draft scientific opinion draft guidance document on … · 2015-08-04 · 22 ecotoxicology...

EFSA Journal 2013;volume(issue):NNNN

Suggested citation: European Food Safety Authority; Guidance Document on Tiered Risk Assessment for Plant Protection Products for aquatic organisms in edge-of-field surface waters. EFSA Journal 2013;volume(issue):NNNN. [181 pp.] doi:10.2903/j.efsa.2013.NNNN. Available online: www.efsa.europa.eu/efsajournal © European Food Safety Authority, 2013

DRAFT SCIENTIFIC OPINION 1 2

DRAFT Guidance Document on tiered risk assessment for plant protection products for 3 aquatic organisms in edge-of-field surface waters 4

EFSA Panel on Plant Protection Products and their Residues (PPR)123 5

European Food Safety Authority (EFSA), Parma, Italy 6 7

SUMMARY 8 9 The European Food Safety Authority (EFSA) asked the Panel on Plant Protection Products and their 10 Residues (PPR) to prepare a revision of the Guidance Document on Aquatic Ecotoxicology under 11 Council Directive 91/414/EEC (SANCO/3268/2001 rev.4 (final), 17 October 2002). The PPR Panel 12 was therefore tasked to prepare a revised Guidance Document and two Scientific Opinions. This 13 Guidance of the PPR Panel is the first of these three requested deliverables as outlined in the Terms of 14 Reference below. The revision of the former Guidance Document on Aquatic Ecotoxicology became 15 necessary mainly due to (1) the entry in to force of the new Regulation (EC) 1107/2009 on 16 authorisation of plant protection products, (2) the revision of the related data requirements and (3) to 17 take into consideration relevant new scientific knowledge. Stakeholders were consulted before the 18 start of the revision process in a public consultation, as well as Risk Managers in a specific 19 consultation, in October-December 2008. The revision of the Guidance Document on Aquatic 20 Ecotoxicology was started in parallel to the revision of the Guidance Document on Terrestrial 21 Ecotoxicology (SANCO/10329/2002, rev.2 final, 17.10.2002) to allow a harmonisation process. As a 22 first step the PPR Panel developed a framework for deriving specific protection goals (EFSA, 2010). 23 The approach outlined in this opinion was the starting point for the development of this updated 24 Guidance Document on Aquatic Risk Assessment. 25 26 The aquatic risk assessment is the combination of the exposure and the effect assessments and there is 27 considerable interaction between these assessments. The focus of this Guidance Document (GD) is on 28 a tiered effect assessment scheme with detailed guidance on Tier 1 and higher tier effect assessments 29 that are mainly based on experimental approaches (Chapters 5-8). A scientific opinion describing the 30 state of the art of mechanistic effect modelling in the aquatic environment (e.g. 31 toxicokinetic/toxicodynamic and population models) will be delivered later under this mandate. The 32 effect assessment guidance is intended to be used for authorisation of active substances at EU level as 33 well as for plant protection products at Member State level. Furthermore, the appropriate linking 34 between exposure and effect assessment is described. The exposure assessment methodology was not 35 reviewed and it is assumed that the current FOCUS surface water exposure assessment methodology 36

1 On request from EFSA, Question No EFSA-Q-2009-00001, draft endorsed for public consultation on 10 December 2012. 2Panel members: Alf Aagaard, Theo Brock, Ettore Capri, Sabine Duquesne, Metka Filipic, Antonio F. Hernandez-Jerez,

Karen I. Hirsch-Ernst, Susanne Hougaard Bennekou,Michael Klein, Thomas Kuhl, Ryszard Laskowski, Matthias Liess, Alberto Mantovani, Colin Ockleford, Bernadette Ossendorp, Daniel Pickford, Robert Smith, Paulo Sousa, Ingvar Sundh, Aaldrik Tiktak, Ton Van Der Linden

3Acknowledgement: EFSA wishes to thank the members of the Working Group Aquatic Ecotoxicology: Alf Aagaard, Paulien Adriaanse, Jos Boesten, Theo Brock, Michael Klein, Matthias Liess, Robert Luttik, Paul Miller, Daniel Pickford, Aaldrik Tiktak, Jan Vanderborght, Lina Wendt-Rasch and EFSA staff: Stephanie Bopp and Maria Arena for the support provided to this scientific opinion.

DRAFT GD on tiered RA for edge-of-field surface water

EFSA Journal 2013;volume(issue):NNNN 2

will continue to be used for exposure assessment at EU level. Only a brief overview of the exposure 37 assessment is included in this GD in Chapter 4, for details reference is made to the related FOCUS 38 surface water guidance (FOCUS, 2001). 39 40 The GD describes first the specific protection goals for aquatic organisms that need to be defined in 41 order to develop an appropriate risk assessment (RA) scheme. Proposed specific protection goals 42 (SPGs) were discussed with Risk Managers in September-November 2012 and are described in 43 Chapter 3. The SPGs overall aim is to protect aquatic plants and animals at the population level in 44 surface water. However, the SPG selected for aquatic vertebrates aims a protection at the individual 45 level, so that visible mortality and suffering due to acute toxicity is avoided. As outlined in the PPR 46 Panel opinion on specific protection goals (EFSA, 2010), the exposure assessment goals also have to 47 be defined in parallel to set the overall level of protection. Since the exposure assessment methodology 48 was not revised in parallel to the effect assessment scheme, definitions for exposure assessment goals 49 are not clear. 50 51 In this GD, the tiered effect assessment procedure and proposals on how to link effects to exposure 52 estimates, are focused on aquatic organisms living in the water column of edge-of-field surface waters. 53 For these organisms, the concentration of the freely dissolved Plant Protection Product (PPP) is chosen 54 as the Ecotoxicologically Relevant Concentration (ERC). This GD also presents the Tier 1 effect 55 assessment procedure for sediment-dwelling organisms when based on water spiked water-sediment 56 toxicity tests. More information for sediment dwelling organisms will be provided in an opinion on the 57 effect assessment for plant protection products on sediment organisms in edge-of-field surface water 58 to be delivered next under this mandate. 59

To protect populations of aquatic organisms, effect assessment schemes are developed that allow to 60 derive Regulatory Acceptable Concentrations (RACs) on basis of two options, i.e. (1) the Ecological 61 Threshold Option (ETO), accepting negligible population effects only, and (2) the Ecological 62 Recovery Option (ERO), accepting some population level effects if ecological recovery takes place 63 within an acceptable time period. In the tiered acute and chronic effect assessment schemes, in 64 principle all tiers are able to address the Ecological Threshold Option, while the model ecosystem 65 approach (Tier 3) under certain conditions is able to also address the Ecological Recovery Option. 66

The GD is structured to give detailed guidance and provide relevant scientific background information 67 for each tier in the respective Chapters, which all end with a section on how to derive RACs and how 68 to perform the risk assessment including decision schemes. Chapter 5 describes the Tier 1 effect 69 assessment based on the revised data requirements. Chapter 6 addresses refinement options based on 70 additional species tested, i.e. the geomean approach and the species sensitivity distribution approach. 71 Chapter 7 addresses higher tier options based on refined exposure laboratory and model ecosystem 72 approaches. This includes guidance on selecting the appropriate refined exposure profiles, on refined 73 exposure laboratory tests, and on designing and evaluating model ecosystem (micro-/mesocosm) 74 studies. Chapter 8 contains detailed guidance on the possible use of non-testing methods, effect 75 assessment for metabolites, and assessment for formulations containing more than one active 76 substance. Chapter 9 addresses other relevant related issues. Chapter 10 provides guidance on 77 addressing the uncertainties in the assessment. 78

Chapter 11 provides an executive summary that joins all guidance and decision schemes in a concise 79 way without the detailed scientific background. This is intended to be helpful to applicants and 80 regulatory authorities providing an overview for day to day use. 81

The guidance was developed based on experience with currently known or approved active substances 82 and plant protection products. When using this guidance, it should be always checked whether the 83 proposed schemes are appropriate for active substances with a new mode of action. 84

85



86

87

KEY WORDS 88 Pesticides, formulations, metabolites, ecotoxicology, aquatic organisms, specific protection goals, regulatory 89 acceptable concentrations 90



TABLE OF CONTENTS 91

Summary .................................................................................................................................................. 1 92 Table of contents ...................................................................................................................................... 4 93 Background as provided by EFSA ........................................................................................................... 8 94 Terms of reference as provided by EFSA ................................................................................................ 9 95 Consideration ......................................................................................................................................... 11 96 1. Introduction ................................................................................................................................... 11 97

1.1. Reading guidance .................................................................................................................. 11 98 1.2. Legislative background ......................................................................................................... 11 99 1.3. Objectives of Guidance Document ....................................................................................... 12 100 1.4. Focus and restrictions of the GD........................................................................................... 13 101

1.4.1. Scope of Risk Assessment ................................................................................................ 13 102 1.4.2. Aquatic organisms living in the water column ................................................................. 13 103 1.4.3. Spatial Scale: edge-of-field- surface waters ..................................................................... 13 104 1.4.4. Use of effect modelling and combination to exposure modelling .................................... 13 105 1.4.5. Use of data on marine organisms...................................................................................... 13 106 1.4.6. Endocrine disruption section ............................................................................................ 13 107 1.4.7. FOCUS exposure assessment methodology ..................................................................... 14 108 1.4.8. Chemical and Biological Monitoring ............................................................................... 15 109 1.4.9. Permanent water bodies vs. water bodies falling temporarily dry .................................... 15 110 1.4.10. Active substances with new modes of action ................................................................... 15 111

2. The tiered approach, risk assessment terminology and linking exposure to effects ...................... 16 112 2.1. Introduction ........................................................................................................................... 16 113 2.2. The tiered approach ............................................................................................................... 16 114 2.3. Terminology in the aquatic risk assessment of PPPs ............................................................ 18 115 2.4. PPP effect assessment scheme .............................................................................................. 20 116 2.5. When to use the peak or a time-weighted average predicted environmental concentration 117 (PEC) in the risk assessment .............................................................................................................. 21 118

2.5.1. When and how (not) to use the PECsw;twa in chronic risk assessments ............................. 21 119 2.5.2. Decision scheme to use the PECsw;max or PECsw;twa in the risk assessment ....................... 22 120

3. Exposure assessment goals and specific protection goals for water organisms ............................ 25 121 3.1. Introduction ........................................................................................................................... 25 122 3.2. The Ecotoxicologically Relevant Concentration (ERC) ....................................................... 25 123 3.3. Exposure assessment goals in edge-of-field surface waters ................................................. 26 124 3.4. Specific Protection Goals for Water Organisms ................................................................... 26 125 3.5. Specific Protection Goal options for aquatic key drivers in edge-of-field surface water ..... 27 126

3.5.1. SPG proposal for algae (e.g. green algae, diatoms, blue-greens) in edge-of-field surface 127 water .......................................................................................................................................... 28 128 3.5.2. SPG proposal for aquatic vascular plants (e.g. dicotyledonous, monocotyledonous) in 129 edge-of-field surface water ............................................................................................................ 29 130 3.5.3. SPG proposal for aquatic invertebrates in edge-of-field surface water (e.g. crustaceans, 131 rotifers, insects, oligochaete worms, molluscs) ............................................................................. 29 132 3.5.4. SPG proposal for aquatic vertebrates in edge-of-field surface water (e.g. fish, 133 amphibians) ................................................................................................................................... 29 134 3.5.5. SPG proposal for aquatic microbes (e.g. bacteria, fungi) ................................................. 30 135

3.6. Vulnerable species ................................................................................................................ 30 136 3.7. Implementation of the specific protection goals in this guidance document ........................ 32 137

4. Exposure Assessment .................................................................................................................... 33 138 4.1. Introduction ........................................................................................................................... 33 139 4.2. FOCUS surface water scenarios and models ........................................................................ 33 140

4.2.1. Description of the different tiers ....................................................................................... 34 141 4.2.1.1. Step 1 ....................................................................................................................... 34 142 4.2.1.2. Step 2 ....................................................................................................................... 36 143 4.2.1.3. Step 3 ....................................................................................................................... 38 144



4.2.1.4. Step 4 ....................................................................................................................... 45 145 4.2.2. Assessment of metabolites by FOCUS surface water modelling .................................... 46 146

5. Data Requirement for active substances and Formulations and Tier 1 effect assessment ............. 47 147 5.1. Introduction to data requirements as laid down in SANCO document 11802 and 148 11803/2012 for approval of active substances and plant protection products and related OECD 149 guidelines ........................................................................................................................................... 47 150 5.2. Standard toxicity tests with aquatic organisms ..................................................................... 48 151

5.2.1. Fish ................................................................................................................................... 49 152 5.2.1.1. Acute toxicity to fish ................................................................................................ 49 153 5.2.1.2. Chronic toxicity to fish ............................................................................................ 50 154

5.2.2. Amphibians ....................................................................................................................... 50 155 5.2.3. Aquatic invertebrates ........................................................................................................ 50 156

5.2.3.1. Toxicity studies with sediment dwelling organisms ................................................ 51 157 5.2.4. Standard toxicity tests with algae ..................................................................................... 52 158 5.2.5. Standard toxicity tests with macrophytes ......................................................................... 52 159

5.3. Deriving regulatory acceptable concentrations (RAC) ......................................................... 53 160 5.4. Further testing on aquatic organisms .................................................................................... 53 161 5.5. Specific requirements for formulated products ..................................................................... 54 162 5.6. Bioconcentration and secondary poisoning .......................................................................... 54 163

5.6.1. Bioconcentration in fish .................................................................................................... 55 164 5.6.2. Secondary poisoning ......................................................................................................... 55 165 5.6.3. Regulatory acceptable concentration (RAC) for biomagnification .................................. 55 166

6. Higher-tier effect assessment on basis of laboratory toxicity tests with standard and additional 167 species .................................................................................................................................................... 57 168

6.1. Introduction ........................................................................................................................... 57 169 6.1.1. Quality check .................................................................................................................... 57 170

6.2. Geometric mean-AF approach .............................................................................................. 57 171 6.2.1. Introduction ...................................................................................................................... 57 172 6.2.2. Approaches considered by EFSA ..................................................................................... 58 173 6.2.3. Derivation acute and chronic RAC ................................................................................... 59 174

6.3. The Species Sensitivity Distribution (SSD) approach .......................................................... 60 175 6.3.1. Introduction to SSD approach .......................................................................................... 60 176 6.3.2. Criteria for the selection of toxicity data to construct SSDs ............................................. 61 177 6.3.3. Selecting toxicity data on basis of toxic mode-of-action of the substance ....................... 63 178

6.3.3.1. Insecticide SSDs ...................................................................................................... 63 179 6.3.3.2. Herbicide SSDs ........................................................................................................ 64 180 6.3.3.3. Fungicide SSDs ........................................................................................................ 65 181

6.3.4. Derivation of acute and chronic RACs for invertebrates and primary producers ............. 66 182 6.3.5. Derivation of acute and chronic RACs for fish/amphibians ............................................. 68 183

7. Higher-tier effect assessment by means of refined-exposure laboratory toxicity tests and 184 experimental ecosystems ........................................................................................................................ 70 185

7.1. Selecting the appropriate exposure regimes when addressing time-variable exposures in 186 higher-tier effect studies..................................................................................................................... 70 187

7.1.1. Introduction ...................................................................................................................... 70 188 7.1.2. Use of predicted exposure profiles for edge-of-field surface waters in higher-tier effect 189 assessments .................................................................................................................................... 71 190 7.1.3. Toxicological (in)dependence of different pulse exposures ............................................. 72 191 7.1.4. The minimum number of toxicological dependent pulse exposures to address in higher-192 tier effect studies. ........................................................................................................................... 72 193 7.1.5. Ecological (in)dependence of different pulse exposures .................................................. 73 194

7.2. Refined exposure laboratory toxicity tests ............................................................................ 75 195 7.2.1. Introduction ...................................................................................................................... 75 196 7.2.2. Reasons to perform refined exposure laboratory toxicity test .......................................... 75 197 7.2.3. Refined exposure tests with standard test species ............................................................ 77 198 7.2.4. Refined exposure tests with additional test species .......................................................... 77 199



7.2.5. Derivation of RAC and calibration of refined exposure laboratory toxicity tests ............ 77 200 7.3. Model ecosystem experiments .............................................................................................. 79 201

7.3.1. Introduction ...................................................................................................................... 79 202 7.3.2. Designing micro-/mesocosm experiments ........................................................................ 79 203

7.3.2.1. Establishment of a representative aquatic community in the test systems ............... 80 204 7.3.2.2. Selection and characterisation of the exposure regime ............................................ 82 205 7.3.2.3. Number of treatments, choice of the doses and replicate test systems per treatment83 206 7.3.2.4. Measurement endpoints ........................................................................................... 83 207 7.3.2.5. Statistical and ecological evaluation of concentration-response relationships ........ 85 208

7.3.3. Interpreting micro-/mesocosm experiments ..................................................................... 86 209 7.3.3.1. Evaluation of the scientific reliability of the micro-/mesocosms test for PPP 210 authorisation .............................................................................................................................. 87 211

7.3.4. Variability in concentration-response patterns between micro/mesocosm experiments 212 exposed to the same PPP ............................................................................................................... 90 213

7.3.4.1. Short-term pulsed exposure ..................................................................................... 90 214 7.3.4.2. Long-term exposure to the same PPP ...................................................................... 91 215

7.3.5. How to derive a RAC from an appropriate micro-/mesocosm experiment and how to link 216 it to PEC......................................................................................................................................... 92 217

7.3.5.1. Selecting and extrapolating micro-/mesocosm results ............................................. 93 218 7.3.5.2. Peak, nominal or TWA concentrations of RAC and PEC used for risk assessment 93 219 7.3.5.3. Deriving a RAC indicative for the ecological threshold option (ETO-RAC) .......... 94 220 7.3.5.4. Deriving a RAC on basis of ecological recovery option (ERO-RAC) .................... 96 221

8. Non-testing methods, metabolites, impurities and formulations with more than one active 222 substance ................................................................................................................................................ 99 223

8.1. Non-testing methods ............................................................................................................. 99 224 8.1.1. Area of use ........................................................................................................................ 99 225 8.1.2. Guidance on (Q)SAR ...................................................................................................... 100 226

8.1.2.1. Model validity ........................................................................................................ 100 227 8.1.2.2. Reliability and adequacy of (Q)SAR prediction .................................................... 101 228

8.1.3. Available (Q)SAR methods, expert systems and read across ......................................... 102 229 8.1.4. Comparison of QSAR model outputs ............................................................................. 105 230 8.1.5. Use of non-testing data in PPP risk assessment .............................................................. 106 231 8.1.6. Decision scheme for use of non-testing systems ............................................................ 107 232

8.2. Metabolites and degradation products ................................................................................ 107 233 8.2.1. Introduction .................................................................................................................... 107 234 8.2.2. Definition of the residue for risk assessment .................................................................. 107 235 8.2.3. Risk assessment scheme for metabolites ........................................................................ 109 236 8.2.4. Alternative information replacing experimental studies ................................................. 110 237 8.2.5. Identification of toxophore ............................................................................................. 111 238 8.2.6. Metabolites structurally similar to the active ingredient and with remaining toxophore.111 239 8.2.7. Metabolites with no toxophore ....................................................................................... 112 240 8.2.8. Non-testing predictions of metabolite toxicity ............................................................... 112 241 8.2.9. Toxicity testing with metabolites.................................................................................... 112 242 8.2.10. Risk Assessment for Metabolites.................................................................................... 113 243 8.2.11. Definition of the residue for monitoring ......................................................................... 113 244

8.3. Combinations of a.s. in formulations (guidance on toxic unit approaches) ........................ 114 245 9. Other Issues ................................................................................................................................. 116 246

9.1. Test batches/impurities ....................................................................................................... 116 247 9.2. Testing poorly soluble and other difficult test substances .................................................. 116 248 9.3. Promising mechanistic effect models ................................................................................. 117 249 9.4. Reduction of (vertebrate) testing......................................................................................... 117 250

9.4.1. Use of limit tests ............................................................................................................. 117 251 9.4.2. Use of non-testing methods ............................................................................................ 118 252

9.5. Differences in risk assessment procedures between Regulation (EC) 1107/2009 and the 253 Water Framework Directive (WFD) ................................................................................................ 118 254



9.5.1. Introduction .................................................................................................................... 118 255 9.5.2. Overview of main differences in risk assessment procedures between Plant Protection 256 Product Regulation and Water Framework Directive .................................................................. 118 257

10. Addressing uncertainties ......................................................................................................... 120 258 10.1. Approaches for characterising uncertainty in higher-tier assessments ............................... 120 259 10.2. Risk characterisation and weight-of evidence assessment .................................................. 122 260 10.3. Uncertainties in extrapolating to real field situations ......................................................... 125 261

11. Executive Summary ................................................................................................................ 128 262 11.1. Aquatic Risks due to toxicity .............................................................................................. 128 263

11.1.1. Introduction .................................................................................................................... 128 264 11.1.2. Summary flow charts for acute and chronic effect/risk assessment ............................... 129 265 11.1.3. Tier 1 RACsw derivation on basis of standard test species ............................................. 132 266 11.1.4. Tier 2 RACsw derivation on basis of additional laboratory toxicity tests ....................... 133 267 11.1.5. Tier 2A: The Geomean-AF approach ............................................................................. 134 268 11.1.6. Tier 2B: The Species Sensitivity Distribution (SSD) approach ..................................... 134 269 11.1.7. Tier 2C: Refined Exposure Laboratory test-AF approach .............................................. 136 270 11.1.8. Tier 3 RACsw derivation on basis of micro-/mesocosm tests ......................................... 137 271

11.2. Bioconcentration and Secondary Poisoning ....................................................................... 141 272 11.3. Non-testing methods ........................................................................................................... 142 273 11.4. Metabolites and degradation products ................................................................................ 143 274 11.5. Combinations of a.s. in formulations (guidance on toxic unit approaches) ........................ 145 275

References ............................................................................................................................................ 147 276 Appendices ........................................................................................................................................... 162 277 A. Elements of the Exposure Assessment Goals related to the choices made in the FOCUSsw 278 scenarios ............................................................................................................................................... 162 279 B. Background of the procedure for partitioning of substance between water and sediment in the 280 FOCUSsw Step2 exposure calculations ................................................................................................. 167 281 C. Comparison of acute rainbow trout toxicity with acute toxicity values for amphibian species .. 169 282 D. Information on Life Cycle Characteristics for Aquatic Organisms ............................................. 175 283 E. Variability in exposure-response relationships between micro-/mesocosm experiments performed 284 with the same PPP ................................................................................................................................ 176 285 Glossary and abbreviations .................................................................................................................. 180 286

287



BACKGROUND AS PROVIDED BY EFSA 288

Member States’ competent authorities were requested by the Director of Sciences of the European 289 Food Safety Authority (EFSA) on 3 July 2006 via the Standing Committee on the Food Chain and 290 Animal Health, to send EFSA a priority list of existing Guidance Documents to be revised and 291 proposals for development of new ones. Answers were received from 15 Member States. 292

Regarding the revision of the Guidance Document on Aquatic Ecotoxicology (SANCO/3268/2001, 293 rev. 4 final, 17 October 2002), five detailed requests were received (FI, DE, NL, DK, SE) highlighting 294 the importance of liaising with the revision of Annex II and Annex III. 295

In 2006 and 2007, EFSA has issued six opinions on the Annexes II and III, two of which related to the 296 ecotoxicological studies (EFSA, 2007a)and the fate and behaviour in the environment (EFSA, 2007b). 297 The rapporteur (UK) has taken these opinions on board in the revision of the Annexes, which are 298 currently with the Commission. It should be considered to generally revise the structure and content of 299 the available Guidance Documents. 300

Member States highlighted the following issues as being particularly important: 301

• More clarity regarding the data requirements for substances expected to be endocrine 302 disrupters is needed; 303

• More guidance should be provided regarding the use of FOCUSSW modelling, e.g. on input 304 parameters or the use of Step 4; 305

• Need for revision in particular with regard to the protection level in adjacent small ditches and 306 main watercourses (in line with the requirements of the Water Framework Directive); 307

• More integrated development of the assessment of exposure modelling and effects; 308

• Conceptual consistency between higher tier assessments in aquatic and terrestrial 309 ecotoxicology needed; 310

• More guidance regarding the assessment of higher tier aquatic studies (assessment of addition 311 of sediments, assessment of quality and quantity of mesocosm studies, assessment of 312 ecotoxicological field studies trigger levels for higher tier studies); 313

• Harmonized endpoints for authorization of plant protection products needed; 314

• A clear and transparent relationship with the Water Framework Directive is wished for. 315

The EFSA PRAPeR Unit emphasised that the aquatic GD needs to be updated regarding the long-term 316 RA to take account of the new exposure data that are the outcome of the FOCUS models. The 317 interaction between exposure and effects need some more guidance. Of course also possible new data 318 requirements in the new regulation that will replace Council Directive 91/414/EEC need to be taken 319 up in the existing GD. 320

Relevant topics and scientific principles of already existing scientific opinions elaborated by the PPR 321 Panel will also be incorporated into the revised Guidance Document. Further, on-going work in other 322 fora, pertinent to the GD will be closely monitored and taken into account where relevant. 323

The public was consulted on the existing GD in October – December 2008 and comments and ideas 324 for the revision by stakeholders will be taken into account during the process. Also comments from a 325 risk manager survey performed October – December 2008 are considered. Furthermore, the activity 326 performed under EFSA-Q-2009-00861 to develop specific protection goals will be used as input to 327 this updated mandate. 328 329



TERMS OF REFERENCE AS PROVIDED BY EFSA 330

EFSA tasks its Scientific Panel on Plant Protection Products and their Residues (PPR Panel) to prepare 331 a revision of the Guidance Document on Aquatic Ecotoxicology under Council Directive 91/414/EEC 332 (SANCO/3268/2001 rev.4 (final), 17 October 2002). 333 334 The Panel is asked to develop a Guidance Document and two Scientific Opinions, as summarised 335 below: 336 337 1. Guidance Document on tiered risk assessment for aquatic organisms in edge-of-field surface 338

waters (by July 2013). 339 340 In particular, the following issues need to be addressed: 341

• Update the current guidance in view of the new Regulation 1107/2009 342

• Update the current guidance in view of the revised data requirements to Regulation 1107/2009 343

• Develop guidance on first tier aquatic effect assessment 344

• Develop guidance on higher tier aquatic effect Assessment (based on laboratory studies and 345 model ecosystem studies, guidance on design and evaluation of higher tier studies) 346

• Guidance on appropriate linking of aquatic exposure and effect assessment 347

This PPR Panel Guidance should be subject to a Public Consultation. 348

349 2. Scientific Opinion of the PPR Panel on the effect assessment for pesticides on sediment 350

organisms in edge-of-field surface waters (2 years after acceptance of the revised mandate, i.e. 351 October 2014) 352 353 A scientific opinion will be provided that describes the state of the art of effect assessment for 354 sediment organisms. 355 356 In particular the following issues will be addressed: 357

358

• Identification of standard test species 359

• Use of the geometric mean approach when toxicity data for a limited number of additional 360 test species are available 361

• Use of Species Sensitivity Distribution approach for sediment organisms 362

• Use of the model ecosystem approach for sediment organisms 363

• Defining the ecotoxicologically relevant concentrations (ERCs) for acute and chronic risk 364 assessment 365

366 3. Scientific Opinion on the state of mechanistic effect modelling approaches for regulatory 367

risk assessment of pesticides for aquatic organisms (3.5 years after acceptance of the revised 368 mandate, i.e. April 2016 ) 369 370 A scientific opinion will be provided that describes the state of the art of mechanistic effect 371 modelling in the aquatic environment. 372 373



In particular the following state of the art of the following types of models will be addressed (for 374 all aquatic water column and sediment dwelling organisms): 375 376

• Describe regulatory questions that can be addressed by effect modelling 377

• Describe model parameters that need to be included in relevant models and that need to be 378 checked in evaluating the acceptability of effect models 379

• Describe available effect models for aquatic organisms, in particular 380

o Toxicokinetic / toxicodynamic models 381

o Mechanistic population models 382

o Mechanistic food web models 383 Secondary poisoning 384 Ecosystem models representative for ditches, ponds and streams 385

• Selection of focal species 386

• Development of ecological scenarios that can be linked to the regulatory defined water bodies 387 in the climatic zones of Europe 388

389 390 This Guidance Document addresses the first part of the Terms of Reference, the two scientific 391 opinions outlined above will follow later. 392

393



CONSIDERATION 394

395

1. Introduction 396

1.1. Reading guidance 397

The GD is structured to give detailed guidance and provide relevant scientific background information 398 for each tier in the respective Chapters, which all end with a section on how to derive RACs and how 399 to perform the risk assessment including decision schemes. Chapter 1 describes the focus and 400 restrictions of the GD. Chapter 2 introduces the tiered approach, terminology and linking of exposure 401 to effects. Chapter 3 describes the specific protection goals. Chapter 4 addresses the exposure 402 assessment according to the current FOCUS surface water methodology. Chapter 5 describes the Tier 403 1 effect assessment based on the revised data requirements. Chapter 6 addresses refinement options 404 based on additional species tested, i.e. the geomean approach and the species sensitivity distribution 405 approach. Chapter 7 addresses higher tier options based on refined exposure laboratory and model 406 ecosystem approaches. This includes guidance on selecting the appropriate refined exposure profiles, 407 on refined exposure laboratory tests, and on designing and evaluating model ecosystem (micro-408 /mesocosm) studies. Chapter 8 contains detailed guidance on the possible use of non-testing methods, 409 effect assessment for metabolites, and assessment for formulations containing more than one active 410 substance. Chapter 9 addresses other relevant related issues. Chapter 10 provides guidance on 411 addressing the uncertainties in the assessment. 412

Chapter 11 provides an executive summary that joins all guidance and decision schemes in a concise 413 way without the detailed scientific background. This is intended to be helpful to applicants and 414 regulatory authorities providing an overview for day to day use. 415

1.2. Legislative background 416

In 2008, the PPR Panel was tasked by EFSA to revise the Guidance Documents (GDs) for Aquatic 417 Ecotoxicology (SANCO/3268/2001 rev 4 final) and Terrestrial Ecotoxicology (SANCO/10329/2002 418 rev. 2 final), which were used in the routine risk assessment of plant protection products in the context 419 of Directive 91/414/EEC. The replacement of Directive 91/414/EEC by Regulation (EC) No 420 1107/2009 (hereafter the Regulation) in June 2011 called for revision of the existing guidance 421 documents in order to take on board new elements in the environmental risk assessment, e.g. cut-off 422 criteria, biodiversity and new endpoints. 423

Moreover, Commission Regulations laying down the data requirements for the dossier to be submitted 424 for the approval of active substances contained in plant protection products (SANCO document 425 11802/2012 4;5) and for the approval of plant protection products (SANCO documents 11803/20129;6) 426 were revised in 2012 to be in line with the new Regulation (EC) 1107/2009. These documents provide 427 information on the basic data requirements for the authorisation of plant protection products. 428

Article 8(5) of Regulation (EC) No 1107/2009 includes a legal obligation to submit scientific peer-429 reviewed open literature data. EFSA has provided guidance (EFSA, 2011) which provides a definition 430 of scientific peer-reviewed open literature. EFSA (2011) also provides guidance on how to identify, 431 select and include scientific peer-reviewed open literature, and how to report the literature search and 432 selection process for the purposes of a dossier. Regulation (EC) 1107/2009 and the Uniform Principles 433 4 The revised data requirements are not yet published and therefore reference is made to the Commission proposals that were

taken note of by the SCFCAH in July 2012: Commission proposal on new data requirements for approval of active substances (SANCO document 11802/2012) and Commission proposal on new data requirements for of plant protection products (SANCO documents 11803/2012). References to the final Regulations on data requirements will be updated after their publication.

5Revision of former Annex II in Directive 91/414/EEC and Regulation (EC) 544/2011 6Revision of former Annex III in Directive 91/414/EEC and Regulation (EC) 545/2011



(Regulation (EC) 546/20117) include the decision-making criteria for the approval of active 434 substances, safeners and synergists at EU level and authorisation of plant protection products at 435 Member State level. A procedure to derive specific protection goals on basis of aquatic key drivers 436 and ecological entities to be protected and the magnitude of the tolerable effect (including its spatio-437 temporal dimension) was proposed by EFSA (2010b) (see also Chapter 3). 438

1.3. Objectives of Guidance Document 439

This Guidance Document (GD) is the first deliverable within the PPR Panel mandate of the revision of 440 the former GD on Aquatic Ecotoxicology (SANCO/3268/2001 rev.4 (final), EC, 2002), EFSA-M-441 2009-0001. An overview of the three deliverables under this mandate is given below: 442

1. Guidance of the PPR Panel on tiered risk assessment for aquatic organisms in edge-of-field 443 surface waters 444

2. Scientific Opinion of the PPR Panel on the effect assessment for pesticides on sediment organisms 445 in edge-of-field surface waters 446

3. Scientific Opinion on the state of mechanistic effect modelling approaches for regulatory risk 447 assessment of pesticides for aquatic organisms 448

The Guidance presented in this document is intended to replace the current GD on Aquatic 449 Ecotoxicology (EC, 2002). The scientific opinions to be delivered as well under the mandate will look 450 in detail at some specific aspects that are outlined in the section on Terms of Reference above. The 451 further information might be incorporated, if needed, in a future review of this GD. 452

A revision of the former GD on Aquatic Ecotoxicology (EC, 2002) became necessary for several 453 reasons, with the main ones listed below: 454 455 1. The requirements of the new Regulation (EC) 1107/2009 and underlying Annexes (EC, 2009) 456

2. The new Regulations laying down the data requirements for the dossiers to be submitted for the 457 approval of active substances contained in plant protection products (SANCO document 458 11802/2012) and for the authorisation of plant protection products (SANCO document 459 11803/2012) 460

3. Recommendations of scientific opinions of the PPR Panel of EFSA with respect to aquatic risk 461 assessment procedures (e.g. EFSA 2005a, 2005b, 2006, 2009a; 2010b) 462

4. Concerns formulated by risk assessors and risk managers of EU Member States and other 463 stakeholders with respect to under- or over-protectiveness of the current aquatic risk assessment 464 procedures and/or risk assessment issues where there is scope for different interpretations (EFSA, 465 2009b) 466

5. Relevant issues identified during the PPP Risk Assessment Peer Review of active substances 467

6. Uncertainties in the current FOCUS surface water scenarios (FOCUS, 2001) and exposure 468 modelling 469

7. Recommendations of recent workshops (e.g. ELINK, Brock et al. 2010c; AMRAP, Maltby et al. 470 2010; LEMTOX, Thorbek et al. 2010; Ampere 2007) 471

8. State-of-the-art new knowledge on the ecotoxicology of PPPs as published in the scientific 472 literature (particularly for PPPs with novel toxic modes-of-action). 473

This GD on the tiered risk assessment (RA) scheme for aquatic organisms in edge-of-field surface 474 waters is intended to provide guidance for applicants and Member States in the context of the 475 authorisation of Plant Protection Products (PPPs) and their active substances under Regulation (EC) 476 1107/2009 (EC, 2009). 477

7Former Annex VI in Directive 91/414/EEC



1.4. Focus and restrictions of the GD 478

1.4.1. Scope of Risk Assessment 479

The GD is intended for use in the risk assessment of single substances or combinations of active 480 substances in one formulation. When a compound or formulation is applied more than once in a 481 growing season, the number of applications is taken into account in the RA. 482

1.4.2. Aquatic organisms living in the water column 483

This GD has its focus on a tiered risk assessment procedure for aquatic organisms living in the water 484 column in edge-of-field surface waters. Nevertheless, also a preliminary (Tier 1) risk assessment 485 procedure for sediment-dwelling organisms on basis of the 28-d water-spiked water-sediment test with 486 Chironomus riparius or Lumbriculus spp. is incorporated, since this concerns a data requirement under 487 the PPP regulation. A later PPR Scientific Opinion in the series mentioned above will deal in detail 488 with the effect assessment for sediment-dwelling organisms by paying attention to a wider array of 489 sediment-dwelling species. 490

1.4.3. Spatial Scale: edge-of-field- surface waters 491

Edge-of-field surface water bodies (ditches, streams and ponds) are defined as surface water systems 492 that are as close to a treated field as possible according to agricultural practice. This GD has its focus 493 on aquatic organisms in edge-of-field surface waters, since appropriate exposure and effects 494 assessment tools for surface waters further downstream and that address environmental risks of PPPs 495 at a larger spatial scale (e.g. catchment) are mainly a research activity to date and new tools and 496 knowledge is expected to become available in the next years only. In section 9.5, however, a short 497 description is given on the main differences in risk assessment procedure between the PPP Regulation 498 and the WFD, and on the basic data requirements that should be provided for PPPs by the applicant to 499 derive WFD Environmental Quality Standards. 500

1.4.4. Use of effect modelling and combination to exposure modelling 501

Promising effect models at different levels of biological organisation are currently under development 502 in different research projects. Most of these modelling approaches are not yet ready for use in 503 regulatory risk assessment. Further guidance on effect modelling (including toxicokinetic / 504 toxicodynamic (TK/TD) modelling, population and food web models) and combined effect and 505 exposure modelling approaches will be provided in a future scientific opinion of the PPR Panel of 506 EFSA. 507

1.4.5. Use of data on marine organisms 508

The data requirements for PPPs mainly focus on data for freshwater organisms for the effect 509 assessment. Only in some cases data for marine species (i.e. Americamysis bahia) are requested. Also 510 the exposure assessment is performed for freshwater only, due to the focus on edge of field surface 511 waters and the available exposure assessment tools. This GD, however, provides some 512 recommendations on how to use additional toxicity data for marine organisms in combination with 513 toxicity data for freshwater organisms (Chapter 6). 514

1.4.6. Endocrine disruption section 515

As set out in Annex II, the new Regulation (EC) 1107/2009 makes specific provision only to approve 516 an active substance, safener or synergist “if it is not considered to have endocrine disrupting properties 517 that may cause adverse effects on non-target organisms unless the exposure […] is negligible” (section 518 3.8.2). It follows from this wording that the regulation takes a hazard-based approach to approval 519 where endocrine disrupters are concerned. 520

Test methods currently required in the data requirement for environmental effects in the PPP 521 Regulation are not designed specifically for identification of endocrine disrupters. However the OECD 522



conceptual framework for testing and assessment of potential endocrine disrupters (OECD 2012c) 523 includes several test guidelines (in levels 3, 4 and 5) designed for identification of endocrine 524 disrupting properties relevant to wildlife. Of relevance to the current Guidance on aquatic toxicity, 525 these include the fish short-term reproduction assay (TG229), 21d fish screening assay (TG230), 526 amphibian metamorphosis assay (TG231), and androgenised female stickleback screen (OECD GD 527 140) at level 3 (in vivo tests providing data about selected endocrine mechanisms/pathways), the fish 528 sexual development test at level 4 (in vivo assays providing data on adverse effects on endocrine-529 relevant endpoints), and the fish lifecycle toxicity test (US EPA OPPTS 850.1500) at level 5 (in vivo 530 assays providing more comprehensive data on adverse effects on endocrine-relevant endpoints over 531 more extensive parts of the life cycle of the organism). OECD 229-231 and 234 also referred to in 532 Commission Communication (SANCO/11844/2010 Publication of test methods and guidance 533 documents under the Regulation). 534

Robust scientific criteria for identification of substances with ‘endocrine disrupting properties’ are 535 therefore needed to inform updated data requirements in support of this provision of the regulation, 536 and by 14 December 2013 the Commission must present a draft proposal for such criteria. Pending 537 these criteria, substances identified under Regulation (EC) 1272/2008 (Classification and Labelling) as 538 C2 and R2, or R2 with evidence of toxicity to endocrine organs, shall be considered to have endocrine 539 disrupting properties, with respect to human health assessment. There is currently no provision for 540 identifying endocrine disruptors within the environmental effects assessment required for the 541 regulation. 542

In support of future inclusion of such provisions in the PPP Regulation and other pieces of EU 543 legislation, EFSA has recently been mandated to elaborate a scientific opinion on this issue, while 544 ensuring involvement of other relevant agencies (e.g. ECHA, EMA) and the Commissions Scientific 545 Committees (SCHER,SCCS and SCENIHR). The opinion, currently being addressed by a Working 546 Group of EFSA’s Scientific Committee, will take stock of the following: 547

• existing scientific criteria for identification of endocrine disrupters 548

• what is an adverse effect (explicit in widely accepted definitions of an endocrine disrupter e.g. 549 Weybridge and WHO/IPCS) and what is physiological modulation? 550

• are existing toxicity test methods appropriately covering effects of endocrine active 551 substances? 552

Given the fact that the EFSA activity on this issue, and similar activities initiated by DG Environment 553 (a ‘Member States Ad hoc Group on scientific aspects related to identification of endocrine disruptors’ 554 (regulatory oriented) and an ‘Advisory Group to provide advice to the European Commission on 555 scientific issues relevant to criteria for the identification of endocrine disrupters’ (scientific oriented, 556 lead by JRC), are ongoing, it is not possible to elaborate guidance on data requirements and their 557 interpretation/next steps in this Guidance Document. The impact of the Regulation (EC) 1107/2009 558 provisions for non-approval of endocrine disrupters on the aquatic toxicity effects assessment will 559 therefore be accommodated in future revisions of this guidance document, and other modules (e.g. 560 terrestrial toxicity), once robust scientific criteria are agreed and available. 561

1.4.7. FOCUS exposure assessment methodology 562

Assessment of risks to organisms is always a combination of an effect assessment and an exposure 563 assessment (EFSA, 2010b). This GD is almost exclusively limited to the effect assessment. The 564 current exposure assessment is based on FOCUS (2001, 2006, 2007a, 2007b). The level of protection 565 achieved by the current FOCUS surface water exposure assessment methodology is unknown since it 566 has not been reviewed during the revision of the GD on Aquatic Ecotoxicology by the PPR Panel of 567 EFSA and exposure assessment goals have not yet been defined for surface water (see also section 3.3 568 and Appendix A). However, the methodology has been used in regulatory decision making throughout 569 the last years and there is currently no alternative standardised exposure assessment methodology. 570 Therefore, it is assumed that the FOCUSsw methodology will continue to be used until updated or new 571



methods become available and adopted by the SCFCAH and will replace the existing tools. FOCUSsw 572 is used for approval of active substances at EU level. It is also used in some Member States for 573 product authorisation, but also different exposure assessment procedures may be used. 574

1.4.8. Chemical and Biological Monitoring 575

Although chemical and biological monitoring could be helpful for validating the exposure and effect 576 predictions within the risk assessment framework, this document does not contain guidance for 577 monitoring. Setting up an appropriate monitoring programme requires a clear definition of both the 578 exposure assessment goals and the ecotoxicological protection goals. The exposure assessment goals 579 have, however, not yet been defined by the SCFCAH. Even if the exposure assessment goals would 580 have been defined, validating the current exposure assessment methodology is not possible since it is 581 not known to which percentile of the statistical population of concentrations the exposure predictions 582 correspond (see previous section). Despite this, the Panel acknowledges the importance of monitoring, 583 for example for the assessment of persistent substances in soil a Tier 5 monitoring approach was 584 recently proposed by the Panel (EFSA Panel on Plant Protection Products and their Residues (PPR), 585 2010a, EFSA 2011). The Panel therefore proposes to develop guidance on monitoring as soon as the 586 current exposure assessment procedure has been reviewed. 587

1.4.9. Permanent water bodies vs. water bodies falling temporarily dry 588

The aquatic exposure assessment is currently performed using FOCUS surface water tools (FOCUS 589 2001), which are based on water bodies (ditches, streams and ponds) with a minimum water depth of 590 30 cm. It can be expected that the annual peak concentration is strongly influenced by this minimum 591 water depth because water depths may be close to the minimum depth during applications of PPPs in 592 spring and summer and the spray drift during these applications may lead to very high predicted 593 environmental concentration (PEC) values for shallow water depths. Therefore, whenever the current 594 exposure assessment methodology will be reviewed, a range of minimum water depths might need to 595 be considered. The ultimate choice of whether such assessment is necessary falls under the 596 responsibility of risk managers. 597 598 The standard effect assessment is based on test organisms and studies that are not designed to 599 specifically cover organisms occurring in water bodies falling temporarily dry. No detailed data is 600 available about potential differences in the sensitivity to PPP exposure of those species compared to 601 those used in the standard risk assessment procedure. Temporary small standing water bodies are 602 known to have characteristic plant and animal communities depending strongly on the specific 603 hydrological conditions. A uniqueness is reported for the communities of plants (e.g. with particular 604 protected fern species) and of invertebrates, which often is determined by the absence of fish as 605 predators. Also endangered faunal groups such as amphibians and branchiopod crustaceans might be 606 particularly abundant in these water bodies (EEA, 2009). Organisms occurring predominantly in 607 temporary ponds are known to be uniquely adapted to the changing environmental conditions, 608 following strategies like dormancy and dispersal to survive. By these adaptations the sensitivity to 609 toxicants of such organisms as well as their potential to recover might be affected (Lahr 1997). 610

Based on the data requirements and standard ecotoxicological tests available, this GD predominantly 611 addresses the risk for organisms occurring in permanent edge-of-field water bodies. 612

1.4.10. Active substances with new modes of action 613

The following guidance and risk assessment scheme are generally recommended for use in the 614 authorisation process of active substances and formulated plant protection products. However, for each 615 compound it should be carefully evaluated that the proposed steps are addressing all relevant questions 616 related to the individual properties of the compound under evaluation. Especially for compounds of 617 new mode of action, a thorough analysis has to be done whether additional or different questions need 618 to be tackled and the scheme proposed here is appropriate. 619

620



2. The tiered approach, risk assessment terminology and linking exposure to effects 621

2.1. Introduction 622

The aquatic risk assessment procedure for PPPs in edge-of-field surface waters consists of two parts; 623

• Exposure assessment, which is the domain of experts in environmental chemistry and 624 exposure modelling, and 625

• Effect assessment, which is the domain of experts in ecotoxicology. 626 627

Within the authorisation procedure of PPPs in the EU, relevant exposure concentrations in edge-of-628 field surface waters are obtained by adopting the exposure assessment endpoints described in Chapter 629 3 and by applying exposure scenarios and fate models to derive predicted environmental 630 concentrations (PECs). For prospective exposure assessment harmonised approaches have been 631 developed that enable to predict realistic worst-case exposure concentrations in edge-of-field waters. 632 These are documented in the “FOCUS Surface Water Scenarios” report (FOCUS, 2001) and are 633 further discussed in Chapter 4. 634

Prospective effects assessment relies on the available specific protection goals (Chapter 3) and 635 relevant ecotoxicological and ecological data. The ecotoxicological data usually concern 636 concentration–response relationships derived from controlled experiments with e.g. standard (Chapter 637 5) and additional aquatic test species (Chapter 6) or refined exposure and micro/mesocosm tests 638 (Chapter 7). The ecological data usually relate to the ‘target image’ of the aquatic community in the 639 relevant surface waters, including ecological traits of the important aquatic species at risk. Examples 640 of ecological scenarios for European streams, ditches and ponds are presented in the ELINK document 641 (Brock et al. 2010c). Assessment factors are usually used to extrapolate the experimental 642 concentration–response relationships in space and time to derive regulatory acceptable concentrations 643 (RACs). In principle, this conventional extrapolation approach may be replaced and/or adjusted by 644 appropriate modelling approaches, which will be addressed in a later scientific opinion of the PPR 645 Panel. 646

2.2. The tiered approach 647

Ideally, when many scientifically underpinned methods are available and costs are not a limiting 648 factor, environmental risk assessments can be performed by applying the best available methods. 649 However, in practice environmental risk assessments are not based on an unlimited number of 650 environmental fate and ecotoxicity data but on factors like pragmatism, costs, and efficacy. When both 651 pragmatism and science drive the assessment, one can understand the development of tiered systems 652 (Posthuma et al. 2008). 653

Tiered approaches are the basis of environmental risk assessment schemes that support the registration 654 of PPPs under the PPP Regulation (see e.g. Campbell et al. 1999; EC 2002; Boesten et al. 2007; EFSA 655 Panel on Plant Protection Products and their Residues (PPR) 2010b). In this context, a tier is defined 656 as a complete effect or exposure assessment resulting in an appropriate assessment endpoint, e.g. PEC 657 (Predicted Environmental Concentration) or RAC (Regulatory Acceptable Concentration). The 658 concept of tiered approaches is to start with a simple conservative assessment and to only do 659 additional more complex work if necessary (so it implies a cost-effective procedure both for industry 660 and regulatory agencies). Note, however, that the higher tiers should result in protective risk 661 assessment decisions not in conflict with the specific protection goals set by the competent authorities. 662 According to Boesten et al. (2007) and Solomon et al. (2008) the general principles of tiered 663 approaches are: 664

• lower tiers are more conservative than higher tiers 665 • higher tiers aim at being more realistic than lower tiers, 666 • lower tiers usually require less effort than higher tiers, 667 • in each tier all available relevant scientific information is used, 668



• all tiers aim to assess the same protection goal. 669 In short, the tiered system as a whole needs to be (i) appropriately protective, (ii) internally consistent, 670 (iii) cost-effective and (iv) address the problem with a higher accuracy and precision when going from 671 lower to higher tiers (see Figure 2.1). 672

673

Figure 2.1: Tiers in the risk assessment process, showing the refinement of the process through the 674 acquisition of additional data (redrafted after Solomon et al. 2008) 675

676

An additional practical aspect of the tiered approach is that there has to be some balance between the 677 efforts and the filtering capacity of a tier. For instance, it does not make sense to define a tier that 678 requires 50% of the efforts of the next higher tier but leads in 95% of the cases to the conclusion that 679 this next tier is needed (Boesten et al. 2007). 680

In PPP risk assessment under the PPP Regulation the basic data requirements for the first tier risk 681 assessment are strictly defined. The EU exposure assessment in edge-of-field surface waters normally 682 consists of FOCUS Steps 1, 2, 3, or 4 (with the restriction for Step 4 that it has to maintain Step 3 683 scenario definitions). These four steps have in common that they can be performed relatively easy and 684 quickly with available tools agreed upon (FOCUS 2001; 2007a; for further details see Chapter 4). 685 Detailed information on data requirements for the first tier effect assessment in the EU can be found in 686 Chapter 5. 687

The ‘unless’-clauses described in the Uniform Principles (Regulation (EU) No 546/2011) offer the 688 possibility to perform higher-tier risk assessments (EC 1997). Procedures for higher-tier effect testing 689 to evaluate the environmental risks of PPPs to aquatic organisms can be found in Chapters 6, 7, 8 and 690 9 of this Guidance Document. 691

A logical consequence of the basic dossier requirements in the process of PPP registration in the EU is 692 that the risk assessment always starts with the first tier. However, the uncertainties and possible risks 693 indicated by the first tier (and other lower-tiers) inform the risk assessors and risk managers on which 694 organisms and methods to focus in the higher-tier risk assessment. Another logical consequence of the 695 principles of the tiered approach, described above, is that higher tiers can be used to calibrate for the 696 lower tiers, because the assessment endpoint derived from a higher tier is closer to the actual 697 objectives of the adopted protection goal (Figures 2.1 and 2.2). In the aquatic effects assessment for 698 PPPs, an appropriate mesocosm test, in combination with an appropriate assessment factor or model 699 for spatio-temporal extrapolation, currently often is the highest tier when invertebrates or primary 700 producers are at risk. Appropriate intermediate tiers may be refined exposure studies with standard test 701 species and the species sensitivity distribution (SSD) approach based on additional toxicity data with 702 potentially sensitive species. 703

2

3

1

4

Simple(few data)

Complex(many data)

Realistic

Conservative

Protection goal

2

3

1

4

Simple(few data)

Complex(many data)

Realistic

Conservative

Protection goal



704

Figure 2.2: Illustration of the relationship between tiers of the risk assessment process and protection 705 goals, in the approach used by the PPR Panel (EFSA Panel on Plant Protection Products and their 706 Residues (PPR) 2010b). 707

As explained above, the uncertainties and possible risks indicated by the first tier determine on which 708 organisms and methods to focus in the higher-tier risk assessment. For example, if the first tier effects 709 assessment for an insecticide indicates that the standard test arthropods are at least an order of 710 magnitude more sensitive than the other standard test species (e.g. algae, fish) the higher-tier tests may 711 focus on aquatic invertebrates by performing further studies such as additional laboratory toxicity tests 712 or microcosm/mesocosm experiments. If these tests lead to a refined RAC for invertebrates, a risk 713 assessor must check whether this refined RAC is still protective for other organisms not at risk in the 714 first tier (e.g. fish). Consequently the tiered approach has to adopt an iterative procedure. 715

2.3. Terminology in the aquatic risk assessment of PPPs 716

Since the aquatic risk assessment for PPPs follows a tiered-approach, characterised by different fate 717 and effect procedures, it is very important to use a transparent risk assessment terminology to facilitate 718 communication between fate experts and ecotoxicologists, between risk assessors and risk managers, 719 and between different stakeholders involved in the administration procedure of PPPs. Before 720 describing the PEC/RAC terminology adopted in this guidance document, definitions of acute and 721 chronic effect/risk assessments is provided. 722

Acute Effect Assessment = Short-term Effect Assessment 723 Assessment of the Regulatory Acceptable Concentration (RAC) for adverse effect of PPP 724 exposure to (non-target) organisms (individuals, populations, communities) occurring within a 725 short period after exposure (hours to weeks; dependent on the life span of the organisms of 726 concern). Note that this is not synonymous with “Assessment of effects due to short-term 727 exposure” since short-term exposure may result in delayed short-term or delayed long-term 728 effects. In current practise of PPP effect assessments, the acute effect assessment scheme 729 starts with the Tier 1 acute toxicity dataset. The acute effect assessment may be refined by 730 addressing additional toxicity data for the most sensitive taxonomic group triggered by the 731 Tier 1 acute core data. 732

Acute Risk Assessment = Short-term Risk Assessment 733 Evaluation/determination of the probability of adverse effects of PPP exposure to (non-target) 734 organisms in the environment is achieved by comparing the RAC of the acute effect 735 assessment scheme with an appropriate Predicted Environmental Concentration (PEC) for the 736



environmental compartment of concern. Within the context of the current guidance document 737 this PEC usually will be the maximum (peak) concentration derived from the predicted 738 exposure profile. 739

Chronic Effect Assessment = Long-term Effect Assessment 740 Assessment of the Regulatory Acceptable Concentration (RAC) for adverse effects of PPP 741 exposure to (non-target) organisms (individuals, populations, communities) that develop 742 slowly and/or have a long-lasting course, and that are caused by short-term exposure (latent 743 effects) or long-term exposure. Consequently, a chronic effect assessment is not synonymous 744 with “Assessment of effects due to long-term exposure”, but does not exclude it. In current 745 practise of PPP effect assessments, the chronic effect assessment scheme starts with the Tier 1 746 chronic toxicity dataset. The chronic effect assessment may be refined by addressing 747 additional toxicity data for the most sensitive taxonomic group triggered by the Tier 1 chronic 748 core data. Since algal tests usually cover the whole life-cycle of the test species, the standard 749 72-92 h toxicity tests with algae can be considered as chronic. 750

Chronic Risk Assessment = Long-term Risk Assessment 751 Evaluation/determination of the probability of adverse effects of PPP exposure to (non-target) 752 organisms in the environment by comparing the RAC of the chronic effect assessment with an 753 appropriate PEC for the environmental compartment of concern. Within the context of the 754 current guidance document this PEC may be the maximum (peak) or a time-weighted average 755 (TWA) concentration derived from the predicted exposure profile. 756

RAC and PEC terminology 757 In this aquatic guidance document, the Predicted Environmental Concentration (PEC) derived from a 758 certain exposure assessment approach (e.g. FOCUS scenarios and models) and the Regulatory 759 Acceptable Concentration (RAC)8 derived from different effect assessment tiers will refer to the 760 environmental compartment to which they apply e.g. surface water (PECsw; RACsw). 761

In addition, the PEC will refer to the type of exposure concentration (peak/maximum or TWA) and the 762 RAC to the type of effect assessment that is addressed (acute (ac) or chronic (ch)). For example: 763

PECsw;max the maximum concentration predicted for surface water, 764 RACsw;ac the regulatory acceptable concentration for surface water within the context of 765

the acute effect assessment scheme, 766

Furthermore, the effect assessment approach, or the protection goal option addressed, may be added as 767 a preposition when referring to a certain RAC, e.g. Geom-RACsw;ch (the RAC for surface water within 768 the context of chronic effect assessment and derived by means of the geometric mean approach), 769 SSD-RACsw;ac (the RAC for surface water within the context of acute effect assessment and derived by 770 means of the SSD approach) or ETO-RACsw,ch (the RAC for surface water within the context of 771 chronic effect assessment and addressing the Ecological Threshold Option (ETO) as derived by means 772 of the model ecosystem approach). 773

To summarise, the following abbreviations are commonly found in subscript following the term PEC 774 or RAC: 775

ac: acute 776 ch: chronic 777 sw: surface water 778 max: maximum 779

8The term RAC was defined by the PPR Panelin EFSA (2006): The Annex VI of Directive 91/414/EEC stipulates that an

authorisation may begranted if e.g. the predicted short-term exposure does not exceed the concentration of the lowest LC or EC50 divided by 100.I.e., such concentration would be considered acceptable under the regulatory criteria of Annex VI, hence this term.



twa: time weighted average 780 The following prepositions are commonly used before the term PEC or RAC: 781 Geom: geometric mean 782 SSD: species sensitivity distribution 783 ETO: ecological threshold option (used in micro/mesocosm studies) 784 ERO: ecological recovery option (used in micro/mesocosm studies) 785 786

2.4. PPP effect assessment scheme 787

From the definitions above it is clear that for PPP effect assessment two distinct effect assessment 788 schemes can be identified that respectively start with the Tier 1 acute toxicity data set (Acute Effect 789 Assessment Scheme) and the Tier 1 chronic toxicity data set (Chronic Effect Assessment Scheme). 790 Since, (1) both the acute and the chronic effect assessment schemes address the same specific 791 protection goal, and (2) the same higher-tier effect study (e.g. micro-/mesocosm test or food web 792 model) may be used in both the acute and the chronic risk assessment, the overall effect assessment 793 scheme presented in Figure 2.3 is a convenient schematic presentation of the tiered approach. A key 794 aspect is that, in PPP effect/risk assessment, both the acute and chronic effects/risks have to be 795 evaluated by starting with the Tier 1 approach. The subsequent tiers that follow may differ for the 796 acute and chronic assessment, depending on the remaining uncertainties. Note the small arrows in the 797 diagram that are placed between Tier 3 (population and community level experiments) and Tier 4 798 (population and food web models), illustrating that the Tier 4 assessment usually concerns a 799 combination of experimental data and modelling to assess population and/or community level 800 responses (e.g. recovery; indirect effects) at relevant spatio-temporal scales. Also note that from a 801 scientific point of view it may be possible to extrapolate results of acute toxicity to assess chronic 802 effects and the other way around, although currently this is not common practise in PPP risk 803 assessment. On basis of Tier 1 and Tier 2 information and appropriate TK/TD models for the species 804 at risk, individual-level effects of time-variable exposure regimes may be assessed. 805 806

807

Figure 2.3: Schematic presentation of the tiered approach within the acute (left part) and chronic 808 (right part) effect assessment for PPPs. For each PPP both the acute and chronic effects/risks have to 809 be assessed. 810

811



2.5. When to use the peak or a time-weighted average predicted environmental 812 concentration (PEC) in the risk assessment 813

A crucial step in the risk assessment is the linking of exposure and effects data. As presented in Figure 814 2.3, PECsw;max values are used in acute risk assessments, whilst in chronic risk assessments, in first 815 instance the PECsw;max, and under certain conditions, a time-weighted average (TWA) PEC (PECsw;twa) 816 may be used. The use of the TWA concentration approach in the risk assessment of PPPs is based on 817 the observation that effects of PPPs on aquatic organisms may be similar when exposed for a short 818 time to a greater concentration or for a longer time to a smaller concentration, a phenomenon referred 819 to as reciprocity (Giesy & Graney 1989). Reciprocity relates to Haber’s law, which assumes that 820 toxicity depends on the product of concentration and time. For example, a 8-day exposure at 10 µg/L 821 may cause the same effects as a 4-day exposure at 20 µg/L or a 2-day exposure at 40 µg/L, an example 822 of linear reciprocity. Linear reciprocity is the basis of the TWA approach where exposure 823 concentration is integrated over time (area under the curve = AUC) and then divided by the duration of 824 the toxicity test. When this approach is applied, different exposure patterns with the same AUC are 825 assumed to have the same effects. Note, however, that for certain PPPs it has been demonstrated that 826 in prolonged acute toxicity tests a higher pulse exposure of a shorter duration may be more detrimental 827 than an equivalent lower pulse exposure of a longer duration. For example, Schultz & Liess (2000) 828 demonstrated that after 240 days a 1-h pulse exposure to the insecticide fenvalerate was acting 829 substantially stronger on emergence and dry weight biomass of emerged Trichoptera than a 10-h pulse 830 that was 10 times lower. If short-term exposure to PPPs results in delayed responses of PPP-831 susceptible species (for other examples see Abel, 1980; Liess, 2002; Brock et al. 2009) this may be a 832 reason to be reluctant in applying the time weighted average PEC in the chronic risk assessment. 833

Theoretically, reciprocity should only apply where both uptake and/or elimination of a compound into 834 the test organism (toxicokinetics) and damage and/or repair processes (toxicodynamics) have reached 835 steady state (Rozman & Doull, 2000). Ashauer et al. (2007) found for the insecticide chlorpyrifos and 836 the crustacean Gammarus pulex, that the TWA approach based on an acute toxicity test greatly 837 underestimated mortality in longer-term exposure studies, whereas it overestimated mortality caused 838 by pentachlorophenol. In long-term toxicity tests that addressed survival/mortality of Gammarus 839 pulex, however, they demonstrated that the TWA concentration approach can be used to extrapolate 840 results of a chronic pulse test to other chronic exposures for both chlorpyrifos and pentachlorophenol. 841 This observation supports the use of the TWA concentration approach in chronic risk assessments. 842 Also for the herbicide metsulfuron-methyl and the aquatic vascular plant Myriophyllum spicatum it 843 was demonstrated that the TWA approach can be used to assess concentration-response relationships 844 under longer-term time-variable exposure conditions (Belgers et al. 2011). In general, the longer 845 duration of chronic tests implies a greater probability that toxicokinetics and toxicodynamics will 846 approach steady state by the end of the study period. 847

Although current scientific knowledge does not support the use of the TWA approach in acute risk 848 assessments (i.e. using a PECtwa), this does not mean that the concentrations of the acute toxicity 849 estimate (the C in the LC50 or EC50) should not be expressed in terms of measured TWA. In general 850 OECD test guidelines recommend that toxicity endpoint (e.g. EC50) should be based on measured 851 concentrations. However, if evidence is available to demonstrate that the measured concentration of 852 the test substance has been satisfactorily maintained within ± 20%of the nominal throughout the test, 853 then the results can be based on nominal. 854

2.5.1. When and how (not) to use the PECsw;twa in chronic risk assessments 855

In chronic risk assessments, the default recommendation is to use PECsw;max, and under certain 856 conditions, a time weighted average (TWA) PEC may be used. However, PECsw;twa should not be used 857 if the following conditions apply (adapted after Brock et al. 2010c): 858

• In chronic risk assessments that use RACs from long-term effect studies where the exposure 859 has not been maintained, loss of the active substance in the test system was relatively fast and 860 the toxicity estimate has been expressed in terms of nominal or initially measured 861



concentration. This may, for example, be the case in the standard 28-d water-spiked test with 862 Chironomus riparius. 863

• When the effect endpoint in the chronic test (used to derive the RAC) is based on a 864 developmental process during a specific sensitive life-cycle stage that may last a short time 865 only (e.g. malformations during metamorphosis, effects caused by endocrine disruption) and 866 evidence exists that the exposure may occur when the sensitive stage is present. 867

• When the effect endpoint in the chronic test (used to derive the RAC) is based on mortality 868 occurring early in the test (e.g. in the first 96 h), or if the acute to chronic ratio (acute EC50 or 869 LC50/chronic NOEC) based on immobility or mortality is < 10. If the acute to chronic ratio is 870 small (< 10) the time to onset of maximum effects likely will be short in the chronic test. 871

• If latency of effects (delayed effects) has been demonstrated in longer-term toxicity tests in 872 which observations continue after the exposure is completed or the organisms have been 873 removed from the stressor, or when latency might be expected on basis of other data such as 874 toxic mode-of-action. In longer-term studies, latency may result from delays in the chain of 875 events between exposure and expression of effects (e.g. in the case of moulting inhibiting 876 insecticides and substances suspected of endocrine disrupting effects). To demonstrate latency 877 it may even be required to make observations on the responses of the offspring. It is advised to 878 address latency if, through analogy to similar substances or knowledge of mechanisms of 879 action, it is expected to occur. In cases where latency is known not to occur in PPPs with a 880 similar toxic mode-of-action, it might be disregarded. Further information on refined exposure 881 laboratory toxicity tests as a mean to address latency is given in section 7.2. 882

883

In cases other than those listed above, the use of the TWA approach in the chronic risk assessment 884 may be an option. Ecotoxicologists must determine, based on knowledge of ecotoxicological data, 885 whether or not the TWA concentration approach is appropriate to be used in the chronic risk 886 assessment, and which time window the TWA should be based upon. 887

For realistic to realistic worst-case risk assessment approaches, the time-window of the TWA PEC 888 should be equal to or smaller than the length of the relevant chronic toxicity test (or life-stage of the 889 species with the highest ecotoxicological concern) that triggered the risk. For invertebrates, fish and 890 macrophytes, a default 7-day TWA time window is proposed if the TWA concentration approach is 891 deemed appropriate (see criteria above) and no further information on the relation between exposure 892 pattern and time to-onset-of the relevant effect is provided. For the time being the PPR Panel of EFSA 893 adopts this pragmatic approach that most likely is relatively worst case. It may be justified to lengthen 894 or shorten the default 7-d TWA period when scientific data are made available that demonstrate that 895 another TWA period is more appropriate. This, for example, may be demonstrated by means of tailor-896 made experiments that allow to compare the effects of different exposure durations (including the 897 onset-of-effects) on the organisms of concern and/or by means of toxicokinetic/toxicodynamic 898 (TK/TD) modelling (e.g. Jager et al. 2011). 899

If the use of the TWA approach in the chronic risk assessment is appropriate, concentration-response 900 relationships observed in toxicity tests with long-term exposure (which may be variable in time), as 901 well as the derived RAC, can be expressed in terms of TWA concentrations. This RAC value can be 902 compared with the appropriate TWA PEC under the condition that the time-window for the PEC 903 estimate is equal or shorter than that for the “C” in the effect estimate. 904

Further information on the use of TWA in combination with endpoints derived from experimental 905 ecosystems is given in section 7.3.5.2. 906

2.5.2. Decision scheme to use the PECsw;max or PECsw;twa in the risk assessment 907

For an appropriate risk assessment of PPPs in a relevant edge-of-field surface water, the minimum set 908 of exposure estimates required are the PECsw;max, the highest PECsw;7d-twa and the annual exposure 909 profile from which exposure characteristics like number of pulse exposures, pulse durations, intervals 910



between pulses and water dissipation DT50 values can be deduced. It may also be convenient to 911 calculate the highest 3-d, 14-d, 21-d and 28-d TWA PECs since these time windows correspond with 912 the test duration of chronic toxicity tests with standard test algae, macrophytes, invertebrates and fish. 913 The following decision scheme may help to determine whether the PECsw;max or the PECsw;twa can be 914 used in the risk assessment. 915

916 1. Acute Risk Assessment 917

Is PECsw;max (of highest available tier) > RACsw;ac (of highest available tier)? 918 Yes: Assessment not sufficient to demonstrate a low risk 919 No: Low acute risk, go to 2 920 921

2. Chronic Assessment 922 Is PECsw;max (of highest available tier) > RACsw;ch (of highest available tier)? 923

Yes: Go to 3 924 No: Low chronic risk 925

926 3. Is the RACsw;ch based on short-term tests and organisms with a generation time < 3 days? (For 927

example, this may concern laboratory tests with unicellular algae) 928 Yes: Go to 4 929 No: Go to 6 930 931

4. Is the observed effect caused by reduction of growth (e.g. not an algicidal effect)? 932 Yes: Go to 5 933 No: PECsw;twa not appropriate (low risk not demonstrated) 934 935

5. Is PECsw;3d-twa (of highest available tier) > RACsw;ch (of highest available tier) ? 936 Yes: Low risk not demonstrated 937 No: Low risk demonstrated 938

939 6. Is the RACsw;ch derived from a long-term test (≥ 7 days) in which (i) loss of the active 940

substance from water is more that 20% of nominal at the end of the exposure period and (ii) 941 the toxicity estimate (e.g. EC10 or NOEC) is expressed in terms of nominal/initially measured 942 concentration of the active substance? (For example, this may concern the 28-d water-spiked 943 Chironomus test and tests with macrophytes) 944

Yes: PECsw;twa not appropriate (low risk not demonstrated) 945 No: Go to 7 946

947 7. Is the RACsw;ch based on treatment-related responses of the relevant test species early in the 948

chronic test (e.g. mortality is observed during the initial 96 h or the time to onset of more than 949 50% of the effect occurs within 7 days in the treatment level above the one from which the 950 RACsw;ch is derived) 951

Yes: PECsw;twa not appropriate (low risk not demonstrated) 952 No: Go to 8 953 954

8. Is it demonstrated by the notifier that for the organisms at risk and the PPP under evaluation 955 and/or PPP with a similar toxic mode-of-action (read-across information), the following 956



phenomena are not likely: (i) latency of effects due to short-term exposure; (ii) the co-957 occurrence of exposure and specific sensitive life stages that last a short time only. 958

Yes: Go to 9 959 No: PECsw;twa not appropriate (low risk not demonstrated) 960

961 9. Is PECsw;7d-twa (of highest available tier) > RACsw;ch (of highest available tier) ? 962

Yes: Go to 8 963 No: Low risk demonstrated 964 965

10. Are experimental (or TK/TD modelling when guidance is available) data available that 966 demonstrate that for the species at risk a larger time window for the PECsw;twa may be used 967 (not exceeding the duration of the Tier 1 chronic test that triggered the risk)? 968

Yes: Go to 7 and replace the PECsw;7d-twa by another appropriate PECsw;twa 969 No: low risk not demonstrated 970

971



3. Exposure assessment goals and specific protection goals for water organisms 972


A risk assessment scheme that addresses a specific protection goal requires a clearly defined 974 ecotoxicologically relevant type of concentration (ERC) that needs to be consistently used in both the 975 exposure and effects assessment procedures within the same risk assessment scheme. Before defining 976 in greater detail the specific protection goals for water organisms in edge-of-field surface waters the 977 relevant ERCs and exposure assessment goals will be described and discussed. 978

3.2. The Ecotoxicologically Relevant Concentration (ERC) 979

Lack of a clear conceptual basis for the interface between the exposure and effect assessment may lead 980 to a low overall scientific quality of the risk assessment. This interface is defined by EFSA (2005a) 981 and Boesten et al. (2007) as the type of concentration that gives an appropriate correlation to 982 ecotoxicological effects, and is called the ecotoxicologically relevant type of concentration (ERC). In 983 the risk assessment the ERC needs to be consistently applied so that field exposure estimates (PECs) 984 and regulatory acceptable concentrations (RACs) can be compared as readily as possible. The 985 ecotoxicological considerations determining the ERC may include the following questions: 986

• In which environmental compartment do the aquatic organisms at risk live (e.g. water or 987 sediment)? 988

• What is bioavailable for the organism (e.g. for pelagic organisms the freely dissolved fraction 989 in water, the sorbed fraction to dissolved organic matter or suspended particulate matter, or the 990 fraction in the food)? 991

• What is the influence of the time-variable exposure pattern on the effects (e.g. do peak or 992 longer-term concentrations explain the responses)? 993

• If latency of effects is of concern or longer-term exposures explain the responses, what should 994 be the relevant time-window of the exposure and effect estimates? 995

996 In ecosystems the ERC may be different for substances that differ in toxic mode-of-action and for 997 different populations of aquatic organisms, life stages of species, and so on. For example, for an 998 aquatic invertebrate living associated with macrophytes in shallow freshwater ecosystems, the ERC 999 could be the maximum concentration over time of the dissolved fraction for a fast-acting insecticide or 1000 some time-weighted average (TWA) concentration for a slow acting fungicide (e.g., 7-day or 21-day 1001 TWA). For detritivores that predominantly dwell on the sediment surface and process particulate 1002 organic matter (POM), the ERC for a hydrophobic substance could be the concentration of the PPP in 1003 the POM consumed. For an aquatic insect that predominantly dwells at the water surface (e.g. water 1004 striders) the ERC of a fast acting insecticide may be the water concentration in the top layer of the 1005 water column, which may be relevant if stratification of the insecticide occurs initially. 1006

After the ERCs for the PPP under evaluation and the aquatic organisms at risk have been determined, 1007 the collected exposure data can be linked to the relevant ecotoxicological data. It is important that 1008 within the same tiered risk assessment scheme (addressing the same specific protection goal) the type 1009 of ERC used to express the “C” in the PEC estimates should not be in conflict with the ERC used to 1010 express the “C” in the RAC estimates, in the sense that a realistic to worst-case risk assessment can be 1011 performed. 1012

This guidance document has its focus on water organisms that dwell in the water column of edge-of-1013 field surface waters. Within the context of this guidance document the concentration of the freely 1014 dissolved PPP (hence, not including PPPs sorbed on suspended matter or sediments9) in the water 1015 column of edge-of-field surface waters is chosen as the most relevant ERC for water toxicity. 1016

9 If the risk is predicted to be via the sediment in a higher tier risk assessment then this should not be ignored, however, detailed guidance will not be provided in this GD.



3.3. Exposure assessment goals in edge-of-field surface waters 1017

The specification of exposure assessment goals has a significant influence on the overall level of 1018 protection to be achieved by the proposed risk assessment procedure. However, as described in 1019 Section 1.3.1, the Panel assumes that the current FOCUS exposure assessment for edge-of-field 1020 surface water systems will continue to be used until updated or new methods become available and 1021 adopted by the SCFCAH. Therefore, no proposals for the specification of the exposure assessment 1022 goals are provided for the time being, instead a description of the related choices in the current 1023 FOCUSsw methodology is provided in Appendix A. 1024

3.4. Specific Protection Goals for Water Organisms 1025

The Regulation requires a high level of protection; e.g. no unacceptable effects on the environment 1026 (preambles 8, 10, 24 and Article 4.3). However, only general protection goals are given in the 1027 legislation and specific protection goals that are needed for risk assessment are not precisely defined. 1028 In the specific protection goal opinion (EFSA, 2010b) a process is described for defining Specific 1029 Protection Goal (SPG) options for key drivers (main groups of organisms) covering ecosystem 1030 services which could potentially be affected by PPPs. The aquatic key drivers, their ecological entities 1031 to be protected and related Tier 1 test species are mentioned in Table 3.2. In general, to ensure 1032 ecosystem services, taxa representative for aquatic key drivers identified need to be protected at the 1033 population level. However, it is proposed to protect aquatic vertebrates (fish, amphibians) at the 1034 individual (in the acute risk assessment to avoid visual mortality) to population level (chronic risk 1035 assessment). To protect the provisioning and supporting services provided by microbes it is proposed 1036 to protect them at least on the functional group level (EFSA, 2010b). 1037

1038 Table 3.2: The aquatic key drivers and their ecological entity to be protected as proposed in EFSA 1039 (2010b) and the current standard aquatic test species related to these key drivers (SANCO document 1040 11802/2012) 1041

Key driver Ecological entity to be protected

Tier 1 taxa mentioned in data requirements (SANCO document 11802/2012)

Aquatic algae Populations Green algae: e.g. Pseudokirchneriella subcapitata Diatoms: e.g. Navicula pelliculosa Blue-greens: e.g. Anabaena flos-aquae

Aquatic vascular plants

Populations Monocot: e.g. Lemna gibba / minor, Glyceria maxima Dicot: e.g. Myriophyllum

Aquatic invertebrates Populations Crustaceans: Daphnia magna /pulex, Americamysis bahia Insects: Chironomus riparius Oligochaets: Lumbriculus spp

Aquatic vertebrates Individuals (in acute risk assessment to avoid visual mortality) – populations (in chronic risk assessment)

Fish: e.g. Oncorhynchus mykiss

Aquatic microbes Functional groups No standard test species 1042 The ultimate goal of the update of the Aquatic Guidance Document is to develop and describe 1043 protective risk assessment schemes based on SPG options for aquatic key drivers (main taxonomic 1044 groups to be protected) defined in terms of the dimensions Degree of certainty (which always needs to 1045 be high), Ecological entity (see Table 3.2), Attribute, Magnitude, Temporal scale and Spatial scale. As 1046 mentioned in EFSA (2010b), interdependency exists between the different SPG dimensions. The 1047



dimension Attribute is closely linked to the dimension Ecological entity. For example functional 1048 groups (ecological entity) are often linked to processes (attribute), populations (ecological entity) to 1049 abundance/biomass (attribute) and individuals (ecological entity) to behaviour/survival/growth 1050 (attribute). Similarly, in edge-of-field surface water the dimension Magnitude is closely linked to the 1051 dimension Temporal scale, for example a larger magnitude of effect may be acceptable only if the 1052 response is short-term and not propagating to the community or a small magnitude of effects may be 1053 considered unacceptable if it is long-term. 1054 1055

3.5. Specific Protection Goal options for aquatic key drivers in edge-of-field surface water 1056

The specific protection goals as proposed below were discussed with Risk Managers (RMs) in the 1057 period of September – November 2012 via the SCFCAH and a dedicated meeting with RMs. 1058 Preliminary agreement was found, however, since the full guidance was not yet available, RMs asked 1059 to wait for the final draft guidance as available in the public consultation for providing their final 1060 decision. 1061 1062 For key drivers in edge-of-field surface waters that need to be protected at the population level or 1063 higher, this report will present assessment schemes that allow derivation of Regulatory Acceptable 1064 Concentrations (RACs) on basis of two options: 1065 1066

(1) Accepting only negligible10 population-level effects (ecological threshold option, ETO). 1067 The reasoning for this approach is based on the consideration that by not accepting population 1068 level effects on representative sensitive populations in edge-of-field surface waters, these 1069 populations will be protected and propagation of effects to the community, ecosystem and 1070 landscape level will not occur. 1071 1072 (2) Accepting some population effects if ecological recovery takes place within an acceptable 1073 time-period (ecological recovery option, ERO). When performing a protective risk 1074 assessment it is nevertheless desirable to not be overly protective. However, when including 1075 recovery to identify (un)acceptable effects all relevant processes that determine population 1076 viability and the propagation of effects to the community-, ecosystem- and landscape level are 1077 to be considered. Only such an integrative assessment can ensure a protective risk assessment. 1078 For example, if a temporal reduction of an invertebrate species of some months is accepted, it 1079 has to be ensured that organisms preying on this invertebrate can use other adequate food 1080 sources that are sufficient to sustain the population of the predator. In addition, if recovery of 1081 populations of short-cyclic water organisms is predicted, it has to be ensured that also species 1082 with contrasting life-cycle traits (i.e. longer generation time) are able to completely recover in 1083 the time available between the exposure events. 1084 1085

Furthermore, the Regulation requires that the risk assessment methodology should account for the 1086 simultaneous use of PPPs (applied in tank mixtures or used in sequence) and that the use of PPPs does 1087 not have any long-term repercussions for the abundance and diversity of non-target species (see EFSA 1088 Panel on Plant Protection Products and their Residues (PPR) 2010b). The selection of option 1 1089 (ecological threshold option), above, for the risk assessment of individual PPPs is more likely to avoid 1090 stress caused by the multiple use of different PPPs. Although, a risk assessment that considers 1091 recovery of sensitive populations may be a reasonable option for surface waters adjacent to crops with 1092 a limited PPP input, it is more uncertain if the option 2, specific protection goal (ecological recovery 1093 option) can be achieved when assessing risks for individual PPPs for their use in crop protection 1094 programmes characterised by intensive PPP use (simultaneous or repeated use of different PPPs). 1095 However, to draw a meaningful conclusion requires a thorough analysis of PPP use in major crops and 1096 the identification of those crops where the ecological risks are unacceptably high due to multi-stress 1097

10 The term negligible is used since it is difficult to demonstrate that no effect is occurring.



by different PPPs (including the recovery of potentially vulnerable populations). No definitive answer 1098 can be given at present. 1099

Below SPG options are provided for the different aquatic key drivers based on the procedure described 1100 in EFSA (2010b) and the two options described above (Table 3.3, Table 3.4 and following text). Note 1101 that in this guidance document the ‘dimension’ Spatial scale is fixed (edge-of-field surface water) and 1102 the ‘dimension’ Degree of certainty always should be high. Consequently, the SPGs for aquatic key 1103 drivers in edge-of-field surface waters are defined in terms of the dimensions: Ecological entity, 1104 Attribute, Magnitude and Temporal scale. 1105

1106

Table 3.3: Overview of proposed specific protection goals for the ecological threshold option 1107

Organism group Ecological entity Attribute Magnitude Time Algae population abundance/biomass

negligible effect not applicable

Aquatic plants population survival/growth abundance/biomass

Aquatic invertebrates population abundance/biomass

Vertebrates individual survival population abundance/biomass

Aquatic microbes functional group Processes (e.g. litter break down)

1108

Table 3.4: Overview of proposed specific protection goals for the ecological recovery option 1109

Organism group Ecological entity Attribute magnitude + duration of effect allowable on most sensitive/vulnerable

population

Algae population abundance/biomass small effect(a) months medium effect(a) weeks large effect(a) days

Aquatic plants(b) population Survival/growth abundance/biomass

small effect(a) months medium effect(a) weeks

Aquatic invertebrates(b) population abundance/biomass

small effect(a) months medium effect(a) weeks large effect(a) days

Vertebrates No recovery option

Aquatic microbes functional group processes RA will not be developed since Tier 1 data requirements are not defined

(a) None of the direct effects should lead to unacceptable indirect effects. 1110 (b) The recovery option will often not be applicable in case organisms with a long life-cycle could be affected and short-1111

term (days) large effects generally will be possible only for short-cyclic organisms that have a high reproduction 1112 capacity. Consequently, strict criteria for (not) allowing the recovery option are given in the further guidance below. 1113

1114

3.5.1. SPG proposal for algae (e.g. green algae, diatoms, blue-greens) in edge-of-field surface 1115 water 1116

For both SPG options algae will be protected at the population level by considering their 1117 abundance/biomass in edge-of-field surface waters. The population level is proposed instead of the 1118 functional group level as indicated in the opinion (EFSA Panel on Plant Protection Products and their 1119 Residues (PPR) 2010b) since clear definitions of functional groups of algae are lacking. Option 1 1120 (ecological threshold option) allows negligible effects on these endpoints only. Option 2 (ecological 1121 recovery option) allows large effect for days, medium effects for weeks, and small effects for several 1122 months on the abundance and/or biomass of vulnerable populations of algae as long as these effects do 1123



not cause persistent indirect effects on other water organisms that depend on algae. In option 2 the 1124 range in acceptable magnitude of effects from small to large is selected because large temporal 1125 changes in abundance/biomass of algal populations are common in (non-stressed) aquatic ecosystems, 1126 due to the short response time to fluctuating environmental conditions such as; light and nutrient 1127 availability and predation by zooplankton. The required assessment endpoints for algae can be 1128 adequately studied in micro-/mesocosms characterised by conditions close to natural in terms of 1129 competition, predation and natural stressors. When these conditions are met, such test systems may be 1130 considered as reference tier to calibrate the lower effect assessment tiers(for details see chapter 7). 1131

3.5.2. SPG proposal for aquatic vascular plants (e.g. dicotyledonous, monocotyledonous) in 1132 edge-of-field surface water 1133

Aquatic vascular plants will be protected at the population level by considering their growth and/or 1134 abundance/biomass in edge-of-field surface waters. Option 1 (ecological threshold option) allows 1135 negligible effects only. Option 2 (ecological recovery option) allows medium effects as long as the 1136 duration of the effect on the abundance and/or biomass of vulnerable populations of macrophytes is 1137 not longer than weeks or small effects when they last for a few months. In option 2 the acceptable 1138 magnitude of effects is small to medium since large effects are not desirable even if recovery can be 1139 demonstrated. Macrophytes play important ecological roles (e.g. as substrate, shelter, food source) on 1140 which many other water organisms depend. As suitable reference tier for aquatic vascular plants 1141 micro-/mesocosm test systems can be used in which bioassays (e.g. potted plants) are introduced. 1142 Exposure conditions of herbicides in micro-/mesocosms that simulate ponds may be relatively worst-1143 case (closed test systems without flow-through). Mechanistic models (e.g. TKTD models) and/or 1144 refined exposure tests with selected species may be used as complimentary tools to address effects of 1145 realistic time-variable exposures. The required assessment endpoints for vascular plants need to be 1146 studied in conditions close to natural in terms of competition, predation and natural stressors in order 1147 to obtain realistic assessment endpoints (Maltby et al. 2010). 1148

3.5.3. SPG proposal for aquatic invertebrates in edge-of-field surface water (e.g. crustaceans, 1149 rotifers, insects, oligochaete worms, molluscs) 1150

Aquatic invertebrates will be protected at the population level by considering their abundance/biomass 1151 in edge-of-field surface waters. Option 1 (ecological threshold option) allows negligible effects on 1152 these endpoint only. Option 2 (ecological recovery option) allows small effects for a few months, 1153 medium effects for weeks and large effects for days on the abundance and/or biomass of vulnerable 1154 populations of invertebrates as long as their reduction does not result in more persistent indirect 1155 effects. In option 2 the range in acceptable magnitude of effects from small to large is selected because 1156 large temporal changes in abundance/biomass of particularly short-cyclic invertebrate populations 1157 (e.g. daphnids, rotifers, and representatives of oligochaete worms and insects) are common even in 1158 more or less pristine aquatic ecosystems. The required assessment endpoints for aquatic invertebrates 1159 can be adequately studied in micro-/mesocosms (reference tier) if the conditions in these test systems 1160 are sufficiently representative of natural ecosystems in terms of species composition, species 1161 interactions (competition, predation) and natural stressors, and the duration of experiment is long 1162 enough to enable detection of delayed effects. It should be carefully evaluated whether representatives 1163 of potentially sensitive and vulnerable species (e.g. uni-/semi-voltine insects with a long life cycle) are 1164 present. To extrapolate results of micro-/mesocosm experiments, it may be an option to also use 1165 population models (not described in this GD) to better address recovery potential of vulnerable 1166 invertebrates. These models, however, should consider how recovery is affected by possible 1167 interference with other populations (species interactions such as predation and competition). 1168

3.5.4. SPG proposal for aquatic vertebrates in edge-of-field surface water (e.g. fish, 1169 amphibians) 1170

Since visible mortality of fish and amphibians due to acute toxicity of PPPs generally is not accepted 1171 by risk managers, and chronic effects on their populations should not be larger than negligible, the 1172 ecological threshold option only is proposed as SPG for aquatic vertebrates. A well-established and 1173



widely accepted reference tier to calibrate the acute and chronic Tier 1 effect assessment for 1174 vertebrates is not (yet) available. A possible reference tier for the ecological threshold option seems to 1175 be the Species Sensitivity Distribution approach (based on acute toxicity data to assess acute effects 1176 and on chronic toxicity data to assess chronic effects). 1177

3.5.5. SPG proposal for aquatic microbes (e.g. bacteria, fungi) 1178

As a default approach we propose a SPG that considers aquatic microbes at the functional group level 1179 and with a focus on ecological processes (e.g. litter breakdown) as main attribute. The magnitude of 1180 effect on these processes should be negligible and not result in more persistent indirect effects (e.g. 1181 inhibition of feeding by detritivores). Overall aquatic microbes show a fast recovery and/or well-1182 developed functional redundancy (important processes can also be performed by less sensitive 1183 microbes). Higher-tier tests (including micro-/mesocosm tests) with fungicides indicate that processes 1184 like organic matter breakdown are sufficiently protected when protecting populations of fish, 1185 invertebrates, algae and macrophytes (Maltby et al. 2009). However, certain species of aquatic fungi 1186 are reported to be more sensitive to specific fungicides (triazoles) than species of fish, invertebrates 1187 and plants usually tested (Bundschuh et al. 2011; Dijksterhuis et al. 2011). If the PPP under 1188 investigation is reported to have a specific toxic mode-of-action affecting aquatic fungi (e.g. triazole 1189 fungicides) it may be an option to select a SPG that also considers population level effects on aquatic 1190 fungi by taking into account ecological recovery (medium to large effects on growth and/or 1191 abundance/biomass that last no longer than weeks to months). 1192

3.6. Vulnerable species 1193

An important follow-up step in the EFSA approach to define specific protection goals (EFSA Panel on 1194 Plant Protection Products and their Residues (PPR) 2010b) is the identification of vulnerable 1195 representatives for each aquatic key driver mentioned in section 3.5. In the aquatic ERA for PPPs the 1196 identification of vulnerable species is important (1) to design and interpret higher tier experiments 1197 (e.g. micro-/mesocosm tests), (2) to identify focal species that need special attention when 1198 constructing ecological scenarios and adopting mechanistic modelling approaches, and (3) to design 1199 and interpret bio-monitoring programmes to evaluate the appropriateness of the prospective risk 1200 assessment procedures. 1201

Vulnerability has been defined as “the degree to which a system, subsystem, or system component is 1202 likely to experience harm due to exposure to a hazard” (Turner et al., 2003). Properties relevant to 1203 define vulnerability are species traits and characteristics that determine (1) susceptibility to exposure, 1204 (2) toxicological sensitivity, and (3) internal and external (re-colonisation) recovery processes (see 1205 Figure 3.4). 1206



1207 Figure 3.4: General framework for ecological vulnerability assessment for hazard. Taken from De 1208 Lange, Sala, Vighi & Faber (2010). 1209 1210 (1) Susceptibility to exposure 1211

Species traits that have been shown to increase exposure are; habitat preference in water-sediment or 1212 air-water boundary layers of aquatic ecosystems where higher concentrations of PPPs may occur (e.g. 1213 Brock et al. 2010a), or selective consumption of food with a high PPP content (e.g. carnivores high in 1214 the food chain; Stäb et al. 1996). Examples of traits that reduce exposure are emergence of insects 1215 before PPP exposure (Liess et al. 2008) and ability to actively migrate to non-exposed patches of 1216 habitat or ecosystems (e.g. Lahr 1997; Liess & Von der Ohe 2005). 1217

(2) Toxicological sensitivity 1218

Inter-specific variation in toxicological sensitivity of aquatic species to PPPs has been widely 1219 documented (e.g. Von der Ohe & Liess 2004; Luttik et al. 2011) and partly can be attributed to the 1220 specific toxic mode-of-action of the PPP of concern. For example, in the case of insecticides, aquatic 1221 arthropods usually are most sensitive (Maltby et al. 2005) while algae and/or macrophytes tend to be 1222 in the case of herbicides (Van den Brink et al. 2006). Many fungicides, however, have general biocidal 1223 properties in that representatives of different taxonomic groups may be categorised as potentially 1224 sensitive (Maltby et al. 2009). Recently efforts are being made to mechanistically link traits with 1225 intrinsic sensitivity of individual species of concern (Rubach et al. 2010). Experimental results 1226 indicate that this approach has promise, but, according to Baird & van den Brink (2007) effort is 1227 needed to compile species trait information to increase the power, precision and taxonomic 1228 representativeness of this approach. Note that toxicological sensitivity of water organisms to a specific 1229 PPP may be influenced by co-occurring other stress-factors (e.g. chemical stressors such as other 1230 PPPs, physico-chemical stressors like low oxygen levels and biotic stressors like pathogens) (Heugens 1231 et al. 2001). 1232

(3) Ecological recovery 1233

As regards the recovery of aquatic organisms from PPP stress, a distinction can be made between 1234 internal and external recovery processes (Caquet et al. 2007; Brock et al. 2010a). Internal recovery 1235 depends on surviving individuals in the stressed ecosystem (e.g. in refuges), or on a reservoir of 1236 resting propagules (e.g., eggs) that remain unaffected by the PPP. Especially species with a short 1237 generation time and a high reproductive output are able to recover quickly from PPP-stress if a small 1238 fraction of the population survives. Laboratory investigations under optimum conditions indicate that 1239



internal recovery at least needs one generation time (Barnthouse 2004). However, field investigation 1240 indicated that internal recovery requires around three generation times under realistic conditions 1241 (Niemi et al. 1990; Liess & Von der Ohe 2005). External recovery depends on the immigration of 1242 individuals from neighbouring ecosystems by active (e.g. flying) or passive (e.g. wind) dispersal. 1243 Especially species with a strong capacity to actively migrate from one site to another are able to 1244 perform this process within a relatively short period of time. However, ecological recovery of sensitive 1245 species may be hampered by other less sensitive and competitive species that increased subsequent to 1246 PPP application, or by physicochemical environmental conditions that adversely changed due to the 1247 stressor. 1248

In short, recovery of affected populations from chemical stress may be rapid if the following 1249 conditions apply: 1250

• The substance is not persistent, the exposure regime is short term or pulsed, and the time 1251 between pulses is sufficient for recovery. 1252

• The physicochemical environment and ecologically important food-web interactions are not 1253 altered by the stressor, or are quickly restored. 1254

• The generation time of the populations affected is short. 1255 • There is a ready supply of propagules of eliminated populations through active immigration 1256

by mobile organisms or through passive immigration by, for example, wind and water 1257 transport. 1258

1259 From the evaluation above it follows that the most vulnerable aquatic species/populations are 1260 characterised by a low ability to avoid exposure in space and time, a high toxicological sensitivity and 1261 life-cycle characteristics that hamper a fast recovery. When opting for the “Ecological threshold 1262 option” considerations for recovery are not necessary in the design and interpretation of higher tier 1263 studies, since there are no indications that within the same taxonomic group species with a low 1264 recovery potential show a higher toxicological sensitivity than species with a high recovery potential 1265 (e.g. Brock et al. 2010b) . When opting for “Ecological recovery option” on basis of micro-/mesocosm 1266 experiments, however, it needs to be critically evaluated whether representatives of potentially 1267 vulnerable populations are sufficiently covered in the micro-/mesocosm experiment. Further guidance 1268 on this is provided in Chapter 7. 1269

3.7. Implementation of the specific protection goals in this guidance document 1270

In the tiered effect assessment scheme developed for this guidance document, in principle all tiers are 1271 able to address the ‘ecological threshold option’ (accepting no population-level effects) (section 3.5). 1272 However, the model ecosystem approach (micro-/mesocosm experiments) also allows to address the 1273 ‘ecological recovery option’ when addressing risks to algae, vascular plants and invertebrates. In 1274 addition, appropriately designed and conducted micro-/mesocosm experiments are considered fit-for-1275 purpose as reference tier to calibrate the lower effect assessment tiers (on basis of laboratory toxicity 1276 tests with standard and additional species). When using results of micro-/mesocosm test to calibrate 1277 the lower tiers, or to derive an Ecological Threshold Option RAC (ETO-RAC), this will be done on 1278 basis of negligible effects on the abundance/biomass of the most sensitive populations in the micro-1279 /mesocosm test system. In this guidance document, negligible effects in appropriately designed and 1280 conducted micro-/mesocosm experiments are equivalent to Effect class 1 or Effect class 2 responses 1281 for the most sensitive populations and the application of an appropriate assessment factor (AF) for 1282 spatio-temporal extrapolation (see chapter section 7.3.5.3 for greater details). 1283

When addressing the ‘ecological recovery option’ in the specific protection goals for algae, vascular 1284 plants and invertebrates in edge-of-field surface waters, the indicated magnitudes and durations 1285 (temporal scale) of effects are characterised by interdependency (section 3.5). Since this 1286 interdependency may be influenced by factors like the vulnerability of the exposed organisms and 1287 possible occurrence of multiple stressors, in this guidance document a prudent but pragmatic approach 1288 is followed when deriving a RAC from micro-/mesocosms on basis of the ‘Ecological Recovery 1289



Option’ (ERO-RAC). It is proposed to base the ERO-RAC on an Effect class 3A concentration (see 1290 section 7.3.3; i.e. a concentration that results in a maximum period of treatment-related effects< 8 1291 weeks) for the most sensitive population level endpoint by applying an appropriate AF to this 1292 concentration (see chapter section 7.3.5.4 for more details). Note that, allowing a treatment-related 1293 effect period of less than 8 weeks (followed by complete recovery) on the most sensitive populations 1294 in appropriately designed and conducted micro-/mesocosm experiments is a relatively conservative 1295 approach (considering exposure from one PPP, see also section 3.5) if, for example, the PPP is applied 1296 repeatedly and/or univoltine invertebrates with a low ability to migrate to isolated micro-/mesocosms 1297 or aquatic vascular plants characterised by a slow growth rate are potentially sensitive. In addition, the 1298 application of the AF is meant to extrapolate the Effect class 3A concentration in space and time so 1299 that other vulnerable populations (section 3.6) that may occur in the edge-of-field surface water at risk 1300 are sufficiently protected. 1301

Although fungicides in particular may affect aquatic microbes it needs to be realised that Tier 1 1302 microbial test species are not mentioned in the revised data requirements (SANCO document 1303 11802/2012). The implication of this is that it is assumed that by protecting populations of aquatic 1304 algae, vascular plants, invertebrates and vertebrates the ecosystem services provided by bacteria and 1305 fungi are sufficiently protected. As there are no Tier 1 data requirements for aquatic microbes 1306 available, no specific RA scheme is developed for them in this guidance document. 1307

1308

4. Exposure Assessment 1309


PPP exposure assessment for the aquatic environment in the European Union is currently based on the 1311 FOCUS methodology (FOCUS, 2001). This is done for approval for active substances at EU level. It 1312 is also used in some Member States for product authorisation, but also different exposure assessment 1313 procedures may be used. The FOCUS surface water methodology has not been reviewed by the PPR 1314 Panel of EFSA during the revision of the GD on Aquatic Ecotoxicology and the overall level of 1315 protection for approval of active substances at EU level is therefore not clear. However, the 1316 methodology has been used in regulatory decision making throughout the last years and there is 1317 currently no alternative standardised exposure assessment methodology. Therefore, it is assumed that 1318 the FOCUS sw methodology will continue to be used until updated or new methods become available 1319 and adopted by the SCFCAH and will replace the existing tools. 1320

This chapter on exposure assessment in surface water is split in two sections, where the first one 1321 (4.2.1) gives a brief overview on how to perform the exposure assessment based on the current 1322 FOCUS surface water methodology and the second section (4.2.2) describes a guidance for the 1323 exposure assessment of metabolites that are formed in water and sediment. The intention of Section 1324 4.2 is only to give a short overview description of the FOCUS methodology. So the Panel has in no 1325 way the ambition to evaluate the methodology nor does the Panel have the intention to endorse the 1326 methodology. The PPR Panel, however, advises to critically evaluate and improve the surface water 1327 exposure assessment. 1328

4.2. FOCUS surface water scenarios and models 1329

The FOCUS Surface Water Modelling Working Group defined a step by step procedure for the 1330 calculation of predicted environmental concentrations in surface water (PECsw) (FOCUS 2001). The 1331 procedure consists of four steps, whereby the first step represents a very simple approach using simple 1332 kinetics, and assuming a loading equivalent to a maximum annual application. The second step is the 1333 estimation of concentrations taking into account a sequence of loadings, and the third step focuses on 1334 more detailed modelling taking into account realistic ‘worst case’ amounts entering surface water via 1335 relevant routes (run-off, spray drift, drainage). The third step considers substance loadings as foreseen 1336 in Step 2, but it also takes into account the range of possible uses. The uses are, therefore, related to 1337



the specific and realistic combinations of cropping, soil, weather, field topography and aquatic bodies 1338 adjacent to fields. The fourth step accounts for risk mitigation measures. Notice that the FOCUS 1339 procedure is a stepped approach, not a tiered approach. The most important reason is that FOCUS 1340 (2001) has not proven that earlier steps are more conservative than later steps. 1341

The aims of FOCUS (2001) for Step 1 and 2 calculations were to represent ‘worst-case loadings’ and 1342 ‘loadings based on sequential application patterns’ respectively but should not be specific to any 1343 climate, crop, topography or soil type. FOCUS (2001) considered the assumptions at both Steps 1 and 1344 2 as very conservative. Spray drift values are essentially based around drift numbers calculated from 1345 BBA (2000) and an estimation of the potential loading of PPPs to surface water via run-off, erosion 1346 and/or drainage. This loading represents any entry of PPP from the treated field to the associated water 1347 body at the edge of the field. 1348

Step 3 requires the use of mechanistic models PRZM, MACRO and TOXSWA. 1349

Already at Step 1 and 2 concentrations can be calculated not only for the active compound but also for 1350 metabolites formed in the soil before run-off/drainage occurs. The user must define the properties of 1351 the metabolite, including the maximum occurrence of the respective metabolite in soil studies and the 1352 ratio of the molecular masses of parent and metabolite. 1353

The fate of metabolites formed in the water body can also be taken into consideration at Step 1 and 2. 1354 The formation will be calculated in a similar way based on the maximum occurrence of the metabolite 1355 in water/sediment studies. 1356

4.2.1. Description of the different tiers 1357

4.2.1.1. Step 1 1358

The background of the FOCUS (2001) scenario properties on Step 1 is a combination of existing 1359 concepts within the EU and Member States and measured datasets. A water depth of 30 cm overlying 1360 sediment of 5 cm depth (density: 0.8 kg/L), however, only 1 cm of the sediment is used in the 1361 calculations when calculating the distribution between water and sediment layer. When calculating 1362 PECsed a depth of 5 cm is used, i.e. a dilution of a factor 5 compared to the 1 cm used for the water 1363 sediment distribution. Sediment with 5 % organic carbon was selected in order to comply with existing 1364 risk assessment approaches within the EU and existing ecotoxicity testing requirements for sediment-1365 dwelling organisms. 1366

1367

Calculation of concentrations resulting from spray drift: 1368

Spray drift deposition is expressed as the mass that enters the water per surface area of water, and 1369 assumed to be a certain fraction of the mass applied per surface area on the treated field. 1370

App: application dose (rate) [g/m²] 1371

Dsd: spray drift deposition as fraction of the application dose, i.e. mass deposited per surface area 1372 of surface water divided by mass deposited per surface area of field [-] 1373

h: surface water depth [m] 1374

C: concentration in surface water [g/m³] 1375

Equation 4.1: hAppDC sd= 1376

1377

Calculation of concentrations resulting from run-off, erosion or drainage: 1378



Run-off, erosion or drainage loadings are defined as the PPP mass that enters the water and is 1379 expressed as a fraction of the total mass applied on the contributing treated field times the surface area 1380 of the contributing field: 1381

App: application dose (rate) [g/m²] 1382

LRO: run-off loss as fraction of the applied PPP mass [-] 1383

Afield: area of the field contributing to run-off [m²] 1384

Asw: surface area of surface water [m²] 1385

C: concentration in surface water [g/m³] 1386

1387

Equation 4.2: sw

ROfield

AhLAApp

C = 1388

1389

An explicit width or length of the water body for the initial step is not defined, because drift loadings 1390 are based upon a percentage of the application rate in the treated field. For run-off, erosion or drainage 1391 entries only a fixed ratio Afield / Asw of 10:1 is defined to reflect the proportion of a treated field from 1392 which PPPs are lost to surface water. This number was calibrated by model runs of PRZM, MACRO 1393 and TOXSWA (FOCUS 2001). 1394

At Step 1 inputs of spray drift, run-off, erosion and/or drainage are evaluated as a single loading to the 1395 water body and ‘worst-case’ surface water and sediment concentrations are calculated. The loading to 1396 surface water is based upon the number of applications multiplied by the maximum single use rate 1397 except for compounds with a short half-life in sediment/water systems. If three times the degradation 1398 half-life (3 x DegT50) (combined water + sediment) is less than the time between individual 1399 applications, the maximum individual application rate is used to derive the maximum PEC as there is 1400 no potential for accumulation in the sediment/water system. For first order kinetics the value of 3 x 1401 DegT50 is comparable to the DegT90 value. Considering run-off loadings only run-off mass is entered 1402 into the stagnant 30 cm water, so no run-off water is added. This implies that exposure caused by run-1403 off entries will be estimated in a conservative way by Step 1. 1404

Four crop groups (arable crops, vines, orchards and hops, representing different types of application 1405 technology) and aerial applications are separated into different drift classes for evaluation at Step 1 1406 and 2. Drift values have been calculated at the 90th percentile from BBA (2000) as summarised in 1407 Table 4.1 (FOCUS 2001). The table indicates that no drift is assumed when the substance is 1408 incorporated or applied as granules or as a seed treatment. EFSA (2004) concluded that dust drift may 1409 occur for such applications and provided computational procedures to estimate this route. 1410

1411

Table 4.1: Step 1 drift input into surface water based on the 90th percentile (from BBA 2000) 1412

Crop Distance (m)(a) Drift (%)(c) Pome / stone fruit, early applications 3 29.2 Pome / stone fruit, late applications 3 15.7 Potatoes 1 2.8 Soybeans 1 2.8 Sugar beet 1 2.8 Sunflower 1 2.8 Tobacco 1 2.8 Vegetables, bulb 1 2.8 Vegetables, fruiting 1 2.8



Vegetables, leafy 1 2.8 Vegetables, root 1 2.8 Vines, early applications 3 2.7 Vines, late applications 3 8.0 Application, aerial 3 33.2 Application, hand (crop < 50 cm) 1 2.8 Application, hand (crop > 50 cm) 3 8.0 No drift (incorporation, granular or seed treatment) 1 0(b)

(a) Distance from edge-of-field to water body 1413 (b) 0 % drift for granular applications and seed treatments is not considered correct by EFSA PPR Panel 1414

(EFSA, 2004) 1415 (c) percentage of the application dose 1416 1417

In contrast, the run-off/erosion/drainage loading to the water body is set at 10% of the application for 1418 all scenarios. 1419

On the day of application drift entries are assumed to be present only in the water phase in order to 1420 estimate a conservative peak concentration. One day later the compound is distributed between water 1421 and sediment. 1422

In contrast, the run-off/erosion/drainage entry is distributed instantaneously between water and 1423 sediment at the time of loading according to the Koc of the compound in order to simulate the process 1424 of deposition of eroded soil particles containing PPPs. In this way compounds are distributed directly 1425 between sediment and water. The relationship between Koc and the distribution between water and 1426 sediment is calculated as follows: 1427

Equation 4.3: Fraction of run-off in water ))(( oceff KocbdSW

W⋅⋅⋅+

1428

1429

where: W = mass of water [30 g] 1430

Seff = mass of sediment available for partition [0.8 g] 1431

oc = mass fraction of organic carbon in sediment [0.05 g g-1] 1432

Koc = PPP organic carbon partition coefficient [cm3 g-1] 1433

bd = bulk density of the sediment [g cm-3]. 1434

1435

4.2.1.2. Step 2 1436

1437 The surface water properties on Step 2 are defined by FOCUS (2001) identically to Step 1 so a static 1438 water body with a water depth of 30 cm, overlying sediment of 5 cm depth (density: 0.8 g cm-3) with 5 1439 % organic carbon. However, only 1 cm of sediment is used in the calculations when calculating the 1440 partitioning between water and sediment. When calculating PECsed a depth of 5 cm is used, i.e. a 1441 dilution of a factor 5 compared to the 1 cm used for the water sediment distribution. 1442

Also at Step 2 the width of the water body is not defined because drift and run-off entries are 1443 calculated in a similar way based on a percentage of the application rate in the treated field. Also, the 1444 same ratio (10:1) is defined to reflect the proportion of a treated field from which PPPs are lost to 1445 surface water. 1446

However, at Step 2 inputs of spray drift, run-off, erosion and/or drainage are evaluated as a series of 1447 individual loadings comprising drift events (number, interval between applications and rates of 1448



application) followed by a loading representing a run-off, erosion and/or drainage event four days after 1449 the final application. Note that, only run-off mass is entered into the stagnant 30 cm water, so no run-1450 off water is added. This implies that peak exposure events caused by run-off entries will be estimated 1451 in a conservative way by Step 2. Degradation is assumed to follow first-order kinetics in soil, surface 1452 water and sediment and the exposure assessor also has the option of using different degradation rates 1453 in surface water and sediment. 1454

In order to prevent multiple worst case assumptions for multiple application patterns FOCUS (2001) 1455 defined different individual drift percentiles, dependent on the total number of applications per season, 1456 which according to FOCUS (2001) represent the overall 90th percentile (see table 4.2). 1457

1458

1459

Table 4.2: Individual drift percentiles for multiple applications on step 2 (FOCUS 2001) 1460

Number of applications Drift percentile(a) 1 90 2 82 3 77 4 74 5 72 6 70 7 69

>8 67 (a)it is assumed that the individual drift events meet the overall 90th percentile 1461 1462 As the procedure may result in lower predicted concentrations for multiple applications than for 1463 individual applications with the 90th drift percentile, the software automatically calculates both 1464 situations so that the user can select the higher value of the two. 1465

Drift inputs are loaded into the water column where they are subsequently distributed between water 1466 and sediment according to the Koc of the active substance. However, the process of adsorption to 1467 sediment at Step 2 is assumed to take longer than 1 day (as assumed at Step 1). It is assumed that, 1468 following a drift event, the PPP is distributed in surface water into two theoretical compartments, 1469 ‘available’ for sorption to sediment and ‘unavailable’ for sorption to sediment. 1470

Equation 4.4: Masw = K · Msw 1471

1472

Equation 4.5: Musw = (1-K) · Msw 1473

1474

where Msw = total mass of PPP in surface water [g], 1475

Masw = PPP mass available for sorption [g], 1476

Musw = PPP mass unavailable for sorption [g] and 1477

K = fraction of PPP mass in water available for sorption [-] 1478

K is set to 2/3 based on comparisons with laboratory sediment/water studies for weakly and strongly 1479 sorbing compounds (FOCUS 2001). 1480

The partitioning between the sediment and the ‘available’ water compartment is calculated with the 1481 equation for the fraction in the run-off water given before using a mass of water of 30 g (so assuming a 1482



30 cm water layer). This is not consistent with assuming that only two thirds (2/3) of the water is 1483 available for sorption as this would require a water depth of 20 cm. However, FOCUS (2001) 1484 considered the approach to be conservative for the PEC in surface water; e.g. for a Koc of 1000 L/kg, 1485 the fraction remaining in the water layer after sorption equilibration is 0.43 for a water depth of 30 cm 1486 and 0.33 for a water depth of 20 cm. 1487

The effect of the 2/3-available/ 1/3-non-available water compartments for sorption is presented and 1488 discussed in more detail in Appendix B. 1489

In contrast to Step 1, the amount of PPP that enters the soil at Step 2 is corrected for crop interception. 1490 For each crop, 4 interception classes have been defined depending on the crop stage. Crop interception 1491 will decrease the amount of PPP that reaches the soil surface and thus ultimately enters the surface 1492 water body via run-off/drainage. 1493

Four days after the final application, a run-off/erosion/drainage loading is added to the water body. 1494 This loading is a function of the residue remaining in soil after all of the treatments [g/ha] and the 1495 region and season of application. The different run-off/drainage percentages applied at Step 2 are 1496 listed in table 4.3. They have been calibrated by FOCUS against the results of Step 3-calculations as 1497 described in the FOCUS (2001) surface water report. For the calculation of the run-off event basically 1498 Eq. 4.2 is used, but instead of the application dose the soil residue (in g/ha) is used which reflects the 1499 situation four days after application. 1500

The user selects from two regions (Northern EU and Southern EU according to the definitions given 1501 for crop residue zones in the SANCO Document 7525/VI/95-rev.7, SANCO, 2001) and three seasons 1502 (March to May, June to September and October to February). 1503

In common with Step 1, the run-off/erosion/drainage entry is distributed between water and sediment 1504 at the time of loading according to the Koc of the compound. An effective sorption depth of 1 cm is 1505 used for the distribution between both phases. In this way compounds of high Koc are mostly added 1506 directly to the sediment whereas compounds of low Koc are mostly added to the water column in the 1507 ‘run-off/drainage’ water. Contrary to spray drift entries, at run-off entries all mass in the water layer is 1508 available for sorption to sediment. 1509

Table 4.3: Step 2 run-off/drainage input into surface water (from FOCUS, 2001) 1510

Region/Season % of soil residue North Europe, Oct. - Feb. 5 North Europe, Mar. - May 2 North Europe, June - Sep. 2 South Europe, Oct. - Feb. 4 South Europe, Mar. - May 4 South Europe, June - Sep. 3

No Run-off/drainage 0 1511

4.2.1.3. Step 3 1512

For Step 3 a selection of scenarios are defined based on a number of broad data sets that cover all 1513 areas of the European Community. According to FOCUS (2001) they should consider representative 1514 realistic worst-case situations and should take into account all relevant entry routes to a surface water 1515 body, as well as considering all appropriate target crops, surface water situations, topography, climate, 1516 soil type and agricultural management practices. However, due to the lack of comprehensive databases 1517 that characterise most of these agro-environmental parameters at a European level, when the scenarios 1518 were defined (1997-2001), they were not selected in a rigorous, statistically-based manner. Instead a 1519 pragmatic approach was adopted, using very basic data sources together with expert judgement. All 1520



scenarios are represented by specific field sites for which monitoring data were available. Table 4.4 1521 shows the inherent agro-environmental characteristics of the scenarios. 1522

Table 4.4: Inherent Agro-environmental characteristics of the Surface water scenarios (from FOCUS 1523 surface water report (2001) table 3.2-6) 1524

Scenario(1)

Meteoro-logical station

Mean spring & autumn temp. (oC)

Mean annual rainfall (mm)

Mean annual recharge

(mm)

Slope (%)

Soil

D1 Lanna <6.6 600 – 800 100 – 200 0 – 0.5 Clay with shallow groundwater

D2 Brimstone 6.6 – 10 600 – 800 200 – 300 0.5 – 2 Clay over impermeable

substrate D3 Vredepeel 6.6 – 10 600 – 800 200 – 300 0 – 0.5 Sand with shallow

groundwater D4 Skousbo 6.6 – 10 600 – 800 100 – 200 0.5 – 2 Light loam over

slowly permeable substrate

D5 La Jailliere 10 – 12.5 600 – 800 100 – 200 2 – 4 Medium loam with shallow groundwater

D6 Váyia, Thiva

>12.5 600 – 800 200 – 300 0 – 0.5 Heavy loam with shallow

groundwater R1 Weiherbach 6.6 – 10 600 – 800 100 – 200 2 – 4 Light silt with

small organic matter

R2 Valadares, Porto

10 – 12.5 >1000 >300 10 – 15 Organic-rich light loam

R3 Ozzano, Bologna

10 – 12.5 800 – 1000 >300 4 – 10 Heavy loam with small organic

matter R4 Roujan >12.5 600 – 800 100 – 200 4 – 10 Medium loam

with small organic matter

(1) D = Drainage, R= Run-off scenario 1525 1526 Inputs to surface water bodies from spray-drift are incorporated as an integral part of all of the 1527 scenarios based on the same tables as for the previous tiers (BBA, 2000). In addition to spray drift the 1528 scenarios are characterised by either run-off/erosions (R) or drainage (D) entries. 1529

For each location a maximum of two water body types is defined as shown in the following Table 4.5. 1530

Table 4.5: Water bodies associated with scenarios (from FOCUS, 2001) 1531

Scenario Inputs Water body type(s)(a) D1 Drainage and drift Ditch, stream D2 Drainage and drift Ditch, stream D3 Drainage and drift Ditch D4 Drainage and drift Pond, stream D5 Drainage and drift Pond, stream D6 Drainage and drift Ditch R1 Run-off and drift Pond, stream R2 Run-off and drift Stream R3 Run-off and drift Stream R4 Run-off and drift Stream



(a)all ditches and streams are assumed to have a length of 100 m, a width of 1 m and 1532 a variable, but minimum depth of 30 cm whereas the ponds are defined by surface 1533 water areas of 30 m x 30 m together with a depth of 100 cm 1534

For calculating substance entries into the surface water and for calculating time dependent 1535 concentration in the surface water bodies, different computer models are used. The currently 1536 recommended models (FOCUS, 2001) are MACRO for estimating the contribution of drainage, 1537 PRZM for the estimation of the contribution of run-off and erosion and TOXSWA for the estimation 1538 of the final PECs in surface waters and SWASH for the estimation of spray drift entries and as the 1539 overall user shell. 1540

To facilitate the calculation of exposure concentrations at Step 3 level, a software tool (SWASH) is 1541 available. It is an overall shell (user interface) combining all models involved in Step 3 calculations. 1542 The main functions of the shell are: 1543 • Maintenance of a central PPP properties database, 1544 • Provision of an overview of all Step 3 FOCUS runs required for use of a specific PPP on a specific 1545

crop, 1546 • Calculation of spray drift deposition onto various receiving water bodies and 1547 • Preparation of input for the models MACRO (drainage entries), PRZM (run-off/erosion entries) 1548

and TOXSWA (fate in surface water). 1549 1550

Calculating drainage entries for TOXSWA with MACRO 1551

MACRO is a general purpose leaching model that includes the effects of macropores (Jarvis, 1994; 1552 Jarvis, 2001). It was chosen by FOCUS to calculate drainage inputs to surface water bodies for the 1553 step 3 simulations because at that time it was the only FOCUS model that was able to simulate PPP 1554 losses through macropore flow. According to FOCUS this model was therefore suitable to cover the 1555 wide range of soil types included in the 6 drainage scenarios. 1556

MACRO considers macropores as a separate flow domain assuming gravity flow of water and a 1557 simple power law function for the conductivity. Solute movement in the macropores is assumed to be 1558 dominated by mass flow, while the concentration of solutes in water entering the macropores at the 1559 soil surface is calculated using the ‘mixing depth’ concept, whereby the incoming rain perfectly mixes 1560 with the soil solution in a given depth of soil. MACRO describes the movement of water through the 1561 soil matrix using Richards’ equation and solute transport with the convection-dispersion equation. 1562 Mass exchange between the flow domains is calculated using approximate first-order expressions 1563 based on an effective diffusion path length. Sorption is described with a Freundlich isotherm, with the 1564 sorption sites partitioned between the two domains. Degradation is calculated using first-order 1565 kinetics. 1566

Drainage from saturated soil layers is given as a sink term to the vertical one-dimensional flow 1567 equation using seepage potential theory (Leeds-Harrison et al., 1986) for saturated layers above drain 1568 depth and the second term of the Hooghoudt equation for layers below drain depth. Perched water 1569 tables are also considered. The bottom boundary condition utilised for the FOCUS surface water 1570 scenarios is a vertical seepage rate calculated as an empirical linear function of the height of the water 1571 table in the soil profile. PPP movement to the drains is calculated assuming perfect mixing in the 1572 lateral dimensions in each saturated soil layer 1573

FOCUS defined a sixteen-month assessment period for simulation of drainage inputs to surface 1574 waters. The weather data for the first 12 months of the assessment period were chosen by FOCUS to 1575 represent the 50th percentile year with respect to annual rainfall (the remaining 4 months were simply 1576 selected as the period following the selected 12-month period). As, especially for persistent 1577 compounds, the travel time of the PPP to the drains can be significantly longer than sixteen months, 1578 FOCUS decided to employ a six-year warm-up period, in the same way as in the FOCUS groundwater 1579 scenarios (FOCUS, 2000). 1580



When using MACRO the user cannot enter explicit application dates. Instead the application has to be 1581 expressed as an application window which is used as input for PAT (so called pesticide application 1582 timer). PAT eliminates a significant number of potential application dates due to the requirement that 1583 at least 10 mm of precipitation be received within ten days following application. According to 1584 FOCUS this criteria in the PAT calculator results in selection of application dates which are the 60th to 1585 70th percentile wettest days for non-irrigated crops and the 50th to 60th percentile wettest days for 1586 irrigated crops (The slightly lower percentile values for irrigated crops are due to the additional 1587 number of wet days created by irrigation events for these crops.). 1588

One of five different application methods can be selected by the user: ground spray, air-blast, granular, 1589 incorporated and aerial. Interception is assumed zero for both granular and incorporated applications. 1590 For air-blast applications and for ground and aerial sprays to perennial crops, the interception is 1591 assumed to always equal the maximum interception fraction. For annual crops a fraction of the dose 1592 specified by the user is calculated as being intercepted by the crop canopy dependent on the 1593 application day(s) calculated by PAT. This is given as a function of the method of application, a 1594 maximum interception reached at the maximum leaf area, and the leaf area index at the time of 1595 application. 1596

Hourly values of water discharges through drains, and the PPP loads in the discharge during the 1597 assessment period are saved to an output file, which is then used as input to the surface water fate 1598 model TOXSWA. 1599

1600

Calculating run-off and erosion entries for TOXSWA with PRZM 1601

The Pesticide Root Zone Model (PRZM) was selected to calculate run-off and erosion loadings for the 1602 the Step 3 FOCUS surface water scenarios. It is a one-dimensional, dynamic, compartmental model 1603 that can be used to simulate chemical movement in unsaturated soil systems within and immediately 1604 below the root zone. It has two major components – hydrology and chemical transport. The hydrologic 1605 component for calculating run-off and erosion is based on the USDA Soil Conservation Service curve 1606 number methodology and a watershed-scale variation of the Universal Soil Loss Equation. 1607 Evapotranspiration is composed of evaporation from crop interception, evaporation from soil and 1608 transpiration from the crop. Water movement is simulated by the use of generalised soil parameters, 1609 including field capacity, wilting point and saturation water content (Carsel et al, 1995). 1610

Hydrologic and hydraulic computations in PRZM are performed on a daily time step. To compensate 1611 for the (compared to the other step 3 models) long time step and to couple the run-off and erosion 1612 results simulated by PRZM with the transient hydrology incorporated in TOXSWA, the daily run-off 1613 and erosion time series output files are post-processed by FOCUS into a series of hourly run-off and 1614 erosion values by distributing the daily values linearly over a number of hours. This number equals the 1615 rainfall event size divided by an average rainfall intensity of 2 mm.h-1, (so, if e.g. there was 18 mm 1616 rainfall causing 4.1 mm run-off the run-off event lasts 18 mm/2 mm.h-1 = 9 h and so, from midnight to 1617 9 am there is 4.1 mm/9 h = 0.46 mm.h-1 run-off). 1618

The erosion loadings and chemical fluxes in run-off and erosion are handled in a similar manner. 1619

The curve numbers used in PRZM were defined by FOCUS as a function of soil type, soil drainage 1620 properties, crop type and management practice. The curve numbers are used to determine a watershed 1621 retention parameter, which in turn determines the daily run-off as follows: 1622

Equation 4.5 S = 1000 mm/RCN – 10 mm 1623

1624

where S = daily watershed retention parameter [mm] 1625



RCN = run-off curve number (dimensionless, adjusted daily depending upon 1626 antecedent moisture) 1627

1628

Equation 4.6 SSMP

SSMPQ⋅++

⋅−+=

8.0)2.0( 2

1629

1630

where Q = daily run-off [mm] 1631

P = daily precipitation [mm] 1632

SM = daily snow melt [mm] 1633

S = daily watershed retention parameter [mm]. 1634

1635

The following equation is used to calculate soil erosion by PRZM: 1636

Equation 4.75 MUSS: PSCLSKAqQXe p ⋅⋅⋅⋅⋅= 009.065.0)(79.0 1637

Where Xe = the event soil loss [metric tonnes day-1] 1638

Q = volume of daily run-off event [mm] 1639

qp = peak storm run-off [mm/h], determined from generic storm hydrograph 1640

A = field size [ha] 1641

K = soil erodability factor [dimensionless] 1642

LS = length-slope factor [dimensionless] 1643

SC =soil cover factor [dimensionless] 1644

P = conservation practice factor [dimensionless]. 1645

This expression depends primarily upon daily run-off volumes and rates as well as the conventional 1646 USLE factors K, LS, C and P. The MUSS equation is a modification of the MUSLE (Modified 1647 Universal Soil Loss Equation), developed by Williams (1975). 1648

When calculating run-off and erosion losses PRZM always runs over 20 years of data. However, 1649 nevertheless FOCUS selected one representative 12-month period for each use pattern being evaluated 1650 in Step 3. The representative years selected for creation of PRZM output files for use by TOXSWA 1651 are given in the Table 4.6. For example, an application to maize, which occurs in June would result in 1652 selection of the following 12-month period for Scenario R3: June 1975 to June 1976. 1653

Table 4.6: Selected years for creation of PRZM to TOXSWA (FOCUS 2001) 1654

Scenario Date of First Application

March to May June to September October to February

R1 1984 1978 1978

R2 1977 1989 1977

R3 1980 1975 1980

R4 1984 1985 1979

1655

According to FOCUS the reason for not using the whole simulation period of 20 years as input was the 1656 computational requirements of TOXSWA. 1657



1658

The Pesticide Application Timer (PAT) in MACRO and PRZM 1659

When working with MACRO and PRZM the user cannot enter application dates directly. Instead this 1660 is done by a similar pesticide application timer (PAT) which uses an application window as input. 1661 PAT then attempts to select appropriate application dates that meet two criteria: 1662

• No more than 2 mm/day of precipitation should occur on any day within two days before or after 1663 an application 1664

• At least 10 mm of precipitation (cumulative) should occur within 10 days after an application 1665 If, however, no dates are found in the meteorological files that meet these criteria, the precipitation 1666 targets and timing in the two rules are progressively relaxed until acceptable application dates are 1667 found. 1668

MACRO as well as PRZM calculate a fraction of the dose as being intercepted by the crop canopy. In 1669 both models the user can select the application methods of ground spray, air-blast, granular, 1670 incorporated and aerial. Interception is assumed zero for both granular and incorporated applications. 1671

Calculating the fate of compounds in surface water with the TOXSWA model 1672

The TOXSWA model describes the behaviour of PPPs in a water body at the edge-of-field scale, i.e. a 1673 ditch, pond or stream adjacent to a single field. It calculates PPP concentrations in the both the water 1674 and sediment layers. In the water layer, the PPP concentration varies in the horizontal direction 1675 (varying in sequential compartments), but is assumed to be uniform throughout the depth and width of 1676 each compartment. In the sediment layer, the PPP concentration is a function of both horizontal and 1677 vertical directions. 1678

TOXSWA considers four processes: (i) Transport, (ii) Transformation, (iii) Sorption and (iv) 1679 Volatilisation. In the water layer, PPPs are transported by advection and dispersion, while in the 1680 sediment, diffusion is included as well. The transformation rate covers the combined effects of 1681 hydrolysis, photolysis (in cases where this is accounted for in the experimental setup used to derive 1682 this parameter value) and biodegradation and it is a function of temperature. Sorption to suspended 1683 solids and to sediment is described by the Freundlich equation. Sorption to macrophytes is described 1684 by a linear sorption isotherm but this feature is not used in the TOXSWA in FOCUS model used for 1685 the FOCUS surface water scenarios. PPPs are transported across the water-sediment interface by 1686 advection (upward or downward seepage) and by diffusion. In the FOCUS surface water scenarios, 1687 transport across the water-sediment interface takes place by diffusion only. 1688

The mass balance equations for the water and sediment layers are solved with the aid of a generalised 1689 finite-difference method. For the numerical solution, the water layer is divided into a number of nodes 1690 in the horizontal direction. Below each water layer node, an array of nodes is defined for the sediment 1691 layer. Distances between the nodes in the water and sediment layers are in the order of magnitude of 1692 metres and millimetres, respectively. 1693

TOXSWA in FOCUS handles transient hydrology and PPP fluxes resulting from surface run-off, 1694 erosion and drainage as well as instantaneous entries via spray drift deposition. In order to simulate 1695 the flow dynamics in an edge-of-field water body in a realistic way, the field-scale system is defined 1696 as the downstream part of a small catchment basin. 1697

The water body system in TOXSWA has been described with the aid of a water balance that accounts 1698 for all incoming and outgoing water fluxes. The incoming fluxes include the discharge from the 1699 upstream catchment basin (base flow component plus run-off or drainage component), the run-off or 1700 drainage fluxes from the neighbouring field. The outgoing fluxes are composed of the outgoing 1701 discharge of the water body and, if desired, a downward seepage through the sediment. The water 1702 fluxes in the modelled system vary in time as well as in space, i.e. with distance in the water body. 1703



The water level in the water body varies in time, but it is assumed to be constant over the length of the 1704 water body. 1705

The TOXSWA model does not simulate the drainage or run-off/erosion processes itself, but uses the 1706 fluxes calculated by other models as entries into the water body system of TOXSWA. For this 1707 purpose the PRZM in FOCUS model for run-off/erosion and the MACRO in FOCUS model for 1708 drainage create output files that list the water and mass fluxes as a function of time on an hourly basis. 1709 TOXSWA uses these output files as input to calculate the hydrologic and PPP behaviour in the 1710 appropriate water body systems. 1711

The variation of the water level in time has been calculated in two ways. For a pond, outflow is 1712 assumed to occur across a weir and the water level in the pond is derived with the aid of a classical 1713 Q(h) relation for a broad-crested weir (Ministère de la Coopération, 1984). In the case of a 1714 watercourse, the water level is calculated by a combination of uniform flow for which the Chézy-1715 Manning equation can be applied in the upstream part of the watercourse and a backwater curve in 1716 front of a weir at the downstream end (Chow, 1959). The water level in the FOCUS stream and ditch 1717 vary with time, but are assumed constant over their 100 m length (Adriaanse and Beltman, 2009). 1718

The FOCUS ditch only occurs in FOCUS drainage scenarios where the land is relatively flat and, in 1719 most cases, relatively slowly drained. The ditch is assumed to be 100 m long and 1 m wide, with a 1720 rectangular cross-section. Its minimum depth is 0.3 m, implying that in all ditches an outflow weir 1721 maintains this minimum water level even during periods of very low discharge. It receives drainage 1722 fluxes from a 1 ha field adjacent to the ditch and from a 2 ha upstream catchment. PPP solute is only 1723 present in drainage waters from the 1 ha field adjacent to the ditch. The upstream catchment basin is 1724 assumed to be not treated with PPPs, therefore it is considered that only one third of the area 1725 considered in the ditch scenarios is treated with PPP. 1726

The FOCUS stream occurs in the FOCUS drainage scenarios as well as the FOCUS run-off scenarios. 1727 Similar to the FOCUS ditch, the stream is assumed to be 100 m long and 1 m wide, with a rectangular 1728 cross-section. Its minimum depth is 0.3 m, implying that also in all streams a weir is located that 1729 maintains the 0.3 m water level even during periods of very low discharge. On one side of the stream a 1730 1ha field is located that delivers its drainage or run-off fluxes into the stream. This field is assumed to 1731 be treated with PPPs. The stream is also fed by the discharge of an upstream catchment basin of 100ha 1732 which delivers its constant base flow plus variable drainage or run-off water fluxes to the stream. A 1733 surface area of 20% of the upstream catchment basin is assumed to be treated with PPPs resulting in 1734 the dilution of edge-of-field drainage or run-off concentrations by an approximate factor of 5 before it 1735 enters the stream. The implications of PPP contribution from the upstream catchment is simplistically 1736 represented via an increase in the drift loading in the TOXSWA input file by a factor of 1.2 (eg 1737 additional 20% loading). 1738

Pond scenarios represent the simplest arrangement. Each 30m x 30m pond receives drainage or run-off 1739 waters with associated PPP in solution from a 4500 m2 contributing catchment. No PPP is present in 1740 the base flow that enters the pond. For run-off scenarios, the pond also receives eroded sediment and 1741 associated PPP from a 20 m ‘corridor’ adjacent to the pond. 1742

Dominance of entry routes 1743

The FOCUS surface water scheme incorporates three potential routes of entry to surface water (spray 1744 drift, run-off and drainage). In aquatic exposure assessment using FOCUSsw modelling (and therefore 1745 risk assessment), active substances applied as sprays can generally be differentiated between those 1746 where spray drift is the dominant potential route of input to surface water and those where run-off 1747 and/or drainflow is the major potential input route. 1748

If substances are applied as a spray and have a high potential for adsorption to soil particles (high 1749 Koc), the spray drift route of input usually dominates. Profiles from Step 3 FOCUSsw for these 1750 compounds give a distinct pulse of exposure for each spray drift event. The duration of this pulse is 1751



shortest for streams (dissipation driven primarily by advection) and longest for ponds (dissipation 1752 driven primarily by degradation and partitioning to sediment). For two reasons the initial PECsw values 1753 of active substances in streams and ditches are higher than for ponds, (1) because of the relative depth 1754 of the systems, and (2) because of the bigger surface water body area of the pond which results in 1755 lower spray drift deposition. This type of PECsw values is usually not sensitive to climate parameters. 1756 For the reasons relating to the method of selecting overall 90th percentile drift inputs for multiple 1757 application uses already discussed at Step 2, where spray drift is the dominant entry route and there is 1758 relatively fast dissipation in the water body between entry events, simulations have to be carried out 1759 for a single application as well as multiple applications, to ensure that appropriate peak concentrations 1760 are generated and available for use in the risk assessment. Unlike the Step 2 tool, the SWASH shell 1761 does not generate these single application simulations automatically. The user has to define the single 1762 application simulations in addition to the multiple application simulations. 1763

If, in contrast the compound is characterised by high solubility in water, low Koc, and relatively long 1764 DegT50 in soil and the method of application favours run-off and drainage entries rather than spray 1765 drift (e.g. soil incorporation, low drift values as in field crops) run-off and/or drainflow becomes the 1766 major potential route of input to surface water. This type of PECsw values is sensitive to the rainfall 1767 pattern shortly after application as both processes are event driven. 1768

4.2.1.4. Step 4 1769

Step 4 simulations are usually performed considering the results of the FOCUS group on Landscape 1770 and Mitigation Measures in Ecological Risk Assessment (FOCUS 2007a, 2007b). 1771

Similar to the other tiers also for Step 4 a software tool (SWAN) is recommended by FOCUS, which 1772 is available and developed on behalf of ECPA. For interpretation of the mitigation of run-off in the 1773 FOCUS surface water scenarios as described by FOCUS in its Landscape and Mitigation report 1774 (FOCUS, 2007a and also see Ter Horst et al, (2009)).The software modifies the input and output files 1775 of the Step 3 models TOXSWA and PRZM to consider drift and run-off buffer zones. The standard 1776 exposure reduction factors for run-off (water volume and PPP mass in run-off water) and erosion 1777 (eroded soil and PPP mass sorbed to eroded soil) as suggested by FOCUS (2007a, 2007b) are shown 1778 in the Table 4.6. 1779

1780

Table 4.6: 90th percentile worst-case values for reduction efficiencies for different widths of 1781 vegetated buffer strips and different phases of surface run-off (taken from FOCUS 2007a) 1782

Buffer width (m) 10-12 18-20 Reduction in volume of run-off water (%) 60 80 Reduction in mass of PPP transported in aqueous phase (%)

60 80

Reduction in mass of eroded sediment (%) 85 95 Reduction in mass of PPP transported in sediment phase (%)

85 95

1783 It should be noted that whilst SWAN can be used to parameterise run-off loadings with greater 1784 reduction values than those indicated in table 4-6, FOCUS (2007a) prescribes a ceiling on run-off 1785 mitigation of 90% runoff entry reduction. Regulatory practice is that this means water volume and 1786 substance solute mass should not be reduced by >80% and mass of eroded sediment / substance mass 1787 in the eroded sediment should not be reduced by >95%. Alternatively, in regulatory practice, it has 1788 been accepted to have all 4 of these parameters reduced by a maximum of 90%. In regulatory practice, 1789 other combinations of reduction approaches might also be accepted, but it is the responsibility of the 1790 applicant to clearly demonstrate that the approach that they have taken, respects the ceiling of not 1791 reducing run-off inputs by more than 90% of that calculated by PRZM at Step 3. 1792



SWAN can also handle drift reduction due to the use of more advanced nozzle techniques (low-drift 1793 nozzles). In addition, to the entry routes considered in the first three tiers the exposure via air for 1794 volatile substances, using the recommendations of the FOCUS air group (FOCUS 2008), can be 1795 considered. 1796

The effect of drift buffer zones (i.e. no-spray buffer zones) can be considered in SWAN for distances 1797 up to 100 m from the surface water body. The model considers the same reduction rates as in the 1798 FOCUS SWASH tool both based on BBA (2000). It should be noted that whilst SWAN can be used to 1799 parameterise drift buffer zones up to 100m and the effects of low drift nozzles can be combined with 1800 drift buffer zones to reduce spray drift inputs still further, that FOCUS 2007a prescribes a ceiling on 1801 spray drift mitigation. This prescription is that spray drift cannot be mitigated such that the mass per 1802 unit area reaching the water body surface is <5% of this value that would be calculated for the crop of 1803 interest using the FOCUS defined baseline distance for that crop (1 - 6 m). I.e. the ceiling for spray 1804 drift mitigation is 95%. 1805

As at Step 3, when a use pattern includes multiple applications, it can also be necessary to simulate a 1806 single application as well as multiple applications at Step 4, to ensure that appropriate peak 1807 concentrations are generated and available for use in the risk assessment. The need for this procedure 1808 to be necessary, reduces, as the extent of spray drift mitigation implemented increases. 1809

1810

4.2.2. Assessment of metabolites by FOCUS surface water modelling 1811

To be added, text will be depending on new releases of FOCUS version control group. 1812

1813 1814



5. Data Requirement for active substances and Formulations and Tier 1 effect assessment 1815

5.1. Introduction to data requirements as laid down in SANCO document 11802 and 1816 11803/2012 for approval of active substances and plant protection products and related 1817 OECD guidelines 1818

The data requirements state that the ecotoxicological risk assessment shall be based on the risk that the 1819 proposed active substance and the formulated plant protection product poses to non-target organisms. 1820 In carrying out a risk assessment it is necessary to compare toxicity with exposure. 1821

The specific data requirements for Regulation (EC) 1107/2009 concerning the placing of plant 1822 protection products on the market are laid down in SANCO document 11802/2012 for the dossier to 1823 be submitted for the approval of active substances contained in plant protection products and in the 1824 SANCO document 11803/2012 for the authorisation of plant protection products. 1825

Test Guidelines 1826

Studies should always be assessed according to recognised test guidelines. Test guidelines are 1827 specified in “SANCO/11844/2010 Commission Communication on test methods and guidance 1828 documents under the Regulation”. Specific guidance for testing difficult substances and mixtures in 1829 aquatic toxicity tests is provided in the OECD guidance document 23 (OECD, 2000). 1830 An overview of the obligatory and additional toxicity tests (that should be provided under certain 1831 circumstances) by the applicant is presented in table 5.1. 1832

1833 Table 5.1: Ecotoxicity studies required for active substances under certain circumstances (for 1834 metabolites see section 8.3) 1835

Acu

te to

xici

ty te

st to

fish

(rai

nbow

trou

t)

Acu

te to

xici

ty te

st to

Dap

hnia

Toxi

city

test

to g

reen

alg

a

Acu

te to

xici

ty te

st to

Chi

rono

mus

ssp

or

Amer

icam

ysis

bah

ia

Toxi

city

test

to a

lgae

(not

gre

en a

lga,

e.g

. dia

tom

N

avic

ula

pelli

culo

sa)

Toxi

city

test

to L

emna

Toxi

city

on

othe

r mac

roph

yte

spec

ies (

e.g.

M

yrio

phyl

lum

or G

lyce

ria)

Fish

ear

ly-li

fe st

age

toxi

city

test

(EL

S)

Long

term

/ chr

onic

toxi

city

test

on

Dap

hnia

mag

na o

r in

cas

e of

two

test

ed sp

ecie

s in

the

acut

e si

tuat

ion

a te

st w

ith th

e sp

ecie

s tha

t sho

wed

the

low

est e

ndpo

int

Chr

onic

spik

ed to

xici

ty te

st o

n C

hiro

nom

us r

ipar

ius

or L

umbr

icul

us sp

p

Fish

shor

t ter

m re

prod

uctiv

e as

say,

21-

day

fish

assa

y or

fish

sexu

al d

evel

opm

ent t

est

Fish

full

life

cycl

e to

xici

ty te

st (F

FLC

)

Every substance x x x Substances with insecticidal mode of action (MOA)

x

Substances with a herbicidal mode of action or plant growth regulators

x x

Substances with aherbicidal MOA for which Lemna is not

x



sensitive or there is expected uptake by the roots of submerged macrophytes Where exposure of surface water is possible and the substance does not hydrolyse instantly (DT90>1d)

x(a) x

Accumulation of the substance in sediment indicated or predicted from fate studies(b)

x

Substances which are suspected to interfere with the moulting hormones (e.g. insect growth regulator)

x

Substances which are identified as a potential endocrine disrupter

x

Substances which have an indication of endocrine mediated effects

x

(a) Unless a Fish Full Life Cycle (FFLC) test is provided 1836 (b) Water/sediment study showed >10% of applied radioactivity at or after day 14 present in the sediment and chronic daphnia 1837

test (or other comparable study with insects) NOEC< 0.1 mg/L. For the time being, the guidance as given in the former 1838 SANCO guidance (2002) should be followed. This might be revised in the future, depending on the PPR Panel opinion on 1839 sediment effect assessment under development (EFSA-Q-2012-00959). 1840

1841

5.2. Standard toxicity tests with aquatic organisms 1842

The following toxicity tests should be submitted for every substance even when it is not expected that 1843 preparations containing it could reach surface water following the proposed conditions of use: acute 1844 toxicity to fish (Oncorhynchus mykiss), acute toxicity for Daphnia species (preferably for Daphnia 1845 magna) and effects on the growth for a green alga (e.g. Pseudokirchneriella subcapitata). 1846

A limit test at 100 mg substance/L may be performed when the results of a range finding test indicate 1847 that no effects are to be expected. In order to minimise fish testing, a threshold approach to an acute 1848 fish testing should be considered. An acute fish limit test should be conducted at 100 mg substance/L 1849 or at an appropriate concentration selected from aquatic endpoints following consideration of the 1850 threshold exposure. When mortality is detected in the fish limit test an acute fish dose-response 1851 toxicity study should be performed to determine an LC50 for use in risk assessment. 1852

The endpoints to be provided for the acute toxicity tests are: 1853 • For fish the tests should provide the 24, 48, 72 and 96-hour toxicity of the substance 1854

expressed as the median lethal concentration (LC50), and where possible the highest 1855 concentration causing no lethal effects and details of observed effects. 1856

• For aquatic invertebrates the test should provide the 24 and 48-hour acute toxicity of the test 1857 substance, expressed as the median effective concentration (EC50) for immobilisation, and 1858 where possible the highest concentration causing no immobilisation and details of observed 1859 effects. 1860

• For algae the test should provide the average specific growth rate inhibition (EC50) (72 1861 hours). 1862



The endpoints required in the “revised data requirements” for the standard tests and the other 1863 requested additional types of tests are EC10, EC20, EC50, together with their 95 % confidence intervals 1864 (or an explanation if they cannot be estimated) and corresponding NOEC values. How these should be 1865 determined is explained below. 1866

Reasoning for introduction of new endpoints (ECx) 1867 Traditionally, the responses measured in laboratory ecotoxicity tests have been expressed as the No 1868 Observed Effect Concentration (NOEC) and/or Lowest Observed Effect Concentration (LOEC). The 1869 PPR Panel recommends that risk assessment for all groups of organisms and based on laboratory 1870 toxicity tests should use the ECx approach in preference to NOEC/LOEC approach (EFSA 2009a). The 1871 recommendations for the choice of ECx for each group of organisms should be based on an analysis of 1872 existing study data for each of these groups and take into consideration the following issues: 1873

• The ECx for each group of organisms should be chosen so as to ensure a level of protection 1874 consistent with the aims of the regulations, taking into account the conservatism of other parts of 1875 the risk assessment. 1876

• The choice of ECx should take into account the reliability of the estimates that can be provided by 1877 standard test methods. 1878

• The procedure for using the chosen ECx in the risk assessment should take account of the quality 1879 of the estimates available for each substance, for example by examining confidence intervals for 1880 the ECx and possibly using these in the risk assessment. 1881

• As existing study methods were not designed to estimate ECx, it is expected that a proportion of 1882 existing studies will not provide a usable estimate for the preferred (or even any) ECx. For reasons 1883 of cost and animal welfare, the procedure should be designed to minimise or, if possible, prevent 1884 any need for retesting in such cases. For example, the procedure could include provision to use 1885 alternative ECx or even the NOEC in such cases, together with appropriate adjustments to the risk 1886 assessment (e.g. different assessment factor) to provide the required level of protection. 1887

• The desirability of harmonising with ECx approaches used under other EU legislation (e.g. 1888 REACH; EC, 2006) should be considered, unless there are good reasons to differ. 1889

1890 In order to provide the flexibility to accommodate these issues in the final procedures and for the 1891 Commission Regulation (EU) No 546/2011 to be revised, the choice of ECx, and whether or not to use 1892 confidence intervals on the ECx should ideally be left open. The possibility to use the NOEC where 1893 necessary should also be retained. This will have the added benefits of (a) providing an indication of 1894 the slope of the dose-response (by comparing the EC10, 20 and 50) and (b) helping the transition from 1895 NOEC to ECx by presenting both together. 1896

The PPR Panel recognises that the lower limit of the confidence interval of the EC20 might be more 1897 appropriate, however, for practicality reasons it is proposed to use for the time being EC10 or when not 1898 available the NOEC in accordance with the Technical Guidance Document 27 to the Water 1899 Framework Directive (EC 2011) until new knowledge on the choice of ECx becomes available. If the 1900 EC10 is used, the same assessment factor should be applied as if a NOEC would be used. 1901

5.2.1. Fish 1902

5.2.1.1. Acute toxicity to fish 1903

A test on rainbow trout (Oncorhynchus mykiss) with the active substance shall always be carried out, 1904 even when it is not expected that the compound will end up in the surface water. In that case it will be 1905 used for classification and labelling. Consideration should be given to allow the reduction of animal 1906 testing (see also section 9.4). 1907

1908



5.2.1.2. Chronic toxicity to fish 1909

Circumstances in which required 1910 A long-term or chronic toxicity study on fish is required for all substances where exposure of surface 1911 water is likely and the substance is deemed to be stable in water, i.e. there is less than 90 % loss of the 1912 original substance over 24 hours via hydrolysis at all relevant pH (hydrolysis DT90>24 hours). An 1913 early-life stage study is always required in these circumstances, however, if a full fish life cycle study 1914 has been generated an early-life stage study is not required. 1915

The fish early-life stage test should determine effects on development, growth and behaviour, and 1916 details of observed effects on fish early-life stages. 1917

A full fish life cycle test may be required depending upon the persistence and bioaccumulative 1918 potential of the substance11. 1919

For substances that fulfil the screening criteria on either of the fish screening assays, or where there 1920 are other indications of endocrine disruption (see paragraph 5.6), appropriate additional endpoints 1921 should be included in the test as recommended by the OECD conceptual framework in support of 1922 testing and assessment of potential Endocrine Disrupter and discussed with the national competent 1923 authorities. 1924

The test conditions should be designed to reflect concerns identified in already available aquatic 1925 toxicity tests, mammalian and bird toxicology studies and other information. The exposure regime 1926 should be selected accordingly, taking account of the rates of application proposed. 1927

5.2.2. Amphibians 1928

Even if the revised data requirements (SANCO document 11802/2012) do not request toxicity tests for 1929 amphibian species, amphibians should be included in the aquatic and terrestrial Risk Assessment of 1930 PPPs. Assessment of the risk to amphibians should be based on any existing relevant information. 1931 Available relevant data, including data from the open literature, for the substance under consideration 1932 should be presented and taken into account in the risk assessment. Terrestrial life stages of amphibians 1933 will be addressed in a future GD on PPP RA for amphibians and reptiles (EFSA-Q-2011-00987) under 1934 the mandate of the revision of the GD on Terrestrial Ecotoxicology. In this guidance, only aquatic life 1935 stages are addressed. 1936

An analysis of acute toxicity data for a large number of amphibian species (Fryday and Thompson, 1937 2012) and comparison to fish acute toxicity data (see Appendix C) shows that the rainbow trout is a 1938 good surrogate test species for predicting the acute toxicity of PPPs for larval stages of amphibian 1939 species living in the aquatic compartment of the environment. By using the same assessment factors as 1940 have been applied for fish, the achieved level of protection will be the same for both groups of 1941 organisms. The assessment is only valid for acute toxicity (mortality) and will not necessarily be 1942 predictive of chronic toxicity. If a refinement of the RA needs to be performed for fish, it needs to be 1943 investigated if the refined RA for fish still covers amphibians. More guidance is given related to the 1944 use of the SSD approach (section 6.3.5) and for refined exposure studies (section 7.2). 1945

5.2.3. Aquatic invertebrates 1946

An acute test with Daphnia magna (e.g. OECD 202) has always to be carried out, even when it is not 1947 expected that the compound will end up in the surface water. In that case it will be used for 1948 classification and labelling. 1949

11In line with the former GD on Aquatic Ecotoxicology (SANCO (2002) the Panel recommends that FFLC-tests may be

required where the BCF is >1000, the elimination during the 14 day depuration phase in the bioconcentration study is <95% or the substance is stable in water or sediment (DT90 >100 days).



In addition, for active substances with an insecticidal mode of action or which show insecticidal 1950 activity12 a second species shall be tested, for example chironomid larvae or mysid shrimps 1951 (Americamysis bahia). The PPR Panel proposes that if Daphnia is an order of magnitude more 1952 sensitive than algae or fish, a second test species is required as well. 1953

A long-term or chronic toxicity study on aquatic invertebrates should be provided for all active 1954 substances where exposure of surface water is likely and the substance is deemed to be stable in water, 1955 that is to say there is less than 90 % loss of the original substance over 24 hours via hydrolysis 1956 (hydrolysis DT90>24 hours, see point 7.2.1.1 of the data requirements). 1957

A chronic toxicity study should be submitted on one aquatic invertebrate species. If acute tests have 1958 been conducted on two aquatic invertebrate species the acute endpoints should be taken into account 1959 (see point 8.2.4 of the data requirements) in order to determine the appropriate species to be tested in 1960 the chronic toxicity study. The PPR Panel proposes to select the more sensitive species in case the 1961 difference in acute toxicity is more than a factor of 10. 1962

If the test substance is suspected of interfering with moulting hormones, i.e. it is an insect growth 1963 regulator, or that has other effects on insect growth and development, an additional study on chronic 1964 toxicity shall be carried out using relevant non-crustacean species such as Chironomus spp. 1965

5.2.3.1. Toxicity studies with sediment dwelling organisms 1966

When accumulation of an active substance in aquatic sediment is indicated or predicted by 1967 environmental fate studies13, the impact on a sediment-dwelling organism shall be assessed. The 1968 chronic risk to Chironomus riparius (OECD 218, 219) or Lumbriculus spp (OECD 225) shall be 1969 determined. 1970

An appropriate alternative test species may be used where a recognised guideline is available. The 1971 active substance should be applied to either the water or the sediment phase of a water/sediment 1972 system and the test should take account of the major route of exposure. 1973

The key endpoint from the study should be presented in terms of mg substance/kg dry sediment and 1974 mg substance/L water. 1975

The PPR Panel recommends to preferably conduct a water spiked study. Sediment-spiked studies 1976 could be part of higher tier testing. This GD focuses on exposure via the water phase. A scientific 1977 opinion addressing the effect assessment for sediment organisms in detail will be developed by the 1978 PPR Panel in the near future. 1979

There are two OECD guidelines available for testing either spiked-water or spiked-sediment (OECD 1980 218 and 219). First instar chironomid larvae are exposed. Chironomid emergence and development 1981 rate is measured at the end of the test. The maximum exposure duration is 28 days for C. riparius, C. 1982 yoshimatsui, and 65 days for C. dilutus (formally C. tentans). 1983

12Data for the non-target arthropods could be used for assessing the potential insecticidal activity of a compound. For most of

the compounds the two standard non target arthropods are tested (Typhodromus pyri and Aphidius rhopalosiphi). When the quotient of the application rate multiplied by a MAF factor and the LR50 is greater than 2 the compound could be considered as having insecticidal activity. In addition efficacy studies with other insects or studies carried out with insects in the screening process could be another source for assessing potential insecticidal activity. In addition data for bees can also be used, if the hazard quotient (HQ) of the application rate in g/ha divided by the acute toxicity (LD50) ≥ 50 the compound can be considered to have insecticidal activity.

13 As described in Table 5.1: Water/sediment study showed >10% of applied radioactivity at or after day 14 present in the sediment and chronic daphnia test (or other comparable study with insects) NOEC< 0.1 mg/L. For the time being, the guidance as given in the former SANCO guidance (2002) should be followed. This might be revised in the future, depending on the PPR Panel opinion on sediment effect assessment under development (EFSA-Q-2012-00959).



An OECD draft guideline is available for assessing the effects of life-long exposure of chemicals to 1984 the freshwater dipteran Chironomus spp, fully covering the first generation and the early part of the 1985 second generation. It is an extension of the existing OECD test guideline 219. 1986

A sediment-spiked test for Lumbriculus variegatus is described in OECD guideline 225. Lumbriculus 1987 spp. are exposed for 28 days and effects on reproduction and biomass are observed. 1988

5.2.4. Standard toxicity tests with algae 1989

Testing should always be carried out on one green alga (such as Pseudokirchneriella subcapitata, 1990 synonym Selenastrum capricornutum). For active substances that exhibit herbicidal activity14, a test 1991 on a second species from a different taxonomic group should be performed such as a diatom, for 1992 example Navicula pelliculosa. 1993

The algal test (OECD guideline 201) is a short-term test although it provides both acute and chronic 1994 endpoints. The preferred observational endpoint in this study is algal growth rate inhibition because it 1995 is not dependent on the test design, whereas an endpoint based on biomass is dependent on both the 1996 growth rate of the test species as well as test duration and other elements of test design. Often both 1997 acute growth rate EC50 (ErC50) and biomass (EbC50) endpoints are reported however the latter should 1998 not be used. The reason is that direct use of the biomass concentration without logarithmic 1999 transformation cannot be applied to an analysis of results from a system in exponential growth 2000 (ECHA, 2008). Where only the EbC50 is reported, but primary data are available, a re-analysis of the 2001 data should therefore be carried out to determine the ErC50. However, if only EbC50values are 2002 presented, this value can be used as well. The result for the endpoint biomass (growth) is generally 2003 somewhat lower than the growth rate and can therefore be considered as a conservative value. In the 2004 revision of the OECD 201 guideline in 2011, the calculation of a toxicity endpoint based on biomass is 2005 no longer recommended. Instead the guideline now provides methods for calculating growth rate and 2006 yield (which may be needed to fulfil specific regulatory requirements in some countries outside of the 2007 EU). 2008

5.2.5. Standard toxicity tests with macrophytes 2009

If a substance is a herbicide, plant growth regulator or shows herbicidal activity13, a test with Lemna 2010 spp. should be carried out. Tests could be performed according to OECD 221 test guideline. In this test 2011 exponentially growing plant cultures of the genus Lemna (Lemna gibba and Lemna minor usually) are 2012 allowed to grow as monocultures over a period of seven days. The objective of the test is to quantify 2013 substance-related effects on vegetative growth over this period based on assessments of selected 2014 measurement variables. This study includes the counting of the frond number and measurement of at 2015 least one other variable (total frond area, dry weight or fresh weight) using the lowest of these 2016 endpoints for the risk assessment. 2017

Additional testing may be required by the Member State competent authorities on other macrophyte 2018 species depending on the mode of action of the substance, or if clear indications of higher toxicity are 2019 apparent to dicotyledonous (for example auxin inhibitor, broad leaf herbicides) or other 2020 monocotyledonous (for example grass herbicides) plant species from efficacy or testing with terrestrial 2021 non-target plants. 2022

Additional aquatic macrophyte species tests may be undertaken on a dicotyledonous species, such as 2023 Myriophyllum spicatum, Myriophyllum aquaticum or a monocotyledonous species, such as aquatic 2024 grass Glyceria maxima, as appropriate. The PPR Panel of EFSA recommends to use Lemna sp. as the 2025 default macrophyte test species and to follow the recommendations of the AMRAP workshop (Maltby 2026 et al. 2010) for testing other macrophyte species. In case Lemna and algae are apparently not sensitive 2027 14If in one or more of the screening or efficacy tests with vascular plant species showing ≥ 50% phytotoxic effects at the

maximum recommended application rate (MRR) or higher, the a.s. is considered to have herbicidal activity (see EPPO scheme for higher plants (EPPO, 2003)).



to the herbicidal product (e.g. EC50 > 1mg/L), or if the herbicide simulates a plant growth hormone, a 2028 rooted macrophyte is recommended (preferably Myriophyllum). Glyceria may be tested in case of a 2029 herbicide that primarily affects monocots in terrestrial plant trials. 2030

5.3. Deriving regulatory acceptable concentrations (RAC) 2031

Tables 5.2 and 5.3 summarise how to derive acute and chronic RAC values for each species. 2032

2033

Table 5.2: Endpoints available from short-term aquatic toxicity tests; basic dossier data are 2034 indicated in bold (based on SANCO document 11802/2012 for approval of active substances). 2035

Taxonomic group

Species/test system Duration Endpoint RAC

fish Oncorhynchus mykiss 96 h LC50 LC50/100 crustaceans Daphnia sp. (D. magna preferred) 48 h EC50 EC50/100 insects/ crustaceans

Additional species, e.g. Chironomus ssp or Americamysis bahia

48 h EC50 EC50/100

2036 2037

Table 5.3: Endpoints available from long-term aquatic toxicity tests; basic dossier data are 2038 indicated in bold (based on SANCO document 11802/2012 for approval of active substances). 2039

Taxonomic group

Species/test system Duration Endpoint RAC

fish early life-stage (ELS) test EC10(NOEC) EC10/10 fish fish full life-cycle test (FFLC) test EC10(NOEC) EC10/10 crustaceans Daphnia sp. or additional species 21 d EC10 (NOEC) EC10/10 insects oligochaete

Chironomus spp (water spiked preferred) Lumbriculus spp

20-28 d 28 d

EC10(NOEC) EC10(NOEC)

EC10/10 EC10/10

algae(a)

green algae (e.g. Pseudokirchneriella subcapitata)

72 h EC50 EC50/10

algae(a) diatom(e.g. Navicula pellucilosa)/ blue green algae

72 h EC50 EC50/10

macrophytes(a) Lemna sp. or Myriophyllum sp. or Glyceria maxima

7 d – 14 d EC50 EC50/10

(a) In other EU directives/regulations (e.g. WFD) the EC50 and an AF of 100 is used in the acute effect assessment and the 2040 EC10 (or NOEC) and an AF of 10 in the chronic effect assessment. 2041

2042

5.4. Further testing on aquatic organisms 2043

Additional studies may be conducted to refine the risk identified or address additional concerns, e.g. 2044 endocrine disrupting effects. Studies should provide sufficient information and data to evaluate 2045 potential impact on aquatic organisms under field conditions. 2046

Additional studies undertaken can take the form of additional species testing (chapter 6), modified 2047 exposure testing (section 7.2), microcosm or mesocosm studies (section 7.3). 2048

Where aquatic acute and/or chronic risk is indicated by the risk assessment, expert judgment shall be 2049 used to decide the type of further assessment and additional studies required. This judgment will take 2050 into account the results of any additional data over and above those required by the present 2051 Regulation. 2052



Before performing these studies, agreement of the competent authorities on the specific aims of the 2053 study to be performed and consequently on the type and conditions of the study to be performed 2054 should be sought. 2055

5.5. Specific requirements for formulated products 2056

The requirement for formulation studies are given in SANCO document 11803/2012. Testing of 2057 formulated products shall be performed where: 2058

a) the acute toxicity of the preparation cannot be predicted on the basis of the data for the active 2059 substance, or 2060

b) the intended use can result in direct exposure to water, or 2061 c) the extrapolation on the basis of available data for a similar preparation is not possible. 2062

In principle acute or short term exposure tests should be carried out on one species from each of the 2063 three to four groups of aquatic organisms (fish, aquatic invertebrates, algae and/or macrophytes) if the 2064 preparation itself may contaminate water. However, where the available information permits to 2065 conclude that one of these groups is clearly more sensitive (factor of 10 difference), tests on only the 2066 relevant group have to be performed. 2067

In addition, in the case of herbicides and plant growth regulators and other substances where there is 2068 reason to suspect effects on plants, tests should be carried out on one (in case several species have 2069 been tested, on the most sensitive) species of aquatic macrophyte, if the preparation itself can 2070 contaminate water. 2071

If the preparation contains two or more active substances, and the most sensitive taxonomic groups for 2072 the individual active substances are not the same, testing on all three/four aquatic groups (i.e. fish, 2073 aquatic invertebrates, algae and macrophytes (if relevant)) is required. 2074

In order to minimise fish testing a threshold approach should be considered for testing acute toxicity in 2075 fish (see Section 5.2 and 9.4). 2076

Chronic studies on fish and invertebrates should be conducted for particular preparations where it is 2077 not possible to extrapolate from data obtained in the corresponding studies on the active substance (i.e. 2078 the preparation is more acutely toxic than the active substance as manufactured by a factor of 10), 2079 unless it is demonstrated that exposure will not occur. 2080

If chronic toxicity studies with the preparation are required, generally, studies similar to those 2081 conducted for an active substance are required. However, static tests may be more useful than flow-2082 through studies as the former provide conditions that are slightly more relevant to exposure occurring 2083 under field conditions concerning the exposure of the formulated product. The results from such a test 2084 can be used to assess whether the formulation results in an increased toxicity, it cannot be compared to 2085 the PECsw without considering the exposure profile of the active substance. An alternative is to 2086 conduct a specifically targeted microcosm study with the preparation to investigate long-term risks. 2087

2088

5.6. Bioconcentration and secondary poisoning 2089

Some compounds in the water have the tendency to accumulate in the tissue of fish or in the tissue of 2090 other organisms. This tendency of a compound is often expressed in a bioconcentration factor (BCF). 2091 The equilibrium concentration for a compound in fish can be estimated by multiplying the 2092 concentration of the compounds in the surrounding water by the fish BCF for that particular 2093 compound. At long exposure times (equilibrium), the BCF also equals the ratio of the uptake constant 2094 (Mackay, 1982). 2095



The bioconcentration of the substance should be assessed where: 2096

- the log Pow is greater than 3 or other indications of bioconcentration (for instance monitoring data 2097 in biota or structural alerts (e.g. functional groups with the potential to ionize at natural relevant 2098 pH)), and 2099

- the substance is considered stable (i.e. there is less than 90% loss of the original substance over 24 2100 hours via hydrolysis). 2101

5.6.1. Bioconcentration in fish 2102

The test should provide a steady-state bioconcentration factor (BCF), uptake rate constants and 2103 depuration rate constants, incomplete excretion, metabolites formed in fish and, if available, 2104 information on organ-specific accumulation. The BCF can either be a calculated BCF based on uptake 2105 rate constants and depuration rate constants or a measured BCF in organism tissue at steady state 2106 (OECD guideline 305). 2107

All data should be provided with confidence limits for each test substance. BCF values should be 2108 reported as growth-corrected and as lipid-normalised values. 2109

Data produced for metabolism, distribution and expression of metabolites in case of the use of active 2110 substances in fish farming may also be relevant in addressing this point. 2111

5.6.2. Secondary poisoning 2112

In addition to potential effects on fish, special attention should also be paid for potential transfer of 2113 lipophilic compounds through the food chain. For organic chemicals, a log Pow≥ 3 indicates a potential 2114 for bioaccumulation. If this condition is met, a risk assessment for secondary poisoning should be 2115 carried out. For the aquatic system this risk is assessed for a fish eating bird with a body weight of 2116 1000 g and a fish eating mammal with a body weight of 3000 g 2117

For the stepwise approach for assessing the bioaccumulation potential, see the risk assessment 2118 methodology according to the guidance document for birds and mammals (EFSA, 2009c). 2119

5.6.3. Regulatory acceptable concentration (RAC) for biomagnification 2120

According to the previous aquatic guidance document (EC, 2002), biomagnification has to be taken 2121 into account for compounds that meet the trigger for a FFLC-test, namely the BCF (whole body) > 2122 1000 and the elimination of radioactivity during the 14 day depuration phase in the bioconcentration 2123 study is < 95% and the substance is stable in water or sediment (DT90> 100 days). The previous 2124 guidance document states that if these triggers are met, detailed food chain modelling should be 2125 performed, or microcosm/mesocosm studies, which implicitly take into account biomagnification, 2126 should be submitted. However, the methodology for food chain modelling as proposed in SANCO 2127 (2002) is very complicated and requires a lot of input data. Furthermore, including fish in 2128 microcosm/mesocosm experiments can present difficulties and needs to be carefully considered. It is 2129 therefore proposed to consider food chain modelling as an option for higher tier assessment. It is 2130 therefore recommended that, as a first tier, the methodology of the Technical Guidance Document 2131 (TGD) (EC, 2003) and EQS-guidance (EC, 2011) are adopted and the risk assessment is performed 2132 using default biomagnification factors. The TGD proposes the following factors, related to BCF and/or 2133 log Kow (Table 5.4). 2134

2135

2136

2137



Table 5.4: Default BMF values for organic substances 2138

BCF (fish) BMF

<2000 or log Pow<3 1

2000–5000 2

>5000 10

2139

From this table it can be seen that biomagnification may be relevant for compounds with a BCF ≥ 2140 2000 L/kg. For these compounds, the appropriate BMF will be selected from Table 5.4 and the RACSP 2141 will be derived according to the following formula: 2142

BMFBCF0.1385NOAELor

BMFBCF159.05NOAELRAC

fish

mammal

fish

birdSP ⋅⋅⋅⋅⋅⋅

= 2143

with 2144 2145 RACsp = Regulatory Acceptable Concentration in water for secondary poisoning [mg/L] 2146 NOAEL = relevant long-term no-adverse-effect-level for birds or mammals [mg/kg bw per d] 2147 BCFfish = whole body bioconcentration factor in fish [L/kg] 2148 BMF = biomagnification factor from Table 5.4 [kg/kg] 2149 2150 This RACsp should be compared with the 21-day TWA PEC in surface water. 2151

If RACsp> 21-d TWA PECsw, no further action is required. 2152 If RACsp< 21-d TWA PECsw, refinement is necessary 2153 2154

Where the need for further refinement is triggered, a higher tier assessment should be carried out and 2155 the food-chain modelling approach of the aquatic guidance document (SANCO, 2002) can be 2156 followed. As bioaccumulation processes often are slow and substances may be persistent, a long-term 2157 assessment is appropriate. Relevant metabolites must also be considered. For background information 2158 with regard to food chain modelling see Romijn et al. (1993, 1994), Traas et al. (1996), Jongbloed et 2159 al. (1996) and Luttik (2003). 2160

2161



6. Higher-tier effect assessment on basis of laboratory toxicity tests with standard and 2162 additional species 2163


When assessing active substances and formulated plant protection products additional laboratory 2165 toxicity data exceeding the regulatory requirements (Regulation (EC) 544/2011 and Regulation (EC) 2166 545/2011) will often be available, due to the legal obligations to submit scientific peer-reviewed open 2167 literature data or the need for additional higher-tier effect data for refinements. Whereas scrutiny of 2168 open literature data should focus on identifying additional toxicity data for relevant species, which are 2169 not captured in the standard test package, addition of supplementary laboratory toxicity studies may 2170 aim to reduce uncertainty in the risk assessment, mainly by addressing more realistic exposure profiles 2171 and/or better capturing of inter species variations in sensitivity. 2172

6.1.1. Quality check 2173

There are some general principles that may be considered when assessing the reliability of the studies 2174 that provide data for establishing or refining risk assessment parameters, e.g. statistical power; 2175 verification of measurement methods and data; control of experimental variables that could affect 2176 measurements; universality of the effects in validated test systems using relevant animal/plant strains 2177 and appropriate routes of exposure; biological plausibility of results; and uniformity among substances 2178 with similar attributes and effects (adapted from Becker et al., 2009). 2179

2180 Freshwater and marine species 2181

It is apparent from Regulation (EC) 1107/2009 that the estuarine and marine water bodies are parts of 2182 the environmental compartment where the level of environmental risk should be assessed, e.g. 2183 transitional-, coastal- and marine water are specifically mentioned in the definition of ‘Environment’ 2184 in Article 3 of the Regulation. 2185

The data requirements, as specified in SANCO document 11802/2012 for the active substance and for 2186 formulated plant protection products in SANCO document 11803/2012, cover only freshwater species, 2187 with the exception of the data requirements for Americamysis bahia (former Mysidopsis bahia), which 2188 is a brackish water species. The obligation to include open literature effect data may, however, 2189 introduce further effect endpoints derived for brackish water or marine species. 2190

An important question to consider is whether marine toxicity data can be used for the effect 2191 assessment in the edge of field surface water. From several papers it seems that the sensitivity 2192 distributions of taxonomically similar freshwater and marine species to organic PPPs do not differ 2193 significantly (Maltby et al, 2005; Brock et al 2008; Klok et al 2012), thus indicating that the data can 2194 be combined. Differences in sensitivity may arise due to the effect of test conditions (e.g. salinity) on 2195 exposure profile or when comparing taxonomically dissimilar datasets. Several marine phyla (e.g. 2196 Echinodermata) are not represented in freshwater and therefore issues regarding protectiveness might 2197 arise for organism groups that are not covered in the standard organisms tested. 2198

Only in cases where toxicity data indicate significant differences in sensitivity among relevant 2199 taxonomic groups of marine and freshwater species, the risk assessment for edge of field surface water 2200 should be exclusively based on toxicity data for freshwater species. 2201

6.2. Geometric mean-AF approach 2202

6.2.1. Introduction 2203

If additional species (not belonging to the standard test species mentioned in chapter 5) are tested, it is 2204 necessary to consider which toxicity value should be used in the risk assessment, at least if the number 2205 of available toxicity data is not high enough to apply the Species Sensitivity Distribution (SSD) 2206



approach (see 6.3). According to the former guidance, “if a considerable number of additional species 2207 was tested in valid studies, then it is possible that the assessment factors that are applied to the lowest 2208 toxicity value could be lowered by up to an order of magnitude” (SANCO, 2002). It is not further 2209 specified how many additional data would be needed to allow for lowering the assessment factor, and 2210 to our experience, this option is not often applied in practice. Although more species are tested and 2211 thus information on the differences in sensitivity between species is available, the risk assessment is 2212 most often still based on the most sensitive species using the default assessment factors. The number 2213 of species to be tested according to PPP legislation effectively sets the first tier level of protection in 2214 the effects assessment. Consequently, when more data are available and the risk assessment is still 2215 based on the lowest value without adjusting the assessment factor, the average level of protection may 2216 exceed the level implied by the provisions of the Regulation for the authorisation of Plant Protection 2217 Products. 2218

6.2.2. Approaches considered by EFSA 2219

In 2005, the EFSA PPR Panel published an opinion on the approaches to deal with additional toxicity 2220 data, taking into account that the same average level of protection should be maintained. 2221

They concluded the level of protection by applying the current approach is not the same for each 2222 group of organisms and depends on the size of the standard deviation. The level of protection is also 2223 not the same for each PPP (see also Luttik et al., 2011). The current level of protection is not specified 2224 in the directive. The Panel therefore developed methods that either maintain the current average level 2225 of protection without specification (option 1 and 2) or that can be applied to achieve any specified 2226 level of protection (options 3 to 5) (EFSA, 2005b). 2227

For taxa where the legislation requires only one species this effectively sets the level of protection in 2228 the effect assessment. When additional species are tested the same average level of protection can be 2229 maintained by taking the geometric mean (rather than the lowest value) and dividing by the current 2230 assessment factor (i.e. option 1 of EFSA, 2005b). 2231

Where the legislation requires at least two species of the same taxonomic group, this implies a higher 2232 level of protection in the effect assessment. In this case, a different procedure (i.e. option 2 of EFSA, 2233 2005b) is required when additional species are tested. The minimum is then replaced by the second or 2234 third lowest toxicity value depending on the sample size available, and divided by the current 2235 assessment factor. 2236

Later research (EFSA, 2008a) showed that this is true for a wide range of distributions that are 2237 symmetric and unimodal (single peak) on a logarithmic scale, and also for asymmetric unimodal 2238 distributions where the long tail is to the left. It is also true for asymmetric distributions with long tails 2239 to the right15 and for some examples of bimodal distributions, provided that the standard assessment 2240 factor includes sufficient allowance for between-species variation in toxicity, which seems likely. 2241

Since the revised data requirements request data for only one fish species, i.e. rainbow trout 2242 (Oncorhynchus mykiss), it is proposed to always follow option 1 (the geometric mean approach). The 2243 geomean approach is also recommended in the case where additional testing has been carried out with 2244 crustaceans for insecticides. The PPR Panel suggests to use the geomean method only if more data are 2245 available than requested in the data requirements for the first tier. 2246

The work described above (EFSA 2005b and 2008) is mainly based on distributions of acute toxicity 2247 data. It remains to be investigated whether the same procedure can be used for chronic toxicity data as 2248 well. NOECs may be over/underestimates (e.g. due to wide dose spacing and limited power to detect 2249 effects often caused by small sample size). The PPR Panel recommended, however, using the 2250 geometric mean for both acute and reproductive toxicity, when multiple species have been tested 2251 15Distributions of acute toxicity data often have long tails to the right on the natural scale, but this is reduced or removed on

the logarithmic scale, which is used for the geometric mean.



within a taxonomic group (EFSA, 2005b). The first tier AF of 10 or 100 should be applied to this 2252 geometric mean value of available toxicity data to derive a RAC. 2253

It should be noted that “taxonomic group” can be interpreted in different ways. For instance, 2254 crustaceans and insects represent different taxonomic groups on the phylum level but are sometimes 2255 grouped into the taxonomic group of arthropods. Other examples are merging amphibians and fish into 2256 a taxonomic group of vertebrates, and merging diatoms, green and blue-green algae into a taxonomic 2257 group of algae or algae and macrophytes into the taxonomic group of primary producers. The default 2258 approach should be to treat these taxa as different groups unless scientific arguments (e.g. read-across 2259 to data-rich compounds with the same mode of action) can be raised to consider them as one group. 2260

Note: there is a possibility that the outcome of the geometric mean approach could be biased 2261 (manipulated) by introducing insensitive species. In case of differences in sensitivity of 1 or 2 orders 2262 of magnitude (factor 10-100) an assessment of this possibility has to be made. If the most sensitive 2263 species is more than a factor of 10 (for plants and chronic tests) or 100 (for acute invertebrate and fish 2264 test) below the geometric mean of all the tested species, a weight of evidence approach should be 2265 applied. Until now, little experience exists in applying the geomean approach in the aquatic RA. It is 2266 an important research topic to calibrate this approach with other RA approaches in the RA scheme. 2267

6.2.3. Derivation acute and chronic RAC 2268

In some cases additional ecotoxicity data may be available, but their number is too low to apply the 2269 SSD approach (see Section 6.3). For this situation, it is proposed to use the geometric mean of the 2270 available toxicity values within a taxonomic group (Option 1 described above; Table 6.1). 2271

It should be noted that the geometric mean approach can only aim at an average level of protection and 2272 cannot address possible substance-specific deviations from average patterns. It is therefore necessary 2273 to spend additional considerations on the whole available information on substance toxicity, in order 2274 to confirm the applicability of the risk prediction obtained with a geometric mean L(E)C50 or 2275 geometric mean EC10/NOEC. To this end, all available data on toxicity should be considered, 2276 including open literature data. Additionally, consideration of dose-response relationships should be 2277 included, e.g. by comparing the RAC of the geometric mean with NOEC from the toxicity studies 2278 (could be from additional toxicity data). If the lowest NOEC is higher than the RAC value, it is 2279 acceptable to use the geometric mean. Otherwise, the lowest toxicity endpoint should be used or more 2280 toxicity data may be generated. 2281

Table 6.1: Proposal for the derivation of RACs for aquatic organisms when a limited number of 2282 additional single species toxicity tests is available. When applying this approach scientific 2283 arguments should be given why the selected toxicity data (on which the geomean is based) 2284 concern the same taxonomic group relevant for the risk assessment. 2285

Taxonomic group Number of toxicity data for different taxa of the relevant taxonomic group

RAC Field exposure concentration (PEC)

Acute risk assessment Aquatic vertebrates(a) < 5 acute LC50’s Geomean LC50/100 (e) PECsw;max Invertebrates (b) < 8 acute EC50’s Geomean EC50/100 (e) PECsw;max Chronic risk assessment Aquatic vertebrates(a) < 5 chronic EC10’s

(or chronic NOECs)Geomean EC10/10 (e) PECsw;max or

PECsw;twa Invertebrates (b) < 8 chronic EC10’s

(or chronic NOECs)Geomean EC10/10 (e) PECsw;max or

PECsw;twa Primary producers(c) < 8 EC50’s (d) Geomean EC50/10 (e) PECsw;max (a) i.e. fish or amphibians 2286 (b) i.e. crustaceans or insects for insecticides, and a more specific taxonomic group for fungicides 2287 (c) i.e. green algae, diatoms, blue-green algae or macrophytes for herbicides or fungicides with a herbicidal mode of action. 2288



(d) In other EU directives/regulations (e.g. WFD) the EC50 and an AF of 100 is used in the acute effect assessment and the 2289 EC10 (or NOEC) and an AF of 10 in the chronic effect assessment. 2290

(e) Of the different taxonomic groups the lowest Geomean value is selected (e.g. the lowest value for insects or crustaceans 2291 in case of insecticides; the lowest value for green algae, diatoms, blue-green algae or macrophytes in case of herbicides) 2292

2293

A benefit of this approach is that all species groups are treated in the same manner. Furthermore, the 2294 method will not have to change if, in the future, more than one standard test species is required for a 2295 particular group of species The first tier AF of 10 or 100 should be applied to the geometric mean 2296 value of available toxicity data to derive a RAC. It should be noted, however, that for neonicotinoids 2297 and insect growth regulators (IGRs) (and for other PPPs with very specific mode of action) the regular 2298 Tier 1 and the geometric approach should be used with caution because it is known that species 2299 sensitivity can vary widely for these types of active substances. For example for neonicotinoids it is 2300 reported that Daphnia magna is more than two orders of magnitude less sensitive than 2301 macrocrustaceans and insects (Beketov and Liess 2008). 2302

2303

6.3. The Species Sensitivity Distribution (SSD) approach 2304

6.3.1. Introduction to SSD approach 2305

2306 Species may vary markedly in their sensitivity to PPPs. This variation in direct toxicity can be 2307 described by constructing a Species Sensitivity Distribution (SSD). The SSD is a statistical 2308 distribution estimated from a sample of laboratory toxicity data and visualised as a cumulative 2309 distribution function (see Figure 6.1). Logistic and lognormal distributions are most often used, 2310 because they require less data than distribution-free methods and are relatively easy to fit with 2311 standard statistical software (Aldenberg & Jaworska, 2000; Aldenberg et al. 2002; Van Vlaardingen et 2312 al. 2004). It is assumed that a SSD follows a normal distribution. Therefore, techniques like bootstrap 2313 should be avoided, since they do not meet the assumption of normality. Note that in SSDs all species 2314 have equal weight and thus are considered of equal importance in assessing the ecotoxicological risks. 2315

SSDs are used to calculate the concentration at which a specified proportion of species are expected to 2316 suffer direct toxic effects. These concentrations, the hazardous concentrations, are expressed as HCx 2317 values and represent the value that affects a specific proportion (x %) of species. For regulatory 2318 purposes usually the HC5 is used, the hazardous concentration to 5% of the species tested. When 2319 compared with the first tier effects assessment on the basis of standard test species, SSDs have the 2320 advantage of making more use of the available laboratory toxicity data for a larger array of species. 2321 They describe the range of sensitivity rather than focusing on a single value, they enable estimates to 2322 be made of the proportion of the species affected at different concentrations, and they can be shown 2323 together with confidence limits showing the sampling uncertainty due to the limited number of species 2324 tested. They can be used in a deterministic risk assessment by taking an appropriate percentile from 2325 the SSD, or in a probabilistic risk assessment by using the whole SSD. For a detailed description of the 2326 underlying assumptions of the SSD approach refer to Posthuma et al. (2002), Forbes & Calow (2002) 2327 and Brock et al. (2011). Note that the median HC5 value is the concentration that with 50% certainty 2328 is lower than the toxicity values (e.g. EC50s or NOECs) for 95% of the species tested, while the lower 2329 limit HC5 provides this concentration with 95% certainty. 2330

2331



2332

Figure 6.1: Graphical presentation of the Species Sensitivity Distribution curve, its 90% confidence 2333 interval, and as dotted arrows the derivation of the lower limit, median and upper limit Hazardous 2334 Concentration to 5% of the species (HC5). Figure from EFSA (2006). 2335

2336

For construction of SSDs and the calculation of HC5 values, programs like ETX2.0 (Van Vlaardingen 2337 et al., 2004) or ToxRat (http://www.toxrat.de/) may be used. These programs also contain several 2338 statistical tools to test the assumptions of normality. It should be noted, however, that the performance 2339 of these tests strongly depends on the number of data. With a relatively low number of data, a 2340 distribution is often accepted as normal, whereas for large datasets deviations from normality will be 2341 more easily detected. The outcome of the tests as such should therefore not be used as a single 2342 criterion to decide whether or not the SSD approach can be applied, or to split datasets to construct 2343 specific SSDs for particular taxonomic groups (see 6.3.3). A thorough evaluation of the individual 2344 data points and visual inspection of the fit may reveal whether or not violation of the assumptions 2345 concerning the distribution is acceptable. For example, violation of the goodness-of-fit test may be 2346 acceptable from a regulatory point of view when the fitted distribution in the tail of the SSD is 2347 relatively worst-case compared to the data points (in the sense that most of the toxicity data around the 2348 HC5 and lower are on the right side of the fitted curve). 2349

6.3.2. Criteria for the selection of toxicity data to construct SSDs 2350

The number of species data used to fit the distribution has to be adequate from a statistical point of 2351 view. Suter et al. (2002) concluded that SSDs could be adequate with data points between 3 and 30, 2352 dependent on the method used. Note, however, that the 90% confidence interval of the SSD generally 2353 will be wider, and consequently the lower limit HC5 lower, with a lower number of toxicity data 2354 included in the SSD. The former aquatic guidance document (EC, 2002) refers to the HARAP 2355 document (Campbell et al., 1999) for the aquatic effect assessment on basis of the SSD approach. The 2356 HARAP document recommends that separate SSDs should be constructed with acute and chronic 2357 toxicity data. In addition, this document also recommends that SSDs to be based upon a minimum of 2358



either eight acute or eight chronic toxicity data for different taxa that are representative for the 2359 sensitive taxonomic group(s), at least if the SSD is not exclusively constructed with toxicity data for 2360 fish. An SSD that addresses the sensitivity of fish should be based on a minimum of 5 toxicity data for 2361 different fish species (Campbell et al., 1999). On basis of published SSD information for PPPs and 2362 aquatic organisms (e.g. Maltby et al. 2005 & 2009; Van den Brink et al. 2006; Giddings et al. 2013) 2363 these HARAP recommendations seem to be pragmatic and appropriate from a plant protection product 2364 regulatory point of view. The PPR Panel recommends to follow these criteria. 2365

Evaluation of scientific literature (Maltby et al, 2005; Brock et al 2008) indicated that toxicity data 2366 from different geographical areas can be combined as long as the SSD is based on the most sensitive 2367 taxonomic group and that the taxonomic profile is appropriate. It is noted however, that toxicity 2368 studies performed in different geographic regions may be conducted under different test conditions, 2369 which may affect exposure profile. The potential effects of test conditions on exposure should be 2370 considered whenever data are collated across different studies, irrespective of the geographical region 2371 in which the data were generated. 2372

The endpoints measured in the toxicity tests on which the SSD is based must be the most sensitive 2373 endpoints that are toxicologically and ecologically relevant. Acute toxicity data mostly address 2374 mortality and immobility as the most frequently studied endpoints for animals, while that is biomass 2375 and growth for primary producers. Chronic toxicity data mostly address reproduction, feeding rate and 2376 growth as the most frequently studied endpoints in animals, and again this is biomass and growth for 2377 primary producers. The use of biochemical/physiological endpoints or biomarkers in SSDs is not 2378 recommended for regulatory purposes due to difficulties in correlating results with tangible ecological 2379 effects (e.g. the protection of populations). The test duration might be a criterion to be applied for the 2380 selection of the toxicity data. Test duration, however, is taxon and guideline dependent and, as a 2381 consequence, a range of test durations for different organisms is often included in the same SSD. 2382

According to Brock et al. (2011) measurement parameters, from which endpoints are calculated, 2383 should preferably be sensitive / responsive in the range of tested concentrations such that SSDs avoid 2384 the use of greater or lower than values. In general, it is not recommended to include unbound values 2385 (greater than- or lower than-values) in the SSD. There are situations, however, where ignoring those 2386 data would lead to a loss of valuable information. When a <-value is lower than the lowest toxicity 2387 endpoint, this means that the other data do not cover the whole range of sensitivities. Leaving out this 2388 information might lead to an HC5 that is under-protective. However, to demonstrate the effect of 2389 including the information in the SSD, the following procedure can be applied: 2390

2391 - If in a set of available toxicity values for a certain species, a greater or lower than value is present: 2392

o this value should not be included in the calculation of the geometric mean value for 2393 that species if the value is inside the range of values, 2394

o this value should be used as such (without the < or > sign) if the value was outside 2395 the range. 2396

- If in a set of available toxicity values for a compound for a particular species, only a greater or 2397 lower than value is present: 2398

o this value should only be used as such (without the < or > sign) in the SSD where the 2399 value is outside the range of all other values (for other taxa). 2400

If an SSD is used in which unbound values are included, this should always clearly justified. 2401

2402

In the risk assessment for PPPs, Species Sensitivity Distributions based on chronic data are scarce. 2403 Acute toxicity data are normally more available than chronic data due to experimental and financial 2404 constraints. Chronic NOEC values and chronic EC10 effect concentrations may be included in a 2405 chronic SSD. Whereas acute toxicity data relate to a limited number of responses and time scales, 2406 chronic toxicity data may include a wide range of responses and test durations, thereby introducing 2407 additional variability into the SSD. The test duration has to be of a chronic duration compared to the 2408



life cycle characteristics of the species group. More specifically, a chronic toxicity test is defined as a 2409 study in which (1) the species is exposed to the PPP for at least one full life-cycle, or (2) the species is 2410 exposed to the PPP during one or more critical and sensitive life-stages (see e.g. Holland, 1996; Brock 2411 et al., 2010c). Consequently what is considered chronic or acute is very much dependent on the 2412 species and endpoint considered. 2413 2414

6.3.3. Selecting toxicity data on basis of toxic mode-of-action of the substance 2415

According to Campbell et al. (1999) and the former aquatic guidance document (EC, 2002) the toxic 2416 mode-of-action of a PPP should be taken into account when constructing SSDs to derive acceptable 2417 concentrations. SSDs used in risk assessment should always be constructed using toxicity data for the 2418 most sensitive taxonomic group. In case of herbicides, vascular plants and/or algae usually comprise 2419 the most sensitive groups. For insecticides, arthropods (crustaceans and insects) usually are most 2420 sensitive. For fungicides, often a range of taxonomic groups are among the sensitive organisms. 2421

The following information can be used to decide which taxonomic groups have to be included in an 2422 SSD for the compound under consideration. If the first tier indicates that one standard test species of 2423 the basic set is considerably more sensitive (differing by a factor > 10) than the others, in first instance 2424 additional toxicity should be gathered that are representative for the sensitive taxonomic group to 2425 which this species belongs. Furthermore, data gathered by read-across on related compounds with 2426 identical or similar toxic mode-of-action may give useful information on the taxonomic groups which 2427 are most likely sensitive for the compound under consideration. Also data in the open literature on the 2428 compound may give information on the sensitive taxonomic groups. In addition, if available, results of 2429 micro-/mesocosm tests with the compound under evaluation may shed light on the sensitive taxonomic 2430 groups, also when these tests studied the effects of relatively high concentrations not suitable to derive 2431 a threshold level of effects. 2432

The next paragraphs give an overview of the sensitive organisms for insecticides, herbicides and 2433 fungicides (adapted after Brock et al. 2011). 2434

6.3.3.1. Insecticide SSDs 2435

Evaluation of the toxicity data of 16 insecticides (including 8 acetyl cholinesterase inhibitors, 5 2436 pyrethroids, 2 organochlor compounds and 1 insect growth regulator) indicates that (1) arthropods are 2437 the preferred taxonomic group to construct acute SSDs, and (2) acute toxicity data for freshwater 2438 arthropods from different geographical regions and different freshwater habitats may be combined 2439 within a single SSD (Maltby et al., 2005). If necessary, toxicity data for freshwater and saltwater taxa 2440 also can be combined in an SSD, but it is important to be aware of differences in taxonomic 2441 composition and possible consequences for HC5 values that are calculated. SSDs constructed using 2442 arthropod species recommended in test guidelines did not differ significantly from those constructed 2443 using non-recommended arthropod species (Maltby et al., 2005). 2444

So, in case of insecticides, arthropods (crustaceans and insects) usually are most sensitive. This 2445 implies that the SSD can focus on these taxonomic groups. Note, however, that for some novel types 2446 of insecticides (e.g. neonicotinoids) insects may be more sensitive than certain micro-crustaceans (see 2447 e.g. Beketov & Liess, 2008). If, for example, the first tier toxicity value for Chironomus is an order of 2448 magnitude lower than that of Daphnia, it is recommended to construct, in first instance, an SSD with 2449 toxicity data for insects, or to explore which insects and crustaceans (e.g. macro-crustaceans) can be 2450 combined in a single SSD on basis of all relevant information available. 2451

Furthermore, for certain insect growth regulators (particularly those that affect moulting) the standard 2452 duration (48-96 h) of the acute toxicity test may not suffice, since latency of effects may occur. For 2453 insect growth regulators it may therefore be necessary to use in the acute SSD results of prolonged 2454 acute toxicity test (e.g., by transferring the test animals to clean water after 48-96 h exposure and to 2455 continue the observations for several days until the incipient effect level is reached). 2456



6.3.3.2. Herbicide SSDs 2457

At environmentally realistic exposure concentrations, herbicides specifically, and mainly, affect 2458 primary producers in aquatic ecosystems, i.e. algae and macrophytes. The AMRAP document (Maltby 2459 et al, 2010) and Giddings et al. (2012) provides guidance on the use of macrophyte toxicity data in the 2460 SSD approach and define areas of uncertainty, which are specifically associated with the selection of 2461 species and endpoints. The uncertainty associated with species and endpoint selection to assess 2462 toxicity for algae is generally less because of the availability of standard protocols that are already 2463 used for a fairly long time. 2464

For some types of herbicides algae and macrophyte data may be combined in the same SSD. Van den 2465 Brink et al. (2006) and Giddings et al. (2012) showed that this is generally possible for photosynthesis 2466 inhibitors. However, herbicides that inhibit amino acid synthesis and herbicides with an auxin 2467 simulation mode-of-action, generally seem to be more toxic to aquatic vascular plants than algae, so 2468 that it may be necessary to construct the SSD with macrophyte data (Giddings et al. 2012). Note, 2469 however, that currently the knowledge with respect to mode-of-action of several other types of 2470 herbicides is too limited to recommend the combination of algae and macrophytes, or not. If in the 2471 first tier dataset the most sensitive macrophyte is an order of magnitude lower than that of algae, a 2472 pragmatic approach is to construct in first instance an SSD with toxicity data for macrophytes only, or 2473 to explore which aquatic vascular plant and algae can be combined in a single SSD on basis of all 2474 relevant information available. 2475

For the construction of macrophyte SSDs the AMRAP document (Maltby et al., 2010) recommends 2476 that, in first, instance a range of morphologically and taxonomically different macrophytes should be 2477 included, unless the mode-of-action of the herbicide primarily affects a specific group of macrophyte 2478 species (e.g. mosses, monocotyledonous or dicotyledonous vascular plants; floating or rooted 2479 macrophytes). Ideally, SSDs should be based on toxicity values for comparable measurement 2480 endpoints generated from tests conducted under similar exposure scenarios and exposure durations, 2481 preferably using standardised protocols. The PPR Panel agrees with these recommendations and 2482 proposes to adopt them in this guidance document. 2483

A more or less similar approach as described above for aquatic macrophytes can be followed for algae. 2484 Ideally, when algae are at risk, the SSD should be constructed with a range of taxonomically different 2485 groups if the two tested algae do not differ more than a factor of 10 (e.g. including green algae, 2486 diatoms, blue-greens etc, and/or different genera representative for these groups). 2487

It appears from the published literature that for aquatic macrophytes a wide array of measurement 2488 endpoints is used. A conservative approach would be to use the lowest endpoint per taxon, no matter 2489 what measurement parameter it is based on. Nevertheless, this wide array of available measurement 2490 endpoints may contribute to the variability in SSDs. The AMRAP document (Maltby et al., 2010) 2491 recommends the use of growth rate endpoints for macrophytes. These growth rate endpoints should be 2492 preferably based on biomass or shoot length, as they potentially provide consistency across time and 2493 species. From a statistical viewpoint, it is preferable that all endpoints used in development of a SSD 2494 are based on common measurement parameters, since each parameter may have a different 2495 distribution. Bergtold & Dohmen (2011) present reasons why toxicity data based on specific growth 2496 rate are more informative and better suited to effect characterisation than toxicity values based on 2497 standing crop/biomass or standing crop/biomass increase (yield) for both algae and macrophytes. Also 2498 according to OECD guidelines for algae (e.g. OECD 201) growth rate data points are preferred. Note 2499 however, that for mathematical reasons an EC50 calculated for growth rate is usually greater than an 2500 EC50 calculated for biomass or biomass increase (yield). The PPR Panel recommends to preferably use 2501 growth rate endpoints when both growth rate and biomass endpoints are available. 2502

2503



6.3.3.3. Fungicide SSDs 2504

Maltby et al. (2009) studied differences in sensitivity between primary producers, invertebrates and 2505 fish to fungicides with different toxic mode-of-action. Although for some fungicides a specific 2506 taxonomic group was most sensitive, the majority of fungicides investigated were classified as general 2507 biocides. For those fungicides that are general biocides, data from all aquatic taxonomic groups are 2508 recommended to be used to construct SSDs (Maltby et al., 2009). The HARAP document (Campbell et 2509 al., 1999) does not specify the taxonomic groups and level of taxonomic resolution when selecting 2510 toxicity data for these generic SSDs. For those fungicides that are general biocides, it is recommended 2511 as default approach to include toxicity data from eight different taxonomic groups including at least 2512 six different orders/families in the SSD. These data include three to five toxicity data already 2513 generated in the first tier (including fish) and five to three additional toxicity data. For those fungicides 2514 for which a certain taxonomic group is clearly more sensitive, it is recommended to construct, in first 2515 instance, an SSD with toxicity data for this taxonomic group, if toxicity value for this most sensitive 2516 test species is at least an order of magnitude lower than that for the other Tier 1 test species. In 2517 addition, when more toxicity data are available, it is advised to always explore which taxonomic 2518 groups can be combined in a single SSD on basis of all relevant information available. In this 2519 procedure SSD curves are generated where a minimum of 8 data points (i.e., taxa) are available. 2520

Initially, for fungicides with a less specific toxic mode-of-action towards aquatic organisms, all 2521 available aquatic toxicity data for a compound are used to generate an SSD and the fit to a log-normal 2522 distribution is assessed using the Anderson-Darling goodness-of-fit test. If the distribution does not 2523 pass the goodness-of-fit test at p = 0.05, separate distributions need to be constructed for vertebrates 2524 and non-vertebrates and the most sensitive distribution is used. If these distributions are not described 2525 by a log-normal model, then the dataset is partitioned further. For example non-vertebrates may be 2526 partitioned in invertebrates and primary produces. Furthermore, invertebrates may be further 2527 partitioned in arthropods and non-arthropods, while primary producers may be further partitioned in 2528 algae and macrophytes (see Maltby et al, 2009). Note that the final SSD curve on basis of the most 2529 sensitive taxonomic group selected should be constructed with a minimum of 8 data points (i.e., taxa) 2530 and that separate SSDs should be constructed with acute or chronic toxicity data. 2531

It should be noted that fungi/microorganisms are not included in the standard dossier dataset as a 2532 specific taxon of interest. As a consequence, data on a potentially sensitive species group may be 2533 missing. Recent research indicates that aquatic fungi may be sensitive for ergosterol inhibiting 2534 fungicides in particular, while for several other types of fungicide the HC5 based on invertebrates, 2535 primary producers and/or fish may be protective for the aquatic fungi tested (Dijksterhuis et al., 2011). 2536 The results indicate that further research into the potential effects on fungi is needed. It should be 2537 noted that the kingdom of fungi is diverse. The selection of relevant species for which standardised 2538 ecotoxicity tests may be developed is therefore identified as a research need. Within this context it is 2539 worthwhile mentioning that several micro-/mesocosm studies with the ergosterol inhibiting fungicide 2540 tebuconazole confirm that aquatic hyphomycetes are amongst the most sensitive endpoints 2541 (Bundschuh et al. 2011; Kosol 2011). Micro-/mesocosm studies with other fungicides that paid 2542 attention to responses of microorganisms at environmentally realistic exposure concentrations, 2543 however, are scarce. A recent study with the dithiocarbamate metiram, revealed that aquatic fungi and 2544 bacteria probably are less sensitive than several populations of aquatic invertebrates and algae (Lin et 2545 al. 2012). 2546

Note that if fish are included in the SSD for general biocides (non-specific fungicides), the aim is to 2547 derive a concentration that is protective at the population/community level. Since for fish a more 2548 stringent protection goal is adopted (see chapter 2), it should always be checked whether the outcome 2549 meets the regulatory lower or higher-tier trigger for fish. If the SSD approach for non-specific 2550 fungicides results in a higher RAC than for fish (for example the Tier 1 RAC for fish or a specific SSD 2551 for fish), the lower RAC value for fish needs to be selected as the final RAC. 2552

2553



6.3.4. Derivation of acute and chronic RACs for invertebrates and primary producers 2554

Compared to the effects considered in microcosm and mesocosms studies, the SSD approach does 2555 neither consider recovery nor indirect effects. However, SSDs might be very useful in risk assessment 2556 as they represent a cost-effective approach for the use of all available laboratory toxicity data for a 2557 larger array of species. From this point of view, hazardous concentrations derived from species 2558 sensitivity distributions for insecticides, herbicides and fungicides were calibrated/validated with 2559 Effect class 1-2 data from micro-/mesocosm studies (Maltby et al., 2005; Brock et al., 2006; Van den 2560 Brink et al., 2006; Maltby et al., 2009). Note, however, that these studies did not consider all PPP 2561 modes of action. 2562

In the insecticide SSD evaluation study by Maltby et al. (2005) the majority of compounds comprised 2563 acetyl cholinesterase inhibitors (organophosphates and carbamates) and pyrethroids, while of the more 2564 novel chemistries only one insect growth regulator (diflubenzuron) and no neonicotinoids or 2565 biopesticides could be considered because at the time of the evaluation no appropriate micro-2566 /mesocosm studies with these compounds were available. 2567

Similarly, in the SSD herbicide evaluation study (Van den Brink et al. 2006) 7 of the 9 compounds 2568 evaluated were photosynthesis-inhibitors (six photosystem (PS) II inhibitors (e.g. atrazine and 2569 metribuzin) and PS I inhibitor (diquat)), 1 compound simulated the growth hormone auxin (2,4-D) and 2570 1 compound was a cell division/cell elongation inhibitor (pendimethalin). At the time of the evaluation 2571 appropriate micro-/mesocosm studies with other types of herbicides were not yet available in the open 2572 literature. 2573

For fungicides a larger variety of modes-of-action could be evaluated (Maltby et al. 2009), but, as 2574 discussed above, hardly no attention was paid to effects on populations of micro-organisms in 2575 fungicide-treated micro-/mesocosm tests. 2576

For the insecticides evaluated (mainly acetyl cholinesterase inhibitors and pyrethroids) the lower limit 2577 HC5 of acute SSDs (constructed with acute L(E)C50’s) was protective for single and repeated pulse 2578 exposures in micro/mesocosm, at least when the effects in these test systems are expressed in terms of 2579 nominal or measured peak concentrations (Maltby et al., 2005). For herbicides evaluated (mainly 2580 photosynthesis inhibitors) the lower limit of the acute HC5 and the median value of the chronic HC5 2581 (based on chronic NOEC/EC10 values) were protective of adverse effects in aquatic microcosms and 2582 mesocosms, even under a long-term exposure regime (Van den Brink et al., 2006). For fungicides, the 2583 derived lower limit HC5 values and the HC1 values were protective of adverse effects in microcosm 2584 and mesocosms studies when effects in these test systems are expressed in terms of nominal or 2585 measured peak concentration (even under more or less long-term exposure regimes) (Maltby et al., 2586 2009). 2587

Maltby et al. (2009), who studied the relationship between HCx concentrations of fungicides and 2588 corresponding threshold concentrations of effects observed in micro-/mesocosms, also reanalysed the 2589 relationships between SSDs constructed with acute toxicity data and threshold concentration derived 2590 from microcosm and mesocosm experiments for insecticides (as published by Maltby et al., 2005) and 2591 herbicides (as published by Van den Brink et al., 2006). It was demonstrated that in general the 2592 median HC5 is protective of short-term exposures, the median HC5 divided by 1.5 is protective of 2593 medium-term exposure regimes and the median HC5 divided by 3 or the HC1 is protective of repeated 2594 longer-term exposure. 2595

Table 6.2 presents a proposal for the derivation of a RAC for edge-of-field surface waters, based on 2596 hazardous concentrations derived from species sensitivity distributions with aquatic invertebrates and 2597 plants for at least 8 different taxa belonging to the relevant sensitive taxonomic group (after Brock et 2598 al. 2011). 2599

The PPR Panel recommends to calculate the SSD-RAC both on basis of the median HC5 (and the 2600 application of an AF of 3) and the lower limit HC5 (AF of 1) and to always select the lower value as 2601



default, unless scientific arguments are provided that the higher SSD-RAC estimate is more plausible. 2602 For chronic RA the same AFs are proposed as pragmatic solution; different AFs might be 2603 recommended if more information becomes available. 2604

2605

Table 6.2: Proposal for the derivation of a RAC in edge-of-field surface waters, based on hazardous 2606 concentrations derived from Species Sensitivity Distributions with aquatic invertebrates 2607 and/or plants. 2608

Type of effect/risk assessment

Relevant PEC Hazardous concentration

AF to derive RAC from Hazardous concentration

Acute and chronic effect/risk assessment for single and repeated pulse exposure (FOCUS step 3 or 4)

PECsw;max Latency of effects not expected(a) Median acute HC5 (based on acute LC50 or EC50 data)(b)

and/or Lower limit acute HC5 (based on acute LC50 or EC50 data) (b) Latency of effects expected Median acute HC5 (based on acute LC50 or EC50 data from prolonged acute toxicity tests (c))

and/or Lower limit acute HC5 (LLHC5) (based on acute LC50 or EC50 data from prolonged acute toxicity tests)

Or

precautionary approach instead of the 2 options

above Apply chronic SSD (see below)

3

1

3

1

Chronic effect/risk assessment with long-term exposure

PECsw;max or PECsw;twa

Median chronic HC5 (based on chronic NOEC and/or EC10 data)

and/or Lower limit chronic HC5 (LLHC5) (based on chronic NOEC and/or EC10 data)

3

1

(a) This has to be demonstrated by the applicant, see further 2.5.1. For example, by read-across for substances with similar 2609 toxic mode of action, prolonged acute toxicity tests, and information from micro/mesocosm studies for similar 2610 compounds with a longer-term observation period after exposure. 2611

(b) For types of PPPs evaluated by Maltby et al. (2005; 2009) and Van den Brink et al. (2006). 2612 (c) In prolonged acute toxicity tests the observation of treatment-related responses is continued after the test organisms are 2613

transferred to clean medium. 2614 2615



If for different taxonomic groups different and valid distributions are available, the most sensitive SSD 2616 is used in the risk assessment. The results of higher tier risk assessments based on a specific SSD have 2617 to be compared again with the results of the first tier to ensure that the RAC based on the specific SSD 2618 is protective for taxa not considered in this SSD. 2619

6.3.5. Derivation of acute and chronic RACs for fish/amphibians 2620

The HARAP document (Campbell et al., 1999) recommends using a minimum of 5 toxicity data to 2621 construct SSDs specific for fish. This lower number of toxicity data is chosen for, among other 2622 reasons, animal welfare considerations and because of the overall lower variability in toxicity data 2623 when e.g. compared with that of invertebrates. In the risk assessment it is sometimes necessary to 2624 consider fish separately and to construct a separate SSD with fish as the most appropriate method to 2625 meet this requirement. For example, constructing a separate SSD for fish may be necessary if the risks 2626 of a PPP to populations of invertebrates and primary producers have been assessed by means of an 2627 appropriate microcosm or mesocosm experiment without fish. In regular mesocosm and microcosm 2628 studies fish are recommended not to be included as the effects of fish might interfere with the effects 2629 of the compound on the macro-invertebrate community (Giddings et al., 2002). If potential risks to 2630 fish cannot be excluded, one appropriate method in risk assessment is to construct a separate SSD for 2631 fish (other options are described in section 7.2). 2632

Currently toxicity values for amphibians are not mentioned as a basic dossier requirement for the 2633 ecotoxicological effect assessment. In addition, hardly any information is available that allows a 2634 systematic comparison of the species sensitivity distributions between fish and amphibians. 2635 Consequently, it will depend on expert judgement whether on basis of the available toxicity data for 2636 fish and amphibians a single or separate SSD has to be constructed for these taxa. When the RAC is 2637 based on a separate SSD for fish, then also a separate RAC for amphibians has to be generated to 2638 make sure they are covered in the refined RA. The separate RAC for the amphibian species may be the 2639 single species assessment approach. Whether the SSD for fish is also representative for amphibians is 2640 a topic for future research. For other options for refined vertebrate RA see section 7.2. 2641

Acute LC10 and acute NOEC values may be used to construct the SSD and to calculate the HC5 and 2642 lower limit of the confidence interval of the HC5 (LLHC5) for fish (and/or amphibians), since a higher 2643 protection level is desired for vertebrates than for invertebrates and plants. Another option is to apply 2644 an extra AF to the HC5 based on acute LC50 or EC50 data. 2645

It is recommended that the following hazardous concentrations and assessment factors are used to 2646 derive a RAC for fish and other aquatic vertebrates (Table 6.3). The rationale behind the suggested 2647 assessment factors is an extrapolation from the assessment factors used for invertebrates which has 2648 been calibrated with micro/mesocosm experiments. However, for fish a more stringent protection level 2649 has been adopted for the acute RA (avoiding visible mortality of individuals) and for that reason an 2650 AF of 3 should be applied on the median HC5 from an SSD constructed with acute NOEC/EC10 2651 values for fish. In order to also derive a SSD RAC for vertebrates based on acute LC50 values (since 2652 these data are usually reported in the dossiers) the panel assumes an overall difference of 3 between 2653 acute LC50 and acute LC10/NOEC values for fish resulting in an assessment factor of 9. For the ratio 2654 between the acute LC50 and chronic NOEC/L(E)C10, usually a factor of 10 is assumed (see e.g. Roex et 2655 al., 2000). Taking this into account, assuming a factor of 3 for the ratio between the acute LC50 and 2656 acute NOEC/LC10 for fish seems to be appropriate. Furthermore, traditionally an assessment factor of 2657 10 has been attributed to the variation in sensitivity between species (for the acute risk assessment) 2658 and hence an assessment factor of 9 harmonises to this assumption. Nevertheless it is acknowledged 2659 that the method proposed needs calibration. 2660

For the chronic risk assessment (focusing on protection of fish populations) the same reasoning and 2661 AF is proposed as for invertebrates. The PPR Panel recommends to calculate the SSD-RAC both on 2662 basis of the median HC5 (and the application of an AF) and the lower limit HC5 and to always select 2663



the lower value as default, unless scientific arguments are provided that the higher SSD-RAC estimate 2664 is more plausible (Table 6.3). 2665

Table 6.3: Proposal for the derivation of a RAC for edge-of-field surface waters, based on hazardous 2666 concentrations derived from Species Sensitivity Distributions with fish (and other aquatic vertebrates). 2667 Information of possible latency of effect may be obtained on basis of knowledge on the specific toxic 2668 mode-of-action and, read across information. To avoid unnecessary testing with aquatic vertebrates for 2669 animal welfare considerations the conduct of prolonged acute toxicity tests to demonstrate latency is 2670 not considered. 2671


Relevant PEC Hazardous concentration

AF to derive RAC from hazardous concentration

Acute effect/risk assessment

PECsw;max Latency of effects not expected(a) Median acute HC5

(based on 96 h NOEC and/or acute LC10 data)

and/or Lower limit acute HC5 (based on 96 h NOEC and/or acute LC10 data) or Median acute HC5

(based on 96h LC50 or EC50 data)

and/or Lower limit acute HC5 (based on 96 h LC50 or EC50 data) or If latency of effects is expected go to chronic effect assessment (see below)

3

1

9

3

Chronic effect/risk assessment


Median chronic HC5

(based on chronic NOEC and/or EC10 data)

or Lower limit chronic HC5 (based on chronic NOEC and/or EC10 data)

3

1

2672 (a) This has to be demonstrated by the applicant, see further 2.5.1. For example, by read-across for substances with 2673

similar toxic mode of action, prolonged acute toxicity tests, and information from micro/mesocosm studies for similar 2674 compounds with a longer-term observation period after exposure. 2675

2676 2677

2678 2679

2680 2681

2682

2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695

2696

2697

2698 2699 2700 2701

EFSA Jou

7.

7.1.

7.1.1. In edge-than thechemicaentry rouwater flexposurepond scmodellinChapter exposurebetweenvalue is in the D5

Figure 7scenariomodellin

urnal 2013;vo

Higher-tieexperimen

Selecting thigher-tier

Introducti

of-field surfe exception. al properties outes (e.g. drow, water de profiles forenario (D5)

ng for use o4). In this

es (predomin pulses diffehighest for t5 pond, whil

7.1: Predictes (D1 and D

ng

olume(issue):N

r effect assental ecosyste

the approprir effect stud

on

face waters, tThe shapes

of the PPP, trift, surface rdepth, pH, lr a hypotheti

are presentof the PPP i

example thnantly due toer between scthe D1 ditchle the pulse d

ed exposure D5) and on

NNNN

essment by mms

iate exposuries

time-variables of the timthe applicatiorun-off, drailight penetraical PPP in twted in Figurin spring cehe predicted o drift), but cenarios. In t, somewhat l

durations incr

profiles forne ditch (D1

DRAF

means of ref

re regimes w

e exposure cme-variable on regime ininage) and pation, biomawo stream scre 7.1. Thesereals (for F

exposure pthe peak con

the exposure lower in the rease when g

r an exampl) and pond

T GD on tiere

fined-exposu

when addres

oncentrationexposures d

n the crop, thproperties of ass of plantscenarios (D1se predictionOCUS scen

profiles are ncentrationsprofiles presstreams (D1

going from s

e PPP used (D5) scena

ed RA for edg

ure laborato

ssing time-v

ns of PPPs ardepend on fhe relative imf the receivins). As an ex1 and D5) anns are basedarios and mcharacterise, pulse duratsented in Fig1 and D5) antream, to dit

in spring cario on basis

ge-of-field surf

ory toxicity t

ariable expo

re more oftenfactors like mportance of ng water bodxample the pnd one ditch d on FOCUSmodel approaed by repeattions and/or gure 7.1 the Pnd considerabtch to pond.

ereals in tws of FOCUS

face water

70

tests and

osures in

n the rule physical-

f different dies (e.g. predicted (D1) and S Step 3 aches see ted pulse intervals

PECsw,max bly lower

wo stream S Step 3



Higher-tier effect assessments do not necessarily need to be performed by simulating constant 2702 exposures normally used in standardised lower tier tests, but may address the exposure regimes 2703 predicted for edge-of-field surface waters. For a straightforward risk and effect assessment, however, 2704 the exposure regime of the PPP in the ecotoxicological test should be realistic to worst-case relative to 2705 the predicted exposure regime in the edge-of-field surface water under consideration, at least if TWA 2706 concentrations cannot be used in the risk assessment (see section 2.5). 2707

If the TWA concentration approach cannot be used in the chronic risk assessment, section 7.2 2708 describes possibilities for refined exposure single species studies and section 7.3 for micro/mesocosm 2709 studies that aim to simulate realistic to worst-case time-variable exposure concentrations, in terms of 2710 height, duration, spacing and frequency of pulse exposures. The refined exposure regime tested should 2711 be guided by relevant exposure predictions for the intended agricultural uses (e.g. as deduced from 2712 FOCUS surface water scenarios or from national exposure scenarios). In the sections below guidance 2713 is given how to select the appropriate exposure regimes in higher-tier effect studies. 2714

7.1.2. Use of predicted exposure profiles for edge-of-field surface waters in higher-tier effect 2715 assessments 2716

Before starting a higher-tier effect assessment on basis of time-variable exposures, the predicted 2717 exposure profile for the PPP of concern in the relevant stream/ditch/pond scenario needs to be 2718 compared with the Tier 1 RACs (based on standard laboratory toxicity data). In Table 7.1 this is done 2719 for the example PPP of which the exposure profiles are presented in Figure 7.1. It appears that 2720 potential acute risks are identified for the D1 and D5 stream scenarios and for the D1 ditch scenario 2721 (PECsw;max > RACsw;ac). For these scenarios the PECsw;max is also larger than the chronic Tier 1 RAC 2722 (RACsw;ch), however, no chronic risks are triggered (PECsw;7d-twa < RACsw;ch) if it is possible to use the 2723 TWA approach. For the pond scenario no acute and chronic risks are triggered since both the acute 2724 and the chronic Tier 1 RAC are higher than the PECsw;max. 2725

Table 7.1: Comparison of PECs (peak and 7d TWA) with the Tier 1 RACs (acute: RACsw;ac; chronic 2726 RACsw;ch) for the hypothetical PPP used in spring cereals and presented in Figure 7.1 2727

Scenario Water body PECsw;max (µg/L)

RACsw;ac (µg/L)

RACsw;ch (µg/L)

PECsw;7d-twa (µg/L)

D1 Stream 0.036 0.020 0.017 0.003 D1 Ditch 0.050 0.020 0.017 0.015 D5 Stream 0.038 0.020 0.017 0.001 D5 Pond 0.002 0.020 0.017 0.002

2728 Since the highest exposure concentrations (both PECsw;max and PECsw;7d-twa) are calculated for the D1 2729 ditch scenario it is logical to evaluate this scenario first when selecting an appropriate exposure regime 2730 for higher-tier effect studies. This can best be done by plotting the Tier 1 RACsw;ac (and/or Tier 1 2731 RACsw;ch) on the predicted D1 ditch exposure profile (Figure 7.2). Note that, when available, the 2732 Geomean-RAC or the SSD-RAC may be used to replace the Tier 1 RAC. In the example (Figure 7.2) 2733 the exposure profile is characterised by a repeated pulse exposure regime (5 pulses) and the peaks of 2734 all pulses exceed for short periods the Tier 1 RACsw;ac. Consequently, to evaluate the potential risks of 2735 these five pulse exposures they should be addressed realistically in higher-tier effect studies, unless the 2736 number of pulses to be studied can be reduced on basis of ecotoxicological information(for details see 2737 below). 2738

It has also to be checked whether the other scenarios have a higher number (with a lower PECsw;max 2739 still above RAC) of peaks or if the duration of the pulses are longer in other scenarios. In the first step 2740 a worst case exposure scenario can be constructed by selecting the maximum number of peaks, the 2741 highest PECsw;max and the longest peak duration among all relevant scenarios. The number of peaks 2742 simulated in the actual test can be lowered based on guidance below. 2743

2744

2745

2746 2747

2748

2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759

2760 2761

2762

2763 2764 2765 2766 2767 2768 2769 2770

2771 2772

2773 2774 2775 2776

EFSA Jou

Figure RACsw;ac

7.1.3. For an exposureindependspan of exposurescenarioif the peindividuwhen indthe first when:

(i) the ilevels, an

(ii) comp

To demopulsed eorganismexposureregime cfor examin Figurelast three

7.1.4.

When adtoxicologfewer puPPP pres

urnal 2013;vo

7.2: Exposuc is plotted.

Toxicologi

appropriate es, in first indent or not (Ethe individu

es. For exam, the three la

eriod of 32 duals, or the dudividuals of two pulses

internal expond

plete repair o

onstrate the texposure testms and PPP es can be deconsisting ofmple, it can be 7.2 are toxe pulses (inc

The minimtier effect s

ddressing thegically depenulse exposuresented in Fig

olume(issue):N

ure profile fo

ical (in)depe

assessment nstance it is EFSA 2005auals of the

mple, in the past pulses canays betweenuration of ththe sensitivecan be consi

osure concen

of damage oc

toxicologicalts (Chapter of concern

emonstrated f fewer pulsebe demonstraxicologically luding the hi

mum numbestudies.

e ecological ndent pulsedes already regure7.2 are t

NNNN

for the exam

endence of d

of risk froimportant to

a). Toxicologsensitive sp

predicted expn be conside

n the second e sensitive lie species (oridered as tox

ntrations in

ccurs betwee

l independensection 7.2) may be usefor the spe

es in higher tated that for t

independentighest peak e

er of toxicolo

threshold opd exposures iesult in the moxicological

DRAF

mple PPPin t

different pul

om exposureo determine gical dependecies is longposure profilered toxicoloand the thirdife-stage, of r sensitive lifxicologically

the individu

en pulse two

nce of differeor toxicoki

d. If the toxcies at risk tier tests thanthe species at from the laexposure) in

ogical depen

ption (ETO)iin higher-tier

maximum effelly dependen

T GD on tiere

the D1 ditch

lse exposure

e profiles chwhether or n

dence of repeg enough tole as presentogically inded pulse is lonthe sensitive

fe-stage) havy independen

ual organism

and three.

ent pulse expinetic/toxicodxicological init may be v

n present in tat risk the firast three pulsthe higher-ti

ndent pulse

it may not br effect studiects. For exa

nt and the firs

ed RA for edg

h scenario o

es

haracterised not the pulseeated pulses o also experted in Figurependent fromnger than thee species at rve a longer lint relative to

ms drop belo

posures, eithdynamic mondependencevalid to adopthe predictedrst two pulseses, it may suier refined ex

exposures t

e necessary es if it can b

ample, if all pst two pulses

ge-of-field surf

on which th

by repeatees are toxicomay occur if

rience repeate 7.2 for the m the first twe average liferisk. In additife-span than

o the last thr

ow critical t

her specially odels for thee of successpt a pulsed d exposure ps of the examuffice to simxposure effec

to address in

to incorporae demonstrapulses of thes already res

face water

72

he Tier 1

d pulsed ologically f the life-ted pulse D1 ditch

wo pulses fe-span of ion, even n 32 days ee pulses

threshold

designed e relevant ive pulse exposure

profile. If, mple PPP

mulate the ct study.

n higher-

ate all the ted that a

e example sult in the



maximum magnitude of effect (e.g. mortality), the last three pulses do not contribute to the magnitude 2777 of the response anymore although the duration of the effect probably will be prolonged by the last 2778 three pulse exposures. 2779

An important question at stake is how to assess the minimum frequency of pulsed exposures, and the 2780 minimum duration of the time-window of the pulsed exposure regime, that will likely result in a 2781 maximum magnitude of effect. The number of PPP applications made in the higher-tier effect 2782 experiment (e.g. refined exposure laboratory toxicity test; micro-/mesocosm study) has to be 2783 considered carefully in relation to the expected biological effects. However, it is considered that the 2784 number of applications in the higher tier studies should be as low as possible and is guided by: (i) the 2785 predicted exposure profile and the number and duration of toxicologically dependent pulsed exposures 2786 exceeding the lower tier RACsw;ac and/or RACsw;ch , 2787

(ii) the time-course of the responses observed in the available laboratory toxicity tests with sensitive 2788 standard and additional test species, 2789

(iii) biological information of the species potentially at risk, and 2790

(iv) read across information for compounds with similar toxic mode-of-action. 2791

For example, if aquatic invertebrates are at risk, the exposure period for the pulsed exposure regime 2792 need not to be longer than the overall duration of the chronic laboratory toxicity for invertebrates (21-2793 28 days; 3 to 4 weekly pulse exposures), if it is likely that the sensitive life-stages of the organisms at 2794 risk are present and the time-to-onset of maximum effect is reached in this period. 2795

Note that in the near future validated/calibrated toxicokinetic/toxicodynamic models may be used to 2796 predict the time course of effects of time-variable exposures and consequently, also to identify 2797 minimum number of toxicologically dependent pulse exposures that has to be addressed in the higher-2798 tier effect study to assure that the maximum magnitude of effects will occur. 2799

7.1.5. Ecological (in)dependence of different pulse exposures 2800

When the toxicologically (in)dependence of successive pulse exposures is sufficiently addressed, it 2801 may be important to also demonstrate their ecological (in)dependence, particularly when ecological 2802 recovery is taken into account in the effects assessment (e.g. to address the ecological recovery option 2803 by means of micro-/mesocosm tests). Successive pulsed exposures may be considered ecologically 2804 independent if peak intervals are greater than the relevant recovery time of the sensitive populations of 2805 concern. The possible ecological independence of pulsed exposures may also be of importance in the 2806 risk assessment if the potentially sensitive species, or specific sensitive life stages of these species, are 2807 not present in the periods that certain pulsed exposures occur (e.g. pulsed exposure in winter because 2808 of drainage). 2809

Evaluating the ecological dependence of successive pulsed exposures will be important when 2810 microcosm and mesocosm tests are used in the risk assessment that aim to derive the ERO-RAC (RAC 2811 on basis of the Ecological Recovery Option) and when the frequency of pulsed exposures(or the 2812 exposure pattern) in the edge-of-field exposure profile deviates from that tested in the micro-2813 /mesocosm experiment (see section 7.1.4 above). In that case the total period of effects may be 2814 estimated by plotting the micro-/mesocosm derived ETO-RAC (RAC on basis of the Ecological 2815 Threshold Option) as well as the ERO-RAC on the field exposure profile. To illustrate this procedure 2816 for the example PPP, these RAC values are plotted on the D1 ditch exposure profile (Figure 7.3). The 2817 mesocosm study from which these RACs are derived is characterised by 3 weekly applications of the 2818 PPP(simulating the last three pulses of the D1 ditch exposure profile). The maximum magnitude of 2819 effects on sensitive invertebrate populations was observed between the second and third weekly 2820 application, while at the treatment level from which the ERO-RAC is derived full recovery of the 2821 affected populations was observed 6 weeks after the last application. So the ERO-RAC is based on a 2822 time-window of effects of approximately 7 weeks, followed by recovery. In both the upper and lower 2823

2824 2825 2826 2827 2828 2829

2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843

2844 2845 2846 2847

2848 2849 2850 2851 2852

EFSA Jou

panel ofprepositiconsidercharacteditch scemesocos

It appearprotectioETO-RAtotal effethe totaleffect perisk thensituationERO-RAperiod ofirst twoapplicatiRAC if tview.

Further ithat deal7.2) motoxicity t

Figure 7values athe uppewhile the

urnal 2013;vo

f Figure 7.3ion ‘provisiring the totalrised by threnario. In thsm experimen

rs from Figuon goal allowAC and the tect-period th effect perio

eriod of apprn the provisin depicted inAC derived of effects wilo pulses, whion. So, in ththe total per

information ls with the c

ore detailed tests will be

7.3: Exposuraddressing reer panel an Ee provisional

olume(issue):N

3 it is assumonal’ is usel period of eee weekly p

he upper pannt is 0.041 µ

ure 7.3 that thws some effetime needed hat might be od most likelroximately 7 ional ERO-R

n the lower pfrom the mell last much hile recoveryhe latter caseiod of estim

on the interpconduct and information provided.

re profile forecovery (EROTO-RAC of l ERO-RAC

NNNN

med that theed since theeffects to be pulse exposunel of Figureµg/L, while th

he authorisatiects followedfor recoveryexpected. F

ly will not bweeks, follo

RAC can bepanel of Figuesocosm stulonger than

y most likely. the provisioated effects

pretation of tinterpretatioon the con

r the examplO-RAC) or t

f 0.041 µg/L is the same.

DRAF

e provisionae definitive expected on

ures and thee 7.3 it is ashat is 0.030 µ

ion of the exd by recovery derived froor the situat

be longer thaowed by fulle upgraded ture 7.3, it caudy (characte

seven weeksy will not bonal ERO-RAwill be long

the provisionon of micro-/nduct and in

le PPP in thethe thresholdand in the lo

T GD on tiere

l ERO-RACERO-RAC

n basis of the total exposssumed that µg/L in the l

xample PPP ory. Evaluatin

om the mesocion depicted

an that derive recovery, iso an official

annot be excerised by thrs, since effecbe observed AC cannot b

ger than acce

nal ERO-RA/mesocosm tnterpretation

e D1 ditch ond level of ef

ower panel an

ed RA for edg

C is 0.052 µcan only b

he results of sure profile the ETO-RAower panel.

only can be gng the exposcosm test, pr

d in the uppeed from the s considered l ERO-RACluded on baree weekly pcts may alreaearlier than

be used to deeptable from

AC will be prtests. In the

n of refined

n which mesffects (ETO-n ETO-RAC

ge-of-field surf

µg/L. Note be establishthe mesocopredicted fo

AC derived f

granted if thesure period arovides insig

er panel of Fmesocosm tto demonstr

C. In contrastsis of the prpulses) that ady be cause

n six weeks erive an offic

a regulatory

rovided in senext chapterexposure la

socosm-deriv-RAC) are p of 0.030 µg

face water

74

that the hed when sm study

or the D1 from this

e specific above the ght in the

Figure 7.3 test. If an rate a low t, for the

rovisional the total

ed by the post last

cial ERO-y point of

ection 7.3 r (section aboratory

ved RAC lotted. In

g/L is use,



7.2. Refined exposure laboratory toxicity tests 2853


Risk assessments based on laboratory toxicity tests performed under constant exposure conditions may 2855 overestimate potential risks. In cases where predicted (modelled) field exposure profiles differ 2856 considerably from exposure regimes in standard toxicity studies it may be appropriate to design 2857 higher-tier laboratory toxicity tests that more closely resemble realistic exposure scenarios. 2858

As described in detail in section 7.1, in designing refined exposure laboratory toxicity tests with 2859 standard and additional aquatic test species, information on the relevant field exposure predictions 2860 needs to be considered. In order to adopt a realistic worst-case exposure scenario in the toxicity test, 2861 the refined exposure regime tested should be deduced from the relevant field exposure scenarios and 2862 the relevant intended agricultural use of the PPP (e.g. FOCUS scenarios or monitoring studies). In 2863 addition, it is necessary to consider whether the first tier procedure triggers acute or chronic risks. In 2864 refined exposure studies supporting acute and chronic risk assessments the peak concentration may be 2865 used in both the PEC and the RAC estimate, if (1) the exposure profile (e.g. height and width of the 2866 pulse exposure) in the refined laboratory toxicity test (on which the RAC is based) is relatively worst-2867 case when compared with that of the relevant predicted (modelled) field exposure profile, and (2) the 2868 duration of the test is long enough to allow the expression of the effects. 2869

Refined exposure laboratory toxicity tests are generally used to address the threshold level of effects 2870 and are less useful to address ecological recovery. Recovery potential of sensitive population within a 2871 realistic community context, however, can be studied in micro-/mesocosm experiments (see chapter 2872 section 7.3) 2873

7.2.2. Reasons to perform refined exposure laboratory toxicity test 2874

Performing refined exposure tests in the risk assessment for PPPs may be done for several reasons, 2875 viz.: 2876

1. To address the effects of time-variable exposures on relevant organisms in case it is 2877 recommended to use the PECsw,max in the chronic risk assessment 2878 If the use of the TWA approach in the chronic risk assessment is appropriate (see section 2.5 2879 of Chapter 2), refined exposure laboratory toxicity tests need not to be performed as a higher-2880 tier option, since the RACsw,ch derived from concentration–response relationships observed in 2881 the standard chronic toxicity tests, can be readily compared with the appropriate PECsw,twa. 2882 However, long-term refined ecotoxicological exposure studies, for example simulating 2883 repeated pulse exposures, may be a higher-tier option if the TWA approach cannot be used. 2884

2885 2. In case of doubt, to demonstrate that for the species at risk (or a representative standard test 2886

species) the PECsw,twa can be used in the chronic risk assessment 2887 If on basis of the available information it is uncertain whether it is appropriate to use the 2888 PECsw,twa in the chronic risk assessment, experiments may be designed to investigate whether 2889 peak concentrations or TWA concentrations better explain the treatment-related responses 2890 observed. These experiments can be performed using different exposure scenarios that are 2891 comparable in the concentration x time factor (e.g. the same 21-d TWA concentration) but 2892 variable in peak exposure concentrations. When the response can be best expressed related to 2893 the experimental TWA concentration/area under the curve, it can be assumed that the 2894 PECsw;twa approach is valid (section 2.5). 2895

2896 3. To address the risks of pulse exposures in addressing possible latency of effects 2897

When designing tests to address risks of (single or several) pulse exposures, generally the 2898 most sensitive life-stage of the test organism of concern and the pulse of exposure should co-2899 occur in the test. If it can be proven that the most sensitive life-stage and the pulse exposure 2900



are not occurring at the same time, a realistic combination of exposure and appropriate life-2901 stage might be chosen. Pulsed exposure on a sensitive life-stage of the test species (e.g. small 2902 juveniles) may already yield the maximum acute and chronic effects. However, latent effects 2903 should be considered as short pulse exposures of PPPs may need a longer time to express the 2904 effect in the test organism as shown in the literature (Duncan et al 2009). Hence, in case of 2905 possible latency of effects it may be necessary that the duration of the refined exposure test 2906 simulating a short-term pulse exposure covers the full life span of the test organism, or at 2907 least the relevant sensitive aquatic life stage (see also section 6.4.5). 2908 2909

4. As a higher-tier effect assessment approach for organisms that usually are not tested in 2910 micro-/mesocosms 2911 In principle, effects of realistic time-variable exposure regimes on populations of freshwater 2912 organisms can be studies in aquatic micro-/mesocosm tests, if test duration covers the full life 2913 span of the test organism as outlined above under bullet point 3 (see also section 7.3). Fish 2914 and other vertebrates like amphibians, however, usually are not introduced into these test 2915 systems because of their undue influence on other populations (e.g. invertebrates). As a 2916 higher-tier approach, however, refined exposure tests may be used to study the individual-2917 level effects of a realistic to worst-case time-variable exposure regime on fish (e.g. juvenile 2918 rainbow trout), amphibians and other water organisms usually not studied in micro-2919 /mesocosms. 2920

2921 5. To demonstrate the toxicological (in)dependence of repeated pulse exposures for the species 2922

at risk (or a representative standard test species) 2923 As discussed in section 7.1, toxicological dependence of repeated pulsed exposures may occur 2924 if the life-span of the individuals of the sensitive species is long enough to experience repeated 2925 pulsed exposures. In case of toxicological dependence a second pulse, even of a smaller 2926 height, may increase the toxicity if between pulses the internal exposure concentration has not 2927 yet dropped below the critical level and/or repair of damage did not yet occur. To demonstrate 2928 (in)dependence of different pulse exposures either special designed pulse exposure tests or 2929 toxicokinetic/toxicodynamic (TK/TD) modelling for the relevant organism and the PPP of 2930 concern is required. Additionally, repeated pulsed exposures may occur in successive 2931 generations within a population. In such cases the overall effect on the following generations 2932 will increase. 2933 2934

6. To gain data for TK/TD model validation/calibration 2935 In principle TK/TD models validated to the taxa of concern are appropriate to address the 2936 risks of time-variable exposure regimes to aquatic organisms, however no detailed guidance 2937 on TK/TD modelling is provided in this document for the time being. Special designed pulsed 2938 exposure tests may be used to generate input data for the species of concern to feed the 2939 TK/TD model or to calibrate/validate the appropriateness of the predictions of the TK/TD 2940 model applied. 2941 2942

7. As additional information to complement results of micro-/mesocosm tests if uncertainties for 2943 a particular population remain 2944 Results of micro-/mesocosm experiments (see Chapter section 7.3) that did not simulate the 2945 appropriate exposure profile, still may be useful for the risk assessment if results of additional 2946 laboratory experiments (or TK/TD models) with the most sensitive species from these micro-2947 /mesocosm experiments allow re-interpretation of the exposure-response relationships 2948 observed. In addition, if in a micro-/mesocosm experiment that simulated an appropriate 2949 time-variable exposure regime a specifically desired species is not present (or present in too 2950 low or erratic densities), additional refined-exposure experiments with this species may 2951 sufficiently complement the higher-tier effect assessment. For example, such an approach 2952



may be performed with Lemna gibba, since this macrophyte species usually does not grow 2953 well under mesotrophic conditions simulated in most micro-/mesocosm experiments. 2954

2955

7.2.3. Refined exposure tests with standard test species 2956

When standard test species are assessed in refined exposure laboratory toxicity tests and these tests are 2957 considered appropriate for the risk assessment, a reduction of the AF is not justified when deriving a 2958 RAC. However, a higher toxicity value (e.g. acute EC50 or chronic EC10 or NOEC) from the refined 2959 exposure study with standard test species and the application of the standard Tier 1 AF (e.g. 100 to 2960 derive the acute RAC and 10 to derive the chronic RAC) may be used in case appropriate observation 2961 times are used (96h for acute tests, full life-cycle for chronic tests or a long-term test in the range of 2962 the life-cycle). 2963

Although refined exposure tests with standard test species that more or less resemble the design of 2964 Tier 1 toxicity studies can be used for RAC derivation, the PPR Panel recommends not to use refined 2965 exposure laboratory tests with populations of invertebrates (e.g. Daphnia) for this purpose when also 2966 recovery is considered. These population-level laboratory experiments with invertebrates are usually 2967 performed with individuals that differ in age and developmental state. As a result a rapid onset of 2968 recovery will occur after contamination under such test scenarios. Recourses for surviving individuals 2969 will increase after contamination and will trigger an unrealistic strong recovery as no competitors are 2970 present (Knillmann et al. 2012b). 2971

7.2.4. Refined exposure tests with additional test species 2972

The use of additional test species enables to estimate better the variability between different test 2973 species. Their inclusion into the risk assessment is therefore highly desirable. In principle the same 2974 considerations as for standard test species are relevant. This refers to timing of exposure and latency of 2975 effect of individuals. Especially related to latency of effect, it has been observed several times in the 2976 literature that effects were only visible a long time after exposure (Duncan et al 2009). For example, 2977 caddis flies exposed for 1 hour to the pyrethroid fenvalerate showed a NOEC of 10 µg/L after 15 days 2978 following exposure, but revealed a mortality-NOEC of 0.01 µg/L eight months after exposure (Liess, 2979 2002). A similar observation was made for the Ephemeroptera Cloen sp. exposed for 1 hour to 2980 fenvalerate. Four days after contamination the NOEC was observed at 1 µg/L; 29 days after 2981 contamination the NOEC was 0.001 µg/L (Beketov and Liess, 2005). When compared to the long-2982 term effects of fenvalerate on Daphnia magna, these results show the high variability between test 2983 species. Additional test species should therefore be selected on the base of a literature research 2984 identifying which groups of species are often showing latency of effect. 2985

7.2.5. Derivation of RAC and calibration of refined exposure laboratory toxicity tests 2986

For the derivation of an acute RAC by means of refined acute toxicity tests (that usually should have a 2987 longer duration than the standard protocol tests) with relevant standard test species, it is proposed to 2988 apply an AF of 100 (for invertebrates and fish) to the LC50 or EC50 (expressed in terms of peak 2989 concentration) under the conditions that: 2990

• The pulse exposure in the refined acute laboratory toxicity test is realistic to worst-case when 2991 compared with the relevant predicted (modelled) field exposure profile. 2992

• The repeated pulse exposures predicted for the field are considered to be toxicologically 2993 independent (section 7.1.3). If not, the repeated pulses should be addressed in the refined and 2994 prolonged acute toxicity test. 2995

• The duration of the acute test is long enough to allow the full expression of the effect. 2996 • The refined acute RAC is compared with the PECsw;max. 2997

2998 Long-term refined exposure tests with standard test species (e.g. simulating repeated pulse exposures) 2999 may be a higher-tier option if the TWA approach cannot be used. For the derivation of a chronic RAC 3000



by means of refined chronic toxicity tests it is proposed to apply an AF of 10 to the NOEC/EC10 for 3001 invertebrates and fish, or to the EC50 for plants, expressed in terms of nominal (if measured peak 3002 exposures do not deviate more than 20% from nominal) or measured peak concentration in the test 3003 systems under the conditions that: 3004

• The (repeated pulsed) exposure regime in the refined laboratory toxicity test is realistic to 3005 worst-case when compared with the relevant predicted (modelled) field exposure profile. 3006

• The duration of the test is long enough to reach the full expression of the effect. 3007 • The refined chronic RAC is compared with the PECsw;max. 3008

3009 A summary of the RACsw;ac and RACsw;ch derivation on basis of refined exposure laboratory tests with 3010 standard test species, and their use in the risk assessment, is presented in Table 7.2. 3011 3012 3013 3014 Table 7.2: Derivation of RACs in edge-of-field surface waters, based on refined exposure laboratory 3015

toxicity tests with standard test species. 3016 Type of effect/risk assessment

Relevant PEC

Endpoint of refined exposure toxicity test with standard test species expressed in terms of peak exposure concentration in test system

RAC


PECsw;max L(E)C50 (animal tests)

L(E)C50/100


PECsw;max EC50 (plant tests) (a) EC10 / NOEC (animal tests)

EC50/10

EC10/10 (a) In other EU directives/regulations (e.g. WFD) the EC50 and an AF of 100 is used in the acute effect assessment and the 3017

EC10 (or NOEC) and an AF of 10 in the chronic effect assessment. 3018 3019 If a refined (prolonged) exposure test with rainbow trout (Oncorhynchus mykiss) is performed to 3020 derive a higher-tier RACsw, this RACsw may also be used to assess the effects for larval stages of 3021 amphibians, since rainbow trout is demonstrated to be a good surrogate species for them (see section 3022 5.2.2 and Appendix C). 3023

In case additional test species are used in refined exposure tests, selection of test species needs to 3024 consider typical species assemblages found in undisturbed water bodies and species sensitive to the 3025 selected compound. When refined exposure studies with several additional test species of the relevant 3026 taxonomic group are available the derived toxicity values might be used as described in sections 6.2 3027 (geomean method) and 6.3 (SSD method), at least when conditions as described above for the 3028 derivation of refined RACs are met. 3029

Note that the PPR Panel proposal predominantly addresses the uncertainty of the ecotoxicological 3030 endpoint. It is assumed that the predicted field exposure profile is sufficiently realistic to worst-case. 3031 Furthermore, note that in a refined risk assessment the uncertainty of the exposure estimate can be 3032 assessed as well. With regard to the calibration of assessment it is considered that long-term 3033 mesocosm experiments containing a realistic community are suitable to validate and calibrate 3034 assessments based on refined exposure laboratory tests. For this, the most sensitive population- or 3035 community-level endpoint should be used (see section 7.3). 3036

3037



7.3. Model ecosystem experiments 3038


Aquatic model ecosystems - usually referred to as microcosms and mesocosms - are bounded systems 3040 that are constructed artificially with samples from, or portions of, natural aquatic ecosystems, or that 3041 consist of enclosed parts of natural surface waters. They usually are characterised by a reduction in 3042 size and complexity when compared with their natural counterparts but they include an assemblage of 3043 organisms representing several trophic levels. Indoor experimental ecosystems are often referred to as 3044 microcosms and outdoor experimental ecosystems as mesocosms, but their difference mainly concerns 3045 their size. Crossland et al. (1993) defined outdoor microcosms as experimental tanks/ponds less than 3046 15 m3 water volume or experimental streams less than 15 m length, and mesocosms as test systems 3047 greater than 15 m3 water volume or 15 m length. The most frequently used freshwater model 3048 ecosystems in European PPP risk assessment are those that mimic shallow, static freshwater habitats 3049 (see Brock & Budde 1994; Caquet et al. 2000), but in recent years more ecotoxicological experiments 3050 with PPPs in artificial streams become available (e.g. Heckmann and Friberg 2005; Mohr et al. 2007; 3051 2008; Liess and Beketov, 2011). 3052

Micro- and mesocosm studies performed for PPP authorisation aim to simulate realistic natural 3053 conditions and environmentally realistic PPP exposure regimes. These studies normally follow 3054 experimental designs to demonstrate causality between treatment and effects, and can also identify 3055 concentration-effect relationships at the population and community level (including structural and 3056 functional endpoints). 3057

The advantage of micro- and mesocosm studies over the other types of experimental higher-tier 3058 studies (e.g. additional laboratory toxicity tests to construct SSD’s; refined exposure studies) is their 3059 ability to integrate more or less realistic exposure regimes with the long-term assessment of endpoints 3060 at higher levels of biological integration (population and community-level effects), and to study intra- 3061 and inter-species interactions and indirect effects in a more or less realistic community. In addition, a 3062 higher number of species and ecological groups are exposed for which dose-response relationships 3063 may be obtained. Since micro-/mesocosm tests can be performed for a relatively long time, and 3064 observations can go on long after the exposure has declined below the threshold level of effects, these 3065 test systems may be used to assess latency of effects and population and community recovery. The 3066 advantage of micro- and mesocosm studies over field monitoring studies is that due to increased 3067 control over confounding factors, causality between PPP exposure and effects is easier to demonstrate. 3068 In addition these kind of studies allow replications and real controls, which would be not possible in a 3069 field study. 3070

It is important to note that communities and environmental condition in micro-/mesocosm represent 3071 only one of the many possible conditions of edge-of-field surface waters. Edge-of-field surface water 3072 bodies potentially at risk vary in community structure (including species composition and life cycle 3073 traits) and abiotic conditions. This should be accounted for in the effect assessment, e.g. by applying 3074 an appropriate AF for spatio-temporal extrapolation of the concentration-response relationships 3075 observed in micro-/mesocosms. 3076

7.3.2. Designing micro-/mesocosm experiments 3077

Useful guidance on experimental design and endpoint selection to conduct a proper micro-/mesocosm 3078 experiment is provided by OECD (2006) and by workshop documents of SETAC Europe (e.g. 3079 Giddings et al. 2002; Brock et al. 2010a; Maltby et al. 2010). The major items for an appropriate 3080 design of an aquatic micro-/mesocosm test within the context of this Guidance Document (focus on 3081 individual PPPs in edge-of-field surface waters) concern: 3082

1. the establishment of an aquatic community in the test systems that is representative for edge-of-3083 field surface waters and can be used in the effect assessment, 3084



2. the appropriate exposure regime of the PPP under evaluation to simulate in the micro-/mesocosm 3085 test system, 3086

3. the number of treatments, choice of the doses and replicate test systems per treatment to derive a 3087 concentration-response relationship, 3088

4. the selection of measurement endpoints (type, number and frequency) for the higher-tier effect 3089 assessment, and 3090

5. the methods for statistical and ecological evaluation of concentration-response relationships. More 3091 detailed guidance on these items will be provided below. 3092

3093

7.3.2.1. Establishment of a representative aquatic community in the test systems 3094

Communities used in micro- and mesocosm studies should be representative for edge-of-field surface 3095 waters. A representative freshwater community for edge-of-field surface waters contains the important 3096 taxonomical groups (not necessarily the same species), trophic groups (e.g. primary producers, 3097 detritivores, herbivores, carnivores) and ecological traits (particularly life cycle characteristics related 3098 to vulnerability as generation time and dispersal ability, see Liess et al. 2008) typical for communities 3099 in ponds, ditches and/or streams. Consulting ecological scenarios for edge-of-field surface waters of 3100 European landscapes may be important and some preliminary guidance on this is provided in chapter 3101 15 of the ELINK document (Wogram 2010; Brock et al. 2010b; Alonso Prados and Novillo-Villajos 3102 2010; Biggs and Brown 2010) and in Gergs et al. (2011) and in section 3.6 on vulnerable species. 3103

Micro-/mesocosm studies can be performed in artificial constructions (mimicking ponds, ditches or 3104 streams) or by enclosing parts of existing aquatic ecosystems (field enclosures). Already established, 3105 uncontaminated, aquatic ecosystems that resemble the required species composition of the micro-3106 /mesocosm test can be used as a source of water, sediment and organisms to seed the artificial test 3107 systems. This will ensure that into the test systems a more or less similar and representative 3108 community (e.g. characterised by zoo- and phytoplankton, pelagic and benthic macro-invertebrates, 3109 and, if necessary, macrophytes) will be introduced. It may be necessary and appropriate to additionally 3110 add certain organisms (e.g. potentially sensitive or vulnerable macro-invertebrates or macrophytes that 3111 are not present in the ‘established source ecosystem’) from other sources if lower tiers or other 3112 information indicate potential risks to specific organisms. It may also be appropriate to use artificial 3113 sediment and water in the tests systems if e.g. the focus of the micro-/mesocosm study is on a specific 3114 group of organisms like aquatic macrophytes (e.g. a herbicide study in an outdoor test system with 3115 potted plants). Using micro-/mesocosms constructed with artificial sediment and water and that aims 3116 to study the effects of PPPs on invertebrates, however, may require a longer acclimatisation period to 3117 develop a realistic pelagic or benthic community. 3118

Artificially constructed model ecosystems require a pre-treatment period of at least several weeks 3119 (plankton-dominated systems) to several months or longer (structurally more complex systems 3120 dominated by long-living macro-invertebrates and macrophytes) in order to allow the establishment of 3121 a community that is recovered from the “construction-stress”, adapted to the conditions in the test 3122 system and characterized by realistic food-web interactions. This will depend on the generation time 3123 of the species considered most relevant (Barnthouse 2004) and may require a time span of few 3124 generation times (Giddings et al. 2002; Liess et al. 2006). However, if during the set-up of the test 3125 system, care is taken to introduce the organisms at natural densities, the acclimatisation period may be 3126 shorter. Currently, there are no micro-/mesocosm studies comparing directly the sensitivity to PPPs of 3127 artificially constructed systems with contrasting periods of establishment. Note that during the 3128 ‘acclimatisation phase’ of micro-/mesocosms other organisms (e.g. aquatic insects), originally not 3129 introduced, may colonise the test systems. This should be considered a normal ecological 3130 phenomenon, provided that the representativeness to edge of field surface water is maintained. 3131

To adequately study potential population and community-level effects for regulatory purposes, it is for 3132 the experimental design of micro-/mesocosm tests important that enough representatives of potential 3133 sensitive (and vulnerable) populations are present. A relevant question at stake is: what should be the 3134



minimum number of potential sensitive/vulnerable populations in micro-/mesocosms from which 3135 concentration-response relationships can be derived to make the study useful for higher-tier RAC-3136 derivation? Note that for RAC-derivation that addresses the ecological threshold option (ETO)it is the 3137 sensitive populations that need most attention, while for RAC-derivation addressing the ecological 3138 recovery option (ERO) most relevant are species that are both sensitive and have a slow recovery 3139 potential (e.g. due to a long life-cycle and/or poorly developed dispersal abilities). In theory, the 3140 required number and identity of potentially sensitive/vulnerable populations that needs to be present in 3141 a micro/mesocosm test system will depend on prior knowledge on the level of sensitivity/vulnerability 3142 of these populations to the PPP of concern. If, for example, laboratory toxicity tests indicate that 3143 certain species are amongst the most sensitive tested (e.g. situated in the tail of SSDs), and individuals 3144 of these species are sufficiently present in the micro-/mesocosm test, the threshold level for toxic 3145 effects derived from these tests may be surrounded by less uncertainty then when this knowledge is 3146 not available. However, species that are identified in laboratory SSDs as most sensitive need not to be 3147 native and likely will not occur in micro-/mesocosms constructed with components of natural 3148 ecosystems. In practice, we often do not know which native species constitute the most sensitive ones. 3149 In contrast it is often known what constitutes the potentially sensitive taxonomic group (e.g. on basis 3150 of lower tiers and read-across for compounds with a similar toxic mode-of-action. In these cases it 3151 seems fair that, besides representatives of different trophic levels, at least 8 different populations of the 3152 sensitive taxonomic group need to be present in the micro-/mesocosm test systems and for which a 3153 concentration-response relationship can be derived. Note that also when applying the SSD approach, 3154 laboratory toxicity data for a minimum number of 8 different taxa of the sensitive taxonomic group are 3155 required in the effect assessment of PPPs for invertebrates and/or primary producers. 3156

Particularly when adopting the ecological recovery concept to derive a RAC it should be carefully 3157 evaluated whether the potentially vulnerable taxa in edge-of-field surface waters (e.g. sensitive 3158 univoltine and semivoltine invertebrates with a low dispersal ability or macrophytes with a relatively 3159 slow growth-rate) are sufficiently represented in the test system (Appendix D). If not, it may be 3160 necessary to apply a higher AF to extrapolate the study specific NOEAEC (No Observed Ecologically 3161 Adverse Effect Concentration), to use modelling tools to extrapolate the observed rate of recovery to 3162 that of more vulnerable field populations, or to derive the RAC on basis of the ecological threshold 3163 concept. Note that species sensitivity distributions constructed with acute EC50’s for aquatic 3164 arthropods and insecticides suggest that the sensitivity of aquatic insects is not correlated with 3165 voltinism of the species involved (see for example Brock et al. 2010b). Although short-cyclic insects 3166 may have similar sensitivity distributions as insects with a more complex life-cycle, it is frequently 3167 reported that sensitive multi-/bivoltine insects recover faster from insecticide-tress than sensitive uni-3168 /semivoltine insects (e.g. Van den Brink et al. 1996; Brock et al. 2009; Liess & Beketov, 2011). 3169 Similarly, sensitive short-cyclic algae may recover faster from herbicide-stress than sensitive 3170 macrophytes with a slower growth-rate (e.g. Coors et al. 2006). If populations of invertebrates and 3171 macrophytes characterised by a low recovery potential are identified to be at risk, efforts should be 3172 undertaken to introduce representatives of these populations when constructing the micro-/mesocosm 3173 test systems. In case populations of these species are affected, the chance that they will recover within 3174 8 weeks is small. Hence, in most cases where uni/semi-voltine invertebrates and/or slow-growing 3175 macrophytes are identified as sensitive groups the recovery option will not bring us further and the 3176 proposed scheme (Figure 7.4) will direct to the threshold option in the effect assessment. When 3177 adopting the threshold concept to derive a RAC it is possible to base it on negligible effect 3178 concentrations for sensitive taxa with short generation times as sensitivity and generation time seems 3179 not linked. Hence to apply the threshold concept it needs not to be a problem when sensitive univoltine 3180 and semivoltine invertebrates with a low dispersal ability or macrophytes with a relatively slow 3181 growth-rate are not sufficiently represented in the test systems. Instead, the availability of data on 3182 negligible effect concentrations for species sensitive to PPPs (high toxicological sensitivity) may 3183 suffice to derive an ETO-RAC (RAC that addresses the ecological threshold option). 3184

When invertebrate and/or plant communities are the principal endpoint of the study, it is 3185 recommended that free-living fish are not included (Giddings et al. 2002). In smaller micro-3186 /mesocosms fish usually cannot be introduced at natural biomass levels appropriate to the abundance 3187



of their prey, and therefore fish can have an undue influence on other populations inhabiting these 3188 confined test systems. However, separate micro-/mesocoms may be used to study the individual-level 3189 effects of a realistic exposure regime on fish (e.g. juvenile rainbow trout; sticklebacks (see further 3190 section 7.2 on higher tier refined exposure tests) 3191

7.3.2.2. Selection and characterisation of the exposure regime 3192

Before designing a micro/mesocosm test for regulatory purposes, in first instance it is important to 3193 evaluate whether the risk assessment using thetier-1 acute or the Tier 1 chronic core data set triggered 3194 the need for the test. If the Tier 1 acute toxicity data triggered the study then the focus usually will be 3195 on effects of peak concentrations. If the Tier 1 chronic toxicity data triggered the study then the focus 3196 may also be on effects of longer-term (time-weighted average) concentrations. 3197

In second instance it is important to evaluate the possible exposure regimes in relevant edge-of-field 3198 surface waters that may result from normal agricultural use of the PPP of concern (e.g. by FOCUS 3199 surface water scenarios and models, multi-year application; see chapter 4), and to identify the relevant 3200 exposure regimes that should be addressed in the acute or chronic effect assessment (see section 7.1). 3201 If the micro-/mesocosm test is triggered by the Tier 1 acute risk assessment and the expected and 3202 relevant field exposure regime is characterised by a single high pulse (e.g. due to drift application), or 3203 by repeated pulses that are toxicologically independent (see for criteria section 7.1.3) a single 3204 application experimental design is an appropriate exposure regime to study in the micro/mesocosm 3205 experiment. The pulse duration should either be equal to larger in the micro-/mesocosm experiment 3206 than that predicted for the field, giving a realistic to worst-case estimation respectively. In these cases, 3207 the RACsw;ac derived from the concentration-response relationships in the micro-/mesocosm tests can 3208 be expressed in terms of the peak concentration of the PPP in the test systems and this RAC-estimate 3209 can be compared with the PECsw;max. 3210

The nominal concentration can be used in the acute effect assessment if, shortly after application, the 3211 measured exposure concentrations in the integrated water column of the test system do not deviate 3212 more than 20% from nominal. Note that during the first hours post application, a heterogeneous 3213 distribution of the test compound in the water column is common which may hamper the proper 3214 measurement of peak concentrations. For fast dissipating compounds the proper measurement of the 3215 actual peak concentration in the test system may be difficult if not performed shortly after application. 3216 An alternative option to estimate the peak concentration in the test systems may be to measure the 3217 concentration in the application solutions, as well as the amounts of application solution applied to 3218 each test system. In repeated application studies the peak concentration may occur immediately after 3219 the last application if the compound does not dissipate completely from the water column between 3220 applications. 3221

If the expected exposure regime in the field triggers concerns of repeated pulsed exposure that are 3222 likely considered to be toxicologically dependent (see section 7.1.3), a repeated exposure regime 3223 should be adopted in the micro/mesocosm experiment to determine a RACsw,ac and/or RACsw,ch 3224 (dependent on the Tier 1 risk assessment that triggered the micro-/mesocosm study). Guidanceon the 3225 minimum number of toxicological dependent pulse exposures to address in the micro-/mesocosm 3226 experiment is provided in section 7.1.4. 3227

In a micro-/mesocosm experiment triggered by the Tier 1 chronic core data, a worst-case approach is 3228 to maintain a more or less constant PPP concentration for at least the duration of the chronic toxicity 3229 test that triggered the micro/mesocosm test, unless the TWA (Time weighted Average) exposure can 3230 be used to express the treatment-related effects (see section 2.5). Alternatively, the long-term exposure 3231 regime simulated in the test systems should be realistic to worst-case relative to the predicted exposure 3232 profile. 3233

To appropriately characterise the exposure regime in the micro-/mesocosm experiment it generally is 3234 not sufficient to report nominal concentrations only. In addition to the dissipation DT50 in the 3235



experiments and the timing of application the following information has to be provided. The analytical 3236 recovery of the test substance in the relevant matrix (water within the context of the current Aquatic 3237 Guidance), exposure concentrations of the dosing solutions used to treat the test systems, and 3238 concentrations measured in water (and sediment) samples at several time points after each application, 3239 should be reported. This allows to express the effect estimates (e.g. NOECs; ECx values) in terms of 3240 the ecotoxicologically relevant concentration (ERC; see section 3.2). 3241

7.3.2.3. Number of treatments, choice of the doses and replicate test systems per treatment 3242

Because of the ecological complexity of micro-/mesocosm experiments it should be realised that 3243 practicality (in terms of man-power and costs) will limit the number of test systems that can be 3244 constructed, treated, sampled and analysed. According to OECD (2006) the number of treatments and 3245 choice of doses, as well as the number of replicate micro-/mesocosm test systems per treatment 3246 depend on the objectives of the study. These objectives may relate to the required effect endpoint (e.g. 3247 EC50, ECx, NOEC values for relevant endpoints), the level of required precision for the effect 3248 estimates (desired power of the test) and the size of the effect which is considered of ecological 3249 significance. For regulatory purposes of PPPs the delegates of the CLASSIC workshop (Giddings et 3250 al. 2002) and OECD (2006) recommended an exposure-response experimental design with preferably 3251 5 or more concentrations, and at least 2, but preferably more, replicates per concentration. An 3252 exposure-response experimental design characterised by a larger number of treatments but a lower 3253 number of replicates per treatment, allows wider use of the data under different regulatory conditions 3254 (e.g. different exposure regimes due to differences in application and mitigation methods, receiving 3255 water bodies etc.) than an ANOVA design with a limited number of treatments but a high number of 3256 replicates per treatment. When adopting an exposure-response experimental design it is common 3257 practise to select more replicates for the controls (often the double amount) than for treatments to 3258 increase the statistical power. Preferably, the lowest test concentration should not result in treatment-3259 related responses, while the highest concentration tested should result in pronounced effects on several 3260 measurement endpoints. This allows the derivation of threshold concentrations for toxic effects, as 3261 well as putting in perspective the possibly more subtle treatment-related responses caused by 3262 intermediate concentration levels. This implies that the selected exposure concentrations should 3263 always be guided by lower-tier effect information (e.g. single species toxicity tests) and the expected 3264 field exposure regime of the substance under evaluation. For this purpose toxicokinetic/toxicodynamic 3265 and/or population models might be used, some models available nowadays are already in a state to 3266 provide this information. Validation and testing of the models itself is not as crucial for designing a 3267 mesocosm experiment as for the direct application in risk assessment. It is expected that using these 3268 model approaches will come up with more precise studies than expert judgement. 3269

7.3.2.4. Measurement endpoints 3270

The appropriate measurement endpoints to study, and the frequency of sampling, in a micro-3271 /mesocosm experiment will depend on the specific protection goals defined for the water organisms 3272 potentially at risk in edge-of-field surface waters (see chapter 3). According to the specific protection 3273 goals defined, the key drivers and their ecological entity to be protected concern populations in case of 3274 aquatic algae, vascular plants and invertebrates, individuals-populations in case of aquatic vertebrates 3275 and populations-functional groups in case of aquatic microbes. This implies that for most organisms at 3276 risk that are studied in micro-/mesocosm tests the selected measurement endpoints should relate to 3277 relevant population-level endpoints, more specifically the attributes survival/growth and 3278 abundance/biomass (see section 3.5). By protecting sensitive/vulnerable populations of primary 3279 producers and invertebrates, community-level effects (including processes) and biodiversity may be 3280 sufficiently ensured, particularly when addressing the ecological threshold option. 3281

The duration of the study and frequency of sampling should be adapted to the treatment-regime (e.g. 3282 more frequent sampling immediately after PPP application), the duration of the life-cycle of the 3283 organisms of concern (e.g. initially more frequent sampling for short-cyclic organisms), and the 3284 objective of the study (e.g. if the study aims to demonstrate recovery or not and the pre-defined period 3285 of acceptable effects). In all cases pre-treatment samples should be taken (at least 1 or 2 pre-treatment 3286



samplings) to demonstrate the suitability of the test systems in terms of similarity of relevant 3287 parameters between test units. For detailed information on test duration and sampling is referred to 3288 OECD (2006) and the CLASSIC document (Giddings et al. 2002). 3289

Microcosm and mesocosm experiments are test systems that allow studying treatment-related effects 3290 of PPPs at the population and community level. Population responses in micro-/mesocosm are usually 3291 studied by means of measurement endpoints that provide information on dynamics in population 3292 abundance, biomass and/or growth. Measurement endpoints to study community-level responses 3293 usually comprise community metabolism endpoints indicative for ecosystem processes like dynamics 3294 of dissolved oxygen in water and rates of decomposition of particulate organic matter (e.g. in litter 3295 bags), but also summary parameters of population-level endpoints, like diversity indices and scores 3296 based on multivariate techniques and trait-based groupings of organisms (for more details see section 3297 7.3.2.5). 3298

The number of taxa/populations occurring in micro-/mesocosms, and consequently the potential 3299 measurement endpoints, may be high. Studying all potential measurement endpoints is very expensive. 3300 For reasons of cost-effectiveness usually a limited number of measurement endpoints are selected. 3301 Available lower-tier studies for the PPP under evaluation (e.g. standard and additional laboratory 3302 toxicity tests) and/or results of model ecosystem experiments with related compounds (characterized 3303 by a similar toxic mode-of-action) may provide insight which structural and functional parameters 3304 should be studied intensively. For example, if the PPP under investigation is a selective herbicide and 3305 the laboratory toxicity tests indicate that algae and the macrophytes Lemna and/or Myriophyllum are at 3306 least an order of magnitude more sensitive than the invertebrates Daphnia and/or Chironomus, the 3307 primary focus of the selected measurement endpoints should be on populations of phytoplankton, 3308 periphyton and macrophytes (structural endpoints for primary producers) and possibly also on 3309 parameters indicative for the functioning of primary producers, such as dissolved oxygen and pH. This 3310 does not mean that no attention should be paid to responses of invertebrates (e.g. to demonstrate 3311 indirect effects) but that it may be enough to select a limited number of measurement endpoints to 3312 monitor the treatment-related effects on taxonomic groups that likely will not be impacted because of 3313 direct toxic effects. If the PPP of concern is an insecticide for which standard toxicity tests and model 3314 ecosystem experiments with related compounds indicate that crustaceans and insects in particular are 3315 sensitive, the focus of the study should be on populations of zooplankton and macro-invertebrates 3316 (possibly including emergent insects and effects of shredder populations on the breakdown of 3317 particulate organic matter) while a limited number of (summary) measurement endpoints for primary 3318 producers may be sufficient. However, if recovery of indirectly affected organisms is of concern it 3319 may be necessary to study endpoints related to this response also in more detail. 3320

In contrast, if the difference in toxicity between the standard test organisms is small, as might be the 3321 case for fungicides with a biocidal mode-of-action, the selected measurement endpoints should include 3322 a variety of taxonomical groups such as populations of primary producers (e.g. algae and vascular 3323 plants) and invertebrates (e.g. zooplankton and macro-invertebrates, including non-arthropods) and 3324 when deemed relevant (e.g. for triazole fungicides; see section 3.5), also microorganisms. Note, 3325 however, that currently validated and internationally recommended (e.g. OECD) methods to study 3326 population-level effects of PPPs on microorganisms are hardly available. The available toxicity tests 3327 with microorganisms usually focus on biomass or processes (Van Beelen 2003) or concern 3328 community-level tests based on molecular methods such as sequencing techniques (Diepens et al. 3329 2013). A way forward for edge-of-field risk assessment of fungicides may be to study the 3330 consequences of previously exposed plant litter on the feeding-behaviour and survival of shredders 3331 (e.g. Bundschuh et al. 2011). Note that leaf-shredding invertebrates prefer leaves conditioned by 3332 microorganisms (particularly aquatic hyphomycetes) as they are more palatable and nutritious 3333 (Bärlocher 1985; Maltby 1992). 3334

3335



7.3.2.5. Statistical and ecological evaluation of concentration-response relationships 3336

Considering the specific protection goals defined for water organisms (see chapter 2) the statistical 3337 analysis of measurement endpoint related to population-level effects are a necessity. Several state-of-3338 the-art techniques are available for univariate analysis (e.g. the Williams test; Kruskal-Wallis multiple 3339 comparison test; Dunnett’s test) to calculate NOECs and LOECs at the population-level. It needs to be 3340 noted that more complex micro-/mesocosm experiments may result in considerable inter-replicate 3341 variation in measurement endpoints, including population densities (Sanderson et al. 2009). 3342 Consequently, to identify the robustness of the LOEC identification the Minimum Detectable 3343 Difference (MDD) should be reported for single species together with the statistical approach used 3344 (see below). At the population level, concentration-response relationships may also be evaluated by 3345 means of logistic or non-linear regression techniques to calculate ECx values. To assure that an effect 3346 of the PPP is treatment-related and not background variability, the ECx value has to be significant, 3347 determined by an adequate statistical test (e.g. Chi2-Test for probit or logistic analysis), while also the 3348 confidence intervals of the ECx estimates need to be reported to evaluate the experimental uncertainty 3349 associated with the ECx estimate. Note that population-level NOECs/LOECs and ECx values usually 3350 can be calculated only for taxa that occur in high enough numbers and that dominate the community. It 3351 is therefore recommended to also perform univariate statistics on aggregated data, for example total 3352 densities of organisms at a higher taxonomic level (e.g. family; order) or on basis of densities of 3353 organisms with specific ecological traits (see e.g. Liess & Von der Ohe, 2005; McGill et al. 2006; 3354 Liess & Beketov, 2011; Gergs et al. 2011). 3355

Univariate statistical tests assume a monotonous concentration-response relationship (increasing effect 3356 with increasing concentration). The population and community-level responses observed in a long-3357 term micro/mesocosm test, however, may be the result of interplay between direct toxic effects and 3358 indirect effects due to shifts in ecological interactions between populations. Factors like indirect 3359 effects may violate the assumption of an increased effect with increasing concentration. For this 3360 reason the identification of treatment-related responses should not only be based on statistics but also 3361 on ecotoxicological knowledge (to identify the direct toxic effects) and ecological knowledge (to 3362 identify possible indirect effects). Statistically significant responses in the same direction (either 3363 decreases or increases) on consecutive samplings should be given special weight. Note that because of 3364 the high number of possible endpoint – sampling date combinations a statistically significant effect on 3365 an isolated sampling may be easily detected while it may be a Type II error. 3366

Micro-/mesocosm tests also allow to study treatment-related responses at the community level. To 3367 evaluate community-level effects multivariate techniques (e.g. Redundancy analysis (RDA) and 3368 Principle Response Curves (PRC) in combination with Monte Carlo permutation tests) may be an 3369 appropriate tool (e.g. Van Wijngaarden et al. 1995; Van den Brink & Ter Braak, 1999). An advantage 3370 of these multivariate techniques is that they also provide species scores that can be used to identify the 3371 most important species/populations explaining the community response. In addition, diversity indices 3372 and approaches to describe the treatment-related response of biological communities in terms of traits 3373 (e.g. the SPEAR index; Liess and Beketov, 2011) may be used e.g. in order to reduce inter-replicate 3374 variability. A review of the NOEC/LOEC of the most sensitive population in the micro-/mesocosm 3375 test system with the NOEC/LOEC values on basis of trait-based groupings, and that include different 3376 types of PPPs (insecticides, fungicides, herbicides), is a topic for future research. These trait-based 3377 groupings may be analysed on basis of univariate and multivariate techniques, the SPEAR-approach, 3378 or a combination of these techniques. We foresee that in the future further guidance can be provided 3379 on the respective advantages of the various methods. 3380

Currently NOECs/LOECs for measurement endpoints are assessed in the majority of micro-3381 /mesocosm experiments (a limited number of model ecosystem experiments focus on deriving ECx 3382 values), but the statistical power to define these endpoints are routinely not reported in micro-3383 /mesocosm reports present in PPP dossiers. Thus also NOECs may be reported for populations (or 3384 aggregated species groups) which are present in the micro-/mesocosm test system, but for which, due 3385 to low abundance and variability between replicates, statistically detecting any effects will be 3386



impossible. So a study can have a long species/taxa list, but for fewer taxa concentration-response 3387 relationships can be evaluated. Also the demonstration of recovery depends on the statistical power. 3388 So in a bad case recovery can be demonstrated just by higher variability in the controls and treatments 3389 during the recovery period shifting the statistical power in a way that no effects can be demonstrated. 3390 Calculating the minimal detectable difference (MDD) allows to report the actual effect which could be 3391 determined in the experiment for a given endpoint at a given time. For applying the MDD concept to 3392 micro-/mesocosm experiments it is noteworthy that the MDD is particularly important if no effect is 3393 observed, since when a LOEC can be calculated the statistical power apparently is high enough to 3394 detect an effect. Additionally it is noteworthy that a high MDD for several measurement endpoints is a 3395 common phenomenon in micro/mesocosm studies (since only a limited number of populations 3396 dominate the community) but this need not to be a reason to reject the study if for several relevant 3397 endpoints/populations (e.g. 8 populations of the sensitive taxonomic group) a statistical evaluation can 3398 be performed. We recommend that in a first step the MDD should be reported together with the NOEC 3399 table for each investigated endpoint in time. We propose to cluster the MDD into 5 groups and the 3400 NOECs in the NOEC table might be shaded (or marked) according to the example shown below 3401 (Table 7.2). It has to be noted that the selection of MDD classes are more or less arbitrary and they 3402 should be revised in future on an appropriate database. A case study of this approach is conducted in 3403 the moment and will be published soon. 3404

Table 7.2: Proposal on classes of minimal detectable differences (MDD) 3405

Class MDD Comment 0 >100% No effects can be determined I 90-100% Only strong effects can be

determined II 70-90 % Strong to medium effects can be

determined III 50-70 Medium effects can be

determined IV 10-50% Low effects can be determined

3406 3407 Use of MDD is introduced to increase transparency and trustability in endpoints derived from micro-/ 3408 mesocosm studies. MDD of critical endpoints should ideally exceed class II. Considering the high 3409 level of biological variance in micro-/ mesocosms (and natural edge-of-field surface waters), 3410 endpoints with lower MDD classes (I-III) may however be considered relevant. For a proper 3411 evaluation also information is required on the normal fluctuations in population densities (for the 3412 organisms of concern) in natural edge-of-field surface waters. It is anticipated that in the coming years 3413 more practical experience will be obtained in applying MDD to evaluate results of micro-/mesocosm 3414 experiments. This practical experience is required before more detailed guidance on MDD and the 3415 interpretation of micro-/mesocosm endpoints can be provided. The PPR Panel advices to prepare a 3416 specific opinion on the use of MDD and the evaluation of micro-/mesocosm studies. 3417

7.3.3. Interpreting micro-/mesocosm experiments 3418

In recent years, discussions shifted towards the awareness of inconsistencies in both the way the same 3419 mesocosm data are interpreted and the assessment factors applied by regulatory experts in different 3420 EU Member States. The Dutch Platform for Assessment of Higher-tier Studies has produced a 3421 guidance document on how micro-/mesocosm data should be presented and evaluated in a uniform and 3422 transparent manner (De Jong et al. 2008). We propose to largely use this document to present and 3423 evaluate micro-/mesocosm studies for regulatory purposes when placing PPPs on the European 3424 market. The main aspects to consider and some deviations are presented below. 3425

3426



7.3.3.1. Evaluation of the scientific reliability of the micro-/mesocosms test for PPP authorisation 3427

On basis of the information presented in section 7.3.2 the following questions should be answered in 3428 the evaluation of the scientific reliability of the micro-/mesocosm experiment. 3429

3430 1. Is the test system adequate and does the test system represent a realistic freshwater community? 3431

[Trophic levels; taxa richness and abundance of (key and sensitive) species; representativeness 3432 of the biological traits with respect to vulnerability] 3433

2. Is the description of the experimental set-up adequate and unambiguous? [ANOVA or 3434 regression design; overall characterization of the experimental ecosystem/community 3435 simulated; measurement endpoints; sampling frequency; sampling techniques] 3436

3. Is the exposure regime adequately described? [Method of application of the test substance; 3437 relevance for predicted exposure profile in the field; concentration in the application solution; 3438 dynamics in exposure concentrations in relevant compartments (e.g. water, sediment); 3439 detection limits] 3440

4. Are the investigated endpoints sensitive and in accordance with the working mechanisms of the 3441 compound, and with the results of the first tier studies? [Compare selected measurement 3442 endpoints with the species potentially at risk as indicated by the lower tiers] 3443

5. Is it possible to evaluate the observed effects statistically and ecologically? [Univariate and 3444 multivariate techniques applied; unambiguous concentration-response relationships; statistical 3445 power of the test; ecological relevance of the statistical output]. 3446

3447 The above mentioned questions could be answered with Yes, Unclear or No, and the answers should 3448 be substantiated with arguments. 3449

A further detailed checklist to assess the scientific reliability of the study is given in the table below, 3450 followed by another table explaining the Reliability index that might be used to classify the overal 3451 quality of the study. 3452

3453

Table 7.3: Checklist for evaluating micro-/mesocosm studies for regulatory purposes (adapted from 3454 De Jong et al 2008) 3455

ITEMS NOTES RELIABILITYIndex 1 – 3 *

METHODOLOGY & TEST DESCRIPTION 1. Substance Properly characterised and reported? 1.1 Concentration [identity and amount of a.s. per litre test water?] 1.2 Formulation and purity

[ingredients in the formulation influencing the working action of the a.s. should be reported]

1.3 Vehicle [in case a vehicle — other than in the formulation — is used, identity and concentration?]

1.4 Chemical analyses [method, LOQ, LOD, recovery] 1.5 Properties [relevant for potential fate and effects in test system] 2. Test site, duration Properly characterised and reported? 2.1 Location [necessary to make a link between the effects and local

environmental conditions, representativeness]




2.2 Test date / duration [application dates and experimental period?] 2.4 General climatic conditions

[necessary to make a link between the effects and local climatic conditions]

3. Application Properly characterised and reported? 3.1 Mode of application [exposure route; spraying or homogenising the a.s. into the test

medium?]

3.2 Dosage and exposure [actual concentrations during the test?] [chemical analysis of dosing solution?]

3.3 Application scheme [necessary to make a link between the test and the intended use of the PPP]

3.4 Conditions during application

[weather conditions during application, wind speed and temperature?]

4. Test design Properly designed and reported? 4.1 Type & size [e.g. outdoor microcosm, outdoor pond or mesocosm;

dimensions]

4.2 Pre-treatment [proper equilibration?] 4.3 Treatment period [number and spacing of treatments? ] 4.3 Post-treatment [period long enough to allow expression of effects and recovery?] 4.4 Untreated control [sufficient number; solvent applied?] 4.5 Replications [sufficient replications for proper statistical analysis?] 4.6 Statistics [univariate and multivariate techniques applied] 4.8 Dose-response [Number of test concentrations for finding a dose-response

relation (controls excl.)]

4.9 Quality assurance [study conducted under GLP?] 5. Biological system Representative and properly reported? 5.1 Populations [enough sensitive/vulnerable species of the relevant taxonomic

group?]

5.2 Community [the community/ecosystem representative and complete?] 6. Sampling Is sampling adequate for risk assessment? 6.1 General features [relevance selected measurement endpoints] 6.2 Actual concentration [actual concentrations measured in medium and other

compartments or biota?]

6.3 Biological sampling [appropriate methods and frequency?] RESULTS 7. Endpoints Properly reported?

7.1 Type [reported endpoints relevant for objective of study?]

7.2 Value [are measured data consistently presented?]

7.3 Verification of endpoint

[test results are verifiable and source data reported]

8. Elaboration of results Are conclusions based on measured data? Methodology correct?

8.1 Statistical comparison [data meet requirements for method used?]

8.2 Dose-effect relationship

[minimal detectable difference; consistence of response]

8.3 Population-level responses

[sufficiently reported?]




8.3 Community-level responses

[sufficiently reported?]

9. Control

9.1 Untreated control [unexpected effects or disappearance of species?]

9.2 Solvent control [possible effects caused by solvent?]

10. Classification of effects Properly derivable?

11. Biological meaning of statistically significant differences

sufficiently explained?

3456 3457 Table 7.4: Definition of the three values of the reliability index 3458 RELIABILITY INDEX (Ri)

DEFINITION

DESCRIPTION

1 reliable All data are reported, the methodology and the description are in accordance with internationally accepted test guidelines and/or the instructions, all other requirements fulfilled

2 less reliable Not all data reported, the methodology and/or the description are slightly deviating from internationally accepted test guidelines or the instructions, without motivation, or not all other requirements fulfilled

3 not reliable Essential data missing, the methodology and/or the description are not in accordance with internationally accepted test guidelines and/or the instructions without motivation, or not reported, or important other requirements are not fulfilled

3459

Based on the questions and checklist above an overall reliability index should be assigned to the 3460 micro-/mesocosm study. Both Reliability Index Ri1 and Ri2 tests might be used in the risk assessment, 3461 but it may be decided to apply a larger AF in the derivation of the micro-/mesocosm RAC when only 3462 an Ri2 study is available on basis of the most relevant (sensitive or vulnerable) population or 3463 community endpoint (see section 7.3.5). 3464

When the micro-/mesocosm study is deemed reliable to use in the effect assessment of the PPP under 3465 evaluation the concentration-response relationships should be evaluated. Below Effect classes to 3466 summarise the concentration-response relationships of micro/mesocosm experiments are given, based 3467 on the definition by Brock et al. (2006) and De Jong et al. (2008) and modified to add the additional 3468 information about the minimal detectable difference (MDD, see section 7.3.2.5). 3469

3470 Effect class 0 (Treatment related effects cannot be evaluated. If this class is consistently assigned to 3471 endpoints that are deemed most relevant for the interpretation of the study the regulatory reliability of 3472 the micro-/mesocosm tests is questionable) 3473

Due to e.g. low abundance and variability the MDD was always larger than 100% so even 3474 very strong effects could not be determined for the endpoint evaluated. 3475 3476

Effect class 1 (No treatment-related effects demonstrated; the endpoints with the lowest Effect class 1 3477 concentrations may be used to derive the overall NOECmicro/mesocosm). 3478

No (statistically and/or ecologically significant) effects observed as a result of the treatment. 3479 Observed differences between treatment and controls show no clear causal relationship. 3480



3481 Effect class 2 (Slight effects). 3482

Effects concern a short-term and/or quantitatively restricted response usually observed at 3483 individual samplings only. 3484

3485 Effect class 3A (Pronounced short-term effects (< 8 weeks, followed by recovery) . 3486

Clear response of endpoint, but full recovery of affected endpoint within 8 weeks after the 1st 3487 application or, in case of delayed responses and repeated applications, the duration of the 3488 effect period is less than 8 weeks and followed by full recovery. Effects observed at some 3489 subsequent sampling instances. Note that the endpoint can be considered to be recovered only 3490 if the MDD during the relevant recover period was at least smaller than 100%. If this is not the 3491 case an appropriate higher class have to be selected. 3492 3493

Effect class 3B (Pronounced effects and recovery within 8 weeks post last application). 3494 Clear response of the endpoint in micro-/mesocosm experiment repeatedly treated with the test 3495 substance and that lasts longer than 8 weeks (responses already start in treatment period), but 3496 full recovery of affected endpoint within 8 weeks post last application. Note that the endpoint 3497 can be considered to be recovered only if the MDD during the relevant recover period was at 3498 least smaller than 100%. If this is not the case an appropriate higher class have to be selected. 3499

3500 Effect class 4 (Pronounced effect in short-term study). 3501

Clear effects (e.g. large reductions in densities of the population) observed, but the study is too 3502 short to demonstrate complete recovery within 8 weeks after the (last) application. 3503

3504 Effect class 5A (Pronounced long-term effect followed by recovery). 3505

Clear response of sensitive endpoint, effect period longer than 8 weeks and recovery did not 3506 yet occur within 8 weeks after the last application, but full recovery is demonstrated to occur 3507 in the year of application. Note that the endpoint can be considered to be recovered only if the 3508 MDD during the relevant recover period was at least smaller than 100%. If this is not the case 3509 an appropriate higher class have to be selected. 3510

3511 Effect class 5B (Pronounced long-term effects without recovery). 3512

Clear response of sensitive endpoints (> 8 weeks post last application) and full recovery 3513 cannot be demonstrated before termination of the experiment or before the start of the winter 3514 period. 3515

3516

7.3.4. Variability in concentration-response patterns between micro/mesocosm experiments 3517 exposed to the same PPP 3518

7.3.4.1. Short-term pulsed exposure 3519

For the interpretation of micro-/mesocosm experiments an important question at stake is whether 3520 concentration-response relationships are reproducible. Effect classes 1 and 2 concentration of the most 3521 sensitive measurement endpoints in the micro-/mesocosm experiment may be used as estimates of the 3522 ecological threshold concentrations of PPPs (not considering ecological recovery) while an Effect 3523 class 3A concentration of the most sensitive measurement endpoints, may be used as the study-3524 specific NOEAEC (No Observed Ecologically Adverse Effect Concentration), an estimate that may 3525 take into account ecological recovery. For a few test substances only at least 3 appropriately 3526 preformed micro-/mesocosm experiments are available when considering the criteria described in 3527



section 7.3.3 and a similar exposure regime. The data presented in Appendix E for the insecticides 3528 chlorpyrifos, lambda-cyhalothrin and esfenvalerate are representative for (short-term) pulsed exposure 3529 regimes. It seems that for these insecticides Effect class 1 - 2 concentrations of the most sensitive 3530 measurement endpoints derived from different micro-/mesocosm experiments show lower variability 3531 than higher Effect classes. Note that the chlorpyrifos, lambda-cyhalothrin and esfenvalerate studies 3532 comprised test systems that considerably varied in dimensions and complexity of community structure 3533 (plankton-dominated; macrophyte-dominated; lentic; lotic). but always contained several dominant 3534 populations of arthropods (but not always well-established populations of insects). Nevertheless, the 3535 comparison of micro-/mesocosm experiments performed with these insecticides suggests that a small 3536 AF may be sufficient to extrapolate Effect class 1 and/or Effect class 2 concentration of the most 3537 sensitive measurement endpoints derived from a well-performed micro-/mesocosm study with a well-3538 defined exposure regime. In addition, for the same PPP and a similar exposure regime these Effect 3539 class 1 and Effect class 2 concentrations do not overlap with the range of concentrations for higher 3540 effect classes (Effect classes 3 to 5). 3541

If an Effect-class 3A concentration (of most sensitive measurement endpoints) for short-term 3542 exposures is considered acceptable, it appears from the data presented in Appendix E that for 3543 chlorpyrifos and lambda-cyhalothrin an AF of 3 may be necessary to cover the variability in observed 3544 concentration-response patterns that include ecological recovery, if a single high quality micro-3545 /mesocosm experiment is available. Also in the case of esfenvalerate, applying an AF of 3 to the 3546 Effect class 3A concentration overall avoids the occurrence of unacceptable class 4-5 effects caused 3547 by pulsed exposures in hydrologically closed systems (lentic micro-/mesocosms or recirculating 3548 experimental streams) (Tables E1.1 to E1.3 in Appendix E). 3549

In accordance with the data for chlorpyrifos, lambda-cyhalothrin and esfenvalerate described in 3550 Appendix E, in lake enclosure studies exploring effects of a single application of pentachlorophenol to 3551 plankton communities in spring, summer, autumn and winter, a low variability in threshold levels for 3552 effects (Effect class 1 concentrations based on peak exposure) was observed. In these lake enclosure 3553 experiments (n = 4) the variability in ecological threshold concentration varied by approximately a 3554 factor of 2 (Willis et al. 2004). 3555

7.3.4.2. Long-term exposure to the same PPP 3556

Again, considering the criteria mentioned in section 7.3, for a few PPPs only three or more appropriate 3557 micro-/mesocosm studies are available mimicking a more or less constant chronic exposure regime. In 3558 lentic test systems the treatment-related responses caused by a long-term chronic exposure regime to 3559 the fungicide carbendazim resulted in similar Effect class 1 concentrations, suggesting little variability 3560 in threshold levels for effects between studies (Table E1.4 in Appendix E). However, long-term 3561 exposure studies with the herbicide atrazine (Table E1.5 in Appendix E) revealed a considerable 3562 overlap between Effect class 1 and Effect class 2 concentrations of the most sensitive measurement 3563 endpoints. In addition, an overlap between Effect class 2 and Effect class 3-5 concentrations was 3564 observed as well for atrazine. As explained in Appendix E, differences in concentration-response 3565 patterns between studies performed with the photosynthesis inhibiting herbicide atrazine might be 3566 explained by differences in light conditions between indoor and outdoor studies presented in Table 3567 E1.5. Nevertheless, if we consider the atrazine data representative for chronic exposure regimes of 3568 other photosynthesis-inhibiting herbicides, and from a regulatory point of view an Effect class 2 3569 response is acceptable as an estimate that approaches the threshold level of effects, an AF of 2 to 3 3570 seems to be necessary to address the variability in concentration-response patterns between well-3571 performed model ecosystem experiment mimicking a chronic exposure regime. Applying an AF of 2 3572 to 3 to Effect class 2 concentrations presented in Table E.5 (see Appendix E) will, with a high 3573 probability, avoid unacceptable class 3 to 5 effects caused by long-term exposure. 3574

3575



7.3.5. How to derive a RAC from an appropriate micro-/mesocosm experiment and how to 3576 link it to PEC 3577

Figure 7.4 and Tables 7.5 to 7.8 present proposals for the derivation of the RACs within acute 3578 (RACsw;ac) and chronic (RACsw;ch) effect assessment schemes on basis of appropriate micro-3579 /mesocosm experiments. A distinction is made in RACs derived on basis of the ecological threshold 3580 option (ETO-RAC) and ecological recovery option (ERO-RAC). 3581

3582

Figure 7.4: Decision scheme for the derivation of RACs from appropriate micro-/mesocosm 3583 experiments on basis of the ecological threshold option (ETO-RAC) or ecological recovery option 3584 (ERO-RAC). 3585



7.3.5.1. Selecting and extrapolating micro-/mesocosm results 3586

For Regulatory Acceptable Concentration (RAC) derivation only those micro-/mesocosm studies 3587 should be selected that largely meet the criteria mentioned in section 7.3.2 and 7.3.3. The RAC may be 3588 derived by applying an AF (for spatio-temporal extrapolation) to the study-specific NOEC (ecological 3589 threshold option) or the NOEAEC that takes into account ecological recovery (ecological recovery 3590 option). The height of this AF should address the model ecosystem to field ecosystem extrapolation. 3591 Edge-of-field surface waters in Europe show a large variability in ecosystem structure and functioning 3592 and a specific microcosm or mesocosm experiment only mimics one of the possible field assemblages. 3593 Within this context it is worthwhile mentioning that most insight in the variability in concentration–3594 response relationships for PPPs is available for lentic communities in micro-/mesocosms. 3595

Addressing the uncertainty with respect to the model ecosystem – field extrapolation for the threshold 3596 level of effects, amongst others (see criteria in section 7.3.3), depend on the relevance of the tested 3597 assemblages for the sensitivity (and vulnerability if the recovery option is selected) of species that 3598 occur in the type of edge-of-field surface water potentially at risk. In addition, other higher-tier 3599 information available (e.g. laboratory toxicity data for additional test species and other micro-3600 /mesocosm experiments) may address this uncertainty. Usually more ecotoxicological data are 3601 available for species occurring in lentic edge-of-field surface waters (ponds, drainage ditches) than for 3602 species typical for edge-of-field streams. For example, the available toxicity data for insecticides and 3603 Ephemeroptera, Trichoptera and Plecoptera may be scarce, while these taxonomical groups often are 3604 more abundant in lotic than in lentic surface waters. If there are reasons to assume that e.g. a lentic 3605 micro-/mesocosm does not sufficiently represent sensitive taxa typical for streams, and these taxa are 3606 likely more sensitive than lentic taxa, this may be a reason to adopt the higher AF in the range 3607 proposed. The same type of reasoning may be valid when extrapolating lotic data to ponds or drainage 3608 ditches when taxonomical groups are sensitive that usually are more abundant in lentic waters (e.g. 3609 floating macrophytes; chironomids, copepods). Note however that Maltby et al. (2005) assessed the 3610 influence of lentic versus lotic habitat on the species sensitivity distributions of arthropods to 8 3611 insecticides and that there was no consistent pattern in the relative sensitivity between lentic and lotic 3612 species. 3613

7.3.5.2. Peak, nominal or TWA concentrations of RAC and PEC used for risk assessment 3614

If the study is triggered by the Tier 1 acute core data and the duration of the pulse exposure in the 3615 micro-/mesocosm experiment appears to be shorter than that predicted for the field, two options might 3616 be explored. First, express the treatment-related responses (short-term exposure effect assessment) in 3617 terms of the initial 48-96 h time weighted average concentration (instead of the measured peak 3618 concentration) as measured in the test systems and compare the final RACsw;ac-estimate to the PECmax. 3619 The duration of the 48-96 h is selected since in the first tier acute effect assessment this time frame 3620 corresponds with the duration of most standard acute tests. In addition, a similar procedure is proposed 3621 to derive a MAC-QS from a micro-/mesocosm test within the context of the Water Framework 3622 Directive (EC, 2011). In this case the TWA approach should not be used for the short-term exposure 3623 assessment, because then the worst-case assumption of the approach is violated, instead the PECsw,max 3624 should be used. Second, decide that the micro-/mesocosm experiment cannot be used in the higher-3625 tier effect assessment for the field exposure-regime under evaluation. Another promising option is to 3626 explore adequate methods to extrapolate concentration-response relationships for shorter pulse 3627 exposures to that of broader ones (e.g. promising may be the use of toxicokinetic/toxicodynamic 3628 models for relevant sensitive organisms), however, no detailed guidance is included here for the time 3629 being. This might be updated based on a future activity of the PPR Panel on aquatic effect modelling. 3630

To evaluate chronic risks (triggered by the Tier 1 chronic core data) either the peak concentration or a 3631 Time Weighted Average (TWA) concentration of the PPP in the relevant matrix (water, sediment) 3632 may be used as estimate of RACsw;ch and/or as PEC estimate (see chapter 2). If the TWA approach is 3633 appropriate (for criteria see chapter 2) and used for the effects assessment, the selected TWA time-3634 window should coincide with the application period of the test substance. For both RAC and PEC 3635 estimation the selection of the length of the TWA time-window should be based on ecotoxicological 3636



considerations (e.g. A/C ratio; time-to-onset-of-effect information; length of the most sensitive life 3637 stage of the organisms at risk) and guided by the length of the relevant chronic toxicity tests that 3638 triggered the micro-/mesocosm experiment. If the TWA approach is considered appropriate (see for 3639 criteria chapter 2) we propose to adopt a default time-window of 7 days for the TWA estimate of the 3640 long-term PEC (= PECsw;7d-twa), at least if no scientific arguments are provided to shorten or lengthen 3641 this default time window. This proposal follows the recommendation of the ELINK workshop. Note 3642 that for a worst-case approach the time-window for the TWA Effect class concentration estimate in the 3643 micro/mesocosm study should not be smaller than the selected TWA time-window for the PEC 3644 estimate in the field. In addition, the time-window for the TWA Effect class concentration estimate in 3645 the micro/mesocosm experiment should not be larger than the period in which the exposure remains 3646 more or less constant (remaining within 20% deviation from nominal concentration) or, in case of a 3647 relatively fast dissipating substance, the application period of the PPP in the micro-/mesocosm study. 3648 The application period is the period in which repeated pulse applications occur. When e.g. a 7-d time-3649 window is adopted for the PEC, the ‘Effect class’ concentrations derived from a micro-/mesocosm 3650 experiment characterised by 3 weekly treatments can be expressed in terms of a TWA concentration 3651 that is ≥ 7 days and ≤ 21 days if in the test systems the PPP is not very persistent. 3652

In case the TWA approach is deemed not to be appropriate in the chronic risk assessment, and 3653 consequently the PECsw;max is used as field exposure estimate, the ‘Effect class’ concentrations derived 3654 from a mesocosm experiment simulating long-term exposure may be expressed in terms of the 3655 nominal, peak or average concentration measured/calculated during the application period (or the 3656 period in which the exposure remains more or less constant in the micro/mesocosm test). Adopting the 3657 nominal or measured/calculated peak concentration only is justified if it can be demonstrated that the 3658 exposure profile in the micro-/mesocosm experiment overall is realistic to worst-case from that in the 3659 relevant field scenario(s). In that case, and if it was demonstrated that the concentration builds up due 3660 to repeated treatments, adopting the nominal concentration during the application period can be 3661 considered as a more conservative approach than adopting the measured/predicted peak concentration. 3662

If the study is triggered by the Tier 1 chronic core data, and the duration of the pulse exposure in the 3663 micro-/mesocosm experiment appears to be shorter than that predicted for the field the same options as 3664 described above may be explored, except that when using the second option it may be necessary to 3665 lengthen the time-window for the TWA concentration to express the concentration-response 3666 relationships observed. The selection of this TWA time-window may be guided by the duration of the 3667 standard chronic toxicity test that triggered the risk. 3668

7.3.5.3. Deriving a RAC indicative for the ecological threshold option (ETO-RAC) 3669

Under the assumptions that (i) the sensitivity of the assemblages in appropriate micro-/mesocosm tests 3670 are representative for those in edge-of-field surface waters and (ii) that the observed variability in 3671 threshold concentrations for effects between micro-/mesocosm tests with the substances chlorpyrifos, 3672 lambda-cyhalothrin, esfenvalerate, pentachlorophenol, carbendazim and atrazine (see section 7.3.4) is 3673 generally valid for PPPs, the AF to address variability in Effect class 1 – 2 concentrations between 3674 freshwater communities that contain sensitive populations potentially at risk may be small. 3675

Table 7.5 presents a proposal for the derivation of the RACsw;ac (triggered by the Tier 1 acute core 3676 data) addressing the ecological threshold option for edge-of-field surface waters on basis of an 3677 appropriate micro/mesocosm experiment. The assessment factors presented in the table below are 3678 proposed for studies that are appropriately designed to address the risk and uncertainty identified at 3679 lower tiers, e.g. studies where a sufficiently low MDD was obtained for adequate number of species 3680 (see section 7.3.2). If this is not the case the AF needs to be adjusted. 3681

3682

3683



Table 7.5: Proposal for the derivation of the RACsw;ac (triggered by Tier 1 acute core data) addressing 3684 the ecological threshold option on basis of an appropriate micro/mesocosm experiment. If in the same 3685 study several treatments resulted in the same ‘Effect class’- response of sensitive measurement 3686 endpoints the highest concentration within the same Effect class is selected. 3687

Assessment factor for ETO-RACsw;ac derivation (ecological threshold option)

Field exposure concentration to compare with the RACsw;ac

Effect class 1 concentration Is rate of dissipation of the active ingredient in test system realistic to worst-case when compared to that predicted for the field? Yes: Base effect estimate on nominal or measured peak concentration in test system. No: Base effect estimate on e.g. the initial 48h TWA concentration in test system or apply appropriate extrapolation techniques

(1 – ) 2(a) Based on expert judgement by considering the criteria mentioned in section 7.2 and 7.3 and ecological information on the type of edge-of-field surface water at risk

PECsw;max


2 – 3(a) Based on expert judgement by considering the criteria mentioned in section 7.3 and ecological information on the type of edge-of-field surface water at risk

PECsw;max

(a) If several adequate micro-/mesocosm studies or other adequate higher tier studies (e.g. monitoring, relevant population 3688 experiments or modelling) are available a lower AF may be applied or the AF is applied to the highest value. The definite 3689 choice of the AF is influenced by the quality of the micro/mesocosm study and the related uncertainties. 3690

3691 3692 Table 7.6 presents a proposal for the derivation of the RACsw;ch (triggered by the Tier 1 chronic core 3693 data) addressing the ecological threshold option for edge-of-field surface waters on basis of 3694 appropriate micro/mesocosm experiments. 3695 3696

Table 7.6: Proposal for the derivation of the RACsw;ch (triggered by Tier 1 chronic core data) 3697 addressing the ecological threshold option on basis of an appropriate micro/mesocosm experiment. If 3698 in the same study several treatments resulted in the same ‘Effect class’- response of sensitive 3699 measurement endpoints the highest concentration within the same Effect class is selected. 3700

Assessment factor for ETO-RACsw;ch derivation (ecological threshold option)

Field exposure concentration to compare with the RACsw;ch

Effect class 1 concentration Based on time weighted average concentration in test system during the application period.

(1 –) 2(a) Based on expert judgement by considering the criteria mentioned in section 7.3 and ecological information on the type of edge-of-field surface


Based on expert judgement by considering the criteria mentioned in chapter 2



water at risk Effect class 1 concentration Based on nominal or peak concentration in test system if the long-term exposure regime (e.g. due to repeated pulses) is realistic to worst-case compared to the predicted field exposure profile

(1 –) 2(a) Based on expert judgement by considering the criteria mentioned in section 7.3 and ecological information on the type of edge-of-field surface water at risk

PECsw;max





Effect class 2 concentration Based on nominal or peak concentration in test system if the long-term exposure regime (e.g. due to repeated pulses) is realistic to worst-case compared to the predicted field exposure profile


PECsw;max

(a) If several adequate micro/mesocosm studies or other adequate higher tier studies (e.g. monitoring, relevant population 3701 experiments or modelling) are available a lower AF may be applied or the AF is applied to the highest value. The definite 3702 choice of the AF is influenced by the quality of the micro/mesocosm study and the related uncertainties. 3703

3704

7.3.5.4. Deriving a RAC on basis of ecological recovery option (ERO-RAC) 3705

Not exceeding the specific protection goals for aquatic key drivers described in chapter 3, Effect class 3706 3A concentrations from appropriate micro-/mesocosm tests might be used to derive a RAC in line with 3707 the recovery option. 3708

Again, under the assumptions that (i) the sensitivity of the assemblages in appropriate micro-3709 /mesocosm tests are representative for those in edge-of-field surface waters and (ii) that the observed 3710 variability in threshold concentrations for effects between micro-/mesocosm tests with the substances 3711 chlorpyrifos, lambda-cyhalothrin, esfenvalerate, pentachlorophenol, carbendazim and atrazine (see 3712 section 7.3.4) is generally valid for PPPs, the AF to address the extrapolation of an Effect class 3A 3713 concentration from a micro/mesocosm test system to the field needs to be larger than that to derive the 3714 RAC representative for the ecological threshold option (see section 7.3.4). 3715

Tables 7.7 and 7.8 present proposals for respectively the derivation of the RACsw;ac (triggered by 3716 Tier 1 acute core data) and the RACsw;ch (triggered by Tier 1 chronic core data) addressing the 3717 ecological recovery option for edge-of-field surface waters on basis of appropriate an micro/mesocosm 3718 experiment. 3719

When using an Effect class 3A concentration to derive the RAC special attention should be paid to the 3720 representativeness of the enclosed community in the test system for potentially sensitive invertebrate 3721 populations with a complex uni-/semivoltine life cycle and limited dispersal abilities and/or for 3722 potentially sensitive macrophytes with a relatively slow growth rate. If these populations are 3723 insufficiently represented in the test system additional information (e.g. additional population studies) 3724 and/or extrapolation techniques (e.g. population models) may be required, or simply the RAC 3725 indicative for the ecological threshold level (ETO-RAC) needs to be used in the risk assessment. 3726



Table 7.7: Proposal for the derivation of the RACsw;ac (triggered by Tier 1 acute core data) addressing 3727 the ecological recovery option on basis of an appropriate micro/mesocosm experiment. If in the same 3728 study several treatments resulted in the same ‘Effect class’- response of sensitive endpoints the highest 3729 concentration within the same Effect class is selected. 3730

Assessment factor for ERO-RACsw;ac derivation (ecological recovery option)


Effect class 3A concentration Maximum magnitude of temporal effects may be medium to large Is rate of dissipation of the active ingredient in test system realistic to worst-case when compared to that predicted for the field? Yes: Base effect estimate on nominal or measured peak concentration in test system. No: Base effect estimate on e.g. the initial 48h TWA concentration in test system or, apply appropriate extrapolation techniques or, consider the ecological threshold option (Table 7.1)


PECsw;max

(a) If several adequate micro-/mesocosm studies or other adequate higher tier studies (e.g. monitoring, relevant 3731 population experiments or modelling) are available a lower AF may be applied or the AF is applied to the 3732 highest value. The definite choice of the AF is influenced by the quality of the micro/mesocosm study and 3733 the related uncertainties. 3734

3735

Table 7.8: Proposal for the derivation of the RACsw;ch (triggered by Tier 1 chronic core data) 3736 addressing the ecological recovery option on basis of an appropriate micro/mesocosm experiment. If 3737 in the same study several treatments resulted in the same ‘Effect class’- response of sensitive 3738 measurement endpoints the highest concentration within the same Effect class is selected. 3739

Assessment factor for ERO-RACsw;ch derivation (ecological recovery option)

Field exposure concentration to compare with the RACsw,ch

Effect class 3A concentration Based on time weighted average concentration in test system during the application period.




Effect class 3A concentration Based on nominal or peak concentration in test system if the long-term exposure regime (e.g. due to repeated pulses) is realistic to worst-case compared to the predicted field exposure profile


PECsw;max

(a) If several adequate micro/mesocosm studies or other adequate higher tier studies (e.g. monitoring, relevant 3740 population experiments or modelling) are available a lower AF may be applied or the AF is applied to the 3741 highest value. The definite choice of the AF is influenced by the quality of the micro/mesocosm study and 3742 the related uncertainties. 3743



3744 Finally, according to ELINK (Brock et al. 2010c) if the derived RAC value from a single- or multiple-3745 application micro-/mesocosm experiment is based on the ecological recovery option (=ERO-RAC), an 3746 appropriate risk assessment can only be performed by plotting this ERO-RAC and the RAC indicative 3747 for the ecological threshold level for effects (= ETO-RAC) on the predicted field exposure profile (see 3748 Figure 7.3). If in the appropriate edge-of-field scenario the pulses are lower than the ERO-RAC value 3749 based on Effect class 3A concentration but higher than the ETO-RAC, the interval between successive 3750 pulses should be carefully considered. If the interval between pulses is smaller than the relevant 3751 recovery time of the sensitive populations of concern, these pulses may be considered as ecologically 3752 dependent. On the basis of this information, the total period of possible effects can be estimated and 3753 considered in the final risk assessment (whether the recovery option can or cannot be used, Figure 3754 7.4). 3755

3756



8. Non-testing methods, metabolites, impurities and formulations with more than one 3757 active substance 3758

8.1. Non-testing methods 3759

This section provides guidance on use of non-testing methods in PPP risk assessment, such as (Q)SAR 3760 (Quantitative Structure-Activity Relationship), expert test systems and analogue read-across as tools 3761 for deriving intrinsic properties of chemicals. All of these approaches have a role in extending and 3762 supplementing the existing information and hereby minimizing the need for additional testing, in 3763 particular with vertebrates (see section 9.4). In addition, non-test methods are important tools for 3764 prioritising chemicals for further consideration, assessment and / or testing and in the planning of 3765 further testing. The development and application of all kinds of non-testing methods is based on the 3766 similarity principle, i.e. hypothesis that similar compounds should have similar biological activities 3767 and the methods may therefore in some cases provide predictions that are so reliable that they can be 3768 used to substitute experimental data for several types of hazard related endpoints, e.g. mortality, 3769 reproduction or endocrine effects. In this regard it should be noted that a positive prediction made by 3770 application of a non-test method for an effect (e.g. reproductive toxicity) may be accepted for use to 3771 avoid further testing while caution should be applied with negative predictions (i.e. lack of effect) 3772 since in most cases not all Modes of Action or mechanisms are covered by the existing non test 3773 method. 3774

When using (Q)SARs, it should be remembered that (Q)SARs are models and are therefore inevitably 3775 associated with a degree of uncertainty. This uncertainty is caused predominantly by two different 3776 reasons: a) the inherent variability of the input data used to establish and validate the QSAR model; 3777 and b) the uncertainty resulting from the fact that a model can only be a partial representation of 3778 reality (in other words does generally not model all possible modes of action or mechanisms and hence 3779 does not represent all types of chemicals). It is noteworthy that these two types of uncertainty is 3780 related to the validation and the applicability domain of the (Q)SAR model respectively. Despite these 3781 uncertainties, it it also noted that a (Q)SAR is not only an empirical model, but that it is associated 3782 with an underlying dataset used to establish and validate the model, a description of the modelled 3783 endpoint, the descriptors and the statistical methods used as well as a characterization of the 3784 applicability domain as well as any appropriate mechanistic understanding of the model. As a 3785 representation of the training dataset for the model, it averages the uncertainty over all chemicals. 3786 Thus, if the models makes reliable predictions within its applicability domain an individual model 3787 estimate will be more accurate than an individual measurement obtained by performing the relevant 3788 test. 3789

8.1.1. Area of use 3790

Use of alternative ‘in silico’ methods such as qualitative or quantitative structure-activity relationship 3791 models (Q)SARs or read across may be used on its own or in combination with expert systems (see 3792 below) as non-testing methods to provide valid endpoints for hazard and risk assessment. However, 3793 often more robust estimates can be generated by using weight-of-evidence approaches where all 3794 available information is taken into account. This could include a combination of (Q)SAR predictions 3795 for the same endpoint by different model systems combined with read across and other available 3796 information like non-standard testing data and toxicodynamic/ toxicokinetic information from 3797 mammals (OECD, 2007a, OECD, 2009 and NAFTA, 201116). 3798

Data provided by non-testing methods shall not be used to substitute experimental data necessary to 3799 fulfil the data requirements (SANCO document 11802 and 11803/2012). However, there may be 3800 situations where non-testing methods can be used to address needs for information, rather than 3801 deriving new experimental data. Situations were non-testing methods may be used on a more regular 3802

16http://www.epa.gov/oppfead1/international/naftatwg/



basis are for metabolites without the toxophore (see 8.2.5) and for impurities17. In addition, QSARs 3803 might together with available test data be used to rank species for identifying the most likely sensitive 3804 taxonomic group to focus experimental testing (EFSA, 2012a). 3805

8.1.2. Guidance on (Q)SAR 3806

The guidance provided here on use of (Q)SAR is specified to the use in relation to PPPs. Guidance on 3807 the validity of (Q)SAR models and reliability and adequacy of (Q)SAR predictions in general can be 3808 found in the ECHA report: Guidance on information requirements and chemical safety assessment 3809 Chapter R.6: QSARs and grouping of chemicals (ECHA, 2008). Only a short summary of the ECHA 3810 guidance is given below on: 3811

• How to establish the validity of (Q)SAR models and how to assess the reliability and 3812 adequacy of (Q)SAR predictions. 3813

• How to document and justify the use of a QSAR model and where do find information on 3814 QSAR models. 3815 3816

More specific guidance on use of (Q)SAR in PPP risk assessment can also be found in the recently 3817 published Guidance Document by the North American Free Trade Agreement18 (NAFTA) Technical 3818 Working Group on PPPs (TWG) (NAFTA, 2012). 3819

It is noted that the field of computational toxicity (including (Q)SAR) is rapidly developing, and 3820 experience in the regulatory use of computational approaches (including their reporting) is increasing., 3821 This guidance document should be considered as a step in a continuously evolving process. 3822

Reporting of validity assessment should follow the OECD QSAR validation format “QSAR Model 3823 Reporting Format” (QMRF). Likewise, the specific prediction should be reported in a “QSAR 3824 Prediction Reporting Format” (QPRF) (see ECHA, 2008). 3825

8.1.2.1. Model validity 3826

QSAR models are not formally validated as is the case for OECD test guidelines. Instead the OECD 3827 has established five internationally agreed principles that can be used to assess the validity of a QSAR 3828 model prediction for a given purpose (OECD 2007a): 3829

Principle 1. A (Q)SAR model should be associated with a defined (measurable) endpoint and the 3830 related experimental protocols. 3831

Principle 2. An unambiguous algorithm is to ensure transparency in the model algorithm that 3832 generates predictions of an endpoint from information on chemical structure and/or 3833 physicochemical properties. 3834

Principle 3. A defined chemical domain of applicability of a (Q)SAR model, for which reliable 3835 predictions can be generated. 3836

Principle 4. Appropriate measures of goodness-of–fit, robustness (internal performance) and 3837 predictivity (determined by external validation). 3838

Principle 5. A mechanistic interpretation of the descriptors used in a model and the endpoint being 3839 predicted (if possible). 3840

3841

17Any component other than the pure active substance and/or variant which is present in the technical material (including

components originating from the manufacturing process or from degradation during storage) [Art. 3 (33) of Regulation (EC) No 1107/2009].

18Formalized framework for partnership between United States Environmental Protection Agency Office of Pesticide Programs (US EPA OPP) and the Pest Management Regulatory Agency (PMRA) of Health Canada to develop common approaches to Integrated Approaches to Testing and Assessment (IATA) for the human health and ecological risk assessment of pesticides.



8.1.2.2. Reliability and adequacy of (Q)SAR prediction 3842

The determination of whether a (Q)SAR result may be used to replace experimental testing can be 3843 broken down into three main steps: 3844

1. an evaluation of the scientific validity (relevance and reliability) of the model 3845 2. an assessment of the applicability of the model to the chemical of interest and the reliability of the 3846

individual model prediction 3847 3. an assessment of the adequacy of the information for making the regulatory decision, including an 3848

assessment of completeness, i.e. whether the information is sufficient to make the regulatory 3849 decision, and if not, what additional (experimental) information is needed. 3850

3851 To be used as a full replacement of an experimental test, all three conditions need to be fulfilled. In 3852 cases where some information elements are missing, (Q)SAR results may still be used in the context 3853 of a weight of evidence approach. 3854

Step 1: 3855 An assessment of the model relevance and reliability follows the five principles mentioned above. 3856 It is noted that there is no unique measure of model reliability. In general, model reliability should 3857 be regarded as a relative concept, depending on the context in which the model is applied. In other 3858 words, a greater or lesser degree of reliability may be sufficient for a given regulatory application. 3859 3860 When evaluating the performance (fitting and external prediction) of QSAR models, several 3861 validation parameters exist, e.g. Predictive Squared Correlation Coefficient (Q2) (Shi et al. (2001), 3862 Schüürmann et al. (2008), Consonni et al. (2009)) and average correlation coefficient r2. Recently 3863 the Concordance Correlation Coefficient (CCC) proposed by Chirico (2011) has been compared 3864 with commonly acceptance thresholds (Q2 = 0.6, average r2 = 0.5). The CCC (Lin, 1989) is similar 3865 to the correlation coefficient (linear alignment), but, in addition, it takes into account the closeness 3866 to the diagonal (perfect match). A CCC threshold value of 0.85 is generally the most restrictive in 3867 the acceptance of QSAR model estimates (Chirico et al., 2012). This indicates that CCC is a 3868 validation parameter in a precautionary approach for assessing accurate predictions. It is noted 3869 however that any validation should always include visual inspection of the experimental versus 3870 predicted plot, in order not to overlooks biases in data set (e.g. location shift and scale shift). 3871 3872

Step 2: 3873 Assessment of model validity is a necessary but not sufficient step in assessing the acceptability of 3874 a QSAR result. Assuming that the model is considered valid, the second and crucial step is to 3875 evaluate the reliability of prediction for a specific compound. The main questions to address are: i) 3876 is the chemical of interest within the scope of the model, according to the defined applicability 3877 domain of the model? ii) is the defined applicability domain suitable for the regulatory 3878 purpose? iii) how well does the model predict chemicals that are similar to the chemical of 3879 interest? iv) is the model estimate reasonable, taking into account other information? 3880

Step 3: 3881 Experience with use of QSAR data in a regulatory context is relatively limited compared with 3882 acceptance of test data (including data on laboratory animals). In a regulatory context, experience 3883 in the regulatory use of non-testing data have often been obtained by following a learning-by-3884 doing approach19, with the learning being documented in draft assessment reports and / or 3885 guidelines/ background documents for the particular regulatory area20. Only limited guidance on 3886 the acceptance of (Q)SARs can be given at this moment. However, three important principles 3887 could be outlined here, as already explained in the TAPIR report (ECB, 2005): i) the principle of 3888 proportionality express the relationship between the amount of information needed and the 3889

19E.g. US or OECD High Production Volume Chemicals (HPVC) programmes and to a lesser extent in the former EU ESR

Programme on existing industrial HPVC 20Experience should be compiled in the next update of this Aquatic Guidance Document



severity of the decision; ii) the principle of caution (or conservativeness) expresses the relationship 3890 between the amount of information needed and the (likely) consequence(s) of the decision based 3891 on that information being wrong; iii) the level of confidence and precision of a non-testing 3892 prediction is higher when the predicted value is close to a regulatory cut off value / decision point 3893 that when it is clearly far away for the cut off. It is noted that the same applies for test data and 3894 that this issue is not dependent of the state of science but rather a consequence of human decision 3895 making systems. 3896 A consequence of these principles, which also applies to test data, is that the relationship between 3897 scientific validity and the reliability of information and hence its regulatory acceptability should 3898 not be regarded as a constant relationship, but a relationship which varies according to the 3899 decision being made and the particular circumstances involved in individual cases. 3900

3901 A formal adoption of (Q)SAR models or other non-testing methods are not foreseen under REACH 3902 (i.e. no official, legally binding list of (Q)SAR methods). Instead, acceptance under REACH will 3903 involve initial acceptance by industry and subsequent evaluation by the authorities, on a case-by-case 3904 basis. The same will be the case for PPP registration. Use of QSARs in such regulatory contexts is also 3905 a learning process, therefore the PPR Panel recommends that a harmonised approach between EU 3906 Member States and different regulatory frameworks is developed. 3907

8.1.3. Available (Q)SAR methods, expert systems and read across 3908

Many (Q)SAR models that estimate the toxicity to aquatic organisms are available, e.g. ECOSAR (US 3909 EPA)21, The Danish (Q)SAR Database22, DEMETRA (EU)23, TOPCAT, (Q)SAR Application Toolbox 3910 (OECD) and/or expert systems proposed, e.g. Escher et al. (2006). 3911

ECOSAR (Ecological Structure Activity Relationships) is a freely available software system (US EPA 3912 2008a) which matches the structure of a query organic substance to one (or more) of its defined 3913 chemical class(es). For most classes, aquatic ecotoxicity values are predicted using available linear 3914 correlations between toxicity and hydrophobicity. ECOSAR predicts acute (short-term) toxicity and 3915 chronic (long-term or delayed) toxicity to aquatic organisms such as fish, aquatic invertebrates, and 3916 aquatic plants. If not available experimentally Kow is estimated for the query molecule using 3917 KOWWIN. In 2012 ECOSAR included 111 organic chemical classes in three main groups, of which 3918 the group ‘Organic Chemicals with Excess Toxicity’ contains several PPP classes (e.g. carbamates, 3919 imidazoles, neonicotinoids, pyrethroids, sulfonyl ureas, triazines). 3920

OECD (Q)SAR Application Toolbox 3921 The Toolbox24 is a software application intended to be used to fill gaps in toxicity and ecotoxicity 3922 data, which are needed for assessing the hazards of chemicals. The Toolbox incorporates databases on 3923 chemical data (e.g. properties), experimental toxicological and ecotoxicological data and estimated 3924 values from a large range of QSAR tools, together with incorporated QSAR modelling and Expert 3925 Systems, built within a regulatory application chassis. This package therefore allows the user to 3926 perform a number of functions: 3927

• Identify analogues for a chemical, retrieve experimental results available for those analogues 3928 and fill data gaps by read-across or trend analysis; 3929

21ECOSAR (Ecological Structure Activity Relationships) is a freely available piece of software that can be downloaded from

the US EPA website (US EPA 2008a) 22A database containing (Q)SAR predictions for the environment and human health for 180.000 chemicals

(http://130.226.165.14/index.html) 23This software tool for the ecotoxicity prediction of pesticides and metabolites was developed as part of an EC funded

project named DEMETRA (http://www.DEMETRA-tox.net). 24 The most recent OECD (Q)SAR Application Toolbox (version 3) was launched in October 2012 at:

http://www.oecd.org/env/chemicalsafetyandbiosafety/assessmentofchemicals/theoecdqsartoolbox.htm#Download_qsar_application_toolbox



• Categorise large inventories of chemicals according to intrinsic chemical properties 3930 (“profilers) related to e.g. physical chemical properties, chemical reactivity related to various 3931 mechanisms or modes of action; 3932

• Functionalities for assessment of metabolites of chemicals even though it does not also 3933 contain probability estimates or toxicokinetic information/predictions of those metabolites. 3934

• Fill data gaps for any chemical by using the library of QSAR models; 3935 • Evaluate the robustness of a potential analogue for read-across; 3936 • Evaluate the appropriateness of a (Q)SAR model for filling a data gap for a particular target 3937

chemical; and 3938 • Build QSAR models. 3939 • Functionalities by which documentation of then performed analysis can be provided 3940

(combination of automated reporting which can be manually improved/ detailed) 3941 3942 The Danish (Q)SAR database 3943 This database is freely available on the Internet and contains (Q)SAR predictions from over 70 3944 (Q)SAR models for approximately 180,000 chemicals. The (Q)SAR models encompass endpoints for 3945 physicochemical properties, fate, eco-toxicity, absorption, metabolism and toxicity. The majority of 3946 estimates are from models developed for mammalian (human) toxicity endpoints in particular 3947 MULTICASE (Multiple Computer Automated Structure Evaluation, Multicase Inc, Ohio, USA). 3948 These predictions also contains simple statements in respect to whether or not the individual prediction 3949 is within the structural applicability domain of the model (yes/no). Estimates for many of the 3950 environmental properties come from the Epiwin software developed by Syracuse Research Co-3951 operation on behalf of the US EPA. However, the acute toxicity models for the environment derive 3952 from models developed by the Danish EPA. Estimates from a few literature based-models are also 3953 included. 3954

3955 DEMETRA 3956 This software tool for the ecotoxicity prediction of PPPs and metabolites was developed as part of an 3957 EC funded project named DEMETRA (http://www.DEMETRA-tox.net). The programme allows 3958 prediction of PPP toxicity in fish, daphnids, bee and quail (oral and dietary exposure) and incorporates 3959 predictive models for five specific endpoints, with each hybrid combinative model incorporating an 3960 intelligent integration of several individual validated QSAR models. The models and software were 3961 developed with the aim of regulatory use and developed according to strict quality criteria according 3962 to OECD guidelines, using only experimental data produced according to official guidelines and 3963 validating using external test sets. The models are applicable to PPPs (and metabolites/impurities) in 3964 general and not specific chemical classes. The predictive models within DEMETRA are a hybrid 3965 combination of two or more individual models, therefore minimum and maximum values are also 3966 computed based on the minimum or maximum predicted values of the individual models (these values 3967 do not refer to the range of the hybrid model). 3968

3969 TOPKAT 3970 TOPKAT contains a range of cross validated QSARs, which are multivariate statistical relationships 3971 between experimentally derived toxicity data and chemical descriptors that quantify chemical 3972 transport properties and biochemical interaction with the target site. It also provides the user with a 3973 measure of whether the query molecule fits within the prediction space of the chosen relationship and 3974 therefore whether the estimation is reliable. 3975

3976 Approach of Escher et al. 3977 The approach proposed by Escher et al. (2006) uses the principle of the toxic ratio (TR) (Verhaar et al. 3978 1992) of the parent PPP to estimate the maximum potency of a metabolite. The TR is the ratio 3979 between baseline toxicity, predicted using QSAR and the toxicity determined experimentally for the 3980



endpoint under investigation (Equation 8.1). Baseline toxicity is the lowest toxicity a chemical can 3981 exhibit, therefore a narcotic chemical would be expected to have a low TR, as long as the baseline 3982 toxicity prediction was accurate. 3983

3984 Equation 8.1 / ,

/ , 3985

3986 Where: 3987 LC/EC50,baseline - baseline (non-polar narcotic) toxicity of a compound estimated using 3988 QSAR [mol/L] 3989 LC/EC50,experimental - toxicity of the compound determined experimentally [mol/L] 3990 3991

The approach proposed by Escher et al. (2006) allows to estimate the ecotoxicological range of a 3992 metabolite. The minimum or baseline ecotoxicity of a metabolite (LC/EC50,baseline) is estimated using a 3993 suitable QSAR, whilst the maximum (or specific) ecotoxicity (LC/EC50,specific) is estimated by 3994 manipulating the baseline metabolite ecotoxicity with the TR of the parent compound (TRparent) 3995 (Equation 8.2). 3996 3997

Equation 8.2 / , / ,

log 3998

3999 4000 The estimation of LC/EC50,specific provides a worst-case estimate for metabolite ecotoxicity, i.e. the 4001 metabolite has the same potency/mode of action as the parent PPP. Therefore the majority of 4002 predictions generated by this technique may over-estimate the potency of a metabolite because there is 4003 an assumption that the metabolite(s) have the same potency as their parent PPP. 4004

4005 Other methods (from OECD (Q)SAR Application Toolbox) 4006 OASIS acute toxicity mode of action profiler was developed by the Laboratory of Mathematical 4007 Chemistry, Bourgas "Prof. As. Zlatarov" University, Bourgas, Bulgaria. It is based on a broader set of 4008 structural alerts gathered primarily from the fathead minnow toxicity testing and defined by Russom et 4009 al. (1997). 4010

The Verhaar classification (Verhaar et al., 1992) was developed utilizing acute toxicity data collection 4011 for guppies and fathead minnows. This scheme based on structural alerts delineated chemicals into one 4012 of five classes. The Verhaar classes include 1) Class 1 or “inert” chemicals, which are nonpolar 4013 narcosis or baseline toxicity, 2) Class 2 or “less inert” chemicals, which are the polar narcotics, 3) 4014 Class 3 or “reactivity” chemicals, which are typically non-selectively, covalently reactive with protein 4015 moieties, 4) Class 4 or “specifically-acting” chemicals, which specific reactivity with receptors, or 5) 4016 Class 5 or “unclassified” chemicals. 4017

Read across 4018 Read-Across for metabolites and impurities should include consideration on: 4019

• Molecular structure of the metabolite/impurity (active part intact or included?); 4020 • The occurrence of metabolites in existing tests with the active substance or major metabolites; 4021 • General knowledge on the relationship between the toxicity of metabolites and their parent 4022

substances; 4023 • Available knowledge on related compounds. 4024

4025



8.1.4. Comparison of QSAR model outputs 4026

The accuracy of DEMETRA in predicting the acute toxicity of 135 PPPs to a standard aquatic test 4027 organism (Daphnia magna) was tested and compared to the performance of ECOSAR and TOPKAT. 4028 DEMETRA was found to provide more accurate predictions than ECOSAR and TOPKAT which were 4029 not designed specifically for PPPs (Porcelli et al., 2008). The study indicated that ECOSAR (55%) and 4030 TOPKAT (40%) gave more false negatives than DEMETRA (20%). It should be noted that more PPP 4031 specific classes has been added to ECOSAR since 2007. 4032

Sinclair (2009) statistically compared measured and estimated acute toxicity data (n=92) for 4033 metabolites of PPP on Daphnia applying DEMETRA, ECOSAR, TOPCAT and Escher et al. expert 4034 system. Results indicated that the simple expert system overall performed best. This is surprising as 4035 this approach is based on a relatively simple concept and indicates that transformation product toxicity 4036 is substantially linked to that of its parent PPP, or at least for transformation products within the 4037 evaluation dataset used in this case (Sinclair and Boxall, 2003). This comparison also indicated that 4038 DEMETRA performed better than TOPKAT and ECOSAR. Again, please note that more PPP classes 4039 have been added to ECOSAR after this comparison. 4040

It is noted that the comparisons provide above are hampered by the fact that the comparison of 4041 estimates and experimental toxicity was only based on acute data derived for Daphnia. Further 4042 validation should include other groups of aquatic organism. 4043

Sinclair (2009) suggested that the simplest way of combining different estimation approaches would 4044 be to generate a conservative estimate of transformation product ecotoxicity, i.e. estimating 4045 ecotoxicity using all approaches and then selecting the most potent prediction (see Figure 8.1a). 4046 Combining approaches in this manner would provide a conservative estimation of ecotoxicity. More 4047 sophisticated methods to aggregate model predictions may be considered in order to better handle 4048 outliers and increase predictive ability, e.g. by calculating the geometric mean of predictive estimates 4049 or by developing rule-based methodologies. However, the likelihood of underestimating hazard may 4050 increase as can be seen from calculation of geometric mean toxicity of five different models in Figure 4051 8.1b. Rule-based aggregation of predictions would require further investigation into: 1) quantifying the 4052 predictive domain of each suitable approach, 2) rationalising the identity of outliers for each approach 4053 and 3) identifying which chemical types/categories are most appropriate for each approach. 4054 Developing such an approach would require a large transformation product dataset that extensively 4055 covers a range of taxa, physico-chemical properties, transformation product chemical classes and 4056 parent pesticidal chemical classes. Such tasks are outside the scope of this guidance document. For the 4057 reasons above, the PPR Panel proposes to use the more conservative endpoint from different model 4058 estimates. 4059



4060

Figure 8.1: A comparison of daphnid acute ecotoxicity data for 92 transformation products; a, the 4061 most potent estimates and b, the geometric means provided by the five evaluated approaches (dashed 4062 line x=y) (from Sinclair 2009). 4063

4064

8.1.5. Use of non-testing data in PPP risk assessment 4065

General recommendations 4066 The major concern in using non-testing data in environmental risk assessment of PPPs is related to the 4067 danger of under-estimating the real toxicity or hazard of a given substance. 4068

No single model/expert system can be recommended for a substance, as the applicability of specific 4069 models depends on the adequacy in relation to this specific substance. However, estimates generated 4070 by the different approaches will vary, and it may therefore increase the likelihood that if a compound 4071 does exhibit a potent ecotoxicity for whatever reason, this will be picked up by at least one approach. 4072

Only suitable models (e.g. covering the right domain) with a high predictive reliability should be used. 4073 This should among others be reflected in the level of statistical significance required for estimates 4074 from (Q)SAR models. Validation parameters should ideally indicate good fits (e.g. Q2> 0.7, ccc > 4075 0.85)25. Estimates of toxicity should where possible26 be assisted by confidence intervals around the 4076 prediction. In case the standard derivation exceeds the predicted value itself, such values should not be 4077 accepted. Generally, the worst case endpoint from several modelling approaches should be used. 4078

Estimates should however be confirmed by using weight-of-evidence approaches where all available 4079 information is taken into account. This could include a combination of the different (Q)SAR model 4080 predictions combined with read across and other available information like non-standard testing data 4081 and toxicodynamic/toxicokinetic information from mammals. 4082

Modelling of impurities 4083 Often ecotoxicological effect data are missing on impurities detected in new sources of active 4084 substances. In case a non-testing approach is considered to provide effect data, the general 4085 recommendation above should be applied. 4086

25For further details consult ECHA (2008) guidance. 26Not all QSAR models provide standard derivations for predictions.



Modelling of metabolites 4087 Whereas ecotoxicological effect data are normally present for the parent substance and major 4088 metabolites, such data are sometimes lacking for other metabolites. QSAR may be used as a mean to 4089 predict effect data to be used directly in hazard and risk assessments, to prioritise metabolites of 4090 highest concern for further testing and/or to identify the likely most sensitive species for further testing 4091 (see section 8.2 on assessment of metabolites). The general recommendations above should be 4092 followed. The same level of strict validation criteria should be applied for long-term QSAR 4093 ecotoxicity predictions as for acute QSAR ecotoxicity predictions. It is noted that less valid QSAR 4094 models are currently available for deriving longer-term toxicity data. 4095

8.1.6. Decision scheme for use of non-testing systems 4096

1. Is the QSAR model valid – i.e. is it relevant and reliable (following 5 OECD principles for 4097 assessing QSAR models). E.g. is prediction accurate enough (recommended assessment 4098 values Q2, CCC, SD) 4099 Yes: Go to 2 4100 No: QSAR not applicable – consider other model 4101 4102

2. Does the substance and model match - i.e. is the chemical of interest within the scope of the 4103 model, according to the defined applicability domain of the model and is the defined 4104 applicability domain suitable for the regulatory purpose? 4105 Yes: Go to 3 4106 No: QSAR not applicable – consider other model 4107 4108

3. Does model predict take into account test information? (E.g. for aquatic toxicity consider 4109 water solubility, log Kow, degradability and volatility) 4110 Yes: Go to 4 4111 No: QSAR not applicable – consider other model 4112 4113

4. Are reliable estimations available from more than one QSAR model 4114 Yes: use lowest predicted QSAR endpoint in risk assessment or as qualifier for testing if 4115 confirmed by weight of evidence approach 4116 No: Single value may be used in risk assessment or as qualifier for testing if clearly 4117 confirmed by weight of evidence approach 4118 4119

4120

8.2. Metabolites and degradation products 4121


Active substances in PPPs may be transformed in the environment by either abiotic or biotic 4123 processes. In Regulation (EC) No 1107/2009 a metabolite is defined as “any metabolite or a 4124 degradation product of an active substance, safener or synergist, formed either in organisms or in the 4125 environment. A metabolite is deemed relevant if there is a reason to assume that it has intrinsic 4126 properties comparable to the parent substance in terms of its biological target activity (presence of 4127 toxophor), or that it poses a higher or comparable risk to organisms than the parent substance or that it 4128 has certain toxicological properties that are considered unacceptable. Such a metabolite is relevant for 4129 the overall approval decision or for the definition of risk mitigation measures;” 4130

8.2.2. Definition of the residue for risk assessment 4131

In the revised data requirement for active substances (SANCO document 11802/2012) it is stated 4132 under Part A point 7.4.1 Definition of the residue for risk assessment that “the residue definition 4133 relevant for risk assessment for each compartment shall be defined to include all components (active 4134 substance, metabolites, degradation metabolites, breakdown and reaction products) that were 4135



identified in accordance with the criteria referred to in this Section. The chemical composition of 4136 residues occurring in soil, groundwater, surface water (freshwater, estuarine and marine) and, 4137 sediment or and air, resulting from use, or proposed use, of a plant protection product containing the 4138 active substance, shall be taken into account.” 4139

The criteria for identification are given in the section on fate and behaviour in the data requirements 4140 for the active substances for the aerobic/anaerobic soil studies (SANCO document 11802/2012 p. 4141 7.1.1.1) and is cited below: 4142

• “identify the individual components present which at any time account for more than 10% of 4143 the amount of active substance added, including, if possible, non-extractable residues; 4144

• identify, if possible, the individual components which in at least two sequential measurements, 4145 account for more than 5% of the amount of active substance added; 4146

• identify, if possible, the individual components (> 5%) for which at the end of the study the 4147 maximum of formation is not yet reached; 4148

• identify or characterise, if possible, other individual components present;” 4149 4150

In addition to these requirements the requirement for the degradation in surface water (SANCO 4151 document 11802/2012 p 7.2.2.2. and p 7.2.2.3) indicates that the study shall “permit, where relevant, 4152 the sediment residue of concern and to which non-target species are or may be exposed, to be 4153 defined”. 4154

Identification, and hence risk assessment is also (implicitly) required for metabolites, degradation and 4155 reaction products which account at any time for more than 10% of the amount of active substance 4156 added in the hydrolysis study (SANCO document 11802/2012 p. 7.2.1.1) and the study on direct 4157 phototransformation in water (SANCO document 11802/2012 p. 7.2.1.2). 4158

In general risk assessment for metabolites formed below 5 % is not considered necessary. However, if 4159 there is reason to believe that a metabolite formed at <5% has intrinsic properties comparable to the 4160 parent substance in terms of its biological target activity, or that it has certain structural properties 4161 indicating high reactivity (i.e. mutagenicity) or endocrine disrupting properties or that it has 4162 unacceptable toxicological properties, then that metabolite may be ecotoxicologically relevant and a 4163 risk assessment is needed. 4164

The following derogation (SANCO document 11802/2012 p. 7.1.1.2.2) from the above requirements 4165 applies for metabolites identified in the soil compartment: 4166

“If, during adequate field studies metabolites, degradation and reaction products which are present in 4167 laboratory studies are below LOQ, which shall not exceed an equivalent of 5% (molar basis) of the 4168 nominal concentration of active ingredient applied, then in principle no additional information on the 4169 fate and behaviour of these compounds shall be provided . A scientifically valid justification for any 4170 discrepancy between laboratory and field appearance of metabolites shall be provided.” 4171

Since no further information on fate and behaviour is necessary under these circumstances, it follows 4172 that no further information on ecotoxicity would be necessary. 4173

In the revised data requirements (SANCO document 11802/2012) the lysimeter study is mentioned (p. 4174 7.1.4.2) as an experimental outdoor study in the framework of a tiered leaching assessment scheme. It 4175 is stated that the lysimeter studies shall be performed, where necessary, to provide information on: the 4176 mobility in soil; the potential for leaching to ground water and the potential distribution in soil. It is 4177 not mentioned that the study should be used to identify metabolites relevant for ecotoxicological risk 4178 assessment. It is required that the degradation and adsorption of metabolites that leach at a 4179



concentration level above 0.1 µg/L in the lysimeter is determined (SANCO document 11802/2012 4180 7.1.1.2.1 and 7.1.2.1). This implies that metabolites formed in a lysimeter is relevant for groundwater 4181 assessment but not for aquatic risk assessment (see also p. 9.2.4 of SANCO document 11803/2012 for 4182 plant protection products). A leachate concentration of 0.1 µg/L in 100 mm percolating water 4183 corresponds with 0.0001 kg/ha, so 0.01% of a dose of 1 kg/ha. So triggering an aquatic risk 4184 assessment for a metabolite based on exceedance of a lysimeter percolate concentration of 0.1 µg/L 4185 would be much more strict than triggering such a risk assessment on the basis of exceeding 5-10% of 4186 the dose of the parent in the soil, surface water (considering mineralisation, hydrolysis or photolysis) 4187 and water-sediment studies mentioned before. Moreover, such a lysimeter study is a higher-tier study 4188 for the leaching assessment and thus not available on a standard basis. It seems therefore inconsistent 4189 to trigger an aquatic risk assessment of a metabolite on the basis of the fact that it exceeds 0.1 µg/L in 4190 lysimeter percolate (of course unless there are ecotoxicological reasons to believe that this metabolite 4191 may cause a problem as described before). 4192

If the metabolite is CO2 or an inorganic compound that is not a heavy metal; or, it is an organic 4193 compound of aliphatic structure, with a chain length of 4 or less, which consists only of C, H, N or O 4194 atoms and which has no "alerting structures" such as epoxide, nitrosamine, nitrile or other functional 4195 groups of known toxicological concern, then no further studies are required and the metabolite is 4196 considered to be not ecotoxicologically relevant and therefore of low risk to the environment. 4197

All metabolites which meet the criteria and definitions described above, would consequently be 4198 included in the “Definition of the residue for risk assessment” and are hereafter called potentially 4199 relevant metabolites. For these, estimation of exposure (Predicted Environmental Concentration) is 4200 necessary for each relevant compartment, see section 4.3.2, as well as information on ecotoxicity. 4201

8.2.3. Risk assessment scheme for metabolites 4202

The decision scheme has been developed in order to facilitate the selection of the most appropriate and 4203 pragmatic assessment route for metabolites. However, possible endocrine disruption properties should 4204 be addressed separately (see section 5.6)). 4205

Sinclair (2009) investigated the toxicity of metabolites in relation to the parent compound of several 4206 PPPs (60 a.s. and 485 transformation products) and demonstrated that the majority (70%) of 4207 transformation products had either a similar toxicity to the parent compound or are less toxic. 4208 However, a significant proportion (30%) were more toxic than their parent compound and 4.2% of 4209 transformation products were more than an order of magnitude more toxic. Over 90% of the observed 4210 increases in toxicity of the metabolite could be explained by the presence of a toxophore (see section 4211 8.2.5), differences in accumulation (i.e. hydrophobicity) or differences in mode of action (for example 4212 active components of pro-PPPs or highly reactive metabolites). Furthermore, the investigation showed 4213 that transformation product that is more hydrophobic than its parent compound and does not have 4214 pesticidal activity is unlikely to be more toxic than its parent to sensitive species that have a receptor 4215 site relevant to the parent mode of action. Hence, the panel has developed an assessment scheme for 4216 metabolites where metabolites for which it is clearly shown that the toxophore is lost in a first step can 4217 be assessed using approximation of toxicity (see section 8.2.7) while testing is required for metabolites 4218 with remaining toxophore (see section 8.2.6). 4219

4220 1. Is the exposure to the metabolite in the toxicity test with the a.s. adequate for assessing the 4221

potential effect of the metabolite (see section 8.2.4)? 4222 Yes: Go to 2 4223 No: Go to 3 4224

4225 2. Perform the risk assessment assuming all the effect observed in the test with the a.s. can be 4226

attributed to the metabolite (see section 8.2.4). Is RACsw;ac > PECsw and RACsw;ch > PECsw? 4227



Yes: low risk 4228 No: Go to 3 4229

4230 3. Is it clear that the toxophore has been lost from the molecule (see section 8.2.5)? 4231

Yes: Go to 6 4232 No or unclear: Go to 4 4233 4234

4. Determine the toxicity to species or taxonomic group27 providing the lowest Tier 1 RACsw;ac 4235 of the a.s. Is the acute metabolite LC50> 10 times the a.s. LC50 (on a molarbasis) (see section 4236 8.2.6)? 4237

Yes: Go to 6 4238 No: Go to 5 4239 4240

5. Determine the toxicity to species or taxonomic group27 providing the lowest Tier 1 RACsw;ch 4241 of the a.s. Is RACsw;ac> PECsw and RACsw;ch > PECsw? 4242

Yes: low risk 4243 No: Consider higher tier refinement 4244 4245

6. Assume that the acute and chronic28 toxicity of the metabolite is equal to the toxicity of the 4246 a.s. for all first tier taxonomic groups (see section 8.2.7)? Is RACsw;ac> PECsw and RACsw;ch> 4247 PECsw? 4248

Yes: low risk 4249 No: Go to 7 4250 4251

7. Are reliable and adequate non testing predictions of toxicity (see section 8.1) available for all 4252 first tier taxonomic groups (fish, plants and invertebrates)? Use prediction to assess if 4253 RACsw;ac> PECsw and RACsw;ch> PECsw ? 4254


4257 8. Determine the acute and chronic28 toxicity is for those taxonomic groups where a valid non-4258

testing predictions of toxicity is not available or for which a risk was identified using 4259 predicted toxicity? Is RACsw;ac> PECsw and RACsw;ch> PECsw ? 4260


8.2.4. Alternative information replacing experimental studies 4264

The principles for assessing metabolites should in essence be the same as those for active substances. 4265 However, in contrast to the active substance, data requirements for metabolites do not always have to 4266 be addressed by experimental studies. Applicants are invited to address the open questions by any 4267 other available information in support of a scientific and rational assessment. Examples of such 4268 information are shortly described below. 4269

27Consider testing with chironomidae if metabolite is distributed in sediment, or other taxonomic group suspected to be most

sensitive (e.g. chironomidae with IGRs) 28if chronic risk assessment is triggered by fate properties of the metabolite



If chemical analysis confirm that the metabolite was present in the test system originally designed for 4270 testing of the a.s. organisms could be considered to have been exposed to the metabolites. The risk 4271 may then be addressed by information from the study with the a.s. assuming that all the observed 4272 effect in the test can be attributed to the metabolite when determining the RAC for the metabolite. 4273 However, this extrapolation is only valid if it is shown that the organisms were exposed to a realistic 4274 or worst-case exposure profile of the metabolite (e.g. compared to FOCUS profile or profile observed 4275 in the water sediment studies). For this extrapolation to be valid it is also important that the time 4276 period after the measured metabolite concentration was of sufficient length for observation of effects. 4277 In general, it will therefore only be possible to use the concentrations of the metabolite measured early 4278 in the test when establishing the RAC. Another possibility could be to prolong the test in order to 4279 lengthen the observation phase from effect occurring due to exposure to the metabolite. 4280

If a metabolite is e.g. formed rapidly via hydrolysis, the toxicity of the metabolite may similarly have 4281 been assessed as part of the standard toxicity studies (this should be supported by analytical 4282 measurements) and addressed as above. However, if a toxicity test is performed at ≥ pH7 and if other 4283 metabolites are formed at pH5 the toxicity of these metabolites need to be addressed separately in 4284 order to cover the risk in more acid waters. Therefore the data from a hydrolysis study should be used 4285 to decide to which extent degradation and toxicity depend on the pH-value of the test medium. 4286

In toxicity studies with intensive lighting (e.g. algae and Lemna tests), it is likely that metabolites 4287 which are formed as a result of photolysis are present in an amount which is relevant for field 4288 conditions and additional toxicity testing with metabolites detected in the photolysis study might not 4289 be warranted. This is particularly the case when static studies have been used. These conclusions 4290 should be supported by analytical measurements and the risk resulting from the metabolite can be 4291 addressed as above. 4292

Furthermore information on the toxicity of metabolites can be derived from studies on the metabolites 4293 on other organisms than those relevant for the aquatic ecosystem (e.g. non target arthropods, terrestrial 4294 plants). 4295

8.2.5. Identification of toxophore 4296

Substances that have a specific mode of action, like PPPs, contain a structural feature or moiety that 4297 gives the toxic property. This structural feature is referred as the toxophore, or toxophoric moiety. The 4298 substance causes toxicity through the interaction of its toxophore with a biomolecular site (e.g., 4299 receptor). Substances that are structurally similar could contain the same toxophore (or may yield a 4300 common toxophore upon metabolism) and may therefore have a common toxic effect. 4301

For the assessment of the metabolite the applicant has to provide a reasoned case as to if the molecule 4302 contain a toxophore or if it has been lost following transformation. Toxophores for each of the major 4303 classes of PPP has been identified by looking for sub-structural similarities within a pesticidal class by 4304 Sinclair et al. (2009), which can be used to support argumentation. A number of ways has been 4305 identified to define domain of applicability, which may be used to decide if toxophores are present or 4306 not (Nikolova and Jaworska, 2003; Dimitrov et al., 2005; Jaworska et al., 2005; Netzeva et al., 2005). 4307 In case it cannot be clearly shown that the toxophore is not present in the molecule it should be 4308 assumed that the toxophore remain and that the molecule has a specific mode of action (see 4309 assessment scheme 8.2.3) 4310

8.2.6. Metabolites structurally similar to the active ingredient and with remaining toxophore. 4311

It is likely that metabolites that are structurally similar to the active ingredient (i.e. the toxophore 4312 remain) are most toxic to the same taxonomic group that was shown to be most sensitive to the a.s. For 4313



such compounds, testing can in a first step be limited to the taxonomic group that was identified to 4314 result in the lowest Tier 1 RACsw,ac and RACsw,ch for the a.s. If however testing with this taxonomic 4315 group shows that this taxonomic group is not sensitive (i.e. the acute end point is > a factor 10 higher 4316 as compared to the parent, on a molar basis29) then the risk assessment needs to be continued assuming 4317 that the molecule does not contain a toxophore most sensitive taxonomic group is unknown and the 4318 risk to all taxonomic groups should be addressed, see section 8.2.7. 4319

If it is unclear if the toxophore remains and the most sensitive group is not known, then the risk 4320 assessment needs to address all taxonomic groups. 4321

8.2.7. Metabolites with no toxophore 4322

If it is clear that the toxophore has been lost from the metabolite, in most cases metabolites are less 4323 toxic to the target organisms than the active substances. As a pragmatic and conservative approach for 4324 metabolites without the toxophore the estimates of exposure could be compared with the RACparent 4325 based on the most sensitive endpoint of the active substance in the relevant compartment. In general 4326 only if this trigger is failed the toxicity needs to be further addressed (see section 8.2.8). 4327

8.2.8. Non-testing predictions of metabolite toxicity 4328

For metabolites which have lost the toxophore the acute and long-term hazard and risk can be 4329 addressed using non-testing predictions of toxicity (see further in section 8.1). In principle, non-testing 4330 methods for predicting toxicity could also be used for specifically acting chemicals, i.e. metabolites 4331 with a toxophore. However, based on the assessment by Sinclair (2009) (see section 8.2.3) it is the 4332 view of the PPR Panel that only metabolites without toxophores should be included for non-testing 4333 estimates. and is therefore expected to be of practical use only for limited number of metabolites. If 4334 the trigger is failed using predicted toxicity then testing is required, see below. 4335

8.2.9. Toxicity testing with metabolites 4336

For metabolites which require experimental studies (see assessment scheme in section 8.2.3), acute 4337 toxicity tests with Daphnia and rainbow trout and an algae should be conducted. In general the same 4338 testing scheme as for active ingredients substances (see table 5.1) is required. Hence testing on 4339 additional species (e.g. additional invertebrate species) may be necessary where the risk to a particular 4340 taxonomic group for the a.s. is considered to be of concern (e.g. additional testing with macrophyte for 4341 a herbicidal metabolite). 4342

In principle, for metabolites found in the sediment of a water-sediment study, the same triggers for 4343 testing should be applied to metabolites as for the active substance (see section 5.2.3.1). That is when 4344 accumulation of a substance in aquatic sediment is indicated or predicted from environmental fate 4345 studies, the impact on a sediment-dwelling organism should be assessed. Clearly the potential to 4346 exclude testing on the basis of toxicity will depend on the data that is available for the metabolite. The 4347 applicant should therefore make a case as to whether sediment testing can be waived based on what is 4348 known about the fate properties and toxicity profile of the metabolite. For example, if risk assessments 4349 with Daphnia indicate that the potential risks are low (taking into account the exposure situation in the 4350 sediment), then no further testing should in general be required. As a first screening step for 4351 metabolites partitioning to the sediment a formula based on equilibrium partitioning theory as outlined 4352 in the TGD part II (EC 2003) section 3.5.3, can be used to indicate if actual testing is needed. Only if a 4353 risk is indicated using this formula actual testing with sediment organisms should be required. This 4354 will be further addressed in a PPR Panel Opinion on sediment risk assessment under development. 4355 29The statement to check whether the LC50 of the metabolite is greater than 10 times the LC of the a.s. on a molar basis

means:

10

where LC50met and LC50ai are mass concentrations (mg/L) of metabolite and active ingredient at 50% mortality and Mmet and Mai are the molar masses (g/mol) of the metabolite and active ingredient.



In order to decide whether chronic testing is necessary the intended uses and the fate and behaviour of 4356 the metabolite should be taken into account. In general chronic/long term tests are required for 4357 metabolites where exposure of surface water is likely and the metabolite is deemed to be stable in 4358 water, as defined in the data requirements, i.e. there is less than 90% loss of the original substance 4359 over 24 hours via hydrolysis under relevant pH conditions (SANCO document 11802 and 4360 11803/2012). 4361

In terms of the choice of taxonomic group(s) to be studied for chronic toxicity, this should take 4362 account of any acute toxicity data on the metabolite. Where information on the acute sensitivity of fish 4363 and invertebrates for a particular metabolite is available, chronic testing should only be required on the 4364 more sensitive group (i.e. that are a factor of 10 more sensitive).If daphnia is suspected to be 4365 insensitive based on the mode of action of the active ingredient, e.g. it is an insect growth regulator or 4366 a neonicotinoid, then it is necessary to conduct a chronic study using the Chironomid, Chironomus 4367 riparius with the metabolite. 4368

For unstable active substances (i.e. there is more than 90% loss of the original substance over 24 hours 4369 via hydrolysis), it may be more appropriate to conduct chronic studies on the stable metabolite instead 4370 of the parent compound. For unstable active substances, where chronic toxicity data for the parent 4371 compound are not available and an environmentally significant metabolite exceeds the persistence 4372 criteria (i.e. there is less than 90% loss of the original substance over 24 hours via hydrolysis), chronic 4373 toxicity data should be submitted for this metabolite regardless of its acute toxicity. 4374

The endocrine disrupting properties of metabolites should also be addressed, however, until common 4375 criteria are developed and agreed by the Commission, it is difficult to give specific guidance on how to 4376 assess EDC in relation to PPPs (see section 1.4.6). Therefore further guidance on how to assess EDC 4377 might be given later when the work of the Commission is finalized. Nevertheless, based o structural 4378 properties of the metabolites and also based on information on related compounds if there are 4379 indications that the metabolite may exhibit endocrine disrupting properties, then chronic/long-term 4380 tests with fish should always be required with this metabolite. 4381

The BCF should be determined as for active substance if the metabolite is stable (i.e. there is less than 4382 90% loss of the original substance over 24 hours via hydrolysis) and has a log Pow>3. 4383

8.2.10. Risk Assessment for Metabolites 4384

In principle, the risk assessment process for metabolites will be similar to that for active substances, 4385 albeit recognizing that risk assessment cases will not always require specific study data for certain 4386 metabolites. If preliminary risk assessments indicate potential concerns then, as for parent molecules, 4387 risk refinement is possible either by refining effect concentrations or by refinement of the exposure 4388 concentration. 4389

If higher-tier studies have been conducted with the active substance, or a relevant formulation, these 4390 studies may have also assessed the risk from the metabolites. It is advised that if a higher-tier study, 4391 e.g. mesocosm study, is being carried out then appropriate analysis should be conducted so that an 4392 assessment of both the exposure and effects of any metabolites can be made. 4393

8.2.11. Definition of the residue for monitoring 4394

Considering the results of toxicological and ecotoxicological testing, the residue for monitoring is 4395 defined to include only those components from the definition of the residue for risk assessment, which 4396 were classified as relevant in those tests (SANCO document 11802/2012 p. 7.4.2). Current practice is 4397 to define as relevant only those metabolites for which a risk has been identified, i.e. metabolites that 4398 pose risk that warrant risk mitigation measures. 4399

For relevant metabolites an analytic method for the compartments of interest should be available 4400 following Annex I listing. 4401



8.3. Combinations of a.s. in formulations (guidance on toxic unit approaches) 4402

The PPR Panel proposes to follow the same approach as outlined in the Guidance Document on Risk 4403 Assessment for Birds & Mammals (EFSA, 2009c) and in the Scientific Opinion on the science behind 4404 the development of a risk assessment of Plant Protection Products on bees (EFSA, 2012b). In those 4405 two documents, in particular in the bee opinion, more detailed background information can be found. 4406

In a recent review for the European Commission (Kortenkamp et al., 2009), the use of the 4407 concentration addition model was proposed as the concept of mixture toxicity that is most relevant for 4408 hazard characterisation and ultimately can be integrated into the legislative process for risk 4409 management purposes. The use of concentration addition has also been discussed by Verbruggen and 4410 van den Brink (2010). There are two reasons that make the use of this model concept attractive for 4411 policy makers. First, the model concept is generally more conservative than the concept of response 4412 addition. Nevertheless, the magnitude of the differences at low levels of exposure between the two 4413 models is usually small and hence, the outcome will not be overly conservative. A second reason for 4414 the use of concentration addition is that the model concept can make use of existing data such as a 4415 NOEC, EC10 or EC50’s by applying the concept of toxic units (TUs). 4416

The concept of TUs has been recently reviewed by the three non-food committees of the European 4417 Commission (the Scientific Committee on Health and Environmental Risks (SCHER), the Scientific 4418 Committee on Emerging and Newly Identified Health Risks (SCENHIR), the Scientific Committee on 4419 Consumer Safety (SCCS)) which defined TUs as “the ratio between the concentration of a mixture 4420 component and its toxicological acute (e.g. LC50) or chronic (e.g. long-term NOEC) endpoint”. In 4421 addition, the toxic unit of a mixture (TUm) has been defined as the sum of TUs of each individual 4422 chemical of that mixture. The committees also noted that the TUs concept only refers to a specific 4423 organism representative of a group of organisms ecologically or taxonomically relevant for the 4424 ecosystem (e.g. algae, daphnids and fish for the freshwater ecosystem) but not to the ecosystem as a 4425 whole (SCHER/SCENIHR/SCCS, 2011). 4426

When comparing estimates using CA, it has been estimated that the majority of estimates do not 4427 deviate by more than a factor of 2 (Deneer, 2000), 2.5 (Warne, 2003; Junghans et al., 2006) or 3 4428 (Kortenkamp et al., 2009). There is also some evidence that this deviation is greatest for mixtures 4429 containing small numbers of chemicals and decreases as the complexity of the mixture increases 4430 (Kortenkamp et al., 2009). 4431

4432

Concentration addition (CA) 4433

The following equation can be used for deriving a surrogate ECx or NOEC value for a mixture of 4434 active substances with known toxicity assuming concentration additivity: 4435

ECx (mix) or NOEC (mix) = ⎟⎟⎠

⎞⎜⎜⎝

⎛∑ )..(__

)..(

ix

i

i saNOECorECsaX

4436

4437 Where: 4438

X(a.s.i) = fraction of active substance [i] in the mixture (please note that the sum Σ X(a.s.i) must be 1) 4439

ECx or NOEC(a.s.i) = toxicity value for active substance [i]. 4440

4441 Where the toxicity value of a formulated product with more than one active substance is available, this 4442 value should be compared with the predicted mixture toxicity assuming concentration additivity. A 4443 different form of the equation is used. 4444



)(__1

)..(__)..(

mixNOECorECsaNOECorECsaX

xi ix

i =∑ 4445

4446 X(a.s.i) = fraction of active substance [i] in the mixture (here: formulation) 4447

ECx or NOEC(a.s.i) = acute toxicity value for active substance [i] 4448

ECx or NOEC(mix) = measured acute toxicity value for the mixture (here: formulation) 4449

A greater value on the right side of the equation indicates that the formulation is more toxic than 4450 predicted from the toxicity of the individual components (active substances and co-formulants of 4451 known toxicity). This may be due to, e.g. further toxic co-formulants, toxicokinetic interaction or 4452 synergism/potentiation of effect. It may also reflect the inherent variability of toxicity testing. In all 4453 these cases, the use of the EC50 for the formulation (together with appropriate exposure estimates, see 4454 Step 4) is recommended for the first tier assessment, because it cannot be excluded that such effects 4455 would also occur after exposure of organisms to residues in the environment. In case the measured 4456 acute toxicity of the formulation indicates a higher toxicity than predicted from the toxicity of the 4457 individual components (i.e. more than a factor of 10) then a chronic test on the formulation may be 4458 required, see further section 5.5. 4459

Dismissing the EC50 of the formulation from the risk assessment would only be acceptable at a higher 4460 tier if any observed greater toxicity in the test could be clearly and unambiguously ascribed to a factor 4461 that would not be relevant under environmental exposure conditions. 4462

If, in contrast, the measured toxicity of a formulation is lower than predicted, the predicted mixture 4463 toxicity should be used in the first tier risk assessment, together with appropriate exposure estimates. 4464

The ECx(mix) or NOEC(mix) should be compared to the sum of the concentrations of the a.s. in order 4465 to calculate the ETRmix. As a first approach it is assumed that the PECsw;max of all a.s. present in the 4466 formulation will occur at the same moment and are not separated in time. In case the trigger value is 4467 not met in higher tiers the predicted exposure patterns can be taken into account (see for example 4468 calculations table 8.1 and table 8.2). 4469

Table 8.1 Example for a Tier 1calculation using highest peaks (PECsw; max) for a mixture of two 4470 compounds (all concentrations in µg/l). Values printed in bold are above the trigger value of 0.01 and 4471 additional risk assessment should be considered (example for acute first tier risk assessment of fish or 4472 invertebrates). 4473

Concentration highest peak compound A 0.12 Concentration highest peak compound B 0.23 Concentration highest peak compounds A + B 0.35 Toxicity compound A 10.0 Toxicity compound B 8.00 Toxicity mixture 8.59 ETR mixture 0.041

Table 8.2 Example for a higher tier calculation for a mixture of two compounds (all concentrations 4474 in µg/l). Values printed in bold are above the trigger value of 0.01 and additional risk assessment 4475 should be considered (example for acute first tier risk assessment of fish or invertebrates). 4476

Days 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Concentration compound A 0.09 0.08 0.07 0.06 0.05 0.04 0.12 0.10 0.08 0.07 0.06 0.05 0.04 0.03 Concentration compound B 0.23 0.12 0.06 0.03 0.01 0.00 0.23 0.12 0.06 0.03 0.01 0.00 0.00 0.00



Days 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Concentration compounds A & B 0.32 0.20 0.13 0.09 0.06 0.04 0.35 0.22 0.14 0.10 0.07 0.05 0.04 0.03 Toxicity compound A 10.00 10.00 10.00 10.00 10.00 10.00 10.00 10.00 10.00 10.00 10.00 10.00 10.00 10.00 Toxicity compound B 8.00 8.00 8.00 8.00 8.00 8.00 8.00 8.00 8.00 8.00 8.00 8.00 8.00 8.00 Toxicity mixture 8.48 8.70 8.97 9.23 9.60 10.00 8.59 8.80 9.03 9.30 9.66 10.00 10.00 10.00 ETR mixture 0.038 0.023 0.015 0.010 0.006 0.004 0.041 0.025 0.016 0.011 0.007 0.005 0.004 0.003

4477 In case the endpoint to be used for the risk assessment is associated with different assessment factors 4478 the calculation of the mixture toxicity can be based on the regulatory acceptable concentration (RAC) 4479 and the following formula could be used: 4480

PECA/RACA + PECB/RACB + ….= SUM 4481

If SUM < 1 the risk is low. 4482

The SUM can be calculated in the acute or chronic risk assessment for the same relevant taxonomic 4483 group (i.e. fish, crustaceans, algae and aquatic plants) using the single species test-RAC, Geomean-4484 RAC or SSD-RAC. In case data from micro-/mesocosms are used, then use the overall ETO-RAC. 4485

4486

9. Other Issues 4487

9.1. Test batches/impurities 4488

Differences in the chemical composition of the test batches might alter the ecotoxicological profile of 4489 the technical material. In Regulation (EC) No 1107/2009, impurities are defined as any component 4490 other than the pure active substance and/or variant which is present in the technical material (including 4491 components originating from the manufacturing process or from degradation during storage) [Art. 3 4492 (33)]. In the European Commission Guidance Document SANCO/10597/2003 –rev. 9 17 June 2011 4493 for assessing the equivalence of technical material, an impurity is considered significant when it 4494 occurs, due to process variability, in quantities ≥ 1 g/kg in the active substance as manufactured, based 4495 on dry weight; an impurity is considered relevant when it is of toxicological and/or ecotoxicological or 4496 environmental concern compared with the active substance, even if present in technical material at < 1 4497 g/kg. 4498

Information on the composition of the test batches used in the ecotoxicological tests should be 4499 available. When the composition of the batches is comparable with the specification of technical 4500 material (SANCO/10597/2003), consisting of the active substance and its associated impurities, no 4501 further assessment is requested. However, when the composition of the batches is different, in terms of 4502 different amount of the same impurities or different impurities or different amount of the active 4503 substances, then the ecotoxicological representativeness of the related endpoints should be addressed. 4504 To determine whether the batches used in the ecotoxicity studies are equivalent to those which will be 4505 approved, it is recommended that the assessment methodology provided in the latest version of the 4506 European Commission Guidance Document SANCO/10597/2003 is followed. It is recommended that 4507 a table with the ecotoxicity tests, the batches analysed and the observed endpoints is included to make 4508 the comparison of the toxicity of different batches easier. 4509

9.2. Testing poorly soluble and other difficult test substances 4510

Detailed guidance on how to deal with poorly soluble substances as well as other substances that are 4511 difficult in aquatic toxicity testing (e.g. volatile or adsorbing substances) can be found in the “OECD 4512 Guidance document on aquatic toxicity testing of difficult substances and mixtures.” (OECD 2000). 4513



9.3. Promising mechanistic effect models 4514

Mechanistic ecological models have been applied to ecotoxicological questions for over 25 years now 4515 (e.g. O’Neill et al. 1982, Kooiijman and Metz 1984), but over the years their acceptance in regulatory 4516 environmental risk assessment (ERA) was very limited. Current risk assessment is based mainly on 4517 statistical (e.g. LOEC, LC50, SSD) and physical models (e.g. Daphnia magna, microcosm) which does 4518 not explicitly rely on systematic understanding on the system of interest. However, recently 4519 importance of mechanistic modelling seems to increase especially in the ERA of PPPs under the 4520 European Regulation (EC) 1107/2009. Mechanistic modelling was mentioned as valuable higher tier 4521 tool in the SETAC workshops AMPERE on mesocosm tests (2007) and AMRAP on macrophyte 4522 testing (Maltby et al. 2010), the ELINK workshop on complex exposure scenarios (Brock et al. 2010c) 4523 and finally the LEMTOX workshop discussing pros and cons of population models (Thorbek et al. 4524 2010). Additionally mechanistic modelling is explicitly mentioned in the EFSA Opinion on Protection 4525 goals (EFSA, 2010b). 4526

Despite the potential power of mechanistic effect modelling to answer important questions within 4527 ERA and its long history in science, its use within ERA is not well tested nowadays and no general 4528 guidance is available. In near future EFSA will work out an opinion on good modelling practice 4529 (EFSA-Q-2011-00989) and a scientific opinion of the PPR Panel on modelling within the aquatic risk 4530 assessment. Since there is a lack of experience and guidance for these approaches in risk assessment 4531 the use of mechanistic modelling within the authorisation of plant protection products has to be 4532 evaluated carefully case by case until special guidance is at hand. 4533

In the future it is expected that mechanistic effect models at three levels of biological organisation will 4534 be used to support the risk assessment of PPPs. On the individual level, toxicokinetics-toxicodynamics 4535 approaches (TK/TD models) simulate survival (Jager et al. 2011) as well as sublethal effects (Jager et 4536 al. 2006) over time, based on uptake and elimination of the toxicant (toxicokinetics) and damage and 4537 repair processes (toxicodynamics) within the organism. Population models allow extrapolation of 4538 lethal and sublethal effects from the individual to the population level (Barnthouse 2004; Forbes and 4539 Calow, 2002; Preuss et al. 2010; Van den Brink et al. 2007). Ecosystem models (Hommen et al. 1993; 4540 Park et al. 2008; Traas et al. 2004) allow the risk characterization within ecosystems, integrating biotic 4541 interactions and thereby indirect effects as found in experimental studies (Duchet et al. 2010). 4542 Population and ecosystem modelling are recognised as important tools through which for example 4543 protection goal fulfilment of lower tier risk assessment can be explored, however, due to the current 4544 state of the art of these models detailed recommendations for their use in RA cannot be given at this 4545 stage. 4546

9.4. Reduction of (vertebrate) testing 4547

Directive 2010/63/EU on the protection of animals used for scientific purposes, describes that “When 4548 choosing methods, the principles of replacement, reduction and refinement should be implemented 4549 through a strict hierarchy of the requirement to use alternative methods.” 4550

Regulation (EC) 1107/2009 clearly requires “The use of non-animal test methods and other risk 4551 assessment strategies should be promoted. Animal testing for the purposes of this Regulation should 4552 be minimised and tests on vertebrates should be undertaken as a last resort.” Therefore, for aquatic risk 4553 assessment alternatives to experimental testing are specifically recommended for fish. 4554

9.4.1. Use of limit tests 4555

The data requirements (SANCO document 11802/2012) generally state for aquatic organisms that a 4556 limit test at 100 mg substance/L may be performed when the results of a range finding test indicate 4557 that no effects are to be expected. 4558

Specifically with respect to minimising vertebrate / fish testing, a threshold approach to acute fish 4559 testing should be considered. An acute fish limit test should be conducted at 100 mg substance/L or at 4560 an appropriate concentration selected from aquatic endpoints following consideration of the threshold 4561



exposure (see also OECD guideline on acute fish testing 203). When mortality is detected in the fish 4562 limit test an acute fish dose-response toxicity study shall be required to determine an LC50 for use in 4563 risk assessment (see also Chapter 5). 4564

A workshop was held in December 2010 in UK to investigate the possibilities of using the threshold 4565 approach for acute fish testing in PPP RA (Creton et al. in preparation). The threshold approach was 4566 considered useful for assessing active substances and further possibilities to reduce the acute fish 4567 testing for formulations are described. 4568

9.4.2. Use of non-testing methods 4569

Non-testing methods such as QSAR and read-across can be applied to fill certain data gaps if no test 4570 data are available. This particularly applies to testing for metabolites or impurities where appropriate 4571 non-testing methods can be recommended in certain cases (see section 8.3). 4572

9.5. Differences in risk assessment procedures between Regulation (EC) 1107/2009 and the 4573 Water Framework Directive (WFD) 4574


In Europe different legislations (Directives, Regulations) have been developed with different 4576 methodologies to assess the aquatic risks of PPPs. In particular, these differences are apparent when 4577 comparing the authorisation criteria for the compartment water according to the plant protection 4578 product Regulation (1107/2009/EC) and the water quality standards according to the WFD 4579 (2000/60/EC). These criteria and standards not only are a reflection of differences in use of data on 4580 environmental fate and ecotoxicology of PPPs, but also on different policy decisions about the 4581 acceptance of risks in relation to formulated protection goals. 4582

The Water Framework Directive aims to maintain and improve the aquatic environment in EU 4583 Member States so that a ‘good ecological status’ and a ‘good chemical status’ is achieved. For a good 4584 status the WFD requires that Environmental Quality Standards (EQSs) are met. These EQSs are one of 4585 the instruments to evaluate water quality and serve as a benchmark to decide whether or not specific 4586 measures are required. A distinction is made in the AA-EQS and the MAC-EQS. The AA-EQS is the 4587 annual average environmental quality standard that aims to protect aquatic ecosystems and organisms 4588 from effects due to long-term exposures. The MAC-EQS is the maximum acceptable concentration 4589 environmental quality standard that aims to protect aquatic ecosystems and organisms from short-term 4590 concentration peaks. The methodology for EQS derivation is described in the ‘Technical Guidance for 4591 Deriving Environmental Quality Standards under the Water Framework Directive’ (EC, 2011). 4592

During the approval of active substances at EU level the relevant data to derive EQS’s need to be 4593 compiled. The data requirements (SANCO document 11802/2012) state that “All of the aquatic 4594 toxicity data shall be used when developing a proposal for environmental quality standards (Annual 4595 Average EQS, AA-EQS; Maximum Acceptable Concentration EQS, MAC-EQS). The methodology for 4596 derivation of these endpoints is outlined in the “Technical Guidance for Deriving Environmental 4597 Quality Standards” for the Water Framework Directive 2000/60/EC.” It should be noted that in 4598 proposing an EQS, the endpoints from some standard and higher-tier studies are interpreted and used 4599 in a different way to that conventionally used in risk assessments. 4600

9.5.2. Overview of main differences in risk assessment procedures between Plant Protection 4601 Product Regulation and Water Framework Directive 4602

Chemical context 4603 Under the umbrella of the WFD EQSs are derived for all toxic chemicals (e.g. metals and all types of 4604 organic pollutants) that are identified as problematic in (one of the) European river basins. In contrast, 4605 Regulation 1107/2009/EC exclusively deals with the risk assessment of plant protection products used 4606 (or intended to use) in the EU. 4607



Protection goals 4608 The protection goals underlying the WFD refer to human and ecosystem health. Within the context of 4609 ecosystem health and EQS setting it is assumed that, (1) ecosystem sensitivity depends on the most 4610 sensitive species (population), and, (2) protecting ecosystem structure protects community function. 4611 EQSs are derived on basis of Predicted No Effect Concentrations (PNECs)for all relevant populations 4612 of water organisms (comparable to the ecological threshold option). Although the generic protection 4613 goals of the WFD and Plant Protection Product Regulation do not differ substantially, the specific 4614 protection goals of the Plant Protection Product Regulation do not exclude that under certain 4615 conditions short-term effects followed by recovery are acceptable (ecological recovery option), while 4616 EQS setting within the context of the WFD in principle is based on the ecological threshold option. 4617

Geographical context 4618 The aquatic risk assessment procedure according to the WFD has its focus on, usually larger, water 4619 bodies within the context of river basins. The aquatic risk assessment procedure under the umbrella of 4620 the Plant Protection Product Regulation has its focus on edge-of-field surface waters in agricultural 4621 landscapes. In some EU Member States a clear differentiation of non-WFD and WFD-water bodies is 4622 implemented. 4623

Exposure assessment 4624 The risk assessment according to the WFD follows a retrospective approach. The products evaluated 4625 are already used (placed on the market) and form potential problems in one or more European water 4626 basins. Comparing chemical monitoring data, based on analysis of discrete chemical monitoring 4627 samples, with the AA-EQS and MAC-EQS is the means by which compliance is assessed. According 4628 to the ‘Technical Guidance for Deriving Environmental Quality Standards under the Water 4629 Framework Directive’ EQSs should be linked to an annual average concentration (AA-EQS) or the 4630 maximum of the measured concentrations (MAC-EQS). 4631

The risk assessment procedure according to the Plant Protection Product Regulation follows a 4632 prospective approach. This approach allows to assess the risks of a PPP before it is placed on the 4633 market. A common, cost-effective approach in the prospective exposure assessment is the use of 4634 harmonised exposure scenarios (FOCUS Surface Water Scenarios). These scenarios, in combination 4635 with models that estimate the emissions to and the fate and behaviour of PPPs in surface waters, intend 4636 to predict realistic worst-case exposure concentrations in edge-of-field surface waters. The RACsw; ac is 4637 compared with the PECsw;max and the RACsw;ch is in first instance compared with the PECsw;max and 4638 under certain conditions with the PECsw;twa. The time-window for the PECsw;twa usually is smaller than 4639 the duration of the standard toxicity test that triggered the risk in Tier 1. The way of linking exposure 4640 to effects is substantially different in the WFD approach. 4641

Effect assessment 4642 In the effect assessment the PPP Regulation follows a tiered approach while the EQS derivation 4643 according to the WFD follows a weight of evidence approach. The main differences in effect 4644 assessment approaches concern (1) the use of toxicity data for algae and macrophytes, (2) the use of 4645 additional toxicity data e.g. to construct SSDs and (3) the way how micro-/mesocosm are used. 4646

1. In the WFD approach the EC50 values for algae and macrophytes are used in the acute effect 4647 assessment and the NOEC/EC10 values in the chronic effect assessment. In the Tier 1 effect 4648 assessment of the PPP regulation the EC50 values of algae and macrophytes are used (in both 4649 the acute and chronic assessment). 4650

2. If besides the base set additional toxicity data are available it is possible to apply the SSD 4651 approach. To apply the SSD approach the procedures developed for the PPP Regulation require 4652 at least 5 (fish) to 8 (plants and invertebrate) toxicity data for different taxa of the sensitive 4653 taxonomic group. The SSD approach developed to derive WFD EQSs requires at least 10 4654 toxicity values for at least 8 different taxonomic groups. So in first instance the specific toxic 4655 mode of action of the PPP will not be considered when constructing the SSD for EQS setting. 4656 The SSD procedure developed for RAC derivation (PPP Regulation) and WFD EQS derivation 4657



is based on calculating the HC5 and by applying an AF. The height of the AF, however, differs 4658 for RAC derivation and EQS derivation. 4659 If the number of additional toxicity data is less than required for the SSD approach the WFD 4660 methodology will select the lowest toxicity value to derive the EQS by applying an AF. For 4661 RAC derivation the Geomean approach may be used. 4662

3. For EQS derivation the threshold levels for effects derived from appropriate micro-/mesocosm 4663 tests may be used by applying an appropriate AF. For RAC derivation on basis of appropriate 4664 micro-/mesocosm tests both the ‘ecological threshold option’ and the ‘ecological recovery 4665 option’ may be followed. In RAC derivation the predicted exposure profile for the edge-of-4666 field surface water of concern plays a prominent role when interpreting results of 4667 micro/mesocosm studies. 4668

A detailed description of EQS derivation procedures can be found in the ‘Technical Guidance for 4669 Deriving Environmental Quality Standards under the Water Framework Directive’ (EC, 2012). In 4670 Brock et al. (2011) a description is given how the EQS derivation can be performed for PPPs, while 4671 also the differences between the WFD and PPR Regulation procedures are explained in greater detail. 4672

4673

10. Addressing uncertainties 4674

Case studies will be performed with the draft guidance. The uncertainties identified then will be 4675 addressed in the update of the draft guidance following the public consultation. 4676

10.1. Approaches for characterising uncertainty in higher-tier assessments 4677

Regulation (EC) 546/2011 states that no authorisation shall be granted unless it is “clearly established” 4678 that no unacceptable impact occurs. The term ‘clearly establish’ implies a requirement for some 4679 degree of certainty. First tier assessments use standardised scenarios and decision rules which are 4680 designed to provide an appropriate degree of certainty. Higher tier assessments are not standardised, 4681 and so the degree of certainty they provide has to be evaluated case by case. The need for risk 4682 assessments to include characterisation of uncertainty has also been emphasised at senior policy levels 4683 in the EU30 (see also Sterling 2010). 4684

Methods for characterising uncertainty can be grouped into three main types: 4685

• Qualitative methods: using words to describe the certainty of an outcome, or to describe how 4686 different the true outcome might be compared to an estimate. 4687

• Deterministic methods: generating deterministic quantitative estimates of impact for a range of 4688 possible scenarios. This shows the range of possible outcomes (e.g. a range of ETRs) and can be 4689 accompanied by qualitative descriptions of their relative probabilities (traditional ‘worst-case’ 4690 assessments are an example of this). 4691

• Probabilistic methods: these give numeric estimates of the probabilities of different outcomes. 4692 These probabilities may be estimated statistically (e.g. when quantifying measurement or sampling 4693 uncertainty, or as outputs from probabilistic modelling). However, they may also be estimated 4694 subjectively, by expert judgement. 4695

All uncertainties affecting an assessment should be considered at least qualitatively. To reduce the risk 4696 of overlooking important uncertainties, it is recommended to systematically consider each part of the 4697 assessment (e.g. different lines of evidence, different inputs to calculations, etc.) and list all of the 4698 sources of uncertainty together with a description of the magnitude and direction of their potential 4699 influence on the expected level of impact. As well as evaluating each individual source of uncertainty, 4700 it is also essential to give an indication of their combined effect. It is recommended to use a tabular 4701 30E.g. “Even though it is not a subject that lends itself easily to quantification, I would urge you to take account of the risk

manager’s need to understand the level of uncertainty in your advice and to work towards a systematic approach to this problem.” (Madelin, 2004).



approach to facilitate and document this process, as illustrated in Table 10.1. This is based on an 4702 approach used in some EFSA opinions (EFSA 2005c; 2007a; 2007b; 2008a), but adapted to increase 4703 clarity by introducing separate columns to describe uncertainties that act in different directions. 4704

Research in social science has shown that there is a general tendency for experts to underestimate 4705 uncertainties. It is therefore important that risk assessors should be aware of the potential magnitude of 4706 common uncertainties in the assessment of risks to non-target organisms. For example, assessors 4707 should be aware of the potential magnitude of measurement uncertainties (e.g. number of dead bees in 4708 dead bee traps and of the potential magnitude of sampling uncertainty associated with small and 4709 moderate sized datasets. 4710

In some cases, a qualitative evaluation of uncertainties may be sufficient to establish clearly (i.e. with 4711 sufficient certainty) that unacceptable levels of impact will not occur, as is required by the ‘unless’ 4712 clause in Regulation (EC) 546/2011. In other cases, a purely qualitative evaluation of uncertainty may 4713 not give a sufficiently clear picture of the range of possible outcomes. In such cases, one option is to 4714 obtain additional data to reduce uncertainty. This may usefully be targeted on the uncertainties that 4715 appeared largest in the qualitative evaluation. However, an alternative option is to refine the 4716 characterisation of the uncertainties progressively, by evaluating some of them using first 4717 deterministic methods and then, if necessary, probabilistic methods. This implies a tiered approach to 4718 the treatment of uncertainties, which starts by evaluating all uncertainties qualitatively and progresses 4719 either by reducing uncertainty (by obtaining additional data) or by refining the evaluation of selected 4720 uncertainties (either deterministically or probabilistically), until the point where it can be ‘clearly 4721 established’ whether an unacceptable impact will occur (as required by the ‘unless clause in 4722 Regulation (EC) 546/2011). 4723

Table 10.1: Tabular approach recommended for qualitative evaluation of uncertainties in refined 4724 assessments. The +/- symbols indicate whether each source of uncertainty has the potential to make 4725 the true risk higher (+) or lower (-) than the outcome of the refined assessment. The number of 4726 symbols provides a subjective relative evaluation of the magnitude of the effect (e.g. +++ indicates an 4727 uncertainty that could make the true risk much higher). If the effect could vary over a range, lower and 4728 upper evaluations are given (e.g. + / ++). If possible, the user should indicate the meaning of different 4729 numbers of symbols (e.g. two symbols might be used to represent a factor of 5, and three symbols a 4730 factor of 10). See section 3 for some practical examples. 4731

Source of uncertainty

Potential to make true risk

lower

Explanation Potential to make true risk higher

Explanation

Concise description of first source of uncertainty

Degree of negative

effect (e.g. - - -)

Short narrative text explaining how this factor could make true risk lower

Second source of uncertainty

Degree of positive effect

(e.g. +++)

Short narrative text explaining how this factor could make true risk lower

Add extra rows as required for additional sources of uncertainty

- Note: many uncertainties may act in both positive and negative directions

+

Overall assessment Narrative text describing the assessor’s subjective evaluation of the overall degree of uncertainty affecting the assessment outcome, taking account of all the uncertainties identified above. The overall assessment should be a balanced judgement and not simply a summation of the plus and minus symbols.



It is unlikely that it will ever be practical – or necessary – to quantify all uncertainties, so every 4732 deterministic or probabilistic assessment should be accompanied by a qualitative evaluation of the 4733 unquantified uncertainties. Also, it should be remembered that deterministic and probabilistic methods 4734 often require assumptions (e.g. about distribution shapes) that are themselves uncertain, and these 4735 additional uncertainties should be included in the qualitative evaluation. Therefore, every refined 4736 assessment should contain at least a qualitative evaluation of uncertainties. 4737

The overall magnitude of uncertainty associated with an assessment will often be very large. This 4738 should not be regarded as implying a failure of risk assessment; on the contrary, it provides essential 4739 information for decision-making (Madelin 2004; Sterling 2010). 4740

It should be noted that for PPPs where several different types of refined assessment are used, the 4741 uncertainties affecting each one will be different. In such cases it is recommended to evaluate the 4742 uncertainties affecting each approach separately. The contribution of the multiple assessment 4743 approaches (multiple lines of evidence) in reducing overall uncertainty can then be evaluated by 4744 weight-of-evidence in the final risk characterisation (see next section). 4745

In summary, it is recommended that: 4746

• Every refined risk assessment should be accompanied by at least a qualitative evaluation of 4747 the uncertainties affecting it, using a systematic tabular approach. In assessments with 4748 multiple lines of evidence, the uncertainties affecting each line of evidence should be 4749 evaluated separately. 4750

• In cases where qualitative evaluation of uncertainty is not sufficient to determine whether it is 4751 clearly established that no unacceptable impact will occur, the assessor may either (a) seek 4752 further data to reduce the uncertainty, or (b) refine the evaluation of the existing uncertainties 4753 using quantitative methods (which can be either deterministic or probabilistic). 4754

10.2. Risk characterisation and weight-of evidence assessment 4755

Risk characterisation is the final step of risk assessment. At this point, all relevant information or 4756 evidence that has been gathered is used to produce an overall characterisation or description of the 4757 risk, in a form that is suitable for decision-making. 4758

To be useful for decision-making, the risk characterisation should focus on evaluating whether the 4759 relevant protection goals are satisfied for the PPP under assessment. 4760

It is therefore recommended to adopt a tiered approach to risk characterisation, as follows: 4761

1. First, to consider whether the evidence provided by the risk assessment is sufficient to satisfy the 4762 protection goal of making any mortality or reproductive effects unlikely. If so, it can be assumed 4763 there is also a high certainty that no visible mortality or long-term repercussions, nor short-term 4764 population effects will occur. This is a more conservative criterion than is implied by the ‘unless’ 4765 clause, but it is more practical to assess and enables firm conclusions to be reached without 4766 requiring more precise definition of the ‘unless’ clause criteria. 4767

2. If the evidence does not satisfy the protection goal of making any effects unlikely, then attention 4768 should shift from establishing the lack of effects to assessing the levels of mortality and 4769 reproductive effects that may occur, as well as their implications for the likelihood of mortality 4770 and long-term repercussions on abundance and diversity. It should be recognised that the 4771 additional uncertainty inherent in this more complex assessment may make it difficult to meet the 4772 Regulation (EC) 546/2011 criterion of ‘clearly establish’. 4773

Often, risk characterisation will involve combining several different types of refined assessment, each 4774 providing a separate indication of the risk. For example, an applicant might submit a refined dietary 4775 exposure assessment, together with some avoidance studies. These need to be integrated in an overall 4776



risk characterisation that takes appropriate account of each, so as to provide the best basis for decision-4777 making. This process of combining available ‘lines of evidence’ to form an integrated conclusion or 4778 risk characterisation is frequently referred to as ‘weight-of-evidence’ assessment (e.g. EC, 2002; Hull 4779 and Swanson, 2006). This term reflects the principle that the contribution of each line of evidence 4780 should be considered in proportion to its weight. 4781

In the context of this document, a line of evidence might be the completed output of any of the 4782 refinement options as described in section 6 and 7. 4783

The PPR Panel recommends a qualitative31 approach to weight-of-evidence assessment, as follows: 4784

• Consider all relevant lines of evidence, including the first tier assessment. Retention of the first 4785 tier assessment is appropriate in all cases, as it is relevant to consider whether it was borderline or 4786 failed by a large margin. 4787

• Evaluate the uncertainties associated with each line of evidence. This should be done by applying 4788 the approaches described in the preceding section to each line of evidence separately. The 4789 characterisation of overall uncertainty for each line of evidence is then used in the weight-of-4790 evidence assessment, as in principle the weight given to each line of evidence should be 4791 proportionate to its certainty. 4792

• Form overall conclusions by using expert judgement to combine all lines of evidence, weighted 4793 according to their certainty, and give more weight to the most certain, but also take due account of 4794 the less certain. High certainty implies high weight. If one line of evidence implies a much 4795 narrower range for the risk than another line of evidence (i.e. higher certainty), then the true risk is 4796 most likely to fall inside the range of the former. 4797

• Be sure to take full account of the uncertainties and to include a fair description of the range of 4798 possible outcomes in the final risk characterisation. Identify the outcome that is considered most 4799 likely, but do not give it more emphasis than is justified by the evidence. 4800

• If different lines of evidence conflict (e.g. a high ETR but no effects in a field study), this should 4801 be considered a form of uncertainty. No line of evidence should be completely discounted unless it 4802 is wholly invalid or irrelevant. Instead, as stated above, each line of evidence should contribute to 4803 the overall conclusion in proportion to its certainty. 4804

• If the overall characterisation of risk is expressed qualitatively, choose words very carefully to 4805 describe the outcome and its uncertainty as clearly as possible. For example the phrase ‘on 4806 balance’ is often used to focus on one of several possible outcomes, e.g. “on balance, it is 4807 concluded there will be no mortality”. This type of statement is not appropriate, because it fails to 4808 communicate the degree of certainty (e.g. ‘on balance’ could mean 51 % certainty, or 99 %)32. 4809

• A weight-of-evidence assessment is inevitably subjective. Different assessors may vary in their 4810 weighing of the evidence, especially when uncertainty is high. Therefore, it is essential to 4811 document the assessment in detail, including the outcome and uncertainty for each lines of 4812 evidence considered, and explaining how they were combined to reach conclusions about the 4813 overall outcome and its uncertainty. 4814

The PPR Panel recommends a systematic tabular approach to documenting the weight-of-evidence 4815 assessment, such as that illustrated in Table 10.2. The tabular format provides a concise yet clear 4816 summary of the lines of evidence considered and how they were combined. It also helps the reader to 4817 evaluate whether the assessment was balanced, and aids consistency of approach between PPPs. 4818

31Quantitative approaches could also be used to combine lines of evidence, but this requires each line of evidence to be

expressed in the same units together with a quantitative measure of its certainty. 32Note that the standard of evidence required by the ‘unless’ clause is ‘clearly establish’, which is much stronger than ‘on

balance’.



It should be noted that Table 10.2 summarises the major types of uncertainty for each line of evidence, 4819 and not just the overall uncertainty. This is recommended because it helps the assessor to take account 4820 of some important strengths and weaknesses of different types of refined assessment (EFSA, 2009c). 4821

The subjectivity of weight-of-evidence assessment can impede the formation of an independent view 4822 when this is based on the assessment of another person. Therefore, when a weight-of-evidence 4823 assessment is submitted by an applicant, it would be prudent for the regulatory authority to conduct 4824 their own weight-of-evidence assessment separately, compare their conclusion with that of the 4825 applicant, and consider the reasons for any differences. 4826

It is sometimes objected that characterising uncertainty is unhelpful in decision-making. In fact, it is 4827 essential for risk assessors to characterise uncertainty, as is clear from Directive 91/414/EEC (‘clearly 4828 establish’) and from policy statements by the European Commission (Madelin, 2004). Furthermore, 4829 practical options exist for dealing with uncertainty in decision-making. Two of the principal options 4830 are to request more data to reduce uncertainty, or to request more refined evaluation or analysis of the 4831 existing uncertainty. A third option is to counter the uncertainty by applying risk mitigation options, so 4832 that the chance of adverse impacts is limited to an acceptable level33. However, choosing between 4833 options for dealing with uncertainty involves risk management considerations outside the scope of this 4834 document such as the acceptability of effects, the degree of certainty required and potentially other 4835 factors such as the cost and time required for further refinement, the need to respect legal deadlines for 4836 authorisations, and the consequences of risk mitigation or non-authorisation (e.g. reduced efficacy, 4837 reduced choice of pest control options in agriculture, risk of resistance, etc.). 4838

In summary, the PPR Panel recommends that: 4839

• Every refined risk assessment should conclude with an overall characterisation of risk, in terms 4840 relevant for decision-making. It is recommended to begin with the consideration of whether the 4841 evidence makes any mortality or reproductive effects unlikely (the surrogate protection goal). 4842 Where this is not satisfied, attention should turn to characterising the levels of mortality and 4843 reproductive effects that may occur, and using this to evaluate whether there is a high certainty of 4844 no mortality and no long-term repercussions on abundance and diversity (the actual protection 4845 goal). 4846

• The overall characterisation of risk should be derived by a qualitative weight-of-evidence 4847 assessment considering all relevant lines of evidence and their uncertainties using a systematic 4848 tabular approach (e.g. Table 10.2). If the overall characterisation is expressed qualitatively (in 4849 words) rather than quantitatively, great care should be taken to describe the outcome and its 4850 uncertainty as clearly as possible. 4851

• The first tier assessment should always be included as one of the lines of evidence, and given 4852 appropriate weight (this will be higher for acute risks of sprayed PPPs than for other types of 4853 assessment). 4854

33“In cases where both the potential risk and scientific uncertainties are high, the risk manager may conclude that a

precautionary approach is appropriate.” (Madelin, 2004).



Table 10.2: Tabular approach recommended for qualitative weight-of-evidence assessment, 4855 summarising the conclusion and uncertainties for several lines of evidence and using them to develop 4856 an overall conclusion. See section 3, Table 9 for a practical example. The +/- symbols indicate 4857 whether each source of uncertainty has the potential to make the true risk higher (+) or lower (-) than 4858 the indicated outcome. The number of symbols provides a subjective relative evaluation of the 4859 magnitude of the effect (e.g. - - - might indicate an uncertainty that could reduce risk by an amount 4860 equivalent to reducing a ETR by about a factor of 10). If the effect could vary over a range, lower and 4861 upper evaluations are given (e.g. - / ++ or + / ++). 4862

Lines of evidence (add more columns if appropriate)

First tier assessment (should always be included)

Second line of evidence

Add one column for each line of evidence

Main contributions to uncertainty:

Concise description of first major source of uncertainty

+ and – symbols (see legend)

Second uncertainty Add one row for each major source of uncertainty

Conclusions for individual lines of evidence

Insert overall assessment for each line of evidence

Overall conclusion Insert overall conclusion giving appropriate weight to each line of evidence, taking account of their relative certainty (more uncertainty = less weight). The overall conclusion should be a balanced judgement and not simply a summation of the plus and minus symbols.

4863 4864

10.3. Uncertainties in extrapolating to real field situations 4865

To meet the requirements of a protective approach in risk assessment (RA) it is necessary to not only 4866 predict, but to also retrospectively monitor the effects of pesticides in the field. As done for every 4867 credible model going beyond a “Gedankenexperiment” (theoretical exercise). Available information 4868 on unacceptable effects of pesticides in the field indicates that care has to be taken using the RA 4869 approach used until now as effects have been identified. This includes: 4870

• The EU/SETAC workshop EPiF identified that „Effects of pesticides were identified in several of 4871 the field studies“ (Liess et al., 2005) 4872

• In Australia pesticide effects on invertebrates were identified in a stream following runoff 4873 (Muschal and Warne, 2003) 4874

• In Germany effects of pesticides on an invertebrate community were observed following runoff. 4875 Possible confounding factors as hydrodynamic stress were experimentally excluded (Liess and 4876 Schulz, 1999). Additionally, in several other streams pesticide effects were identified (Liess and 4877 von der Ohe, 2005). 4878

• In South Africa Dabrowski et al. (2001) identified pesticide effects on a stream. 4879 • A recent meta-study of Schäfer et al. (2012) identified pesticide effects in 8 data sets comprising 4880

exposure data and effect data. 4881 4882



Possible reasons for the effects observed, and solutions to solve the problems identified, could be: 4883

• Non-compliance with good agricultural practice and risk mitigation measures so that the exposure 4884 is higher than predicted. 4885

• In the current exposure assessment, certain exposure routes might not be covered. 4886 • Effects are not covered by the current prospective risk assessment (first tier or higher tier). 4887 • The possible shortcomings of the prospective risk assessment procedure should, at least in part, be 4888

addressed by an appropriate implementation of the Water Framework Directive and the 4889 Sustainable Pesticide Use Directive. If in chemical monitoring programs certain pesticides 4890 (regularly) exceed the WFD water quality standards (AA-EQS; MAC) this should have 4891 consequences for the re-registration if it can be demonstrated that the problems are caused by a 4892 specific pesticide use, or other appropriate measures should be taken (e.g. punishing the farmers 4893 that do not follow label instruction; implementing more strict risk mitigation measures). The 4894 results presented above may give important evidence that in practice, the RA based on existing 4895 methodology is in some cases not protective enough for aquatic non-target communities in edge-4896 of-field water bodies. If there is some concern about the safety of a product, competent authorities 4897 could request chemical as well as biological post-authorisation monitoring in edge-of-field water 4898 bodies. Results from this monitoring could be considered in relation to the authorisation of the 4899 substance. Note that WFD environmental quality standards are stricter than the RACs, since the 4900 WFD also aims to protect aquatic communities in larger surface waters and different procedures in 4901 the linking of exposure to effects are used. It may be a future option in the registration procedure 4902 to always require a chemical (and biological) monitoring program in edge-of-field surface waters 4903 for a few years if a new substance is placed on the market. Appropriate monitoring programmes 4904 can, however, only be set up once the exposure assessment goals are defined. 4905

• Environmental stress may alter effects of toxicants on populations and communities by a factor of 4906 more than 10. Examples include investigations by Foit et al. (2012), Knillmann et al. (2012a), 4907 Liess and Beketov (2011), Reynaldi et al. (2011), and Stampfli et al. (2011). In addition to the 4908 effect, also recovery may be influenced (Foit et al. 2012; Knillmann et al. 2012b). 4909

• A mixture of the possibilities mentioned above. 4910 • Surely another plausible explanation for the failure of RA to have predicted/precluded these 4911

effects, is that the effects reported were not due to single pesticide exposure. This is in the realm 4912 of eco-epidemiology, and similar uncertainties apply with regard to isolating potential causative 4913 factors and eliminating confounding factors. Examples within the framework that certainly will 4914 underestimate the risk are tank mixtures, while multiple exposure (serial application of several 4915 PPP, multi-crop, multi-year) and long-term delayed effects of pesticides may be underestimated in 4916 the RA procedure described in this GD. 4917

4918

Establishing a firm link between the exposure concentration of a single pesticide and its effect in the 4919 field faces the problem that mostly several pesticides occur simultaneously in streams. However, in 4920 small agricultural streams (and other edge-of-field surface waters) one or two substances are in most 4921 of the cases strongly dominating the toxicity at the same time and/or in sequence. This has been shown 4922 in field experiments using realistic application rates of the total package of pesticides used in a wheat 4923 crop (Auber et al. 2011), a potato crop (Arts et al. 2006) and a tulip crop (Van Wijngaarden et al. 4924 2004) and in field monitoring studies of agricultural areas in Australia (Muschal & Warne, 2003), 4925 Germany (Liess and Schulz, 1999; Liess and von der Ohe, 2005), France (Schäfer et al. 2007) and the 4926 USA (Belden et al. 2007). Hence, observed effects are in many cases related to the effects of one or 4927 two active substance(s) (that may be characterised by repeated pulse exposure) in one year. This may 4928 often concern substances that differ in toxic mode-of-action (e.g. insecticide and fungicide) if effects 4929 are caused by more active substances. 4930

4931

4932



Conclusions: 4933

• Long-lasting effects of pesticides were identified in several field studies. Therefore, as mentioned 4934 above, uncertainty should be taken into account and if there is some concern about the safety of a 4935 product, competent authorities could request chemical and possibly biological post-authorisation 4936 monitoring in edge-of-field water bodies. 4937

• As long as there is no guidance on how to include additional stressors into the RA we need to 4938 reflect the related uncertainty. Hence it should be realized that in the first tier an AF of 100 for 4939 acute and 10 for chronic toxicity may not be protective in 100% of the cases. Consequently this 4940 will also be the case for all effect tiers. Therefore, there is a need to validate / calibrate the RA 4941 scheme to the field situation. When there is a systematic deviation for substances with a specific 4942 mode of action, this should trigger a revision of the RA scheme. 4943

• Care should be taken when extrapolating from tier 1 and higher tiers such as mesocosms 4944 (substitute reference tiers) to the field, which is the ultimate reference tier. This extrapolation will 4945 benefit from ecological modelling (when available in the future) including all relevant processes 4946 necessary for extrapolation (including environmental stress and biological interaction). 4947

4948

Research needed: 4949

• Field investigations need to exemplarily verify exposure and effect predictions. See also Artigas et 4950 al. (2012). 4951

• The link between results obtained by 1st tier tests with the situation in the field need to be 4952 strengthened. This includes investigations on the degree to which effects from single species tests 4953 can be altered by the environmental context. 4954

• The link between results obtained by mesocosms with the situation in the field need to be 4955 strengthened. This includes investigations on the degree to which effects from field and mesocosm 4956 can be altered by the environmental context. 4957

4958

4959



11. Executive Summary 4960

11.1. Aquatic Risks due to toxicity 4961


The tiered effect assessment procedure, and proposals how to link effect to exposure estimates, 4963 presented in this Aquatic Guidance Document has its focus on aquatic organisms living in the water 4964 column of edge-of-field surface waters. For these organisms, the concentration of the freely dissolved 4965 Plant Protection Product (PPP) is chosen as the Ecotoxicologically Relevant Concentration (ERC). 4966 This GD also presents the Tier 1 effect assessment procedure for sediment-dwelling organisms when 4967 based on water spiked water-sediment toxicity tests. 4968

The aquatic risk assessment is the combination of the exposure and the effect assessments and there is 4969 considerable interaction between these assessments. This guidance document assumes that the current 4970 exposure assessment procedure (FOCUS surface water scenarios and models) is continued to be used 4971 at the EU level for approval of active substances and does not include further guidance for the 4972 exposure assessment. To date the Panel did not evaluate the current exposure assessment procedure. 4973 The level of protection achieved by the exposure assessment procedure is not clear. The overall level 4974 of protection of aquatic organisms is determined by the combination of the Specific Protection Goals 4975 for the organisms and the exposure assessment goals. Since the exposure assessment methodology was 4976 not revised in parallel to the effect assessment scheme, definitions for exposure assessment goals are 4977 not clear. The PPR Panel advises to critically evaluate and improve the surface water exposure 4978 assessment. 4979

Two distinct effect assessment schemes are identified that respectively start with the Tier 1 acute and 4980 the Tier 1 chronic toxicity data set (see Figure 11.1 for a schematic overview; for a more detailed 4981 description see sections 2.3 and 2.4). The acute and chronic effect assessment schemes address the 4982 same Specific Protection Goals (SPGs). These SPGs overall aim to protect aquatic plants and animals 4983 at the population level in surface water. However, the SPG selected for aquatic vertebrates aims 4984 protection at the individual level, so that visual mortality and suffering due to acute toxicity is avoided 4985 (see for more details section 3.5). 4986

4987 Figure 11.1: Schematic presentation of the tiered approach within the acute (left part) and chronic 4988 (right part) effect assessment for PPPs. For each PPP both the acute and chronic effects/risks have to 4989 be assessed. PECsw = Predicted Environmental Concentration in surface water; RACsw = Regulatory 4990 Acceptable Concentration for surface water; ac= acute; ch= chronic; max = predicted peak 4991 concentration; twa= predicted time weighted average concentration. 4992 4993

4994 4995 4996 4997 4998

4999 5000 5001 5002 5003 5004 5005 5006

5007

5008 5009 5010 5011 5012 5013

5014 5015 5016 5017 5018

EFSA Jou

AlthoughToxicokmainly Acceptabsubject o

To protederive Rnegligiblpopulatiosection 3principleapproachpotential

11.1.2. The basiin edge-effect/rischaptersinformatpesticide

DecisionChaptersdecision

urnal 2013;vo

h in the effekinetic/Toxico

updates guible Concentrof a future sc

ect populatioRACs on basle populationon level eff3.5). In the e all tiers arh (Tier 3) ul vulnerable

Summary

ic flow chart-of-field sursk assessme/sections of tion on sciene both the acu

n scheme s/sections of

n making, as w

olume(issue):N

ects assessmeodynamic (Tidance on erations (RAC

cientific opin

ons of aquatisis of two on effects onlfects if ecoltiered acute

re able to adunder certainspecies) is ab

flow charts

ts for the tierface water ent is presthe GD are ntific rationaute and chro

A: Summaf the Aquatiwell as its sc

NNNN

ent schemes TK/TD), popexperimental Cs). Effect mnion of the PP

ic organismsoptions, viz. ly, and (2) Togical recovand chronic

ddress the En conditionsble to also ad

for acute an

red proceduris presented

sented in Dmentioned w

ale, can be onic effect/ris

ary flow ic Guidance cientific unde

DRAF

presented inpulation and

approachesmodels in thePR Panel of

s, effect asse(1) The Ec

The Ecologicvery takes pc effect assecological Th

s (e.g. possiddress the Ec

nd chronic e

re to assess d in DecisioDecision scwhere detailfound. Note

sk assessmen

chart for Document

erpinning, ca

T GD on tiere

n Figure 11.1food-web m

s (Tiers 1, e aquatic riskEFSA.

essment scheological Thrcal Recoveryplace within ssment schehreshold Optibility to excological Re

effect/risk as

acute effectson scheme

cheme B. Iled guidancee that in thent procedures

acute effecare mention

an be found.

ed RA for edg

1 also modelmodels) are in

2 and 3) tk assessmen

emes are devreshold Optiy Option (ER

an acceptabmes presenttion, while t

xtrapolate obcovery Optio

ssessment

s/risks of preA. The prn these sc

e for decisione registrations have to be p

ct/risk assesned where d

ge-of-field surf

lling proceduncorporated,o derive Re

nt of PPPs w

veloped that ion (ETO), aRO), acceptible time pered in Figurethe model ecbserved respon.

edicted PPP rocedure forchemes the n making, an procedure performed.

ssment of detailed guid

face water

129

ures (e.g. , this GD egulatory

will be the

allow to accepting ing some riod (see

e 11.1, in cosystem

ponses to

exposure r chronic

specific as well as

for each

toxicity. dance for

5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047

EFSA Jou

DecisionChaptersdecision In the aexposurePECsw;maexposureeffects oby shortunder cewater). Apresente

1.

2.

3.

urnal 2013;vo

n scheme s/sections of

n making, as w

acute risk ase occurring ax (predictede profile. Inof pesticide et-term exposertain conditiA decision d below.

Acute Risk AIs PECsw;maxYes: AssessNo: Low ac

Chronic AssIs PECsw;max

Yes:No:

Is the RACsexample, th

olume(issue):N

B: Summaf the Aquatiwell as its sc

ssessment, thwithin a re

d peak expon the chronicexposure thature or long-ions with thescheme whe

Assessment x (of highest sment not sucute risk, go

sessment x (of highest a: Go to 3 Low chroni

sw;ch based onhis may conce

NNNN

ary flow chic Guidance cientific unde

he RACsw;acelatively shoosure concenc risk assesst develop slo-term exposue PECsw;twa (en to use th

available tierufficient to do to 2

available tier

ic risk

n short-term ern laborator

DRAF

hart for chDocument

erpinning, ca

(RAC in suort period afntration in sment, the Rowly and/or hure) is in firs(predicted timhe PECsw;max

r) > RACsw;ademonstrate

r) > RACsw;ch

tests and orgry tests with

T GD on tiere

hronic effecare mention

an be found.

urface water fter exposuresurface wat

RACsw;ch (RAhave a long-lst instance cme-weightedx or the PE

ac (of highest e a low risk

h (of highest

ganisms withunicellular a

ed RA for edg

ct/risk assened where d

for adversee) is alwayster) derived AC in surfaclasting cours

compared wid average conCsw;twa in th

available tie

available tie

h a generationalgae)

ge-of-field surf

ssment of detailed guid

e effects of s compared

from the pce water forse and that arth the PECswncentration i

he risk asses

er)?

er)?

n time < 3 da

face water

130

toxicity. dance for

pesticide with the predicted r adverse re caused w;max, and is surface ssment is

ays? (For



Yes: Go to 4 5048 No: Go to 6 5049

5050 4. Is the observed effect caused by reduction of growth (e.g. not an algicidal effect)? 5051



Yes: Low risk not demonstrated 5056 No: Low risk demonstrated 5057

5058 6. Is the RACsw;ch derived from a long-term test (≥ 7 days) in which (i) loss of the active 5059

substance from water is more that 20% of nominal at the end of the exposure period and (ii) 5060 the toxicity estimate (e.g. EC10 or NOEC) is expressed in terms of nominal/initially measured 5061 concentration of the active substance? (For example, this may concern the 28-d water-spiked 5062 Chironomus test and tests with macrophytes) 5063


5066 7. Is the RACsw;ch based on treatment-related responses of the relevant test species early in the 5067

chronic test (e.g. mortality is observed during the initial 96 h or the time to onset of more than 5068 50% of the effect occurs within 7 days in the treatment level above the one from which the 5069 RACsw;ch is derived) 5070


5073 8. Is it demonstrated by the notifier that for the organisms at risk and the PPP under evaluation 5074

and/or PPP with a similar toxic mode-of-action (read-across information), the following 5075 phenomena are not likely: (i) latency of effects due to short-term exposure; (ii) the co-5076 occurrence of exposure and specific sensitive life stages that last a short time only. 5077



Yes: Go to 8 5082 No: Low risk demonstrated 5083

5084 10. Are experimental (or TK/TD modelling when guidance is available) data available 5085

that demonstrate that for the species at risk a larger time window for the TWA PEC may be 5086 used (not exceeding the duration of the Tier 1 chronic test that triggered the risk)? 5087

Yes: Go to 7 and replace the PECsw;7d-twa by another appropriate PECsw;twa 5088 No: low risk not demonstrated 5089

5090

For invertebrates, fish and macrophytes, a default 7-day time window is proposed for the PECsw,twa if 5091 the TWA approach is deemed acceptable. It may be justified to lengthen or shorten the default 7-d 5092 TWA period of the PEC if justified with appropriate scientific data (for further details see section 5093 2.5.1). 5094

In the sections below a concise summary will be presented of Effect Tiers 1 to 3. However, before 5095 starting the acute and chronic effect assessment for RAC derivation it is essential to gather information 5096 on the predicted exposure profiles for the pesticide of concern in the relevant edge-of-field water 5097 bodies on basis of FOCUS surface water (see chapter 4) and/or Member State specific exposure 5098 scenarios. This is in particular of importance when evaluating refined exposure tests (Tier 2C) and 5099 micro-/mesocosm experiments (Tier 3), since these studies are able to address the effects of the 5100



predicted field exposure profile in a more realistic way. The rationale for this is presented in section 5101 7.1. 5102

11.1.3. Tier 1 RACsw derivation on basis of standard test species 5103

The specific data requirements for Regulation (EC) 1107/2009 concerning the placing of plant 5104 protection products on the market are laid down in SANCO document 11802/2012 for the dossier to 5105 be submitted for the approval of active substances contained in plant protection products and in 5106 SANCO document 11803/2012 for the authorisation of plant protection products. The obligatory 5107 toxicity tests that should be provided for pesticides are presented below in Tables 11.1-11.3. A more 5108 detailed description of the Tier 1 data requirements is presented in chapter 5. Note that in the tables 5109 below the tests with algae and macrophytes are placed under the chronic risk assessment since these 5110 test comprise the complete, or a large part of, the life-cycle of these organisms, although the toxicity 5111 endpoints selected is the EC50. 5112

5113

Table 11.1: The obligatory toxicity tests that should be provided for pesticides with an insecticidal 5114 mode-of-action. 5115 Standard test species Duration Endpoint RAC

Acute effect assessment

- Daphnia sp. (D. magna preferred) - Chironomus sp. or Americamysis bahia - Oncorhynchus mykiss

48 h 48 h 96 h

EC50 EC50

LC50

EC50/100 EC50/100 LC50/100

Chronic effect assessment

- Daphnia sp. or additional arthropod (a) - Chironomus sp. (water spiked test preferred) (b) - Early life stage test (ELS) or full life-cycle test (FFLC) with fish (c) - Green alga (e.g. Pseudokirchneriella subcapitata)

21 d 20-18 d - 72 h

EC10 (NOEC) EC10 (NOEC) EC10 (NOEC) ErC50

(d)

EC10/10 EC10/10 EC10/10 ErC50/10

(a) Preferably the most sensitive standard test arthropod (Daphnia, Chironomus, Americamysis) should be selected as test 5116 species in the chronic effect assessment and obligatory if in the acute assessment this species is a factor of 10 more sensitive. 5117 (b) Obligatory only if the substance accumulates in sediment and/or when the substance interferes with moulting hormones 5118 (e.g. insect growth regulators) 5119 (c) A fish FFLC is always required if there are indications that the substance has endocrine mediated effects. If a FFLC is 5120 available a fish ELS needs not to be supplied 5121 (d) In other EU directives/regulations (e.g. WFD) the EC50 and an AF of 100 is used in the acute effect assessment and the 5122 EC10 (or NOEC) and an AF of 10 in the chronic effect assessment. 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143



Table 11.2: The obligatory toxicity tests that should be provided for pesticides with a herbicidal 5144 mode-of-action. 5145 Standard test species Duration Endpoint RAC


- Daphnia sp. (D. magna preferred) - Oncorhynchus mykiss

48 h 96 h

EC50 LC50

EC50/100 LC50/100


- Green alga (e.g. Pseudokirchneriella subcapitata) - Additional non-green alga (e.g. diatom Navicula pelliculosa) - Lemna sp or Myriophyllum sp. or Glyceria maxima (a)

- Daphnia sp. - Early life stage test (ELS) or full life-cycle test (FFLC) with fish (b) - Chironomus sp. or Lumbriculus sp. (water spiked test preferred) (c)

72 h 72 h 7 – 14 d 21 d - 20-28 d

ErC50 (d)

ErC50

(d) EC50

(d) EC10 (NOEC) EC10 (NOEC) EC10 (NOEC)

ErC50/10 ErC50/10 EC50/10 EC10/10 EC10/10 EC10/10

(a) Usually Lemna sp. will be the default macrophyte test species. In case Lemna and algae are apparently not sensitive to the 5146 herbicidal product (e.g. EC50 > 1mg/L), or if the herbicide simulates a plant growth hormone, a rooted macrophyte is 5147 preferred (preferably Myriophyllum). It is advised to test Glyceria in case of a herbicide that primarily affects monocots in 5148 terrestrial plant trials. 5149 (b) A fish FFLC is always required if there are indications that the substance has endocrine mediated effects. If a FFLC is 5150 available a fish ELS needs not to be supplied 5151 (c) Substance accumulates in sediment 5152 (d) In other EU directives/regulations (e.g. WFD) the EC50 and an AF of 100 is used in the acute effect assessment and the 5153 EC10 (or NOEC) and an AF of 10 in the chronic effect assessment. 5154 5155 5156 Table 11.3: The obligatory toxicity tests that should be provided for other pesticides. 5157 Standard test species Duration Endpoint RAC


- Daphnia sp. (D. magna preferred) - Oncorhynchus mykiss

48 h 96 h

EC50 LC50

EC50/100 LC50/100


- Green alga (e.g. Pseudokirchneriella subcapitata) - Daphnia sp. - Early life stage test (ELS) or full life-cycle test (FFLC) with fish (a) - Chironomus sp. or Lumbriculus sp (water spiked test preferred) (b)

72 h 21 d - 20-28 d

ErC50 (c)

EC10 (NOEC) EC10 (NOEC) EC10 (NOEC)

ErC50/10 EC10/10 EC10/10 EC10/10

(a) A fish FFLC is always required if there are indications that the substance has endocrine mediated effects. If a FFLC is 5158 available a fish ELS needs not to be supplied 5159 (b) Substance accumulates in sediment 5160 (c) In other EU directives/regulations (e.g. WFD) the EC50 and an AF of 100 is used in the acute effect assessment and the 5161 EC10 (or NOEC) and an AF of 10 in the chronic effect assessment. 5162 5163

11.1.4. Tier 2 RACsw derivation on basis of additional laboratory toxicity tests 5164

If besides the basic data requirements presented in Tables 11.1-11.3 additional laboratory toxicity tests 5165 are provided a Tier 2 Effect assessment may be performed. In the GD three different Tier 2 effect 5166 assessments are described, viz.; 5167

• Tier 2A: The Geomean-AF approach (see section 6.2 for a detailed description) 5168 • Tier 2B: The Species Sensitivity Distribution (SSD) approach (see section 6.3) 5169 • Tier 2C: Refined Exposure Laboratory test-AF approach (see section 7.2) 5170



11.1.5. Tier 2A: The Geomean-AF approach 5171

The PPR panel of EFSA advices to apply the Geomean-AF approach as summarised in Table 11.4. In 5172 this approach the geomean L(E)C50 or geomean NOEC/EC10 values for species belonging to the same 5173 taxonomic group (e.g. crustaceans, insects, fish, green algae, diatoms, macrophytes) are calculated and 5174 the AF of the Tier 1 effect assessment is applied. 5175

Table 11.4: Derivation of RACs for aquatic organisms when a limited number of additional single 5176 species toxicity tests is available. 5177 Taxonomic group Number of toxicity data

for different taxa of the relevant taxonomic group

RAC Field exposure concentration (PEC) for risk assessment

Acute risk assessment Aquatic vertebrates(a) < 5 acute LC50’s Geomean LC50/100 (e) PECsw;max Invertebrates (b) < 8 acute EC50’s Geomean EC50/100 (e) PECsw;max Primary producers (c) < 8 EC50’s Geomean EC50/10 (e) PECsw;max

Chronic risk assessment Aquatic vertebrates(a) < 5 chronic EC10’s

(or chronic NOECs) Geomean EC10/10 (e) PECsw;max or PECsw;twa

Invertebrates (b) < 8 chronic EC10’s (or chronic NOECs)

Geomean EC10/10 (e) PECsw;max or PECsw;twa

Primary producers(c) < 8 EC50’s (d) Geomean EC50/10 (e) PECsw;max (a) i.e. fish or amphibians 5178 (b) i.e. crustaceans or insects for insecticides, and a more specific taxonomic group for fungicides 5179 (c) i.e. green algae, diatoms, blue-green algae or macrophytes for herbicides or fungicides with a herbicidal mode of action. 5180 (d) In other EU directives/regulations (e.g. WFD) the EC50 and an AF of 100 is used in the acute effect assessment and the 5181

EC10 (or NOEC) and an AF of 10 in the chronic effect assessment. 5182 (e) Of the different taxonomic groups the lowest Geomean value is selected (e.g. the lowest value for insects or crustaceans 5183

in case of insecticides; the lowest value for green algae, diatoms, blue-green algae or macrophytes in case of herbicides) 5184 5185 5186

11.1.6. Tier 2B: The Species Sensitivity Distribution (SSD) approach 5187

The PPR panel of EFSA advices to apply the SSD approach as summarised in Table 11.5. In this 5188 approach the median HC5 (Hazardous Concentration to 5% of the tested species) and the Lower Limit 5189 HC5 values (see section 6.3.1) are derived from the SSD curve that are constructed with at least 8 5190 representative toxicity data for different non-vertebrate species or with at least 5 representative 5191 toxicity data for different fish and/or amphibian species. For RACsw;ac derivation acute toxicity data of 5192 the relevant sensitive taxonomic groups should be used to construct the SSD. If the specific toxic 5193 mode-of action of the compound will likely result in latent affects (e.g. as demonstrated for some 5194 insect growth regulators) the SSD should be constructed with EC50 values derived from prolonged 5195 acute toxicity tests. In prolonged acute toxicity tests the observation of treatment-related responses is 5196 continued after the test organisms are transferred to clean medium. Chronic toxicity data of the 5197 relevant sensitive taxonomic groups should be used to construct the SSD for RACsw;ch derivation (for 5198 selection of toxicity data see section 6.3.2). In Tables 11.5 and 11.6 a distinction is made between 5199 RAC derivation for vertebrates and non-vertebrates on basis of the SSD approach, since a higher 5200 protection level is required in the acute effect assessment for fish and amphibians (avoidance of visual 5201 mortality and suffering due to pesticide exposure). 5202

5203

5204

5205



Table 11.5: Derivation of RACs in edge-of-field surface waters, based on hazardous concentrations 5206 derived from Species Sensitivity Distributions with aquatic invertebrates and/or plants. Note that the 5207 L(E)C50 values used to construct the SSD concern acute toxicity data for animals and chronic toxicity 5208 data for plants 5209


Relevant PEC

Hazardous concentration

AF to derive RAC from HC5

Acute and chronic effect/risk assessment for single and repeated pulse exposure (FOCUS step 3 or 4)

PECsw;max Latency of effects not expected(a) Median (acute) HC5 (based on LC50 or EC50 data) (b)

and/or Lower limit (acute) HC5 (based on LC50 or EC50 data) (b) Latency of effects expected Median (acute) HC5 (based on acute LC50 or EC50 data from prolonged acute toxicity tests)

and/or Lower limit (acute) HC5 (LLHC5) (based on acute LC50 or EC50 data from prolonged acute toxicity tests)

or precautionary approach instead of the two options above by using a chronic SSD (see below)

3 1

3

1

Chronic effect/risk assessment with long-term exposure



and/or Lower limit chronic HC5 (LLHC5) (based on chronic NOEC and/or EC10 data)

3

1 (a) This has to be demonstrated by the applicant, see further section 2.5.1. For example, by read-across for substances with 5210

similar toxic mode of action, prolonged acute toxicity tests, and information from micro/mesocosm studies for similar 5211 compounds with a longer-term observation period after exposure. 5212

(b) For types of pesticides evaluated by Maltby et al. (2005; 2009) and Van den Brink et al. (2006). 5213 5214 5215

Table 11.6: Derivation of RACs in edge-of-field surface waters, based on hazardous 5216 concentrations derived from Species Sensitivity Distributions with fish (and other aquatic 5217 vertebrates). 5218 Type of effect/risk assessment

Relevant PEC

Hazardous concentration

AF to derive RAC from HC5


PEC sw;max Latency of effects not expected(a) Median acute HC5 (based on 96 h NOEC and/or acute LC10 data)

and/or Lower limit acute HC5 (based on 96 h NOEC and/or acute LC10 data)

or Median acute HC5 (based on 96h LC50 or EC50 data)

and/or Lower limit acute HC5 (based on 96 h LC50 or EC50 data)

or If latency of effects is expected go to chronic effect assessment (see below)

3

1

9

3




PEC sw;max or PECsw;twa


or Lower limit chronic HC5 (based on chronic NOEC and/or EC10 data)

3

1

(a) This has to be demonstrated by the applicant, see further section 2.5.1. For example, by read-across for substances 5219 with similar toxic mode of action, prolonged acute toxicity tests, and information from micro/mesocosm studies for 5220 similar compounds with a longer-term observation period after exposure. 5221

5222

11.1.7. Tier 2C: Refined Exposure Laboratory test-AF approach 5223

The PPR panel of EFSA proposes to explore a higher tier RAC derivation on basis of the Refined 5224 Exposure Laboratory-AF approach if predicted (modelled) exposure profiles for edge-of-field surface 5225 waters differ considerably from exposure regimes in standard toxicity studies and if the PECsw;twa 5226 cannot be used in the chronic risk assessment (for a more detailed description see section 7.2). Refined 5227 Exposure Laboratory tests usually are performed with the Tier 1 standard test species that drive the 5228 aquatic risks, and are designed in such a way that the exposure in these test more realistically resemble 5229 the field exposure conditions. Nevertheless, in order to consider them as an appropriate higher-tier 5230 effect assessment approach, the refined exposure tests should simulate a realistic-worst case exposure 5231 relative to that predicted for the edge-of-field, as well as should be long enough to allow the 5232 expression of the maximum effects. In acute risk assessments this usually requires prolonged acute 5233 refined exposure toxicity tests. In chronic risk assessments the duration of the refined exposure tests 5234 usually is similar to that of the chronic Tier 1 standard toxicity test, but may be larger (e.g. in case of 5235 tests with algae). Note that the RACs derived from refined exposure toxicity tests should always be 5236 expressed in terms of peak exposure concentration in these tests, and that these RACs should always 5237 be compared with the PECsw;max. 5238

A summary of the RACsw;ac and RACsw;ch derivation on basis of refined exposure laboratory tests, and 5239 their use in the risk assessment, is presented in Table 11.7. 5240

Table 11.7: Derivation of RACs in edge-of-field surface waters, based on refined exposure laboratory 5241 toxicity tests with standard test species. 5242


Relevant PEC

Endpoint of refined exposure toxicity test with standard test species expressed in terms of peak exposure concentration in test system

RAC


PECsw;max L(E)C50 (animal tests)

L(E)C50/100


PECsw;max EC50 (plant tests) (a) EC10 / NOEC (animal tests)

EC50/10

EC10/10 (a) In other EU directives/regulations (e.g. WFD) the EC50 and an AF of 100 is used in the acute effect assessment and the 5243

EC10 (or NOEC) and an AF of 10 in the chronic effect assessment. 5244 5245 5246 5247 5248 5249 5250 5251



11.1.8. Tier 3 RACsw derivation on basis of micro-/mesocosm tests 5252

The requirements for the conduct and interpretation of micro-/mesocosm tests for RACsw derivation 5253 are described in detail in section 7.3. To evaluate the scientific reliability of micro-/mesocosm 5254 experiments the following questions should be addressed: 5255

1. Is the test system adequate and does the test system represent a realistic freshwater community? 5256 [Trophic levels; taxa richness and abundance of (key and sensitive) species; representativeness 5257 of the biological traits with respect to vulnerability] 5258

2. Is the description of the experimental set-up adequate and unambiguous? [ANOVA or 5259 regression design; overall characterization of the experimental ecosystem/community 5260 simulated; measurement endpoints; sampling frequency; sampling techniques] 5261

3. Is the exposure regime adequately described? [Method of application of the test substance; 5262 relevance for predicted exposure profile in the field; concentration in the application solution; 5263 dynamics in exposure concentrations in relevant compartments (e.g. water, sediment); 5264 detection limits] 5265

4. Are the investigated endpoints sensitive and in accordance with the working mechanisms of the 5266 compound, and with the results of the first tier studies? [Compare selected measurement 5267 endpoints with the species potentially at risk as indicated by the lower tiers] 5268

5. Is it possible to evaluate the observed effects statistically and ecologically? [Univariate and 5269 multivariate techniques applied; unambiguous concentration-response relationships; statistical 5270 power of the test; ecological relevance of the statistical output]. 5271

5272 Furthermore, the criteria for RAC derivation from micro-/mesocosm tests on basis of the Ecological 5273 Theshold Option (ETO) and the Ecological Recovery Option (ERO) are presented below in Decision 5274 scheme C. In Decision scheme C reference is made to Effect class concentrations for the most 5275 sensitive measurement endpoints derived from micro-/mesocosm tests. The following Effect classes 5276 are important in the derivation of ETO-RACsw and ERO-RACsw values. 5277

5278 Effect class 1 (No treatment-related effects demonstrated; the endpoints with the lowest Effect class 1 5279 concentrations may be used to derive the overall NOECmicro/mesocosm). 5280

No (statistically and/or ecologically significant) effects observed as a result of the treatment. 5281 Observed differences between treatment and controls show no clear causal relationship. 5282

Effect class 2 (Slight effects). 5283 Effects concern a short-term and/or quantitatively restricted response usually observed at 5284 individual samplings only. 5285

Effect class 3A (Pronounced short-term effects (< 8 weeks), followed by recovery) . 5286 Clear response of endpoint, but full recovery of affected endpoint within 8 weeks after the 1st 5287 application or, in case of delayed responses and repeated applications, the duration of the 5288 effect period is less than 8 weeks and followed by full recovery. Effects observed at some 5289 subsequent sampling instances. 5290

5291 A more elaborate description of all Effect classes is given in section 7.3.3.1. 5292

The PPR Panel of EFSA proposes the procedure presented in Tables 11.8 and 11.9 to derive an ETO-5293 RACsw on basis of micro-/mesocosm tests. 5294

The PPR Panel of EFSA proposes the procedure presented in Tables 11.10 and 11.11 to derive an 5295 ERO-RACsw on basis of micro-/mesocosm tests. 5296

5297



5298

Decision scheme C: Summary flow chart for the derivation of RACs from appropriate micro-5299 /mesocosm experiments on basis of the ecological threshold option (ETO-RAC) or ecological 5300 recovery option (ERO-RAC) 5301

5302



Table 11.8: Acute risk assessment: Derivation of the RACsw;ac addressing the ecological threshold 5303 option on basis of an appropriate micro/mesocosm experiment. 5304 5305

Assessment factor for ETO-RACsw;ac derivation (ecological threshold option)



(1 –) 2(a) Based on expert judgement by considering the criteria mentioned in sections 7.2 and 7.3 and ecological information on the type of edge-of-field surface water at risk

PECsw;max



PECsw;max


5309 5310 5311 Table 11.9: Chronic risk assessment: Derivation of the RACsw;ch addressing the ecological threshold 5312 option on basis of an appropriate micro/mesocosm experiment. 5313 5314

Assessment factor for ETO-RACsw;ch derivation (ecological threshold option)

Field exposure concentration to compare with the RACsw;ch


(1 –) 2 (a) Based on expert judgement by considering the criteria mentioned in section 7.3 and ecological information on the type of edge-of-field surface water at risk



(1 –) 2 (a) Based on expert judgement by considering the criteria mentioned in section 7.3 and ecological information on the type of edge-of-field surface water at risk

PECsw;max








PECsw;max


5318 5319 5320 5321 Table 11.10: Acute risk assessment: Derivation of the RACsw;ac addressing the ecological recovery 5322 option on basis of an appropriate micro/mesocosm experiment. 5323

Assessment factor for ERO-RACsw;ac derivation (ecological recovery option)


Effect class 3A concentration Is rate of dissipation of the active ingredient in test system realistic to worst-case when compared to that predicted for the field? Yes: Base effect estimate on nominal or measured peak concentration in test system. No: Base effect estimate on e.g. the initial 48h TWA concentration in test system or, apply appropriate extrapolation techniques or, consider the ecological threshold option


PECsw;max


5327

5328

5329

5330

5331

5332



Table 11.11: Chronic risk assessment: Derivation of the RACsw;ch addressing the ecological recovery 5333 option on basis of an appropriate micro/mesocosm experiment. 5334

Assessment factor for ERO-RACsw;ch derivation (ecological recovery option)

Field exposure concentration to compare with the RACsw,ch

Effect class 3A concentration Based on time weighted average concentration in test system during the application period.



Effect class 3A concentration Based on nominal or peak concentration in test system if the long-term exposure regime (e.g. due to repeated pulses) is realistic to worst-case compared to the predicted field exposure profile


PECsw;max


5338

11.2. Bioconcentration and Secondary Poisoning 5339

Some compounds in the water have the tendency to accumulate in the tissue of fish or in the tissue of 5340 other organisms. This tendency of a compound is often expressed in a bioconcentration factor (BCF). 5341 The equilibrium concentration for a compound in fish can be estimated by multiplying the 5342 concentration of the compounds in the surrounding water by the fish BCF for that particular 5343 compound. At long exposure times (equilibrium), the BCF also equals the ratio of the uptake constant 5344 (Mackay, 1982). 5345

Bioaccumulation often correlates with lipophilicity. Thus, for organic chemicals, a log Pow of ≥3 5346 indicates a potential for bioaccumulation. The stability of a compound is another indicator for 5347 accumulation. The compound is considered stable when less than 90% loss of the original substance 5348 over 24 hours via hydrolysis has been noted (see section 5.7). 5349

The regulatory acceptable concentrations for secondary poisoning (RACsp) for birds and mammals 5350 eating fish out of the surface water contaminated with a PPP can be assessed in the following way (see 5351 section 5.7.3): 5352

fish

mammal

fish

birdSP BCF0.1385

NOAELorBCF159.05

NOAELRAC××××

= 5353

with 5354

RACsp = Regulatory Acceptable Concentration in water for secondary poisoning [mg/L] 5355

NOAEL = Relevant long-term no-adverse-effect-level for birds or mammals [mg/kg body weight per 5356

d] 5357

BCFfish = whole body bioconcentration factor in fish [L/kg] 5358

5 is the assessment factor; 0.159 and 0.138 are multiplication factors based on a 3000-g mammal 5359 eating 415 g fish per day and a 1000-g bird eating 159-g fish per day. 5360



11.3. Non-testing methods 5361

Guidance, largely following ECHA recommendations (ECHA, 2008), is provided on use of non-5362 testing methods in PPP risk assessment, such as (Q)SAR (Quantitative Structure-Activity 5363 Relationship), expert test systems and analogue read-across as tools for deriving intrinsic properties of 5364 substances. Non-testing methods may be used to estimate endpoints for metabolites without the 5365 toxophore and for impurities. In addition, QSARs might together with available test data be used to 5366 rank species for identifying the most likely sensitive taxonomic group to focus experimental testing 5367 (EFSA, 2012a). For a detailed description of non-testing methods see chapter 8. 5368

Only suitable models (e.g. covering the right domain) with a high predictive reliability should be used 5369 (see section 8.1.3). This should among others be reflected in the level of statistical significance 5370 required for estimates from (Q)SAR models. Validation parameters should ideally indicate good fits 5371 (e.g. Q2 > 0.7, ccc > 0.85)34. Estimates of toxicity should where possible35 be assisted by confidence 5372 intervals around the prediction. In case the standard derivation exceed the predicted value it self, such 5373 values should not be accepted. Generally, the worst case endpoint from several modelling approaches 5374 should be used. 5375

Estimates should be confirmed by using weight-of-evidence approaches where all available 5376 information is taken into account. This could include a combination of the different (Q)SAR model 5377 predictions combined with read across and other available information like non-standard testing data 5378 and toxicodynamic/ toxicokinetic information from mammals. 5379

To date, most experience is gained with QSAR models that predict acute toxicity. It is noted that less 5380 valid QSAR models are currently available for deriving chronic toxicity data. 5381

A decision scheme for use of non-testing systems is presented below. 5382

1. Is the QSAR model valid – i.e. is it relevant and reliable (follow 5 OECD principles for 5383 assessing QSAR models). E.g. is prediction accurate enough (recommended assessment 5384 values Q2, CCC, SD) (see sections 8.1.2.1 and 8.1.2.2) 5385 Yes: Go to 2 5386 No: QSAR not applicable – consider other model 5387 5388

2. Does the substance and model match - i.e. is the chemical of interest within the scope of the 5389 model, according to the defined applicability domain of the model and is the defined 5390 applicability domain suitable for the regulatory purpose? (see sections 8.1.2.1 and 8.1.2.2) 5391 Yes: Go to 3 5392 No: QSAR not applicable – consider other model 5393 5394

3. Does model predict take into account test information? (E.g. for aquatic toxicity consider 5395 water solubility, log Kow, degradability and volatility) (see section 8.1.5) 5396 Yes: Go to 4 5397 No: QSAR not applicable – consider other model 5398 5399

4. Are reliable estimations available from more than one QSAR model (see section 8.1.5) 5400 Yes: use lowest predicted QSAR endpoint in risk assessment or as qualifier for testing if 5401 confirmed by weight of evidence approach 5402 No: Single value may be used in risk assessment or as qualifier for testing if clearly 5403 confirmed by weight of evidence approach 5404 5405

5406 5407 34 For further details consult ECHA (2008) guidance. 35 Not all QSAR models provide standard derivations for predictions.



11.4. Metabolites and degradation products 5408

The panel has developed an assessment scheme for risk assessment of metabolites where metabolites 5409 for which it is clearly shown that the toxophore is lost in a first step can be assessed using 5410 approximation of toxicity (see section 8.2.7) while testing is required for metabolites with remaining 5411 toxophore (see section 8.2.6). The scheme has been developed in order to facilitate the selection of the 5412 most appropriate and pragmatic assessment route for metabolites. However, possible endocrine 5413 disruption properties should be addressed separately (see section 5.6). 5414

5415 1. Is the exposure to the metabolite in the toxicity test with the a.s. adequate for assessing the 5416

potential effect of the metabolite (see section 8.2.4)? 5417 Yes: Go to 2 5418 No: Go to 3 5419

5420 2. Perform the risk assessment assuming all the effect observed in the test with the a.s. can be 5421

attributed to the metabolite (see section 8.2.4). Is RACsw;ac > PECsw and RACsw;ch > PECsw? 5422 Yes: low risk 5423 No: Go to 3 5424

5425 3. Is it clear that the toxophore has been lost from the molecule (see section 8.2.5)? 5426

Yes: Go to 6 5427 No or unclear: Go to 4 5428 5429

4. Determine the toxicity to species or taxonomic group36 providing the lowest Tier 1 RACsw;ac 5430 of the a.s. Is the acute metabolite LC50 > 10 times the a.s. LC50 (on a molar basis) (see 5431 section 8.2.6)? 5432

Yes: Go to 6 5433 No: Go to 5 5434 5435 5436

5. Determine the toxicity to species or taxonomic group6 providing the lowest Tier 1 RACsw;ch of 5437 the a.s. Is RACsw;ac > PECsw and RACsw;ch > PECsw ? 5438


6. Assume that the acute and chronic37 toxicity of the metabolite is equal to the toxicity of the 5442 a.s. for all first tier taxonomic groups (see section 8.2.7)? Is RACsw;ac > PECsw and RACsw;ch 5443 > PECsw? 5444

Yes: low risk 5445 No: Go to 7 5446 5447

7. Are reliable and adequate non testing predictions of toxicity (see section 8.1) available for all 5448 first tier taxonomic groups (fish, plants and invertebrates)? Use prediction to assess if 5449 RACsw;ac > PECsw and RACsw;ch > PECsw ? 5450


36 Consider testing with Chironomidae if metabolite is distributed in sediment, or other taxonomic group suspected to be most

sensitive (e.g. Chironomidae with IGRs) 37 if chronic risk assessment is triggered by fate properties of the metabolite



5453 8. Determine the acute and chronic7 toxicity is for those taxonomic groups where a valid non-5454

testing predictions of toxicity is not available or for which a risk was identified using 5455 predicted toxicity? Is RACsw;ac > PECsw and RACsw;ch > PECsw ? 5456

Yes: low risk 5457 No: Consider higher tier refinement 5458

5459

For the assessment of the metabolite the applicant has to provide a reasoned case whether the molecule 5460 contains a toxophore or if it has been lost following transformation. In case it cannot be clearly shown 5461 that the toxophore is not present in the molecule it should be assumed that the toxophore remains and 5462 that the molecule has a specific mode of action (see assessment scheme above). 5463

For metabolites with remaining toxophore testing can in a first step be limited to the taxonomic group 5464 that was identified to result in the lowest Tier 1 RACsw;ac and RACsw,ch for the a.s. If, however, testing 5465 with this taxonomic group shows that this taxonomic group is not sensitive (i.e. the acute end point is 5466 > a factor 10 higher as compared to the parent, on a molar basis38) then the risk assessment needs to be 5467 continued assuming that the most sensitive taxonomic group is unknown and the risk to all taxonomic 5468 groups should be addressed. If it is unclear whether the toxophore remains and the most sensitive 5469 group is not known, then the risk assessment needs to address all taxonomic groups. 5470

If it is clear that the toxophore has been lost from the metabolite, in most cases metabolites are less 5471 toxic to the target organisms than the active substances. As a pragmatic and conservative approach for 5472 metabolites without the toxophore the estimates of exposure could be compared with the RACparent 5473 based on the most sensitive endpoint of the active substance in the relevant compartment. In general 5474 only if this trigger is failed the toxicity needs to be further addressed. For metabolites which have lost 5475 the toxophore the acute and long-term hazard and risk can be addressed using non-testing predictions 5476 of toxicity (see further in section 8.1). If the trigger is failed using predicted toxicity then testing is 5477 required, see below. 5478

For metabolites which require experimental studies (see assessment scheme above), acute toxicity 5479 tests with Daphnia and rainbow trout and an alga should be conducted. In general the same testing 5480 scheme as for active substances (see table 5.1) is required. Hence testing on additional species (e.g. 5481 additional invertebrate species) may be necessary where the risk to a particular taxonomic group for 5482 the a.s. is considered to be of concern (e.g. additional testing with macrophyte for a herbicidal 5483 metabolite). 5484

In principle, for metabolites found in the sediment of a water-sediment study, the same triggers for 5485 testing should be applied to metabolites as for the active substance. However as a first screening step 5486 for metabolites partitioning to the sediment a formula based on equilibrium partitioning theory as 5487 outlined in the TGD part II (EC 2003) section 3.5.3, can be used to indicate if actual testing is needed. 5488 Only if a risk is indicated using this formula actual testing with sediment organisms should be 5489 required. This will be further addressed in a PPR Panel Opinion on sediment risk assessment under 5490 development. 5491

38 The statement to check whether the LC50 of the metabolite is greater than 10 times the LC of the a.s. on a molar basis

means:

50 10 50

where LC50met and LC50ai are mass concentrations (mg/L) of metabolite and active ingredient at 50% mortality and Mmet and

Mai are the molar masses (g/mol) of the metabolite and active ingredient.



In order to decide whether chronic testing is necessary the intended uses and the fate and behaviour of 5492 the metabolite should be taken into account. In general chronic/long term tests are required for 5493 metabolites where exposure of surface water is likely and the metabolite is deemed to be stable in 5494 water, as defined in the data requirements, i.e. there is less than 90% loss of the original substance 5495 over 24 hours via hydrolysis under relevant pH conditions (SANCO document 11802/2012). 5496

In terms of the choice of taxonomic group(s) to be studied for chronic toxicity, this should take 5497 account of any acute toxicity data on the metabolite. Where information on the acute sensitivity of fish 5498 and invertebrates for a particular metabolite is available, chronic testing should only be required on the 5499 more sensitive standard test species (i.e. a factor of 10 more sensitive than the other standard test 5500 species). If Daphnia is suspected to be insensitive based on the mode of action of the active ingredient, 5501 e.g. it is an insect growth regulator or a neonicotinoid, then it is necessary to conduct a chronic study 5502 using the chironomid Chironomus riparius with the metabolite. 5503

5504

11.5. Combinations of a.s. in formulations (guidance on toxic unit approaches) 5505

5506 The PPR Panel proposes to use of the concentration addition model to assess effects of combinations 5507 of active substances in a formulation. The following equation can be used for deriving a surrogate ECx 5508 or NOEC value for a mixture of active substances with known toxicity assuming concentration 5509 additivity: 5510

5511

ECx (mix) or NOEC (mix) = ⎟⎟⎠

⎞⎜⎜⎝

⎛∑ )..(__

)..(

ix

i

i saNOECorECsaX

5512

5513 Where: 5514

X(a.s.i) = fraction of active substance [i] in the mixture (please note that the sum Σ X(a.s.i) must be 1) 5515

ECx or NOEC(a.s.i) = toxicity value for active substance [i]. 5516

5517 Where the toxicity value of a formulated product with more than one active substance is available, this 5518 value should be compared with the predicted mixture toxicity assuming concentration additivity. A 5519 different form of the equation is used. 5520

5521

)(__1

)..(__)..(

mixNOECorECsaNOECorECsaX

xi ix

i =∑ 5522

5523 X(a.s.i) = fraction of active substance [i] in the mixture (here: formulation) 5524

ECx or NOEC(a.s.i) = acute toxicity value for active substance [i] 5525

ECx or NOEC(mix) = measured acute toxicity value for the mixture (here: formulation) 5526

A greater value on the right side of the equation indicates that the formulation is more toxic than 5527 predicted from the toxicity of the individual components (active substances and co-formulants of 5528 known toxicity). This may be due to, e.g. further toxic co-formulants, toxicokinetic interaction or 5529 synergism/potentiation of effect. It may also reflect the inherent variability of toxicity testing. In all 5530



these cases, the use of the EC50 for the formulation (together with appropriate exposure estimates, see 5531 Step 4) is recommended for the first tier assessment, because it cannot be excluded that such effects 5532 would also occur after exposure of organisms to residues in the environment. In case the measured 5533 acute toxicity of the formulation indicates a higher toxicity than predicted from the toxicity of the 5534 individual components (i.e. more than a factor of 10) then a chronic test on the formulation may be 5535 required, see further section 5.5. 5536

Dismissing the EC50 of the formulation from the risk assessment would only be acceptable at a higher 5537 tier if any observed greater toxicity in the test could be clearly and unambiguously ascribed to a factor 5538 that would not be relevant under environmental exposure conditions. 5539

If, in contrast, the measured toxicity of a formulation is lower than predicted, the predicted mixture 5540 toxicity should be used in the first tier risk assessment, together with appropriate exposure estimates. 5541

The ECx (mix) or NOEC(mix) should be compared to the sum of the concentrations of the a.s. in order 5542 to calculate the ETRmix. As a first approach it is assumed that the PECsw;max of all a.s. present in the 5543 formulation will occur at the same moment and are not separated in time. In case the trigger value is 5544 not met in higher tiers the predicted exposure patterns can be taken into account (see further section 5545 8.3). 5546

In case the endpoint to be used for the risk assessment is associated with different assessment factors 5547 the calculation of the mixture toxicity can be based on the regulatory acceptable concentration (RAC) 5548 and the following formula could be used: 5549

PECA/RACA + PECB/RACB + ….= SUM 5550

If SUM < 1 the risk is low. 5551

The SUM can be calculated in the acute or chronic risk assessment for the same relevant taxonomic 5552 group (i.e. fish, crustaceans, algae and aquatic plants) using the single species test-RAC, Geomean-5553 RAC or SSD-RAC. In case data from micro-/mesocosms are used, then use the overall ETO-RAC. 5554

5555

5556



REFERENCES 5557 5558 Abel PD, 1980. Toxicity of hexachlorcyclohexane (Lindane) to Gammarus pulex; mortality in relation 5559

to concentration and duration of exposure. Freshwater Biology, 10, 251-259. 5560

Adriaanse PI and Beltman WHJ, 2009. Transient water flow in the TOXSWA model (FOCUS 5561 versions): concepts and mathematical description. Statutory Research Tasks Unit for Nature and 5562 the Environment, WOT report 101, Wageningen, The Netherlands. 5563

Aldenberg T, Jaworska JS, 2000. Uncertainty of hazardous concentrations and fraction affeceted for 5564 normal species sensitivity distributions. Ecotoxicology and Environmental Safety, 46,1-18. 5565

Aldenberg T, Jaworska JS and Traas TP, 2002. Normal species sensitivity distributions in probabilistic 5566 ecological risk assessment. In: Posthuma L, Traas TP, Suter GW, (eds). Species Sensitivity 5567 Distributions in Risk Assessment. Boca Raton (FL), CRC Press, pp 49-102. 5568

Alonso Prados E and Novillo-Villajos A, 2010. Ecological characterization of permanent and 5569 ephemeral streams of a typical Mediterranean agricultural landscape (east and southeast of Iberian 5570 peninsula). In: Brock TCM, Alix A, Brown CD, Capri E, Gottesbüren BFF, Heimbach F, Lythgo 5571 CM, Schulz R, Streloke M, (EDs), 2010. Linking Aquatic Exposure and Effects: Risk Assessment 5572 of Pesticides. SETAC Press & CRC Press, Taylor & Francis Group, Boca Raton, FL, pp 288-303. 5573

Ampere 2007 http://www.systemecology.eu/AMPERE/Start.html 5574

Artigas J, Arts G, Babut M, Caracciolo AB, Charles S, Chaumot A, . . . Williams R, 2012. Towards a 5575 renewed research agenda in ecotoxicology. Environmental Pollution, 160, 201-206. 5576

Arts GHP, Buijse-Bogdan LL, Belgers JDM, Van Rhenen-Kersten CH, Van Wijngaarden RPA, 5577 Roessink I, Maund SJ, Van den Brink PJ and Brock TCM, 2006. Ecological impact in ditch 5578 mesocosms of simulated spray drift from a crop protection programme for potatoes. Integrated 5579 Environmental Assessment and Management, 2,105-125. 5580

Ashauer R, Boxall ABA and Brown CD, 2007. New ecotoxicological model to simulate survival of 5581 aquatic invertebrates after exposure to fluctuating and sequential pulses of pesticides. 5582 Environmental Science & Technology, 41(4), 1480-1486. 5583

Auber A, Roucate M, Togola A and Caquet T, 2011. Structural and functional effects of conventional 5584 and low pesticide input crop-protection programs on benthic macroinvertebrates communities in 5585 outdoor pond mesocosms. Ecotoxicology 20: 20142-2055 5586

BBA, 2000. Bekanntmachung über die Abtrifteckwerte, die bei der Prüfung und Zulassung von 5587 Pflanzenschutzmitteln herangezogen werden. (8. Mai 2000) in : Bundesanzeiger No.100, amtlicher 5588 Teil, vom 25. Mai 2000, S. 9879. 5589

Baird DJ and van den Brink PJ, 2007. Using biological traits to predict species sensitivity to toxic 5590 substances. Ecotoxicology and Environmental Safety, 67, 296–301. 5591

Bärlocher F, 1985. The role of fungi in the nutrition of stream invertebrates. Botanical Journal of the 5592 Linnean Society, 91, 83-94. 5593

Barnthouse LW, 2004. Quantifying population recovery rates for ecological risk assessment. 5594 Environmental Toxicology & Chemistry, 23, 500-508. 5595

Becker RA, Janus ER, White RD, Kruszewski FH and Brackett RE, 2009. Good Laboratory Practices 5596 and Safety Assessments. Environmental Health Perspectives, 117, 482-483. 5597

Belden JB, Gilliom RS, Martin JD and Lydy MJ, 2007. Relative toxicity and occurence patterns of 5598 pesticide mixtures in streams draining agricultural watersheds dominated by corn and soybean 5599 production. Integr Environ Assess Manag 3:90-100. 5600

Beketov MA and Liess M. 2005. Acute contamination with Esfenvalerate and food limitation: Chronic 5601 effects on the mayfly Cloeon dipterum. Environmental Toxicology and Chemistry. 24, (5) 1281–5602 1286. 5603



Beketov MA and Liess M, 2008. Acute and delayed effects of the neonicotinoid insecticide thiacloprid 5604 on seven freshwater arthropods. Environmental Toxicology & Chemistry, 27,461-470. 5605

Belgers JDM, Aalderink GH, Arts GHP and Brock TCM, 2011. Can time-weighted average exposure 5606 concentrations be used to assess the risks of metsulfuron-methyl to Myriophyllum spicatum? 5607 Chemosphere, 85, 1017-1025. 5608

Bergtold M and Dohmen GP, 2011. Biomass or growth rate endpoint for algae and aquatic plants: 5609 relevance for the aquatic risk assessment of herbicides. Integrated Environmental Assessment and 5610 Management, 7,237-247. 5611

Biever RC, Giddings JM, Kiamos M, Annunziato MF, Meyerhoff R and Racke K, 1994. Effects of 5612 chlorpyrifos on aquatic microcosms over a range of off-target drift exposure levels. Brighton Crop 5613 Protection Conference- Pests and Deseases, pp. 1367 - 1372. 5614

Biggs J and Brown CD, 2010. Ecological characterization of water bodies in clay landscapes in the 5615 United Kingdom. Brock TCM et al. (Eds): Linking Aquatic Exposure and Effects: Risk 5616 Assessment of Pesticides. SETAC Press & CRC Press, Taylor & Francis Group, Boca Raton, FL, 5617 p. 304-320. 5618

Blenkinsop S, Fowler HJ, Dubus IG, Nolan BT and Hollis JM, 2008. Developing climatic scenarios 5619 for pesticide fate modelling in Europe. Environmental Pollution, 154, 219-231. 5620

Boesten JJTI, Köpp H, Adriaanse PI, Brock TCM and Forbes VE 2007. Conceptual model for 5621 improving the link between exposure and effects in the aquatic risk assessment of pesticides. 5622 Ecotoxicology and Environmental Safety, 66:291-308. 5623

Brock TCM and Budde BJ, 1994. On the choice of structural parameters to indicate responses of 5624 freshwater ecosystems to pesticide stress. In: Hill IA, Heimbach F, Leeuwangh P, Matthiesen P 5625 (Eds), Freshwater Field Tests for Hazard Assessment of Chemicals Lewis Publishers, Michigan, pp 5626 19-56. 5627

Brock TCM, Arts GHP, Maltby L and Van den Brink PJ, 2006. Aquatic Risks of Pesticides, 5628 Ecological Protection Goals and Common Aims in European Union Legislation. Integrated 5629 Environmental Assessment and Management, 2, E20-E46. 5630

Brock TCM, Maltby L, Hickey CW, Chapman J and Solomon K, 2008. Spatial extrapolation in 5631 ecological effect assessment of chemicals. In: Solomon KR, Brock TCM, De Zwart D, Dyer SD, 5632 Posthuma L, Richards SM, Sanderson H, Sibley PK and Van den Brink PJ (Eds), Extrapolation 5633 Practice for Ecotoxicological Effect Characterization of Chemicals, SETAC Press & CRC Press, 5634 Boca Raton, USA, pp 223 – 256. 5635

Brock TCM, Roessink I, Belgers JDM, Bransen F and Maund SJ, 2009. Impact of a benzoyl urea 5636 insecticide on aquatic macro-invertebrates in ditch mesocosms with and without non-sprayed 5637 sections. Environmental Toxicology and Chemistry, 28,2191-2205. 5638

Brock TCM, Belgers JDM, Roessink I, Cuppen JGM and Maund SJ, 2010a. Macroinvertebrate 5639 responses to insecticide application between sprayed and adjacent non-sprayed ditch sections of 5640 different sizes. Environ. Toxicol. Chem. 29: 1994-2008. 5641

Brock T, Arts G, Belgers D and Van Rhenen-Kersten C 2010b. Ecological characterization of 5642 drainage ditches in the Netherlands to evaluate pesticide stress. In: Brock TCM, Alix A, Brown 5643 CD, Capri E, Gottesbüren BFF, Heimbach F, Lythgo CM, Schulz R and Streloke M, (EDs), 2010. 5644 Linking Aquatic Exposure and Effects: Risk Assessment of Pesticides. SETAC Press & CRC 5645 Press, Taylor & Francis Group, Boca Raton, FL, pp 269-287. 5646

Brock TCM, Alix A, Brown CD, Capri E, Gottesbüren BFF, Heimbach F, Lythgo CM, Schulz R and 5647 Streloke M (Eds), 2010c. Linking Aquatic Exposure and Effects: Risk Assessment of Pesticides. 5648 SETAC Press & CRC Press, Taylor & Francis Group, Boca Raton, FL, 398 pp. 5649

Brock TCM, Arts GHP, Ten Hulscher TEM, De Jong FMW, Luttik R, Roex EWM, Smit CE and Van 5650 Vliet PJM, 2011. Aquatic effect assessment for plant protection products: A Dutch proposal that 5651



addresses the requirements of the Plant Protection Product Regulation and Water Framework 5652 Directive. Alterra Report No 2235. 5653

Brockway DL, Smith PD and Stancil FE, 1984. Fate and effects of atrazine in small aquatic 5654 microcosms. Bulletin of Environmental Contamination and Toxicology, 32, 345-353. 5655

Bundschuh M, Zubrod JP, Kosol S, Maltby L, Stang C, Duester L and Schulz R, 2011. Fungal 5656 composition on leaves explains pollutant-mediated indirect effects on amphipod feeding. Aquatic 5657 Toxicology, 104,32-37. 5658

Campbell PJ, Arnold DJS, Brock TCM, Grandy NJ, Heger W, Heimbach F, Maund SJ and Streloke 5659 M, 1999. Guidance document on higher-tier aquatic risk assessment for pesticides (HARAP). 5660 Brussels (BE): SETAC-Europe, 179 p. ISBN 90-5607-011-8. 5661

Caquet T, Lagadic L and Sheffield SR, 2000. Mesocosms in ecotoxicology. 1. Outdoor aquatic 5662 systems. Reviews of Environmental Contamination & Toxicology, 165, 1-38. 5663

Caquet C, Hanson M, Roucaute M, Graham D and Lagadi L, 2007. Influence of isolation on the 5664 recovery of pond mesocosms from the application of an insecticide. II Benthic macroinvertebrate 5665 responses. Environmental Toxicology and Chemistry, 26 (6),1280–1290. 5666

Carsel RF, Imhoff JC, Hummel PR, Cheplick JM and Donigian Jr AS, 1995. PRZM-3. A Model for 5667 Predicting Pesticide and Nitrogen Fate in the Crop Root and Unsaturated Soil Zones. Users Manual 5668 for Release 3.0. National Exposure Research Laboratory, U.S. Environmental Protection Agency, 5669 Athens, GA, USA. 5670

Chirico N, 2011. Real external predictively of QSAR model is: How to evaluate it? Comparison of 5671 different validation criteria and proposal of using the Concordance Correlation Coefficient. Journal 5672 of Chemical Information and Modeling, 51, 2320-2335. 5673

Chirico C, Papa E and Gramatica P, 2012. Externally Predictive QSAR Models: Thresholds of 5674 Acceptance by Various External Validation Criteria and Critical Inspection of Scatter Plots. Poster, 5675 SETAC Berlin, 2012. 5676

Consonni et al, 2009. Comments on the Definition of the Q2 Parameter for QSAR Validation. Journal 5677 of Chemical Information and Modeling, 49, 1668-1678. 5678

Coors A, Kuckelkorn J, Mammers-Wirtz and Strauss T, 2006. Application of in-situ bioassays with 5679 macrophytes in aquatic mesocosm studies. Ecotoxicology, 15, 583-591. 5680

Crank J, 1967. The mathematics of diffusion. Oxford University Press, Oxford, UK. 5681

Creton S, Clook M and Wheeler JR, in preparation. Application of the threshold approach for acute 5682 fish toxicity testing to plant protection products: a proposed framework. 5683

Crossland NO, Heimbach F, Hill IR, Boudou A, Leeuwangh P, Matthiessen P and Persoone G, 1993. 5684 European Workshop on Freshwater Field tests (EWOFFT), Summary and recommendations. 5685 Workshop held in Potsdam, Germany, June 25 -26, 1992. 5686

Cuppen JGM, Van den Brink PJ, Camps E, Uil KF and Brock TCM, 2000. Impact of the fungicide 5687 carbendazim in freshwater microcosms. I. Water quality, breakdown of particulate organic matter 5688 and responses of macroinvertebrates. Aquatic Toxicology, 48, 233-250. 5689

Daam MA, Crum SJH, Van den Brink PJ and Nogueira AJA, 2008. Fate and effects of the insecticide 5690 chlorpyrifos in outdoor plankton-dominated microcosms in Thailand. Environmental Toxicology & 5691 Chemistry, 27, 2530-2538. 5692

Daam MA, Satapornvanit K, Van den Brink PJ and Nogueira AJA, 2009a. Sensitivity of 5693 macroinvertebrates to carbendazim under semi-field conditions in Thailand: Implications for the 5694 use of temperate toxicity data in a tropical risk assessment of fungicides. Chemosphere, 74, 1187-5695 1194. 5696



Dabrowski JM, Peall SKC, Reinecke AJ, Liess M, Schulz R, 2001. Runoff-related pesticide input into 5697 the Lourens river, South Africa: Basic data for exposure assessment and risk mitigation at the 5698 catchment scale. Water, Air and Soil Pollution, 135(1-4), 265-283. 5699

De Jong FMW, Brock TCM, Foekema EM and Leeuwangh P, 2008. Guidance for summarizing and 5700 evaluating aquatic micro- and mesocosm studies. RIVM Report 601506009, Bilthoven, the 5701 Netherlands, 59 p. 5702

De Lange M, Sala S, Vighi M and Faber J, 2010. A framework for applying ecological vulnerability in 5703 risk assessment. 20th Annual Meeting SETAC Europe (Society of Environmental Toxicology and 5704 Chemistry). 23-27 May 2010, Sevilla, Spain. 5705

Deneer JW, 2000. Toxicity of mixtures of pesticides in aquatic systems. Pest Management Science, 5706 56, 516-520. 5707

DeNoyelles Jr. F, Dewey SL, Huggins DG and Kettle WD, 1994. Aquatic mesocosms in ecological 5708 effects testing: detecting direct and indirect effects of pesticides. In: Graney RL, Kennedy JH, 5709 Rodgers JH jr. (eds) Aquatic mesocosm studies in ecological risk assessment. Lewis Publishers, 5710 Boca Raton, FL, pp 577-603. 5711

Detenbeck NE, Hermanutz R, Allen K and Swift MC, 1996. Fate and effects of the herbicide atrazine 5712 in flow-through wetland mesocosms. Environmental Toxicology & Chemistry, 15, 937-946. 5713

Dijksterhuis J, Van Doorn T, Samson R and Postma J, 2011. Effects of seven fungicides on non-target 5714 aquatic fungi. Water, Air & Soil Pollution, 222, 421-425. 5715

Diepens NJ, Arts GHP, Brock TCM, Van den heuvel-Greve MJ, Smidt H, Van den Brink PJ and 5716 Koelmans AA. 2013. Improved protocols for sediment toxicity testing: A review. Critical Reviews 5717 in Environmental Science and Technology (accepted). 5718

Dimitrov SD, Dimitrova GD, Pavlov TS, Dimitrova N, Patlewicz GY, Niemela J and Mekenyan OG, 5719 2005a. A stepwise approach for defining the applicability domain of SAR and QSAR models. J. 5720 Chem. Inf. Model. 45:839-849. 5721

Duchet C, Coutellec MA, Franquet E, Lagneau C and Lagadic L, 2010. Population-level effects of 5722 spinosad and Bacillus thuringiensis israelensis in Daphnia pulex and Daphnia magna: comparison 5723 of laboratory and field microcosm exposure conditions. Ecotoxicology, 19, 7, 1224-1237. 5724

Duncan J, Hinchcliffe A and Palmquist K, 2009. Evidence of potential long term effects in (aquatic 5725 and terrestrial) invertebrates after short term pulsed exposure, Literature reviews on ecotoxicology 5726 of chemicals with special focus on plant protection products. Reference: FT/EFSA/PPR/2008/01. 5727

EEA, 2009. Small water bodies – Assessment of the status and threats of standing small water bodies. 5728 EEA/ADS/06/001 – Water, Version 1.1. 5729

EC (European Commission), 1997. Council Directive 97/57/EC of 22 September 1997 establishing 5730 Annex VI to Directive 91/414/EEC concerning the placing of plant protection products on the 5731 market. Official Journal of the European Union C 240:1-23. 5732

EC (European Commission), 1992. Council Directive 92/43/EEC of 21 May 1992 on the conservation 5733 of natural habitats and of wild fauna and flora (OJ L 206, 22.7.1992, p. 7). 5734

EC (European Commission), 2000. Directive 2000/60/EC of the European Parliament and of the 5735 Council of 23 October 2000 establishing a framework for Community action in the field of water 5736 policy. OJ L 327/1, 22.12.2000, p. 1-72. 5737

EC (European Commission), 2002. Guidance Document on Aquatic Ecotoxicology in the context of 5738 the Directive 91/414/EEC (SANCO/3268/2001)rev.4 final, 17.11.2002, p. 1-62. 5739



EC (European Commission – DG Health and Consumer Protection), 2002. Guidance Document on 5740 Risk Assessment for Birds and Mammals under Council Directive 91/414/EEC. 5741 SANCO/4145/2000 – final 25 September 2002, pp 74.39 5742

EC (European Commission), 2003. Technical Guidance Document on risk assessment in support of 5743 Commission Directive 93/67/EEC on risk assessment for new notified substances and Commission 5744 Regulation (EC) No 1488/94 on risk assessment for existing substances and Directive 98/8/EC of 5745 the European parliament and of the council concerning the placing of biocidal products on the 5746 market. 5747

EC (European Commission), 2006. Regulation (EC) No 1907/2006 of the European Parliament and of 5748 the Council of 18 December 2006concerning the Registration, Evaluation, Authorisation and 5749 Restriction of Chemicals (REACH),establishing a European Chemicals Agency, amending 5750 Directive 1999/45/EC and repealing Council Regulation (EEC) No 793/93 and Commission 5751 Regulation (EC) No 1488/94 as well as Council Directive76/769/EEC and Commission Directives 5752 91/155/EEC, 93/67/EEC, 93/105/EC and 2000/21/EC. OJ L 136/3, 29.5.2007, p. 1-287. 5753

EC (European Commission), 2008. Regulation (EC) No 1272/2008 of the European Parliament and of 5754 the Council of 16 December 2008 on classification, labelling and packaging of substances and 5755 mixtures, amending and repealing Directives 67/548/EEC and 1999/45/EC, and amending 5756 Regulation (EC) No 1907/2006. EN OJ L 353/1, 31.12.2008, p. 1-1355. 5757

EC (European Commission), 2009. Regulation (EC) No 1107/2009 of the European Parliament and of 5758 the Council of 21 October 2009 concerning the placing of plant protection products on the market 5759 and repealing Council Directives 79/117/EEC and 91/414/EEC. OJ L 309/1, 24.11.2009, p. 1-50. 5760

EC (European Commission), 2011. Guidance Document No. 27 Technical Guidance For Deriving 5761 Environmental Quality Standards; Technical Report - 2011 – 055; Common Implementation 5762 Strategy for the Water Framework Directive (2000/60/EC). 5763

EC (European Commission), 2011. Regulation (EC) No 544/2011 of 10 June 2011 implementing 5764 Regulation (EC) No 1107/2009 of the European Parliament and of the Council as regards the data 5765 requirements for active substances. 5766

EC (European Commission), 2011. Regulation (EC) No 545/2011 of 10 June 2011 implementing 5767 Regulation (EC) No 1107/2009 of the European Parliament and of the Council as regards the data 5768 requirements for plant protection products 5769

EC (European Commission), 2011. Regulation (EC) No 546/2011 of 10 June 2011 implementing 5770 Regulation (EC) No 1107/2009 of the European Parliament and of the Council as regards uniform 5771 principles for evaluation and authorisation of plant protection products 5772

EC (European Commission), 2011. Technical Guidance for Deriving Environmental Quality 5773 Standards, Guidance Document No: 27 under the Common Implementation Strategy for the Water 5774 Framework Directive (2000/60/EC). Technical Report - 2011 – 055. 5775

EC (European Commission), 2011. Guidance Document, Guidelines on comparability, extrapolation, 5776 group tolerances and data requirements for setting MRLs. SANCO 7525/VI/95 - rev.9, March 5777 2011. 5778

EC (European Commission), 2011. Guidance Document on the assessment of the equivalence of 5779 technical materials of substances regulated under Regulation (EC) No 1107/2009. 5780 SANCO/10597/2003 –rev. 9, 17 June 2011. 5781

ECHA, 2008. Guidance on information requirements and chemical safety assessment. Chapter R.7b: 5782 Endpoint specific guidance. Version 1.1. Helsinki, Finland: European Chemicals Agency. 234 p. 5783

EFSA, 2004. Opinion of the Scientific Panel on Plant health, Plant protection products and their 5784 Residues on a request from EFSA on the appropriateness of using the current FOCUS surface 5785

39 Available at: http://ec.europa.eu/food/plant/protection/evaluation/guidance/wrkdoc19_en.pdf



water scenarios for estimating exposure for risk assessment in aquatic ecotoxicology in the context 5786 of Council Directive 91/414/EEC. EFSA Journal, 145, 1-31. 5787

EFSA, 2005a. Opinion of the Scientific Panel on Plant health, Plant Protection Products and their 5788 Residues on a request from the EFSA related to the evaluation of dimoxystrobin. The EFSA 5789 Journal 178, 1-45. 5790

EFSA, 2005b. Opinion of the Scientific Panel on Plant health, Plant Protection Products and their 5791 Residues on a request from the EFSA related to the assessment of the acute and chronic risk to 5792 aquatic organisms with regard to the possibility of lowering the assessment factor if additional 5793 species were tested. The EFSA Journal 301, 1-45. 5794

EFSA (European Food Safety Authority), 2005c. Opinion of the Scientific Panel on Plant Health, 5795 Plant Protection Products and their Residues (PPR) on a request from EFSA related to the 5796 evaluation of pirimicarb. The EFSA Journal (2005) 240, 1-21. 5797

EFSA (European Food Safety Authority), 2006. Opinion of the Scientific Panel on Plant health, Plant 5798 Protection Products and their Residues on a request from the EFSA related to the aquatic risk 5799 assessment for cyprodinil and the use of a mesocosm study in particular. The EFSA Journal 329, 1-5800 77. 5801

EFSA (European Food Safety Authority), 2007a.Opinion of the Scientific Panel on Plant protection 5802 products and their Residues on a request from the Commission on acute dietary intake assessment 5803 of pesticide residues in fruit and vegetables. 19 April 2007. The EFSA Journal (2007) 538, 1-88. 5804

EFSA (European Food Safety Authority), 2007b. Opinion of the Scientific Panel on Plant protection 5805 products and their Residues on a request from the Commission on the risks associated with an 5806 increase of the MRL for dieldrin on courgettes. 12 September 2007. The EFSA Journal (2007) 554, 5807 1-48. 5808

EFSA (European Food Safety Authority), 2008a. Risk Assessment for Birds and Mammals - Revision 5809 of Guidance Document under Council Directive 91/414/EEC (SANCO/4145/2000 – final of 25 5810 September 2002) [1] - Scientific Opinion of the Panel on Plant protection products and their 5811 Residues (PPR) on the Science behind the Guidance Document on Risk Assessment for birds and 5812 mammals. EFSA Journal, 734,1-181. 5813

EFSA (European Food Safety Authority), 2008b. Opinion of the Scientific Panel on Plant Protection 5814 products and their Residues to evaluate the suitability of existing methodologies and, if appropriate, 5815 the identification of new approaches to assess cumulative and synergistic risks from pesticides to 5816 human health with a view to set MRLs for those pesticides in the frame of Regulation (EC) 5817 396/2005. 15 April 2008. The EFSA Journal (2008) 704, 1-3. 5818

EFSA (European Food Safety Authority), 2009a. Scientific Opinion of the Panel on Plant Protection 5819 Products and their Residues on a request from EFSA updating the opinion related to Annex II & 5820 III: Ecotoxicological studies. The EFSA Journal, 1165, 1-25. 5821

EFSA (European Food Safety Authority), 2009b. Outcome of the Public Consultation on the existing 5822 Guidance Documents on Aquatic and Terrestrial Ecotoxicology under Directive 91/414/EC. The 5823 EFSA Journal 2009, 7(11), 1375. 5824

EFSA (European Food Safety Authority), 2009c. Guidance Document on Risk Assessment for Birds 5825 and Mammals on request of EFSA. EFSA Journal 2009; 7(12):1438 [358 pp.]. 5826

EFSA Panel on Plant Protection Products and their Residues (PPR), 2010a. Scientific Opinion on 5827 outline proposals for assessment of exposure of organisms to substances in soil. EFSA Journal 5828 2010; 8(1):1442 [38 pp.]. 5829

EFSA Panel on Plant Protection Products and their Residues (PPR), 2010b. Scientific opinion on the 5830 development of specific protection goal options for environmental risk assessment of pesticides, in 5831 particular in relation to the revision of the Guidance Documents on Aquatic and Terrestrial 5832



Ecotoxicology (SANCO/3268/2001 and SANCO/10329/2002). EFSA Journal 2010; 8(10):1821 5833 [55 pp.]. 5834

EFSA (European Food Safety Authority), 2011. Submission of scientific peer-reviewed open literature 5835 for the approval of pesticide active substances under Regulation (EC) No 1107/2009. EFSA Journal 5836 2011;9(2):2092 [49 pp.]. 5837

EFSA Panel on Plant Protection Products and their Residues (PPR), 2012a. Scientific Opinion on 5838 Evaluation of the Toxicological Relevance of Pesticide Metabolites for Dietary Risk Assessment. 5839 EFSA Journal 2012;10(07): 2799. [187 pp.] doi:10.2903/j.efsa.2012.2799. 5840

EFSA Panel on Plant Protection Products and their Residues (PPR), 2012b. Scientific Opinion on the 5841 science behind the development of a Risk Assessment of Plant Protection Products on bees (Apis 5842 mellifera, Bombus spp. and solitary bees). EFSA Journal 2012;10(5):2668 [275 pp.]. 5843

EFSA Panel on Plant Protection Products and their Residues (PPR), 2012c. Scientific Opinion on the 5844 science behind the guidance for scenario selection and scenario parameterisation for predicting 5845 environmental concentrations in soil. EFSA Journal 2012, 10, 2562. 5846

EPPO, 2003. Environmental risk assessment scheme for plant protection products. Chapter 12: Non-5847 target terrestrial higher plants. EPPO Bulletin 33, 239-244. 5848

Escher BI, Bramaz N, Richter M, Lienert J, 2006. Comparative ecotoxicological hazard assessment of 5849 beta-blockers and their human metabolites using a mode-of-action based test battery and a QSAR 5850 approach. Environmental Science and Technology, 40, 7402-7408. 5851

Fairchild JF, La Point TW, Schwartz TR, 1994. Effects of herbicide and insecticide mixture in aquatic 5852 mesocosms. Archives of Environmental Contamination and Toxicology, 27, 527-533. 5853

Farmer D, Hill IR, Maund SJ, 1995. A comparison of the fate and effects of two pyrethroid insecticides 5854 (lambda-cyhalothrin and cypermethrin) in pond mesocosms. Ecotoxicology, 4, 219-244. 5855

FOCUS, 2000. FOCUS groundwater scenarios in the EU review of active substances. Report of the 5856 FOCUS Groundwater Scenarios Workgroup, EC Document Reference SANCO/321/2000 rev.2, 5857 202 pp. 5858

FOCUS, 2001. FOCUS Surface Water Scenarios in the EU Evaluation Process under 91/414/EEC. 5859 Report of the FOCUS Working Group on Surface Water Scenarios, EC Document Reference 5860 SANCO/4802/2001-rev.2, 245 pp. 5861

FOCUS, 2007a. Landscape and mitigation factors in aquatic risk assessment. Volume 1. Extended 5862 Summary and Recommendations”. Report of the FOCUS Working Group on Landscape and 5863 Mitigation Factors in Ecological Risk Assessment, EC Document Reference SANCO/10422/2005 5864 v2.0. 169 pp. 5865

FOCUS, 2007b. Landscape And Mitigation Factors In Aquatic Risk Assessment. Volume 2. Detailed 5866 Technical Reviews. Report of the FOCUS Working Group on Landscape and Mitigation Factors in 5867 Ecological Risk Assessment, EC Document Reference, SANCO/10422/2005 v2.0. 436 pp. 5868

FOCUS, 2008. Pesticides in Air: Considerations for Exposure Assessment. Report of the FOCUS 5869 Working Group on Pesticides in Air, EC Document Reference SANCO/10553/2006 Rev 2 June 5870 2008. 327 pp. 5871

FOCUS, 2009. Assessing Potential for Movement of Active Substances and their Metabolites to 5872 Ground Water in the EU. Report of the FOCUS Ground Water Work Group, EC Document 5873 Reference Sanco/13144/2010 version 1, 604 pp. 5874

Foit K, Kaske O and Liess M, 2012. Competition increases toxicant sensitivity and delays the recovery 5875 of two interacting populations. Aquatic Toxicology, 106-107, 25-31. 5876

Forbes VE and Calow P, 2002. Species Sensitivity Distributions revisited: a critical appraisal. Human 5877 and Ecological Risk Assessment, 8,473-492. 5878



Fryday S and Thompson H, 2012. Toxicity of pesticides to aquatic and terrestrial life stages of 5879 amphibians and occurrence, habitat use and exposure of amphibian species in agricultural 5880 environments. Supporting Publications 2012:EN-343. [348 pp.]. 5881

Gergs A, Classen S, Hommen U and Preuss TG, 2011. Identification of realistic worst case aquatic 5882 macroinvertebrate species for prospective risk assessment using the trait concept. Environmental 5883 Science and Pollution Research, 18, 1316-1323. 5884

Giesy JP and Graney RL, 1989. Recent developments in and intercomparisons of acute and chronic 5885 bioassays and bioindicators. Hydrobiologia 188/189, 21-60. 5886

Giddings JM, Brock TCM, Heger W, Heimbach F, Maund SJ, Norman S, Ratte H-T, Schäfers C and 5887 Streloke M (eds). 2002. Community-level aquatic system studies-interpretation criteria. 5888 (CLASSIC) Pensacola (FL): SETAC 44 p. 5889

Giddings J, Arts G and Hommen U, 2013. The Relative Sensitivity of Macrophyte and Algal Species 5890 to Herbicides and Fungicides: An Analysis Using Species Sensitivity Distributions". IEAM (in 5891 review after minor revisions). 5892

Gruessner B and Watzin MC, 1996. Response of aquatic communities from a Vermont stream to 5893 environmentally realistic atrazine exposure in laboratory microcosms. Environmental Toxicology 5894 & Chemistry, 15, 410-419. 5895

Guasch H and Sabater S, 1998. Light history influences the sensitivity to atrazine in periphytic algae. 5896 Journal of Phycology, 34, 233-241. 5897

Heckman LH and Friberg N, 2005. Macroinvertebrate community response to pulse exposure with the 5898 insecticide lambda-cyhalothrin using in-stream mesocosms. Environmental Toxicology & 5899 Chemistry, 24, 582-590. 5900

Heugens E, Hendriks A, Dekker T, Van Straalen NM and Admiraal W, 2001. A review of the effects 5901 of multiple stressors on aquatic organisms and analysis of uncertainty factors for use in risk 5902 assessment. Critical Reviews in Toxicology, 31,247-284. 5903

Hill IR, Runnalls JK, Kennedy JH and Ekoniak P, 1994. Lambda-cyhalothrin: A mesocosm study of its 5904 effects on aquatic organisms. In: Aquatic mesocosm studies in ecological risk assessment. R.L. 5905 Graney, J.H. Kennedy, J.H. Rogers (Eds.). Special Publication of the Society of Environmental 5906 Toxicity and Chemistry, Michigan: Lewis Publishers, pp. 403-467. 5907

Holland PT, 1996. Glossary of terms relating to pesticides (IUPAC Recommendations 1996). Pure and 5908 Applied Chemistry, 68,1167-1193. 5909

Hommen U, Poethke HJ, Dulmer U, Ratte HT, 1993. Simulation-models to predict ecological risk of 5910 toxins in fresh-water systems. ICES Journal of Marine Science, 50, 4, 337-347. 5911

Hull RN and Swanson S, 2006. Sequential analysis of lines of evidence – an advanced weight of 5912 evidence approach for ecological risk assessment. Integrated Environmental Assessment and 5913 Management, 2, 302-311. 5914

Jaworska J, Nikolova-Jelizkova N, and Aldenberg T, 2005. Review of methods for QSAR 5915 applicability domain estimation by the training set. A.T.L.A. 33: 445-459. 5916

Jager T, Albert C, Preuss TG, Ashauer R, 2011. General unified threshold model of survival – a 5917 toxicokinetic-toxicodynamic framework for ecotoxicology. Environmental Science and Technology, 5918 45, 2529-2540. 5919

Jarvis NJ, 1994. The MACRO model (Version 3.1). Technical Description and sample simulations. 5920 Reports and Dissertations, 19, Department of Soil Sciences, Swedish University of Agricultural 5921 Sciences, Uppsala, Sweden, 51 pp. 5922

Jarvis NJ, 2001. The MACRO model (version 4.3). Technical description. 5923

Johnson BT, 1986. Potential impact of selected agricultural chemical contaminants on a northern 5924 prairie wetland: a microcosm evaluation. Environmental Toxicology & Chemistry, 5, 473-485. 5925



Jongbloed RH, Traas, TP and Luttik R, 1996. A probabilistic model for deriving soil quality criteria 5926 based on secondary poisoning of top predators. II. Calculations for dichlorodiphenyltrichloroethane 5927 (DDT) and cadmium. Ecotoxicology and Environmental Safety, 34, 279-306. 5928

Junghans M, Backhaus T, Faust M, Schloze M and Grimme LH, 2006. Application and validation 5929 ofapproaches fort he predictive hazard assessment of realistic pesticide mixtures. 5930 AquaticToxicology, 76, 93-110. 5931

Jüttner I, Peither A, Lay JP, Kettrup A and Ormerod SJ, 1995. An outdoor mesocosm study to assess 5932 ecotoxicological effects of atrazine on a natural plankton community. Archives of Environmental 5933 Contamination and Toxicology, 29,435-441. 5934

Klok C, de Vries P, Jongbloed R and Tamis J, 2012. Literature review on the sensitivity and exposure 5935 of marine and estuarine organisms to pesticides in comparison to corresponding fresh water 5936 species. Supporting Publications 2012:EN-357. [157 pp.]. 5937

Knillmann S, Stampfli NC, Beketov MA and Liess M, 2012a. Intraspecific competition increases 5938 toxicant effects in outdoor pond microcosms. Ecotoxicology, 1-10. Retrieved from Google Scholar. 5939

Knillmann S, Stampfli NC, Noskov YA, Beketov MA and Liess M (2012b). Interspecific competition 5940 delays recovery of daphnia spp. Populations from pesticide stress. Ecotoxicology. 5941

Kosinski RJ, 1984. The effect of terrestrial herbicides on the community structure of stream 5942 periphyton. Environmental Pollution Series A, 36, 165-189. 5943

Kosinski RJ, Merkle MG, 1984. The effect of four terrestrial herbicides on the productivity of artificial 5944 stream algal communities. Journal of Environmental Quality, 13, 75-82. 5945

Kosol S, 2011. The effects of pesticides on aquatic hyphomycet community structure and functioning 5946 – implications for ecological risk assessment. Thesis University of Sheffield. 5947

Kortenkamp A, Backhaus T and Faust M, 2009. State of the art report on mixture toxicity. Final 5948 report, study Contract number 070307/2007/485103/ETU/D.1. European Commission, Brussels, 5949 Belgium, 391 pp. 5950

Krieger KA, Baker DB, Kramer JW, 1988. Effects of herbicides on stream Aufwuchs productivity and 5951 nutrient uptake. Archives of Environmental Contamination and Toxicology, 17, 299-306. 5952

Lahr J, 1997. Ecotoxicology of organisms adapted to life in temporary freshwater ponds in arid and 5953 semi-arid regions. Archives of Environmental Contamination and Toxicology, 32, 50-57. 5954

Leeds-Harrison PB, Shipway CJP, Jarvis NJ and Youngs EG 1986. The influence of soil macroporosity 5955 on water retention, transmission and drainage in a clay soil. Soil Use and Management 2, 47-50 5956

Liess M and Schulz R, 1999. Linking insecticide contamination and population response in an 5957 agricultural stream. Environmental Toxicology and Chemistry, 18, 1948-1955. 5958

Liess M, Brown C, Dohmen P, Duquesne S, Hart A, Heimbach F, Kreuger J, Lagadic L, Maund S, 5959 Reinert W, Streloke M and Tarazona, J, (2005). Effects of pesticides in the field – EPIF (p. 136). 5960 Brussels, Belgium: SETAC Press. 5961

Liess M, 2002. Population response to toxicants is altered by intraspecific interaction. Environmental 5962 Toxicology & Chemistry, 21, 138-142. 5963

Liess M and von der Ohe PC 2005. Analyzing effects of pesticides on invertebrate communities in 5964 streams. Environmental Toxicology and Chemistry. 24, (4): 954-965. 5965

Liess M, Pieters B and Duquesne S, 2006. Long-term signal of population disturbance after pulse 5966 exposure to an insecticide - rapid recovery of abundance, persistent alteration of structure. 5967 Environmental Toxicology and Chemistry, 25, 1326-1331. 5968

Liess M, Schäfer R and Schriever C. 2008. The footprint of pesticide stress in communities - species 5969 traits reveal community effects of toxicants. Science of the Total Environment, 406, 484-490. 5970



Liess M and Beketov M, 2011. Traits and stress: keys to identify community effects of low lvels of 5971 toxicants in test systems. Ecotoxicology, 20, 1328-1340. 5972

Lin A, 1989. Concordance Correlation Coefficient to evaluation reproducibility. Biometics, 45, 255-5973 268. 5974

Lin R, Buijse L, Dohmen P, Kosol S, Maltby L, Rocha Dimitrov M, Roessink I, Sinkeldam JA, Smidt 5975 H, Van Wijngaarden RPA and Brock TCM (2012). Effects of the fungicide metiram in outdoor 5976 freshwater microcosms: Responses of invertebrates, primary producers and microbes. 5977 Ecotoxicology 21:1550-1569. 5978

López-Mancisidor P, Carbonell G, Marina A, Fernández C and Tarazona JV, 2007. Zooplankton 5979 community responses to chlorpyrifos in mesocosms under Mediterranean conditions. 5980 Ecotoxicology and Environmental Safety, 71,16-25. 5981

López-Mancisidor P, Carbonell G, Fernández C and Tarazona JV, 2008. Ecological impact of repeated 5982 applications of chlorpyrifos on zooplankton community in mesocosms under Mediterranean 5983 conditions. Ecotoxicology, 17, 811-825. 5984

Luttik R, 2003. Risk Assessment Scheme for the Impact of Plant Protection Products on Birds and 5985 Mammals. Thesis, Leiden University, The Netherlands. 5986

Luttik R, Hart A, Roelofs W, Craig P and Mineau P, 2011. Variation in the level of protection 5987 afforded to birds and crustaceans exposed to different pesticides under standard risk assessment 5988 procedures, Integrated Environmental Assessment and Management, 7, 459-465. 5989

Mackay D, 1982, Correlation of bioconcentration factors: Environmental Science and Technology, 16, 5990 274-278. 5991

Madelin R, 2004. The importance of scientific advice in the Community decision making process. 5992 Opening address to the Inaugural Joint Meeting of the members of the Non-Food Scientific 5993 Committees. Directorate General for Health and Consumer Protection, European Commission, 5994 Brussels. 5995

Maltby L, 1992. Detritus Processing In Calow P & Petts GE (Eds), The Rivers Handbook Vol. 1. 5996 Blackwell Scientific Publications, Oxford, UK, pp 331-353. 5997

Maltby L, Blake N, Brock TCM and Van den Brink PJ, 2005. Insecticide species sensitivity 5998 distributions: the importance of test species selection and relevance to aquatic ecosystems. 5999 Environmental Toxicology and Chemistry, 24, 379-388. 6000

Maltby L, Brock TCM and Van den Brink PJ, 2009. Fungicide risk assessment for aquatic ecosystems: 6001 Importance of interspecific variation, toxic mode of action and exposure regime. Environmental 6002 Science & Technology, 43, 7556-7563. 6003

Maltby L, Arnold D, Arts G, Davies J, Heimbach F, Pickl C and Poulsen V (Eds), 2010. Aquatic 6004 Macrophyte Risk Assessment for Pesticides. SETAC Press & CRC Press, Taylor & Francis Group, 6005 Boca Raton, FL, 140 pp. 6006

McGill BJ, Enquist BJ, Weiher E and Westoby M, 2006. Rebuilding community ecology from 6007 functional traits. Trends in Ecology & Evolution, 21,178-185. 6008

Mohr S, Berghahn R, Feibicke M, Meinecke S, Ottenströer, Schmiedling I, Schmiediche R and 6009 Schmidt R, 2007. Effects of the herbicide metazachlor on macrophytes and ecosystem function in 6010 freshwater pond and stream mesocosms. Aquatic Toxicology, 82, 73-84. 6011

Mohr S, Feibicke M, Berghahn R, Schmiediche R and Schmidt R, 2008. Response of plankton 6012 communities in freshwater pond and stream mesocosms to the herbicide metazachlor. 6013 Environmental Pollution, 152, 530-542. 6014

Muňos I, Real M, Guasch H, Navarro E and Sabater S, 2001. Effects of atrazine on periphyton under 6015 grazing pressure. Aquatic Toxicology, 55, 239-249. 6016



Muschal M and Warne MS, 2003. Risk posed by pesticides to aquatic organisms in rivers of northern 6017 inland new south wales, australia. Human and Ecological Risk Assessment, 9(7), 1765-1787. 6018

NAFTA, 2011. (Quantitative) Structure Activity Relationship [(Q)SAR] Guidance Document. North 6019 American Free Trade Agreement (NAFTE) Technical working Group on Pesticides (TWG). 6020

Netzeva TI, Worth A, Aldenberg T, Benigni R, Cronin MT, Gramatica P, Jaworska JS, Kahn S, 6021 Klopman G, Marchant CA, Myatt G, Nikolova-Jeliazkova N, Patlewicz GY, Perkins R, Roberts D, 6022 Schultz T, Stanton DW, van de Sandt JJ, W Tong, Veith G and Yang C, 2005. Current Status of 6023 Methods for Defining the Applicability Domain of (Quantitative) Structure-Activity Relationships. 6024 The report and recommendations of ECVAM Workshop 52. Altern. Lab. Anim. 33(2): 155-173. 6025

Niemi GJ, De Vore P, Detenbeck N, Taylor D, Lima A and Pastor J, 1990. Overview of case studies 6026 on recovery of aquatic systems from disturbance. Environmental Management 14:571-587 6027

Nikolova N and Jaworska J, 2003. Approaches to measure chemical similarity – a review. QSAR & 6028 Comb. Sci. 22: 1006-1026. 6029

Nystrőm B, Paulsson M, Almgren K and Blank H, 2000. Evaluation of the capacity for development 6030 of atrazine tolerance in periphyton from a Swedish freshwater site as determined by inhibition of 6031 photosynthesis and sulfolipid synthesis. Environmental. Toxicology & Chemistry, 19, 1324-1331. 6032

OECD, 2000. Guidance document on aquatic toxicity testing of difficult substances and mixtures, 6033 OECD series on testing and assessment Number 23, ENV/JM/MONO(2000)6 6034

OECD, 2006. Guidance Document on Simulated Freshwater Lentic Field tests (outdoor microcosms 6035 and mesocosms). Series on Testing and Assessment, No 53, ENV/JM/MONO(2006)17, OECD 6036 Environment Directorate, Paris, 37 pp. 6037

OECD, 2007a. Guidance document on the validation of (Quantitative) Structure-Activity Relationship 6038 [(Q)SAR] models. ENV/JM/MONO(2007)2. 6039

OECD, 2007b. Guidance on grouping of chemicals. Series on testing and assessment Number 80. 6040

OECD, 2009. Guidance Document for using the OECD (Q)SAR application toolbox to develop 6041 chemical categories according to the OECD Guidance on grouping of chemicals. Series on testing 6042 and assessment Number 102. 6043

OECD, 2012a. Detailed review paper on the state of the science on novel in vitro and in vivo 6044 screening and testing methods and endpoints for evaluating endocrine disruptors. Series on testing 6045 and assessment Number 178. 6046

OECD, 2012b. Revised draft fish toxicity testing framework. Final adoption, revised as appropriate, 6047 by the Annual meeting of the National Coordinators of the OECD Test Guideline Program in April 6048 2012 6049

OECD, 2012c. Guidance document on standardised test guidelines for evaluating chemicals for 6050 endocrine disruption. Series on Testing and Assessment, No. 150 (ENV/JM/MONO(2012)22) 6051

O'Neill RV, Gardner RH, Barnthouse LW, Suter GW, Hildebrand SG and Gehrs CW, 1982. 6052 Ecosystem risk analysis: A new methodology. Environmental Toxicology and Chemistry 1, 167-6053 177. 6054

Park RA, Clough JS and Wellman MC, 2008. AQUATOX: Modeling environmental fate and 6055 ecological effects in aquatic ecosystems. Ecological Modelling, 213, 1, 1-15. 6056

Porcelli C, Boriani E, Roncaglioni A, Chana A and Benfenati E, 2008. Regulatory Perspectives in the 6057 Use and Validation of QSAR. A Case Study: DEMETRA Model for Daphnia Toxicity. 6058 Environmental Science and Technology, 42, 2, 491-496. 6059

Posthuma L, Suter GWI and Traas TP (eds). 2002. Species-Sensitivity Distributions in Ecotoxicology. 6060 CRC Press, Lewis, Boca Raton, FL, 587 pp. 6061



Posthuma L, De Zwart D, Solomon KR and Brock TCM, 2008. Guidance on the application of 6062 extrapolation methods in ecological exposure and effects characterization of chemicals. In: 6063 Solomon KR, Brock TCM, De Zwart D, Dyer SD, Posthuma L, Richards SM, Sanderson H, Sibley 6064 PK and Van den Brink PJ. (Eds). (2008) Extrapolation Practice for Ecotoxicological Effect 6065 Characterization of Chemicals, SETAC Press & CRC Press, Boca Raton, Fl, USA, pp 281-322. 6066

Pratt JR, Bowers NJ, Niederlehner BR and Cairns Jr. J, 1988. Effects of atrazine on freshwater 6067 microbial communities. Archives of Environmental Contamination and Toxicology, 17, 449-457. 6068

Preuss TG, Hammers-Wirtz M and Ratte HT, 2010. The potential of individual based population 6069 models to extrapolate effects measured at standardized test conditions to relevant environmental 6070 conditions-an example for 3,4-dichloroaniline on Daphnia magna. Journal of environmental 6071 monitoring, 12, 11, 2070-2079. 6072

Pusey BJ, Arthington AH and McClean J, 1994. The effects of a pulsed application of chlorpyrifos on 6073 macroinvertebrate communities in an outdoor artificial stream system. Ecotoxicology and 6074 Environmental Safety, 27, 221 - 250. 6075

Reynaldi S, Meiser M and Liess M, 2011. Effects of the pyrethroid fenvalerate on the alarm response 6076 and on the vulnerability of the mosquito larva culex pipiens molestus to the predator notonecta 6077 glauca. Aquatic Toxicology, 104(1-2), 56-60. doi:10.1016/j.aquatox.2011.03.017 6078

Roessink I, Arts GHP, Belgers JDM, Bransen F, Maund SJ and Brock TCM, 2005. Effects of lambda-6079 cyhalothrin in two ditch microcosm systems of different trophic status. Environmental Toxicology 6080 and Chemistry, 24, 1684-1696. 6081

Roex EWM, Van Gestel CAM, Van Wezel AP and Van Straalen NM. 2000. Ratios between acute 6082 aquatic toxicity and effects on population growth rates in relation to toxicant mode of action. 6083 Environmental Toxicology & Chemistry, 19,685-693. 6084

Romijn CAFM, Luttik R, v.d. Meent D, Slooff W and Canton JH, 1993. Presentation of a General 6085 Algorithm to Include Effect Assessment on Secondary Poisoning in the Derivation of 6086 Environmental Quality Criteria. Part 1. Aquatic Food Chains. Ecotoxicology and Environmental 6087 Safety, 26, 61-85. 6088

Romijn CAFM, Luttik R, and Canton JH, 1994. Presentation of a general algorithm to include effect 6089 assessment on secondary poisoning in the derivation of environmental quality criteria. 2. Terrestrial 6090 food chains. Ecotoxicology and Environmental Safety, 27, 107-127. 6091

Rozman KK and Doull J (2000). Dose and time as variables of toxicity. Toxicology, 144,169–178. 6092

Rubach MN, Baird DJ and Van den Brink PJ, 2010. A new method for ranking mode specific 6093 sensitivity of freshwater arthropods to insecticides and its relationships to biological traits. 6094 Environmental Toxicology and Chemistry, 29, 1-12. 6095

Russom CL, Bradbury SP, Broderius SJ, Hammermeister DE and Drummond RA, 1997. Predicting 6096 modes of toxic action from chemical structure: acute toxicity in the fathead minnow(Pimephales 6097 promelas), Environmental Toxicology and Chemistry, 16:948–967. 6098

SANCO document 11802/2012, Commission proposal on new data requirements for approval of 6099 active substances. 6100

SANCO document 11803/2012, Commission proposal on new data requirements for approval of plant 6101 protection products. 6102

Sanderson H, Laird B, Brain R, Wilson CJ and Solomon KR, 2009. Detectability of fifteen aquatic 6103 micro/mesocosms. Ecotoxicology 18: 838-845. 6104

Schäfer RB, Caquet T, Siimes K, Mueller R, Lagadic L, Liess M, 2007. Effects of pesticides on 6105 community structure and ecosystem functions in agricultural streams of three biogeographical 6106 regions in europe. Science of the Total Environment, 382(2-3), 272-285. 6107



Schäfer RB, von der Ohe PC, Rasmussen J, Kefford BJ, Beketov MA, Schulz R, Liess M, 2012. 6108 Thresholds for the effects of pesticides on invertebrate communities and leaf breakdown in stream 6109 ecosystems. Environmental Science & Technology, 46(9), 5134-5142. doi:10.1021/es2039882 6110

Schriever C, Ball MH, Holmes C, Maund S, Liess M, 2007. Agricultural intensity and landscape 6111 structure: Influences on the macroinvertebrate assemblages of small streams in northern Germany. 6112 Environmental Toxicology and Chemistry, 26, 346–357. 6113

Schüürmann et al, 2008. External validation and prediction employing the predictive squared 6114 correlation coefficients Test Set Activity means vs Training Set Activity Mean. Journal of 6115 Chemical Information and Modeling, 48, 2140-2145. 6116

Seguin F, Leboulanger C, Rimet F, Druart JC and Berard A, 2001. Effects of atrazine and nicosulfuron 6117 on phytoplankton in systems of increasing complexity. Archives of Environmental Contamination 6118 and Toxicology, 40, 198-208. 6119

Shi et al, 2001. QSAR models using a large diverse set of estrogens. Journal of Chemical Information 6120 and Computer Science, 41, 186-195. 6121

Siefert RE, Lozano SJ, Brazner JC and Knuth ML, 1989. Littoral enclosures for aquatic field testing of 6122 pesticides: effects of chlorpyrifos on a natural system. Miscellaneous Publication Series 6123 Entomological Society of America, 75: 57-73. 6124

Sinclair CJ and Boxall ABA, 2003. Assessing the ecotoxicity of pesticide transformation products, 6125 Environmental Science and Technology, 37, 4617-4625. 6126

Sinclair CJ, 2009. Predicting the environmental fate and ecotoxicological and toxicological effects of 6127 pesticide transformation products. PhD Thesis. University of York, Environment Department 6128

Slijkerman DME, Baird DJ, Conrad A, Jak RG and Van Straalen NM, 2004. Assessing structural and 6129 functional plankton responses to carbendazim toxicity. Environmental Toxicology & Chemistry, 23, 6130 455-462. 6131

Smit, C.E., 2005. Energy and moisture content and assimilation efficiency of bird and mammal food. 6132 RIVM report 601516013, pp. 57–71. 6133

Solomon KR, Brock TCM, De Zwart D, Dyer SD, Posthuma L, Richards SM, Sanderson H, Sibley PK 6134 and Van den Brink PJ. (Eds), 2008. Extrapolation Practice for Ecotoxicological Effect 6135 Characterization of Chemicals, SETAC Press & CRC Press, Boca Raton, Fl, USA, 380 pp 6136

Stäb JA, Traas TP, Stroomberg G, Van Kesteren J, Leonards P, VanHattum B, Brinkman UAT and 6137 Cofino WP, 1996. Determination oforganotin compounds in the foodweb of a shallow 6138 freshwaterlake in The Netherlands. Archives of Environmental Contamination and Toxicology 6139 31:319–328 6140

Stampfli NC, Knillmann S, Liess M and Beketov MA, 2011. Environmental context determines 6141 community sensitivity of freshwater zooplankton to pesticides. Aquatic Toxicology, 104, 116-124. 6142

Stay FS, Katko A, Rohm CM, Fix MA and Larsen DP, 1989. The effects of atrazine on microcosms 6143 developed from four natural plankton communities. Archives of Environmental Contamination and 6144 Toxicology, 18, 866-875. 6145

Sterling A, 2010. Keep it complex. Nature 468, 1029–1031, doi:10.1038/4681029a 6146

Suter GW, Traas TP and Posthuma L. 2002. Issues and practises in the derivation and use of species 6147 sensitivity distributions. In: Posthuma L, Traas TP, Suter GW (eds), Species Sensitivity 6148 Distributions in Risk Assessment. Boca Raton (FL), CRC Press, pp 473-474. 6149

Ter Horst MMS, Adriaanse PI and Boesten JJTI, 2009. Mitigation of runoff in the FOCUS Surface 6150 Water Scenarios. Note of the fate group of the Environmental Risk Assessment team of Alterra on 6151 the interpretation of the mitigation of runoff in the FOCUS Landscape and Mitigation report 6152 (2007). Alterra rapport 1794. Wageningen 6153



Thorbek P, Forbes VE, Heimbach F, Hommen U, Thulke H-H, Van den Brink PJ, Wogram J and 6154 Grimm V (Eds), 2010. Ecological Models for Regulatory Risk Assessment of Pesticides. 6155 Developing a Strategy for the Future. SETAC Press & CRC Press, Taylor & Francis Group, Boca 6156 Raton, FL, 125 pp. 6157

Traas TP, Luttik R, and Jongbloed RH, 1996. A probabilistic model for deriving soil quality criteria 6158 based on secondary poisoning of top predators. I. Model description and uncertainty analysis. 6159 Ecotoxicology and Environmental Safety, 34, 264-278. 6160

Traas TP, Janse JH, Van den Brink PJ, Brock TCM and Aldenberg T, 2004. A freshwater food web 6161 model for the combined effects of nutrients and insecticide stress and subsequent recovery. 6162 Environmental Toxicology and Chemistry, 23, 2, 521-529. 6163

Turner II BL, Kasperson RE, Matson PA, McCarthy JJ, Corell RW, Christensen L, Eckley N, 6164 Kaperson JX, Luers A, Martello ML, Polsky C, Pilsipher A and Schiller A, 2003. A framework for 6165 vulnerability analysis in sustainability science. PNAS, 100, 8074-8079. 6166

US EPA, 2008. Ecological Structure Activity Relationships. 6167 http://www.epa.gov/oppt/newchems/tools/21ecosar.htm, United States Environmental Protection 6168

Agency, Washington DC, USA. 6169

Van Beelen P, 2003. A review on the application of microbial toxicity tests for deriving sediment 6170 quality guidelines. Chemosphere, 53, 795-808. 6171

Van den Brink PJ, Van Donk E, Gylstra R, Crum SJH and Brock TCM, 1995. Effects of chronic low 6172 concentrations of the pesticides chlorpyrifos and atrazine in indoor freshwater microcosms. 6173 Chemosphere, 31, 3181-3200. 6174

Van den Brink PJ, Van Wijngaarden RPA, Lucassen WGH, Brock TCM and Leeuwangh P, 1996. 6175 Effects of the insecticide Durban 4®E (a.i. chlorpyrifos) in outdoor experimental ditches: II. 6176 Community responses and recovery. Environmental Toxicology and Chemistry, 15, 1143-1153. 6177

Van den Brink PJ and Ter Braak CJF,1999. Principal response curves: Analysis of time-dependent 6178 multivariate responses of biological communities to stress. Environmental Toxicology and 6179 Chemistry, 18,138-148. 6180

Van den Brink PJ, Hattink J, Bransen F, Van Donk E and Brock TCM, 2000. Impact of the fungicide 6181 carbendazim in freshwater microcosms. II. Zooplankton, primary producers and final conclusions. 6182 Aquatic Toxicology, 48,251-264. 6183

Van den Brink PJ, Hartgers EM, Gylstra R, Bransen F and Brock TCM, 2002. The effects of a mixture 6184 of two insecticides on freshwater microcosms. II. Water quality, responses of zooplankton, 6185 phytoplankton and periphyton and ecological risk assessment. Ecotoxicology, 11, 181-197. 6186

Van den Brink PJ, Blake N, Brock TCM and Maltby L, 2006. Predictive value of species sensitivity 6187 distributions for effects of herbicides in freshwater ecosystems. Human and Ecological Risk 6188 Assessment, 12, 645-674. 6189

Van den Brink PJ, Baveco JM, Verboom J and Heimbach F, 2007. An individual-based approach to 6190 model spatial population dynamics of invertebrates in aquatic ecosystems after pesticide 6191 contamination. Environmental Toxicology and Chemistry, 26, 10, 2226-2236. 6192

Van Vlaardingen PLA, Traas TP, Wintersen AM and Aldenberg T, 2004. ETX 2.0. A program to 6193 calculate Hazardous Concentrations and Fraction Affected, based on normally distributed toxicity 6194 data. Bilthoven, The Netherlands: National Institute for Public Health and the Environment 6195 (RIVM). Report 601501028. http://www.rivm.nl/bibliotheek/rapporten/601501028.pdf. Software 6196 available via http://www.rivm.nl/rvs/risbeoor/Modellen/ETX.jsp 6197

Van Wijngaarden RPA, Van den Brink PJ, Oude Voshaar JH and Leeuwangh P, 1995. Ordination 6198 techniques for analysing the responses of biological communities to toxic stress in experimental 6199 ecosystems. Ecotoxicology, 4, 61-77. 6200



Van Wijngaarden RPA, Cuppen JGM, Arts GHP, Crum SHJ, Van den Hoorn MW, Van den Brink PJ 6201 and Brock TCM, 2004. Aquatic risk assessment of a realistic exposure to pesticides used in bulb 6202 crops: A microcosm study. Environmental Toxicology & Chemistry, 23, 1479-1498. 6203

Van Wijngaarden RPA, Brock TCM and Douglas MT, 2005. Effects of chlorpyrifos in freshwater 6204 model ecosystems: the influence of experimental conditions on ecotoxicological thresholds. Pest 6205 Management Science, 61, 923 – 935. 6206

Van Wijngaarden RPA, Brock TCM, Van den Brink PJ, Gylstra R and Maund SJ, 2006. Ecological 6207 effects of spring and late summer applications of lambda-cyhalothrin in freshwater microcosms. 6208 Archives of Environmental Contamination and Toxicology, 50, 220 – 239. 6209

Verbruggen, EMJ and van den Brink PJ, 2010. Review of recent literature concerning mixture toxicity 6210 of pesticides to aquatic organisms. RIVM report 601400001. The National Institute for Public 6211 Health and the Environment (RIVM), BA Bilthoven, The Netherlands, 34 pp. 6212

Verdonck FAM, Aldenberg T, Jaworska J and Vanrolleghem PA, 2003. Limitations of current risk 6213 characterization methods in probabilistic environmental risk assessment. Environmental 6214 Toxicology and Chemistry 22: 2209–2213. 6215

Verhaar HJM, van Leeuwen CJ, and Hermens JLM, 1992. Classifying environmental Pollutants. 1: 6216 Structure-activity relationships for prediction of aquatic toxicity. Chemosphere, 25, 471-491. 6217

Von der Ohe PC and Liess M, 2004. Relative sensitivity distribution of aquatic invertebrates to 6218 organic and metal compounds. Environmental Toxicology and Chemistry, 23,150-156. 6219

Warne MStJ, 2003. A Review of the ecotoxicity of mixtures, approaches to, and recommendations for, 6220 their management. In: Proceedings of the 5th national workshop on the assessment of site 6221 contamination. Eds Langley A, Gilbey M and Kennedy B. National Environment Protection 6222 Council (NEPC), Adelaide, South Australia, 253-276. Available on 6223 http://www.ephc.gov.au/sites/default/files/ASC_WkshopPaper__19_Mix_Warne_Ecotoxicity_2006224 301.pdf 6225

Williams JR, 1975. Sediment Yield Prediction with Universe Equation Using Runoff Energy Factor. 6226 In: Present and Prospective Technology for Predicting Sediment Yields and Sources. U.ES. Dept. 6227 of Agriculture, Washington, DC. ARS-S-40. 6228

Willis KJ, Van den Brink PJ and Green JG, 2004. Seasonal variation in plankton community responses 6229 of mesocosms dosed with pentachlorophenol. Ecotoxicology, 13, 707-720. 6230

Wogram J, 2010. Ecological characterization of small streams in northern and central Germany. In: 6231 Brock TCM, Alix A, Brown CD, Capri E, Gottesbüren BFF, Heimbach F, Lythgo CM, Schulz R, 6232 Streloke M, (EDs), 2010. Linking Aquatic Exposure and Effects: Risk Assessment of Pesticides. 6233 SETAC Press & CRC Press, Taylor & Francis Group, Boca Raton, FL, pp 250-268. 6234

6235



APPENDICES 6236

A. ELEMENTS OF THE EXPOSURE ASSESSMENT GOALS RELATED TO THE CHOICES 6237 MADE IN THE FOCUSSW SCENARIOS 6238

6239

A.1. Introduction 6240

The aquatic risk assessment is the combination of the exposure and the effect assessments and there is 6241 considerable interaction between these assessments. EFSA (2010b) indicated that a specification of the 6242 spatio-temporal statistical population of exposure concentrations together with the percentile to be 6243 taken from this spatio-temporal population are essential parts of the protection-goal dimensions 6244 because the risk is only assessed for the spatio-temporal variability of the systems that are included (so 6245 for the remaining systems it cannot be excluded that high exposure concentrations leading to 6246 unacceptable effects will occur). 6247

For this guidance document the PPR Panel assumes that the current exposure assessment procedure 6248 (FOCUSSW scenarios and models) will continue to be used and does not include further guidance for 6249 the exposure assessment. EFSA’s PPR Panel does also not expect to evaluate the current exposure 6250 assessment (FOCUS, 2001, 2007a,b) in the coming years due to limited resources. As described 6251 above, the overall level of protection of aquatic organisms is determined by the combination of the 6252 Specific Protection Goals for the organisms and the exposure assessment goals. So without description 6253 of the exposure assessment goals, the overall level of protection for aquatic organisms is undefined 6254 (EFSA, 2010b). Therefore, we describe below the elements of the exposure assessment goals and add 6255 as far as possible the choices made by FOCUS (2001, 2007a,b). 6256

The exposure assessment by FOCUS is based on diffuse sources of pollution that will occur if the 6257 plant protection product is applied and used following the rules of good agricultural practice (FOCUS, 6258 2001, 2007a,b), so not including point sources resulting from inappropriate agricultural practices such 6259 as cleaning of spraying equipment and discharging the contaminated water directly into surface water 6260 systems. This is in line with Regulation (EC) 1107/2009 which requires that plant protection products 6261 are authorised based on application consistent with Good Plant Protection Practice and having regard 6262 to realistic conditions of use. 6263

Since around 2000, it has become common practice in the SCFCAH to accept exposure assessments 6264 based on 90th percentile concentrations, as this is considered to be ‘realistic worst case’. The definition 6265 of the exposure assessment goal then has to focus on the types of concentrations to be considered: e.g. 6266 a spray drift event will cause a much higher concentration in a shallow stream that is 1 cm deep than in 6267 a stream that is 30 cm deep. This specification is in the next sections split into (i) the spatial unit, (ii) 6268 the spatial statistical population of spatial units, and (iii) the temporal statistical population of 6269 concentrations. At the end the value of the percentile and its determination is discussed. 6270

A.2. Definition of the spatial unit 6271

The definition of the spatial unit splits into two aspects: the type of spatial unit (e.g. edge-of-field 6272 water bodies that temporarily fall dry or that are permanent; macrophyte-dominated water bodies or all 6273 water bodies) and the size or area of this unit over which exposure concentrations may be averaged. 6274

FOCUS (2001) developed 12 ditch and stream scenarios which all were 100x1 m and had a minimum 6275 water depth of 30 cm. FOCUS (2001) developed also three pond scenarios that were 30×30 m and at 6276 least 1 m deep. So these are all permanent water bodies, which is consistent with the effect assessment 6277 which is also based on permanent water bodies (see Section 1.3.6). 6278



FOCUS (2001) considered the concentration at the end of a ditch or stream that received spray drift, 6279 run-off or drainage from an adjacent field over a length of 100 m. So high local concentrations caused 6280 by outlets of individual drainpipes or by high local spray drift depositions (resulting e.g. from vertical 6281 spray boom movements) were ignored. This approach is likely to provide concentrations that are close 6282 to the concentration averaged over a ditch or stream length of 100 m. This averaging length is related 6283 to the mobility of the organism whose effect is assessed: in principle immobile organisms such as 6284 macrophytes would require averaging lengths as short as 1 m whereas for larger fish an averaging 6285 length of 1000 m could be defensible. However, the specific protection goal for the macrophytes 6286 indicates that these organisms are protected at the population level. So also for macrophytes this 6287 averaging over 100 m seems to be consistent with the proposed effect assessment. 6288

FOCUS (2001) considered only average concentrations over the full surface area of the ponds. 6289

6290 A.3. The spatial statistical population of the spatial units 6291

After the spatial units have been defined, the spatial statistical population of these units can be defined. 6292 The first step is to specify the total area to be considered: for example the whole EU, one of the 6293 regulatory zones, a zone based on climate properties, a Member State or a major agricultural area 6294 within the EU like the Po valley. This total area is related to the purpose of the regulatory decision 6295 making (e.g. EU registration, zonal registration or national registration). 6296

We consider here only registration at EU level. Both FOCUS groundwater and FOCUS surface water 6297 scenarios were developed for some ten locations distributed over the EU (then the EU-15; FOCUS, 6298 2000, 2001) representing a range of climatic conditions. This was considered sufficient to identify a 6299 safe use of significant size. So the FOCUS surface water scenarios could be considered to apply to 6300 some ten EU zones based on climate (and possibly soil) properties. The total area to be considered for 6301 each scenario would then be one of these ten zones. 6302

In the second step of the definition of the spatial statistical population it has to be decided whether all 6303 water systems in this area should be considered (i.e. a landscape-level approach) or only those 6304 adjacent to fields grown with the crop or crop group considered (i.e. the edge-of-field approach). 6305 FOCUS (2001) developed edge-of-field scenarios whereas FOCUS (2007a,b) described also (in great 6306 detail) methodologies for landscape-level exposure assessment. This guidance document deals only 6307 with edge-of-field water systems so we recommend not to use the landscape-level exposure 6308 assessment of FOCUS (2007a,b) in combination with this guidance document. 6309

For the exposure assessment of edge-of-field surface waters FOCUS (2001) assumed that the crop is 6310 grown and the substance is applied as close to the water as is possible considering good agricultural 6311 practice; this was supplemented by guidance by FOCUS (2007a,b) on emission reduction measures 6312 such as run-off buffer strips or spray-free zones. So the convention is that the crop is grown as close to 6313 the water as possible considering good agricultural practice. 6314

FOCUS (2001) limited the population to water systems adjacent to fields treated with this active 6315 substance, so excluding systems adjacent to fields treated which other active substances. 6316

FOCUS (2001) developed drainage scenarios for six locations (D1 to D6) and run-off scenarios for 6317 four locations (R1 to R4). The drainage scenarios receive only input from spray drift and drainpipes 6318 and the run-off scenarios receive only input from spray drift and run-off. The spray drift deposition 6319 values used by FOCUS were based on drift measurements that are downwind. FOCUS (2001) did not 6320 define the spatial populations on which the scenarios are based but the approach followed may be 6321 consistent with a statistical population that is further reduced based on the occurrence of entry routes 6322 (e.g. not considering surface water systems that are upwind during application or surface water 6323 systems that get only drainage or run-off inputs). 6324



A.4. Multi-year temporal statistical population of concentrations 6325

The concentrations do not only vary in space but also from year to year (see Figure A.1). So it has also 6326 to be defined which years are included in the statistical population and which not. If a use in a single 6327 crop has to be evaluated, this is straightforward: i.e. include only the application years because these 6328 will nearly always generate the highest concentrations and because including all the years without 6329 applications in a rotation sequence does not make sense. However, if the same active substance is used 6330 in different crops in a crop rotation sequence, the exposure assessment becomes more complicated. It 6331 may of course happen that the peak concentration for crop nr 2 becomes higher if crop nr 1 is included 6332 in the exposure assessment (e.g. because drainpipe leaching resulting from application in crop nr 1 6333 leads to a higher concentration peak in the surface water due to spray-drift at the time of application of 6334 crop nr 2). Such combinations of uses in different crops are usually not included in exposure 6335 assessments at EU or national level. However, the definition of the temporal statistical population has 6336 to be clear also in case risk managers for some reason would want to assess such combinations in the 6337 future. Therefore it is advisable to include in the definition of the temporal statistical population of 6338 concentrations also use of the same active substance in a crop rotation sequence. 6339

Consider the following complicated but realistic example of a 4 year application sequence for a certain 6340 active substance: 6341

- year 1: 1 kg/ha in maize and 0.5 kg/ha in carrots 6342 - year 2: 0.7 kg/ha in sugar beet 6343 - year 3: no applications 6344 - year 4: no applications 6345 - year 5: 1 kg/ha in maize and 0.5 kg/ha in carrots 6346 - … etc. 6347 6348

Let us assume that the effect assessment is based on annual peak concentrations. So each year in such 6349 a sequence generates one concentration and it has to be defined which concentration-year 6350 combinations are part of the temporal statistical population. FOCUS (2001) developed scenarios 6351 considering only one evaluation year in which a spray drift event takes place and the water system is 6352 also loaded with run-off and drainage. So years with low concentrations are not considered. This is 6353 consistent with including only the year with the highest concentration in the sequence in the statistical 6354 population (so only the years with the arrows in Figure A.1). 6355

6356

Figure A.1: Hypothetical sequence of annual peak concentrations for a period of 16 years with a 6357 four-year recurrence period of the application sequence (as in the example in the text). Dashed lines 6358 indicate the four-year periods and the arrows indicate the concentrations that are included in the 6359 temporal statistical population. 6360



A.5. Value of the percentile and its determination from the resulting combined spatio-temporal 6361 statistical population of concentrations 6362

The concentration in the water systems is a function of both space and time. Let us assume for 6363 example that we have a population of 100 ditches and a 10-year time series of annual peak 6364 concentrations. Then we have a population of 100 x 10 = 1000 concentrations (further called PECs). 6365 Let us assume that the 90th percentile is the target of the exposure assessment. The simplest procedure 6366 is just to derive the PEC considering all 1000 values as one pool (i.e. one cumulative frequency 6367 distribution). Then the 90th percentile can be approximated by taking the 900th of the ranked PEC 6368 values. However, such an approach does not distinguish between space and time and has the 6369 consequence that the 100 PEC values above the 90th percentile PEC can be from all the years in 10 6370 ditches thus accepting that the RAC may be exceeded at 10% of the locations in all years considered. 6371 But these 100 PEC values can also be from 1 year in 100 ditches, thus accepting that the RAC may be 6372 exceeded at all locations at 10% of the time (see Figure A.2 for an illustration of the possible 6373 combinations of spatial and temporal percentiles all giving an overall 90th percentile). 6374

6375

6376

Figure A.2. Example of contour diagram of percentiles of exposure concentrations as a function of the 6377 spatial and temporal percentiles. The pink line shows all combinations that give an overall 90th 6378 percentile and the grey lines show a possible combination of a spatial and a temporal percentile giving 6379 an overall 90th percentile. 6380

6381

An alternative approach would be to impose additional restrictions to selecting a percentile: for 6382 instance, do not assess the overall 90th percentile PEC but assess the 90th percentile PEC in space at the 6383 50th percentile in time at all locations; thus accepting exceedence of the RAC for more than 50% of the 6384 years only at 10% of the locations. A priori it is unknown whether the overall 90th percentile PEC is 6385 lower than the 90th percentile PEC in space at the 50th percentile in time. FOCUS (2001) did not 6386 consider these aspects of the selection of a percentile. 6387

The SCFCAH accepted FOCUS (2000) which based the assessment of leaching to groundwater on a 6388 90th percentile. The SCFCAH accepted also FOCUS (2001, p. 7) which described the FOCUS surface 6389 water scenarios as ten realistic worst-case scenarios which collectively represent agriculture in the EU. 6390 So FOCUS (2001) did not specify a value of the overall percentile, but used, e.g. for spray drift, 90th 6391 percentile values from the experimental drift dataset they selected. So in view of both FOCUS (2000) 6392 and FOCUS (2001) it seems that it was the intention of FOCUS (2001) to assess a 90th percentile. This 6393 is more or less supported by the statements on p. 109 of FOCUS (2001): (i) ‘The various assumptions 6394 and ‘worst-case’ assessments summarised above show that, for many of the scenario factors that 6395 determine the magnitude and duration of pesticide residues in water bodies, a 90th+ percentile worst-6396



case has been adopted.’ and (ii) ‘The highest PECsw estimates from the ten scenarios are likely to 6397 represent at least a 90th percentile worst-case for surface water exposures resulting from agricultural 6398 pesticide use within the European Union.’ 6399

As described before, EFSA PPR did not yet evaluate whether FOCUS (2001) achieved this 90th 6400 percentile protection level. Furthermore, Risk Managers did not yet take any decision on the exposure 6401 assessment goals for surface water exposure assessment at EU level. 6402



B. BACKGROUND OF THE PROCEDURE FOR PARTITIONING OF SUBSTANCE BETWEEN 6403 WATER AND SEDIMENT IN THE FOCUSSW STEP2 EXPOSURE CALCULATIONS 6404

The effect of the 2/3-available/ 1/3-non-available water compartments for sorption is shown for a 6405 strongly adsorbing, non degrading compound (Koc 10000 L/kg) in Figure B.1. In the example 6406 simulation, a sequence of 5 applications, each with 1 kg/ha, is assumed (field crop, no run-off/drainage 6407 entry, no crop interception). 6408

6409

Figure B.1: Effect of the sediment/water distribution coefficient in the FOCUS STEPS2 model 6410

6411

As shown in Figure B.1 the concept of available/non-available water compartments delays the 6412 partitioning between water and sediment compared to the normal equilibrium approach. 6413

After the run-off/drainage event (which occurred on day 32 in Figure B.1) the standard equilibrium 6414 equation between water and sediment is used by setting K, the fraction of PPP mass in water available 6415 for sorption, to 1). Consequently, the two lines in the diagram totally overlap after the run-off/drainage 6416 event. 6417

The figure shows also that the steady-state value of the system with the special sediment/water 6418 distribution coefficient is higher than that of the normal equilibrium approach for the loads due to 6419 spray drift. This can be understood as follows. If Feq is defined as the equilibrium fraction in the water 6420 phase as defined by Eqn 4.3 and Favailable is defined as the fraction in the system in the available pool at 6421 steady state (so sum of available water and sorbed to sediment). In the steady-state situation, the 6422 fractions in the different compartments have to be as follows: 6423

Available water: Feq · Favailable (consequence of definitions of Feq and Favailable) 6424

Unavailable water: 0.5 · Feq · Favailable (because it is 50% of the available water) 6425

Sediment: (1- Feq ) · Favailable (consequence of definitions of Feq and Favailable) 6426

6427

The sum of the three compartments is (1+ 0.5 Feq) Favailable which is by definition 1.0 . 6428

Effect of different pesticide mass fractions in water available for sorption before the runoff event

0

5

10

15

20

25

30

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38

Days after 1st application

Cw

ater

(µg/

L)

fraction of pesticide mass in water available for sorption before runoff = 2/3

fraction of pesticide mass in water available for sorption before runoff = 1



So the following is obtained: 6429

Favailable = 1/(1+ 0.5 Feq ) 6430

Thus the fraction in the water Fwater in Step 2 after spray applications can be calculated with 6431

Fwater = 1.5 Feq / (1 + 0.5 Feq) = 3 Feq / (2 + Feq) 6432

6433

This can be illustrated with the first concentration peak in Figure B.1. Feq for this system is 0.070, so 6434 the equilibrium concentration is 6 times 0.070 = 0.42 µg/L. The above equation then gives an Fwater for 6435 this system of 0.101 so the equilibrium concentration in Step 2 is 6 times 0.101 = 0.61µg/L. 6436

In FOCUS Step 3 the partitioning between water and sediment is described with TOXSWA which 6437 assumes a perfectly mixed water layer and diffusion into sediment. It is interesting to know to what 6438 extent the decline in water in Step 2 is comparable to that in the equivalent stagnant TOXSWA 6439 system. The Panel tested this by considering a system without degradation in water and sediment. For 6440 such a system, there is an analytical solution available (Eqn. 4.43 of Crank, 1967). Step 2 calculations 6441 were made for Koc values of 10, 1000 and 100 000 L/kg, a dosage of 1 kg/ha and an application that 6442 generated 12.1 % spray drift with no degradation in water and sediment. These were compared with 6443 the analytical solution assuming a dry bulk density of 0.8 kg/L, a porosity of 0.68, a tortuosity of 6444 0.565, a diffusion coefficient in the liquid phase of 0.43 cm2/d and the same dosage (and of course a 6445 water layer of 30 cm and an organic carbon content of 5 %). The porosity was calculated from the dry 6446 bulk density and the organic carbon content. The tortuosity was calculated using the same equation as 6447 in TOXSWA. The results in Figure B.2 indicate that for Koc = 10 L/kg there is almost no substance 6448 going to the sediment in both systems. For Koc = 1000 and 100 000 L/kg Step 2 shows a much faster 6449 decline than the analytical solution. 6450

6451

Figure B.2: Comparison between time course of the concentration in water calculated with (i) Step 2 6452 and (ii) the analytical solution for a perfectly mixed water layer with diffusion into the sediment (Eqn 6453 4.43 of Crank, 1967). Calculations are for a dosage of 1 kg/ha and an application that generated 12.1 6454 % spray drift with no degradation in water and sediment. Dashed lines and black symbols are 6455 calculations with Step 2 and solid lines and red symbols are calculations with the analytical solution 6456 for the three Koc values as indicated. 6457



C. COMPARISON OF ACUTE RAINBOW TROUT TOXICITY WITH ACUTE TOXICITY VALUES 6458 FOR AMPHIBIAN SPECIES 6459

C.1 Introduction 6460

Although the new data requirements (SANCO document 11802/2012) do not request specific toxicity 6461 tests for amphibian species it is stated that available and relevant data, including data from the open 6462 literature for the active substance of concern, regarding the potential effects to amphibians shall be 6463 presented and taken into account in the risk assessment. It is unclear from this whether and when it 6464 would be necessary to test substances of concern on amphibians. 6465 6466 In the new data requirements (SANCO document 11802/2012) it is stated that a test on rainbow trout 6467 (Oncorhynchus mykiss) shall be carried out. In contrast with the old data requirements no additional 6468 species to the rainbow trout is required. To determine whether standard tests with fish required for the 6469 dossier would be likely to cover the potential risk to amphibians present in the surface water. acute 6470 toxicity values for fish and amphibians have been compared. 6471 6472 This comparison used the data collected by Fryday and Thompson (2012) on amphibian species 6473 exposed in water. In particular, data used were generated in tests with an exposure duration of 96 6474 hours and employing either flow-through or static-renewal exposure system. 6475 6476 Ideally the corresponding data for rainbow trout would have been taken from the dossiers, but the 6477 EFSA database on endpoints derived from the conclusions on pesticides included rainbow trout LC50 6478 values for only 5 compounds that overlap with the Fryday and Thompson amphibian database. The 6479 second source for fish data.was the Footprint IUPAC Pesticide Property Database (PPDB), and where 6480 this database did not provide an LC50 for rainbow trout, Safety Data Sheets of the industry were used. 6481 6482

C.2 Data used for the comparison 6483

In total, 253 data points for amphibian species with corresponding rainbow trout values were 6484 available, from tests on a plant protection product performed either under flow-through or static-6485 renewal system. For 48 different species a toxicity test with a plant protection product was available 6486 (see Table C.3). Most of the tested species belong to the subclass of Anura (frogs and toads) and 7 of 6487 the tested species to the subclass Caudata (salamanders). 34 % of the tests were carried with Xenopus 6488 laevis, the African clawed frog. All individual values can be found in Fryday and Thompson (2012). 6489 6490 Tests are available for 60 different plant protection products: 7 fungicides, 19 herbicides, 32 6491 insecticides, 1 plant growth regulator and 1 synergist (Table C.4). Only for two compounds was no 6492 LC50 value for the rainbow trout available, and one study had only 2 days of exposure instead of the 6493 standard period of 4 days. 6494 6495 C.3 Results 6496

Figure C.1 depicts the comparison of each amphibian toxicity value with the corresponding toxicity 6497 value for the rainbow trout (Oncorhynchus mykiss). The black line is the 1:1 line: the line indicating 6498 where toxicity to rainbow trout and amphibian would be exactly the same. 6499 In 62% of the cases the rainbow trout is more sensitive than the amphibian species (Table C.1 and C.2, 6500 points above the 1:1 line on figure C1). The red line in figure C1 represents an assessment factor of 6501 100, i.e. where toxicity to amphibian would be exactly 100x higher than toxicity to the rainbow trout. 6502 Only in 2% of the cases is the amphibian test species more than a factor of 100 more sensitive than the 6503 rainbow trout (values below the red line in figure C.1). Only in those cases the RAC for amphibians 6504 would be lower than the RAC based on the rainbow trout. 6505 6506



The dataset of 253 tests with amphibians includes several life stages e.g. tadpoles (including the 6507 Fryday and Thompson category ‘larvae’) and embryos. Repeating this analyses but split by lifestage 6508 (i.e. keeping embryos and larvae separate) gives a comparable to the assessment on the whole dataset 6509 (see Table C.2). Therefore, the results are considered to be valid for both embryos and larvae. The 6510 amphibian toxicity values compared with rainbow trout values did not include adult life stages, and the 6511 level of protection of rainbow trout acute test data for adult amphibians therefore remains uncertain. 6512 However, it should be noted that for the purposes of the surface water risk assessment in field margins, 6513 fully aquatic life stages of amphibians (i.e. embryos and larvae) can be considered the most relevant. 6514 6515 6516

6517

Figure C.1: Comparison of each amphibian toxicity value with the respective toxicity value for 6518 rainbow trout (Oncorhynchus mykiss). The black line is the 1:1 line, i.e. the line indicating that the 6519 outcome for rainbow trout and amphibians would be exactly the same. The red line considers the 6520 Assessment Factor of 100 applied in the acute RA of fish. 6521

6522

Table C.1: Difference between amphibian species (embryos and tadpoles) and rainbow trout 6523

Difference between amphibians and rainbow trout (n = 253)

Amphibian species more sensitive than rainbow trout

Rainbow trout more sensitive than amphibian species

More than factor of 1000 0.4 % 2.8 % Between 100 -1000 times 1.6 % 5.5 % Between 10 -100 times 15.4 % 18.2 % Between 1 and 10 times 20.6 % 35.6 % Less than a factor of 1 62.0 % 38.0 %

6524

6525

6526

Toxicity values for rainbow trout and amphibian species for the same plant protection products

0.00

0.00

0.00

0.01

0.10

1.00

10.00

100.00

1000.00

10000.00

0.00001 0.0001 0.001 0.01 0.1 1 10 100 1000 10000

LC50 values for rainbow trout (mg/l log scale)

LC

50 fo

r A

mph

ibia

n sp

ecie

s (m

g/l l

og sc

ale)



Table C.2: Differences between amphibian embryos or tadpoles, and rainbow trout 6527

Amphibians versus rainbow trout (Embryos n = 54 and tadpoles n = 171)

Amphibian embryos more sensitive than rainbow trout

Amphibian tadpoles more sensitive than rainbow trout

> 1000 times 0 % 1 % 100 -1000 times 4 % 4 % 10 -100 times 13 % 11 % 1 and 10 times 11 % 37 % < 1times 72 % 47 %

6528

Table C.3: Amphibian species included in the database/analysis and number of tests carried out 6529 with each species. 6530

Amphibian species Number of tests Ambystoma gracile 3 Ambystoma laterale 1 Ambystoma maculatum 2 Ambystoma mexicanum 9 Ambystoma opacum 1 Bombina bombina 1 Bufo americanus 6 Bufo boreas 2 Bufo bufo gargarizans 9 Bufo japonicus formosus 6 Bufo melanostictus 1 Centrolene prosoblepon 1 Crinia insignifera 2 Cynops pyrrhogaster 7 Dendrosophus microcephalus 1 Engystomops pustulosus 1 Fejervarya limnocharis 1 Hyla japonica 6 Hyla versicolor 2 Hypsiboas crepitans 1 Hypsiboas pulchellus 2 Limnonectes limnocharis 1 Lithobates catesbeianus 1 Litoria moorei 2 Microhyla ornata 11 Notophthalmus viridescens 1 Physalaemus biligonigerus 1 Pseudacris crucifer 2 Pseudacris regilla 6 Rana cascadae 2 Rana catesbeiana 8 Rana clamitans 15 Rana cyanophlyctis 4 Rana hexadactyla 4 Rana limnocharis 1 Rana nigromaculata 7 Rana pipiens 9 Rana spinosa 1



Rana sylvatica 1 Rhacophorus arboreus 4 Rhinella arenarum 12 Rhinella granulosa 1 Rhinella marina 1 Rhinella typhonius 1 Scinax nasicus 1 Scinax ruber 1 Silurana tropicalis 3 Xenopus laevis 87

6531

Table C.4: Compounds tested for Amphibians, fish toxicity values 96-hrs LC50 in mg/l and 6532 references (website where values were found including day of downloading). 6533

Compound LC50 Remark Day of References mg/l downloading 2,4-D 250 11-11-2012 http://www.kellysolutions.com/erenewals/documentsubmit/

KellyData%5COK%5Cpesticide%5CMSDS%5C42750%5C42750-107%5C42750-107_PD_2_5_8_2007_1_39_51_PM.pdf

Acephate 110 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Acetochlor 0.36 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Acrolein 0.15 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Acrylonitrile 70 11-11-2012 http://www.petrochemistry.net/ftp/pressroom/Microsoft%20

Word%20-%20MSDS%20generic%203%2008.pdf Aldoxycarb 35.3 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Arsenous oxide

18.8 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/

Atrazine 4.5 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Azinphos-methyl


Butachlor 0.52 11-11-2012 http://www.tlongagro.com/news_en/images/100728_1.pdf Carbendazim 0.19 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Carbofuran 22 11-11-2012 http://www.mingdouchem.com/Enweb/UploadFiles/201009

02102011404.pdf Chlorimuron-ethyl

1000 11-11-2012 http://msds.dupont.com/msds/pdfs/EN/PEN_09004a3580169c5a.pdf

Chlorpyrifos 0.0013 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Copper sulfate


Cypermethrin 0.0028 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ DDD 0.07 11-11-2012 http://webwiser.nlm.nih.gov/getSubstanceData.do;jsessionid

=D71974FAF99B1421B5256A8607EFFA77?substanceID=30&displaySubstanceName=pp-TDE&UNNAID=&STCCID=&selectedDataMenuItemID=79

DDT 7 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Deltamethrin 0.0002

6 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/

Diazinon 3.1 10-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Dieldrin 0.0012 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Dimethoate 30.2 16-11-2012 EFSA database Diuron 5.6 11-11-2012 http://nuturf.com.au/commerce/nuturf/msds133.pdf



Compound LC50 Remark Day of References mg/l downloading Endosulfan 0.002 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Endrin 0.0007


Esfenvalerate 0.00007

11-11-2012 http://msds.dupont.com/msds/pdfs/EN/PEN_09004a358060a40d.pdf

Fenitrothion 1.3 16-11-2012 EFSA database Gamma-HCH, alpha-HCH, lindane


Glyphosate 38 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Glyphosate isopropylammonium

7.5 11-11-2012 http://technical.nufarm.co.uk/documents/Herbicide/Safety/Clinic%20Ace.pdf

Imidacloprid 211 16-11-2012 EFSA database Malathion 0.018 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Mancozeb 0.074 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Maneb 0.2 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Mefenacet 6 Carp (a) 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Mercuric chloride

0.814 11-11-2012 http://scialert.net/fulltext/?doi=pjbs.2007.1098.1102

Myclobutanil 2 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Naphthalene 0.11 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Nicosulfuron 65.7 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Nonanoic acid

91 11-11-2012 http://apps.echa.europa.eu/registered/data/dossiers/DISS-9d8487fe-dd23-01ab-e044-00144f67d249/AGGR-3b0f8bc9-3bd9-4dac-b15e-7d14f920da40_DISS-9d8487fe-dd23-01ab-e044-00144f67d249.html

Paclobutrazol 23.6 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Paraquat 19 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Paraquat dichloride


Parathion-methyl, methyl parathion


Pentachlorophenol


Permethrin 0.0125 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Phosmet 0.23 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Piperonyl butoxide

6.12 11-11-2012 http://www.kellysolutions.com/erenewals/documentsubmit/KellyData%5COK%5Cpesticide%5CMSDS%5C655%5C655-665%5C655-665_Pyronyl_Ul_100_Concentrate_Prentox_9_22_2005_2_11_08_PM.pdf

Pirimicarb 79 16-11-2012 EFSA database Prochloraz 1.5 16-11-2012 EFSA database Profenofos 0.08 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Rotenone 0.0019 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Simetryn 7 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Sodium pentachloro-phenoxide

0.17 48 hrs 11-11-2012 http://pmep.cce.cornell.edu/profiles/fung-nemat/febuconazole-sulfur/pentachlorophenol/prof-pentachlorophenol.html



Compound LC50 Remark Day of References mg/l downloading Sulfometuron-methyl


Temephos 3.49 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ Thiobencarb 1.1 11-11-2012 http://www.bayercropscience.com.au/resources/uploads/ms

ds/file7380.pdf?201211470908 Trichloro-acetic acid

2000 Lepomis macro-chirus

11-11-2012 http://www.biovision.com/manuals/K860_MSDS.pdf?osCsid=sr2be5r0v8ocfvjbufqsj7bul2

Triclopyr-butotyl

0.65 11-11-2012 http://www.clarence.nsw.gov.au/content/uploads/grazon_dsmsd.pdf

Trifluralin 0.088 9-11-2012 http://sitem.herts.ac.uk/aeru/iupac/ (a) According to http://www.agropages.com/agrodata/Detail-891.htm the tested species is a carp 6534 6535



D. INFORMATION ON LIFE CYCLE CHARACTERISTICS FOR AQUATIC ORGANISMS 6536

This Appendix provides some indicative information on life cycle characteristics for aquatic 6537 organisms available in the scientific literature and on the internet. This collection is not comprehensive 6538 and intended only to support e.g. 6539

• to decide if individuals may be exposed to repeated pulses of the a.s. during their life-span 6540 (section 7.1.3) 6541

• whether species are to be considered uni/semivotine (section 7.3.2.1) 6542 This information should be however used with caution since the life-cycle characteristics and overall 6543 life span may vary e.g. depending on the climatic zone. 6544 6545 6546 Generation time for various groups of aquatic organisms as derived from the PondFX Aquatic Life 6547 Database (www.ent.orst.edu/PondFX ) and Barnthouse (2004) 6548 6549 Taxon Generation time in days

Mean (range)

Phytoplankton 1

Lemna 3

Rotifera 8 (6 – 35)

Cladocera 14

Copepoda 61 (14 – 73)

Oligochaeta 105 (51 – 730)

Amphipoda 73 (105 – 250)

Ostracoda 121 (51 – 362)

Gastropoda 513 (105 – ?)

Bivalvia 256 (105 – ?)

Coleoptera (209 – ?)

Diptera (81 – 503)

Ephemeroptera (81 – 730)

Hemiptera (81 – 503)

Trichoptera (162 – 1,264)

Fish (short life cycle) 181

Fish (long life cycle) 1,673

6550

More information on life cycle traits can be found in the SPEAR documentation: 6551 http://www.systemecology.eu/spear/ 6552

Within this calculator life cycle traits as well as relative sensitivity values of around 1400 taxa are 6553 stored and can be retrieved. 6554

6555



E. VARIABILITY IN EXPOSURE-RESPONSE RELATIONSHIPS BETWEEN MICRO-6556 /MESOCOSM EXPERIMENTS PERFORMED WITH THE SAME PPP 6557

6558 For a few PPPs only, more than 3 micro-/mesocosm experiments have been performed that studied a 6559 similar exposure regime. The information available for the organophophorous insecticide chlorpyrifos 6560 in particular allows the evaluation of effects of a single pulse exposure regime (Table E.1). 6561 6562 Table E.1: Effect class concentrations (in µg/L) of the most sensitive measurement endpoint in 6563 micro/mesocosm experiments that studied the impact of single pulse, repeated pulse and chronic 6564 exposures of the insecticide chlorpyrifos. The Effect classes are expressed in terms of nominal 6565 concentrations. These nominal concentrations generally were within 20% of the exposure 6566 concentrations on basis of measurements in the application solutions or in the water column of the test 6567 systems. 6568 Exposure regime

Effect class 1

Effect class 2

Effect class 3A

Effect class 4-5

Type of test system

Reference; Country

Single pulse (peak)

0.1 0.3 1.0 3.0 Outdoor lentic microcosm

Biever et al. 1994; USA

Single pulse (peak)

- 0.1- - 0.9 Outdoor lentic mesocosms

Van den Brink et al. 1996; NL

Single pulse (peak)

0.1 - - 1.0 Outdoor lentic mesocosms

Lopez-Mancisidor et al. 2007; Spain

Single pulse (peak)

0.1 - - 1.0 Outdoor lentic mesocosm

Daam et al. 2008; Thailand

Single pulse (peak)

0.1 - (5*) - Outdoor lotic mesocosm

Pusey et al. 1994; Australia

Single pulse (peak)

- - 0.5 6.3 Outdoor lentic mesocosm

Siefert et al. 1989; USA

Single pulse (peak)

0.1 - 1.0 10 Indoor lentic cosm; 16 °C, mesotrophic

Van Wijngaarden et al. 2005; NL

Single pulse (peak)

0.1 - 1.0 - Indoor lentic cosm; 26 °C, mesotrophic


Single pulse (peak)

0.1 - - 1.0 Indoor lentic cosm; 26 °C, eutrophic


Repeated pulse (4x)

0.033 0.1 1 Outdoor lentic mesocosms

Lopez-Mancisidor et al. 2008; Spain

Constant chronic (28 d)

- - - 0.1 Indoor lentic microcosm

Van den Brink et al. 1995

Constant chronic (28 d)

- 0.01** - 0.1** Indoor lentic microcosm


* recovery is fast because of constant input of propagules in experimental stream after pulse exposure 6569 ** exposure to a mixture of chlorpyrifos and lindane; all treatment-related effects were assigned to chlorpyrofos 6570 6571 For 8 aquatic micro-/mesocosm experiments, performed in different parts of the world and/or under 6572 different experimental conditions, an Effect class 1-2 response was observed at a peak concentration 6573 of 0.1 µg chlorpyrifos/L. Note that this is partly due to the fact that similar exposure concentrations 6574 were selected by the different experimenters. Nevertheless, the similarity between Effect class 1-2 6575 responses between different studies can be explained by the fact that both crustaceans and insects are 6576



sensitive to this insecticide and that the communities of the micro-/mesocosm test systems used all 6577 contained a reasonably high diversity of these arthropods. 6578 6579 It appears that differences in these Effect class 3A concentrations between studies are relatively large. 6580 Note, however, that from a regulatory point of view it is fair to make a distinction in recovery of 6581 sensitive arthropods between hydrologically isolated test systems (lentic micro-/mesocosms: Effect 6582 class 3A concentrations < 1.0 µg/L) and the outdoor stream in which a more or less constant inflow of 6583 sensitive stream invertebrates was possible (resulting in an Effect class 3A concentration of 5 µg/L). It 6584 also appears from the chlorpyrifos data presented in Table E.1 that the threshold concentration (Effect 6585 class 1) of the repeated (4x) pulse exposure study is a factor of approximately 3 lower than that of the 6586 single exposure studies. Treatment-related effects due to a constant chronic exposure probably occur at 6587 concentrations equal to higher than 0.01 µg chlorpyrifos/L. 6588 6589 For the pyrethroid insecticide lambda-cyhalothrin the majority of micro-/mesocosm experiments 6590 available concerns repeated application studies (Table E.2). It appears that the variability in Effect 6591 class 1 (n=2) and Effect class 2 (n=4) responses between different studies is remarkably low, while 6592 that for Effect class 3A (n=6) responses is somewhat higher. 6593 6594 Table E.2: Effect class concentrations (in ng/L) of the most sensitive measurement endpoint in 6595 micro/mesocosm experiments that studied the impact of pulsed exposures of the insecticide lambda-6596 cyhalothrin. The Effect classes are expressed in terms of nominal peak concentrations. In most studies 6597 the nominal concentrations were in accordance with measurements of the test substance in the 6598 application solutions. 6599 Exposure regime

Effect class 1

Effect class 2

Effect class 3A

Effect class 4-5

Type of test system

Reference; Country

Single pulse

-

- 50 - Outdoor lotic mesocosms

Heckmann & Friberg 2005; Denmark

Repeated pulse (12x)

2.7*

- - 27* Outdoor lentic mesocosms

Hill et al. 1994; USA

Repeated pulse (2x)

4.0** - 16** 85** Outdoor lentic mesocosms

Arts et al. 2006; NL

Repeated pulse (5x)

- 10** 25** 50** Indoor lentic microcosms


Repeated pulse (3x)

- 10 10 25 Outdoor lentic microcosm

Roessink et al. 2005; NL

Repeated pulse (3x)

- 10 50 - Outdoor lentic microcosm

Roessink et al. 2005; NL

Repeated pulse (3x)

- 10 10 - 25 50 Outdoor lentic microcosms


Repeated pulse (3x)

- - - 17 Outdoor lentic mesocosms

Farmer et al. 1995; UK

* experiment was characterized by both spray-drift (nominal 1.6 and 16 µg/L) and run-off applications (nominal 4.7 and 47 6600 µg/L). As exposure concentration the median value for the spray drift and run-off application was used. 6601 ** exposure to a realistic package of different PPPs used in a specific crop including lambda-cyhalothrin; all treatment-6602 related effects were assigned to lambda-cyhalothrin 6603 6604 For the pyrethroid insecticide esfenvalerate the majority of micro-/mesocosm experiments available 6605 concerns pulsed exposures (single or repeated applications) (Table A.3). It appears that the variability 6606 in Effect class 1-2 (n=2) concentrations of the two repeatedly exposed complex mesocosm studies 6607 (including many insect populations) are similar (0.01 µg/L). In addition, an Effect class 3A 6608



concentration of 0.03 was observed for the most sensitive endpoint reported (in the paper) for a simple 6609 plankton dominated microcosm test (NSH-NH microcosms of Stampfli et al (2011). Note that when an 6610 additional stressor in the form harvesting a considerable proportion of zooplankton (NSH-H 6611 microcosm) coincided with the esfenvalerate-stress an Effect class 2 concentration of 0.03 µg/L was 6612 observed for the most sensitive endpoint reported, while the additional stressor of shading the 6613 microcosms (SH-NH microcosms) resulted in a similar LOEC of 0.03 µg/L on basis of the most 6614 sensitive endpoint reported. 6615 6616 Table E.3: Effect class concentrations (in µg/L) of the most sensitive measurement endpoint in 6617 micro/mesocosm experiments that studied the impact of (short-term) pulsed exposures of the 6618 insecticides esfenvalerate. The Effect classes are expressed in terms of nominal peak concentrations. 6619 In most studies the nominal concentrations were in accordance with measurements of the test 6620 substance in the application solutions.*= focus on plankton organisms only; # additional stressor 6621 Exposure regime

Effect class 1

Effect class 2

Effect class 3A

Effect class 4-5

Type of test system

Reference; Country

Repeated pulse (2x) Esfenvalerate

- 0.01 - 0.08 Littoral enclosures

Lozano et al., 1992; USA

Repeated pulse (10x) Esfenvalerate

0.01 - - 0.25 Outdoor mesocosms

Webber et al., 1992 USA

Singe pulse Esfenvalerate

- - 0.03* 0.3* NSH-NH microcoms

Stampfli et al. 2011

Single pulse Esfenvalerate

-

0.03*# - 3*# NSH-H microcosms

Stampfli et al 2011 Germany

Single pulse Esfenvalerate

- 0.03*# 3*# SH-NH microcosms

Stampfli et al. 2011

6622 6623 There appears to be limited information on PPP-treated model ecosystems comparing Effect class 1 or 6624 Effect class 2 concentrations for direct toxic effects as a result of more or less constant chronic 6625 exposure. The limited microcosm/mesocosm information available for the persistent fungicide 6626 carbendazim suggests little variation in Effect class 1 concentrations between experiments as a result 6627 of a long-term chronic exposure regime (Table E.4). 6628 6629 Table E.4: Effect class concentrations (in µg/L) of the most sensitive measurement endpoint in 6630 micro/mesocosm experiments (fish not present) that studied the impact of more or less constant 6631 exposure of the fungicide carbendazim. 6632 Exposure regime

Effect class 1

Effect class 2

Effect class 3

Effect class 4

Type of test system

Reference; Country

Long term 2.6

- - 26.4 Outdoor microcosms

Daam et al. 2009a Thailand

Long term 2.2

- - 20.7 Outdoor mesocosms

Slijkerman et al. 2004, NL

Long term 3.3 - - 33.0 Indoor microcosms

Cuppen et al. 2000; Van den Brink et al. 2000; NL

6633 Only for the persistent herbicide atrazine a large data set is available (Table E.5). 6634 6635 6636 6637 6638 6639



Table E.5: Effect class concentrations (in µg/L) of the most sensitive measurement endpoint in 6640 micro/mesocosm experiments that studied the impact of more or less constant long-term exposure of 6641 the herbicide atrazine. 6642 Exposure regime

Effect class 1

Effect class 2

Effect class 3B

Effect class 4-5

Type of test system

Reference

Long-term - 2 - 30 Outdoor lentic mesocosms

Seguin et al. 2001;

Long-term 5

- - - Indoor lentic microcosms


Long-term - 5 - - Indoor lotic microcosms

Gruessner & Watzin, 1996

Long-term 5 10 - 22 Outdoor lentic microcosms

Jüttner et al. 1995

Long-term - 10 - 100 Indoor lentic microcosm

Johnson 1986

Long-term 5 - 50 100 Indoor lentic microcosm

Brockway et al. 1984

Long-term 10 - - 32 Indoor lentic microcosms

Pratt et al. 1988

Long-term - - - 10 Indoor lotic microcosms

Kosinsky 1984, Kosinsky & Merkle, 1984

Long-term 14 25 - 80 Indoor lotic microcosms

Nyström et al. 2000


Muňos et al. 2001

Long-term - - - 15 Experimental swamp

Detenbeck et al. 1996

Long-term - - - 20 Outdoor lentic mesocosms

DeNoyelles et al. 1994 (and literature cited)

Long-term - 20 - 100 Indoor lentic microcosms

Stay et al. 1989


Krieger et al. 1988

Long-term - - - 50 Outdoor lentic mesocosms

Fairchild et al. 1994

6643 Data available for atrazine suggest a larger variability in class 1 and class 2 effect concentrations 6644 between chronic exposure experiments; however, also a larger number of studies is available. Effect 6645 class 1 concentrations could be derived from 5 different atrazine studies, and Effect class 2 6646 concentrations for 6 studies (Table E.5). 6647 6648 The relatively high variability in Effect class 1 – 2 concentrations for chronic studies with atrazine 6649 when compared with those with pulsed exposures to chlorpyrifos and lambda-cyhalothrin might be 6650 explained by differences in toxic mode-of-action between these substances. Atrazine is a photosystem 6651 II inhibitor. According to Guasch & Sabater (1998), inhibition of photosynthesis by atrazine is 6652 influenced by ambient light conditions, which most probably considerably varied between the 6653 different micro-/mesocosm studies reported in Table E.5. Consequently, a question at stake is whether 6654 the results from the chronic micro-/mesocosm studies with atrazine are representative for PPPs with 6655 another toxic mode-of-action. 6656



GLOSSARY AND ABBREVIATIONS a.s. active substance

AMRAP Aquatic Macrophyte Risk Assessment for Pesticides, SETAC Europe, 2009 2nd SETAC Europe Special Science Symposium, Brussels, Belgium, 2009-09-17/ 2009-09-18

AMPERE Aquatic Mesocosms in Pesticide Registration in Europe: Recent Experiences (AMPERE) SETAC Europe Workshop, Leipzig, Germany, 24 - 25 April 2007

CA Concentration Addition

CLASSIC Community Level Aquatic System Studies Interpretation Criteria (CLASSIC). SETAC Europe/OECD/EC/BBA/UBA Workshop, 1999

DG SANCO Directorate General for Health and Consumer Affairs

ECHA European Chemicals Agency

ECx Concentration where x% effect was observed/calculated

EFSA European Food Safety Authority

ELink Linking Aquatic Exposure and Effects in the Registration Procedure of Plant Protection Products (Brock et al 2010c)

ERA Environmental Risk Assessment

ERC Ecotoxicologically Relevant Concentration

EQS Environmental Quality Standards

ETR Exposure-Toxicity Ratio

EU European Union

Exposure profile The course of time of the concentration on a relative concentration scale (an effect study is usually carried out at different concentration levels but for the same exposure profile).

FOCUS FOrum for the Co-ordination of pesticide fate models and their USe

GD Guidance Document

HARAP Higher-tier Aquatic Risk Assessment for Pesticides (HARAP), SETAC Europe workshop

HCx Hazardous concentration for x % of the species of a SSD

LLHC5 Lower limit of the confidence interval of the hazardous concentration for 5% of the species of a SSD

LOEC Lowest Observed Effect Concentration



NOEC No Observed Effect Concentration

Metabolite Any metabolite or a degradation product of an active substance, safener or synergist, formed either in organisms or in the environment (thus including also oxidation products which may have a larger molecular mass than the parent substance) (EFSA, 2012c).

MDD Minimal Detectable Difference

OECD Organization for Economic Cooperation and Development

QSAR Quantitative structure-activity relationship

PEC Predicted Environmental Concentration

PPP Plant Protection Product

PPR Panel EFSA’s Panel on Plant Protection Products and their Residues

RA Risk Assessment

RAC Regulatory Acceptable Concentration

REACH Registration, Evaluation, Authorisation and Restriction of Chemical substances

Ri Reliability Index

SCFCAH Standing Committee on the Food Chain and Animal Health

SETAC Society for Environmental Toxicology and Chemistry

SPG Specific Protection Goal

SSD Species Sensitivity Distribution

TK/TD Toxicodynamics/toxicokinetic

TU Toxic Unit

TWA Time weighted average

WFD Water Framework Directive