Pesticides in Perspective Introduction Although consumers Ærst interest in connection with pesticides is the safety of the food that they eat they are increasingly becoming aware of and concerned about the effects pesticides may have on the environment. The previous article in this column described the measures the UK government takes to monitor the environment for such effects. A signiÆcant number of other world governments undertake similar monitoring to a greater or lesser extent. However a signiÆcant amount of experimental and modelling work has to be conducted by industry before compounds are registered for use to prove that such pesticide use does not result in harmful side effects on nontarget organisms in the environment Ecological Risk Assessment for Agricultural Pesticides{ Introduction Pesticides are designed to control target pest disease and weed organisms.In agriculture these chemicals are used to protect crops from competition from weeds and attack by plant pathogens insects and other pests. They have the potential to impact on non-target organisms both within the target area and through movement of the chemical or treated organisms off-target. Furthermore the control of pest species may impact indirectly on other species which rely on the pests as a food source. Ever since Rachel Carson's book Silent Spring the potential effects of pesticides on our environment has often had a high proÆle in the media and is discussed in very emotive terms witness the last line of a previous article in this series ``The environmental impact of pesticides is a far sharper nail {The opinions expressed in the following article are entirely those of the author and do not necessarily represent the views of either The Royal Society of Chemistry the Editor or Editorial Board of JEM Zeneca Agrochemicals or those of the Column Editor.104N J. Environ. Monit. 2000 2 This journal is # The Royal Society of Chemistry 2000 such as birds bees and aquatic organisms. The current article gives an overview of the ecological risk assessment that is required for pesticides globally prior to approval being granted. The author of this article is Mick Hamer who has worked for Zeneca Agrochemicals for 22 years since joining with a degree in Zoology from Oxford University in 1978.His practical research experience covers the fate and effects of pesticides in aquatic ecosystems and includes aquatic ecotoxicology (laboratory and Æeld) bioconcentration and bioavailability in sediments. His current role is as Technical Advocate within the Ecological Risk Assessment Section preparing and presenting risk assessments and technical positions for for the pesticides cofÆn than residues in food''.1 Another article in the series ``Monitoring of Pesticides in the UK Environment''2 described the various schemes that through a combination of chemical and biological monitoring together with incident reporting ensure compliance with regulatory standards and identify potential adverse effects.Of course monitoring for adverse effects and taking subsequent regulatory action whilst a very important tool can be seen as `shutting the stable door after the horse has bolted' as these monitoring programmes are only possible post-registration. As adverse effects are by deÆnition undesirable it is vital that the registration process has the ability to identify potential risks screening out unacceptable chemicals and/or use patterns and allowing for risk mitigation such that risks are reduced to an acceptable level. This article looks at such schemes and how recent developments in ecotoxicity testing and exposure modelling are being incorporated into them. It will also highlight areas where risk assessment Pesticides discussions with regulators pesticide distributors users and any other interested parties.He has been involved in many technical workshops on aquatic and sediment ecotoxicology and risk assessment mostly through SETAC (Society of Environmental Toxicology and Chemistry) of which he is a European Council member. He has presented at many ecotoxicological and risk assessment courses including shortcourses on probabilistic risk assessment for agrochemicals at SETAC US and World conferences. Dr Terry Clark Column Editor Zeneca Agrochemicals UK E-mail Terry.Clark@aguk.zeneca.com procedures are still under development and anywhere it might be considered that deÆciencies exist within such schemes.Before discussing different approaches to ecological risk assessment it is customary to clarify what is meant by risk assessment and a working deÆnition is below. Also given is a deÆnition of risk management which highlights the distinction between the two.3 Risk assessment is a science based process that consists of hazard identiÆcation and exposure assessment and that ultimately integrates hazard and exposure to characterise risk. Risk management is a policy-based activity that deÆnes risk assessment questions and endpoints to protect human health and the environment. It takes the scientiÆc risk assessment and incorporates social economic political and legal factors that impinge or inØuence the Ænal decision and selects regulatory actions.Risk assessment should therefore be a scientiÆc process and not some sort of black art. Risk management has a much wider brief of which science is only one part. However as a risk assessor I have to admit that we frequently have trouble seeing beyond the science and therefore of appreciating the difÆculties faced by the risk manager. Risk can be considered as the probability of an adverse event occurring. Of course deÆning an adverse event is a difÆcult thing to do and it is all too easy to default to the easy option of any effect being looked upon being unacceptable. However this can remove much of the science from the process and should be strongly resisted particularly when we are discussing effects in agriculture which in itself will bring changes to the environment whether chemicals are used or not.Pesticide risk assessment schemes and registration Pesticide risk assessment schemes have been established for some time in the USA under FIFRA (Federal Insecticide Fungicide and Rodenticide Act) and Europe under the EU Pesticide Directive 91/414/EEC. These in common with other procedures for assessing the risk of toxicants adopt a tiered approach. Tier 1 is essentially a screen there to identify low risk uses or those groups of organisms at low risk. It is important to realise that it is not the ``pesticide'' which is being assessed and registered but the use which includes the chemical and its use pattern (application method rate and timing) together with any relevant environmental fate data.So for example a chemical may present an unacceptable ecological risk if applied by air-blast in hops or citrus at rates required to give effective pest control and yet be perfectly acceptable applied by tractor-mounted sprayer for use to control cereal pests. Tier 1 also focuses the assessment on the organisms that might be at risk for example it might identify low risk to terrestrial organisms whilst indicating potential risks to aquatic invertebrates and Æsh. Failure to ``pass'' at tier 1 does not mean that the pesticide use presents a risk rather that there is insufÆcient information to reach a conclusion. The options exist therefore to (i) remove the issue by not allowing registration; (ii) reduce the risk to acceptable levels by reducing exposure through limiting application rates numbers of applications or the use of no-spray zones; (iii) reÆne the risk assessment.The Ærst two options above can be seen as risk management tools. The Ærst option however may deny the obvious beneÆts social and/or economic that the pesticide might bring. Equally an industry which has already invested hugely in the development of a chemical is likely to look for an alternative solution. It is of course in industry's interests to develop chemicals which can be registered and given the importance of having pesticides with acceptable environmental proÆles assessing environmental proÆles is now part of the decision making process during selection of potential pesticides.Given this it should therefore be unlikely that registration of a new active ingredient is totally refused on ecological risk grounds although risk mitigation measures may be necessary. However regulatory requirements evolve rapidly and sometimes unpredictably. One problem therefore can be the time-lag of 5±10 years between pesticide invention development and registration making it is necessary to look into the future and anticipate future regulatory concerns. The process is somewhat different for chemicals already registered however it is important to ensure that chemicals are re-evaluated to ensure they reach current environmental standards.Mitigation measures that result in reducing non-target exposure are an approach that can be adopted although limiting application rates for a speciÆc use is often not a viable option as rates tend to be optimised on efÆcacy. More likely is that certain uses which require higher application rates or result in higher non-target exposures may not be registered. The use of no-spray zones is a useful tool in reducing exposure to nontarget organisms outside the target area for example in aquatic environments or hedgerows. When considering exposure of organisms within the target area which can include birds mammals and other wildlife together with earthworms and beneÆcial arthropods no sprayzones are not an effective mitigation measure.Therefore it is necessary to include in registration schemes the potential for risk assessment reÆnement. This involves further work to remove some of the uncertainties that are inherent in initial risk characterisation to obtain a more realistic estimate of the risk. This work can involve (i) further interrogation of the data already available; (ii) further ecotoxicity testing; (iii) better characterisation of exposure/effects of the chemical. As stated above failure to do any further work will result in regulatory action to mitigate the risk ranging from not registering the use or restricting the Pesticides use such that trigger values are not exceeded. Representative groups of organisms are assessed for risk to pesticides including from the terrestrial environment birds and mammals bees and beneÆcial arthropods earthworms soil micro-organisms non-target plants and from the aquatic environment Æsh aquatic invertebrates and plants.All these organisms are assessed in Europe under 91/414/EEC whereas the US EPA concentrates on birds and mammals bees non-target plants and aquatic organisms. Tier 1 Screening out low risk uses The quotient approach. Tier 1 screens out low risk uses/organisms and most commonly uses the quotient approach. This involves comparing an estimate of toxicity derived from a standard test under laboratory conditions with a worst-case estimate of exposure a predicted environmental concentration (PEC) based on proposed label uses.For aquatic organisms it is assumed exposure is through the water phase and PECs (w/v) in water for example in mg l21 are compared to a toxicity value (LC50 NOEC) expressed in the same units. For birds and mammals it is assumed that exposure is through eating treated food items and residue concentrations (w/w) in mg kg21 are compared to toxicity data (dietary LC50 NOEC) again in the same units. The estimate of exposure at tier 1 is generally a maximum modelled residue concentration regardless of how long the exposure in the toxicity tests which could be up to 2 years in some of the longer term chronic tests. The US EPA derives a risk quotient (RQ) as RQ~ exposure toxicity and the EU derives a toxicity/exposure ratio (TER) as TER~ toxicity exposure For toxicity to bees a similar situation exists whereby the exposure (Æeld application rate) is divided by the toxicity (mg bee21) to derive a hazard quotient.If the resulting RQ is less than a certain value the level of concern (LOC) the risk is characterised as low and no further action is required. The same is true if the TER exceeds a trigger value. Effectively in common with many 105N J. Environ. Monit. 2000 2 Pesticides other risk assessment schemes a safety factor is being applied to cover uncertainty. In ecological risk assessment for pesticides the factor applied can vary between 1 and 100 depending on the organisms being assessed and whether the toxicity endpoint is acute based on short-term lethal or immobility effects (LD/LC/ EC50) or chronic based on no observed effect generally from longer term tests and based on both lethal and relevant sub-lethal (growth reproduction) parameters.How conservative or worstcase the exposure assessment is can also affect the safety factor applied. It is reasonable to assume that these safety factors have a sound scientiÆc basis however this is not always the case. Safety factors are supposed to cover all the uncertainty in the initial risk assessment including inter- and intra-species sensitivities inter- and intra-laboratory variation and laboratory-to-Æeld extrapolations. These factors will vary according to the mechanism of toxicity of a chemical together with its physico-chemical properties and environmental fate.Thus tier 1 safety factors are generic whereas the parameters to do with the uncertainty are mostly chemical speciÆc. A good discussion of safety factors including the precautionary principle is presented by Chapman et al. `A critical evaluation of safety (uncertainty) factors for ecological risk assessment'.4 Safety factors have been chosen to be protective they have worked in the past and consequently are required to be large enough to ensure that those uses classiÆed as low risk at tier 1 indeed are and that there is only a low probability of incorrectly concluding that there will be no signiÆcant effects in the environment.At the risk of being repetitive it is not a requirement to achieve a toxicity/ exposure or an exposure/toxicity ratio such that these safety factors are achieved or triggers not exceeded although that can be an option through reÆning the exposure estimates. All too often papers are published which state that residue levels have exceeded ``safe'' levels or that acute/chronic effects may be expected based on some residue monitoring programme with reference to ``maximum permissible concentrations'' (MPC) or ``environmental quality standards'' (EQS) derived from the lowest toxicity value together with a generic safety factor. Often for pesticides additional information will be available to reÆne the risk assessment and whilst residue 106N J.Environ. Monit. 2000 2 levels should not be declared to be ``safe'' they might present a perfectly acceptable level of risk whilst exceeding MPC or EQS values. Alternative tier 1 approaches. With birds mammals bees earthworms and aquatic organisms the tier 1 procedures are relatively well established with both the standard tests and the tier 1 assessments well characterised and understood. The beneÆcial arthropod area is less well deÆned however tier 1 procedures are under review by an expert group5 and it is likely that some hazard quotient approach will be included at tier 1 in the near future.6 For assessing risk to beneÆcial arthropods at tier 1 the current procedure under 91/414/EEC requires indicator species to be tested in the laboratory with the results expressed as % effect at the Æeld rate.If the effect on these indicator species is less than a trigger value the chemical is deemed to be harmless to beneÆcials at the proposed rate the risk is acceptable and no further work is required. In this scheme the safety factor is built into the sensitivity of the tier 1 tests and the fact that they overestimate toxic effects in the Æeld. This is due to the acknowledged sensitivity of some of the tier 1 test species and the extreme exposure. For example tests may be done with direct application at Æeld rates onto glass plates or bare soil there is no crop interception for foliar applied products and no allowance for dilution in a three-dimensional environment.Another area where the risk assessment procedures are under development include the potential risk to non-target plants. Assessment for effects on plants at the Æeld rate and multiples or fractions of the Æeld rate can indicate the potential for effects on non-target plants. However a number of risk assessment questions remain. For example particularly for a herbicide what is a non-target plant? If any non-crop plants in the target area are considered pests and therefore legitimate targets what about those off-crop? What effect and exposure level is it appropriate for use at tier 1? What constitutes a signiÆcant effect on a plant? Whilst it is perhaps easy to see the signiÆcance of effects on seedling germination or emergence it is more difÆcult to interpret effects when the endpoint is a qualitative or quantitative effect on growth or vigour.ReÆning the risk assessment Tier 2 and beyond A tier 1 risk assessment clearly identiÆes uses and organisms that are low risk and provides a focus for further assessments. Given the type of chemicals being used their bio-efÆcacy and the wide range of non-target organisms studied tier 1 ecological risk assessments rarely result in an absence of areas requiring further consideration. Some applications do so on the basis of there being no or limited exposure of the chemical to non-target organisms through the registered use such as stored grain products or seed treatments.If tier 1 criteria are not met registration can be denied unless the risk assessment can be reÆned to demonstrate that the risk is acceptable. Under Directive 91/414/EEC Annex VI states that for aquatic organisms if tier 1 triggers are not met ``ºno authorisation shall be granted ººunless it is clearly established through an appropriate risk assessment that under Æeld conditions no unacceptable impact on the viability of exposed species occurs.'' Similar statements albeit with different trigger values are included for other organisms. Unacceptable impacts should not occur and from this it can be concluded that there is such a thing as an acceptable impact. It is unrealistic to expect that the use of pesticides would result in no impact and indeed against the background of the impact of agriculture and other environmental impacts natural and anthropogenic it is unrealistic to require this.Unlike the initial risk characterisation in many instances there are no standard procedures available at tier 2 and risk reÆnement is conducted very much on a case-by-case basis. Lately however there has been a lot of activity in this higher tier area. This has involved groups of academics regulators and industry getting together to share both their experience and their opinions of these higher tier assessments. The outputs from such groups range from guidance documents in study design and interpretation to tools such as exposure models which can be used conduct these assessments.These include (i) ECOFRAM Ecological Committee on FIFRA Risk Assessment Methods;7 (ii) FOCUS Forum for the Coordination of Pesticide Models and Their Use; (iii) HARAP Higher Tier Aquatic Risk Assessment for Pesticides;8 (iv) ESCORT2 European Standard Characteristics of Non-target Arthropod Regulatory Testing;5 (v) CLASSIC Community Level Aquatic System Studies±Interpretation Criteria.9 The above list is not meant to be comprehensive rather to give an indication of the amount of interest and effort in this area. ECOFRAM covers the entire area of terrestrial (using avian as an example but using principles which would be transferable to other vertebrates) and aquatic exposure and effects using four different subgroups.The other groups are more focused on the different areas described in their titles. Given the toxicity/exposure quotients used it seems obvious to split the discussion of how to reÆne the risk assessment into these two areas. I would like to discuss brieØy the approaches to reÆning the risk assessment starting with time a parameter which is relevant to both exposure and effects. The importance of time Although the studies conducted for the initial risk assessment clearly have a time component for example ``96 h LC50 to rainbow trout'' ``5 day dietary LC50 to bobwhite quail'' or ``21 day Daphnia magna life-cycle study'' the LC50/NOEC endpoint is used in the tier 1 assessment without reference to time.Generally it is not just exposure to a particular residue concentration that produces an effect the time of exposure is equally important in determining the dose to an organism. A combination of the level of exposure and the time of exposure is generally what produces the observed effect and effects manifest themselves after a particular time. Although not normally expressed an LT or ET50 (the time taken to kill or affect 50% of the test population) at a speciÆed concentration is a valid alternative endpoint to an LC or EC50 at a speciÆed time. The exposure value used at tier 1 is usually a peak concentration from a worst-case modelled scenario regardless of how long this level is maintained. From this maximum level of exposure the only way the exposure level is going to change is for it to reduce which will affect the potential dose an organism will receive and may modify the potential effects.Thus one of the Ærst considerations in reÆning the ecological risk assessment can be establishing the relationship between exposure/effect/time. Standard laboratory studies are designed to determine hazard and generally use maintained exposures. The difference between laboratory and Æeld exposures particularly for compounds that dissipate rapidly in the environment can mean that complex long-term studies have little or no value in risk assessment. A valuable initial source of information on the link between time of exposure and effects can be the simple tier 1 studies which will often contain more information than simply the data used to generate the reported endpoint.For example in a 96 h LC50 study assessments would normally be made after 24 48 and 72 h and these observations together with knowledge of the mechanism of toxicity or toxicokinetics can help establish the relationship between time of exposure and effects. Alternatively simple studies can be designed to deÆne the relationship between time of exposure and effect. Any number of PEC values can be developed for a compound in the environment maximum values time weighted average concentrations over any period of time or even annualised means. Establishing the relationship between time of exposure and time to effect can indicate which PEC is the most appropriate.ReÆning exposure Where possible reÆnements of exposure estimates are commonly used at higher tiers in risk assessment. There is a tendency to jump straight from tier 1 to chemical monitoring in the environment and generate ``real world'' data. This is sometimes because of the mistrust of computer models and a feeling that measuring real world concentrations will determine real world exposures. However this approach has its limitations. Care must be taken when trying to extrapolate from any monitored situation to other similar situations and there is always the problem with establishing where the monitored chemical came from. Monitoring is only a snapshot in time and monitoring programmes rarely give sufÆcient information about concentrations over time which is often necessary to determine exposure.These problems can be overcome with the use of higher tier modelling. However modelling has its own problems most notably that of model validation. These modelling procedures are best developed and validated within the US by the EPA for exposure in aquatic environments and in Europe the FOCUS Pesticides surface water group is working on similar approaches. The US EPA use an established generic tier 1 exposure model (GENEEC generic estimated environmental concentration) into which chemical speciÆc parameters and application data are entered. Entry into a non-target water body through run-off and spray drift produces a single worstcase entry rate concentration together with concentrations in the water over time.From this it is possible to move to tier 2 modelling (PRZM/EXAMS pesticide root zone model/experimental analysis modelling systems) which is speciÆc to a particular crop and region. This uses historic weather data and is able to produce concentrations over time typically 36 years allowing distributions of concentrations to be developed and probabilistic estimates of exposure to be made. ECOFRAM discusses this fully and the further developments at tiers 3 and 4 of modelling. At the highest tiers of exposure modelling the real world starts to be addressed. Earlier modelling makes assumptions about the proximity of water bodies to crops normally that they are adjacent.Using satellite imagery and geographic information systems (GIS) a picture can be built up of the area in which a chemical is or may be used. The proximity of water bodies to crops the different types of water bodies soil types slopes and other features can all be added in to determine where in relation to the chemical application are the water bodies we are trying to protect and what is the probability of exposure. The above procedures are relevant to aquatic environments but similar procedures can be used for terrestrial environments where exposure is often via residues on crops or potential food items. Tier 1 would assume a bird or mammal would consume food containing maximum residues throughout its lifetime.Data on the decline of residues on crops can be used to better characterise exposure and the assumption that 100% of an animals diet consists of treated food can begin to be challenged. ReÆning toxicity/effects There are a number of approaches available for reÆning the toxicity input into the risk assessment. Some of these simply use the quotient approach from tier 1 modifying toxicity values through generating toxicity data under more realistic conditions. For example instead of maintaining concentrations 107N J. Environ. Monit. 2000 2 Pesticides throughout a test the chemical could be introduced at the start of the study and allowed to dissipate and behave in a similar way to how it would in the environment.Calculating the toxicity based on the initial exposure will give a value which can be compared to modelled peak exposure concentrations giving a better estimate of the relationship between environmental exposures and effects. This approach is an alternative to reÆning exposure estimates according to time and calculating PECs other than initial maximum levels. ReÆning risk is all about reducing uncertainties and often the biggest uncertainty in ecological risk assessment is with regard to inter-species sensitivity. How representative of the organisms we are trying to protect in the Æeld are those individuals/species tested in the laboratory? We are usually not trying to protect individual organisms but populations and communities of different species.If we really wish to understand the potential risks from a chemical it is important to understand the sensitivities of different groups of organisms. Tier 1 safety factors are generic and if the laboratory species are particularly sensitive to a pesticide then they could be overprotective and any possible beneÆts from the chemical could be lost in needless mitigation. Indeed sensitivity in response to chemicals is one of the criteria used in selecting test organisms as surrogates both the life stage and the species. For a relatively small investment by testing the toxicity of a wide range of species much of the uncertainty can be removed from the assessment.What should be done with species sensitivity data and how should it be incorporated into the risk assessment? One approach is simply to acknowledge the reduction in the uncertainty and reduce the safety factor applied in the assessment. Without this there is no incentive to generate data beyond the limited base-set requirements to understand the potential risks. There are other ways of using these data. Understanding inter-species sensitivity is a major part of the aquatic risk assessment process being developed in ECOFRAM. This proposes the use of species sensitivity distributions in a probabilistic effects assessment. The principle behind this is the same as for other probabilistic assessments in that it assumes that the species tested are from a distribution of sensitivities in a universe of species.The available data are Ætted to a model to describe the 108N J. Environ. Monit. 2000 2 distribution of sensitivities that would be expected to occur. From this distribution exposure levels that would protect 90 95 99% or indeed any percentage of the species can be determined. Of course there are a number of concerns such as what level if any of species affected might be acceptable which species might be affected how might they be affected and are they economically ecologically or otherwise important. It is important to establish that the species are from the same sensitivity distribution and what is an appropriate model to Æt to that distribution.Despite these and other questions this approach has been used for many years in some form. Indeed it is an attractive idea that with this approach probabilistic estimates of effects can be combined with those of exposure to give joint probabilities. All of the effects assessments discussed so far have involved laboratory studies when our stated goal is protection of species and communities in the Æeld. Field studies still have a place in effects assessments although for the same reason as for chemical monitoring biological monitoring in the Æeld should not be undertaken lightly. Establishing cause and effect provision of adequate controls and applicability to other situations are always problematic.Controlled Æeld or semi-Æeld studies are however a very useful tool and may be the only tool currently available to assess recovery in affected systems. Recovery particularly time to recovery plays an important part in determining whether an effect is acceptable. Another tool under development is population modelling to determine whether populations are affected given certain input parameters (for example the sensitivity of different life stages to the chemical) and assumptions. These models allow the running of different scenarios however they are often limited by the availability of certain basic life-history data. Limitations of ecological risk assessments for agrochemicals and do they work? There are a number of areas in which the current risk assessment schemes are commonly considered deÆcient in certain areas.Some of these will be discussed albeit brieØy below. One criticism is that some groups of organisms are not represented. Quite clearly this is going to be the case at tier 1 where currently only 2 bird species honey bees 2 Æsh an aquatic invertebrate and a freshwater green alga together with some 10 non-target plant species might be studied for US registration. Risk to wild mammals uses data generated for the toxicology assessment. In addition for registration in Europe data are also required for non-target arthropods earthworms and the effects on soil micro-organisms. Protection of the environment and any exposed species is the goal and so it is easy to think of organisms that are not represented.Most commonly mentioned are reptiles and amphibians. There is particular concern over amphibians as there has undoubtedly been a decline in populations globally and pesticides have been cited as one of the contributing factors. It could be argued that vertebrates as a sub-phylum are actually fairly well represented with 3 classes being tested and there should perhaps be more legitimate concerns over invertebrates particularly aquatic invertebrates where the tier 1 requirement is a single freshwater crustacean zooplankton species (Daphnia). Whilst perhaps not so vocal as those advocating more protection for reptiles and amphibians there are people championing various different aquatic invertebrates.It would not seem unreasonable to add aquatic insects to the list of species tested at tier 1 if testing an insecticide; for example in the same way as herbicides trigger additional testing on plants. What is important is not ensuring that organisms are represented in the battery of tests rather that the battery of tests is protective of all the organisms of concern. For example there is little value in testing an amphibian or an aquatic protozoan if they are already protected by those surrogate species tested together with the applied safety factors at tier 1. Whether this is the case it is probably not possible to say currently and there is perhaps some basic research and information gathering required in this area.An ecological risk assessment for pesticides concentrates on direct impacts on exposed species and it would ideally demonstrate low risk of any effect. If direct effects are anticipated the potential for indirect effects does need to be considered and should be addressed by community level studies. One area not covered is the potential for indirect effects on wildlife as a result of the intended effect of the chemical such as removal of unwanted pests. A previous article cited the example of the decline in farmland birds associated with intensive farming as an indirect effect of pesticides.2 This is perhaps a good example which shows the importance of not just looking at ecological risk assessment in isolation but the importance of an integrated crop management (ICM) approach towards sustainable agriculture.Ecological risk assessment for pesticides concentrates on single chemicals whereas in the environment organisms exposure might be exposed to more than one pesticide or other chemical mixtures including natural toxins and a variety of other stressors. Concern is often expressed for the effects of complex mixtures on exposed organisms and that this is not considered in the standard ecological risk assessment procedures. It would be impossible to consider the potential range of mixtures that occur in the environment and therefore in risk assessment the pragmatic approach is to evaluate chemicals individually.Pesticides are tested in mixtures to representative organisms when formulated together in a product and so therefore a relatively large database exists on mixtures. Chemical mixtures may act in different ways they may act independently (in which there is no interaction) and this might be expected if the chemicals have differentmechanisms of toxicity. Additive effects might be expected if the chemicals have the same mode of action. Greatest concern is expressed that there could be a synergistic effect that is the effect of the chemicals together is greater than that predicted from the parts the converse of this is an antagonistic effect,which is not likely to be of concern in a risk assessment.Whilst it is easy to talk about these different possible effects of mixtures detecting them or differentiating between them in standard ecotoxicity tests is extremely difÆcult.Databases of toxicity tests with mixtures show independence or additivity to be the most common mechanism. Independence of action might be expected frommixtures of pesticides as formulations because chemicals with different modes of action and mechanisms of toxicity are generally mixed to produce different effects or hit different targets. It is less likely that chemicals with the same mode of action would be sprayed together and so whilst chemicals with similar modes of action might be found together in areas remote from the target (for example in a river estuary) it is less likely in the worst-case inÆeld or edge-of-Æeld scenarios (for example in a small headwater stream) straight after application which are those situations which drive the risk assessment.There have been reports of synergy from mixtures of pesticides. Perhaps the best example is the interactive effects between the ergosterol-biosynthesis inhibiting (EBI) fungicides and organophosphorus and pyrethroid insecticides in birds10 and bees.11 The biochemical basis for this is well established and it is readily demonstrated in the laboratory with controlled doses and the correct timing of the dose of the chemicals. It is unlikely that these laboratory effects would translate into effects in the Æeld nevertheless the possibility of synergism exists and whilst it is impossible to test for all situations consideration should be given to the possibility for such effects through the known biochemical modes of action.As a cautionary note a synergistic effect between fungicides used in bananas was initially publicised as the cause of Taura Syndrome a disease of cultured shrimps most common in Ecuador. Fungicides are used extensively in bananas and can be found in some water monitoring samples. A claim had been made that Taura Syndrome was generated in the laboratory from exposure to mixtures of chemicals. However these laboratory claims could not be substantiated and eventually a viral agent was identiÆed as the cause.12 If the process of ecological risk assessment for pesticides works why is there such concern for the environmental risks? There are a number of reasons for this.Perhaps the most important is that there are still a number of older compounds being used some of which (for example some organochlorine insecticides) are of environmental concern. This is particularly true in less developed countries where these chemicals are readily made cheap and because of this risk management decisions might be different than in some developed nations. Some compounds are not being supported through reregistration procedures in the US and the EU. It is important that the newer compounds with a better environmental proÆle are allowed to come through the registration process to replace older compounds registered when requirements were less comprehensive if they do not meet current standards.Programmes such as the US EPA Reduced Risk Registration Scheme explained by Rick Tinsworth13 will be valuable in ensuring this is the case. Requirements for registration are now extensive and whilst it is not possible to state that any pesticide use presents no risk it is possible to conclude that the use of compounds which have recently been through a regulatory ecological risk assessment are of low or negligible risk to the environment when used according to label recommendations. That is not to say Pesticides that we can be complacent and this where the incident reporting and monitoring discussed by Pepper and Carter2 in their earlier article are important.If it is true that current pesticides are of low risk to the environment why is it that Professor Ian Shaw can state ``The environmental impact of pesticides is a far sharper nail for the pesticides cofÆn than residues in food''?1 This statement might be true for some pesticides although perhaps not for the majority of those currently registered. What we have is a similar situation to that of the relevance of pesticide residues in our diet presented in excellent discussions by Ian Shaw and Naresh Atreya.14 There is a gap between public perception of risks and reality. A gap due to ineffective risk communication a gap which will not be easy to bridge but one which the industry together with other stakeholders will have to address as part of the progress towards a sustainable agricultural system. References 1 I. Shaw J. Environ. Monit. 2000 2 34N. 2 T. Pepper and A. Carter J. Environ. Monit. 2000 2 83N. 3 SETAC Pesticide Risk and Mitigation Final report of the Aquatic Risk Assessment and Mitigation Dialogue Group SETAC Foundation for Environmental Education Pensacola FL 1994. 4 P. M. Chapman et al. Environ. Toxicol. Chem. 1998 17 99. 5 M. P. CandolÆ et al. Guidance document on regulatory testing and risk assessment procedures for plant protection products with non-target arthropods SETAC Europe Brussels 2nd ed. in preparation. 6 P. J. Campbell et al. J. Pest. Sci. 2000 in press. 7 ECOFRAM US EPA http:// www.epa.gov/oppefed1/ecorisk/ 8 P. J. Campbell et al. Guidance document on higher-tier aquatic risk assessment for pesticides (HARAP) SETAC-Europe Brussels 1999. 9 W. Heger et al. Proceedings of the CLASSIC workshop (Community Level Aquatic System Studies - Interpretation Criteria) SETAC-Europe Brussels in preparation. 10 G. Johnston et al. Environ. Toxicol. Chem. 1994 4 621. 11 E. D. Pilling Aspects Appl. Biol. 1992 31 43. 12 K. W. Hasson et al. Diseases of Aquatic Organisms 1995 23 115. 13 R. Tinsworth J. Environ. Monit. 2000 2 63N. 14 N. Atreya J. Environ. Monit. 2000 2 53N Mick Hamer Ecological Risk Assessment Section Zeneca Agrochemicals Bracknell UK 109N J. Environ. Monit. 2000 2