Comments on this paper (or requests for copies sent as e-mail attachments formatted in MS Word) can be addressed to Dr. Lois Levitan. Questions/comments about the IPM Symposium and other papers presented there can be directed to Dr. Sarah Lynch at USDA/Economic Research Service.
Agriculture intentionally disturbs the natural ecosystem and imposes a managed system that has multiple direct and indirect environmental consequences from those of unmanaged systems. Given the uncertainty and complexity of these consequences, a number of different approaches for assessing the impacts of agricultural practices on the environment have been proposed and discussed. All these methods can be viewed as attempts to answer the question "What are the environmental consequences of agricultural management decisions?" IPM investigators are currently being challenged to respond to this question as part of their research and as one means of assessing the success of IPM. Previously, IPM has been judged primarily in terms of the cost and efficacy of IPM practices. To the extent that environmental impact was considered, it was assessed primarily by reduction in pesticide use or by indicators important to implementing IPM (for example, the impacts on beneficial arthropods).
The objectives of this paper are twofold: first, to encourage IPM investigators to think more deeply about the potentials, limitations and complexities of environmental impact assessment, and second, to acquaint IPM investigators with the range of current approaches they might use to evaluate environmental impacts of their IPM programs. The paper is divided into four sections. The first section discusses the meaning of environmental impact. Our purpose is to inspire researchers to think broadly when considering environmental impacts, and to illustrate some of the consequences of a narrow view of the environment. The second section describes a number of challenges in conducting environmental impact assessments. The point of this section is to encourage researchers to recognize problems with current environmental assessment methods, and use these as a motivation for improving assessment tools. The third section presents a typology of approaches to environmental assessment. We discuss the objectives, strengths and limitations of various assessment methodologies but do not evaluate particular environmental assessment methods. This section is meant to encourage researchers to consider how different types of assessment methods may or may not be suitable for their project. The last section considers some practical issues which researchers face in deciding which assessment method to use. These include determining who the assessment is suppose to serve and trade offs in ease-of-use versus complexity. The aim of this section is to encourage researchers to consider these issues explicitly before choosing an environmental assessment method.
1. Defining Environmental Impacts
1.1. Chemical to Biocriteria
One of the reasons that the classic ecotoxicological model has been widely used is that it is easier to set goals and write regulations related to chemical levels (e.g. in terms of the concentration of pesticide in groundwater) than in terms of impacts on ecosystems. Objections have been raised to the individual-species toxicity tests which are integral to this model. These objections include the limited array of species used may not be most sensitive, the same species is not most sensitive to all chemicals, and species may respond differently when not isolated from other species (Cairns 1995). Micro- and mesoscale testing systems have been developed to overcome some of these objections. The results of these tests have been considered by some too inconsistent to be practicable, though Cairns (1995) believes this approach may have been too easily dismissed. More generally, the classic ecotoxicological model fails when the acceptable level of a chemical in the environment as established from test endpoints does not correlate with the environmental impacts of interest to the public. Another shortcoming of applying the classic ecotoxicology model to assessments of agricultural impact is that people are generally not directly concerned with the level of a chemical in the environment per se, even if this level is lethal to 50% of a specific organism in a test. What is of interest to them is the impact of management decisions on such components of the environment as populations of biota and the functioning of ecosystems (Karr 1995), which are sometimes referred to as assessment endpoints (Suter 1995). We will use the term 'decision endpoints' in referring to these environmental components that are of actual interest to various decision-making groups.
In response to some of the limitations of the classic ecotoxicological model, with its focus on chemical criteria, some scientists suggest using field-measured biological criteria that can be more directly related to decision endpoints (Karr 1995) rather than single species toxicity tests (Fig. 1b). The use of biological criteria as indicators of environmental impact has both a public and scientific tradition. For centuries people have been concerned about fish supplies and, more recently have expressed concern for the preservation of other 'wildlife' (Policansky 1993). There is increasing public and scientific interest in the more general notion of environmental integrity, and a recognition by the scientific community that single-species toxicity is not necessarily indicative of system-level responses (Policansky 1993, Barbour et al. 1995, Cairns 1995). Characterizing environmental integrity generally requires measures of an array of biological attributes. These can include use of habitat indices, conditions of individual organisms (i.e. diseases, anomalies or metabolic processes), community structure measures (i.e. taxa richness and trophic dynamics) and productivity measures. In environmental assessment, this approach has probably been taken furthest in evaluating the integrity of water resources (Barbour et al. 1995).
Although biocriteria are important indicators of environmental impact, their use raises several problems. There is not currently a widely-accepted, multi-dimensional measure of biological integrity/ecosystem quality (Barbour et al. 1995). An index of biotic integrity (IBI) has been developed using biosurvey data to construct a multimetric index of heterogeneous variables (Karr 1981, Simon and Lyons 1995). Criticism of this index approach include ambiguity, eclipsing of one metric by another, arbitrary variance , unreality involved in combining unlike metrics, post hoc justification, single linear scale of response, inability to use in diagnostics, and nonsense results. Simon and Lyons (1995) attempt to defend IBI in the face of these criticisms, but many of Suter's concerns are inherent to such indices and therefore should be taken seriously.
A second problem in the use of biocriteria is in defining appropriate reference conditions, particularly in terrestrial ecosystems (Policansky 1993, Hughes 1995). The problems encountered in defining reference conditions can be easily illustrated by issues in restoration ecology -- to what condition should derelict or degraded land be restored? Both in restoration ecology and in defining an acceptable biological status of an ecosystem, it has been recognized that human values must be taken into consideration. Diamond (1987), in his studies of restoration ecology, points out that different segments of the population hold different values and therefore different views of appropriate restoration conditions. Hughes' (1995) position is that "The [biological] reference condition must be politically palatable and reasonable. In other words, it must be acceptable and understandable by persons most concerned with nature for its own sake and those unconcerned with nature or only concerned with what it can provide humans. If the process for determining the reference condition is acceptable and understandable by only one of these groups, it will not be broadly implemented by the majority of persons who fall between these two extremes."
Another important concern with the use of biocriteria in environmental impact assessment is that it is often difficult to infer from measures of biological integrity the cause of biological impairment. Changes in biological integrity may be due to one or more environmental stresses produced by any number of management decisions. Recently, multimetric approaches have been proposed to develop thresholds using biocriteria that may be useful in identify different types of stresses (Barbour et al. 1995). However, it will likely prove difficult to develop fate or process models that can relate the impact of a particular farm management decision to the biological integrity of nearby streams and lakes. So while the environmental impact assessment model summarized in figure 1b has the advantage of using decision rather than test endpoints, a disadvantage lies in the difficulty of linking specific farm management practices to perturbations in environmental integrity.
The EPA has been providing guidance to the states on the development and use of biological criteria (Southerland and Stibling 1995). Though at first glance biological criteria may appear complicated to implement in IPM assessment programs, IPM researchers and practitioners are already using biological indicators in their research on beneficial organisms and predator-prey relationships as indicators of community structure and trophic dependencies.
1.2. Spatial and Temporal Scales
Spatial and temporal scales are also important to consider when data are transferred between disciplines, when data are used to infer trends, and when data generated at one scale or in a narrowly-defined system are used to interpret studies at a different scale or in a wider system, such as a landscape. Impacts of agriculture are generally experienced at spatial and temporal scales much larger than those at which environmental measurements are made. Processes in the landscape occur over a wide range of scales, but sampling is usually restricted to scales of time and space determined by sampling procedures and the time frame of a research or monitoring project. For example, soil scientists, measure and monitor chemical concentrations at scales ranging from soil profile to field, during experiments which rarely last more than a few years.
How should we approach measurement and monitoring at larger scales? Applying conventional measurement techniques to more sites for longer time periods can provide useful information, but requires excessive effort and is costly. We need to rethink the way in which we approach such broad-scale projects, starting with an assessment of pathways and impacts, tailoring monitoring strategies to the whole system, rather than a few arbitrary points in it. Field monitoring and measurement strategies for broad-scale projects should be carefully planned and evaluated, taking into account both temporal and spatial variability. Techniques for parameter estimation, monitoring and modeling should change as we move from point of application to catchment or regional scales, and attempt to predict responses and impacts over decades rather than months.
1.3 Natural Resource Use and Sustainability
Choices of agricultural pest management practices may have long-term impacts on atmospheric and soil quality. For example, United Nations scientists estimate that methyl bromide -- which is used primarily as a soil fumigant in agriculture -- is responsible for 5-10% of the thinning of the stratospheric ozone layer. Thinning of the ozone shield is an indicator of physical change in the environment that has been related to human health problems, to effects on non-human biota, and to marine and agricultural productivity (Allen et al. 1995; UNEP 1992, 1994, 1995).
On a global scale, fossil energy resources are finite and non-renewable, although their use has quite different economic and social ramifications as a cost of production in different political jurisdictions. Fossil energy is used in agriculture directly as a fuel, and indirectly as embodied in farm machinery, transportation, pumped irrigation, synthetic pesticides and chemical fertilizers. When quantities of fossil inputs are converted to energy units such as calories, joules and BTU's it can be seen that the ratio of energy input to output in agriculture has changed significantly over time and with changing priorities and options in production and distribution. Fossil energy and electricity use on US farms had increased more than 6-fold between the turn of the century and the late 1970's when oil price shocks spurred energy conservation throughout the economy. At peak usage in 1978, direct and indirect energy use on farms was equivalent to 5% of total US energy consumption, while energy inputs to the entire food system (including distribution and processing) have been estimated at three to four times that amount. By 1990, however, energy productivity in agriculture had doubled from the minimum levels of the mid-1970's due to conservation, reduced acreage tilled, and greater use of diesel fuel, which delivers more mechanical energy per unit than gasoline (Cleveland 1995).
The significance of energy as an economic cost of production is of course recognized by growers, but we stress it here because energy analysis is a means of making a link between socioeconomic factors and environmental consequences. It is estimated that domestic sources of high quality fossil energy will be depleted within the lifetimes of people who are now middle aged (Hall, Cleveland and Kaufmann 1986). This will likely have serious, widespread ramifications on our environment and way of life, affecting the scale and location of agricultural production, the delineation of marketscapes and food systems, the demand for agricultural land and labor, the use of synthetic (fossil-based) pesticides and nutrients, and interest in promoting non-fossil-based alternatives in pest control and fertilization. Despite the relatively short time scale of these projected changes, we have seen stops and starts in developing policies and pricing systems that inspire more efficient use of these resources. Therefore, we suggest that evaluating environmental consequences of non-renewable and slowing renewable resources use may provide additional insights and leverage in policy formation.
1.4. Summary: What is Environmental Impact Assessment?
2. Challenges in Assessing Environmental Impacts
2.1. Choosing Environmental Indicators and Deciding How to Integrate Them
Another challenge to creating a composite assessment of environmental impacts of agricultural strategies is finding a meaningful common currency to describe different types of impacts. In answering many questions about environmental impacts, monetary values do not adequately describe non-market costs -- such as the loss of an individual life, loss of biodiversity, impacts on 'non-game' species, disruption of an ecosystem, future costs of current soil erosion, or loss of non-replaceable resources. Ongoing research in several disciplines (and inter-disciplines) is aimed at devising means of valuing environmental and other non-market goods; much of this work falls under the rubric of 'resource ecological economics' (Daly 1991; Daly and Townsend 1993; Daly and Cobb 1994; Guinee and Heijungs 1995; Krishnan, Harris and Goodwin 1995).
In some agricultural impact assessment systems, both environmental parameters and on-farm economic costs are rated on a unitless scale; in others, on-farm costs are quantified in monetary terms and environmental costs are indexed separately and 'flagged' to indicate a hazard or high risk. In a number of other systems, monetary values are imputed to a range of environmental impacts using one of several methods such as replacement or remediation costs, lost productivity, or 'willingness-to-pay' (contingent valuation) as the basis for assigning value to impacts. The drawback to remediation or replacement cost accounting is that money is only a useful measure of impact if the environmental parameters or organisms in question are of intrinsic economic interest to people, or if the costs of previous remediation efforts are known (see Pimentel et al. 1992). Contingent valuation is a useful measure only if the group surveyed for their willingness-to-pay are realistically able to assign monetary values to the non-market goods in question and are not swayed by thinking there will be possible economic or regulatory ramifications from answers that are biased high or low. Surveys to find out how much money individuals would be willing to pay for a non-market good are valid only when the sample represents the population that will bear most of the associated costs or reaps most of the associated benefits. To give an example illustrating this last point: a farmer's willingness-to-pay to avoid polluting water with a toxic pesticide or fertilizer runoff is not a reasonable or accurate way to value this environmental damage because all of society suffers from the results of such pollution and pays the costs of remediation. On the other hand, a survey assessing farmers willingness-to-pay to avoid toxic risk to pesticide applicators may indeed be a reasonable method of valuation because this environmental cost affects farmers disproportionately. In designing assessment systems, it is important to remember that willingness-to-pay does not measure the existence or extent of an environmental problem; rather it measures attitude toward a problem, and whether the problem bothers a particular stakeholder enough to pay for an alternative (Levitan et al. 1995).
Another challenge of creating composite assessments of environmental impacts is due to the reality that there is no one set of social or environmental indicators that is most appropriate to use in assessing impacts of agriculture. Different circumstances and objectives prioritize different indicators and interpretations. One may answer the question of how to integrate, weight and value impacts in the context of one assessment scenario, but these issues will re-emerge when the question of environmental impacts is asked on a different scale or with different objectives. For example, the types of data required to create a decision model for a farmer to use in the field in choosing a 'least impact' but efficacious pest control method, may not be the same as the data required for a national policy model assessing agricultural practices. To illustrate: while IPM farmers want to avoid using pesticides that harm parasites and predators specific to the crop pests in their fields, these producers might be misled by a decision model based on the more generic information about impacts of pesticides on beneficials that might be used in a national model of environmental impacts of IPM. Were the national model to consider impacts on beneficials at all, it would most likely rely on EPA data on acute toxic impacts of pesticides to honey bees, which are the only beneficials included in EPA's Ecological Effects dataset (US EPA 1996). Even if the toxic dose responses were comparable for honey bees and other beneficials, the significance of these effects might be quite different. When honey bees are repelled from a field by pyrethroid pesticides, for example, they survive and move on to another nectar source; however, if beneficial parasites and predators are repelled from a location, they are not then available to work as biological control agents. The design of an assessment system must, therefore, be appropriate to the objectives of the audience served.
2.2. Bias Against Future as Compared to Present Impacts
A second manner in which we can be biased against the future as compared to the present is by not considering impacts associated with future events (Garetz 1993) such as leaking of improperly stored pesticides in the future. Assessing future impacts of future events can be more uncertain than assessing impacts of current events, but this does not mean that such impacts are less important. For example, the Superfund Program and Hazardous Waste Program were established primarily on the basis of future rather than current risks.
Another problem for current assessments is that as environmental systems change or become better understood in the future, the impact of IPM and other farm management systems may be assessed differently. This implies that assessors must be aware of new information and problems, and be prepared to modify or change their assessment methods to account for changes in our knowledge base.
2.3. Data Limitations
Toxicological and ecological effects datasets of pesticides are incomplete. In addition, some of the existing toxicity data are inappropriate to use as the basis for assessing relative impacts of different agricultural management strategies because they were not collected using standardized protocols and, therefore, are not comparable (Levitan et al. 1995). Moreover there are very limited data and no standardized datasets on new biocides, such as microbial and fungal pesticides. The scientific community is only beginning to develop tools and to collect data for assessing positive and negative environmental impacts of bio-intensive IPM practices. The reasons for this are twofold. First, there are many inter-linked physical, chemical and biological processes which play a role in IPM, and it would be unusual for all of these processes to be fully understood and quantified for specific evaluations. Second, natural systems are inherently variable, both in space and time, and in order to characterize both their average behavior as well as their variability, high-intensity sampling is required. Since it is often the occasional extreme occurrences which may lead to environmental damage, it is important to be able to predict the likelihood of these events (Wagenet and Hutson, 1994; Jury and Gruber, 1989).
As we note in an earlier section, most available data on pesticide environmental impacts originate from toxicity tests on single species of biota. In addition to limitations associated with testing single species of organisms, these data are also of limited value because the pesticides tested are generally applied in single doses of individual active ingredients. Impacts to the environment, however, are from mixtures of active ingredients -- whether tank mixes or mixes of residues in the environment -- which can be greater or less than the sum of impacts from individual toxins. Cumulative impacts from repeated or extended exposures can also be different than impacts of single, larger exposures. Little is known about cumulative impacts and interactive effects, particularly in terrestrial systems, even though both human and non-human biota are virtually always exposed to chemical mixes and amounts that change spatially and over time (Yang 1994). Yang concludes that the toxicology of long-term, low level exposures to chemical mixtures produces subtle effects, unlike acute toxic responses to higher doses; that such toxic interactions are possible at environmentally-realistic levels; that the toxic responses may be from unconventional endpoints that are not usually tested; that there is a possibility that residual effects may become interactive with later exposures; and that these exposures may pose a safety risk to the public. While these comments are intended to apply to human subjects, we can extrapolate these principles and concerns to non-human biota, some populations of which may be more vulnerable to such risks because of limited mobility and physiological factors.
2.4. Summary: Challenges in Assessing Environmental Impacts
3. Methodologies for Impact Assessment
In this section we review several categories of environmental impact assessment methods, including (3.1.) surveys and monitoring, (3.2.) fate models, and (3.3.) categorical indices of impacts. In each case we discuss the objectives, strengths and limitations of the methodology. All of these approaches have been used in environmental assessments of agriculture. The aim of this section of the paper is to encourage IPM researchers to actively consider the objectives and assumptions of the methods they are using and to refine methods where feasible, rather than mechanically adopting methods without appropriate adaptations. In this way, researchers will not only increase the usefulness of their assessment, but may also contribute to the development of environmental assessment methods.
3.1. Sampling and Monitoring
Of all the methodologies we will be discussing, sampling and monitoring are the most familiar to IPM researchers. Sample surveys are used in many fields to characterize populations (used broadly here to include biotic and abiotic phenomena) that are too large to census. Monitoring of various components of the environment usually involves repeating sample surveys over time. However, there are cases when monitoring involves measuring changes in the entire population of interest, rather than a sample of that population, for example when monitoring changes in a population of some endangered species. In any case, the major objective of monitoring is to address questions concerning the present status, changes and future trends in the population that is being monitored (Larsen 1995).
On the national level, the US Geological Survey, the USDA Soil Surveys and the national network of weather stations have long been engaged in surveying the physical resource base of the nation and in providing this information to the public. More recently there has been a growth in the use of surveys to characterize the natural and agricultural resource base. Examples include the National Agricultural Statistical Survey, the Forest Inventory Assessment, the National Wetlands Inventory, and the National Acidic Precipitation Program's survey of lakes and streams. Surveys conducted over time add a temporal dimension to survey data, thus moving beyond a 'snapshot' approach to resource inventory and essentially becoming a monitoring exercise. The US-EPA Environmental Monitoring and Assessment Program (EMAP) is an example of a program designed to track changes in important environmental indicators that have been selected to characterize the condition of the nation's ecosystems. Another example of an environmental monitoring program is the Swiss National Soil Monitoring Network (Desaules 1993).
IPM researchers are familiar with sampling and monitoring of the environment at the local level since these activities are a major part of IPM research and practice. The strengths and weaknesses of surveying and monitoring are similar at local and regional levels. Surveys based on population samples make it feasible to characterize environmental resources, such as soil, lakes and streams, as well as biotic populations that are too large to census. Otherwise the status of a population would have to be inferred from an indicator or other species or simulation modeling. Monitoring can also be used to provide data for evaluating whether a system is changing and to predict future trends.
Obvious problems with sampling and monitoring are those of cost, convenience and extrapolation. Often, so many samples must be taken to validly describe a population that the cost of sampling may become prohibitive. At other times, it can be impractical to choose a valid sample population. For example, farmers who are interested in working with extension agents and researchers to implement new pest management strategies are not necessarily representative of the entire population of farmers who are using more conventional techniques. Given the voluntary nature of such arrangements, it may not be practical to select an unbiased sample of farmers. Lastly, without using other tools, the results of the sampling and monitoring work cannot be used to draw inferences about other populations (i.e. other farms, other practices, other components of the environment).
There are several other problems associated with monitoring beyond those of cost, convenience and inability to extrapolate to populations not represented by the sample. Much of the rational for monitoring lies in trend detection. However, in some environments trend detection has been likened to 'looking for a needle in a haystack' -- with the needle often being very small changes representing a trend lost in the haystack of measurement error and natural random fluctuations in time and space (Oliver 1993). Clearly, knowledge of natural fluctuations in time (e.g. seasonal effects) and space (e.g. soil types or soil depth) need to be considered in designing a monitoring system (Oliver 1993). Dynamic simulation models can be used to predict temporal and spatial fluctuations and potentially improve the design of a monitoring system. When the trend is very small compared to natural fluctuations in time and space, then other approaches need to be considered. An interesting improvement over standard monitoring is the combination of regional mass balances with monitoring data by the soil monitoring network in Switzerland mentioned above (Bader and Baccini 1993, von Streiger and Obrist). The approach used in the Swiss study is to identify various categories of farms and then apply a model that distributes system inputs and outputs by farm category using regional average data. This method was used to identify agricultural land at high risk for copper contamination (in this case it was 11.9% of the total cultivated land) and then to focus monitoring activity on this smaller area of cultivated land at high risk. Such an approach can guide those responsible for monitoring and influence how often and where samples should be collected.
3.2. Fate Models
Integrating and extrapolating physical, chemical and biological processes in the environment is an essential part of assessing impacts of agriculture. Natural systems are dynamic. Models identify the relative importance of various dissipation pathways, and allow estimation of flux densities, concentrations, residence times and exposure. Since most data collection is performed at detailed scales, simulation models are an attractive option for extending these data to broader space and time scales. Models may be viewed as repositories for dynamic processes, analogous to databases, which are often repositories for static data only.
Dynamic simulation models vary in their scope and complexity (Addiscott and Wagenet 1985), falling into broad use categories of education, screening, regulation, and research. The simplest of these models require few data and sometimes contain overly simplistic assumptions, but are easy to run and are useful for demonstrating the principles of environmental interaction. Screening models are usually used to rank chemicals in terms of potential environmental impact, and generally compare the relative impact of a different chemicals against a constant environmental background. Models currently used for pesticide registration include environmental dynamics (rainfall, temperature, etc) but exclude processes which may be important but which are currently difficult to quantify, such as sorption kinetics. In regulatory models, processes are often represented as simply as possible, consistent with current knowledge and available data. Regulatory models make extensive use of libraries of existing databases and are structured to perform multiple executions easily. Research models are the most detailed in terms of their representation of processes. Their data demands are usually high, and considerable knowledge and experience are required to use them effectively.
The complexity and dynamic nature of environmental processes make simulation particularly attractive. The use of computer simulation models is increasing despite controversy over their validity and applicability. The controversy arises from opposing views of how models should be used. At one extreme there are those who feel that models should contain only processes that have been proved valid, and that they should not be applied outside a range of situations for which they are applicable. At the other extreme are those who would apply models even though the processes or data are known to be inapplicable to the situation under study. Useful applications probably lie between these two extremes, especially when combined with a critical and insightful evaluation of the output. Hauhs (1990) suggests that models should be applied until they are shown to be invalid, as they represent the current level of knowledge. However, if evidence from measurement, monitoring or experience suggests that the model is deficient or inappropriate, then the scientific foundation of the model should be re-examined and improved.
When using a model outside the situation in which it has proven applicable, it is important to remember that the model is a hypothesis, and that subsequent measurement may prove it invalid or incomplete. Other approaches and available data should be reviewed before embarking on a modeling exercise. Such a review will highlight areas where there are insufficient data, thus highlighting the role of model output as a possible substitute. During this evaluation process, major mass balance components may be estimated, and deemed sufficiently accurate to satisfy demands of other disciplines.
Environmental evaluation often consists of the application of established scientific principles or models from several disciplines to larger-scale systems. The models employed at this larger scale are based on processes determined at the research scale. Processes that control responses at the larger (e.g., catchment) scale should be included, but are not necessarily present in smaller-scale models. At larger, more complex levels, direct cause and effect relationships are more difficult to establish, and existing process-based models may become inadequate. Long-term experience and monitoring may become the sole measures of behavior at larger scales. But if models are viewed as providing hypotheses about system response at the larger scale, then it may be possible to design experiments or measurement exercises which can help assess the models. In this way we may develop a science at the larger environmental scale which does not depend completely on scaling-up of local-scale research.
3.3. Index or Ranking of Impacts of Pest Control Products and Methods
Some indexing systems use categories such as high, moderate, low or no risk; in others these categories are translated into the colors at a stop light: 'red' for high hazard, impact or risk; 'yellow', where there are moderate impacts and the practice should be used with caution; and 'green' for 'go ahead' -- to indicate there is little or no impact from the practice. In some systems these categories are scored, and the scores serve as the 'common currency' to be weighted and summed in creating a composite assessment of impact from the practice. In other systems, continuous numerical ratings are used rather than discrete categorical interpretations of the data on impact. These numbers may be derived directly from toxicity tests (such as an LD50 value); may be a numerical test result modified by an exposure factor or other situation-specific property; or be a ratio of environmental concentration to an effective concentration that causes a measurable impact (such as an LD50 or EC50). In other systems, such as the World Wildlife Fund's assessment of adoption of IPM practices described by Hoppin (this volume), the categories are 'behavioral' -- in terms of types of IPM practices (low level IPM, medium, bio-intensive IPM) -- rather than categories of magnitude of impact. In such 'behavioral' systems, a relationship is assumed between certain behaviors or practices and the impacts of these practices.
Indexing and ranking systems are well-suited for comparing relative impacts of similar pest management options, such as comparing toxicity of different pesticides, each of which have been assessed for the same endpoints at similar levels of exposure. Due to the conceptual difficulties in integrating different measures and indicators of impact, there is a 'greater margin of creative interpretation' when indexing is used to compare impacts of quite different options. Some examples are comparing impacts of herbicides to control weeds versus tillage, or comparing regional food production systems where pesticides may be used to the environmental impacts of transporting organically-produced food from a different agricultural region. Such systems are well-suited for evaluation with hybrid assessment tools that draw on the strengths of both indexing and simulation methods.
Indexing systems are useful for evaluating many types of environmental variables, not only those which can be sampled, monitored or mathematically modeled. This enables the leap from assessments based on test endpoints to the development of systems for assessing decision endpoints. We return to the example of the impact of different pesticides on honey to illustrate the difference: The measurement of toxicity to an organism is a test endpoint that provides data on the rate of pesticide application lethal to bees, or the rate at which certain behaviors (such as nectar-collecting activity) will change. However, what a beekeeper is more likely to want to know is the combination of factors affecting hive survival or crop pollination. Management decisions of farmers and beekeepers could be affected by knowing how the impact on honey bees might be reduced by using a different pesticide, a lower dosage or a different time of application.
In this example, acute toxicity to adult honey bees may not be the crucial variable for the beekeeper's decision, because the most toxic pesticides may rapidly kill worker bees in the field, or repel them away from the field (as pyrethroid insecticides do), whereas somewhat less-acutely toxic pesticides may mix with the nectar or pollen and be brought back to the hive and fed to the brood, which are the next generation of workers. Or the less-acutely-toxic pesticide may have a sub-lethal impact on the adults, reducing their activity level and decreasing long-term chances of hive survival. Indexing systems have the potential of integrating test endpoints and ranking decision endpoints. A decision-making aid for determining whether a situation is hazardous to hive survival or pollination success might require the integration of a number of test. Decision models for efficient and safe management practices for farmers, growers, livestock managers and beekeepers might differ from each other, and also be different than assessment models intended to summarize long-term and off-farm impacts to the environment and society. Without modifications such as those described in this example to incorporate site- and situation-specific factors, ranking systems reflect a 'generalized' condition. In pesticide ranking systems site and situation specific factors include dose, time of day and season of application, and qualities of the formulated product.
A challenge in developing indexing systems is that the integration of impacts on specific endpoints into a composite assessment of impacts on the environment involves value judgment. The challenge is in justifying these judgments and in creating assessment tools which are sufficiently transparent and flexible to enable situation-specific modifications in the integrating algorithm. As methods are developed to incorporate situation-specific sensitivity to impacts, the value of indexing systems will improve.
3.4. Directions and Trends in Impact Assessment Systems
Systems will be developed which are more transparent and flexible in setting impact criteria, in determining which variables to include in the model, and in weighting relative importance of these variables in the system. With improved input data, and these other modifications, assessment models will be able to portray a more holistic picture of environmental impacts.
4. Choosing an Assessment Method
4.1 Identifying Decisions, Values, and Assessment Endpoints
Throughout this paper, we have emphasized that environmental impact assessment has no single, well-defined methodology. In the first section we emphasized that there are numerous environmental assessment endpoints of interest to various groups. In the next section, we raised questions suggesting that it is still not possible to conduct a complete (i.e. an 'holistic') environmental assessment. In the third section we discussed the objectives, strengths and limitations of some existing methodologies for environmental assessment of agriculture, pointing out limitations to each of these methods. How then should IPM researchers determine an appropriate approach to use in assessing the environmental impact(s) of the management systems they are promoting? Suter (1995) states that the selection of an appropriate environmental assessment method that will lead to an informed decision must involve not only the assessors but also must be guided by an understanding of the public values involved in the decision. He suggests that selecting the appropriate method requires addressing four questions: 1. What is the nature of the decision? 2.What societal values are involved in the hazard to be assessed? 3. How can those values be operationally defined as assessment endpoints? 4. What combination of models, test endpoints and other data will most efficiently provide an assessment of the assessment endpoints in a form suitable for the decision? In the next few paragraphs we discuss these and other questions related to choosing a particular environmental assessment method.
Before selecting an environmental assessment method, it is critical to determine who is expected to use the assessment method and the information it generates. Is the information to be used by government agencies to assess policy impacts, or by growers to inform them of the potential environmental consequences of management decisions? Because many pest management systems involve multiple decisions, IPM assessments potentially involve contrasting the impact of a range of decisions (the impact of the application of different pesticides, at different rates, at different times and at different places) rather than just contrasting the standard use of a pesticide with no use of a pesticide.
There can be multiple societal values involved in estimating hazards of pesticide use. Excluding human health concerns, farmers are concerned about the impacts of pesticides on beneficials and the inducement of pesticide resistance in target populations. Regulatory agencies are concerned with how farm management decisions may impact benchmark values for pesticide levels in water and air. Other government agencies may be interested in endpoints that are important on a global scale and thus subject to international negotiations (Cairns 1995). Many in the general public are concerned with the impacts of pesticides on non-target organisms, while environmentalists are also concerned with long-term, ecosystem-level impacts that may not be safeguarded by current standards. Scientists are concerned with potentially significant, unstudied impacts. Depending on the environmental values of the assessment developers and target audience, assessments of environmental impact of alternative decisions could be primarily focused on the short-term versus the long-term consequences and on site-specific versus regional or national impacts. Some groups may be interested in potential negative environmental consequences of proposed practices and want these to be compared to the environmental impact of standard production practices. Thus the assessment or decision endpoints of most interest are likely to differ among different groups (Suter 1995). A quotation earlier in the paper (Hughes 1995) suggests that an environmental assessment of IPM should include assessment endpoints of interest to a broad spectrum of interested parties. Cairns (1995), in an article dealing with future trends in ecotoxicology, argues that ecotoxicological information will need to be more site specific and generated more rapidly.
The implications of Suter's questions referred to at the beginning of this section are that only once the nature of the decision(s), societal values involved and assessment endpoints are identified, can the models, test endpoints and data necessary to assess the endpoint be determined. As Suter points out, despite this ideal, most assessments have to rely on standard test endpoints available from existing toxicity data. These generally are not the assessment endpoint. In this case, the role of the assessor must include tailoring the assessment to the decision. When considering use of an existing environmental assessment tool, it is important to determine whether the assumptions and data used in developing the tool are appropriate to conditions or systems under which it will now be applied. For example, a pesticide hazard rating developed for apple orchards may not be appropriate for vegetable or grain crop systems. There may be a need for further measurements and it may also be necessary to refine or further develop the assessment tool.
4.2. Choice of a Model
Choice of a model will depend on the reason for modeling, i.e. the questions we expect to answer. For example, a screening model may provide all the information required if the objective is merely to rank chemicals in terms of their potential for reaching groundwater. However, if a site-specific assessment is required, then data pertaining to that site and its weather have to be included, which necessitates a more complex model. In a scientific study of isolated and controlled processes, a simple model is likely to be successful, whereas more complex models which include many processes are required for large-scale simulations. Regardless of the application, an intelligent selection of a model requires the user to have a clear understanding of how well the processes included in the candidate models describe the processes likely to be important in the field.
At the outset we need to recognize that the processes included in models are usually elucidated under highly controlled conditions. Interactions between processes and their behavior under changing environmental conditions are rarely studied, except in field experiments limited both in space and time. Thus models are constructed to predict behavior under field conditions, and to extrapolate processes to other soils and over longer times. Because it is impossible to measure everything, it is inevitable that models will be used to provide an extension of empirical knowledge.
4.3. Towards a Holistic Approach to Environmental Impact Assessment of Agriculture
We will close by referring back to the objectives reflected in the title of this paper: Environmental Impact Assessment: the Quest for a Holistic Picture, but with this quest modified somewhat by the conceptual challenges and technical limitations we have described. We have stressed the point that no single assessment system could include all of the environmental parameters we have mentioned, and do so accurately at all scales of operation from decisions made on a farmer's fields, to evaluating regional or watershed impacts, to national policy models, to planetary assessments. Nevertheless, in designing and implementing assessment systems, we believe it is preferable to think about the implications and ramifications of an agricultural practice on all of a system, rather than to think only about a limited portion of the system, while believing or implying that it is an assessment of impact on the entire system. We need to remember that environmental processes are ongoing even if they are not being monitored, sampled or included in the assessment model.
In creating decision tools from assessment systems, we must think broadly about 'environmental impacts', and develop methods for integrating environmental costs, public health costs, social costs and on-farm costs -- without losing valuable information about each set of issues. What this suggests is that both environmental impacts (non-target costs) and farm cost data (target impacts) need to be collected, but analyzed independently. Conclusions from an analysis of the monetary costs of pest control should not influence or mitigate assessments of non-target (environmental or social) costs. After all, environmental degradation and resource depletion resulting from a given practice do not decline because the economic costs of doing without a pesticide are high. Environmental impacts do not go away just because there are few alternative practices or products available. However, while the environmental assessment should not be mitigated by production cost data, the decision about which production strategy to follow must of course weigh the information gleaned about on-farm costs as well as environmental impacts. These decisions should not be made in a 'black box'. When the economic costs of environmental protection are high, society perhaps needs to consider whether and how to shift that economic burden from the farmer or the consumer to a larger group. In order to have this discussion, the methodology and results of impact assessment systems must remain visible (Fig. 3).
So what can be expected from environmental impact assessment systems? As we have implied, there are many ways to evaluate the environment and many ways to integrate a summary of impacts from specific agricultural strategies. We suggest that one of the greatest values of developing environmental impact assessment systems is that they will facilitate rational social discourse about the effects, implications and sustainability of agricultural production and marketing systems. It is our hope and prediction that good assessment systems will draw a broader group of better informed parties into that discussion.
References
Addiscott, T.M. and R. J. Wagenet. 1985. "Concepts of solute leaching in soils: a review of modeling approaches". J. Soil Sci. 36, 411-424.
Allen, Will, Jo Ann Baumgartner, Kalee Kreider and Susan Pollack. 1995. Out of the frying pan, avoiding the fire: Ending the use of methyl bromide. An analysis of methyl bromide use in California and the alternatives. Washington DC: Ozone Action, Inc. 37 pp.
Bader, Hans-Peter and Peter Baccini. 1993. "Monitoring and control of regional material fluxes". In R. Schulin, A. Desaules, R. Webster and B. von Steiger (eds.), Soil monitoring: early detection and surveying of soil contamination and degradation. Basel, Switzerland: Birkh-user Verlag Basel, pp. 25-34.
Barbour, M.T., J. B. Stribling and J.R. Karr. 1995. "Multimetric approach for establishing biocriteria and measuring biological conditions". In W.S. Davis and T.P. Simon (eds.), Biological assessment and criteria: tools for water resource planning and decision making. Boca Raton, Fl: Lewis Publishing, pp.63-77.
Cairns, John Jr. 1995a. "The genesis of ecotoxicology". In J. Cairns, Jr. and B.R. Niederlehner (eds.), Ecological toxicity testing: Scale, complexity and relevance. Boca Raton, FL: Lewis Publishing, pp. 1-10.
Cairns, John Jr. 1995b. "Future trends in ecotoxicology". In J. Cairns, Jr. and B.R. Niederlehner (eds.), Ecological toxicity testing: Scale, complexity and relevance. Boca Raton, FL: Lewis Publishing, pp. 217-222.
Cleveland, Cutler J. 1995. "The direct and indirect use of fossil fuels and electricity in USA agriculture, 1910-1990". Agriculture, ecosystems & environment 55: 111-121.
Cook, Earl. 1976. Man, energy, society. San Francisco: Freeman. 476 pp.
Cooper, William. 1993. "Ecological health risks: introduction". In C.R. Cothern (ed.), Comparative environmental risk. Boca Raton, FL: Lewis Publishing, pp. 35-36.
Cottrell, Fred. 1955. Energy and society. NY: McGraw-Hill.
Daly, Herman E. 1991. Steady-state economics: the economics of biophysical equilibrium and moral growth. 2nd ed. Washington DC: Island Press. 302 pp.
Daly, Herman E. and Kenneth N. Townsend. 1993. Valuing the earth: economics, ecology, ethics. Cambridge, MA: MIT Press. 387 pp.
Daly, Herman E. and John B. Cobb, Jr. with Clifford W. Cobb. 1994. For the common good: redirecting the economy toward community, the environment, and a sustainable future. Boston, MA: Beacon. 534 pp.
Desaules, A. 1993. "Soil monitoring in Switzerland by the NABO-network: objectives, experiences and problems". In R.Schulin, A. Desaules, R.Webster and B. von Steiger (eds.), Soil monitoring: early detection and surveying of soil contamination and degradation. Basel, Switzerland: Birkhauser Verlag Basel, pp.7-24.
Diamond, J. 1987. "Goals: theory and practice in restoration ecology". In W.R. Jordan, M.E. Gilpin and J.D. Aber (eds.), Restoration ecology: a synthetic approach to ecological research. NY: Cambridge University Press, pp. 329-336.
Fava, James E., Richard Denison, Bruce Jones, Mary Ann Curran, Bruce Vigon, Susan Selke and James Barnum (eds.). 1991. A technical framework for life-cycle assessment. Pensacola, FL: Society of Environmental Toxicology and Chemistry and SETAC Foundation for Environmental Education, Inc. 134 pp.
Fava, James E., Frank Consoli, Richard Denison, Kenneth Dickson and Tim Mohin (eds.). 1993. A conceptual framework for life-cycle impact assessment. Pensacola, FL: Society of Environmental Toxicology and Chemistry and SETAC Foundation for Environmental Education, Inc. 160 pp.
Guinee, Jeroen and Reinout Heijungs. 1993. "A proposal for the classification of toxic substances within the framework of life cycle assessment of products". Chemosphere 10:1925-1944.
Guinee, Jeroen and Reinout Heijungs. 1995. "A proposal for the definition of resource equivalency factors for use in product life cycle assessment". Environmental toxicology and chemistry 14:917-925.
Garetz, William V. 1993. "Current concerns regarding the implementation of risk-based management: how real are they?" In C.R. Cothern (ed.), Comparative environmental risk. Boca Raton, FL: Lewis Publishing, pp.11-31.
Hall, Charles A.S. (ed.) 1995. Maximum power: the ideas and applications of H.T. Odum. Niwot, CO: University Press of Colorado. 393 pp.
Hall, Charles A. S., Cutler J. Cleveland and Robert Kaufmann. 1986. Energy and resource quality: the ecology of the economic process. NY: Wiley-Interscience. 577 pp.
Hauhs, M. 1990. "Ecosystem modelling: science or technology?" J. Hydrol. 116: 25-33.
Helsel, Zane R. (ed.) 1987. Energy in plant nutrition and pest control. Energy in world agriculture series. Amsterdam and NY: Elsevier Science Publications. 444 pp.
Hoppin, Polly. 1996. "Reducing pesticide reliance and risk through adoption of IPM: an environmental perspective". This Volume.
Hughes, Robert. M. 1995. "Defining acceptable biological status by comparing with reference conditions". In W.S. Davis and T.P. Simon (eds.), Biological assessment and criteria: tools for water resource planning and decision making. Boca Raton, FL: Lewis Publishing, pp. 31-47.
Jury, W.A. and Gruber, J. 1989. "A stochastic analysis of the influence of soil and climatic variability on the estimate of pesticide and groundwater pollution potential". Water Resour. Res., 25:2465-2474.
Karr, James R. 1995. "Protecting aquatic ecosystems: clean water is not enough". In W.S. Davis and T.P. Simon (eds.), Biological assessment and criteria: tools for water resource planning and decision making. Boca Raton, FL: Lewis Publishing, pp.7-13.
Karr, James R. 1981. "Assessment of biotic integrity using fish communities". Fisheries 6:21-27.
Krishnan, Rajaram, Jonathan M. Harris and Neva R. Goodwin (eds.). 1995. A survey of ecological economics. Frontier issues in economic thought series. Washington DC: Island Press. 384 pp.
Larson, David P. 1995. "The role of ecological sample surveys in the implementation of biocriteria". In W.S. Davis and T.P. Simon (eds.), Biological assessment and criteria: tools for water resource planning and decision making. Boca Raton, FL: Lewis Publishing, pp.287-300.
Levitan, Lois, Ian Merwin and Joe Kovach. 1995. "Assessing the relative environmental impacts of agricultural pesticides: the quest for a holistic method." Agriculture, ecosystems & environment 55: 153-168.
Odum, Howard T. 1971. Environment, power, and society. NY: Wiley-Interscience. 336 pp.
Odum, Howard T. 1983. Systems Ecology: An Introduction. NY: Wiley-Interscience. 643 pp.
Pimentel, David (ed.). 1980. Handbook of energy utilization in agriculture. Boca Raton, FL: CRC Press. 475 pp.
Pimentel, David, Herbert Aquay, Michael Biltonen, P. Rice, M. Silva, J. Nelson, V. Lipner, S. Giordano, A. Horowitz and M. D-Amore. 1992. "Environmental and economic costs of pesticide use." BioScience 42:750-760.
Policansky, David. 1993. "Application of ecological knowledge to environmental problems: ecological risk assessment." In C.R. Cothern (ed.), Comparative environmental risk. Boca Raton, FL: Lewis Publishing, pp. 37-51.
Schroll, H. 1994 "Energy-flow and ecological sustainability in Danish agriculture". Agriculture, ecosystems & environment 51:301-310.
Simon, T.P. and J. Lyons. 1995. "Application of the index of biotic integrity to evaluate water resource integrity in freshwater ecosystems". In W.S. Davis and T.P. Simon (eds.), Biological assessment and criteria: tools for water resource planning and decision making. Boca Raton: Lewis Publishing, pp. 245-262.
Southerland, M.T. and J. B. Stribling 1995. "Status of biological criteria development and implementation". In W.S. Davis and T.P. Simon (eds.), Biological assessment and criteria: tools for water resource planning and decision making. Boca Raton, FL: Lewis Publishing, pp. 81-96.
Suter, G. W. 1995. "Endpoints of interest at different levels of biological organization". In J. Cairns, Jr. and B.R. Niederlehner (eds.), Ecological toxicity testing: scale, complexity and relevance. Boca Raton, FL: Lewis Publishing, pp. 35-48.
Suter, G.W. 1993. "A critique of ecosystem health concepts and indexes". Environmental toxicology and chemistry 12: 1533-1539.
UNEP 1992. Montreal protocol assessment supplement: methyl bromide: Its atmospheric science, technology and economics. Synthesis report of the methyl bromide interim scientific assessment and methyl bromide interim technology and economic assessment. Executive summary. Nairobi, Kenya. United Nations Environment Programme, Ozone Secretariat. 5 pp.
UNEP 1994. Report of the methyl bromide technical options committee. Nairobi, Kenya. United Nations Environment Programme, Ozone Secretariat.
UNEP 1995. Cited in Pesticide Action Network Update, March 3, 1995 "UN confirms significant ozone destruction". On-line distribution (paninfopubs@igc.apc.org).
US EPA. Office of Pesticide Programs. Ecological effects database. 1996. Contact: Brian Montague.
von Steiger, B. and J. Obrist. 1993. "Available databases for regional mass balances in agricultural land". In R.Schulin, A. Desaules, R.Webster and B.von Steiger (eds.), Soil monitoring: early detection and surveying of soil contamination and degradation. Basel, Switzerland:Birkh-user Verlag Basel, pp.35-46.
Wagenet, R.J. and J.L. Hutson. 1994. Computer simulation models as an aid in estimating the probability of pesticide leaching. In: 5th International Conference on Computers in Agriculture, Orlando, FL, Feb. 5-9, 1994, pp. 786-794.
Yang, Raymond S. H. (ed.) 1994. Toxicology of chemical mixtures: case studies, mechanisms, and novel approaches. San Diego: Academic Press. 720 pp.
Notes and Acknowledgements
Figures (not included)
Figure 1b.
Figure 2. Space and time scales of environmental studies.
Figure 3. Integrated assessment. Tradeoffs betwen economic cost and
environmental impact can be calculated and visualized.
Dr. Lois Levitan
Department of Fruit and Vegetable Science
162 Plant Sciences
Cornell University
Ithaca, New York 14853
Phone: (607) 255-3033
FAX: (607) 255-0599
e-mail: LCL3@Cornell.Edu
