Aquatic risks of pesticides, ecological protection goals, and common aims in european union legislation

Authors

  • Theo CM Brock,

    Corresponding author
    1. Alterra, Centre for Water and Climate, Wageningen University and Research Centre, P.O. Box 47, 6700 AA, Wageningen, The Netherlands
    • Alterra, Centre for Water and Climate, Wageningen University and Research Centre, P.O. Box 47, 6700 AA, Wageningen, The Netherlands
    Search for more papers by this author
  • Gertie HP Arts,

    1. Alterra, Centre for Water and Climate, Wageningen University and Research Centre, P.O. Box 47, 6700 AA, Wageningen, The Netherlands
    Search for more papers by this author
  • Lorraine Maltby,

    1. Department of Animal and Plant Sciences, University of Sheffield, Sheffield, United Kingdom
    Search for more papers by this author
  • Paul J Van den Brink

    1. Alterra, Centre for Water and Climate, Wageningen University and Research Centre, P.O. Box 47, 6700 AA, Wageningen, The Netherlands
    2. Department of Aquatic Ecology and Water Quality Management, Wageningen University, PO. Box 8080, 6700 DD Wageningen, The Netherlands
    Search for more papers by this author

Abstract

This discussion paper presents a framework for spatiotemporal differentiation in ecological protection goals to assess the risks of pesticides in surface waters. It also provides a proposal to harmonize the different scientific approaches for ecotoxicological effect assessment adopted in guidance documents that support different legislative directives in the European Union (Water Framework Directive and Uniform Principles). Decision schemes to derive maximum permissible concentrations in surface water are presented. These schemes are based on approaches recommended in regulatory guidance documents and are scientifically underpinned by critical review papers concerning the impact of pesticides on freshwater organisms and communities. Special attention is given to the approaches based on standard test species, species sensitivity distribution curves, and model ecosystem experiments. The decision schemes presented here may play a role in the “acceptability” debate and can be used as options in the process of communication between risk assessors and risk managers as well as between these risk experts and other stakeholders.

INTRODUCTION AND PROBLEM FORMULATION

Plant protection products (PPPs) are chemicals deliberately released into the environment to control pest species that harm agricultural crops. Aquatic ecosystems may be contaminated with PPPs as a result of spray-drift, leaching, runoff, and/or accidental spills, and because aquatic ecosystems contain species related to the target organisms of PPPs, undesirable side effects may occur. Therefore, governmental authorities have set criteria to protect aquatic life from pesticide stress. These criteria, however, often are debated because of the high economic consequences of too strict—and the high ecological consequences of too weak— environmental risk assessment procedures. Consequently, the ecological relevance of estimated risk levels is an important item in recent ecotoxicological research with PPPs (see, e.g., Crane and Giddings 2004; De Jong et al. 2005).

Most regulatory documents that deal with PPPs are based on policy goals that are ambiguous or difficult to define or measure. In the European Union (EU) Uniform Principles (UP; Annex VI to Directive 91/414/EEC; EU 1997) for example, it is stated that

  • The influence of PPPs on the environment should not be unacceptable (i.e., leaving room for interpretation of the degree of impact that is acceptable);

  • Member States shall ensure that use of PPPs does not have any long-term repercussions for the abundance and diversity of nontarget species (i.e., not excluding that shorter-term impacts followed by recovery may be acceptable); and

  • No authorization shall be granted unless it is scientifically demonstrated that under field conditions, no unacceptable effect on the environment occurs (i.e., suggesting a science-based risk assessment with a tiered approach).

In the EU Guidance Document on Aquatic Ecotoxicology (SANCO 2002a), a document in support of Directive 91/414/EEC, procedures are described for prospective higher-tier testing to evaluate the ecological risks of PPPs before their marketing. However, besides Directive 91/414/EEC, other European legislation exists that deals with the derivation of Environmental Quality Standards (EQS) of chemicals in surface waters, such as the Water Framework Directive (WFD). The WFD (2000/60/EC; EU 2000a) largely follows a retrospective approach and aims to achieve “good status” for European surface waters, which will be achieved, in part, by protecting the populations of water organisms from chemical stress. The WFD (Article 16) demands complete cessation or phasing out of the discharge of priority hazardous substances and the progressive reduction of priority substances. In the current list of priority (hazardous) substances, several PPPs are mentioned. When comparing the administration of PPPs within the context of Directive 91/414/EEC and the derivations of EQS values within the context of the WFD, it appears that different procedures may be used to derive acceptable concentrations (see next section). Ideally, the different scientific methods proposed in guidance documents in support of different EU Directives should not be in conflict with each other. It can even be argued that these methods need to be harmonized. It should be realized, however, that the underlying protection goals of the different directives have a different focus.

Although both EU Directives mentioned above aim to protect the aquatic environment from pesticide stress, their approaches are different. The aim of the present paper is to discuss the relation between ecological protection goals and endpoints that can be used to evaluate the risks of PPPs, with reference to the effect assessment for aquatic organisms and freshwater ecosystems. We investigate the potential to harmonize the different scientific approaches adopted in guidance documents that support both EU legislation directives (UP and WFD), and we propose decision schemes that incorporate these harmonized approaches to generate options for deriving acceptable PPP concentrations in surface waters. We have written this discussion paper in our role of scientists as informers of the acceptability debate. The aim of the present paper is not to discuss what is acceptable, because this is a societal question.

EU LEGISLATION AND RISK ASSESSMENT PROCEDURES FOR PESTICIDES

This section provides an overview regarding the current status of the risk assessment procedures under EU Directives 91/414/EEC and 2000/60/EC. A discussion of the science underlying these procedures is presented in the following sections.

Uniform Principles (Directive 91/414/EEC)

Risk assessment within the context of Directive 91/414/EEC has its main focus on relatively small bodies of water in the direct vicinity of the agricultural fields where PPPs are applied, follows a tiered approach, and clearly distinguishes the risks of short-term from those of long-term exposure (SANCO 2002a). To address the possible risks of PPPs to freshwater organisms, the lower-tier assessment requires

  • A Daphnia test to generate an acute 48-h EC50;

  • A Daphnia test to generate a chronic 21-d NOEC (if dissipation time 50% [DT50]water > 2 d or the exposure regime concerns repeated applications);

  • Two fish tests (a cold-water and a warm-water species) to generate acute 96-h LC50s;

  • A fish bioconcentration study (if Pow > 3 and dissipation time 90% [DT90]water-sediment > 10 d and/or the exposure regime concerns repeated applications);

  • A chronic 28- to 60-d NOEC for fish (if DT50water > 2 d or the exposure regime concerns repeated applications; the duration of the test is dependent on the values for DT90water-sediment and/or the bioconcentration factor [BCF]);

  • An algae toxicity test to generate a 72- to 120-h EC50 (2 taxa if the PPP is a herbicide);

  • A Lemna toxicity test to generate a 7-d EC50 (if the PPP is a herbicide);

  • A Chironomus riparius test to generate an acute 48-h EC 50 (if it concerns an insecticide to which aquatic insects are suspected to be more sensitive than Daphnia); and

  • A C. riparius test to generate a long-term, 28-d NOEC (if a chronic Daphnia test is required and in acute tests Chironomus is at least 10-fold more sensitive than Daphnia; this test also is required if risks for sediment-dwelling organisms are triggered).

In the lower-tier risk assessment, acute toxicity data are divided by maximum predicted exposure concentration (PECmax) values to generate acute toxicity exposure ratios (TERshort-term)- The TERshort-term should be at least 100 for the most sensitive animal species and at least 10 for the most sensitive plant species, which are numbers that can be considered as assessment factors (AFs). In addition, the chronic NOEC values are divided by either the PECmax or the predicted time-weighted average concentration (PECTWA) values to derive chronic TERs (TERlong-term). For the PECTWA values, usually a period of 21 and 28 d is taken into account to assess the risks to invertebrates and fish, respectively, but shorter time frames also may be chosen (e.g., dependent on time-to-effect information derived from toxicity tests). The PECmax is used to calculate the TERlong-term if it concerns a long-term static test (e.g., the 28-d C. riparius test in a water-sediment system in which the water is spiked) or if evidence suggests that in the long-term test, the time-to-effect period is short for relevant chronic endpoints like reproduction. In other cases, PECTWA values are used. The TERlong-term should be at least 10 for the most sensitive animal species; the lower-tier risks to plants are assessed by the TERshort-term method only.

In case a substance does not pass the trigger value, the “Unless” clauses described in the UP allow the possibility to refine the exposure and effect assessment. Higher-tier effect assessments may comprise laboratory toxicity tests with additional species, modified exposure studies with standard test species, population-level laboratory studies, indoor microcosm experiments, and outdoor mesocosm experiments. Possible methods for higher-tier risk assessment are described in the HARAP Guidance Document (Campbell et al. 1999). An overview of the current use of lower- and higher-tier risk assessment procedures for the administration of PPPs is given in the Guidance Document on Aquatic Ecotoxicology (SANCO 2002a).

The most frequently used higher-tier effect assessment procedures for the administration of PPPs are the species sensitivity distribution (SSD) approach and the model ecosystem approach. According to the HARAP Guidance Document (Campbell et al. 1999), the toxic mode of action should be taken into account when constructing SSDs to derive acceptable concentrations. If the lower-tier indicates that 1 species of the basic set is considerably more sensitive, an SSD should be constructed that is representative for the sensitive taxonomic group. According to the HARAP Guidance Document, toxicity data for at least 8 different species from the sensitive taxonomic group are recommended to construct SSDs. In case of herbicides, vascular plants and algae usually comprise the most sensitive group; in case of insecticides, arthropods usually are most sensitive. For fish, the HARAP Guidance Document recommends using a minimum of 5 toxicity data to construct SSDs specific for fish. This lower number of toxicity data is chosen for, among other reasons, animal welfare considerations. For PPPs with biocidal properties, such as several fungicides for which the basic set of standard test species shows a more or less equal sensitivity, toxicity data for at least 8 different taxonomic groups should be used. The HARAP Guidance Document, however, does not specify the taxonomic groups and level of taxonomic resolution when selecting toxicity data for this generic SSD. According to the Guidance Document on Aquatic Ecotoxicology (SANCO 2002a), the lower-tier AFs may be reduced if additional sensitive species are tested. A statistical extrapolation technique (e.g., the method described in Aldenberg and Jaworska 2000) also can be used to calculate the concentration at which a specified proportion of species (p) is expected to suffer direct toxic effects, referring to as the hazardous concentration (HC) to p% of the species (HCp). The SSD from which the HCp is derived can be based on either acute or chronic toxicity data. However, the smaller the number of data available for the calculation, the larger the confidence interval around the SSD (and the HCp). The HARAP Guidance Document (Campbell et al. 1999) mentions HC5 and HC10 values as possible assessment endpoints (Figure 1). However, in the Guidance Document on Aquatic Ecotoxicology (SANCO 2002a), no established guidance is currently provided regarding which HCp is appropriate for assessments under Directive 91/414/EEC

Figure Figure 1..

Graphical presentation of the species sensitivity distribution curve, its 95% confidence interval, and the derivation of the lower limit and median hazardous concentration for 5% of the species (HC5).

Several guidance documents have been published that deal with the conduct and interpretation of aquatic micro- and mesocosm tests (SETAC 1991; Giddings et al. 2002). In the Guidance Document on Aquatic Ecotoxicology (SANCO 2002a), effect classes are presented that aim to facilitate an objective interpretation of model ecosystem experiments for regulatory purposes (Table 1). When evaluating micro- and mesocosms, important assessment endpoints are the NOECpopulation, the NOECmicro/mesocosm (based on the most sensitive measurement endpoint/population), the NOECcommunity (usually based on multivariate analysis of community responses), and the no-observed-ecological-adverse-effect concentration (usually based on the recovery potential of sensitive populations). According to SANCO (2002a), these assessment endpoints can be used to derive an ecologically acceptable concentration. The ecologically acceptable concentration aims to take into account the spatiotemporal extrapolation of the results of the specific model ecosystem experiment, including information concerning the potential recovery of sensitive populations. The derivation of an ecologically acceptable concentration usually follows a case-by-case approach based on expert judgment.

Water Framework Directive (2000/60/EC)

The WFD has its main focus on relatively large bodies of water (>50 ha or 10 km2) and, among other goals, aims to produce a predicted-no-effect concentration (PNEC) for all types of toxic chemicals that pose risks to aquatic ecosystems, including PPPs (EU 2000a). The PNECs are used to define the good ecological status to be achieved both from short-term (peaks) and long-term exposures to these chemicals. The methodology for deriving PNECs for the freshwater and marine environment has been detailed in reports from the Fraunhofer Institute (Lepper 2002) and is based on the Technical Guidance Document of the European Commission (TGD; EU 2003). The TGD is a guideline for the risk assessment of new notified substances (Directive 93/67), existing substances (Regulation 1488/94), and biocides (Directive 98/8), but it does not include the risk assessment for PPPs.

Environmental risk assessment in the context of the WFD is based on 2 assumptions (EU 2003),

  • Ecosystem sensitivity depends on the most sensitive species, and

  • Protecting ecosystem structure protects community function.

According to the TGD (EU 2003), these assumptions are necessary to justify extrapolations from single-species toxicity tests to ecosystem effects. These extrapolations are made by applying AFs, the size of which (10–1,000) depends on the confidence with which a PNEC can be derived from the available data. This confidence increases if toxicity data are available for a larger number of trophic levels and taxonomic groups. Lower AFs are allowed with larger and more relevant data sets. The measured or predicted concentrations should be lower than the PNEC. Because for most modern PPPs a basic toxicity data set for at least 3 standard test species has to be provided within the context of 91/414/EEC, Lepper (2002) proposed to use an AF of 100 to the lowest acute toxicity value and an AF of 10 to the lowest chronic toxicity value to derive 1st-tier PNECs for pesticides with a complete dossier.

In case a substance does not pass the 1st-tier trigger value, further testing/information may lead to a revision of the exposure concentration and/or PNEC before a final conclusion regarding the acceptability of the risk is drawn. In the effect assessment, this further testing includes additional single-species studies to construct SSDs (see elaboration in EU 2003). According to the TGD, deriving a higher-tier PNEC by means of the SSD method requires preferably more than 15 but at least 10 chronic NOECs for different species covering at least 8 taxonomic groups. For comparable data regarding the same ecotoxicological endpoint for a particular species, the geometric mean should be used as input. Lepper (2002) recommends the median estimate of the 5th percentile value (i.e., HC5) of the SSD for calculation of the PNEC (Figure 1). The AF that reflects the further uncertainties identified should be between 5 and 1 (to be judged on a case-by-case basis). Criteria are the quality of the database, the endpoints covered, the diversity of the taxonomic groups, the statistical uncertainties around the HC5 estimate, the outcome of comparison between field/mesocosm studies, and the HC5 itself.

Information from micro- or mesocosm studies also are acceptable if the test design meets quality requirements. However, evaluation criteria of micro- and mesocosm studies are not elaborated in the TGD (EU 2003) or in Lepper (2002). It is only stated that micro- and mesocosm studies enable more precise AFs to be calculated and applied and that the size of the AF is decided on a case-by-case basis.

Table Table 1.. Classes of pesticide effects to evaluate treatment-related responses observed in aquatic micro/mesocosm studiesa
Effect classbDescriptionCriteria
  1. a Brock, Lahr, et al. (2000); Brock, Van Wijngaarden, et al. (2000); and SANCO (2002a).

  2. b Effects classes aim to facilitate an objective interpretation of model ecosystem experiments performed for regulatory purposes.

1Effects could not be demonstrated (NOECmicro/mesocosm)No (statistically significant) effects observed as a result of the treatment.
  Observed differences between treatment and controls show no clear causal relationship.
2Slight effectsEffects reported as “slight,” “transient,” or other similar descriptions
  Short-term and/or quantitatively restricted response of 1 or a few sensitive endpoints, and only observed at individual samplings.
3 (A or B)Pronounced short-term effectsClear response of sensitive endpoints, but full recovery of affected endpoints within 8 weeks after the 1st (effect class 3A) or last (effect class 3B) application (from recovery in treated test systems).
  Effects reported as “temporary effects on several sensitive species,” “temporary effects on less sensitive species/endpoints,” or other similar descriptions.
  Effects observed at some subsequent sampling instances.
4Pronounced effects in short-term studyClear effects (e.g., large reductions in densities of sensitive species) observed, but the study is too short to demonstrate complete recovery within 8 weeks after the (last) application.
5 (A or B)Pronounced long-term effects in long-term study, but full recovery not observed within 8 weeks after last applicationClear response of sensitive endpoints and recovery time longer than 8 weeks after the last application, but full recovery is demonstrated to occur in the year of application (effect class 5A) or cannot be demonstrated before termination of the experiment or before the start of the winter period (effect class 5B).
  Effects reported as “long-term effects followed by recovery on several sensitive and less sensitive species/endpoints” or other similar descriptions.
  Effects observed at various subsequent samplings, but not at the end of the study period.

Within the context of the WFD, the derived PNEC values, either by means of the lower-tier or refined risk assessment procedure, are the EQS values. Among other goals, they aim to provide sufficient protection to populations of aquatic organisms from exposure to priority (hazardous) substances. In addition, these EQS values aim to safeguard a healthy use of water resources by humans. Within the WFD, a distinction is made between the annual average EQS (AA-EQS) and the maximum allowable concentration EQS (MAC-EQS) values. The AA-EQS aims to ensure protection against long-term exposure and may be compared with the annual average concentration of measurements for a pollutant in the water of the ecosystem of concern. The current proposal is to calculate the annual average concentration as the arithmetic mean of the concentrations measured in the samples taken. The MAC-EQS aims to ensure protection against short-term exposure. Peak concentrations in water should never exceed the MAC-EQS in the ecosystems of concern. The EU aims to list and publish AA-EQS and MAC-EQS values for all current priority (hazardous) substances. The procedures to derive these EQS values are described in a proposal by Lepper (2002), but the final decisions are only taken after consultation of all EU Member States and important stakeholder groups.

Summary of main differences in ecological risk assessment procedures between WFD and UP

The main differences in ecological risk assessment procedures between the WFD and the UP can be summarized as follows:

  • The risk assessment within the context of the WFD has its main focus on relatively large bodies of water (> 50 ha or 10 km2), whereas that of the UP has its focus on relatively small bodies of water (e.g., drainage ditches, small streams, and ponds) in the direct vicinity of agricultural fields.

  • The WFD addresses the potential ecological risks of all toxic chemicals (including PPPs), whereas the UP deals with PPPs only.

  • The WFD follows a general approach and largely makes use of retrospective exposure assessments (monitoring data). The UP follows a prospective approach, and the risk assessment is based on specific uses of PPPs in certain crops.

  • Because for most modern PPPs a basic toxicity data set for at least 3 standard test species has to be provided within the context of 91/414/EEC, Lepper et al. (2002) propose to use an AF of 100 to the lowest acute toxicity value and an AF of 10 to the lowest chronic toxicity value to derive 1st-tier PNECs. This implies that the 1st-tier risk assessment procedure based on laboratory toxicity tests of standard test species often is similar between UP and WFD, except when standard algae or Lemna are the most sensitive test species.

  • Under the umbrella of the UP, a larger number of different methods for higher-tier risk assessment are described and used (and often experimentally validated) compared with the refined risk assessment procedures in the context of the WFD.

  • Both the WFD and the UP allow the SSD approach; however, different criteria are used in the construction of the SSD (number of taxa and taxonomic groups).

  • Documents in support of the WFD provide little guidance regarding how to derive permissible concentrations for short-term exposure (focus on chronic risks) and how to use micro- and mesocosm experiments in the derivation of norm concentrations, in contrast to guidance documents in support of the UP.

  • In the higher-tier risk assessment of PPPs in the context of the UP, the potential for recovery of sensitive populations is taken into account when deriving acceptable concentrations. The PNECs in the context of the WFD aim always to protect sensitive populations; consequently, recovery is not considered.

RECENT SCIENTIFIC DEVELOPMENTS IN THE TIERED EFFECT ASSESSMENT PROCEDURE

The tiered approach

As described above, effect assessment procedures within the context of EU Directive 91/414/EEC may be based on a tiered approach. Important assumptions of the tiered approach are

  • Lower tiers are more strict than higher tiers, and higher tiers are more realistic than lower tiers. (This hierarchical approach implies that the highest tier should be closest to the requirements as set by the adopted protection goal);

  • Lower tiers require less effort than higher tiers;

  • The information of lower tiers is used to design the (experimental) setup of higher tiers;

  • Each tier acts as a sieve and has sufficient distinctive power; and

  • The tiered approach includes a willingness to accept any relevant information.

In the effect assessment procedure of PPPs, the 1st tier is based on the standard test species-AF approach. A frequently used intermediate tier is based on the SSD approach (Forbes and Calow 2002; Posthuma et al. 2002; Van den Brink et al. 2002). The model ecosystem approach (Hill, Heimback, et al. 1994; Brock, Lahr, and Van den Brink 2000; Brock, Van Wijngaarden, and Ven Geest 2000; Caquet et al. 2000; Van Wijngaarden, Brock, and Van den Brink 2005) often is used as the highest tier. These 3 approaches will be discussed in greater detail below, although we realize that other tiers (e.g., approaches based on population level studies, food-web modeling, and field observations) are possible.

The standard test species approach

A quantitative analysis of the Ist-tier trigger values with threshold concentrations for direct toxic effects (effect classes 1 and 2 in Table 1) in aquatic micro- and mesocosm experiments suggests that the standard test species approach (applying fixed AFs to toxicity data of standard test species) is sufficiently conservative for most pesticides (see, e.g., the quotients effect class 1 and 2/0.01 × EC50 in Figure 2). Van Wijngaarden, Brock, and Van den Brink (2005) demonstrated for organophosphorous, carbamate, and pyrethroid insecticides that the effect class 1 and 2 concentration (based on the most sensitive measurement endpoint in a micro- and mesocosm test) usually is a factor of 10-fold or more higher than 0.01 × (48-h EC50 of Daphnia magna), even when the insecticide is applied repeatedly. The 1st-tier trigger also seems to be sufficiently protective for photosynthesis-inhibiting herbicides. Brock, Lahr, and Van den Brink (2000) showed that only slight effects on the most sensitive measurement endpoints may be expected at exposure concentrations of less than 0.1 toxic unit (i.e., 0.1× EC50 of the standard test alga), even under long-term exposure regimes. In addition, preliminary data indicate that the 1st-tier trigger concentration for fungicides usually also is lower than the effect class 1 and 2 concentration observed in micro- and mesocosm studies (L Maltby, Department of Animal and Plant Sciences, University of Sheffield, Sheffield, UK, unpublished data).

The SSD approach

As shown in Figure 1, an SSD is a statistical distribution estimated from a sample of toxicity data and visualized as a cumulative distribution function. In the scientific literature dealing with the SSD approach, the most frequently calculated HC from the SSD curve is the HC5 (Posthuma et al. 2002; Maltby et al. 2005). One may question the geographical extrapolation of SSD curves and the derived HC5 values. However, the analyses conducted by Maltby et al. (2005) suggest that although the composition of freshwater communities varies across biogeographical regions, climatic zones, and habitat types, the distribution of species sensitivities does not vary markedly. For example, no evidence indicates that lotic arthropod assemblages generally are more sensitive than lentic arthropod assemblages to insecticides (Maltby et al. 2005). Schroer et al. (2004) demonstrated that SSDs and HC5 values derived from single-species tests with the insecticide lambda-cyhalothrin and freshwater arthropods were very similar between independent studies performed in The Netherlands and the United Kingdom.

Figure Figure 2..

Geometric means and 95% confidence intervals of quotients obtained by dividing effect class 1 and 2 concentrations observed in micro- and mesocosms by either the L(E)C50/100 value for the most sensitive standard test species including algae (EC50 × 0.01), the lower limit HC5 (LL HC5), or the median HC5 (HC5). The LL HC5 values are based on species sensitivity distributions (SSDs) constructed with acute toxicity data. A distinction is made between micro- and mesocosm experiments treated once and repeatedly with the pesticide. Basic data for these quotients are reported in Tables 2 and 3.

Recently, calibration of the relationship between HC5 values and results of semifield experiments for insecticides and herbicides was provided by Maltby et al. (2005) and by Van den Brink et al. (2006). In their evaluations, both the median HC5 value and the lower-limit HC5 estimate were used (Figures 1 and 2). The key point is the focus on a specific taxonomic group in which the assessment concerns a PPP with a specific toxic mode of action. Maltby et al (2005) demonstrated for insecticides and aquatic arthropods that the lower-limit HC5 estimate derived using acute toxicity data provides a conservative estimate of the ecological threshold concentration in micro- and mesocosms not only for single but also for multiple and continuous application of the insecticide (see Figure 2 and Tables 2 and 3 for a more quantitative presentation). The median HC5 estimate based on acute toxicity for freshwater arthropods generally is protective of single insecticide applications and of continuous and multiple applications when a safety factor of at least 5 is applied (Maltby et al. 2005).

Van den Brink et al. (2006) showed that for herbicides and primary producers, the lower limit of the acute HC5 and the median value of the chronic HC5 were protective of adverse effects in aquatic micro- and mesocosms even under a long-term exposure regime (Figure 2 and Tables 2 and 3). The median HC5 estimate based on acute toxicity data of herbicides was protective of adverse effects in aquatic micro- and mesocosms when a short-term exposure regime (pulse application in flow-through system; single application of a nonpersistent [DT50water < 10 d] herbicide in stagnant test system) was studied (Van den Brink et al. 2006).

Current research aims to compare the results of the SSD approach and the model ecosystem approach in the effect assessment for fungicides (L Maltby, Department of Animal and Plant Sciences, University of Sheffield, Sheffield, UK, unpublished data). Preliminary investigations indicate that the lower limit of the HC5 based on acute toxicity data for the relevant taxonomic group (which may be a wider array of species tested because of the general biocidal properties of certain fungicides) results in a concentration that generally is lower than the effect class 1 and 2 concentration in a micro- and mesocosm test when treated repeatedly with the same fungicide (Table 3). The median HC5 estimate based on acute toxicity data was protective of single fungicide application in 2 of the 4 cases (Table 2). The persistent compound triphenyltin acetate bioaccumulates in the food chain, which may explain this exception (Roessink et al. 2006).

The summary data for all pesticides tested (Figure 2) indicate with high certainty that the median acute HC5 value is lower than the effect class 2 concentration observed in a micro- and mesocosm experiment treated once with the pesticide. The corresponding lower-limit HC5 value is, with high certainty, lower than reported effect class 1 and effect class 2 concentrations, even when the pesticide is applied repeatedly.

The model ecosystem approach

Given the natural variability in the structure and function of freshwater communities, it is reasonable to question the geographical extrapolation of results of model ecosystem experiments with PPPs (Crane and Giddings 2004). Because most model ecosystems enclose parts of—or have been seeded with components of— natural communities, the geographical location of micro- and mesocosms will determine their species composition and, potentially, their sensitivity.

Within the context of the derivation of norm concentrations, the question at stake is how unique such test systems are with respect to their ecological threshold levels for toxic effects. It appears from the model ecosystem experiments performed with the nonpersistent insecticides chlorpyrifos (single applications) and lambda-cyhalothrin (repeated applications) that concentrations specifically in the range of “no” to “slight and transient” (effect class 1 and effect class 2, respectively) are remarkably consistent (Tables 4 and 5). Also, the review by Van Wijngaarden, Brock, and Van den Brink (2005) cites similar effect class 1 and 2 concentrations for different model ecosystems treated once with azinphosmethyl or repeatedly with the fast-dissipating esfenvalerate. Furthermore, lake enclosure studies exploring effects of a single application of pentachlorophenol to plankton communities in spring, summer, autumn, and winter indicated that threshold levels for effects (effect class 1) varied little with season (24–54 μg/L; Willis et al. 2004).

Table Table 2.. First-tier lowest L(E)C50 value, median and lower limit values of hazardous concentration to 5% of the species (HC5) based on acute toxicity data for the most sensitive taxonomic group, and effect class 1, 2 and 3 concentrations for the most sensitive measurement endpoint as observed in micro- and mesocosms simulating ponds or recirculating streams that were treated once with the pesticidea
 Single speciesbSSD approachcModel ecosystem approach
CompoundLowest L(E)C50 (μg/L)Median HC5 (μg/L)Lower-limit HC5 (μg/L)Effect class 1 (μg/L)Effect class 2 (μg/L)Effect class 3 (μg/L)
  1. a If for the same microcosm or mesocosm study more effect class 1 or effect class 2 concentrations are available, the highest value is reported. Effect classes are expressed in terms of nominal peak concentrations. If for the same compound more than 1 study is available, L(E)C50 data are reported as geometric means, and micro- and mesocosm data are reported as the geometric mean value and the range. Data are based on Maltby et al. (2005), Van den Brink et al. (2006) and L Maltby, Department of Animal and Plant Sciences, University of Sheffield, Sheffield, UK, unpublished data. The fungicide data are preliminary.

  2. b Cv = 72-h Chlorella vulgaris; Dm = 48-h Daphnia magna; Om = 96-h Onchorhyncus mykiss; Sc = 72-h Selenastrum capricornutum; Ss = 72-h Scenedesmus subspicatus.

  3. c Species sensitivity distributions (SSDs) are based on the following: AR = arthropod species; AS = all species; M = macrophyte species; NV = nonvertebrate species; PP= primary producers (algae and macrophytes); VE = vertebrates.

Insecticides
 Azinphos-methyl1.3 (Dm)0.09 (AR)0.020.20.721.0
 Carbaryl6.7 (Dm)2.94 (AR)1.37.0<20
 Carbofuran41 (Dm)0.22 (AR)0.03525
 Chlorpyrifos0.8 (Dm)0.08 (AR)0.050.1 (0.1–0.1)0.30.84 (0.5–1)
 Diflubenzuron6.7 (Dm)0.05 (AR)0.00050.3<0.7
 Fenitrothion4.7 (Dm)0.44 (AR)0,91.118.7
 Fenvalerate0.45 (Dm)0.043 (AR)0.0070.010.05
 Methoxychlor9.1 (Dm)0.37 (AR)0.1435<20
Herbicides
 2,4-Dichlorophenoxyacetic acid3,793 (Dm)71 (M)7.110<100
 Atrazine55 (Ss)13 (PP)5.856.7 (2–20)<30
 Diuron1 5 (Sc, Ss)12.0 (PP)7.63<30
 Metamitron852 (Sc)667 (PP)22628011204480
 Metribuzin12.3 (Cv)7.4 (PP)4.05.61856
 Pendimethalin8.8 (Sc)2.0 (PP)0.20.231.14.9
 Simazine129 (Ss)52 (PP)1871 (50–100)100150
Fungicides
 Azoxystrobin5.5 (Dm)37 (AR)5.4 1010 to <30
 Carbendazim391 (Dm)24.6 (NV)2.23.030
 Pentachlorophenol148 (Om)35.0 (PP)14.832.5 (24–54)<54 (36–81)
 Triphenyltin acetate6(Sc)0.95 (NV)0.380.31

Whether the robustness in ecological threshold concentrations also is the case for more or less constant, chronic exposure regimes needs to be investigated. Data available for the herbicide atrazine suggest a larger variability in class 1 and 2 effects between experiments (Table 6). This, however, may be caused by the fact that it concerns a persistent photosynthesis inhibitor that has been used on a large scale and for a long time. Adaptation of the community to atrazine may have influenced the treatment-related responses that were observed. Also, differences in light conditions between experiments may have played a role. According to Guasch and Sabater (1998), inhibition of photosynthesis by atrazine is comparable to limiting light conditions, and photosystem II inhibition is lower for plants that are adapted to low-light conditions.

Table Table 3.. First-tier lowest L(E)C50 value, median and lower limit values of hazardous concentration to 5% of the species (HC5) based on acute toxicity data for the relevant taxonomic group, and effect class 1, 2 and 3 concentrations for the most sensitive measurement endpoint as observed in micro- and mesocosms simulating lotic and lentic freshwater ecosystems that were treated repeatedly with the pesticidea
 Single speciesbSSD approachcModel ecosystem approach
CompoundLowest L(E)C50 (μg/L)Median HC5 (μg/L)Lower-limit HC5 (μg/L)Effect class 1 (μg/L)Effect class 2 (μg/L)Effect class 3 (μg/L)
  1. a If for the same microcosm or mesocosm study more effect class 1 or effect class 2 concentrations are available, the highest value is reported. Effect classes are expressed in terms of nominal treatment levels. If for the same compound more than 1 study is available, L(E)C50 data are reported as geometric means, and micro- and mesocosm data are reported as the geometric mean value and the range. Data are based on Maltby et al. (2005), Van den Brink et al. (2006), and L Maltby, Department of Animal and Plant Sciences, University of Sheffield, Sheffield, UK, unpublished data. The fungicide data presented are preliminary.

  2. b Dm = 48-h Daphnia magna; Lm = 7-d Lemna minor; Om = 96-h Onchorhyncus mykiss; Sc = 72-h Selenastrum capricornutum; Ss = 72-h Scenedesmus subspicatus.

  3. c Species sensitivity distributions (SSDs) are based on the following: AR = arthropod species; AS = all species; AS-M+ all species except macrophytes; M = macrophyte species; NV = nonvertebrate species; PP = primary producers (algae and macrophytes); VE = vertebrates.

Insecticides
 Azinphos-methyl1.3 (Dm)0.09 (AR)0.020.220.95
 Diflubenzuron6.7 (Dm)005 (AR)0.00050.1<1
 Fenvalerate0.45 (Dm)0.043 (AR)0.0070.01<0.1
 Lambda-cyhalothrin0.25 (Dm)0.003 (AR)0.0010.0033 (0.003–0.004)0.01 (0.01–0.01)0.027 (0.016–0.05)
 Lindane28 (Om)1.7 (AR)0.740.251.0<4
 Parathion-ethyl1.4 (Dm)0.24 (AR)0.12  0.5
Herbicides
 Atrazine55 (Ss)1 3 (PP)5.87.7 (5–14)15.8 (10–25)<14 to 50
 Linuron7 (Lm)5.8 (PP)0.740.558.7 (5–15)
Fungicides
 Carbendazim391 (Dm)24.6 (NV)2.23.3<33
 Chlorothalonil31 (Om)8.8 (NV)3.01030
 Fluazinam63 (Om)6.9 (AS)2.011030
 Kresoxim-methyl63 (Sc)6.4 (NV)0.713.3<66.6
 Mancozeb910 (Dm)24.8 (NV)0.61032
 Picoxystrobin21 (Dm)15.0 (NV)4.4624.0
 Tolylfluanid27 (Om)11.5 (AS-M)2.31021
 Trifloxistrobin5.3 (Ss)5.0 (NV)1.46.76.7–21.6

Studies of other pesticides that compared threshold concentrations for direct toxic effects as a result of chronic exposure between model ecosystem experiments are not known to us. However, evidence suggests that threshold concentrations for chronic exposure to the surfactant dodecyl trimethyl ammonium chloride do not differ much between model stream experiments (range of NOECecosystem values, 180–300 μg/L; n = 5; Versteeg et al. 1999). Similarly, only a 3-fold difference in threshold concentrations is reported for long-term exposure to copper (hardness adjusted to 50 mg/L as CaCC>3) derived from 1 lentic mesocosm and 6 artificial stream studies conducted in the United States and Europe (Versteeg et al. 1999).

The data presented above illustrate that the spatiotemporal extrapolation of ecological threshold concentrations (based on the most sensitive endpoints in micro- and mesocosms) because of exposure to nonpersistent pesticides seems to be possible with relatively low uncertainty. Keep in mind, however, that even standardized laboratory tests performed with D. magna and the same chemical may easily vary a factor of 2 or 3 within and between laboratories (Baird et al. 1989). A larger variability in responses generally is observed for different types of pesticides studied at exposure concentrations well above their threshold concentration for direct toxic effects. For example, in the indoor plankton-dominated microcosms treated with chlorpyrifos (Van Wijngaarden, Brock, and Douglas 2005), some clear, indirect effects were observed after treatment with 1 μg/L (i.e., algal blooms and increase in Rotifera), but no indirect responses could be detected in the more complex outdoor experimental ditches after treatment with 0.9 and 6 μg/L (Van den Brink et al. 1996). Apparently, when evaluating indirect responses of chemical stress, laboratory microcosms sometimes show exaggerated responses to high concentrations of toxicants. In more structurally complex outdoor test systems, however, a greater number of feedback mechanisms may be available that dampen indirect effects. Also, Roessink et al. (2005) concluded that at higher concentrations of lambda-cyhalothrin, the magnitude and duration of effects differed between plankton-dominated enclosures (with a community characterized by short-lived organisms) and structurally more complex, macrophyte-dominated enclosures. This indicates that once clear effects are caused by exposure to non-persistent pesticides, the variability in the rate of recovery may be relatively high between test systems (see, e.g., effect class 3 in Tables 4 and 5)

Table Table 4.. Effect class concentrations of the most sensitive measurement endpoint observed in model ecosystem experiments that studied the impact of short-term exposure (single pulse in experimental streams or single application in letic test systems) to the insecticide chlorpyrifosa
ExperimentEffect class 1Effect class 2Effect class 3Effect class 4Effect class 5Reference (type of test system)
  1. a For an explanation of effect classes, see Table 1.

  2. b Recovery is relatively fast because of constant input of propagules in experimental streams.

a0.1 μg/L(5.0 μg/L)bPusey et al. (1994) (experimental streams)
b0.1 μg/L0.3 μg/L1.0 μg/L3.0 μg/LBiever et al. (1994) (outdoor microcosms)
c0.1 μg/L0.9 μg/LVan den Brink et al. (1996) (experimental ditches)
d0.1 μg/L1.0 μg/L10 μg/LVan Wijngaarden, Brock, and Van den Brink (2005) (lab microcosms; mesotrophic; ∼16°C)
e0.1 μg/L1.0 μg/LVan Wijngaarden, Brock, and Van den Brink (2005) (lab microcosms; mesotrophic; ∼26°C)
f0.1 μg/L1.0 μg/LVan Wijngaarden, Brock, and Van den Brink 2005) (lab microcosms; eutrophic; ∼26°C)
g0.5 μg/L6.3 μg/LSiefert et al. (1989) (pond enclosures)

Can aquatic micro/mesocosm experiments be used to predict effects in the field?

Experimental aquatic ecosystems have become widely used tools in ecotoxicology, because they allow a greater degree of control, replication, and repeatability than can be achieved in natural ecosystems. The test systems in use vary from small indoor microcosms to large and complex outdoor experimental ecosystems. However, natural freshwater systems also may vary considerably in size and ecological complexity. Nevertheless, differences in model ecosystem size and complexity are reported to have a profound effect on the enclosed community (Caquet et al. 2000). Important aspects of ecosystem and community functions may be controlled by key organisms that are too large or too mobile to be confined in experiments smaller than the ecosystem of concern, such as large, predatory fish. In addition, problems in interpreting micro- and mesocosm experiments may be caused by inadequate or erroneous scaling of sediment-water interactions and potential artefacts associated with containerization (wall effects and water renewal times). Furthermore, large freshwater ecosystems usually are characterized by a diversity of habitats differing in abiotic and biotic properties (e.g., the pelagic or the littoral zone of lakes), whereas most artificial aquatic ecosystems usually simulate one of these habitats only.

Belanger (1997) analyzed data from more than 150 studies using model stream ecosystems ranging in size from 0.2 to 540 m in length, 0.05 to 4.3 m in width, and 1.5 to 8 × 105 L in volume. He concluded that although larger systems could be sampled more intensively and were more likely to contain fish, no relationship existed between test system size and species richness of invertebrate, algal, or protozoan assemblages. Few studies have compared assemblages in model streams and natural streams, but those that have indicate that assemblages in model streams are representative of the natural streams from which they are derived (Belanger et al. 1995; Wong et al. 2004). Model ecosystems that simulate lentic aquatic ecosystems, however, usually contain species that are characteristic of the deeper parts of freshwater ponds and often lack the species assemblages that are typical of littoral zones (Williams et al. 2002). However, it appears from several model ecosystem experiments with insecticides that threshold levels for effects may be very similar between lentic test systems that differ considerably in complexity, at least when they contain representatives of sensitive taxonomic groups (in this case, arthropod populations; Tables 4 and 5).

Table Table 5.. Effect class concentrations of the most sensitive measurement endpoint observed in model ecosystem experiments that studied the impact of repeated short-term exposure (from multiple applications) to the insecticide lambda-cyhalothrina
ExperimentEffect class 1Effect class 2Effect class 3Effect class 4Effect class 5Reference (type of test system; number of applications)
  1. a For an explanation of effect classes, see Table 1. In experiment a, the application comprised both spray-drift and runoff applications of lambda-cyhalothrin. In experiments b and c, fungicides and herbicides also were applied, but in this evaluation, all effects were assigned to the insecticide lambda-cyhalothrin.

  2. b Experiment was characterized by both spray-drift and runoff applications. As exposure concentration, the median value of spray-drift and runoff concentrations was used.

a2.7 ng/Lb27 ng/LbHill, Runnalls, et al. (1994) (experimental ponds; 6–12)
b4.0 ng/L16 ng/L85 ng/LArts et al. (2006) (experimental ditches; 2)
c10 ng/L25 ng/LVan Wijngaarden et al. (2004) (laboratory microcosms; 5)
d10 ng/L25 ng/LRoessink et al. (2005) (ditch enclosures, plankton-dominated, spring; 3)
e10 ng/L50 ng/LRoessink et al. (2005) (ditch enclosures, macrophyte-dominated, spring; 3)
f10 ng/L25 ng/L50 ng/LVan Wijngaarden et al. (2006) (ditch enclosures, macrophyte-dominated, summer; 3)
g17 ng/LFarmer et al. (1995) (pond mesocosms; 4)

WHAT CONSTITUTES A SUSTAINABLE FRESHWATER ECOSYSTEM?

One of the key steps in problem formulation is the statement of what is to be protected and which assessment endpoints to use as operational measure of the protection goal. When assessing the aquatic risks of the agricultural use of PPPs, it is important to have scientifically sound ideas of what constitutes an important ecological effect of these chemicals in surface waters and what constitutes a sustainable freshwater ecosystem. This scientific insight is important input for the societal debate regarding which protection goals to adopt and what level of risk to consider as acceptable.

To “sustain” is to hold, keep alive—literally, “able to last.” Sustainability of freshwater ecosystems may concern not only their ecological properties but also their economic and social functions. In surface waters adjacent to agricultural fields, it is not realistic to keep the freshwater communities in a condition comparable to that in nature reserves. Nevertheless, we recognize the multifunctional character of freshwater ecosystems in the agricultural landscape, including their ecological functions. The interest in ecological risk assessment is to contribute to the sustainable management of freshwater ecosystems. However, sustainable freshwater ecosystems may include both pristine and disturbed multifunctional ecosystems (e.g., agroecosystems), preventing a simple definition of a sustainable ecosystem.

Within the context of the sustainability of freshwater communities and associated ecological functions and services are 3 general categories of effects of PPPs in the environment. These relate to ecosystem structure, function, and aesthetic as well as economic values to humans (see, e.g., Calow 1998; Brock and Ratte 2002). The structure of an ecosystem is a combination of which and how many organisms are present, whereas function relates to what the organisms do in the ecosystem. Changes in structure generally are expressed in terms of overall species richness and densities as well as population densities of key and indicator species. Changes in ecosystem functioning usually are expressed as changes in the rate of biogeochemical cycles (e.g., changes in primary productivity, processing of nutrients, and mineralization of organic matter). Changes in perceived aesthetic and economic values usually concern the real and perceived benefits and values of the ecosystem and its organisms to humans.

Table Table 6.. Effect class concentrations of the most sensitive measurement endpoint observed in model ecosystem experiments that studied the impact of long-term exposure to the herbicide atrazinea
ExperimentEffect class 1Effect class 2Effect class 3Effect class 4Effect class 5Reference (type of test system)
  1. a Atrazine is very persistent in water (DT50 > 90 d) so that a single application in closed test systems already results in long-term exposure. For an explanation of effect classes, see Table 1. DT50 = dissipation time 50%.

a2 μg/L30 μg/LSeguin et al. (2001) (experimental ponds)
b5 μg/LVan den Brink et al. (1995) (lab microcosms)
c 5 μg/LGruessner and Watzin (1996) (recirculating streams)
d5 μg/L10 μg/L22 μg/LJüttner et al. (1995) (pond enclosures)
e 10 μg/L100 μg/LJohnson (1986) (lab microcosms)
f5 μg/L50 μg/L100 μg/LBrockway et al. (1984) (lab microcosms)
g10 μg/L32 μg/LPratt et al. (1988) (lab microcosms)
h 10 μg/LKosinsky (1984); Kosinsky and Merkle (1984) (recirculating streams)
i14 μg/L25 μg/L80 μg/LNyström et al. (2000) (pond enclosures)
j14 μg/LMuños et al. (2001) (experimental streams)
k15 μg/LDetenbeck et al. (1996) (experimental swamp)
l20 μg/LDeNoyelles et al. (1982, 1989, 1994); Dewey (1986); Kettle et al. (1987) (experimental ponds)
m20 μg/L1 00 μg/LStay et al. (1989) (lab microcosms)
n24 μg/LKrieger et al. (1988) (recirculating streams)
o50 μg/LFairchild et al. (1994) (experimental ponds)

Overall, the following ecological impacts may be considered important from a scientific and stakeholder point of view.

Decrease in biodiversity

Decrease in biodiversity concerns the several specific types of declines.

Decline in overall species richness and densities—This may be expressed as the number of taxa or diversity indices for the total community or for taxonomic or functional groups. According to ecological theory, biodiversity usually is highest at intermediate levels of disturbance (Southwood 1988; LaPoint 1994). Experimental research and computer simulations (e.g., Scheffer et al. 2003) indeed confirm that externally imposed variability to aquatic ecosystems can allow many species to coexist by counteracting competitive exclusion. Competitive exclusion is the phenomenon that in homogeneous environments, species that compete for the same resource cannot coexist. Counteracting competitive exclusion may even be caused by periodic exposure to low levels of PPPs (see, e.g., Hanazato 1998).

Decline in population densities of ecological key species—Ecological key species are species that play a major role in ecosystem performance, productivity, stability, and resilience. Their impact on the other components of the community often is disproportionately large relative to their abundance (Power et al. 1996). It may concern species that are critical determinants in trophic cascades (e.g., top predators) or species with properties of ecological engineers, having great impact on the physical properties of the habitat (e.g., macrophytes). In this context, it is important to realize that the concern rarely is for individual organisms but, rather, for the viability of the total population in the habitat of concern.

Decline in population densities of indicator species—This may concern species with a high information content for monitoring purposes, species protected by law, or regionally rare or endangered species

Impact on ecosystem functioning

Impact on ecosystem functioning concerns negative effects on biogeochemical cycles and energy flow. According to ecological theory, the protection of community structure will ensure the maintenance of ecosystem function (Levine 1989), whereas the loss of certain species in the community does not necessarily affect ecosystem processes because of the redundancy in roles and functions provided by the surviving species (Lawton 1994). However, the observation that several redundant species usually exist within functional groups of aquatic organisms (e.g., shredders, collectors, scrapers, and predators) may be valid only for a relatively restricted spatiotemporal scale. For example, when a species becomes locally extinct because of a chemical stressor and when recovery from nearby unpolluted sites is hampered, it cannot be excluded that on a larger time scale, other stressors may affect the remaining species within the functional group. Consequently, on a larger spatiotemporal scale, the best insurance for maintaining the life-support functions of ecosystems is to protect biodiversity (Vandermeer et al. 1998). In addition, at the scale of the watershed, there may be, in comparison with any single patch, a greater range of environmental stress resulting from exposure to a higher number of different chemical stressors. However, when the stressor concerns a nonpersistent chemical in a relatively small patch of a larger water system, ecosystem functioning usually is not affected when biodiversity of the community, both temporarily and locally, declines. Key is the preservation of the ecosystem's capacity to keep on functioning, and to keep the ability for self-regulation, under a range of environmental conditions that fall within the normal operating range (in space and time) of the system (Jeffers 1997; Swift et al. 2004).

Decrease in perceived aesthetic value and functionality to humans

A decrease in the perceived aethestic value and functionality to humans can be caused by the following:

  • Disappearance of species that are highly valued and/or have a popular appeal (eg., dragonflies and water lilies).

  • Visual mortality of individuals of fishes, frogs, waterfowl, and other vertebrates. Human beings often perceive effects on vertebrates as being less acceptable than those on invertebrates (Crane et al. 2006).

  • Decrease in overall water quality (e.g., taste and odor problems, oxygen depletion, and symptoms of eutrophication).

  • Decrease in harvestable resources and recreation value (e.g., drinking water, sport fishing, and game species).

Figure Figure 3..

The domains of science, ethics, and aesthetics all play an important role in the perception of the ecological risks of toxicants and in the political decisions on the management of the agricultural use of plant protection products.

The perception and acceptance of the risks of PPPs is only partly based on scientific facts and figures. From an ethical perspective, the intrinsic value of all the species that inhabit freshwater habitats in the agricultural landscape may be an important motive to protect them. On the other hand, from an economic perspective, the profits of agroecosystems may be valued more than the conservation of populations of vulnerable species in these systems. Generally accepted public values as well as views and emotions of stakeholder groups are fundamental in the acceptability debate and, consequently, influence the management decisions regarding the agricultural use of PPPs (Figure 3).

Only society can determine what levels of disturbance to freshwater ecosystems will be acceptable, and this is strongly affected by the human demands for economic benefits from these ecosystems. However, there may be disagreements between different sectors of society as to what constitutes a desirable benefit (see, e.g., Frewer 1999). According to Bier (2001) communication between risk assessors and risk managers as well as between these risk experts and the public should be interactive to ensure that the conduct of the risk assessment process aligns with the information needs of risk managers and the public. An important role of scientists in this process is to present options, along with the potential environmental and ecological consequences of these options. These options can be considered by stakeholders and responsible governmental authorities to underpin their views and decisions.

ECOLOGICAL PROTECTION GOALS AND RISK ASSESSMENT

The decision to allow the agricultural use of PPPs implies the acceptance of effects on target pest organisms and, inevitably, the acceptance of certain effects on nontarget populations in the agroecosystem as well. Certain types of surface waters (e.g., drainage ditches) can be considered as an integral part of the agroecosystem. Susceptible nontarget organisms in aquatic ecosystems often are taxonomically related to the pest organisms of concern.

It can be argued that in agricultural landscapes, the ecological risk assessment of PPPs for freshwater organisms should be based on a dynamic view, in which populations and communities of water organisms are considered in their temporal and spatial contexts within the landscape. In this context, the multifunctionality of most surface waters in the agricultural landscape cannot be ignored (recognizing their ecological importance, but also their economic functions). However, to maintain a high (potential) biodiversity in surface waters usually is a desired management goal at the spatial scale of the watershed. In addition, in many parts of Europe (e.g, in The Netherlands), nature reserves and areas with intensive agriculture often are hydrologically interconnected, such as via groundwater flows and/or the dense network of surface waters (e.g., ditches, lowland streams, and rivers). In dry summers, agricultural drainage water may even be used to supply nature reserves with water.

It might be argued that a differentiation in the protection level of aquatic habitats may contribute to a more focused risk assessment that takes into account perceived differences in functionality and intrinsic value of surface waters as well as the different aims of the EU Directives. When considering the risks of PPPs in surface waters, it may be convenient to distinguish 4 principles (Brock 2001) that allow for temporally and spatially differentiated protection goals,

  • 1.The Pollution Prevention Principle.
  • 2.The Ecological Threshold Principle.
  • 3.The Community Recovery Principle.
  • 4.The Functional Redundancy Principle.

The Pollution Prevention Principle

The Pollution Prevention Principle presupposes that conservative approaches are necessary to protect the environment. This is because it is not possible to exclude the possible presence of extremely sensitive populations in the ecosystems at risk, multiple stresses resulting from low concentrations of more than 1 substance, latency of effects, or unexpected effects, especially when the chemicals are toxic, persistent, bioaccumulative, and mobile in the environment. The Pollution Prevention Principle should not be confused with the Precautionary Principle, as adopted by the EU (EU 2000b). The Pollution Prevention Principle simply aims to prevent pollution, whereas the Precautionary Principle is more subtle and is based on precautionary action if the uncertainty of the risk is too great. In that case, the measures taken should be proportionate and temporary, accompanied by efforts to reduce the uncertainty, and reviewed again when further information becomes available.

When applying the Pollution Prevention Principle, emission of substances to nontarget sites should be prevented as much as technologically and socioeconomically feasible. In this respect, special attention should be paid to chemicals that are persistent, bioaccumulative, and toxic (the so-called PBT substances), although most of the pesticides with PBT properties (e.g., DDT and other organochlorine pesticides) have been phased out in the EU. In addition, when aiming to protect aquatic ecosystems, special attention also should be paid to PPPs that are water soluble, persistent, and consequently, very mobile in the aquatic environment (see, e.g., several triazine and urea herbicides). These compounds frequently violate the drinking water criterion of 0.1 μg/L at monitoring sites (EU 1998; Schultz 2004). An option in line with the Pollution Prevention Principle is always to use conservative procedures to assess risks of PPPs with PBT properties and/or that are very mobile in the aquatic environment. The derivation of EQS values for priority (hazardous) substances within the context of the WFD tends toward this approach (see Statement 44 of the WFD).

A key question is where in the landscape should water quality be in line with the EQS values. For priority (hazardous) substances currently identified within the WFD legislation, it seems that their exposure concentration in surface waters always should be in line with the officially published AA-EQS and MAC-EQS values (with an ecotoxicological basis), irrespective of the type of water body or the EU Member State of concern. The MAC-EQS for the abstraction of drinking water should account for those localities in the watershed where water is abstracted for the production of drinking water. The priority (hazardous) substances include the pesticides alachlor, atrazine, chlorfenvinphos, chlorpyrifos, diuron, endosulfan, lindane, isoproturon, simazine, and trifluralin (for the most recent information regarding EQS values of priority (hazardous) substances, see http://europa.eu.int).

In this discussion paper, a decision scheme for the derivation of EQS values for priority (hazardous) substances is not given, because the methods to derive these EQS values are still under discussion in the EU (see the proposal of Lepper 2002). In addition, the procedure to derive these EQS values is based not only on scientific arguments but also on political consensus between EU Member States.

Most PPPs currently used in Europe do not have PBT properties and are not considered as priority (hazardous) substances. However, water-quality objectives for some of these PPPs may be defined per river basin district. In accordance with the WFD, EU Member States have to set quality standards for river basin-specific pollutants. Ideally, these water-quality standards should guarantee an adequate protection of the structure and functioning of aquatic communities, whereas the methods used to derive these standards should be based on scientific understanding. These methods may be the same as those used to derive EQS values for priority (hazardous) substances (see the proposal of Lepper 2002). However, several ecological risk assessment methods developed in support of the administration procedure of pesticides also may be used to derive water-quality standards that sufficiently protect the structure and functioning of freshwater communities. The Ecological Threshold Principle may be the umbrella under which these methods are harmonized.

The Ecological Threshold Principle

The Ecological Threshold Principle considers a certain concentration of a substance to be acceptable if the most sensitive structural and functional endpoints of the exposed communities either are not or are only briefly affected. Consequently, the Ecological Threshold Principle aims always to protect sensitive populations and processes in surface waters potentially exposed to the PPP of concern. This opinion is in line with the Rivet Hypothesis (Ehrlich and Ehrlich 1981) that presumes each loss of a species affects ecosystem integrity to a small extent; if too many rivets are lost, the system collapses.

A tiered strategy may be adopted to derive the maximum permissible concentrations in which, besides the standard test species approach, the SSD approach and the model ecosystem approach also play a role. As mentioned already, the scientifically underpinned methods currently used may differ when deriving EQS values within the context of the WFD or acceptable concentrations within the context of Directive 91/414/EEC.

Here, an attempt is undertaken to describe a tiered approach of scientifically underpinned methods that can be used to derive EQS values for PPPs at WFD monitoring sites and aquatic ecosystems with an important nature conservation function. The EQS in line with the Ecological Threshold Principle will be denoted as the EQSlong-term and EQSshort-term, respectively. The EQSlong-term aims to ensure protection of ecosystem structure and functioning against long-term exposure. This norm concentration may be compared with the annual average concentration or a chosen probability (e.g., 90th percentile values) of measurements at monitoring sites in specific bodies of water (WFD approach) or with time-weighted average (TWA) concentrations calculated for a certain realistic, worst-case time window (e.g., 7, 21, or 28 d), as is common practice in the authorization procedure of PPPs. The EQSshort-term aims to ensure protection against short-term exposure and should be compared with measured or calculated peak concentrations in the water system of concern.

Ideally, the procedures to derive EQSlong-term or EQSshort-term values for PPPs should be transparent and based on information already available in the dossiers used for the administration of these chemicals. When harmonizing the Scientific methods to derive EQSlong-term or EQSshort-term values, it is important to accept that most PPPs concern chemicals with a specific toxic mode of action and that this information should be used in the risk assessment procedure. In addition, it should be realized that nonpersistent PPPs in the aquatic environment often cause acute risks from short-term exposure only. For this reason, the methodologies to assess ecological risks of long-term and short-term exposures need equal attention.

The methods described in Table 7 to derive EQSshort-term and EQSlong-term values for PPPs follow a tiered approach. The 1st tier makes use of acute and chronic toxicity data as obtained for standard aquatic test species and as specified in the Guidance Document on Aquatic Ecotoxicology (SANCO 2002a).

The 2nd tier described in Table 7 is based on the SSD approach. When using the SSD approach to derive EQSshort-term or EQSlong-term values, we propose to use the lower limit of the HC5 following the statistical method described by Aldenberg and Jaworska (2000; see Figure 2 for a quantitative presentation). The lower-limit HC5 is the lower value of the 95% confidence interval around the median HC5 value and is the concentration that with 95% certainty protects at least 95% of the species tested (Figure 1). In our proposal, acute toxicity data (e.g., acute EC50 values) are used to construct the SSD used to derive the EQSshort-term, whereas for the EQSlong-term, the SSD is based on chronic toxicity data (e.g., chronic NOEC or chronic EC10 values). Alternatively, an AF (e.g., 10 if the acute to chronic ratios of the Ist-tier standard test organisms support this) may be used to extrapolate an EQSshort-term to an EQSlong-term.

The 3rd tier described in Table 7 makes use of micro- and mesocosm experiments. The comparison of model ecosystem experiments performed with chlorpyrifos (Table 4) and lambda-cyhalothrin (Table 5) suggests that the threshold levels for no to slight effects earn confidence as an indicator of safe concentrations in the field, at least for short-term (single or repeatedly pulsed) exposure regimes. On the other hand, it appears from these data that the margin between the effect classes 2 and 3 may be a factor of only 2 (see Tables 4 and 5). For this reason, a small AF of 3, to be applied to the effect class 2 concentration, seems to be justified to derive an EQSshort-term for a PPP if a single high-quality model ecosystem is available.

It appears from the model ecosystem experiments with the persistent herbicide atrazine that concentrations in the range of “no” (effect class 1) to “slight and transient” (effect class 2) show some variability (Table 6). In addition, atrazine exposure concentrations that in 1 study resulted in class 1 or 2 effects caused class 4 or 5 effects in other experiments. We are not aware of experiments with other pesticides that allow a comparison between model ecosystem experiments of threshold concentrations for direct toxic effects as a result of a chronic exposure regime. In deriving an EQSlong-term value for PPPs, we propose to use an AF of 3 to 5 to be applied to an effect class 2 concentration if a single high-quality model ecosystem experiment is available that studied the impact of a long-term exposure regime. When applying an AF of 3 or 5 to the highest effect class 2 concentration observed for atrazine, the derived acceptable concentration is considerably lower than the lowest effect class 3 to 5 concentration reported (Table 6).

We are aware that our proposal implies a strong judgment about the representativeness of chlorpyrifos, lambda-cyhalothrin, and atrazine. Given the rather limited number of substances for which threshold levels for direct toxic effects can be derived from several model ecosystem experiments, it is prudent to recommend that the proposed AF of 3 for the EQSshort-term and of 3 to 5 for the EQSlong-term be revisited when more data are available.

Community Recovery Principle

The Community Recovery Principle presupposes that an ecosystem can absorb and endure a certain amount of pollution because of ecological recovery processes. The stressor should be limited to an intensity or concentration that causes reversible impacts only on the most sensitive populations. From a scientific point of view, periodically occurring declines in population densities can be considered as a normal phenomenon in ecosystems. In the course of their evolution, organisms have developed a large variety of strategies to survive and cope with temporally variable and unfavorable conditions, such as desiccation, flooding, temperature shocks, shading, oxygen depletion, food limitations, toxins in food, and anthropogenic stressors (Ellis 1989). In some cases, but certainly not in all, the stress caused by a PPP may more or less resemble that of a natural stress factor. The use of the “normal operating range” of population densities and functional endpoints in specific ecosystems has been suggested as a baseline against which to assess pesticide-induced changes (Domsch et al 1983). In other words, effects of PPPs for which the bioavailable fraction is restricted in space and time may, in certain habitats, be regarded as ecologically unimportant when they are of a smaller scale than changes caused by other natural or anthropogenic stresses.

The “Unless” clauses formulated within the context of the UP (Directive 91/414/EEC) tend toward the Community Recovery Principle (see also Introduction and Problem Formulation), at least for the multifunctional aquatic ecosystems adjacent to the sites of application (e.g., drainage ditches). Within the context of Directive 91/414/EEC, it seems to be logical to compare the maximum permissible concentrations with the estimated short-term (peak) and long-term (TWA) exposure concentrations in water as assessed for the different Forum for the Coordination of pesticide fate models and their use (FOCUS) scenarios (SANCO 2001).

Table Table 7.. Proposed approaches and assessment factors to derive predicted-no-effect concentration (PNEC) values in line with the Ecological Threshold Principle for Plant Protection Products (PPPs)a
 Assessment factor
ApproachesEQSshort-termEQSlong-term
  1. a Higher-tier procedures overrule lower-tier procedures (going from I to III). The measured/estimated peak concentration should be lower than the short-term Environmental Quality Standard (EQSshort-term), whereas measured/estimated long-term concentrations (e.g., arithmetic mean or time-weighted average (TWA) concentrations during relevant time windows) should be lower than the long-term Environmental Quality Standard (EQSlong-term).

  2. b Because it concerns risks to vertebrates, it may be decided to use a small AF or to use the acute EC10 or acute NOEC values to construct the SSD.

  3. c For a description of effect classes to evaluate population and community responses in micro- and mesocosm, see Table 1.

  4. d The size of the AF may be based on the structural complexity of the experimental ecosystem and the type and number of measurement endpoints studied.

1st tier: Standard test species approach
Ia: Based on acute toxicity data
Lowest acute L(E)C50 value of at least 3 species representing 3 trophic levels (selection out of Daphnia, Chironomus, alga, Lemna, and fish). [Method in accordance with that described by Lepper (2002) and, except for plants, also with the Guidance Document on Aquatic Ecotoxicology (SANCO, 2002a). SANCO (2002a) describes an assessment factor (AF) of 10 to be multiplied with the EC50 of algae and Lemna.]100
Ib: Based on chronic toxicity data
Lowest chronic NOEC of at least 3 to 5 species representing 3 trophic levels (selection out of Daphnia, Chironomus, alga, Lemna, and fish). [Method in accordance with that of Lepper (2002) and, except for plants, also with Guidance Document on Aquatic Ecotoxicology (SANCO, 2002a). SANCO (2002a) addresses the risks to plants on the basis of acute toxicity data only.]10
2nd Tier: Species sensitivity distribution approach
IIa: Based on acute toxicity data
Lower-limit hazardous concentration for 5% of the species (HC5) estimate based on a minimum of 10 L(E)C50 values for at least 8 taxonomic groups in case the PPP has biocidal properties, [Species selection in accordance with Technical Guidance Document (EU 2003). The proposal to use acute toxicity data and the lower limit of the HC5 is new.]1
or, in case of a PPP with a specific toxic mode of action.
Lower-limit HC5 estimate based on a minimum number of 8 acute EC50 values of a representative and sensitive taxonomic group of either invertebrates or plants, [Species selection in accordance with HARAP Guidance Document (Campbell et al. 1999). The proposal to use the lower limit of the HC5 is new and scientifically underpinned by Maltby et al. (2005) and Van den Brink et al. (2006).]1
and/or.
Lower-limit HC5 estimate based on a minimum number of 5 acute LC50 values for fish and/or other vertebrates (5 taxa because of animal welfare considerations) [Number of taxa in accordance with HARAP Guidance Document (Campbell et al. 1999)]1–3b
Note: An appropriate AF may be used to extrapolate the lower-limit acute HC5 estimate to a lower-limit chronic HC5 estimate, such as an AF of 10 if the acute to chronic ratios of the 1 st-tier standard test species support this. Case by case
IIb: Based on chronic toxicity data
Lower-limit HC5 estimate based on a minimum of 10 chronic NOEC values for at least 8 taxonomic groups in case the PPP has biocidal properties, [Species selection in accordance with Technical Guidance Document (EU 2003). The proposal to use the lower limit of the HC5 is new.]1
or, in case of a PPP with a specific toxic mode of action.
Lower limit HC5 estimate based on 8 or more chronic NOEC/EC10 values of a re presentative and sensitive taxonomic group of either invertebrates or plants, [Species selection in accordance with HARAP Guidance Document (Campbell et al. 1999). The proposal to use the lower limit of the HC5 is new and scientifically underpinned by Van den Brink et al. (2006).]1
and/or.
Lower limit HC5 based on a minimum number of 5 chronic NOEC/EC10 values for fish (5 taxa because of animal welfare considerations). [Number of taxa in accordance with HARAP guidance document (Campbell et al. 1999).]1
3rd Tier: Model ecosystem approach
IIIa: Short-term exposure
Effect class 1c concentration (based on peak levels) for the most sensitive relevant endpoint assessed in a high-quality micro- and mesocosm experiment studying an acute exposure regime.1–3d
or,
Effect class 2c concentration (based on peak levels) for the most sensitive relevant endpoint assessed in a high-quality micro and mesocosm experiment studying an acute exposure regime, [Effect classes in accordance with Guidance Document on Aquatic Ecotoxicology (SANCO 2002a) and proposal for size of assessment factor based on data provided in the present paper.]3
or,
Effect class 1 and 2 concentrations (based on peak levels) can be derived from several high-quality micro- and mesocosm experiments.Case by case 
Effect class 1 concentration (based on long-term exposure concentration during representative time window) for the most sensitive endpoint assessed in a high-quality micro- and mesocosm experiment studying a chronic exposure regime.3
or,
Effect class 2 concentration (based on long-term exposure concentration during re presentative time window) for the most sensitive endpoint assessed in a high-quality micro- and mesocosm experiment studying a chronic exposure regime, [Effect classes in accordance with Guidance Document on Aquatic Ecotoxicology (SANCO 2002a) and proposal for height of assessment factor based on data provided in this paper.]3–5d
or,
Effect class 1 and 2 concentrations (based on long-term exposure concentration during representative time window) can be derived from several high-quality micro- and mesocosm experiments studying a chronic exposure regime.Case by case

In accordance with the Guidance Document on Aquatic Ecotoxicology (SANCO 2002a), a tiered approach is adopted to derive short-term and long-term maximum permissible concentrations in line with the Community Recovery Principle. In the present paper, they will be referred to as regulatory acceptable concentrations (RACs)—more specifically, the short-term RAC (RACshort-term) and the long-term RAC (RAQlong-term). In the present paper, we avoid using the concept of ecologically acceptable concentration, as used in guidance documents related to Directive 91/414/EEC (Campbell et al. 1999; Giddings et al. 2002; SANCO 2002a) because of associated semantic issues (see Crane and Giddings 2004). From a philosophical point of view, it may be argued that ecology as a science has no moral principles and that, consequently, something like “ecologically acceptable” does not exist and should not be mixed up with “regulatory acceptable.” In the present paper, we consider the time window of observed effects (including the time needed for recovery) to be an important criterion in the higher-tier assessment of the RAC.

The 1st-tier approach in line with the Community Recovery Principle (Table 8) is based on the standard test species-AF approach as described in the UP (EU 1997) and the Guidance Document on Aquatic Ecotoxicology (SANCO 2002a).

The 2nd-tier approach mentioned in Table 8 is the SSD approach and the calculation of the HC5 (Aldenberg and Jaworska 2000). Assuming that the substances evaluated in Tables 2 and 3 are representative of other pesticides, it appears the probability is high that the median HC5 value of a pesticide (based on acute toxicity data of relevant taxonomic groups) is lower than the effect class 2 concentration derived from a micro- and mesocosm experiment to which the same (nonpersistent) pesticide is applied once (Figure 2). For pesticides that are applied repeatedly to micro- and mesocosm experiments, it appears the probability is high that the lower-limit HC5 value (based on acute toxicity tests) is lower than the effect class 1 and effect class 2 concentrations (Figure 2). In addition, Maltby et al. (2002) and Van den Brink et al. (2006) showed for insecticides and herbicides that the median value of the HC5 based on chronic toxicity data (for arthropods and primary producers, respectively) was protective of adverse effects in aquatic micro- and mesocosms under a long-term exposure regime. However, the database for this comparison of chronic studies was relatively small.

As a 3rd tier, the model ecosystem approach is mentioned in Table 8. The effect classes described in Table 1 are used to facilitate the interpretation of concentration-response relationships for relevant measurement endpoints. The Community Recovery Principle presupposes that a certain concentration of a PPP is acceptable as long as effects on sensitive species are short term (in the sense that actual recovery of these species occurs within an acceptable time frame). In the present paper, the acceptable period for actual recovery in high-quality micro- and mesocosm experiments mimicking a short-term exposure regime is set at 8 weeks postapplication (effect class 3). In the review papers that introduced the effect classes (Brock, Lahr, and Van den Brink 2000; Brock, Van Wijngaarden, and Ven Geest 2000), a period of 8 weeks was chosen, because in most of the micro- and mesocosms evaluated, macroinvertebrates were sampled at intervals of 2 to 4 weeks. Consequently, a period of 8 weeks after the last application allows a few sampling dates to show recovery. We recognize, however, that choosing the acceptable time frame for recovery, which may differ for different taxonomic groups, is in fact a risk management decision. On the basis of consensus reached among risk managers, the effect classes mentioned in Table 1 can be adapted accordingly.

It appears from the data presented in Tables 4 and 5 that an AF of at least 3 is necessary for spatiotemporal extrapolation of an effect class 3 peak concentration from a single high-quality model ecosystem experiment mimicking an acute exposure regime. On the basis of the data presented in Tables 4 and 5, applying an AF of 3 to the highest effect class 3 concentration avoids the occurrence of unacceptable class 4 and 5 effects because of short-term (pulsed) exposure in other systems.

Concomitantly, an AF of at least 3 seems to be necessary to extrapolate an effect class 2 concentration from a single high-quality model ecosystem experiment that mimics a more or less constant, chronic exposure regime. On the basis of the data presented in Table 6, applying an AF of 3 to the highest effect class 2 concentration observed will, with a high probability, avoid unacceptable class 4 and 5 effects because of long-term exposure.

When assessing the risks of repeated short-term exposure because of a multiple application scenario, the total time window of potential effects during the application period should be considered in concert with the time needed for recovery. If in the multiple-application scenario the total time window of effects is too long because of a high number of applications, it is an option to use a higher AF or to consider the effect class 1 and 2 concentration as acceptable.

Again, considering the limited number of pesticides in the data set used, it is prudent to recommend that the proposed AF of 3 to derive higher-tier RAC values on the basis of micro- and mesocosm experiments be revisited when more data are available.

Functional Redundancy Principle

The Functional Redundancy Principle presupposes that for sustainable functioning of the agroecosystem, a decrease in biodiversity can be tolerated as long as key species and their functions (e.g., nutrient cycling and mineralization of organic matter) are not affected beyond an unacceptable level. This principle rests on the redundancy in roles and functions provided by the surviving species in the community, and it is in line with the functional redundancy hypothesis (Lawton 1994).

When adopting the Functional Redundancy Principle, the emphasis is on ecosystem processes; impacts are considered to be acceptable when functional attributes are not changed, despite possible effects on community structure. At the community and ecosystem level, functional endpoints rarely are more sensitive than structural ones (e.g., Kersting 1994). Effects on functional endpoints indicate the limit of functional redundancy within the stressed community. Once ecosystem processes have changed as a result of contamination, this usually is an indication of truly severe effects on structural endpoints.

The Functional Redundancy Principle may be used to evaluate the acceptability of the impact of pesticides in areas that are considered to have a main function in the production of crops and food (Van der Linde et al. 2006). This protection goal is focused on the protection of the life-support function of the agroecosystem or hydroculture to allow the growth of the crop or the fish/prawn culture and to protect its quality. The text in the EU Guidance Document on Terrestrial Ecotoxicology (SANCO 2002b) with respect to the environmental risk assessment of PPPs in agricultural soils of agroecosystems is in line with the Functional Redundancy Principle. In the aquatic environment, the Functional Redundancy Principle may be used to derive maximum permissible concentrations of PPPs applied in systems like rice paddies and in fish breeding ponds.

Table Table 8.. Approaches and (proposed) assessment factors to derive regulatory acceptable concentrations (RACs) for Plant Protection Products (PPPs) in line with the Community Recovery Principlea
 Assessment factor (AF)
ApproachesRACshort-termRAClong-term
  1. a Higher-tier procedures may overrule lower-tier procedures (going from tier 1 to tier 3). The measured/estimated peak concentration should be lower than the short-term RAC (RACshort-terrn), whereas measured/estimated long-term exposure concentrations during relevant time windows should be lower than the long-term RAC (RAClong-term). The approaches are based on the possible tiers described in the HARAP Guidance Document (Campbell et al. 1999).

  2. b For insecticides with a specific toxic mode of action, use insect species like Chironomus.

  3. c Because it concerns risks to vertebrates, it may be decided to use a small AF or the acute EC10 or acute NOEC values to construct the species sensitivity distribution.

  4. d Effect classes to evaluate population and community responses in micro- and mesocosm are described in detail in Table 1.

1st tier: Standard test species approach
Ia: Based on acute toxicity data
Lowest acute LC50 value of at least 2 standard fish species (cold- and warm water).100
Acute EC50 value of at least 1 or 2 standard invertebrate species (Daphnia, Chironomusb), and100
EC50 value of at least 1 standard primary producer species (usually an alga; for herbicides, 2 algae and a representative macrophyte, e.g., Lemna). [Approach in accordance with Uniform Principles and Guidance Document on Aquatic Ecotoxicology (SANCO, 2002a).]1010
Ib: Based on chronic toxicity data
Lowest chronic NOEC of at least 2 standard test species representing invertebrates (Daphnia, Chironomus) and fish (required if dissipation time 50% (DT50)water > 2 d, or repeated applications). [Approach in accordance with Uniform Principles and Guidance Document on Aquatic Ecotoxicology (SANCO, 2002a).]10
2nd Tier: Species sensitivity distribution approach
IIa: Based on acute toxicity data and risk concerns a single pulse exposure or a single application of a relatively fast-dissipating PPP (DT50water < 10 d) in a lentic system
Median hazardous concentration for 5% of the species (HC5) based on a minimum of 8 acute L(E)C50 values for at least 8 taxonomic groups in case the PPP has biocidal properties,1
or, in case of a PPP with a specific toxic mode of action
Median HC5 based on a minimum number of 8 acute EC50 values of a representative taxonomic group of either invertebrates or plants, [Proposed method scientifically underpinned by Maltby et al. (2005) and Van den Brink et al. (2006).]1
and/or,
Median HC5 based on a minimum number of 5 acute LC50 values for fish (5 taxa because of animal welfare considerations). [The number of species according to Campbell et al. (1999).]1–3c
IIb: Based on acute toxicity data and risk caused by repeated pulse exposures or by a single application of a relatively persistent PPP in a lentic system (DT50water > 10 d)  
Lower-limit HC5 based on a minimum of 8 acute L(E)C50 values for at least 8 taxonomic groups in case the PPP has biocidal properties.1
or, in case of a PPP with a specific toxic mode of action.
Lower-limit HC5 based on a minimum number of 8 acute EC50 values of a representative taxonomic group of either invertebrates or plants, [Proposed method scientifically underpinned by Maltby et al. (2005) and Van den Brink et al. (2006).]1
and/or,
Lower limit HC5 based on a minimum number of 5 acute LC50 values for fish (5 taxa because of animal welfare considerations) [The number of species according to Campbell et al. (1999).]1–3c
or,
Procedure of tier 3a (median HC5 estimates) and it is demonstrated in tests with representative sensitive species that the risk of a repeated application is not larger than that of a single application. [New proposal.]Case by case
Note: An AF may be used to extrapolate the median acute HC5 to a median chronic HC5, such as an AF of 10 if the acute to chronic ratios of the tier 1 species support this.Case by case
IIc: Based on chronic toxicity data
Median HC5 based on a minimum of 8 chronic NOEC/EC10 values for at least 8 taxonomic groups in case the PPP has biocidal properties.1
or, in case of a PPP with a specific toxic mode of action.
Median HC5 based on a minimum number of 8 chronic NOEC/EC10 values of a representative taxonomic group of either invertebrates or plants, [Proposed method scientifically underpinned by Maltby et al. (2005) and Van den Brink et al. (2006).]1
and/or,
Median HC5 based on a minimum number of 5 chronic NOEC/EC10 values for fish (5 taxa because of animal welfare considerations).1
3rd tier: Model ecosystem approach
IIIa: Model ecosystem experiments mimicking a realistic short-term or pulsed exposure regime
Effect class 1 and 2 concentration (based on peak concentration) for the most sensitive relevant endpoint assessed in a high-quality micro- and mesocosm experiment.1
or,
Effect class 3d concentration (based on peak concentration) for the most sensitive endpoint in a high-quality (lentic) micro- and mesocosm experiment (see Note below).3
or,
Effect class 3 concentrations can be derived from several high-quality micro- and mesocosm experiments studying an acute exposure regime.Case by case
Note: In multiple-application scenarios, the effects observed during the application period also may be considered as part of the total time window of effects. If the number of applications is high and, consequently, the time window of effects is large, it may be decided to use either a higher AF or to base the assessment on effect class 1 and 2 concentrations [Procedure in accordance with CLASSIC Guidance Document (Giddings et al. 2002) and proposal for size of AF based on data in the present paper]  
IIIb: Model ecosystem experiments mimicking a realistic long-term exposure regime
Effect class 1d concentration (based on long-term exposure concentration during relevant time window) for the most sensitive endpoint in a high-quality micro- and mesocosm experiment.1
Effect class 2 concentration (based on long-term exposure concentration during relevant time window) for the most sensitive endpoint assessed in a high-quality micro- and mesocosm experiment.3
or,
Effect class 2 concentrations (based on long-term exposure concentrations) for the most sensitive endpoint can be derived from several high-quality micro- and mesocosm experiments [Procedure in accordance with CLASSIC Guidance Document (Giddings et al. 2002) and proposal for size of AF based on data in the present paper.]Case by case

When adopting the Functional Redundancy Principle, an important evaluation criterion should be the length of time required for reversibility of the effect, because complete irreversible change in agroecosystem use should be guarded against. This, however, may hamper crop rotation and future changes in system use. In aquatic agroecosystems (e.g., rice paddies and fish breeding ponds), an effect class 2 and 3 concentration (Table 1) based on functional attributes (e.g, oxygen metabolism, nutrient cycling, and decomposition of organic matter) may be considered to be acceptable if the PPP of concern does not harm the survival of the crop plants or animals. In addition, affected nontarget populations should recover within an acceptable time frame to allow crop rotation. In the present paper, no attempt is undertaken to provide a decision scheme in line with the Functional Redundancy Principle, because the current guidance documents that deal with the protection of freshwater ecosystems mainly focus on structural endpoints when assessing risks.

CASE STUDIES WITH EXAMPLE DATA SETS

In this section, we use available ecotoxicological data for the insecticide chlorpyrifos, the herbicide metribuzin, and the fungicide carbendazim to illustrate the risk assessment procedures described in Table 7 (derivation of EQS values) and Table 8 (derivation of RAC values).

A summary of the relevant 1 st-tier aquatic toxicity data for chlorpyrifos, metribuzin, and carbendazim, as well as the 1st-tier hazard assessment procedure by means of the standard test species approach, is presented in Table 9. These data reveal that the crustacean Daphnia is the most sensitive standard test species for the insecticide chlorpyrifos, whereas the green alga Chlorella vulgaris is the most sensitive for the herbicide metribuzin. The specific toxic mode of action of these compounds is clearly reflected in the toxicity values for standard test species. In contrast, in the case of the fungicide carbendazim, the acute toxicity data reveal that not only Daphnia but also the fish Oncorhynchus and the alga Chlorella are relatively sensitive, suggesting a biocidal mode of action by this fungicide. Nevertheless, in our carbendazim database, the most sensitive standard test species was D. magna, in both acute and chronic tests. For chlorpyrifos and carbendazim, the 1st-tier EQS values and the 1st-tier RAC values are similar, because the procedures described in Tables 7 and 8 follow the same approach as that when Daphnia is the most sensitive standard test species. In contrast, for the herbicide metribuzin, the 1st-tier EQS and RAC values differ by up to a factor of 10, because different AFs and/or toxicity data are used in the 1st-tier effect assessment procedures if an alga is the most sensitive standard test species.

For the 3 compounds, a large enough number of acute toxicity data for different aquatic species is available, allowing application of the SSD approach and calculation of acute HC5 values (Table 10). For chlorpyrifos, the effect assessment by means of the SSD approach is based on no less than 92 arthropod acute toxicity data. For this reason, the median acute HC5 value and the lower-limit acute HC5 value do not differ much. For metribuzin, the SSD is based on 19 primary producer acute toxicity data (10 algae and 9 aquatic vascular plants). Because of the biocidal properties of carbendazim, the SSD for this compound is based on toxicity data representing different taxonomic groups. For carbendazim, it was possible to construct an acute SSD based on 11 nonvertebrate taxa (2 algae and 9 invertebrates) and an acute SSD based on 17 taxa, including both aquatic nonvertebrates and vertebrates (all taxa, including 4 fish and 2 amphibians). In addition, for chlorpyrifos and metribuzin, enough data are available to construct SSDs based on toxicity data for fish (Table 10). Because the HC5 values calculated for chlorpyrifos are based on a larger number of toxicity data, differences between 2nd-tier EQS and RAC values are smaller for chlorpyrifos than for metribuzin (Table 10). For carbendazim, EQS and RAC values derived from HC5 estimates are given for both nonvertebrates and all taxa. When comparing Tables 9 and 10, it appears that for chlorpyrifos and metribuzin, the risk assessment based on the SSD approach is less conservative than that based on the standard test species approach. For carbendazim, however, the 1st-tier triggers (Table 9) were somewhat higher than the EQSshort-term, EQSlong-term, and RACshort-term values based on the SSD approach (Table 10).

Table Table 9.. Summary of acute and chronic toxicity values (geometric means) of aquatic standard test species found for the insecticide chlorpyrifos, the herbicide metribuzin, and the fungicide carbendazim in the databases used by Maltby et al. (2005), L Maltby, Department of Animal and Plant Sciences, University of Sheffield, Sheffield, UK, unpublished data, and Brock et al. (2004)a
 ChlorpyrifosMetribuzinCarbendazim
 Acute EC50/LC50Chronic NOECAcute EC50/LC50Chronic NOECAcute EC50/LC50Chronic NOEC
Invertebrates
 Daphnia magna0.84 μg/L0.057 μg/L14,983 μg/L3,298 μg/L391 μg/L13 μg/L
Fish
 Oncorhynchus my kiss18.8 μg/L0.51 μg/L53,497 μg/L6,068 μg/L431 μg/L18 μg/L
 Lepomis macrochirus9.0 μg/L     
Primary producers
 Scenedesmus subspicatus  28.9 μg/L1.8μg/L54,000 μg/L10,000 μg/L
 Selenastrum capricornutum  39.7 μg/L5.4 μg/L1 3,000 μg/L500 μg/L
 Chlorella vulgaris  12.3 μg/L   
 C. pyrenoidosa    484 μg/L 
Skeletonema costatum403 μg/L     
 Lemna minor/gibba  36.5 μg/L   
1 st-Tier effect assessment (standard test species approach)
 ChlorpyrifosMetribuzinCarbendazim
 Short-term exposureLong-term exposureShort-term exposureLong-term exposureShort-term exposureLong-term exposure
  1. a The lower part of the table presents the 1 st-tier effect assessment in line with the Ecological Threshold Principle (Table 7) and Community Recovery Principle (Table 8). EQS = Environmental Quality Standard; RAC = regulatory acceptable concentration.

EQSshort-term0.0084 μg/L (

Table 7

; la)
 0.123 μg/L (

Table 7

; la)
 3.91 μg/L (

Table 7

; la)
 
EQSlong-term 0.0057 μg/L (

Table 7

; lb)
 0.18 μg/L (

Table 7

; lb)
 1.3 μg/L (

Table 7

; lb)
RACshort-term0.0084 μg/L (

Table 8

; la)
 1.23 μg/L (

Table 8

; la)
 3.91 μg/L (

Table 8

; la)
 
RAQlong-term 0.0057 μg/L (

Table 8

; lb)
 1.23 μg/L (

Table 8

; la)
 1.3 μg/L (

Table 8

; lb)
Table Table 10.. Summary of the hazardous concentrations to 5% of the species (HC5) and the lower limit of the 95% confidence interval (in parentheses) for the insecticide chlorpyrifos, the herbicide metribuzin, and the fungicide carbendazima
 ChlorpyrifosMetribuzinCarbendazim
 Acute EC50/LC50 (μg/L)Chronic NOEC (μg/L)Acute EC50/LC50 (μg/L)Chronic NOEC (μg/L)Acute EC50/LC50 (μg/L)Chronic NOEC (μg/L)
HC5: Sensitive0.08 (0.05) 7.4 (4.0)1.4 (0.2)14.2 (1.7) 
taxonomicn = 92 n = 19n = 8n = 17 
groupArthropoda PlantsPlantsAll taxa 
     24.6 (2.2) 
     n = 11 
     Nonvertebrates 
HC5: Fish0.58 (0.15) n = 300.07(0.01) n = 73315 (615) n = 11   
Higher-tier effect assessment (species sensitivity approach)
 ChlorpyrifosMetribuzinCarbendazim
 Short-term exposureLong-term exposureShort-term exposureLong-term exposureShort-term exposureLong-term exposure
  1. a The HC5 values were calculated on the basis of acute or chronic toxicity data for species belonging to the relevant, sensitive taxonomic group of aquatic organisms and to aquatic vertebrates. Most data were retrieved from Maltby et al. (2005), Brock et al. (2004) and Van Wijngaarden et al. (1998). In addition, the lower part of the table presents the higher-tier predicted-no-effect concentrations in line with the Ecological Threshold Principle (Table 7) and Community Recovery Principle (Table 8) and obtained by means of the species sensitivity distribution approach. EQS = Environmental Quality Standard; RAC = regulatory acceptable concentration; DT50 = dissipation time 50%.

EQSshort-term0.05 μg/L (

Table 7

, IIa; lower-limit acute HC5 for Arthropoda)
 4.0 μg/L (

Table 7

, IIa; lower-limit acute HC5 for plants)
 1.7–2.2 μg/L (

Table 7

, IIa; lower limit acute HC5 for all taxa or nonvertebrates)
 
EQSlong-term 0.005 μg/L (

Table 7

, IIa; lower-limit acute HC5 Arthropoda and AF of 10)
 0.2 μg/L (

Table 7

, IIb; lower-limit chronic HC5 for plants)
 0.17–0.22 μg/L(

Table 7

, IIa; lower-limit acute HC5 for all taxa or nonvertebrates and AF of 10)
RACshort-term0.08 μg/L (

Table 8

, IIa; median acute HC5 for Arthropoda)
 7.4 μg/L (

Table 8

, IIa; median acute HC5 for plants)
 1.7–2.2 μg/L (

Table 8

, 2b; lower-limit acute HC5 for all taxa or nonvertebrates)
 
RAClong-term 0.008 μg/L (

Table 8

, IIb; median acute HC5 for Arthropoda and AF of 10)
 1.4 μg/L (

Table 8

, IIc; median chronic HC5 for plants)
 1.42–2.46 μg/L (

Table 8

, IIb; median acute HC5 for all taxa or nonvertebrates and AF of 10)
Field DT50waterApproximately 1Approximately 7.1 d>20 d
Table Table 11.. Summary of class 1, 2 and 3 effects (see Table 1) observed in model ecosystems treated with the insecticide chlorpyrifos, the herbicide metribuzin. or the fungicide carbendazima
 ChlorpyrifosMetribuzinCarbendazim
 Short-term exposure (single application; field DT50water ∼1 d)Short-term exposure (single application; field DT50water ∼7.1 d)Short- to long-term exposure (single application; field DT50water > 20 d)Long-term exposure (constant)
Effect class 10.1 μg/L n = 6 (

Table 4

)
5.6 μg/L n = 1 (Brock et al. 2004)3.0 μg/L n = 1 (Slijkerman et al. 2004)3.3 μg/L n = 1 (Cuppen et al. 2000; Van den Brink et al. 2000)
Effect class 20.3 μg/L n = 1 (

Table 4

)
18 μg/L n = 1 (Brock et al. 2004)  
Effect class 30.5–1.0 μg/L n = 4 (

Table 4

)
56.0 μg/L n = 1 (Brock et al. 2004)30 μg/L n = 1 (Slijkerman et al. 2004) 
Higher-tier effect assessment (model ecosystem approach)
 ChlorpyrifosMetribuzinCarbendazim
 Short-term exposureShort-term exposureShort- to medium-term exposureLong-term exposure
  1. a See Table 1 for an explanation of effects class 1, 2, and 3. The lower part of the table presents the higher-tier acceptable concentrations in line with the Ecological Threshold Principle (Table 2) and Community Recovery Principle (Table 6) and obtained by means of the model ecosystem approach. In the model ecosystems, fish was not present. EQS = Environmental Quality Standard; RAC = regulatory acceptable concentration; DT50 = dissipation time 50%.

  2. b Because several studies resulted in an effect class 1 concentration of 0.1 μg/L, no assessment factor is applied (Table 4).

EQSshort-term0.1 μg/L (

Table 7

, IIIa)b
6.0 μg/L (

Table 7

, IIIa)
1.0–3.0 μg/L (

Table 7

, IIIa)
 
EQSlong-term   1.1 μg/L (

Table 7

, IIIb)
RACshort-term0.3 μg/L (

Table 8

, Ilia)
18.7 μg/L (

Table 8

, IIIa)
10 μg/L for nonvertebrates (

Table 8, IIIa) Overruled by 4.31 μg/L for fish (1st tier; acute LC50/100; Table 9

)
 
RAClong-term   3.3 μg/L for nonvertebrates (

Table 8, III) Overruled by 1.8 μg/L for fish (1 st tier; chronic NOEC/EC10; Table 9

)

The effects of short-term exposure to chlorpyrifos and metribuzin, and of medium-term and long-term exposure to carbendazim, have been studied in micro- and mesocosms (Table 11). In these semifield experiments, fish was not present, but as demonstrated in Tables 9 and 10, fish are less sensitive overall than arthropods in toxicity tests with chlorpyrifos, and fish are much less sensitive than plants in tests with metribuzin. Again, procedures described in Tables 7 and 8 are followed to derive EQS and RAC values, respectively. Overall, the higher-tier effect assessment by means of the model ecosystem approach is less conservative than that of the SSD approach (Tables 10 and 11). However, for carbendazim, the RAC values derived for the non-vertebrate populations in the model ecosystem experiments are overruled by the Ist-tier trigger for fish (Tables 9 and 11).

The case studies with the example data sets reveal that our proposed decision schemes result in a more or less logical effect assessment procedure, and that it is rewarding to provide more ecotoxicological data. The carbendazim data, however, also show that exceptions cannot be ruled out. We are aware that the validation status of the risk assessment methods incorporated in Tables 7 and 8 is still poor for pesticides with a mode of action not represented in our data set (see Tables 2 and 3).

DISCUSSION

In the present paper, the decision schemes to derive maximum permissible concentrations for PPPs in surface water are based on methods recommended in regulatory guidance documents and are scientifically underpinned by critical review papers regarding the impact of PPPs on freshwater organisms and aquatic communities. The present paper focuses on lower- and higher-tier approaches in the effect assessment, with reference to the standard test species approach (1st tier), the SSD approach (2nd tier), and the model ecosystem approach (3rd tier). However, more effect assessment approaches are mentioned in guidance documents (Campbell et al. 1999; SANCO 2004). These approaches include modified exposure studies, population-level studies, and effect model simulations. For these approaches, little guidance has been developed, and review papers that compare the outputs of these approaches with those of other, well-established tiers are lacking. As a consequence, the choice and interpretation of these approaches are highly dependent on expert judgment. To determine population-level effects, both modeling and experimental approaches can be used (see Boxall et al. 2002 for examples). Population-level studies can play a role in the regulatory procedure of pesticides only if other studies have clearly demonstrated that the specific population being tested is representative for the populations at risk. As a possible higher tier, model simulations may be used to extrapolate the results of experimental studies to a more realistic spatiotemporal context of the agricultural landscape (SANCO 2004). Food-web models might be used to predict the potential consequences of a temporal decline in sensitive species on the dynamics of indirect effects and ecosystem functioning under variable environmental circumstances (see, e.g., Traas et al. 2004). Recovery models may be the appropriate tool to evaluate the potential consequences of the local decline of species that differ in life-cycle characteristics (see, e.g., Barnthouse 2004). Finally, observations regarding the impact of pesticides in the field may be used as a further line of evidence to adjust the effect assessment procedures based on laboratory or micro- and mesocosm experiments (Liess et al. 2005).

The ecological protection goals and associated decision schemes to evaluate the hazards of PPPs in the aquatic environment described in the present paper have been presented from a scientific perspective. They may be used as options that can be considered by stakeholders and responsible governmental authorities to underpin their views and decisions, and they may act as the input for a wider debate. In addition, these protection goals and decision schemes may be used in discussions of spatial differentiation in the protection level of aquatic habitats by taking into account perceived differences in functionality and intrinsic value of surface waters as well as the different aims of EU directives. The Ecological Threshold Principle may be the umbrella under which the aims and effect assessment procedures in line with both EU Directive 200/60/EC and EU Directive 91/414/EEC are harmonized to derive water-quality objectives for WFD monitoring sites. In the decision scheme in line with the Community Recovery Principle, the current EU Directive 91/414/EEC procedures to assess the effects of PPPs in surface waters at edge-of-field situations is incorporated. The derivation of an acceptable norm concentration in drainage ditches, however, should not be in conflict with desired water-quality standards downstream in the watershed. The AFs proposed and applied in the present paper to extrapolate results of model ecosystem experiments are based on the current state of knowledge and on the data available. A review of these AFs may be necessary if more data become available or if other compounds with a specific mode of action have to be incorporated in the decision schemes.

Regulatory schemes for the authorization of PPPs currently are based on assessments of individual compounds (EU 1997). It is common practice, however, for several different pesticides to be applied during the growing season to protect crops (De Zwart 2005). The potential for interactions between different pesticides therefore has been suggested as a potential uncertainty in the risk assessment process. Relatively few studies have investigated the effects of realistic combinations of pesticides in aquatic ecosystems that are related to treatment programs used in specific crops (Van Wijngaarden et al. 2004; Arts et al. 2006). Those studies have revealed that most of the observed effects are consistent with the results from higher-tier studies using individual compounds. Apparently, for the pesticide packages studied, the risk assessment based on the individual compounds was sufficiently protective for their use in a crop-protection program (Van Wijngaarden et al. 2004; Arts et al. 2006).

The present paper predominantly addresses ecotoxicological and ecological effects of stress by PPPs. To assess environmental risks, information regarding actual exposure concentrations and/or PECs also is required. Pesticide exposure may be another important source of uncertainty in a risk assessment; consequently, higher-tier procedures also should be developed and implemented to refine estimates of exposure. This may warrant improvement of the validation status of the fate models and chemical monitoring procedures currently applied. The use of probabilistic approaches when interpreting chemical monitoring data at the watershed level may be promising (Hart 2001; Giddings et al. 2005). In addition, harmonizing the aims of the EU Directives 91/414/EEC and 2000/60/EC may need the development of fate and exposure models at the scale of the watershed.

Acknowledgements

The writing of this discussion paper was financially supported by the Dutch Ministry of Agriculture, Nature Conservation and Food Safety (Research Program 416). The authors thank Peter van Vliet and 3 anonymous reviewers for critical comments on an earlier draft of this paper.

Ancillary