close these gaps, estimation methods for corresponding LCIs such as ref 10 should be improved, e.g. by taking into account statistical analysis of empirical data (25). Such developments are likely to substantially improve data quality and reduce the uncertainty of production data. Finally, methods for modeling emissions of active substances from the field should preferably consider spatial and temporal dependencies of impacts, as well as further exposure pathways, such as food (17, 23). LCA may be used during various stages of product development. However, LCAs carried out in early stages may lead to different conclusions than LCAs based on improved knowledge at later stages. This is a risk associated with any assessment method when used in the early stages, including commonly applied risk assessment and economic evaluations. An overview of typical pitfalls and uncertainties of LCA applied in early stages could be obtained by the routine use of LCA in early as well as later stages of pesticide development. Methods could then be improved for optimum robustness of results obtained with differing levels of knowledge. Until then, LCA results based on the restricted knowledge in the evaluation stage of pesticide development should be regarded with care. By contrast, LCA-based break-even points used to aid decision-making during the development stage reflect environmental strengths and weaknesses of pesticides quite accurately. The concept of break-even points may also be used to assess the environmental performance of agricultural means other than pesticides, such as fertilizers or irrigation.

Acknowledgments We gratefully acknowledge the expertise and data of Syngenta Crop Protection and the helpul comments of three anonymous reviewers.

Supporting Information Available Information about the calculation of environmental breakeven points, inventory data, and the field study on yield increases. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Syngenta: Syngenta Annual Review 2000. www.syngenta.ch. (2) Aventis, 2004. www.aventis.com.

(3) BASF: Sustainability 2004. http://corporate.basf.com/en/ sustainability (4) Directive 91/414/EEC Concerning the Placing of Plant Protection Products on the Market. European Commission: Bruxelles, 1991. (5) ISO 14040: Environmental ManagementsLife Cycle AssessmentsPrinciples and Framework. European Committee for Standardization (CEN): Bruxelles, 1997. (6) Guine´e, J. B.; Gorre´e, M.; Heijungs, R.; Huppes, G.; Kleijn, R.; de Koning, A.; van Oers, L.; Wegener Sleeswijk, A.; Suh, S.; de Haes, H. A. U. CML-Guide to Life Cycle Assessment, Centre of Environmental Studies; Leiden University (CML): Leiden, 2001. (7) Weidenhaupt, A.; Hungerbu ¨ hler, K. Integrated product design in chemical industry. A plea for adequate life-cycle screening indicators. Chimia 1997, 51, 217-221. (8) Hauschild, M.; Wenzel, H. Environmental Assessment of ProductssMethodology, Tools and Case Studies in Product Development; Chapman & Hall: London, 1997; Vol. 1. (9) Heinzle, E.; Weirich, D.; Brogli, D.; Hoffmann, V. H.; Koller, G.; Verduyn, M. A.; Hungerbu ¨ hler, K. Ecological and economic objective functions for screening in integrated development of fine chemical processes. 1. Flexible and expandable framework using indices. Ind. Eng. Chem. Res. 1998, 37, 3395-3407. (10) Geisler, G.; Hofstetter, T. B.; Hungerbu ¨ hler, K. Production of Fine and Speciality Chemicals: Procedure for the Estimation of LCIs. Int. J. LCA 2004, 9, 101-113. (11) Mu ¨ ller, F., Ed. Agrochemicals, 1st ed.; Wiley-VCH: Weinheim, 2000. (12) Regev, U.; Gotsch, N.; Rieder, P. Are fungicides, nitrogen and plant growth regulators risk-reducing? Empirical evidence from Swiss wheat production. J. Agric. Econ. 1997, 48, 167-178. (13) Huijbregts, M. A. J.; Gilijamse, W.; Ragas, A. M. J.; Reijnders, L. Evaluating uncertainty in environmental Life-Cycle Assessment. Environ. Sci. Technol. 2003, 37, 2600-2608. (14) Geisler, G.; Hellweg, S.; Hungerbu ¨ hler, K. Uncertainty Analysis in Life Cycle Assessment (LCA): Case Study on Plant-Protection Products and Implications for Decision Making. Int. J. LCA 2004 (Online First http://dx.doi.org/10.1065/lca2004.09.178), 10 (4), in press. (15) Schaltegger, S.; Burritt, R. Contemporary Environmental Accounting; Greenleaf Publishing: Sheffield, U.K., 2000. (16) EPI suite, 10 ed.; U.S. Environmental Protection Agency: Washington, DC, 2001; v 3. (17) Geisler, G.; Hellweg, S.; Liechti, S.; Hungerbu ¨ hler, K. Variability Assessment of Groundwater Exposure to Pesticides and its Consideration in Life Cycle Assessment (LCA). Environ, Sci. Technol. 2004, 38, 4457-4468. (18) Hauschild, M. In Agricultural data for life cycle assessments; Report no. 2.00.01; Weidema, B. P., Meeusen, M. J. G., Eds.; Agricultural Economics Research Institute (LEI): The Hague, 2000; Vol. 2, pp 64-79. (19) Syngenta, C. P. Moddus der Lagerschutz (Moddus for Protection against Lodging); Basel, 2001. (20) Preisliste 2000, PflanzenschutzsRatgeber (Price List and Crop Protection Handbook); Novartis Crop Protection: Dielsdorf, CH, 2000. (21) Pesticide Residues in Food, Report of the Joint FAO/WHO Meeting of Experts, Food and Agriculture Organisation (FAO); World Health Organisation (WHO) Joint Meeting on Pesticide Residues (JMPR), Rome, 1994. (22) Frischknecht, R. E.; Bollens, U.; Bosshart, S.; Ciot, M.; Ciseri, L.; Doka, G.; Hischier, R.; Martin, A.; Dones, R.; Ganter, U. O ¨ koinventare von Energiesystemen (LCIs of Energy Systems), 3rd ed.; Swiss Federal Office of Energy: Bern, 1996. (23) Hellweg, S.; Geisler, G. Life Cycle Impact Assessment of PesticidessConference Report on the 19th LCA Discussion Forum. Int. J. LCA 2003, 8, 310-312. (24) Brentrup, F. Life Cycle Assessment to Evaluate the Environmental Impact of Arable Crop Production; Universita¨t von Hannover: Hannover, 2003. (25) Capello, C.; Hellweg, S.; Badertscher, B.; Hungerbu ¨ hler, K. Lifecycle Inventory of Waste Solvent Distillation: Statistical Analysis of Empirical Data. Environ. Sci. Technol. Submitted for publication.

Received for review June 7, 2004. Revised manuscript received December 24, 2004. Accepted January 4, 2005. ES049145M VOL. 39, NO. 7, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2413