(Oilseed Rape)?

From: Pywell, Richard F. Sent: 19 June 2013 11:50 To: @syngenta.com; @syngenta.com Cc: ; Bullock, James M. Subject: RE: Landscape study and bee health - CEH comments and costs

Dear ,

Below are some comments on the proposed NNI landscape-scale study and an indication of our costs.

1. Overall impression A very timely study with a strong scientific basis. We agree this should be undertaken by five independent research organisations across Europe as this adds power to the study. CEH would be very keen to be involved in this. A minor comment on the presentation. Would be best to make the project outcomes (slide 2) more objective, i.e. don’t say ‘Demonstrate NNIs have no impacts on bees…’ say something like ‘Measure/quantify the effects of NNIs on wild bees and honeybees in real farming situations etc…’

2. Project management One of the five partner countries needs to take on an overall leadership and co-ordination role to ensure consistency of design and data collection. It would also be worth involving a qualified statistician in the design to add validity.

3. Experimental design You may not need a fully factorial design to answer your research questions (i.e. all combinations of +/- NNI with +/- BMP). The following might suffice and importantly would allow more replication (within each country): a) -NNI with -BMP (current situation with the EU ban) b) +NNI with -BMP (previous situation before the ban) c) +NNI with +BMP (research treatment) Comparison of a) with b) will tell you what effect NNI has Comparison of b) with c) will tell you if you can mitigate the effects of NNI (if there are any) A treatment of –NNI with +BMP is telling you about mitigation without NNIs so might not be needed.

We would recommend some limited replication of these treatments within each country (such that each country could report separately if required and to provide a margin for loss of sites etc). Five replicates of the three treatments per country would seem reasonable. (Total 15 sites/plots per country)

Each treatment ‘plot’ should comprise a radius of 1 km of arable land containing a high proportion of oilseed rape to which the treatment is applied. This size is a compromise between mean foraging range of pollinators and cost/practical ability to apply good quality BMP to large areas of farmland. There should be separation between treatment plots such that bees cannot fly between them (say 5km).

Blocking of the treatment plots should be by region/landscape to take account of landscape structure etc. It is recommended that the structure / composition of each landscape is quantified using earth observation data at the design stage. CEH could undertake this using the approaches developed for the UK oilseed rape study.

4. Response variables (standard measures across countries) Pollinators – aim to cover response of all major pollinator groups: Honeybee colony health/productivity etc. CEH would need to go into partnership with FERA in the UK as they are the experts in this. Bombus terrestris commercially produced colonies (large number per site as response is variable) – productivity, reproductives etc Osmia rufa commercially produced / or country-specific solitary bees e.g. consider using Andreana sp. For the UK currently cultured by ? Secondary measures – abundance counts, traps nests for solitary bees etc

5. Practical issues Ironically we think the European ban on NNIs makes this project more feasible because we are asking selected farmers to use NNI which will benefit their yield (as opposed to the pre-Ban situation where we would ask some farmers to not use NNI at a major cost to their yield). Syngenta should consider supplying all the seed to each farmer as this will ensure a consistent variety of OSR is grown in each country. It will also encourage the no NNI farmers to join the project. BMP should be a similar concept for each country but may need to be tweaked. Operation Pollinator is a good starting point. Syngenta should consider supplying the seed for this. It is important to ensure all the BMP measures (either habitat or honeybee husbandry) are done to a consistently high standard. This would mean advice and visits to the participating farms and should be costed in. In the UK we would use Wildlife Farming Company to do this as a subcontractor.

6. Costs If CEH were to take on the lead for this project it would require a senior scientist full time (in addition to the time of the field teams, CEH statistician, FERA subcontract, WFC subcontract etc). A ball park estimate for this would be in the region of total for the 3 years.

This is a very exciting project and I would be happy to discuss further with you both

Best wishes

Richard

From: @syngenta.com [mailto: @syngenta.com] Sent: 17 June 2013 15:17 To: Pywell, Richard F.; @syngenta.com Subject: Fw: Landscape study and bee health 5.pptx

Fyi

From: To: Sent: Mon Jun 17 12:58:46 2013 Subject: Landscape study and bee health 5.pptx

Here are a few thoughts, mainly to confirm alignment with your points.

We are planning to have a combination of a scientific study with a wider demonstration initiative. This provides a platform to interact with national institutes and supports communication of scientific data to a wider (non- scientific) community, in a credible way.

While the statistical robustness is critical for the design of the study and the acceptability of data, this doesn’t imply a regulatory study.

Having on board some local institutes should support a direct dialogue with the Commission and EFSA and a blessing on the methodology.

We are planning just one single crop, as this would be a great simplification. We may need to check this with the local institutes.

Also, other companies will be made aware of the project and have the possibility to join, if interested.

CRO capacity, and costs are the real issues. We are waiting for news from our procurement that is currently understanding the availability of few CROs for the demonstration part.

I will continue to keep you updated and ask your advice.

Best regards,

Syngenta UK Limited, Registered in England No 00849037 Registered Office : CPC 4, Capital Park, Fulbourn, Cambridge, CB21 5XE, United Kingdom

This message may contain confidential information. If you are not the designated recipient, please notify the sender immediately, and delete the original and any copies. Any use of the message by you is prohibited. From: @syngenta.com Sent: 19 June 2013 12:13 To: Pywell, Richard F.; @syngenta.com Cc: Bullock, James M. Subject: Re: Landscape study and bee health - CEH comments and costs

Dear Richard,

Thank you for this comprehensive answer.

I have a review with out top management tonight. I expect to be able to come back to you soon, with some comments.

Best regards,

From: Pywell, Richard F. To: Cc: @ceh.ac.uk>; Bullock, James M. Sent: Wed Jun 19 12:50:10 2013 Subject: RE: Landscape study and bee health - CEH comments and costs

Dear

Below are some comments on the proposed NNI landscape-scale study and an indication of our costs.

This is a very exciting project and I would be happy to discuss further with you both

Best wishes

Richard

From: @syngenta.com [mailto: @syngenta.com] Sent: 17 June 2013 15:17 To: Pywell, Richard F.; @syngenta.com Subject: Fw: Landscape study and bee health 5.pptx

Fyi

From: To: GBFB Sent: Mon Jun 17 12:58:46 2013 Subject: Landscape study and bee health 5.pptx

Here are a few thoughts, mainly to confirm alignment with your points.

While the statistical robustness is critical for the design of the study and the acceptability of data, this doesn’t imply a regulatory study.

Having on board some local institutes should support a direct dialogue with the Commission and EFSA and a blessing on the methodology.

We are planning just one single crop, as this would be a great simplification. We may need to check this with the local institutes.

Also, other companies will be made aware of the project and have the possibility to join, if interested.

CRO capacity, and costs are the real issues. We are waiting for news from our procurement that is currently understanding the availability of few CROs for the demonstration part.

I will continue to keep you updated and ask your advice.

Best regards,

Syngenta UK Limited, Registered in England No 00849037 Registered Office : CPC 4, Capital Park, Fulbourn, Cambridge, CB21 5XE, United Kingdom

This message (and any attachments) is for the recipient only. NERC is subject to the Freedom of Information Act 2000 and the contents of this email and any reply you make may be disclosed by NERC unless it is exempt from release under the Act. Any material supplied to NERC may be stored in an electronic records management system.

This message may contain confidential information. If you are not the designated recipient, please notify the sender immediately, and delete the original and any copies. Any use of the message by you is prohibited. From: @syngenta.com Sent: 20 June 2013 16:46 To: Pywell, Richard F. Cc: Bullock, James M.; @syngenta.com; @syngenta.com Subject: RE: Landscape study and bee health - CEH comments and costs Attachments: Landscape study and bee health 7.pptx

Dear Richard,

The meeting with our COO went well. He appreciated very much your comments and your offer to coordinate the project.

Next step is now on Monday with the Regional Head – I am sure he will provide us with the expected final blessing.

Let me share some of the suggestions/ options to be evaluated that emerged from yesterday’s meeting:

· Limit the field margins only to some of the farms – demonstrate that this is not a way to offset the negative impact of NNIs

· Add a blind experimental protocol to test the seeds (systemic or NNIs Vs. non systemic) · Report the progress through agriculture magazines · Add, as you suggested, a statistician to support the design

I suggest, I will call you next week to finally confirm that we are ready to start.

Also, please find attached an updated plan.

Thank you again.

Best wishes,

From: Pywell, Richard F. [mailto:[email protected]] Sent: Mittwoch, 19. Juni 2013 12:50 To: Cc: Bullock, James M. Subject: RE: Landscape study and bee health - CEH comments and costs

Dear ,

Below are some comments on the proposed NNI landscape-scale study and an indication of our costs.

This is a very exciting project and I would be happy to discuss further with you both

Best wishes

Richard

From: @syngenta.com [mailto: @syngenta.com] Sent: 17 June 2013 15:17 To: Pywell, Richard F.; @syngenta.com Subject: Fw: Landscape study and bee health 5.pptx

Fyi

From: To: Sent: Mon Jun 17 12:58:46 2013 Subject: Landscape study and bee health 5.pptx

Here are a few thoughts, mainly to confirm alignment with your points.

While the statistical robustness is critical for the design of the study and the acceptability of data, this doesn’t imply a regulatory study.

Having on board some local institutes should support a direct dialogue with the Commission and EFSA and a blessing on the methodology.

We are planning just one single crop, as this would be a great simplification. We may need to check this with the local institutes.

Also, other companies will be made aware of the project and have the possibility to join, if interested.

CRO capacity, and costs are the real issues. We are waiting for news from our procurement that is currently understanding the availability of few CROs for the demonstration part.

I will continue to keep you updated and ask your advice.

Best regards,

Syngenta UK Limited, Registered in England No 00849037 Registered Office : CPC 4, Capital Park, Fulbourn, Cambridge, CB21 5XE, United Kingdom

This message may contain confidential information. If you are not the designated recipient, please notify the sender immediately, and delete the original and any copies. Any use of the message by you is prohibited. From: @syngenta.com Sent: 26 June 2013 22:33 To: Pywell, Richard F.; @syngenta.com Cc: Bullock, James M. Subject: RE: Landscape study and bee health - CEH comments and costs Attachments: Landscape study and bee health 10.pptx

Dear Richard,

Sorry for the delay. We have finally the green light to proceed with the study.

The study will be sponsored by Bayer Crop Science (BCS) and Syngenta (SYT). From the second year we expect also our Industry association to join.

The request of our management is to start immediately in order to include the next sowing campaign (August).

The overall leadership and co-ordination role would be with CEH. BCS and SYT will be the sponsors and will follow the development of the project without interfering with its management. As coordinator, you will have full control on the protocol and the design of the research project as well as on the selection of the institutes.

We are ready to supply seeds for both crop and field margins. Though it would be convenient to test the scenario +NNI with +BMP (research treatment) on plots/ farms that have already got field margins.

In each country, experts and subcontractors should work under the coordination of the Institute.

As you say, this is a very exciting project and I would be really pleased to do it with you and your organization.

Attached, please find an updated draft charter.

If you agree on the initiative I suggest that we organize a call ASAP, also involving BCS.

Best regards,

From: Pywell, Richard F. [mailto:[email protected]] Sent: Mittwoch, 19. Juni 2013 12:50 To: Cc: Bullock, James M. Subject: RE: Landscape study and bee health - CEH comments and costs

Dear ,

Below are some comments on the proposed NNI landscape-scale study and an indication of our costs.

fi honeybee husbandry) are done to a consistently high standard. This would mean advice and visits to the participating farms and should be costed in. In the UK we would use Wildlife Farming Company to do this as a subcontractor.

This is a very exciting project and I would be happy to discuss further with you both

Best wishes

Richard

From: @syngenta.com [mailto @syngenta.com] Sent: 17 June 2013 15:17 To: Pywell, Richard F.; @syngenta.com Subject: Fw: Landscape study and bee health 5.pptx

Fyi

From: To: Sent: Mon Jun 17 12:58:46 2013 Subject: Landscape study and bee health 5.pptx

Here are a few thoughts, mainly to confirm alignment with your points.

While the statistical robustness is critical for the design of the study and the acceptability of data, this doesn’t imply a regulatory study.

Having on board some local institutes should support a direct dialogue with the Commission and EFSA and a blessing on the methodology.

We are planning just one single crop, as this would be a great simplification. We may need to check this with the local institutes.

Also, other companies will be made aware of the project and have the possibility to join, if interested.

CRO capacity, and costs are the real issues. We are waiting for news from our procurement that is currently understanding the availability of few CROs for the demonstration part.

I will continue to keep you updated and ask your advice.

Best regards,

Syngenta UK Limited, Registered in England No 00849037 Registered Office : CPC 4, Capital Park, Fulbourn, Cambridge, CB21 5XE, United Kingdom

Dear all,

Below are some issues we would like to discuss at the 11th July teleconference. Please add any items to this:

1. Contract arrangements 2. Project governance 3. Project management 4. Costings 5. Experimental treatment clarification 6. Next steps 7. AOB

Talk to you all at 11:30 am Swiss time and 10:30am UK time.

Teleconference details:

Please note that the meeting’s start time is 10:30 BST, 11:30 CEST. You will be able to join the conference from 10 minutes before the meeting’s start time.

 The conference can be accessed by calling:

595990 from Wallingford offices +44(0)1524 595990 for all external callers

 Conference ID:

 Conference Password:

When you are asked to press ‘the pound key’, please use the # button on the phone keypad.

Please let me know if you require any further information or assistance to join the conference.

Best wishes

Richard

Professor Richard Pywell Science Area Lead: Sustainable Land Management Section Head: Biodiversity Patterns and Processes ______NERC Centre for Ecology and Hydrology Maclean Building, Benson Lane Crowmarsh Gifford Wallingford Oxfordshire OX10 8BB

:01491 692356  : [email protected]

______

From: Pywell, Richard F. Sent: 12 July 2013 15:03 To: @syngenta.com; @syngenta.com; @SYNGENTA.COM; @syngenta.com Cc: Bullock, James M.; Shore, Richard F.; Subject: Response to issues raised in 11 July neonicotinoid-pollinator experiment teleconference

Dear ,

Thank you for the constructive discussion of the issues around the proposed neonicotinoid-pollinator experiment earlier today.

I must re-iterate that CEH are committed to the highest quality independent scientific research and monitoring, and believe it is the only way to address the contentious neonicotinoid and pollinator issue. I would also emphasise that money is not a strong motivating factor for us.

Below I have addressed the key points from our discussion and suggest a way forward:

The challenge Previous field studies on the impacts of neonicotinoids on pollinators have not shown clear cut effects, and have been criticised for poor replication and lack of a proper control treatment. Such control has been difficult or impossible to achieve because of the prevalence of neonicitoids in the agricultural landscape and the ability of pollinators to forage over substantial areas. Laboratory studies, on the over hand, have shown clear effects and highlighted the susceptibility of pollinators other than honey bees, but it is not clear whether the doses used are typical of those experienced by free-living insects or if the observed effects are likely to be ecologically significant in the ‘real world’. The challenge therefore is to undertake a scientifically credible, properly controlled field trial that is capable of detecting any effects of neonicotinoids on pollinators in the real world. The complex mixture of factors affecting pollinator populations will require a very carefully designed and implemented experiment that minimises other sources of variability and focuses just on the effects of the pesticide.

The risks If this pan-European experiment is poorly designed and implemented, the results will not be accepted by the regulators, the scientific community or the public. Also, we would have little credibility if we (or others) were to repeat the experiment at a later date with a different design (i.e. we have one chance to get this right). Such outcomes would do serious reputational damage to Syngenta, Bayer and CEH.

The need for a scientifically rigorous control treatment The use of neonicotinoids is very widespread (with up to 80% of oilseed rape crops treated in the UK) and is likely to remain so until post-harvest 2014. Also, many crop pollinators are inherently mobile species foraging over many kilometres. This means that achieving a scientifically rigorous ‘no neonicotinoid’ control treatment would be very difficult and expensive to maintain before the ban comes into effect. However, once the ban comes into being all rape will in effect be the control treatment (the norm). Moreover, farmers will be keen to join the project to apply neonicotinoids under experimental licence. We therefore think the only credible scientific approach is to wait until the neonicotinoid ban comes into effect across all the study countries post-harvest 2014.

The need for careful control of agronomy, crop variety and mitigation measures There is a limited budget for the project, and field measures and chemical analysis will be expensive. We are therefore likely to only have limited replication of treatments within each country. It is thus essential to minimise variability in response of pollinators due to factors such as crop variety, agronomy, and mitigation measures. For example, different oilseed rape varieties are known to vary in attractiveness to bees; it is possible dwarf varieties may have higher concentrations of neonicotinoids etc. Also, recent research suggests interactive effects of other pesticides on pollinators. Disentangling the effects of variable agronomy, varieties and mitigation measures would require huge replication of treatments and would attract criticism as any discerned effects would just be correlative.

The need for rigorous statistical design (including power testing) It is also critical that we understand the statistical power of any studies to detect effects both within country and across the whole study. For example, if scientific understanding suggests an x% reduction in queen production was ecologically significant, our study design would have to be sensitive enough to detect whether such an effect might have occurred. If this is not done, our studies would be of limited or no help to regulators who will have to determine whether there is acceptable or unacceptable risk. It is important that we gather preliminary information to be able to test the likely power of potential study designs. This might be a valuable function of pilot work in France discussed at the teleconference.

What can realistically be achieved between Aug and Dec 2013? Farmers are 2-3 weeks away from harvesting oilseed rape so no pollinator monitoring is possible this season. The following is a list of what we believe are necessary and realistic activities within this short timeframe:

 Agree contract terms and conditions with Syngenta and Bayer (we cannot begin the project until terms are agreed. This will inevitably take some time)  Establish an international Project Advisory Committee (PAC) comprising EU stakeholders, industry and regulators to oversee and advise the project. Again this will require careful selection of members and will take time to convene. In order to be credible they will need to see and comment on the study design and monitoring protocols at an early stage.  Seek permission for experimental licences to apply neonicotinoids in the five study countries.  Undertake statistical power analyses to determine how many sites are required and how many bee hives are needed to detect a response etc. This is not a trivial task and will require pooling data from a wide variety of sources. This approach was used to give credibility to the Farm-scale Evaluations of GM crops.  Draft design the experiment and monitoring for consultation with PAC  Finalise the experimental design  Write outline specification to circulate to EU scientific partner countries  Agree terms with the EU partners  Using the pooled expertise of the five scientific partner organisation write detailed protocols: - Site selection - Crop agronomy - Mitigation management measures - Pollinator recording  Begin to identify suitable experimental farms

Spring 2014

 Complete site selection  Visit farms to collect baseline data, plan location of mitigation measures with crop rotations for next 3 years  Farmer training in implementation of mitigation  Agronomist training in protocols  Begin implementation of mitigation measures  Pilot testing of field monitoring protocols  Identify potential demonstration farms

Post-harvest 2014

 Complete implementation of mitigation measures  Begin monitoring proper in all countries

Note that we would recommend that the experiment is run for at least 3 years post ban (spring 2015, 2016, 2017) as a minimum to allow for inter annual variability due to weather and other factors, and to test the consistency of any effect (or lack of effect).

Recommended experimental design

Countries: UK, France, Germany, Poland, Hungary Study three to five countries depending on budget

Treatments: SYNGENTA (use Syngenta OSR variety) a) No neonic (control) – test to ensure no neonic residues from soil b) Syngenta neonic c) Syngenta neonic PLUS mitigation measures

BAYER (use Bayer OSR variety) d) No neonic (control) – test to ensure no neonic residues from soil e) Bayer neonic f) Bayer neonic PLUS mitigation measures

Replication of treatments within the same country is essential but will depend on the budget

Crop agronomy:

Agronomy No neonicotinoid (control) Neonicotinoid Autumn Contact insecticide ×2 - Contact fungicide ×2 Contact fungicide ×2 Spring Contact insecticide ×2 Contact insecticide ×2 Contact fungicide ×2 Contact fungicide ×2 Summer Desiccant Desiccant Note: same oilseed rape variety in no neonic and neonic treatment

Further input We are aware that there is considerable expertise in designing and running field-scale neonicotinoid pollinator studies within the product safety division of Syngenta ( ). It would be worth seeking their input into these issues.

Speak to you on Monday afternoon.

Best wishes

Richard

:01491 692356

: [email protected]

______

From: @syngenta.com Sent: 12 July 2013 15:27 To: Pywell, Richard F.; @syngenta.com; @syngenta.com; @syngenta.com Cc: Bullock, James M.; Shore, Richard F.; @syngenta.com Subject: Re: Response to issues raised in 11 July neonicotinoid-pollinator experiment teleconference

Dear Richard,

Thank you very much for your support in this project and your quick answer.

CEH is a point of reference in the scientific community and it is very useful and a great pleasure to count on your collaboration and guidance.

Your solid scientific reputation and independence are the reasons why we have requested your help and are important assets in the prosecution of the overall project.

Best regards,

From: Pywell, Richard F. To: Cc: Bullock, James M. ; Shore, Richard F. ;

Sent: Fri Jul 12 16:03:16 2013 Subject: Response to issues raised in 11 July neonicotinoid-pollinator experiment teleconference

Dear ,

Thank you for the constructive discussion of the issues around the proposed neonicotinoid-pollinator experiment earlier today.

Below I have addressed the key points from our discussion and suggest a way forward:

· Agree contract terms and conditions with Syngenta and Bayer (we cannot begin the project until terms are agreed. This will inevitably take some time)

· Establish an international Project Advisory Committee (PAC) comprising EU stakeholders, industry and regulators to oversee and advise the project. Again this will require careful selection of members and will take time to convene. In order to be credible they will need to see and comment on the study design and monitoring protocols at an early stage.

· Seek permission for experimental licences to apply neonicotinoids in the five study countries.

· Undertake statistical power analyses to determine how many sites are required and how many bee hives are needed to detect a response etc. This is not a trivial task and will require pooling data from a wide variety of sources. This approach was used to give credibility to the Farm-scale Evaluations of GM crops.

· Draft design the experiment and monitoring for consultation with PAC

· Finalise the experimental design

· Write outline specification to circulate to EU scientific partner countries

· Agree terms with the EU partners

· Using the pooled expertise of the five scientific partner organisation write detailed protocols: - Site selection - Crop agronomy - Mitigation management measures - Pollinator recording Begin to identify suitable experimental farms Spring 2014 Complete site selection Visit farms to collect baseline data, plan location of mitigation measures with crop rotations for next 3 years Farmer training in implementation of mitigation Agronomist training in protocols Begin implementation of mitigation measures Pilot testing of field monitoring protocols Identify potential demonstration farms Post-harvest 2014 Complete implementation of mitigation measures Begin monitoring proper in all countries Note that we would recommend that the experiment is run for at least 3 years post ban (spring 2015, 2016, 2017) as a minimum to allow for inter annual variability due to weather and other factors, and to test the consistency of any effect (or lack of effect).

Recommended experimental design Countries: UK, France, Germany, Poland, Hungary Study three to five countries depending on budget

Treatments: SYNGENTA (use Syngenta OSR variety)

a. No neonic (control) – test to ensure no neonic residues from soil b. Syngenta neonic c. Syngenta neonic PLUS mitigation measures

BAYER (use Bayer OSR variety)

d. No neonic (control) – test to ensure no neonic residues from soil e. Bayer neonic f. Bayer neonic PLUS mitigation measures Replication of treatments within the same country is essential but will depend on the budget Crop agronomy: Agronomy No neonicotinoid (control) Neonicotinoid Contact insecticide ×2 - Autumn Contact fungicide ×2 Contact fungicide ×2 Contact insecticide ×2 Contact insecticide ×2 Spring Contact fungicide ×2 Contact fungicide ×2 Summer Desiccant Desiccant Note: same oilseed rape variety in no neonic and neonic treatment

Speak to you on Monday afternoon.

Best wishes

Richard

':01491 692356

*: [email protected]

______This message (and any attachments) is for the recipient only. NERC is subject to the Freedom of Information Act 2000 and the contents of this email and any reply you make may be disclosed by NERC unless it is exempt from release under the Act. Any material supplied to NERC may be stored in an electronic records management system.

Dear ,

As discussed in the teleconference earlier.

Outline of the methodological paper:

As an early deliverable from the neonicotinoid-pollinator project by spring 2014 we would submit a methodological paper to an open-access, peer reviewed journal such as PLoS One. This paper would review the evidence of neonicotinoid impacts on pollinators and describe the most appropriate design of a field experiment to objectively detect any effects. This would be based on a statistical power analysis of the pollinator population measures (e.g. reproductive success, population recovery time) required by the EU regulators to inform the sample size and sensitivity of the design to detect effects of different magnitude. It would also put down a clear marker as to the scientifically rigorous approach we are proposing to take for this project.

This approach was used to great effect for the Farm scale Evaluation of GM herbicide- tolerant crop project (see attached paper).

Best wishes

Richard

:01491 692356

: [email protected]

Journal of Applied BlackwellOxford,JAPPLJournal0021-8901British2401METHODOLOGICAL 2003 Ecological ofUK Science, Applied Society,Ltd Ecology 2003 INSIGHTS BiometricalJ. N. Perry et issues al. of GMHT FSE Ecology 2003 40, 17–31 Design, analysis and statistical power of the Farm-Scale Evaluations of genetically modiﬁed herbicide-tolerant crops

JOE N. PERRY, PETER ROTHERY*, SUZANNE J. CLARK, MATT S. HEARD* and CATHY HAWES† Rothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UK; *NERC Centre for Ecology and Hydrology, Monks Wood, Abbots Ripton, Huntingdon, Cambridgeshire PE28 2LS, UK; and †The Scottish Crop Research Institute, Invergowrie, Dundee DD2 5DA, UK

Summary 1. The effects on British farmland wildlife of the management of four genetically modified herbicide-tolerant crops are currently being studied in a 5-year trial termed the Farm-Scale Evaluations (FSE), the first 4 years of which are completed. The FSE is controversial and extensive. There has been intense scrutiny of the experimental design and proposed analysis, and of the estimated statistical power to detect effects of a given magnitude, should any exist. 2. For each crop, the FSE is a form of on-farm trial with a single composite null hypothesis and a simple randomized block experimental design. This has statistical implications for the imposition of treatments by growers and the need for proper randomization. The choice of a half-field experimental unit was based on field availability, the focus on herbicide management, the need to reduce variability and efficiency gains in sampling effort. Farms and fields were selected to represent the range of variability of geography and intensiveness across Britain for each crop. 3. Results of a power analysis suggested that the planned replication of the FSE of about 60 fields per crop over 3 years would be sufficient to provide useful information, from which valid statistical inferences could be drawn. The achieved replication for spring crops in the FSE exceeded, by more than threefold, that in any of 82 comparable terrestrial manipulative ecological experiments undertaken previously. 4. Here, we exemplify a range of analyses including covariates, interactions between various factors including years and treatments, diagnostic procedures to aid selection of the most efficient statistical model, the estimation of power from coefficients of variation, a novel and apparently robust test statistic and the calculation of overall variance from within- and between-unit variability. Preliminary results indicated that a simple log-normal model appeared adequate for most analyses. 5. Synthesis and applications. Statistical challenges created by the scope of the FSE were resolved from a sound knowledge of good experimental design. There is an urgent need for further statistical studies to develop experimental designs or modelling approaches that allow similar studies of genetically modified (GM) crops, at reduced cost. However, this power analysis has shown that this cannot be achieved at the expense of adequate replication, essential for all risk assessment studies. Key-words: abundance, biodiversity, biometry, GM crops, plant and invertebrate populations, statistical power analysis. Journal of Aplied Ecology (2003) 40, 17–31

18 fields have been sown, over four crops and 3 years. Introduction J. N. Perry et al. There has therefore been intense scrutiny of its design The first genetically modified (GM) crops under con- and analysis and of its estimated statistical power to sideration for commercial planting in the UK have been detect effects of a given magnitude. This study focused on altered to make them less sensitive to broad-spectrum three major biometrical issues: design, analysis and power. herbicides. In 1998, English Nature, the statutory body An overview of the project is given by Firbank et al. (2003). set up to promote the conservation of England’s wildlife, raised concerns that the management of these Design genetically modified herbicide-tolerant (GMHT) crops could result in reductions of plant and invertebrate    populations on which farmland birds and other farmland wildlife depend (Anonymous 1998). In statistical terms the chosen design for the FSE is There is evidence that farmland wildlife has already straightforward. It is a paired-comparison experiment been affected deleteriously by the intensification of in a randomized block design, with a single treatment agriculture (Krebs et al. 1999; Robinson & Sutherland factor at two levels. Each block is a single field; in any 2002). On the one hand, the introduction of GMHT year, each field is sited on a different farm. The two levels crops might exacerbate this situation by allowing of the treatment factor are GMHT crop management greater use of herbicide in farmland. This would result and conventional crop management. There are two in fewer plants for insects to live on, and consequently experimental units per block, comprising two halves of fewer insect prey for farmland birds. Alternatively, the the same field. The GMHT crop is allocated randomly use of GMHT crops may allow more precise weed con- to one half-field and the conventional crop to the other. trol, allowing plants to remain longer in the crop. GMHT herbicide management might thereby increase   the abundance and diversity of farmland wildlife compared with herbicide use in equivalent conventional The form of the FSE null hypothesis dictated that treat- crops. To distinguish between these alternatives the ments be chosen deliberately to represent a composite Department for the Environment, Food and Rural of agronomic effects, not a single ecological process. Affairs (DEFRA) and the Scottish Executive have Any feature of the crop itself, such as a varietal trait, funded a 5-year study termed the Farm-Scale Evaluations and any concomitant agronomic practice linked to the (FSE) to provide a thorough understanding of the envi- crop concerned, such as recommended herbicide ronmental effects of growing GMHT crops (Firbank usage, would contribute towards the potential treat- et al. 1999). This is being conducted by a consortium ment effect being measured. Such practices, tied to the of public sector research institutes (Firbank et al. 2003). crop, had therefore to be allocated to units as part of It began in spring 1999 with a pilot year to develop pro- the identical process whereby the treatments were ran- tocols; the evaluations proper began in spring 2000. domized. Composite null hypotheses are often used in Three spring-sown crops, spring oilseed rape, fodder initial studies, to demonstrate the existence and esti- maize and beet (sugar and fodder), were studied in each mate the magnitude of effects and thereby to screen out of 3 years, 2000, 2001 and 2002. Also, one autumn-sown those unworthy of further interest. In such experi- crop, winter oilseed rape, was sown in those years and ments, the most important property is of realism and the third year’s data for this crop will be collected in applicability, so that the results relate unequivocally to 2003. The taxa studied are plants and invertebrates. the system that is studied. The FSE was designed as a For each crop the FSE aims to test a specific null large-scale experiment of this kind. hypothesis: that there is no difference between the management of GMHT varieties and that of comparable - . -: conventional varieties, in their effect on the abundance     and diversity of arable plants and invertebrates. The alternative hypothesis is that there is a difference in An issue for discussion before the design was finalized abundance or diversity, in either possible direction; all concerned the size and location of experimental units. tests are therefore two-tailed. Specifically, should farms act as blocks and units be Effects are likely to be indirect resulting from crop whole-fields, paired within the farm to be as alike in management, rather than from the direct effect of the biodiversity as possible? Alternatively, should a single use of GM plant breeding technology. Indeed, had her- field be divided into two halves, again as alike as pos- bicide resistance been introduced to the experimental sible, defining a unit as a half-field? The arguments in crops by traditional breeding, the design of the study favour of the alternative approaches involved scientific, would have been the same. Farmers grow and manage statistical and practical issues. © 2003 British both GM and conventional crops as closely as possible A strong argument for the half-field design was the Ecological Society, Journal of Applied to commercial practice. The FSE is one of the most potential for reduction in variability. The two halves of Ecology, 40, controversial ecological experiments proposed in Brit- a field are much more likely to be similar, in previous 17–31 ain and perhaps the most extensive ever attempted; 272 management, soil type and surrounding habitat, than

19 two different fields. Residual variation is reduced by to minimize problems. Any chosen design would have Biometrical issues choosing blocks such that experimental units within to attempt to match field-margin biodiversity between of GMHT FSE them are matched, as far as practicable, for the meas- experimental units. Such margins are important habitat ured variable (Perry 1997). Under this argument, halv- in arable ecosystems as reservoirs for plants and over- ing fields should enhance the statistical power to detect wintering sites for insects, cover and food for birds, and differences between treatments, and increase the preci- may affect invertebrate distributions (Lewis 1967). sion with which they are estimated. The FSE aims to compare GMHT and conventional However, ecological relationships measured at one varieties of each of the four crops grown in realistic spatial scale may not have the same parameters or per- commercial conditions, which might favour the use of tain at all at other scales (Heads & Lawton 1983; whole-fields. Against this was the practical issue that in Norowi et al. 2000). Caution is required in extrapolat- the pilot year there was a lack of candidate fields, vital ing the results of a study on half-fields to a larger to choose pairs sufficiently well-matched for previous whole-field scale. Duffield & Aebischer (1994), Perry management and cropping history; this strongly (1997) and Kennedy et al. (2001) have shown how the favoured the use of half-fields. Also, half-fields reduce use of relatively small plots close to one another has greatly the sampling effort, as recorders travel less to affected the interpretation of experiments for relatively collect data. Accuracy might be improved if there is less mobile species such as carabid beetles; this argument time pressure; experience during the pilot year revealed could favour the use of paired whole-fields. Indeed, birds this as an important consideration at particular times and small mammals were excluded from comparison of the year when sampling overlapped between taxa. within the main FSE precisely because their territories Unfortunately, very few data exist on the relative and foraging areas often extend beyond half- or whole- variability between whole-fields within farms and that fields (Firbank et al. 2003). More generally, tritrophic between half-fields within whole-fields. Surveys have interactions between the chemical ecology of plants, her- been used to assess the environmental effects of intensive bivores and their natural enemies are subtle (Vet 1999) and agriculture within the UK for decades (Potts & Vickerman Schuler et al. (1999) highlighted many potential indirect 1974) but designed experiments are relatively recent effects of GM plants on arthropod natural enemies. and lack adequate replication of realistic-sized units Movement of individuals between the two halves of (Sotherton, Jepson & Pullen 1988; Aebischer 1990; Perry the same field might bias the estimated difference 1997; Moller & Raffaelli 1998; Raffaelli & Moller 2000). between treatments, especially if movement was related Lennon (1998) listed nine recent European projects to the effect of crop management. For example, in- on integrated pest management and noted that each creased mortality on one half of the field could be suffered difficulties with inference that resulted from compensated by density-dependent immigration from either inadequate replication or complications due to the other half. An individual carabid may easily travel crop rotations. Unfortunately, the crops studied in the the order of 300 m, the breadth of a square 10-ha field, well-designed MAFF LINK Integrated Farming Systems in two nights (Kennedy 1994). Duffield & Aebischer Study (Ogilvy et al. 1995) were largely different to those of (1994) noted that the recovery from pesticide applica- the FSE. However, some data from the Game Conserv- tion of invertebrate populations would proceed at a ancy and Allerton Research and Educational Trusts slower rate when entire fields were treated, compared (Boatman & Brockless 1998), from up to five winter oilseed with within-field plots of an identical size. Despite rape fields on the demonstration farm at Loddington, limited replication in the largest of their plots, they Leicestershire, UK, provided information on compon- suggested that small-scale within-field trials to evaluate ents of variation (Perry 1989) within eight abundant pesticides would in many cases fail to predict accurately suction-sampled invertebrate groups. Some fields were the impact of commercial pesticide management. halved, yielding information from 1994 to 1996 on Despite these caveats, useful information may still be between- and within-field variation, that could be used obtained from half-fields for highly mobile species, to compare the likely efficiency of half-field and paired- such as bees and butterflies, as long as direct inferences field designs. The variability of paired-fields was often concerning abundance are not made from counts. similar to that for half-fields, but sometimes, especially Instead, treatment differences relate to foraging pref- during 1995, was much greater (Fig. 1). It was not pos- erences towards flowering plants. These problems of sible, due to constraints of proper randomization and interpreting data concerning bees, butterflies and, to a insufficient replication, to use data from the FSE pilot lesser extent, some carabids must be seen in the context year (1999) to inform the choice of design, although an of the ecology of the taxa studied, relative to the treat- informal inspection suggested that half-fields were ments imposed. Direct effects of herbicide manage- inherently less variable than paired-fields. ment regimes are most likely to impinge on vegetation; The final choice of a half-field design was based on effects on invertebrates will probably be indirect. the availability of fields, the associated difficulty of © 2003 British Care must be taken to avoid interference between ex- obtaining suitably matched paired fields, the probable Ecological Society, Journal of Applied perimental units that are close together, for example from major effect of herbicide being on weeds rather than Ecology, 40, spray drift. Here, the separation distances, of 50 m for rape invertebrates, the need to reduce variability and effi- 17–31 and maize and 6 m for beet, between half-field units help ciency gains in sampling effort. The choice was made

20 of geographical spread, intensiveness and biodiversity, J. N. Perry et al. and to identify underrepresented strata. The approach in the FSE was not to sample farms in proportion to their frequency of occurrence according to some factor. For example, low-intensity farms are relatively rare but they may contribute proportionately more to biodiversity than intensively managed farms (Watkinson et al. 2000). The consortium sought to include a disproportionately large sample of such low- intensity farms. Analyses will seek to identify a possible interaction between the treatment effect and intensity, for which there are ample degrees of freedom available. Note that the randomization to half-fields within Fig. 1. Comparison of estimated coefficients of variation each field is distinct from the ability to scale-up from (CV) between half- and paired-fields from 1994 to 1996 from the experiment to some wider population, which requires Loddington Farm. The data from the Allerton Project that the experimental units within fields, and the fields (Boatman & Brockless 1998), run by the Game Conservancy themselves, must be representative. The larger the pool Trust for the Allerton Research and Educational Trust, were of farms, the more likely it was that a suitable set of supplied by Dr Nicholas Aebischer (Game Conservancy Trust). Symbols represent annual values for the eight most farms could be selected. However, there is no require- abundant invertebrate groups in suction samples: Collembola ment per se for such selection, either to ensure validity (C), aphids (A), Homoptera (H), Thysanoptera (T), para- of the statistical test or for the ability to scale up. sitoids (P), staphylinid larvae (S), Coleoptera adults (C) and Rather than the statistical tests of the null hypo- Coleoptera larvae (L). thesis, other approaches are to extrapolate the results of the FSE through explanatory, mechanistic modelling (Firbank & Forcella 2000; Watkinson et al. 2000) or with the proviso that half-fields should fall within the multivariate community-based analysis; such work is range of field sizes used commonly for each crop, and not considered here. should not compromise realistic growing conditions.           :      Randomization of allocation of the GMHT and con- An important requirement of the FSE is that its results ventional varieties to the two halves of the field safe- should apply to the British agricultural ecosystem and guarded against selection bias, for example GMHT landscape as a whole. This raises the question of the crops being applied to the weedier half of the field. It representativeness of the farms included and the issue also provided statistical validity for the test of the null of farm and field selection. For example, it would be hypothesis, and for the estimates of the precision of the unsatisfactory if there were no fields within the FSE magnitude of any differences. and it allowed differences growing spring oilseed rape in Scotland, where a large detected to be ascribed causally as treatment effects. acreage of the crop is grown. For the pilot study, fields The randomization protocol for the trial required a came from a limited self-selected set of growers who structured dialogue between the recorder from the con- were willing to grow GMHT crops. Within the FSE sortium and the grower, so that the choice of boundary proper, the issue of representativeness was addressed line to halve the field for sowing was made on scientific by attempting to select fields that encompassed the full grounds not agronomic convenience. The optimum range of variation, in various variables, likely to be choice of boundary should result in two half-field units found in commercial practice. The current status as alike as possible over the range of factors that con- within Britain for each crop was summarized with tribute to the variability of wildlife within the field. The regard to its geographical distribution, usual agron- protocol also guarded against any preference a grower omy, soil types and field sizes. This profile was then had for what side of the field should receive the GMHT compared with more detailed information on specific treatment. Thus, treatment allocation was predeter- candidate farms and fields, obtained from a question- mined by project statisticians who assigned one treat- naire issued by the consortium to each grower who ment at random to the label ‘A’ and the other to label expressed interest in taking part (Firbank et al. 2003). ‘B’. This allocation was provided to the recorder (but Estimates were made of the intensiveness of the unknown to him or her) in a sealed envelope. After the grower’s inputs and the extent to which the farmers boundary line was agreed between recorder and grower, © 2003 British managed their land in ways that might favour biodiver- the half-field unit towards the north (for an east–west Ecological Society, Journal of Applied sity. Potential growers required early notification of boundary line) or towards the west (for a north–south Ecology, 40, whether their farm was selected, so a sequential approach boundary line) was labelled as ‘A’, and the other as ‘B’, 17–31 was used to monitor the structure of the sample in terms and drawn on a rough map. The envelope was then

21 opened and the treatments noted on this map. With this usually identified to species or family level, analyses Biometrical issues auditable procedure none of the recorder, statistician focused initially on totals over all species, functional of GMHT FSE or grower could influence the randomization. groups and over large groupings such as total mono- cotyledons. Several samples may have been taken for each taxonomic group through the year; these may      have been aggregated to give a single annual total or  analysed separately. Whilst this present study is not In some respects, the FSE experimental design has intended to provide an exhaustive list of potential ana- much in common with on-farm trials carried out by lyses, it is likely that measures of species richness and farmers, on their own land, in studies on third-world diversity will also be compared between treatments. agriculture (Buzzard 2000). The control crop variety was selected by the farmer according to local condi- Statistical power tions, and varied between farms. Both GMHT and conventional systems were managed by growers as      closely as possible according to their current commer-  cial practice, although within this constraint management practices were kept as similar as possible. Any The choice of the number of fields in the FSE was con- pesticide seed treatment was the same on both treat- troversial because it represented the first occasion on ments at a farm. Where non-herbicide treatments were which GMHT crops were sown on this large scale in imposed on both GMHT and the conventional vari- Europe and it was deemed inappropriate to grow a eties, they were applied at the same time unless there was large area. Also, the cost of any publicly funded experi- good agronomic reason, for example if there were more ment must be constrained within limits and the more pests on one half-field than the other. Growers took fields sown and sampled, the greater the cost. Against usual decisions for weed control on the conventional this, sufficient replication was required to detect effects. variety; this might or might not involve the use of con- The statistical power that comes from proper control of sultant agronomists. However, usual practice remains variation and adequate replication (Perry 1986) is difficult to define for GMHT varieties, because none important in regulatory trials, which seek to study has yet been grown commercially within Britain. Pro- whether there are any deleterious environmental effects cedures that ensured that the treatment applied within of new products (Anonymous & Perry 1999). A statis- each management regime was applicable are outlined tical power analysis, which quantifies the likely effi- by Firbank et al. (2003). Such considerations are vital ciency of an experiment, was essential for the FSE. to enable valid inference, and are equally important The statistical power of a significance test is the as biometrical issues of design and analysis; no treat- probability of rejecting the null hypothesis when some ment randomization can allow for biases arising from given alternative hypothesis is true. The power meas- inappropriate management of the GMHT variety. ures the chance of detecting an effect of a known mag- Note that it was possible for there to be no herbicide nitude using the specified experimental design, and applications to either half-field unit if, for example, varies according to the magnitude of the effect speci- there were no weeds to treat. fied. It is often difficult for biologists to specify this Some agronomic practices, such as the increased quantitatively but without an answer to the question use of direct drilling or changes to normal rotations, ‘Precisely what degree of treatment effect do you con- might become associated with GMHT technology if sider important?’ any power analysis is uninformative. it were commercialized. The FSE cannot, at this early Power depends also on sample size, the degree of stage in the use of GMHT, evaluate efficiently such random variation between experimental units and the events within an experimental framework of imposed chosen significance level of the test (Sokal & Rohlf treatments. However, the FSE will provide data to para- 1981). Power is a continuum that varies non-linearly meterize predictive models in which such scenarios and gradually with sample size. There is no threshold may be studied. level of replication below which an experiment is too poorly resourced to be worth conducting and above which it is satisfactory.    Power was estimated for the FSE over scenarios that Details of the range of farmland wildlife taxa studied encompassed a range of treatment differences, num- in the FSE are given by Firbank et al. (2003). Both bers of fields and degrees of random variability. For density and biomass of plants were recorded, as well as data that approximately follow a normal distribution, the seed bank and seed return. Invertebrates counted the power of standard tests, such as Student’s t-test, can included carabid beetles in pitfall traps; butterflies and be calculated routinely. However, more complex calcu- © 2003 British bees sighted along transects; other arthropods, such as lations are required when, as here, ecological count Ecological Society, Journal of Applied Collembola and Heteroptera, in Vortis suction samples; data are collected that have an asymmetric frequency Ecology, 40, crop pests counted on plants; and gastropods in refuge distribution and vary in relation to mean abundance. 17–31 traps and verge searches. Although samples were Power was estimated both for a standard, simple,

22 model based on a logarithmic transformation of counts, ﬁeld effect Fj and a treatment effect ti. These combined

J. N. Perry et al. and also for an extended model developed to be more to give the expected value of the response variable, cij, realistic for the form of ecological count data collected, on natural logarithmic scales: with a large proportion of zero counts possible for γ θ some species. loge E[cij] = + Fi + ti = ij, say eqn 2

Treatment and field effects were assumed multiplica-  :  , -  tive on the natural count scale. Suppose there were j = 1, … , n fields, with two experi- The random component of the extended model mental half-field units per field. The treatment factor, reflected variability from unit to unit, i.e. between half- GMHT vs. conventional, had two levels, denoted i = fields and within fields. A negative binomial distribu-

1,2. In the simple model, the observed response variate, tion was assumed for the counts, cij, but the shape

the count cij, was transformed to lij = ln(cij + 1). Then, a parameter of the distribution, kij, for each treatment on standard randomized block  was done, with each ﬁeld was constrained (Perry et al. 1998) to follow

fields as blocks, on the transformed values, lij; the treat- equation 1. The mean of this negative binomial distri- ment effect was assessed with a t-test. This approach bution, for each treatment on each field, was denoted φ φ θ assumed a normal distribution for lij and therefore an ij, where ij = exp( ij). For each particular submodel φ approximately log-normal distribution for cij; this was the value of kij( i) for each count was determined from: termed the log-normal model (Table 1). φ φ αφ β−1 − φ φ ∞ kij( ij) = ij/( ij 1), for kij( ij) > 0, and kij( ij) = , otherwise eqn 3  :  ,    Three submodels were studied, with a different single The extended model was designed to allow for value of β for each submodel: β = 1, 1·5 and 2, respectively (Table 1). These values of β were chosen to many small or zero values of cij and for the observed dependence of variance, V, upon mean abundance, µ, incorporate relationships typical of those observed for ecological count data, often expressed through a for ecological count data. When β = 1, variance was pro- power-law (Taylor 1961), with parameters α and β: portional to the mean. For efficient analysis, a generalized linear model (GLM) with a logarithmic link V = αµβ eqn 1 would be assumed typically, with Poisson errors and estimated scale parameter (McCullagh & Nelder Model I above explicitly assumed that variance 1989); this was termed the log-linear model. When is homogeneous after transformation, and therefore β = 2, the coefficient of variation was theoretically con- implicitly assumed that the exponent, β, was close to stant, and the simple model with a logarithmic trans- 2·0. This followed from the result based on first-order formation provided an efficient analysis. Values close Taylor series approximation (Cochran 1938), that the to β = 1·5 lacked such mathematically tractable inter- variance on a natural log scale is approximately equal pretation, but were typical of exponents encountered to the variance on the untransformed scale divided by for many species in field data (Taylor, Woiwod & Perry

the square of the untransformed mean (i.e. var[ln(cij)] 1978). ≈ V/µ2). To generate the counts, given speciﬁed values of µ, the

The systematic effects in the extended model explic- ﬁeld effect Fi and the treatment effect ti (see below), φ itly allowed for variability in the blocking structure. the value of ij was found from equations 1 and 2. Then, The effects represented by the parameters of the given values of α and β for a particular submodel, the µ γ extended model were: an overall mean, = e , say, a value of kij was found from equation 3. This equation

Table 1. Summary of statistical models used in the study. Models differ in their assumed variance-mean relationship, measured through the power-law function, V ∝ µβ. Parametric methods use F-tests.  denotes analysis of variance. GLM denotes generalized linear model; for β = 1 with Poisson errors and log-link, for β = 1.5 with user-deﬁned error and log-link. Entry in column detailing the non-parametric randomization test is the test-statistic used: d is mean difference in logarithmically-

transformed count; r is logarithmically-transformed ratio of arithmetic means of counts; dw is a weighted version of d. For further details see text

Data analysis Power estimate

Non-parametric Non-parametric β Parametric randomization test Parametric randomization test © 2003 British Ecological Society, 2 Log-normal,  d Log-normal d Journal of Applied 1·5 GLM dw – dw Ecology, 40, 1 Log-linear, GLM r – r 17–31 23 treated the rare case of a negative simulated value of k Table 2. Summary of runs used in statistical power simu- Biometrical issues as indicating effectively Poisson variation, for which lations. In each case the variance, V, of the count is related µ αµβ of GMHT FSE case the value of k was set to infinity and the Poisson to the mean, , through V = . Means and coefficients of variation (CV) are expected values, which take no account of distribution replaced the negative binomial for simu- any additional variability induced by field and treatment lation. A negative binomial variate with mean µ and effects variance µ + µ2/k was simulated using the following two-step procedure (Morgan 1984). First, a random Reference αβ variate, say g, was sampled from a gamma distribution number of run µ CV% with mean µ and variance µ2/k; secondly, a count 1 1 1·0 1 100 from a Poisson distribution with mean g was sampled. 2 1 1·5 1 100 Random gamma and Poisson variates were generated 3 1 2·0 1 100 using subroutines from the NAG library (Numerical 4 5 1·0 5 100 Algorithms Group 1997). 5 1 2·0 10 100 6 7·07 1·5 50 100 7 1·79 1·5 5 80      8 6·4 1·0 10 80 9 0·64 2·0 50 80    10 0·25 2·0 5 50 The model was used to generate sets of count data for 11 0·79 1·5 10 50 12 12·5 1·0 50 50 specified combinations of parameter values and different magnitudes of the treatment effects (Table 2). Mean abundance, µ, was studied for four values: 1, 5, ± 10, 50; the first three values were chosen because, for the other. For example, values of ti = 0·203 were used these simulations, attention was focused largely on the to represent a multiplicative difference of R = 1·5

case of smaller counts. The ﬁeld effect, Fi, modelled the between the treatments. Then, for a mean count on the effect of the variability between ﬁelds and contributed logarithmic scale of 0·0, the expected value, back- to the variability of counts, at larger than unit scales, transformed onto the natural count scale under treat- through equation 2. Field effects were simulated as ment 1, would be 1·225. For treatment 2 the value

ﬁxed effects, such that Fi = –M + ( j – 1)q, where j = 1, would be 0·816. This yielded a multiplicative difference

… , n and q = 2M/(n – 1), with M = loge 10 = 2·303, and of 1·5-fold between the two treatments; it may be n specified the number of fields. For example, for n = 20, viewed either as a 50% increase or as a 33·33% decrease − − this resulted in the series: Fi = 2·303, 2·182, … , of one relative to the other. −0·364, −0·121, 0·121, 0·364, … , 2·182, 2·303. This scheme ensured two orders of magnitude variation in     the blocking factor representing the field effect, so the expected values of the response variable for the two The power of three statistics was computed, each based extreme fields varied by 100-fold. The variation in on Monte Carlo paired randomization tests (Manly mean abundance, above, when combined with the field 1994), and applied to each set of simulated data effect, gave expected ranges of simulated abundance of (Table 1). Briefly, this entailed recording the value of 0·10–10, 0·50–50, 1–100 and 5–500; and means of 2, the ‘observed’ statistic computed for each generated set 12, 23 and 115. of 2n counts, and comparing this value against 199 The degree of random variation between the experi- other ‘randomized’ values of the statistic, recomputed mental half-field units was varied through the parameter after random relabelling of the treatment codes for α. Values of α were chosen to achieve coefficients of each of the n pairs of counts. If the observed statistic variation (CV) of 50%, 80% and 100%. The coefficient exceeded the upper 5th centile of the ranked random- of variation, √V/µ, here equated to α1/2µ(β/2−1). It provided ized values then the null hypothesis was rejected for a useful way of specifying baseline variability that that generated set. Two-tailed tests were performed by permitted direct comparison with characteristic values using the absolute value of the test-statistic. The pro- for a particular taxon, perhaps calculated from previous cess was then repeated for 500 sets and the power experiments. However, the theoretical values listed above, estimated as the proportion of rejections. of 50%, 80% and 100%, were different from those values The three statistics studied reflected the three forms actually realized by the simulations, which were subject of variance–mean relationships, characterized by the to random variation. exponent β. The first, d, closely related to the log-

Three different values of treatment effect, ti, were normal model, was the simple mean of the differences speciﬁed, representing multiplicative differences of between the two treatments on the logarithmic scale, × 1·3-fold, × 1·5-fold and × 2-fold. If treatment effects d = Σ [l – l ]/n. This should have relatively high power © 2003 British i 1i 2i were denoted as t = 0·5 ln(R) and t = −0·5 ln(R) on the when variance is proportional to the square of the Ecological Society, 1 2 β Journal of Applied natural scale, then this corresponded to a multiplicative mean ( = 2). The second statistic, r, closely related to Ecology, 40, difference of R, expressed either as a (R – 1)% increase the log-linear or Poisson regression model for count − 17–31 or a (1 − R 1)% decrease, of one treatment relative to data (McCullagh & Nelder 1989), was the logarithm of 24 Table 3. Estimated type I errors (5% level) for Monte Carlo Table 4. Statistical power for detecting a × 1·3-fold difference J. N. Perry et al. paired randomization tests using n = 20 pairs. Estimates based for simulated count data. Details of run parameters and the on 1000 sets of simulated data (SE = 0·7%). The three test three test-statistics are in Table 2. Estimates based on 500 sets

statistics for testing treatment difference were d, r and dw, of simulated data. The power for the log-normal model was where d denotes the mean difference in logarithmically based on a paired t-test. Values of power exceeding 80% transformed count; r denotes logarithmically transformed shown in bold. See text for further details

ratio of arithmetic mean counts; dw is a weighted version of d. For further details see text Reference Number of pairs (n) number Test Type I error (%) of run CV% statistic 20 30 40 60 90 Reference test statistic number 1 100 d 16 19 27 41 55 βµ of run CV% drdw r 15 22 35 48 69

dw 19 23 35 49 67 1 1·0 1 100 5·3 3·5 5·6 2 100 d 10 15 22 31 43 2 1·5 1 100 4·6 3·9 5·5 r 10 12 19 27 40

3 2·0 1 100 4·0 3·6 4·9 dw 10 16 22 31 46 4 1·0 5 100 5·5 4·4 4·8 3 100 d 6 14131622 5 2·0 10 100 5·4 5·3 5·4 r 5 10111516

6 1·5 50 100 4·9 5·6 4·7 dw 6 12131622 7 1·5 5 80 5·6 4·4 5·1 4 100 d 15 19 25 39 53 8 1·0 10 80 4·3 4·2 3·6 r 20 25 35 53 70

9 2·0 50 80 4·8 4·4 4·9 dw 20 27 33 56 70 10 2·0 5 50 6·3 4·9 5·5 5 100 d 11 11 11 19 26 11 1·5 10 50 6·5 4·8 5·6 r 10 11 10 16 21

12 1·0 50 50 5·1 4·2 4·7 dw 11 12 12 21 29 Mean 5·2 4·4 5·0 6 100 d 16 17 19 28 40 r 13 19 17 30 42

dw 17 21 21 36 51 Log-normal 100 16 22 28 40 55 the ratio of the overall arithmetic means of the two 780d 14 20 28 38 47 Σ Σ treatments, r = ln [ ic1j/ ic2j]. This should have rela- r 14 20 26 36 44 tively high power when variance is proportional to dw 15 24 32 41 52 880d 20 25 34 46 60 the mean (β = 1). The third statistic, d , derived by P.R., w r 30 39 53 74 88 was introduced to try to accommodate the intermedi- dw 29 37 50 69 86 ate case (β = 1·5). It was a weighted version of d, with 980d 12 20 27 31 46 weights based on the approximate variance, assuming r 9 13171928 β = 1·5, of the difference in logarithmically trans- dw 11 17 22 28 39 Log-normal 80 20 29 37 52 70 formed counts, i.e. d = Σ w [l − l ]/(Σ w ), where w j j 1j 2j j j 10 50 d 20 30 39 57 75 −0·5 −0·5 −1 wj = [(1 + c1j) + (1 + c2j) ] . r 19 25 30 43 61

Power was estimated for treatment differences of dw 22 32 39 58 77 magnitude R = 1·3, 1·5 and 2, for 12 different com- 11 50 d 32 47 57 76 88 binations of the parameter values α, β and µ. It was r 31 50 59 80 92 d 36 58 68 88 94 decided to study power assuming equal sample sizes of w 12 50 d 24 38 46 62 76 20 and 30 ﬁelds per year. Over a 2-year experiment r 60 80 91 98 100

these would give sample sizes of n = 40 and 60, and over dw 53 72 87 97 100 3 years n = 60 and 90. Hence, the range of sample sizes Log-normal 50 39 55 68 85 96

used, n = 20, 30, 40, 60 and 90, covered all combinations of 20 and 30 ﬁelds per year, for periods from 1 to 3 years. This power study simulated a total of more calculated using the statistical package Minitab Release than 8·5 million negative binomial random variables, 13 (Moultine & Bluman 2001). For a treatment effect in a total of 90 000 sets of data. In addition, the type of R = 1·5, n = 60 pairs and a CV = 50%, the power I error of each test was checked using 1000 sets of exceeded 90% in all but one case. When the treatment simulated data. effect represented a doubling or halving of density and R = 2, for n = 60, for values of CV = 50%, 80% and 100% and for values of µ ≥ 5, the power exceeded      85% in all but one case. Table 3 shows that in each case the type I error was As expected, the power of the r-statistic was higher close to its nominal value of 5%. Tables 4–6 show the than that of the d-statistic when the variance was estimated power for values of R = 1·3, 1·5 and 2, respec- proportional to the mean (β = 1) and vice-versa when © 2003 British tively, and, for comparison, the corresponding power of the variance was proportional to the square of the Ecological Society, β β Journal of Applied a paired t-test for the situation when lij was assumed to mean ( = 2). The dw-statistic performed best for = 1·5, Ecology, 40, have a normal distribution, i.e. the simple log-normal but also appeared agreeably robust, maintaining com- 17–31 model referred to earlier. The power of the t-test was parable power to d for β = 2 and to r for β = 1. 25 Table 5. Power for detecting a × 1·5-fold difference using Table 6. Power for detecting a × 2-fold difference using Biometrical issues simulated count data (see Table 4) simulated count data (see Table 4) of GMHT FSE Reference Number of pairs (n) Reference Number of pairs (n) number Test number Test of run CV% statistic 20 30 40 60 90 of run CV% statistic 20 30 40 60 90

1 100 d 32 42 53 72 91 1 100 d 67 87 94 100 100 r 37 54 65 85 97 r 75 94 99 100 100

dw 40 53 65 83 97 dw 77 94 98 100 100 2 100 d 25 31 42 55 72 2 100 d 51 69 82 96 100 r 22 28 41 53 71 r 49 66 83 97 99

dw 26 34 46 59 77 dw 55 74 86 98 100 3 100 d 13 22 22 34 47 3 100 d 26 44 51 74 89 r 12 18 20 29 44 r 21 38 46 65 86

dw 14 23 24 35 49 dw 27 45 54 75 90 4 100 d 29 46 50 75 86 4 100 d 69 86 95 98 100 r 35 60 69 88 97 r 77 96 99 100 100

dw 37 61 70 89 97 dw 82 97 99 100 100 5 100 d 16 21 27 39 58 5 100 d 37 53 64 85 95 r 12 16 23 30 42 r 31 42 49 71 84

dw 16 22 28 38 55 dw 37 52 65 85 95 6 100 d 21 38 43 56 77 6 100 d 52 69 84 95 100 r 25 35 43 58 81 r 54 73 85 95 100

dw 30 42 50 66 88 dw 60 82 90 99 100 Log-normal 100 31 45 56 75 90 Log-normal 100 70 88 95 99 100 780d 28 43 51 68 84 780d 67 83 92 99 100 r 28 40 51 70 86 r 63 82 91 99 100

dw 33 46 59 77 91 dw 71 90 96 99 100 880d 38 53 65 77 92 880d 77 93 97 100 100 r 55 75 90 97 100 r 94 99 100 100 100

dw 54 74 88 96 100 dw 92 99 100 100 100 980d 25 39 46 63 82 980d 57 76 88 97 100 r 17 26 31 44 59 r 40 58 69 87 100

dw 24 35 38 59 78 dw 51 72 85 97 100 Log-normal 80 41 58 71 87 97 Log-normal 80 84 96 99 100 100 10 50 d 44 63 77 91 98 10 50 d 91 96 99 100 100 r 37 50 65 80 93 r 81 91 97 100 100

dw 47 64 77 93 99 dw 89 98 100 100 100 11 50 d 57 79 88 98 100 11 50 d 97 100 100 100 100 r 65 83 92 99 100 r 98 100 100 100 100

dw 73 92 97 100 100 dw 100 100 100 100 100 12 50 d 50 71 80 93 99 12 50 d 88 98 100 100 100 r 91 99 100 100 100 r 100 100 100 100 100

dw 88 98 100 100 100 dw 100 100 100 100 100 Log-normal 50 73 90 96 100 100 Log-normal 50 99 100 100 100 100

Because of its complexity, the power estimated from ting field and treatment effects in the  of lij. In the negative binomial model varied more than that addition, the solid line in Fig. 2 shows the power for the based on the log-normal model. Some of this variation simple log-normal model, which, because it was plot- could be accounted for by deviations of the actual real- ted vs. ∆, was independent of the quantity R/σ and so ized CVs from the theoretical target values given in may be utilized generally to estimate the power when- Tables 4–6. ever the magnitude of effect is expressed as a multiple To summarize this complex situation power was number of standard errors. The log-normal model examined in relation to a standard statistical ‘non- therefore provided a useful baseline against which to centrality parameter’ (Pearson & Hartley 1976; section assess the effect of assuming negative binomial counts, 14.5, tables 27 and 30), when a consistent pattern i.e. the effect of the discrete, variable and sometimes emerged (Fig. 2). The non-centrality parameter used, small counts encountered commonly in ecology. Inter- ∆, was the true difference divided by the standard error pretation was aided by noting that for the log-normal of the estimated difference, d; it therefore had much in model the percentage coefficient of variation was equal common with the simple t-statistic, well-known in to 100√[exp(σ2) − 1] (Hastings & Peacock 1975), so the © 2003 British ecology. Specifically, ∆ was calculated as log R(2σ 2/n) −1/2, four solid circles in Fig. 2 corresponded to CV of 50%, Ecological Society, e σ 2 Journal of Applied where was the variance of lij. For the negative binomial 80%, 100% and 156%. Clearly, most of the simulated 2 Ecology, 40, model, σ was estimated from the simulated data by powers fell slightly below the solid line, so the addi- 17–31 calculating the average residual mean square after fit- tional variability resulted, as expected, in a small 26 their contribution will be mainly analysed through J. N. Perry et al. indices of diversity.

     The consortium sowed 272 fields over four crops, an average of n = 68 per crop. The power analysis indicated that replication of 20 fields per crop per year over 3 years (n = 60) should have provided adequate power (> 80%) to detect multiplicative differences of R = 1·5- fold, so long as CV did not exceed 50% and mean abun- Fig. 2. Statistical power (%) for detecting an R = ×1·5-fold dance exceeded 5·0. There was no need for strictly difference with a scheme of 20 fields over 3 years (i.e. n = 60 equal replication of 20 fields per crop per year, as it is pairs), based on simulated data from a negative binomial the total replication that is important. Estimates of CV model with theoretical, target CV of 50% (triangles), 80% from sets of data made prior to the start of the FSE (open circles) and 100% (squares), using the d-statistic (see were, in most cases, close to the figure of 50% used Table 5), for values of µ ≥ 5. Solid lines show power for log- normal model; solid circles on these lines relate to CV, from in the power analysis. Mean CV in half-fields from left to right, of 156%, 100%, 80% and 50%. Non-centrality the Allerton project (Boatman & Brockless 1998) for ∆ ∆ σ 2 −1/2 σ 2 parameter, , calculated as = loge R(2 /n) , where is different taxonomic groups were (Fig. 1): Collem- the variance of the logarithmically transformed count (see bola (38%), aphids (45%), Homoptera adults (46%), text). Thysanoptera (55%), Parasitica (59%), staphylinid larvae (47%), Coleoptera (65%) and Coleoptera larvae (128%). Frampton (1999) reported other suction sample reduction in power. This approach also provided a data in winter wheat for which the total Collembola direct method of linking the theoretical power calcula- count had a CV of 51% for variation between different tions to the analysis of actual data, via an estimate of σ 2 fields.

from a simple  of lij. For any desired value of R and The weighted test-statistic dw, may provide a promising projected value of n, we may use this future estimate of basis for future analyses of data with a variance-mean σ 2 to derive a value of ∆; an approximate prediction of exponent, β, between 1 and 2. The results reinforced power may then be made for the log-normal model, the importance of reducing the CV between experi- perhaps with some slight downward adjustment for mental units, and of the limitations in analyses of variates the effect of extra variability. For the log-normal model with small means. The need to reduce the CV supported the power for detecting a difference of three times the the choice of half-field over paired-fields in the FSE standard error was about 85%, and about 98% for a dif- design. ference of four standard errors. There are some cases, mainly for small µ and large Analysis CV, where the recommended minimum replication level of 20 fields per crop per year over 3 years (n = 60)    had low power for R = 1·5. Tables 4–6 suggested, for µ ≥ 5, that this reduction in power could sometimes be In the FSE, the crops are considered and will be ana- offset by using the r-statistic and the log-linear model lysed, at least initially, separately. Here we illustrate a for analysis. However, if the mean count was as small as mode of analysis that follows the simple and extended µ = 1, even in the ideal case of a completely random models and their associated test-statistics. The ration- distribution of counts power would always be limited. ale is to provide a range of analyses that (i) address the For example, for the log-linear model the standard null hypothesis and allow estimation of treatment error of the estimate of log R, s, was approximated by effects; (ii) are appropriate to the data; (iii) match the ≈ Σ −1 Σ −1 1/2 s {A[( jc1j ) + ( jc2j) ]} , where A was an esti- simplicity of the design; (iv) provide results that are mate of the overdispersion. Consider a scheme with transparent and easily understood; (v) allow for hetero- µ µ n = 60 pairs; mean counts per half-field of 1 = 1 and 2 geneity and other deviations from model assumptions; = 0·7 for the two treatments, so ln(R) = 0·36; and with and (vi) have the flexibility to allow for the inclusion no overdispersion, so A = 1. Then s = 0·20, and the dif- of covariates and multifaceted extensions to the basic ference of 0·36 was a mere 1·8 times the standard error. analysis. Currently, a two-stage process is envisaged. Hence it was not surprising that the power for detecting The first relates to a basic analysis, which conforms an effect would be small. This emphasizes the import- to criteria (i–v) above, and will be almost identical © 2003 British ance of having adopted protocols that yielded suffi- for all variates and crops. Extensions to allow for Ecological Society, Journal of Applied ciently large means per half-field unit, and of focusing criterion (vi) will build on this basic analysis. What Ecology, 40, on the analysis of the abundance of common individual follows is presented to exemplify the sort of analysis 17–31 species or groups. Rare species are important too, but that might be performed on 3 years of data from the 27 FSE. The basic analysis will consist of three com- Table 7. Skeleton analysis of variance for 60 beet fields over 3 Biometrical issues ponents, each with a parametric and non-parametric years, with levels equally apportioned over three blocking of GMHT FSE form (Table 1). In the first of the parametric analyses factors representing intensity (I), years (Y) and crop type (sugar and fodder beet, C), with two, three and two levels, each count is assumed to vary proportionally to the respectively. All the main effects and interactions measured by square of its mean. This leads naturally to a logarithmic these blocking factors are estimated in the fields stratum. The transformation for efficient analysis, and hence to the main treatment factor, comparing GMHT vs. conventional, is log-normal model. An analysis of variance is used to represented by T. The main effect of T, and all two-, three- and provide a test of the null hypothesis, with the fields as four-factor interactions involving T are estimated in the half- field units stratum. All F-tests are based on 48 residual degrees blocks; the magnitude of d is estimated. In the second, of freedom the variability of counts is assumed to vary proportionally to the mean. The log-linear GLM described above Source of variation d.f. SS MS F is used, with an analogous analysis of deviance; the magnitude of r is estimated. Both these models are used Fields stratum ubiquitously in the ecological literature to analyse I1 plant and insect counts; both provide an F-statistic to Y2 test the null hypothesis. An extension of this GLM, to C1 All interactions between 7 the case where the variability of counts is assumed I, Y and C to vary as a power of 1·5 of the mean, is also fitted. Residual 48 RSSs RMSs The models underlying the parametric approach Total 59 might lack sufficient robustness to deviations from the Half-field units stratum assumed distributions. The distributional assumptions Total fields (blocks, 59 from above) implicit in the different variance–mean relationships Main effect of T 1 FT adopted by the models will be tested informally, using × T I interaction 1 FTI × a suite of diagnostic plots of residuals (Carroll & T Y interaction 2 FTY × Ruppert 1988). In addition, formal Monte Carlo random- T C interaction 1 FTC T × I × Y interaction 2 F ization tests will also be done by random relabelling of TIY × × T I C interaction 1 FTIC the treatment codes for each of the observed pairs × × T Y C interaction 2 FTYC of counts. Such randomization tests are useful when × × × T I Y C interaction 2 FTIYC

there are many small and/or zero counts in a data set, Residual 48 RSSu RMSu for which parametric tests might be too liberal. The Total 119

test-statistics are, respectively, d, r and dw (Table 1). The randomization tests use 999 random permutations within each run to estimate P-values. interaction, all F-tests of interest are computed with 48 residual degrees of freedom. There would, of course, be sufficient flexibility to fit other covariates of interest.        Examples of analyses using FSE data from Here an illustration is given of an extension to the basic year 2000 analysis to answer additional questions, subsidiary to the main null hypothesis but adding to the interpret- After the first year of the FSE, data were available to ative power of the FSE. One is the possible interaction estimate CV for a range of taxa from each protocol, between treatment and years. Another is the require- and to reassess the statistical power calculations. Some ment to confirm with experimental data that sugar and example analyses follow. fodder beet can, as claimed, be treated as effectively the same crop for the FSE; this involves testing the inter- -  -- action between treatment and crop type. With so many      two-factor interactions that require testing, it might be   prudent to test some higher-order interactions as well, and the FSE design provides plenty of degrees of free- The first data considered were weed seeds and weed dom. For example, consider the analysis of a variable seedlings on fields where no pre-emergence herbicide for 60 beet fields, with, for convenience, levels equally was used. This ensured that for all these data, at the apportioned over three factors representing intensity, time of sampling, each half-field had exactly the same years and crop type (i.e. sugar and fodder beet), with operations; there was therefore no reason to expect any two, three and two levels, respectively. There would treatment effect. For the seeds, samples were taken therefore be five fields representing each combination from n = 4, and for the seedlings from n = 12 transects © 2003 British of levels of intensity, years and crop type, with the main per half-field (Firbank et al. 2003). This analysis took a Ecological Society, Journal of Applied treatment factor occurring at both levels on each field. components of variance approach (Perry 1989) to Ecology, 40, In a skeleton analysis of variance (Table 7), with all distinguish the component governed by sampling error,

17–31 possible interactions ﬁtted up to the full four-factor arising from variation (Vt) between transects within 28 Table 8. Hierarchical nested analysis of variance of total weed seedling and total weed seed counts for each transect, pooled over

J. N. Perry et al. all crops. Data pooled over all four crops sown in 2000. Counts, c, were transformed to loge(c + 1) prior to analysis. Variance component Vt was estimated directly as the between-transect, within-half-ﬁeld MS. Variance component Vh was estimated from:

Vh = (between-half-field MS – within-half-field MS)/n, with n = 12 for seedlings and n = 4 for seeds. Overall variance of the total √ count per half-field was estimated as Vo = Vh + Vt/n, and approximate CV% as 100 Vo

Source of variation d.f. SS MS F (P) Estimated variance component (% of total)

Weed seedlings Between-ﬁelds, within-crops 17 687·3 40·43 23·0 (< 0·001)

Between-half-ﬁelds, within-ﬁelds 20 35·18 1·76 3·26 (< 0·001) Vh = 0·102 (16)

Between-transects, within-half-ﬁelds 440 235·8 0·536 Vt = 0·536 (84)

Total 447 958·3 Vo = 0·147, approx. CV = 38% Weed seeds Between-ﬁelds, within-crops 68 366·9 5·40 8·10 (< 0·001)

Between-half-ﬁelds, within-ﬁelds 72 47·97 0·666 1·84 (< 0·001) Vh = 0·075 (17)

Between-transects, within-ﬁelds 432 157·6 0·365 Vt = 0·365 (83)

Total 572 572·4 Vo = 0·166, approx. CV = 41%

half-ﬁelds, from the variation between half-ﬁelds

within fields (Vh). We then studied whether within-half- field sampling intensities were sufficiently large to

reduce the overall variance (Vo, where Vo = Vh + Vt /n) to an acceptable level, when expressed as a CV. Data were pooled over all four crops sown in 2000 (Table 8). If, say, only n = 1 transect per ﬁeld been sampled, then the predicted CV would have been approximately 80% and 66% for seedlings and seeds, respectively. This jus- tiﬁed the multiple transects used, which reduced the estimated CV to 38% and 41%, respectively.

     The next analysis was of total weed seed abundance for Fig. 3. Total weed seed abundance from the beet crop seed the 24 fields of the beet crop sown in 2000. Samples bank, for year 2000, plotted for treatment 2 vs. treatment 1, on were taken after halving the field, but before sowing, so a logarithmic (base 10) scale. Equality line shown for no treatment had been applied and no difference was guidance. Horizontal and vertical lines around points show expected between the treatments, here denoted as 1 and approximate standard errors for each estimated total. 2. The geometric mean abundance was 93 seeds, and the estimated CV was 52%. Such abundance and variability would be entirely satisfactory according to the power analysis. A scatterplot (Fig. 3) showed, as Table 9. Analysis of variance and tables of means for weed seed abundance in the seed bank for the 24 beet fields in year expected, no gross differences between the treatments. 2000. The main treatment factor, comparing GMHT vs. The range of values for both treatments exceeded 1·5 conventional, is represented by the factor T. The F-test is orders of magnitude, comparable with the two orders based on 23 residual degrees of freedom. The column headed of magnitude assumed for the power analyses. P gives the F-probability. The SED is the standard error of the The estimate of R, the multiplicative factor by which difference between two means

one treatment was greater than the other was, for the log-normal model, ×1·33 (approximate SE = 0·199; P Source of = 0·068; Table 9). By contrast, the estimate for the log- variation d.f. SS MS F (P) linear model was ×1·25 (approximate SE = 0·183; P = Fields 23 6·093 0·265 5·23 0·139). The difference in these estimates of R emphasized T 1 0·186 0·186 3·67 (0·068) the importance of discrimination between the two models Residual 23 1·165 0·051 with diagnostic residual plots. In fact, the log-linear Total 47 7·443 model demonstrated a clear increase in the variability Tables of means (logarithmic scale, base 10) © 2003 British of the residuals as the fitted values increased (Fig. 4a), Mean Value Replication SED Ecological Society, Grand mean 1·972 24 Journal of Applied indicating the assumed variance–mean relationship was wrong and the need for a more skewed distribution. Treatments 1 2 Ecology, 40, 2·034 1·910 12 0·0650 17–31 By contrast, the diagnostic plots implied a preference 29 provided similar results. For maize, the geometric mean Biometrical issues density over the 13 fields was 122, and the estimated CV of GMHT FSE was 42%. For spring oil seed rape there were two out- lying large values that both exceeded 400, while the geometric mean density over the 14 fields was 89. The estimated CV was 37%. Again, for both maize and spring oil seed rape the diagnostic plots showed that the log-normal model appeared appropriate while the log- linear model displayed variance heterogeneity.

       All analyses considered thus far have used abundance as the response variable, but other responses are possible. The number of species, S, measured in the seed bank for the beet crop data above, provided an alternative quantitative comparison of biodiversity, as long as abundance was similar between treatments, as here. S ranged from six to 30 species over the fields, averaging Fig. 4. (a) Standardized residuals from log-linear model 14·75 for each treatment. The estimated CV was 16%. analysis of total weed seed abundance from the beet crop seed Diagnostic plots showed both models appeared appro- bank, for year 2000, plotted against fitted values on natural scale. The lack of trend in the graph indicates there is no priate. The null hypothesis requires both abundance systematic lack of fit, but variability clearly increases with and diversity measures to be addressed by the FSE. fitted values, so the value of unity for the exponent in the assumed variance–mean relationship is too small. (b) Residuals plotted against fitted values, both on a logarithmic Discussion scale, for the same data analysed using the log-normal model. There is no trend, and the logarithmic transformation has The ambitious scope of the FSE has created numerous equalized the residuals, supporting the assumed value of 2 for challenges, many of them concerned with quantitative the exponent in the variance–mean relationship. issues. Those outlined in this study have largely been resolved from sound knowledge of good experimental design and biometrical practice in ecology (Hairston for the log-normal model for which residuals, plotted 1989; Perry 1989, 1997; McArdle 1996). against fitted values, appeared homogeneous (Fig. 4b). Some problems arose specifically from the large extent This was also indicated by the comparisons between of the study. These included the need for database man- the parametric and non-parametric analyses for the agement, data verification, punching, storage, integrity log-normal and log-linear models. The above probabil- and extraction. The number of protocols is large, but it ity of 0·068 for the log-normal model was close to the is desirable to have a common approach to analysis where value of 0·073 estimated by the randomization test for possible. The development of statistical models that the d-statistic. However, this was not the case for the underpin the analysis was driven by the availability of log-linear model, for which the randomization prob- data that built up slowly over a 3-year period; this resulted ability was estimated as 0·248. This discrepancy cor- in a gradual evolution of analytical methods. These have roborates the implication that the log-linear model been made available through the provision of standard should be viewed with caution for these data. Indeed, Genstat 5 software (Payne & members of the Genstat 5 in similar analyses, the randomization tests were often Committee 1993), running in interactive or batch mode. more conservative than the equivalent parametric tests, The plethora of possible analyses at each stage of the particularly when there were many zero values, as for project imposed a requirement for these to be audited rarer species. Although the P-value of 0·068 was fairly and results stored for later comparison. close to the usual critical value of 0·05, any difference A major statistical issue is whether one of the variance– between treatments was difficult to explain on ecological mean relationships studied will prove clearly more grounds because the samples were taken before any appropriate than others. The possible advantage of treatments had been applied. The multiplicative dif- the log-normal model, evident through diagnostic plots ference of ×1·33 was probably a chance effect; data from and in its agreement between parametric and non- other years should help to clarify the interpretation. parametric analyses, will be examined carefully. The

apparent robustness of the dw-statistic (Rothery, Clark & Perry 2002) may result in its wide use for other ecolo- © 2003 British       gical studies involving count data. Ecological Society,     Journal of Applied Some motivation for the chosen degree of replication Ecology, 40, Analyses of total weed seed abundance from the seed of fields per crop might have been drawn from the lit- 17–31 bank of the two other spring-sown crops during 2000 erature. The planned replication for the FSE exceeds, 30 by more than threefold, any of the comparable 82 estimate how reliable a study it would have been with J. N. Perry et al. terrestrial manipulative ecological experiments under- reduced replication. taken previously, for all plot sizes, reviewed by Moller & Raffaelli (1998) and Raffaelli & Moller (2000). Those Acknowledgements are slightly different from the FSE because they refer to what are termed ‘press experiments’ in animal ecology, The FSE is funded by the Department for Envir- in which animals, often predators, were added to, or onment, Food and Rural Affairs and the Scottish removed from plots. However, they represent the best Executive; further details are given on the website recent set of unbiased data to compare with the FSE. http://www.environment.defra.gov.uk/environment/fse/ However, it is the power analyses that provide the index.htm. We thank all of our colleagues in the consor- confidence that replication is neither too small to detect tium for supplying FSE data and for their help, encourage- obvious effects that might be present, nor so great that ment and advice. We thank Professor Mick Crawley experimental resources could easily be redirected. At (Imperial College at Silwood Park) and Dr Nicholas any stage in the project, available data may be used to Aebischer (Game Conservancy Trust) of the Scientific derive a current estimate of power for a particular vari- Steering Committee for their comments, help in formu- σ 2 ate by estimating from a simple  of lij, specify- lating ideas, and for supplying data that contributed ing a desired value of R and a projected value of n, and towards the initial power calculations. We thank Professor thereby deriving a value of ∆. For the log-normal Chris Pollock for his forbearance during lengthy dis- model, the power for detecting a difference of three (or cussions of mind-numbing statistical questions, and four) times the standard error is about 85% (or 98%). for his constant encouragement to try to keep these All the results described in this study suggest that if simple and clear. Rothamsted Research receives grant- data were available for about 60 fields per crop the FSE aided support from the Biotechnology and Biological would be replicated sufficiently, and should provide Sciences Research Council of the United Kingdom. useful information from which valid statistical inferences may be drawn. This may subsequently be checked References by plotting the logarithm of the estimated multiplica- Aebischer, N.J. (1990) Assessing pesticide effects on non- tive treatment ratio vs. the logit-transformed P-value target invertebrates using long-term monitoring and from the randomization test. The planting of extra fields time-series modeling. Functional Ecology, 4, 369–373. was a sensible insurance against unforeseen losses. Anonymous (1998) Genetically Modified Crops Threaten The fact that power is a continuous function of Wildlife. Press Release EN/98/20, 27 March 1998. English sample size, not a step function, does not weaken the Nature, Peterborough, UK. Anonymous & Perry, J.N. (1999) Design and Analysis of Effi- argument for adequate replication. However, it does cacy Evaluation Trials. EPPO Guideline for the Efficacy strengthen the argument against naive claims that an Evaluation of Plant Protection Products, PP 1/152 (2). Vol. 1. experiment would be useless if there were a marginal Introduction, General and Miscellaneous Guidelines, New failure to achieve some arbitrarily chosen target level of and Revised Guidelines. EPPO/OEPP, Paris, France. replication. This would be the case even if there were Boatman, N.D. & Brockless, M.H. (1998) The Allerton Project: farmland management for partridges (Perdix perdix, only a single variate of interest. It is even more strongly Alectoris rufa) and pheasants (Phasianus colchicus). Perdix the case when there is a very large number of variates of VII: International Symposium on Partridges, Quails and interest, all of which vary differently. It will always Pheasants (eds M. Birkan, L.M. Smith, N.J. Aebischer, be the case that for some of these power will be large F.J. Purroy & P.A. Robertson), pp. 563–574. Gibier Faune while for others it will be small. In any event, even when Sauvage, Paris, France. Buzzard, C. (2000) Lessons Learned from on-Farm Trials: the the magnitude of the effect required to be detected PD/A CRSP Experience. PD/A CRSP, Oregon State Uni- has been specified quantitatively, there remains the versity, Corvallis, OR. difficulty of interpreting what this means in terms of Carroll, R.J. & Ruppert, D. (1988) Transformation and environmental impact, given the buffering and resilience Weighting in Regression. Chapman & Hall, New York, NY. in ecosystems. Another desirable aspect of an experi- Cochran, W.G. (1938) Some difficulties in the statistical analysis of replicated experiments. Empire Journal of ment is consistency of results, especially one such as Experimental Agriculture, 6, 157–175. the FSE in which there are many protocols, some of Duffield, S.J. & Aebischer, N.J. (1994) The effect of spatial which themselves involve many taxa. scale of treatment with dimethoate on invertebrate popula- Future studies to assess the ecological effects of GM tion recovery in winter-wheat. Journal of Applied Ecology, crops on a large scale may be required for other crops, 31, 263–281. Firbank, L.G. & Forcella, F. (2000) Agriculture – genetically in other countries, and of alternative GM traits. It is modified crops and farmland biodiversity. Science, 289, unlikely that there will often be sufficient funding for 1481–1482. experiments as intensive as the FSE. Therefore, there is Firbank, L.G., Dewar, A.M., Hill, M.O., May, M.J., Perry, an urgent need for further statistical methodological J.N., Rothery, P., Squire, G.R. & Woiwod, I.P. (1999) Farm- © 2003 British studies to develop experimental designs or modelling scale evaluation of GM crops explained. Nature, 399, 727–728. Ecological Society, Firbank, L.G., Heard, M.S., Woiwod, I.P., Hawes, C., Haughton, Journal of Applied approaches that allow efficient study at reduced cost. A., Champion, G., Scott, R., Hill, M.O., Dewar, A., Squire, Ecology, 40, One possible initial approach, available during late G.R., May, M., Brooks, D.R., Bohan, D., Daniels, R.E., 17–31 2003, might be to backcast results from the FSE to Osborne, J.L., Roy, D., Black, H.I.J., Rothery, P. & Perry, J.N. 31 (2003) An introduction to the Farm-Scale Evaluations of J.M., Prew, R.D. & Spink, J. (1995) LINK integrated farm- Biometrical issues genetically modified herbicide-tolerant crops. Journal of ing systems: a considered approach to crop protection. of GMHT FSE Applied Ecology, 40, 2–16. Integrated Crop Protection: Towards Sustainability? (eds Frampton, G.K. (1999) Spatial variation in non-target effects R.W. McKinley & D. Atkinson), pp. 331–338. BCPC, of the insecticides chlorpyrifos, cypermethrin and pirimi- Farnham, UK. carb on Collembola in winter wheat. Pesticide Science, 55, Payne, R.W. & members of the Genstat 5 Committee (1993) 875–886. Genstat 5 Release 3 Reference Manual. Oxford University Hairston, N.G.S. (1989) Ecological Experiments: Purpose, Press, Oxford, UK. Design and Execution. Cambridge University Press, Cam- Pearson, E.S. & Hartley, H.O. (1976) Biometrika Tables for bridge, UK. Statisticians, Vol. 2. Griffin, High Wycombe, UK. Hastings, N.A.J. & Peacock, J.B. (1975) Statistical Distribu- Perry, J.N. (1986) Multiple-comparison procedures: a dissent- tions. Butterworths, London, UK. ing view. Journal of Economic Entomology, 79, 1149–1155. Heads, P.A. & Lawton, J.H. (1983) Studies on the natural Perry, J.N. (1989) Review: population variation in entomo- enemy complex of the holly leaf-miner – the effects of scale logy: 1935–50. I. Sampling. Entomologist, 108, 184–198. on the detection of aggregative responses and the implica- Perry, J.N. (1997) Statistical aspects of field experiments. tions for biological control. Oikos, 40, 267–276. Methods in Ecological and Agricultural Entomology (eds Kennedy, P.J. (1994) The distribution and movement of D.R. Dent & M.P. Walton), pp. 171–201. CAB Interna- ground beetles in relation to set-aside arable land. Cara- tional, Wallingford, UK. bid Beetles: Ecology and Evolution (eds K. Desender, Perry, J.N., Parker, W.E., Alderson, L., Korie, S., Blood- M. Dufrene, M. Loreau, M.L. Luff & J.-P. Maelfait), Smyth, J.A., McKinlay, R. & Ellis, S.A. (1998) Simulation pp. 439–444. Kluwer Academic Publishers, Dordrecht, the of counts of aphids over two hectares of Brussels sprout Netherlands. plants. Computers and Electronics in Agriculture, 21, 33–51. Kennedy, P.J., Conrad, K.F., Perry, J.N., Powell, D., Aegerter, Potts, G.R. & Vickerman, G.P. (1974) Studies on the cereal J., Todd, A.D., Walters, K.F.A. & Powell, W. (2001) Com- ecosystem. Advances in Ecological Research, 8, 107–197. parison of two field-scale approaches for the study of effects Raffaelli, D. & Moller, H. (2000) Manipulative field experi- of insecticides on polyphagous predators in cereals. Applied ments in animal ecology: do they promise more than they Soil Ecology, 17, 253–266. can deliver? Advances in Ecological Research, 30, 299–338. Krebs, J.R., Wilson, J.D., Bradbury, R.B. & Siriwardena, Robinson, R.A. & Sutherland, W.J. (2002) Post-war changes G.M. (1999) The second silent spring? Nature, 400, 611– in arable farming and biodiversity in Great Britain. Journal 612. of Applied Ecology, 39, 157–176. Lennon, M. (1998) Design and analysis of multiple site, large Rothery, P., Clark, S.J. & Perry, J.N. (2002) Design and plot field experiments. Unpublished PhD Thesis. University analysis of farm-scale evaluations of genetically modified of Reading, Reading, UK. herbicide-tolerant crops. Proceedings of the XXith Inter- Lewis, T. (1967) The horizontal and vertical distribution of national Biometric Conference, 21–26 July 2002, Freiburg, flying insects near artificial windbreaks. Annals of Applied Germany, Invited Papers (ed. M. Schumacher), pp. 351–364. Biology, 60, 23–31. German Region of the International Biometric Society, McArdle, B.H. (1996) Levels of evidence in studies of com- Freiburg, Germany. petition, predation and disease. New Zealand Journal of Schuler, T.H., Poppy, G.M., Kerry, B.R. & Denholm, I. Ecology, 20, 7–15. (1999) Potential side effects of insect-resistant transgenic McCullagh, P. & Nelder, J.A. (1989) Generalized Linear plants on arthropod natural enemies. Trends in Biotechno- Models. Chapman & Hall, London, UK. logy, 17, 210–216. Manly, B.F.J. (1994) Randomization, Bootstrap and Monte Sokal, R.R. & Rohlf, F.J. (1981) Biometry: The Principles and Carlo Methods in Biology, 2nd edn. Chapman & Hall, Practice of Statistics in Biological Research. W.H. Freeman, London, UK. New York, NY. Moller, H. & Raffaelli, D. (1998) Predicting risks from new Sotherton, N.W., Jepson, P.C. & Pullen, A.J. (1988) Criteria organisms: the potential of community press experiments. for the design, execution and analysis of terrestrial non- Statistics in Ecology and Environmental Monitoring: Risk target invertebrate field tests. BCPC Monograph, 40, 183–190. Assessment and Decision Making in Biology (eds D.J. Taylor, L.R. (1961) Aggregation, variance and mean. Nature, Fletcher, L. Kavalieris & B.J.F. Manly), pp. 131–156. 189, 732–735. Otago University Press, Dunedin, New Zealand. Taylor, L.R., Woiwod, I.P. & Perry, J.N. (1978) The density- Morgan, B.J.T. (1984) Elements of Simulation. Chapman & dependence of spatial behaviour and the rarity of randomness. Hall, London, UK. Journal of Animal Ecology, 47, 383–406. Moultine, G. & Bluman, A.G. (2001) Minitab Manual for Use Vet, L.E.M. (1999) From chemical to population ecology: with Elementary Statistics. McGraw-Hill, New York, NY. infochemical use in an evolutionary context. ISCE Silverstein- Norowi, H.M., Perry, J.N., Powell, W. & Rennolls, K. (2000) Simeone Lecture Award, 1 May 1998. Journal of Chemical The effect of spatial scale on interactions between two Ecology, 25, 31–49. weevils and their parasitoid. Ecological Entomology, 25, Watkinson, A.R., Freckleton, R.P., Robinson, R.A. & 188–196. Sutherland, W.J. (2000) Predictions of biodiversity response Numerical Algorithms Group (1997) The NAG Fortran to genetically modified herbicide-tolerant crops. Science, Library Manual Mark 18. Numerical Algorithms Group 289, 1554–1557. Ltd, Oxford, UK. Ogilvy, S.E., Turley, D.B., Cook, S.K., Fisher, N.M., Holland, Received 5 June 2002; final copy received 26 October 2002

© 2003 British Ecological Society, Journal of Applied Ecology, 40, 17–31 From: @syngenta.com Sent: 16 July 2013 10:00 To: Pywell, Richard F. Cc: Bullock, James M.; @syngenta.com; @syngenta.com; @syngenta.com; @syngenta.com Subject: RE: Scientific paper outlining design of neonicotinoid-pollinator experiment

Dear Richard,

Thank you very much for the outline of the methodological paper. We discussed already the concept with and we are all very positive about it.

I will come back to you after my call with tomorrow.

Best wishes,

From: Pywell, Richard F. [mailto:[email protected]] Sent: Montag, 15. Juli 2013 18:29 To: Cc: Bullock, James M. Subject: Scientific paper outlining design of neonicotinoid-pollinator experiment

Dear ,

As discussed in the teleconference earlier.

Outline of the methodological paper:

This approach was used to great effect for the Farm scale Evaluation of GM herbicide- tolerant crop project (see attached paper).

Best wishes

Richard

':01491 692356

*: [email protected]

______

This message may contain confidential information. If you are not the designated recipient, please notify the sender immediately, and delete the original and any copies. Any use of the message by you is prohibited. From: Pywell, Richard F. Sent: 25 September 2013 14:46 To: Wells, Nick; Cc: Bullock, James M. Subject: URGENT - FW: CEH Wallingford Confidentiality (reciprocal) 20th Sept 2013.doc Attachments: CEH Wallingford Confidentiality (reciprocal) 20th Sept 2013.doc

Dear Nick and

Please see attached a reciprocal data licensing agreement btw Syngenta and CEH to allow us access to their field trials data on neonicotinoids and bees for Mark’s signature.

This is to allow us to do statistical power testing to design a major pan European field experiment.

It is part of a contract between CEH, Bayer and Syngenta we are currently negotiating.

On quick scanning the wording seems okay to me.

URGENT - Please can you let me know if you think the terms are acceptable as I need to get back to Syngenta.

Many thanks

Richard

From: @syngenta.com [mailto: @syngenta.com] Sent: 20 September 2013 15:09 To: Pywell, Richard F. Cc: @syngenta.com; @syngenta.com; @syngenta.com; @syngenta.com Subject: FW: CEH Wallingford Confidentiality (reciprocal) 20th Sept 2013.doc

Good afternoon Richard,

This is the SYN proposal for the data confidentiality part only (and only between CEH and Syngenta) to allow you to receive our data. It is kindly put together by (my thanks) and I have suggested a wide wording in the technical part so that the agreement will not need refining should you require more but slightly different data.

Should this be acceptable to you, I think we could move pretty quickly on it- regards,

From: Sent: 20 September 2013 14:28 To: Subject: CEH Wallingford Confidentiality (reciprocal) 20th Sept 2013.doc

Syngenta Limited, Registered in England No 2710846 Registered Office : Syngenta Limited, European Regional Centre, Priestley Road, Surrey Research Park, Guildford, Surrey, GU2 7YH, United Kingdom

This message may contain confidential information. If you are not the designated recipient, please notify the sender immediately, and delete the original and any copies. From: Pywell, Richard F. Sent: 11 October 2013 14:52 To: @SYNGENTA.COM'; @bayer.com Cc: @SYNGENTA.COM; @SYNGENTA.COM; Bullock, James M. Subject: RE: Defra document Attachments: Bee health monitoring project_v2_rpywell.docx

Dear ,

Please see attached my comments.

I have cc.d James Bullock as he may have a view.

My key point is that the stated aim of the project must be an open test of the effects of NNIs on pollinators (see my text), not as you state ‘to show that modern beekeeping, as well as natural pollinator diversity and intensive farming, are compatible in a tangible, compelling and scientifically grounded way.’

Best wishes

Richard

From: @SYNGENTA.COM [mailto: @SYNGENTA.COM] Sent: 11 October 2013 07:03 To: Pywell, Richard F.; c @bayer.com Cc: @SYNGENTA.COM; @SYNGENTA.COM Subject: Defra document

Dear ,

We received a request yesterday from the UK Government to share more information about the bee health monitoring project (Syngenta & Bayer). Defra will have an internal meeting on Monday.

It would be very useful if you could review this document and send your comments to by today.

Apologies in advance for the short notice.

Thank you very much!

Best regards,

Syngenta Crop Protection AG WRO-1008.7.25 Schwarzwaldallee 215 CH-4058 Basel Switzerland @syngenta.com www.syngenta.com

Cc: Shore, Richard F.; Bullock, James M.; Hails, Rosemary S. Subject: Draft agenda for Neonicotinoid Experiment Workshop 18-19 Nov Wallingford

Dear all,

Please see attached a draft agenda for the workshop for comment.

Based on this agenda I will prepare a powerpoint presentation with all the issues we need to address over the two days.

We would like to gather sufficient information from the Workshop to prepare a detailed experimental design and monitoring protocol for approval by the Project Advisory Group.

Best wishes

Richard

:01491 692356

: [email protected]

______

Pan-European study of neonicotinoid effects on pollinators

Planning Workshop at CEH Wallingford 12:00 pm 18-19 November 2013 Draft Agenda

Present: CEH - Richard Pywell, ; Syngenta – Bayer -

Apologies: Richard Shore, James Bullock,

Aim: to produce a detailed design and monitoring protocol for the experiment

1. Introductions & domestic arrangements

2. Project background - aim, governance, budget, organisation

3. Existing trials data - brief description - statistical power analysis / meta-analysis

4. Experimental design - study countries, landscape, replication, treatments, plot area, mitigation measures, seed dose rates, crop agronomy etc

5. Response measures - chemical, agronomic, biological

6. Review of actions

7. Next steps

8. AOB

From: Pywell, Richard F. Sent: 29 November 2013 10:44 To: @SYNGENTA.COM; @SYNGENTA.COM'; @bayer.com' Subject: FW: EFSA

Dear ,

For information:

James Bullock spoke with & colleagues from the Pesticide Unit at EFSA

Email address: @efsa.europa.eu

Best wishes

Richard

From: Bullock, James M. Sent: 29 November 2013 10:15 To: Pywell, Richard F. Subject: FW: EFSA

Hi Richard, see below. The contact was

From: Bullock, James M. Sent: 20 November 2013 11:09 To: Pywell, Richard F. Subject: EFSA

Hi Richard, I've just had a useful meeting with & colleagues from the Pesticide Unit. They were friendly & appreciated being consulted. Some points which it is easier to list by email.

1) Concerning representation on the Advisory Group, they thought it would not be appropriate from anyone from the Unit itself (ie EFSA employees) to be involved as they would be ultimately evaluating what we have done & there may be conflict of interest. They suggested however that someone from the PPR Panel (Scientific Panel on Plant Protection Products and their Residues) might be on the AG, ie an independent expert with knowledge of EFSA needs. They suggested we email to discuss possibilities.

2) & colleagues would be interested to see our protocol when developed, so they can keep updated with progress

3) They suggested we might contact the EC itself, ie DG SANCO, about this project to see if they would like to be involved. They gave me the names of (Head of Unit).

4) They said the CRD has a protocol checking role, which we might use - apparently under 'Applicant Advice' on the CRD website.

4) They pointed out that there is a lot of info on past studies which is linked to the EFSA conclusions on NNIs. Go to the relevant Conclusion, click on Register of Questions & then on Peer Review Report

5) They suggested we use maximum relevant dosage & sowing rates in each location - ie the maximum of the range usually used - to maximise the chance of picking up effects. Indeed they criticised some field studies which have not done so.

6) They liked the mitigation strips idea.

7) For the mitigation & control they brought up the issue of residues from past plantings, but also said that there are issues of NNI drift in dust. Whether we sample dust(!) or assess this by looking at NNI levels in bees is unclear.

8) All docs for the final decision will be in by 25 May 2015, so they appreciate we will have no results by then.

I can chat now or when I return ************************* Prof James Bullock Centre for Ecology & Hydrology Benson Lane Wallingford OX10 8BB UK

. From: @syngenta.com Sent: 04 December 2013 14:45 To: @bayer.com Cc: Pywell, Richard F.; Keightley, Stephen C.; .; @syngenta.com Subject: RE: Draft Agreement with Syngenta and Bayer - Pan-European study of the effects of neonicotinoids on pollinator populations Attachments: Cooperation Agreement Bee Monitoring_Updated NERC CM RDV 04122013.docx

Dear ,

Yes, I am very happy with both format and content. I have some minor changes to propose for the contract, while, for the annex I have reported the changes that we already agreed on another document that is describing the project.

I suggest we wait for Bayer’s comment and then we can have a consolidated review.

Thank you.

Best wishes,

From: Sent: Mittwoch, 4. Dezember 2013 12:38 To: ; @bayer.com Cc: Pywell, Richard F.; Keightley, Stephen C.; Subject: FW: Draft Agreement with Syngenta and Bayer - Pan-European study of the effects of neonicotinoids on pollinator populations

Dear

I am just checking, on behalf of my colleague , to see if you are in a position yet to confirm whether you are happy with the final format and content of the amended contract?

Many thanks

NERC Legal Team Polaris House North Star Avenue Swindon SN2 1EU

www.nerc.ac.uk Planet Earth Online Follow @NERCscience

______From: Sent: 20 November 2013 15:50 To: Pywell, Richard F.; @bayer.com; @SYNGENTA.COM Cc: @SYNGENTA.COM; Keightley, Stephen C. Subject: RE: Draft Agreement with Syngenta and Bayer - Pan-European study of the effects of neonicotinoids on pollinator populations

Dear

Please find attached the amended contract.

Please let me know if you are happy with the final format and content.

Many thanks.

www.nerc.ac.uk

Planet Earth Online

Follow @NERCscience

______From: Pywell, Richard F. Sent: 11 November 2013 10:00 To: @bayer.com; @SYNGENTA.COM Cc: @SYNGENTA.COM; ; Keightley, Stephen C. Subject: Draft Agreement with Syngenta and Bayer - Pan-European study of the effects of neonicotinoids on pollinator populations

Dear ,

from NERC contracts team will be sending an amended contract to you in the near future.

Best wishes

Richard

':01491 692356

*: [email protected]

<< OLE Object: Picture (Device Independent Bitmap) >>

This message may contain confidential information. If you are not the designated recipient, please notify the sender immediately, and delete the original and any copies. Any use of the message by you is prohibited. From: @syngenta.com Sent: 12 December 2013 13:50 To: Pywell, Richard F. Cc: Subject: Pan-European study of the effects of neonicotinoids on pollinators - Anticipation of the value of work completed in 2013

Importance: High

Dear Richard,

I would like to touch base with you on two important elements.

The first one: at a meeting with UK DEFRA Chief Scientist , had the opportunity to share information on our Pan-European study. He was really happy with the general proposed work programme and the idea of a Project Advisory Group chaired by . Also, he suggested that in order to allow UK Government Research Institutions like BBSRC, NERC, ESRC and DEFRA to contribute towards such a programme that everybody should pay into a “Collective Fund” which is then used to fund the research Project. The “Independent Advisory Group” would then steer the project and help “socialising” of the protocol.

The second, for which I prepared the message below is to help us to prepare the accruals for the end of the year. It is clearly just a draft that you can amend as appropriate. Can you please send it back to me by tomorrow?

Is there any possibility to have a quick call?

Thank you.

Best wishes,

------

Dear

Our joint efforts to develop a research and demonstration project which aims to assess potential effects of commercial, field- scale application of neonicotinoid seed treatments on pollinator populations will continue in 2014, 2015 and 2016 as planned. With this email I would like to highlight that some main components within the project have been advanced.

i) Pan-European Field-Scale Experiment a. Establishment of a first nucleus of the consortium of independent EU research organizations led by the NERC Centre for Ecology and Hydrology b. Access to the Syngenta registration dossier on NNI. c. Preparation of a methodological paper describing the rigorous design of the experiment.

ii) Demonstration farm network a. Establishment of the demonstration farm network with 15 demonstration fields in five countries. b. Definition of the research protocol.

However, we have not yet sent to you any invoices for the work undertaken in 2013. The invoices will be sent in January either directly from CEH or from partners or from a “Collective Research Fund” in case we establish one as suggested by DEFRA.

The expected value of work undertaken in 2013 is as follow:

· Total project value: . · Expected value of work completed in 2013: · Given the fact that Syngenta and Bayer are equal funders of the work, the total expected value of work completed in 2013 for Syngenta is:

Best wishes, Richard Pywell

This message may contain confidential information. If you are not the designated recipient, please notify the sender immediately, and delete the original and any copies. Any use of the message by you is prohibited. From: @syngenta.com Sent: 17 December 2013 15:47 To: Pywell, Richard F. Subject: FW: Pan-European study of the effects of neonicotinoids on pollinators - Anticipation of the value of work completed in 2013

As discussed. Best regards,

From: Pywell, Richard F. To: Cc: Bullock, James M. Sent: Mon Dec 16 19:33:35 2013 Subject: RE: Pan-European study of the effects of neonicotinoids on pollinators - Anticipation of the value of work completed in 2013

Dear ,

As requested here is a brief update on progress regarding the research and demonstration project to measure the effects of neonicotinoids on pollinator populations we plan to undertake in 2014, 2015, 2016 (and finishing in spring 2017 with final assessment of overwinter mortality of exposed honeybee colonies): i) Pan-European Field-Scale Experiment (CEH) a. Co-ordinate a two day technical workshop to discuss the design and monitoring of the experiment b. Produce a detailed draft experimental protocol for comment and feasibility analysis c. Undertake a preliminary statistical power analysis using the EFSA (2013) guidelines d. Establish an independent Project Advisory Group to oversee the project e. Agree access to Syngenta commercial field trial data to inform project feasibility and design ii) Demonstration farm network (Syngenta, Bayer, Eurofins) a. Establishment of the demonstration farm network with 15 demonstration fields in five countries. b. Develop draft pollinator monitoring protocol.

We will not send any invoices for the work undertaken during the project planning and feasibility phase (1 Aug- 31 Dec 2013).

Subject to contract agreement and project feasibility, invoices will be sent from 1 January 2014 either directly from CEH. Alternatively, in the event of other funders joining the project (e.g. Defra), payment may be made to a “Collective Research Fund”.

The expected value of work undertaken in 2013 is as follow: • Total project value (2013-2017): . • Expected value of work completed in 2013: . • Given the fact that Syngenta and Bayer are equal funders of the work, the total expected value of work completed in 2013 for Syngenta is:

Best wishes

Richard

From: Pywell, Richard F. Sent: 18 December 2013 16:41 To: @syngenta.com' Cc: Hails, Rosemary S. Subject: RE: Pan-European study of the effects of neonicotinoids on pollinators - Anticipation of the value of work completed in 2013

Dear ,

It is my understanding that this estimate of expenditure is to commit or ‘ring fence’ the Syngenta project budget for 2013-2017 for your accountants. CEH will invoice Syngenta for a nominal payment for the period 1 Jan to 31 Mar 2014 ( reflecting actual staff time expenditure in this period. We will expect to invoice for significantly larger amounts in the following quarters as the project get underway.

As requested here is a brief update to date (Aug-Dec 2013) on progress regarding the research and demonstration project to measure the effects of neonicotinoids on pollinator populations we plan to undertake in 2014, 2015, 2016 (and finishing in spring 2017 with final assessment of overwinter mortality of exposed honeybee colonies): i) Pan-European Field-Scale Experiment (CEH) a. Co-ordinate a two day technical workshop to discuss the design and monitoring of the experiment b. Produce a detailed draft experimental protocol for comment and feasibility analysis c. Undertake a preliminary statistical power analysis using the EFSA (2013) guidelines d. Establish an independent Project Advisory Group to oversee the project e. Agree access to Syngenta commercial field trial data to inform project feasibility and design ii) Demonstration farm network (Syngenta, Bayer, Eurofins) a. Establishment of the demonstration farm network with 15 demonstration fields in five countries. b. Develop draft pollinator monitoring protocol.

We will not send any invoices for the work undertaken during the project planning and feasibility phase (1 Aug- 31 Dec 2013).

Subject to contract agreement and project feasibility, invoices will be sent for payment from 1 January 2014 directly from CEH. Alternatively, in the event of other funders joining the project (e.g. Defra), payment may be made to a “Collective Research Fund”.

The expected value of work undertaken in 2013 is as follow: • Total project value (2013-2017): . • Expected value of work completed in 2013: • Given the fact that Syngenta and Bayer are equal funders of the work, the total expected value of work completed in 2013 for Syngenta is:

Best wishes

Richard

:01491 692356

: [email protected]

______

From: @syngenta.com Sent: 10 January 2014 14:33 To: Pywell, Richard F. Cc: @bayer.com Subject: Revised Agreement.docx Attachments: Revised Agreement.docx

Dear Richard,

As discussed in our recent teleconference, please find here the suggested amendments to make the agreement compliant with what was suggested by DEFRA.

The impact of the changes is minimal as the majority of the elements were already included in the previous approved version.

I have highlighted the major elements related to the governance (e.g. independence of CEH) and to the differentiation of the consortium members (e.g. research entities and sponsors).

While our lawyers will have, as usual, the possibility to wordsmith the text, I believe this draft offer the possibility for Prof. Rosie Hails to fully reassure Prof. Ian Boyd on forthcoming meeting on 20th Jan.

I hope it helps.

Best wishes,

This message may contain confidential information. If you are not the designated recipient, please notify the sender immediately, and delete the original and any copies. Any use of the message by you is prohibited. From: Pywell, Richard F. Sent: 23 January 2014 12:13 To: @syngenta.com'; ' ' Cc: @SYNGENTA.COM'; @bayer.com'; @syngenta.com'; @SYNGENTA.COM; @bayer.com; Bullock, James M.; Shore, Richard F.; Hails, Rosemary S.; Subject: NNI Experiment costed feasibility study

Dear ,

CEH have now completed a costed feasibility study for the NNI experiment (attached).

In this we explore various design options for the field experiment based on a detailed statistical power analysis and assessment of costs. We have also taken account of the possibility of additional funding from other project partners based on recent discussions with Defra and NERC.

From this we make a number of recommendations that we would like you to consider.

Hopefully this document will inform decisions regarding the final design of the experiment going forward to the Scientific Advisory Committee.

It would probably be advantageous for a core group of us to meet to discuss this face-to-face in the near future (perhaps at Jealott’s Hill or Wallingford as it is close to Heathrow?).

I look forward to hearing from you.

Best wishes

Richard

:01491 692356

: [email protected]

______

From: @syngenta.com Sent: 23 January 2014 16:09 To: Pywell, Richard F. Subject: RE: NNI Experiment costed feasibility study

Dear Richard,

Thank you for sending the feasibility study. Your recommendations require proper consideration and discussion both internally and with BCS. Therefore, I may need a few days to be able to come back to you. I will ask to investigate a possible date to meet face-to-face in the UK.

Best wishes,

From: Pywell, Richard F. [mailto:[email protected]] Sent: Donnerstag, 23. Januar 2014 13:13 To: Cc: @bayer.com; @bayer.com; Bullock, James M.; Shore, Richard F.; Hails, Rosemary S.; Subject: NNI Experiment costed feasibility study

Dear ,

CEH have now completed a costed feasibility study for the NNI experiment (attached).

From this we make a number of recommendations that we would like you to consider.

Hopefully this document will inform decisions regarding the final design of the experiment going forward to the Scientific Advisory Committee.

It would probably be advantageous for a core group of us to meet to discuss this face-to-face in the near future (perhaps at Jealott’s Hill or Wallingford as it is close to Heathrow?).

I look forward to hearing from you.

Best wishes

Richard

':01491 692356

*: [email protected]

______

Options for an experiment to quantify the impacts of neonicotinoids (NNIs) on domesticated and wild bees

, Richard Shore, James Bullock, Rosie Hails, & Richard Pywell NERC Centre for Ecology and Hydrology

January 2014

1. Background

In July 2013 Syngenta and Bayer approached the NERC Centre for Ecology and Hydrology (CEH) to design and implement an experiment to quantify objectively the impact on domesticated and wild bees of two neonicotinoid (NNI) seed treatments in commercially grown oilseed rape crops (‘Clothianidin’ Bayer CropScience and ‘Thiamethoxam’ Syngenta). At subsequent meetings in Aug and Nov 2013 detailed requirements of the experiment were agreed. The budget for this work is currently over three years starting January 2014.

This paper presents the findings of a project feasibility study based on: 1) statistical power analysis to determine the experimental replication to achieve different levels of effect delectability, and 2) detailed costing of the experiment based on the best available estimates. Both follow the EFSA guidance on the risk assessment of plant protection products on bees1 as far as practically feasible.

From this analysis we make a number of recommendations for the design of the experiment depending on potential funding scenarios from the project partners, including Bayer/Syngenta, and in the UK the Department for Environment, Food and Rural Affairs (Defra) and the Natural Environmental Research Council (NERC).

2. Power analysis

2.1 Statistical power analyses have been undertaken to estimate the minimum sample size required to detect an effect of a given size. In this context an effect size refers to the percentage change in a population mean (of the response variable of interest, such as bee numbers) that would be expected to be detected with probability p<0.05 in response to the use of Neonicotinoids. Power analysis can also be used to calculate the minimum effect size that is likely to be detected with a certain probability where resources place an upper limit on the degree of replication. In the following discussion, a replicate block refers to a cluster of three sites (or farms). Within each replicate block one each of the three sites would have one of the ‘Neonicotinoid’ treatment levels represented by Control, Clothianidin and Thiamethoxam treated oilseed rape. In the context of this study it is important that the experiment has sufficient statistical power to detect an effect of a specified magnitude, if present, with a probability of at least 0.8. In the absence of a power analysis, it is not

1 EFSA (2013). Guidance on the risk assessment of plant protection products on bees (Apis mellifera, Bombus spp. and solitary bees). EFSA Journal 11:3295, 266.

possible to place a rejection of the null hypothesis into context; it could just be that the experiment failed to detect an impact of the Neonicotinoids due to lack of statistical power.

2.2 EFSA guidelines recommend that a power analysis be undertaken prior to the design of any field trial. EFSA provide a methodology to achieve such a power analysis and propose that the number of replicate blocks should be sufficiently high to detect a 7 % effect size1.

2.3 We undertook a power analysis using this EFSA methodology with what at present represents the best available data from field trials of oilseed rape and monitoring of honeybees (Apis mellifera) based on published Syngenta field trials2 and published CEH studies of native bumblebees (Bombus terrestris)3. Full details of the power analysis and a description of the data sets used for this feasibility study are provided in Annex 1 (see below). In all cases we have restricted where possible power analyses to honey bee hives or bumblebee colonies found in association with oilseed rape fields as this will be representative of the study system proposed in the current project.

2.4 The preliminary analysis considered two key response variables, namely colony strength of honeybees (Liebefeld estimate of worker numbers) and cumulative worker counts for B. terrestris. There remains considerable scope for further power analysis using additional data sets (e.g. Bayer field trials) and for different end-points (e.g. overwinter survival), and it should be noted that these variables may well require additional replication to achieve desired power. However, colony strength and cumulative worker counts represent core population measurements for honey bees and bumblebees respectively.

2.5 High between and within site variation in colony strength for honeybees suggests that detection of a 7 % change in worker bee numbers in response to application of the neonicotinoid seed treatments would require 154 replicate blocks (triplicates of Control, Clothianidin and Thiamethoxam treated sites); this assumes 5 Honey bee hives per site.

2.6 Due to large between colony variability, detection of 7 % change in B. terrestris worker numbers in response to neonicotinoid seed treatments would require 650 replicate blocks (triplicates of Control, Clothianidin and Thiamethoxam treated sites) assuming 7 colonies per site.

2.7 Detection of a 7 % effects size in the context of a field experiment is considered to be unrealistic given the degree of within and between hive / colony variance. We believe that reducing the power of statistical tests (i.e. accepting that the study would only be able to detect a larger than 7 % drop from the population mean in response to neonicotinoid use) is a more pragmatic and realistic approach. Because the relationship between replicate block number and detection power is not linear, it is still possible to design relatively sensitive field trials with a more manageable level of replication, as shown below.

2 Pilling E., Campbell P., Coulson M., Ruddle N. & Tornier I. (2013). A four-year field program investigating long-term effects of repeated exposure of honey bee colonies to flowering crops treated with thiamethoxam. PLoS ONE, 8.

3 Carvell, C., Rothery, P., Pywell, R. F., & Heard, M. S. (2008). Effects of resource availability and social parasite invasion on field colonies of Bombus terrestris. Ecological Entomology, 33(3), 321-327. 2

• Assuming 5 honey bee hives per site, the power analysis suggests that the following numbers of replicate blocks (triplicates of Control, Clothianidin and Thiamethoxam treated sites) would be needed to detect 7, 15 or 20% declines in peak worker numbers:

. 7 % effect size: 154 replicate blocks . 15 % effect size: 20 replicate blocks . 20 % effect size: 13 replicate blocks

• Assuming 7 B. terrestris colonies per site, the power analysis suggests that the following number of replicate blocks (triplicates of Control, Clothianidin and Thiamethoxam treated sites) would be needed to detect 7, 15 or 20% declines in total worker numbers

. 7 % effect size: 650 replicate blocks . 15 % effect size: 38 replicate blocks . 20 % effect size: 23 replicate blocks

3. Costs of monitoring

3.1 Project costs were calculated on the basis of the experiment being undertaken in the UK and used CEH daily staff costs (inclusive of travel to field sites & consumables required ) for just the core experiment (Tier 1) which consists of three treatments: • Control: No Neonicotinoids • Treat 1: Clothianidin • Treat 2: Thiamethoxam

3.2 Costs were divided into a) overall project management and co-ordination, and b) field measurements. These were restricted to just the EFSA-recommended primary response measures or practical adaptations of these for Honey bee and B. terrestris colonies only (excluding measures of other native pollinators). It should be noted that costings for most variables were based on a significantly lower frequency of monitoring than recommended by EFSA for practical reasons. As such they should be considered the minimum requirement, though this would need to be discussed with EFSA. Full details of the proposed measures and associated costing are provided in Annex 2.

3.3 There is considerable uncertainty around costs for some activities and there would be merit in sharing knowledge with Bayer and Syngenta staff with experience of undertaking field exposure trials on pollinators.

3.4 This is a high profile and complex project so overall project co-ordination and management represents a significant but nevertheless essential charge to the budget (estimated to be 18%- see Appendix 2).

3.5 To undertake even the basic EFSA-recommended primary response measures will be costly in terms of consumables and staff time. We estimate it will cost p.a. to implement and monitor the experiment at an individual site, of which three sites are present in the basic experimental replicate block (composed of a triplicate of Control, Clothianidin and

Thiamethoxam treated sites). Each replicate block will therefore cost to run for a single year.

4. Feasibility of current project

4.1 The current budget for the project is £ at current exchange rates). This would fund a Tier 1 core experiment with 13 replicate blocks (39 sites in total) in just one year.

4.2 Based on the statistical power analysis this experiment would have the power to detect a 20 % drop in peak honey bee worker numbers and a 30 % drop in for B. terrestris worker numbers (significance of p=0.05 and a power of β=0.80).

4.3 This level of detection, particularly in the case of B. terrestris, poses a risk to the project in that there will be low confidence in a ‘no effect result’ of Neonicotinoids. EFSA recommend detection of a 7 % effect size, and while such a degree of detection may be unrealistic in field trials, this issue needs to be acknowledged. Note, the detection of a 20 % effect size is biologically justifiable in the context of ecological field trials where high levels of between and within site variance make the detection of smaller effect sizes (<15 %) impractical.

4.4 If we proceed on this basis then CEH recommends the study is undertaken in a single country (the UK) as working in additional countries will increase further the variation between sites and so reduce the power of detection and increase the replication required. Working in more than one country will also increase project management and co-ordination costs.

4.5 There may also be a logistical advantage in splitting the resource such that 6 replicates are monitored in 2015 (oilseed rape to be sown in 2014) (ideally 7 replicates should resources become available) and the balance of 7 replicates are measured in 2016 (oilseed rape to be sown in 2015). Should additional resources become available for additional replicate blocks the study would benefit from monitoring the crop over a three year period. This would allow the experimental design to: a) eliminate statistically the differences between years in the analysis; b) mitigate against unusual conditions in any one year.

5. Alternative strategies for the Experiment

5.1 Scenario 1: Alternative experimental design (no additional funding)

If no additional funding is available then the one practical means of increasing the power of the experiment to detect a smaller effect size would be to compare just one neonicotinoid product (Clothianidin or Thiamethoxam) to the no neonicotinoid control. This reallocation of current resources from replicate blocks composed of three sites (triplicates of Control, Clothianidin and Thiamethoxam treated sites) to replicate blocks of two sites (pairs of Control and a single Neonicotinoid treated sites) would increase replication. This would allow us to increase the number of replicate blocks to 19 (Control and a single Neonicotinoid treated sites).

This approach would mean that the power of detection would increase so that:

• For 19 replicate blocks (Control and one neonicotiniod) we would be able to detect a c. 16 % drop in the peak abundance of honey bee worker numbers (significance of p=0.05 and with a power of 0.8). • For 19 replicates we would be able to detect a c. 23 % effect size drop in the total number of worker B. terrestris (significance of p=0.05 and with a power of 0.80).

5.2 Scenario 2: Additional funding from Bayer/Syngenta

Scenario 2a) Additional replicates of the Tier 1 experiment The detection of a 7 % effects size is unfeasible given the practical constraints of real world variation in the hive and colony sizes of both the honeybees and B. terrestris. However, additional funding that increases the number of replicate blocks (i.e. triplicates of Control, Clothianidin and Thiamethoxam treated sites) to 20 would result in a considerable increase in the ability of the study to detect smaller changes in worker numbers in response to the use of Neonicotinoid pesticides. Increasing the number of replicate blocks to 20 would allow the experiment to detect a 15 % drop in peak honey bee worker numbers and a 22% drop in total B. terrestris worker numbers. Subject to the support of the independent Scientific Steering Group, we would argue that this is a realistic and defensible level of detection. The cost of an additional seven replicate blocks undertaken in the UK (and with monitoring of only core variables) would be an additional on the existing budget.

Scenario 2b) Tier 2 experiment to quantify the role of supplementary floral resources to: i) reduce the impacts of bees exposed to neonicotinoids, and ii) enhance biodiversity supporting crop production There are currently insufficient resources in the project to address the secondary hypothesis relating to how the introduction of supplementary floral resources (flower rich field margins equivalent to 3- 5 % of the area of treated crop) may: i) influence effects of neonicotinoids on wild pollinator populations if they are identified under the Tier 1 experiment; and ii) enhance native farmland biodiversity, and specifically that associated with crop production. This could potentially include testing up to THREE ADDITIONAL treatment levels within a single replicate of the Tier 1 core experiment: • non neonicotinoids + supplementary resources • Clothianidin + supplementary floral resources • Thiamethoxam + supplementary resources

The cost of funding 20 replicates of a single Tier 2 treatment (e.g. Clothianidin + supplementary floral resources) would be

Note that Scenarios 2a and 2b are independent. Additional replication plus testing of the impact of supplementary floral resources would require complementary increases in funding.

5.3 Scenario 3: Funding from Defra

Initial discussions with the UK Department for Environment, Food and Rural Affairs (Defra) suggest they may be willing to contribute a small amount of funding to the project. A potential use of this funding could be to carry out additional measures of neonicotinoid effects on native pollinators (e.g. solitary bees) in the Tier 1 experiment in the UK. The current experimental design only considers changes in B. terrestris colony size and so the inclusion of additional sampling on other wild pollinators would considerably enhance the breath and relevance of the study. Such funding is not currently secured and would only apply to the UK.

5.4 Scenario 4: Funding from NERC

NERC (the parent body of CEH) has expressed an interest in funding the Tier 1 experiment to increase the scientific robustness of the study. Negotiations suggest NERC is interested in supporting additional replication to increase statistical power of the core Tier 1 experiment in the UK and additional monitoring of native pollinators (e.g. solitary bees) which are not in the scope of the current design. Again, such funding is not currently secured and would only apply to the UK. In addition, NERC may fund a pre-field scoping study using CEH existing, long-term national monitoring data to assess the impacts of neonicotinoids on domesticated and wild bees in the UK. This exploratory study will test whether wild pollinator declines are related to patterns of neonicotinoid use and will take into account distributional trends in bees prior to their widespread introduction in the UK. It will analyse national, long-term data sets held by the CEH Biological Records Centre (BRC) on the distribution and trends in oilseed rape crop pollinators together with data on the use of neonicotinoids from the Defra Pesticide Utilisation Survey (http://www.fera.defra.gov.uk /landUseSustainability/surveys/). This analysis will provide an important foundation for the large-scale field experiments that will allow us to understand the links between environmental factors and ecological mechanism underpinning the response of wild bees to the potential impacts of Neonicotinoid pesticides (Tier 1). This analysis will also provide base data to develop strategies for optimising pollinator abundance in the farmed environment (Tier 2).

6. The way forward

There are many possible outcomes to the various funding scenarios. However, CEH would ask Bayer and Syngenta to consider these two alternatives:

1) Assuming no additional funding from NERC/Defra then c.a. would support the additional replication of the priority Tier 1 experiment to increase levels of detection of neonicotinoid effects. Further funding would be required for the Tier 2 experiment; 2) Assuming additional funding from NERC (and possibly Defra) then c.a. would secure a limited Tier 2 experiment comprising one additional treatment with supplementary floral resources.

In our opinion this represents the best way of satisfying the requirements of the different funding partners and the most effective use of any additional resources available to the project.

Annex 1: Details of the statistical power analysis

A1.1 Preliminary power analysis based on Bombus terrestris colonies

Data source: Carvell, C., Rothery, P., Pywell, R. F., & Heard, M. S. (2008). Effects of resource availability and social parasite invasion on field colonies of Bombus terrestris. Ecological Entomology, 33(3), 321-327.

Summary: Data derived from 24 experimental colonies of the bumblebee Bombus terrestris established in association with oilseed rape fields on a 1000 ha arable farm at Hillesden, Buckingham, UK. The colonies were split between 6 replicate 75-90 ha parcels of land (4 colonies per parcel) managed under either cross-compliance, Entry Level Agri-Environment or Enhanced Entry Level Agri-Environment Schemes. The farm grows oilseed rape in rotation with winter wheat in large contiguous block under conventional intensive management. Note we selected colony data only from areas of the farm growing oilseed rape. All data colony was collected using the same methodologies. We use a fundamental measure of colony strength (maximum worker number) to assess overall mean colony size (μ=18.2) workers), variance between colonies within sites (σ2 = 62.6), and variance in mean worker number between sites (τ2= 6.4).

Power analysis.

We follow the power analysis provided by EFSA (2013) (Eq. 1). This analysis is designed to determine the required number of replicate pairs (N) of treated (with neonicotinoid) and control (no neonicotinoid) oilseed rape fields required to determine an effect size of 7 % (e.g. in B. terrestris colony maximum worker number) at a significance level of α=0.05 and a power of β=0.80. The analysis incorporates information on both within site variation in the colony population parameter (σ2) and between site variation (τ2). We rearrange this power equation and simulate for different numbers of possible replicate pairs (N = 5, 10, 15, 20, 30, 50, 75) of experimental control and treatment sites how the predictive power of an analysis would change (Fig.1). While we keep within site variation in colony strength fixed on the observed value described above (σ2 = 62.4), we vary between site variation in colony strength ± 100 % of its observed value (τ2= 6.4 ± 6.4). This is intended to provide an indication of how the predictive power of the analysis will vary under a range of possible between site differences in colony strength variance that may be encountered in the field. We repeat these simulations for two different numbers of within site replications of colonies (7 and 15 per site). While EFSA provide a recommendation that a 7 % effect size should be detectable at a significance of α=0.05 and with a power of β=0.80, we show how a larger and perhaps more biologically realistic level of detectable effect size of 15 % also responds to N and τ2. Note, that as the focal comparison of interest is between the control and either of the two neonicotinoid treatment (Clothianidin or Thiamethoxam) the inclusion of an additional treatment level does not change the underlying predictions of the number of replicate pairs required to achieve the described power.

Eq. 1. Where N = number of independent pairs of observation (treated and control sites); α = significance level

of test; zα = α quantile of standard normal distribution; β = power of test; zβ = β quantile of standard normal distribution; p = logarithmic treatment effect; σ2 = variation between sites; τ2 = variation between colonies within a site; n = number of colonies per field.

Number of B. terrestris colonies per field, n=7.

a) Detection of 7 % effect size b) Detection of 15 % effect size 1.00 1.00 N = 75

0.80 0.80 N=50

0.60 0.60 N=30

0.40 0.40 N=20 N=10

Power of test 0.20 0.20 N= 5 0.00 0.00 0 5 10 15 0 5 10 15

Between site variance B. terrestris cumulative worker number (τ2)

Number of B. terrestris colonies per field, n = 10

c) Detection of 7 % effect size d) Detection of 15 % effect size

N = 75 1.00 1.00 N=50 0.80 0.80

0.60 0.60 N=30

0.40 0.40 N=20

Power of test Power 0.20 0.20 N=10

0.00 0.00 N= 5 0 5 10 15 0 5 10 15

Between site variance B. terrestris cumulative worker number (τ2) Fig. 1. Simulations showing the predicted power of analyses attempting to identify both 7 % and 15 % effect sizes in B. terrestris colony strength (maximum worker numbers) between replicated control and treatment oilseed rape sites in response to between site variance in this parameter. Simulations are made for stated examples of number of sites (N) and different numbers of colonies within sites (n). The observed between site variation in colony strength is (τ2= 6.4) and is indicated by a vertical dashed line on each graph. Note, we simulate τ2 for ± 100 % of this value.

Summary: Due to large between colony variability in B. terrestris worker numbers detecting a 7 % change in this in response to the application of neonicotinoid seed treatments on oilseed rape at a significance level of α=0.05 and a power of β=0.80 would not be possible at the maximum

investigated number of replicate pairs of sites (N=75). This failure to reach the required 0.80 power is true where 7 or 10 colonies are present within an individual site (Fig. 1a, c). Only where the detectable effect size is increased to 15 % (Fig. c, d) does the number of replicate pairs of sites start to differentiate in terms of the detectable power of the analyses. This suggests 38 replicate pairs of sites would be required where 7 B. terrestris colonies are present at each site (or 31 replicate pairs of sites for 10 B. terrestris colonies).

Cumulative Number of hives Variance maximum worker Sites Treatment number Cross compliance 1 21.0 4 75.3 Cross compliance 3 18.5 4 15.0 Entry Level Scheme 1 17.3 4 101.5 Entry Level Scheme 2 17.3 4 40.9 Entry Level Scheme Extra 1 20.8 4 110.2 Entry Level Scheme Extra 3 14.3 4 15.0 Overall mean (μ) 18.2 Mean Variation between 62.7 colonies within a site (σ2) Between site variance 6.4 (τ2) From Carvell et al (2008) Data derived from 24 experimental colonies of the bumblebee Bombus terrestris established in association with oilseed rape fields on a 1000 ha arable farm at Hillesden

A1.2 Preliminary power analysis based on Honeybee colonies

Data source: Martin, S.J. and D. Kemp, Average number of reproductive cycles performed by the parasitic mite Varroa jacobsoni in Apis mellifera colonies. J. Apicultural Research, 1997. 36: p. 113-123.

Pilling, E., et al., A Four-Year Field Program Investigating Long-Term Effects of Repeated Exposure of Honey Bee Colonies to Flowering Crops Treated with Thiamethoxam. PLoS ONE, 2013. 8(10): p. e77193.

Summary: We use two different data sets to derive: 1) Mean colony size (μ ) and between site variance in maximum honeybee colony strength (τ2) based on data sets in Pilling et al. (2013). Note we selected data from sites linked with oilseed rape (Picardie & Alsace) as those associated with maize alone would have far greater variability in local foraging resources and so introduce unduly large variance to the data set; and 2) between colony variation within a site (σ2) based on data presented in Martin and Kemp (1997). In all cases colony strength reflects the number of workers in each hive, and is expressed in units of 1000 workers (e.g. × 103). Pilling et al. (2013) presents a simple paired experimental design of neonicotinoid treated crops (oilseed rape and maize) vs control treated fields. These pairs (treatment and control) are replicated in five regions in France (Maize: Avignon, Alsace and Loraine; Oilseed Rape: Picardie and Alsace). Each site was recorded in four years (2005 - 2008) and mean colony strength (without a measure of variance) based on six hives a field is presented. Data is derived directly from graphs where we determine the maximum mean number of observed workers at a site for each year. Mean colony size (μ = 21.73) and between site variance is determined for each year, with an average of this value values used in subsequent power analyses (τ2=7.71± 100%). In Martin and Kemp (1997) the variance in 15 colonies of honeybees kept

at a single site is recorded for seven sampling dates from May to September in 1996. We use this data to determine average within site variance in colony strength for honeybee worker number (σ2= 16.29). Again this data is derived directly from presented graphs.

Number of

Month Year Mean worker number (x 103) hives Variance June ear 1996 33.7 15 36.0 June late 1996 30.7 15 23.7 Jul ear 1996 35.1 15 7.70 Jul late 1996 35.0 15 15.8

Aug 1996 28.5 15 4.41 sep ear 1996 31.1 15 20.2 Sep late 1996 33.7 15 6.12

Average between colony variation (× 103 Bees) 16.2 Average colony size (× 103bees) 32.5 Betw een colony variation in worker bee numbers (× 103) within a site (σ2) based on data presented in Martin and Kemp (1997).

Crop Location Treatment Year Overall 2005 2006 2007 2008 Average 20.9 23.6 17.8 16.7 Oilseed Picardie Neonicotinoid 17.1 32.4 22.7 20.4 Oilseed Picardie Control 18.2 31.8 22.4 16.1 Oilseed Alsace Neonicotinoid 21.2 30.3 18.2 17.9 Oilseed Alsace Control 19.4 29.5 20.3 17.8 μ = 21.7 Mean 4.0 16.2 7.0 3.6 τ2=7.7 Variance Peak colony size (μ ) and between site variance (τ2) in maximum honeybee worker numbers (× 103) based on data sets in Pilling et al. (2013)

Number of Honeybee hives per field (n) = 5

a) Detection of 7 % effect size b) Detection of 15 % effect size 1.00 1.00 N = 75

N=50 0.80 0.80

0.60 0.60 N=30

0.40 0.40 N=20

N=10 Power of test 0.20 0.20 N= 5 0.00 0.00 0 5 10 15 0 5 10 15

Between site variance Honeybee worker number (τ2)

Number of Honeybee hives per field (n) = 10

c) Detection of 7 % effect size d) Detection of 15 % effect size

1.00 1.00 N = 75

0.80 0.80 N=50

0.60 0.60 N=30 0.40 0.40 N=20

Power of test Power 0.20 0.20 N=10

N= 5 0.00 0.00 0 5 10 15 0 5 10 15

Between site variance honeybee worker number (1000 workers) (± 100 %) Fig. 2. Simulations showing the predicted power of analyses attempting to identify both 7 % and 15 % effect sizes in honeybee colony strength (worker numbers). Simulations are made for example numbers of replicate pairs of control and treatment sites (N) and different numbers of colonies within sites (n). The observed between site variation in colony strength is (τ2= 7.71) and is indicated by a vertical dashed line on each graph. Note, we simulate τ2 for ± 100 % of this value.

Summary: Both high between and within site variation in colony strength for honeybees suggests that to detect a 7 % change in the number of worker bees in response to the application of neonicotinoid seed treatments at a significance level of α=0.05 and a power of β=0.80 would not be possible for the maximum investigated number of replicate pairs of sites (N=75). This failure to reach the required 0.80 power is true where 5 or 10 hives are present within an individual site (Fig. 2a, c). Only where the detectable effect size is increased to 15 % (Fig. c, d) does the number of replicate pairs of required sites start to differentiate in terms of the detectable power. Where 5 12

hives are located at a site, c. 19 - 20 replicate pairs of sites. In the context of the experimental design proposed (replicate blocks of Control, Clothianidin and Thiamethoxam treated sites) this would also equate to c. 19 - 20 replicate blocks to detect differences at α=0.05.

References

EFSA (European Food Safety authority) (2013) Guidance on the risk assessment of plant protection products on bees (Apis mellifera, Bombus spp. and solitary bees). EFSA Journal 2013; 11(7):3295, 266pp.

Annex 2: Detailed costing of experimental activities

1. Project cost summary Cost (£) Total budget ($ 3 million) Total buget - less overall project costs (2. below) Per site cost (see 3. below) Per replicate block cost (each replicate block is composed of three sites: Control, Clothianidin & Thiamethoxam)

No. replicate blocks (three sites: control, Clothianidin & Thiamethoxam)

2. Overall project management Staff Band B7 B6 B5 B4 Locating sites, design Data management Spatial data analysis Data analysis Project meetings Project management, protocols, QA Reporting

Total days Daily cost (£) including travel Total cost

Grand total

3. Site costs Consumables Time requirements

Tier 1 (per year) Cost per unit (£) unit per Cost Dayssampling of effort

Activity TOTAL consumable cost (£) Seed Area OSR sown (ha) 50

Apis hive management Apis Hives per site 6 10

Cost of transferring hives to and from site (?)

Exposure tents (NNI concentrations in crop) Exposure tents (tent cost) 1 9 Nectar NNI chemical (caged Apis) 2 Pollen NNI chemical (caged Apis) 2

Honeybee hive monitoring Dead bee traps (3 x counts) 5 1

Liebefeld method (3 x monitoring OSR flowering) 4 Liberfeld method (2 x monitroing pre & post overwinter for overwintering mortality) 2

Colony weight sensors 5 2

Nectar / pollen / wax chemical analsysis (2 occasions) 30 0

Bee activity: Bee counter (still checking into this. Expensive so 1 per site) 1 1

Bombus terrestris hives Bombus colonies per site (7 per site) 7 3 Protective housing 7

Initial weighing of colonies and checks 1

4 week field check and weighing of colonies 1

Processing time for frozen colonies (counts of workers, reproductive, pollen nectar samples etc) 5 Nectar / pollen / wax chemical analsysis (end of season) 21

Behavioural transects 2 transects: Sampled on 3 ocasions a site 2

Additional material for chemical anaysis (Archive) Time for collection in field 1 Time for subsequent storage and processing 1

Additional equipment / postage etc for colonies

Other staff costs Farmer liaison 2 Collection agronomy data 2 Rapid landscape assessment (cropping pattern, buffering etc) 3 Data input (database & digitising) 2

From: Pywell, Richard F. Sent: 28 January 2014 18:18 To: @syngenta.com'; Cc: @SYNGENTA.COM'; @bayer.com Subject: FW: neonicotinoids and bees: an external request for an appointment

Dear ,

Please see attached message sent to EFSA today requesting a meeting.

Best wishes

Richard

From: Bullock, James M. Sent: 28 January 2014 17:58 To: Cc: Pywell, Richard F.; Bullock, James M. Subject: RE: neonicotinoids and bees: an external request for an appointment

Dear Domenica,

As you will remember from our informal discussion on 20th November 2013, the UK’s Centre for Ecology & Hydrology is currently undertaking a large scale field study investigating the impacts of two neonicotinoid seed dressings (‘Clothianidin’ Bayer CropScience and ‘Thiamethoxam’ Syngenta) on honeybees and bumblebees foraging on oilseed rape. Whilst this study is funded by Syngenta and Bayer, and the UK Natural Environmental Research Council, it will be independent with the results and methodologies being audited by an independent scientific advisory group.

The first part of the study has involved a detailed assessment of feasibility and cost of the project largely based on the guidance produced by EFSA (EFSA, 2013. Guidance on the risk assessment of plant protection products on bees. EFSA Journal 11:3295, 266). Key to this has been a power analysis following EFSA guidelines to identify the level of replication of the experimental treatments required to determine a 7 % effect size for the impact of neonicotinoids on bee populations.

We would welcome the opportunity to present the findings of the feasibility study to the EFSA Panel on Plant Protection Products and their Residues. We would particularly welcome their feedback on various aspects of the proposed experimental design and monitoring protocols. This would help us to ensure that our results will be useful to EFSA.

We understand that any discussion would be informal and would not be interpreted as EFSA approval or endorsement of the experiment. EFSA would ultimately assess our findings independently.

Ideally we would like to implement the field study in the summer of 2014 so we would prefer to meet the PPR Panel at the next available opportunity; possibly at the next plenary meeting?

Yours sincerely

James Bullock and Richard Pywell

************************* Prof James Bullock Centre for Ecology & Hydrology Benson Lane Wallingford OX10 8BB UK Tel: +44 7824 460866

From: Bullock, James M. Sent: 29 November 2013 10:15 To: Cc: Pywell, Richard F. Subject: RE: neonicotinoids and bees: an external request for an appointment

Dear , it was good to meet you and your colleagues last week. I will follow up with your suggestion to contact about representation on our project advisory group (e.g. from the PPR panel). First we will meet wuith the chair of our advisory group - - to discuss ToR. He will invite individuals on to the advisory group.

Best wishes, James

************************* Prof James Bullock Centre for Ecology & Hydrology Benson Lane Wallingford OX10 8BB UK Tel: +44 7824 460866

From: Sent: 18 November 2013 09:58 To: Bullock, James M. Cc: Subject: neonicotinoids and bees: an external request for an appointment

Dear James Concerning your request, I got a positive reply from my colleagues of PRAS unit for a chat on the topic I put you in contact with one of them, (reading in CC) in order not to spam too many people

After lunch, you will be located with your group in the meeting room of the first floor (FMT SEAT 01/B15 (8p), where we could discuss together the received feedbacks. or/and her colleagues can jump directly to the room or call in order to organize a coffee for an informal chat when most convenient to both of you.

I hope it helps

Have a nice day

Plant Health Unit European Food Safety Authority - EFSA Via Carlo Magno 1/a, 43126 Parma - Italy

http: //www.efsa.europa.eu

Save the planet: please don't print this message

From: Pywell, Richard F. Sent: 11 February 2014 10:34 To: @syngenta.com'; Cc: @SYNGENTA.COM'; ; Bullock, James M.; ; Hails, Rosemary S.; Shore, Richard F.; Subject: FW: neonicotinoids and bees: an external request for an appointment

Dear ,

Below is a reply from from EFSA regarding the peer review of the experimental design/protocol.

Clearly if you would like EFSA to undertake a peer review then your companies will need to submit the protocols via the procedure agreed between the manufacturers and EFSA.

Best wishes

Richard

':01491 692356

*: [email protected]

From: efsa.europa.eu] Sent: 10 February 2014 16:18 To: Pywell, Richard F. Cc: Bullock, James M.; Hails, Rosemary S.; Subject: RE: neonicotinoids and bees: an external request for an appointment

Dear Richard, for the point number 1, it is fine for us.

For the point number 2, I can confirm the procedure – but please note that the applicants only need to submit the protocols by the 28th February, if they wish for them to be peer reviewed. The applicants are not legally requested to submit any protocols.

Best regards, Domenica

From: Pywell, Richard F. [mailto:[email protected]] Sent: 10 February 2014 15:50 To: Cc: Bullock, James M.; ; Hails, Rosemary S. Subject: RE: neonicotinoids and bees: an external request for an appointment

Dear

Thank you for your response.

1) Power analysis We think that the most efficient way to deal with the relatively minor technical issues around the power analysis is to submit them to you in writing in the first instance rather than travel to EFSA on 25 February. We will be in a position to send a letter to you in the next day or so. If we are unable to find solutions using this approach then we would be happy to meet face to face to discuss.

We apologise for any inconvenience caused by this.

2) Review of the experimental design/protocol for the NNI field experiment Please could you confirm our understanding of the review procedure is correct:

The sponsors of the study (Bayer/Syngenta) need to submit the detailed experimental protocol to the designated Rapporteur Member State (RMS) by 28 February. A peer review of the protocol will then be organised by EFSA followed by an MS experts’ meeting for discussing the critical issues and to make recommendations in April.

Best wishes

Richard

':01491 692356

*: [email protected]

______

terial supplied to NERC may be stored in an electronic records management system. From: Pywell, Richard F. Sent: 25 February 2014 14:38 To: @syngenta.com' Cc: @SYNGENTA.COM' Subject: Recent pesticide conference

:01491 692356

: [email protected]

______

The Impact of Pesticides on Bee Health

Joint meeting of the 22 – 24 January 2014 British Ecological Society, Biochemical Society Charles Darwin House, London, UK and the Society for Experimental Biology The British Ecological Society was established in 1913, making it the oldest ecological learned society in the world. We have a membership of 5,000 across 80 countries, publish ﬁve world renowned journals, organise a wide portfolio of events, fund numerous grants and have far-reaching education and policy schemes. Members are our lifeblood and we encourage them to be active in their Society, e.g. by joining a Special Interest Group, commenting on consultations, proposing events, joining a committee, attending our meetings. Email: [email protected] Twitter: @BritishEcolSoc

The Biochemical Society exists for the advancement of the molecular and cellular biosciences, both as an academic discipline and to promote its impact on areas of science including biotechnology, agriculture, and medicine. Biochemistry helps to play a key role in tackling global issues such as improving lifelong health, treatment of disease, biotechnology and food security. We achieve our mission though our publications and journals, scientiﬁc meetings, educational activities, policy work, awards and grants to scientists and students.

The Society for Experimental Biology (SEB) believes that the broad nature and lack of “ology” boundaries implicit in Experimental Biology give it a pivotal role in the development of Life Sciences which are of considerable beneﬁt to its members and to society. In particular, Experimental Biology contributes to knowledge that can be applied to the development of agriculture and medicine and to understanding the impacts of human activity on living organisms and ecosystems.

2 Contents

Society Information ______2

Welcome and Introduction ______4

Presentation Programme and Talks Wednesday 22 January ______5 Thursday 23 January ______12 Friday 24 January ______20

Poster Talks Wednesday 22 January ______8 Thursday 23 January ______16

Author Index ______22

Delegate Information ______23

Facilties Information ______26

Useful Contacts ______27

Notes ______28

3 Welcome! Introduction

We are delighted to welcome you to Charles Darwin We extend a warm welcome to the third joint meeting House for the “The Impact of Pesticides on Bee Health” of the Biochemical Society, the British Ecological – the third joint meeting between the three societies. Society and the Society for Experimental Biology. This is a very timely meeting, given the recent The Biochemical Society, the British Ecological implementation of the EU moratorium on the specific Society and the Society for Experimental Biology have application of neonicotinoids to bee-friendly crops. I a long history of organizing and supporting meetings am sure that the benefits, shortfalls and risks of this on all aspects of biochemistry, cell biology and policy will be discussed at the meeting. ecology. However, the separation of these disciplines is becoming increasingly blurred as techniques and The aims of the meeting are challenging as we ideas spread more widely; all are critical if we are to attempt to present and discuss the evidence for, succeed in understanding real, complex, dynamic, and against, the use of pesticides and in particular biological and ecological systems. the neonicotinoids. A particular challenge is the disparate evidence obtained from laboratory and field We believe this joint meeting between the societies studies. In the laboratory, conditions can be highly (to be held at the joint headquarters of the three controlled and confidence in the result strong, but this societies) will succeed in bringing together experts is produced under artificial conditions that may not from around the world, who are solving biological reflect the real environment. In contrast, field studies problems using multiple integrated approaches. We are poorly controlled and many other environmental hope that this meeting will be a tremendous success factors can influence the findings. However, they do and form part of many future meetings between our represent an example of a real environment where three societies. conditions may mimic those encountered in real life. The Biochemical Society, the British Ecological Together, these studies may identify some common Society and the Society for Experimental Biology. ground, improvements for future experimentation and key knowledge gaps. Understanding the impact of pesticides on insect pollinators is a perfect example of the need for multidisciplinary approaches and a wide knowledge base that is offered by such joint society meetings and this is reflected in the diverse backgrounds of all participants (speakers and attendees). Importantly, this includes scientists from academia, industry and government (DEFRA) and covers a range of different scientific evidence and opinions. It is a great credit to all participants that you are here to engage in polite scientific debate on a complex and highly topical issue. The contribution that you will all make is of enormous benefit, not just to our scientific understanding, but also that of our policymakers, stakeholders and the general public. We hope that you will all enjoy this meeting and the opportunities it offers to develop your own ideas and make valuable contacts. Christopher Connolly (University of Dundee, UK) and Geraldine Wright (Newcastle University, UK).

4 Programme & Talks: Wednesday 22 January Chair: Chris Connolly 11:00 Registration, with lunch from 12:00 B14.1 13:00 Mr Randolf Menzel 13:00 Neonicotinoids interfere with navigation in honeybees Randolf Menzel (Free University of Berlin, Germany), (Free University of Berlin, Germany) Johannes Fischer (Institut für Bienenkunde Polytechnische Neonicotinoids interfere with navigation in Gesellschaft, Germany), Bernd Grünewald (Institut für honeybees Bienenkunde Polytechnische Gesellschaft, Germany), Teresa [B14.1] Müller (Institut für Bienenkunde Polytechnische Gesellschaft, Germany), Uwe Greggers (Freie Universität Berlin, Germany) 13:30 Prof Dave Goulson Three neonicotinoids, imidacloprid, clothianidin and (University of Sussex, United Kingdom) thiacloprid, agonists of the nicotinic acetylcholine receptor in Impacts of pesticides on bumblebee colonies the central brain of insects, were applied at sublethal doses in order to test their effects on honeybee navigation. A catch- [B14.2] and-release experimental design was applied in which feeder 14:00 Dr Nigel E Raine trained bees were caught when arriving at the feeder, treated with one of the neonicotinoids, and released 1.5 hours later (Royal Holloway University of London, United at a remote site. The flight paths of individual bees were Kingdom) tracked with harmonic radar. The initial flight phase controlled The impacts of pesticides on bumblebees: from by the recently acquired navigation memory (vector memory) individual behaviour to colony function was less compromised than the second phase that leads the [B14.3] animal back to the hive (homing flight). The rate of successful return was significantly lower in treated bees, the probability 14:30 Refreshment Break of a correct turn at a salient landscape structure was reduced, and less directed flights during homing flights were 15:00 Dr G. Christopher Cutler performed. Since the homing phase in catch-and-release (Dalhousie University, Canada) experiments documents the ability of a foraging honeybee Semi-field and field studies examining impacts of to activate a remote memory acquired during its exploratory neonicotinoid insecticides on bee colony health orientation flights, we conclude that sublethal doses of the three neonicotinoids tested either block the retrieval of [B14.4] exploratory navigation memory or alter this form of navigation 15:30 Mr Galen P Dively memory. These findings are discussed in the context of the application of neonicotinoids in plant protection. (University of Maryland, United States) Email: [email protected] Field exposure, in-hive fate and impact of imidacloprid on honey bee colony health B14.2 [B14.5] 13:30 Impacts of pesticides on bumblebee colonies Dave Goulson (University of Sussex, United Kingdom) 16:00 The causes of pollinator decline, and in particular the relative (Syngenta, United Kingdom) contribution of pesticides in driving these declines, remains A four year field program investigating long term controversial. Here I will provide a review of the evidence, effects of repeated exposure of honey bee colonies with a particular focus on neonicotinoids and their impacts on bumblebee colonies. A coherent and consistent body of to flowering crops treated with thiamethoxam research demonstrates that field-realistic exposure is likely to [B14.6] be causing significant harm to wild bumblebee populations. The prophylactic use of these broad-spectrum pesticides 16:30 Mr Olivier Samson-Robert goes against the long-established principles of Integrated (Laval University Horticulture Research Centre, Pest Management (IPM), leading to environmental concerns Canada that go beyond bees. I will discuss these broader issues Neonicotinoid-coated seeds: impacts on honey relating to persistence of neonicotinoids in the environment, bees, bumble bees and water contamination. their accumulation in soils and their occurrence in waterways and non-target vegetation. [B14.7] Email: [email protected] 17:00 Posters and Wine Reception 18:00 Buffet Dinner 19:00 Pat Willmer (University of St Andrews, United Kingdom) A wider perspective: Conserving and supporting all the pollinators. [B14.8]

20:00 End of Session

5 Talks: Wednesday 22 January

B14.3 B14.5 14:00 The impacts of pesticides on bumblebees: from individual 15:30 Field Exposure, in-hive fate and impact of imidacloprid on behaviour to colony function honey bee colony health Nigel E Raine (Royal Holloway University of London, United Galen P Dively (University of Maryland, United States Kingdom), John Bryden (Royal Holloway University of Imidacloprid is widely used on many pollinated agricultural London, United Kingdom), Richard J Gill (Royal Holloway crops, and increasing evidence indicates that it moves to University of London Imperial College, United Kingdom), some extent into pollen and nectar. Three studies reported Vincent A A Jansen (Royal Holloway University of London, here characterized exposure and addressed the risk United Kingdom), Dara A Stanley (Royal Holloway University assessment of imidacloprid on colony health. One study of London, United Kingdom) showed that potential exposure of neonicotinoid residues in Bumblebees are essential pollinators of many important pollen and nectar to pollinators depends on environmental agricultural crops and wild plants. While foraging in farmed conditions and the method and timing of application relative landscapes bees are likely to be exposed to pesticides, and to flowering. Another study tracked the movement and other agrochemicals, applied for crop protection. Although degradation of imidacloprid within whole colonies exposed bees typically encounter these pesticides at sublethal levels, to known levels of imidacloprid residues entering the hive, exposure can still have impacts on individual behaviour and quantified the actual exposure dose to worker bees, such as foraging or navigation. As social insects, bumblebee brood and the queen via honey, brood food and honey jelly. A colonies depend on the collective performance of many third study examined the chronic sublethal effects on whole individual workers. Therefore, while field-level pesticide honey bee colonies fed supplemental pollen diet containing exposure may have relatively subtle effects on individual imidacloprid at field relevant doses for 12 weeks. Various behaviour, it is poorly understood whether bee societies can endpoints of colony performance and foraging behavior were buffer such effects or whether they result in severe cumulative measured during and after exposure, including winter survival. effects at the colony level. We have shown that chronic Email: [email protected] exposure to both neonicotinoid and pyrethroid pesticides B14.6 have significant impacts on bumblebees (Bombus terrestris) at field-relevant levels, including reduced individual foraging 16:00 A four year field program investigating long term effects efficiency and colony growth rate. Combined exposure to both of repeated exposure of honey bee colonies to flowering pesticides significantly increased chances of colony failure. crops treated with thiamethoxam We have modelled pesticide stress on individual bees, which (Syngenta, United Kingdom), Ed Pilling impairs colony function, and shown how some colonies fail (JSC International, United Kingdom), while others thrive. Results from our computer models closely (Syngenta, United Kingdom), Natalie Ruddle (Syngenta, predict the dynamics of colony failure from our experiments. United Kingdom), (Eurofins, Germany) This suggests our model can explain the enigmatic aspects of Neonicotinoid residues in nectar and pollen from crop plants bee colony failures, highlighting an important role for sublethal have been implicated as one of the potential factors causing stress in colony declines. the declines of honey bee populations. Indeed the European Bryden J, Gill RJ, Mitton RAA, Raine NE, Jansen VAA (2013). Commission has introduced a 2 year moratorium for the use Chronic sublethal stress causes bee colony failure. Ecology of Imidacloprid, Thiamethoxam and Clothianadin on selected Letters 16: 1463-1469. bee attractive crops. However, much of the data that has Gill RJ, Ramos-Rodriguez O, Raine NE (2012). Combined implicated neonicotinoids in the decline of honey bee health pesticide exposure severely affects individual- and colony- has been generated either under laboratory conditions or level traits in bees. Nature 491: 105-108. have used unrealistic exposure conditions. In this study, Email: [email protected] conducted under field conditions, the long-term risk to honey bee colonies was investigated following four years consecutive B14.4 single treatment crop exposures to flowering maize and 15:00 Semi-field and field studies examining impacts of oilseed rape grown from thiamethoxam treated seeds at rates neonicotinoid insecticides on bee colony health recommended for insect control. During the study honey bee G. Christopher Cutler (Dalhousie University, Canada) mortality, foraging behavior, colony strength, colony weight, Neonicotinoid insecticides are widely used plant-systemic brood development, food storage levels and over wintering compounds that contain the active ingredients imidacloprid, success are monitored and reported. The results confirm a thiamethoxam, and clothianidin. This class of insecticide low risk to honey bees from systemic residues in nectar and has perhaps been subject to more scrutiny than any other pollen following the use of thiamethoxam as a seed treatment potential cause of recent honey bee and wild pollinator on oilseed rape and maize. These results contribute towards declines. Laboratory-based studies have indeed shown reducing the gap in our understanding of exposure and risk to that neonicotinoids may elicit various acute or chronic honey bees from the use of neonicotinoids as seed treatments effects on bees. However, higher-tier studies, where dietary under field conditions. exposure to pollen and nectar occurs through plants grown Email: @syngenta.com from soil or seed-treatment applications, have failed to demonstrate significant effects. In this talk, I will describe results from semi-field and field studies we have conducted with honey bees and bumble bees that suggest exposure to neonicotinoid-treated crops has no significant effect on colony health. I will frame these results within the context of other semi-field and field studies, and other observations and analyses regarding declines and health of pollinators over the past few decades. Email: [email protected]

6 Talks: Wednesday 22 January

B14.7 16:30 Neonicotinoid-coated seeds: impacts on honey bees, bumble bees and water contamination. Olivier Samson-Robert (Laval University Horticulture Research Centre, Canada), Geneviève Labrie (CÉROM Grain Research Centre, Canada), Madeleine Chagnon (UQAM Department of Biological Sciences, Canada), Valérie Fournier (Laval University Horticulture Research Centre, Canada) Multiple paths of exposure to neonicotinoid insecticides have been identified for pollinators. Contacts with sowing dust and collection or consumption of contaminated pollen, nectar and water are the most common. Neonicotinoid- coated seeds are ubiquitous, yet we understand very little of their impacts on pollinators in natural conditions. In this 2-year study, mortality levels of commercial apiaries (Apis mellifera) were monitored during the corn planting period. Exposed sites were located within 500 meters of fields with coated seeds whereas control sites were 3 km distant. Liquid chromatography tandem mass spectrometry was used to analyse samples of dead honey bees (N=70) and puddle water from nearby fields (N=74). In addition, one colony of bumble bees (Bombus impatiens) was installed at each study site and foragers were analyzed with quantitative PCR (N=82) to determine the expression level of acethylcholinesterase. Results showed a significant honey bee mortality increase in sites exposed to coated seeds. All water samples from exposed sites had neonicotinoid compound while all control sites’ samples were clear. Clothianidin and thiamethoxam concentrations found in water samples collected one month following planting were well above honey bee lethal doses. Finally, bumble bees from exposed sites showed a significant increase in acetylcholinesterase expression. These findings provide evidence that extensive use of neonicotinoid-coated seeds exacerbates honey bee mortality levels and affects bumble bee neuronal activity. Finally, this is the first study to identify puddle water as a novel route of potential exposure to neonicotinoids for pollinators. Email: [email protected] B14.8 19:00 A wider perspective: Conserving and supporting all the pollinators. Pat Willmer (University of St Andrews, United Kingdom) Most of the key pollinators worldwide have been declining, as ecologists have known for at least 30 years; the recent explosion of interest in honeybee decline has at last initiated a wider discussion of consequences, and what can or should be done. But we must address all the potential causes, in context, and avoid measures that might affect one kind of animal-plant interaction, or one pollinator group, at the expense of others. Honeybees are terrific generalist flower visitors, and give a managed pollination service to key food crops worldwide. But they are often less efficient than the natives (especially bumblebees, other bees and hoverflies), with which they may compete; and they cannot pollinate some crops at all. So there is a somewhat controversial case to be made that, whatever the causes, some reductions in introduced honeybee numbers may be no bad thing. But if - AND ONLY IF - we also build on excellent recent research and take measures to enhance all the native pollinator communities, in both agricultural and more urban settings. Supporting current generations of both farmers and beekeepers of course matters, and we certainly cannot afford to add to the honeybees’ and bumblebees’ ongoing problems (on which this conference focusses). But let’s grasp a longer perspective: take steps to conserve all the bees, and all the other pollinators, for the benefit of future biodiversity and our own food security.

7 Poster Session: Wednesday 22 January

B14.25 Little and often makes much: testing for persistence of B14.29 Could early-stage honey storage behaviour influence dietary pesticides in bees. pesticide spread in Apis mellifera L. colonies? Philippa J Holder (University of Exeter, United Kingdom), Mark K Greco (University of Bath, United Kingdom), Edward James E Cresswell (University of Exeter, United Kingdom) Feil (University of Bath, United Kingdom), Nicholas Oriest Bumblebee and honey bee populations are exposed to (University of Bath, United Kingdom) pesticide residues in the nectar and pollen of treated Decision making in honeybees is based on information flowering crops on which they feed. If a pesticide is which is acquired and processed in order to make choices persistent, even small, sublethal residues could build up in between two or more alternatives. These choices lead to a bee’s body over time and be detrimental to their health. the expression of optimal behaviour strategies such as The negative impacts of persistent pesticide residues floral constancy. Optimal foraging strategies such as floral could threaten pollination services to wildflowers and constancy improve a colony’s chances of survival however crops. Using a proposed new EU protocol, three pesticides to our knowledge there has been no research on decision found in nectar and pollen, two of which are included in making based on optimal storage strategies. Here we show, the recent EU 2-year ban for use on bee-attractive crops, using diagnostic radioentomology, that decision making in were tested for their persistence. Thiamethoxam and storer bees is influenced by nectar sugar concentrations cypermethrin were found to act as non-persistent toxicants and that, within 48 hours of collection, honeybee workers in both bumblebees and honeybees, whereas fipronil store carbohydrates in groups of cells with similar sugar was persistent in both species. Current EU regulations concentrations in a non-random way. This behaviour, as for pesticide risk assessment for bees do not include evidenced by patchy spatial cell distributions, would by persistence testing. The findings presented here suggest default separate, nectar pesticides in the hive via cells of that the protocol adopted would be a useful addition in the similar sugar concentrations. Thus, colonies which exhibit EU’s regulatory process for pesticide approval. optimal storage strategies such as these would have an Email: [email protected] evolutionary advantage and improved colony survival expectations over less efficient colonies and it should be B14.26 Drivers of honeybee population change: An evaluation by plausible to select colonies that exhibit these preferred traits. Hill’s causality criteria Email: [email protected] Emma L Wright (University of Exeter, United Kingdom), (Syngenta, United Kingdom), James E Cresswell B14.30 Bumblebees are not deterred by nectar secondary (University of Exeter, United Kingdom) compounds Whilst globally numbers are increasing, the number of Erin Jo Tiedeken (Trinity College Dublin, Ireland), Jane C managed honey bee colonies has fallen in the USA and Stout (Trinity College Dublin, Ireland), Philip C Stevenson parts of Europe with annual losses estimated at around (University of Greenwich, United Kingdom), Geraldine A 30% in recent years compared to previous rates of 5-10%. Wright (Newcastle University, United Kingdom) Various factors could be responsible including pesticides, Bees visit flowers to collect nectar and pollen that contain pathogens, and forage availability. We used Hill’s causality nutrients and simultaneously facilitate plant sexual criteria to evaluate the importance of some of the most reproduction. Paradoxically, nectar produced to attract commonly identified factors as potential drivers of change pollinators often contains deterrent or toxic compounds. in regional stocks of managed honey bees. This analytical “Toxic nectar” can arise via two processes: 1.) systemic method was first designed to determine whether there was pesticides applied as seed dressings leach into nectar sufficient evidence to link certain detrimental agents, like or 2.) natural plant compounds usually thought to cigarettes, to human disease. Hill identified eight kinds deter herbivores are produced in nectar. The functional of non-experimental evidence, which he segregated into significance of both natural and synthetic nectar toxins is ‘criteria’. For each criterion, we use evidence from the not fully understood, but they may have a negative impact scientific literature to assign a level of conviction to the on pollinator behaviour and health, and ultimately plant proposition that a specified factor is a driver of change in pollination. This study investigates whether a generalist honey bee stocks. We conclude that the strongest driver bumblebee, Bombus terrestris, can detect field realistic of honey bee stock change examined so far is market concentrations of potentially deterrent nectar toxins. forces. If the cost of bee keeping is low and profit from Using paired-choice experiments, we identified deterrence selling products or pollination is high stocks can increase. thresholds for five natural plant compounds (quinine, The strongest driver for colony declines so far is varoosis, caffeine, nicotine, amygdalin, and grayanotoxins) and one the combination of varroa mites and viruses. This agrees systemic neonicotinoid pesticide (imidacloprid) found in the with the results of many other studies and the views of nectar of bee-pollinated plants. The deterrence threshold bee keepers. Dietary neonicotinoids scored negatively was determined when bumblebees significantly preferred suggesting that, from the existing evidence, this is not a a sucrose solution over a sucrose solution containing driver of honey bee declines. the compound. Bumblebees had the lowest deterrence Email: [email protected] threshold for the neonicotinoid imidacloprid, (0.001mM); all other compounds had higher deterrence thresholds, above the natural concentration range in floral nectar. Our data combined with previous work using honeybees indicate that generalist bee species have poor acuity for the detection of nectar toxins. The fact that bees cannot detect nectar relevant concentrations of naturally-produced compounds suggests it is difficult for them to learn to avoid flowers presenting toxic nectar, maintaining this trait in plant populations and exposing bees to potential sub-lethal effects of toxins. Email: [email protected]

8 Poster Session: Wednesday 22 January

B14.32 The Varroa destructor RDL receptor pore lining region: a B14.34 Distribution scenario of a xenobiotic inside a honeybee novel target for bee-friendly acaricides? colony Sarah Lummis (University of Cambridge, United Kingdom), Ulrike Riessberger-Gallé (Institute of Zoology University Kerry Price (University of Cambridge, United Kingdom), Mona of Graz, Austria), Wolfgang Schuehly (Institute of Zoology Alqazzaz (University of Cambridge, United Kingdom), Martin Graz, Austria), Javier Hernández-López (Institute of Zoology Hailstone (University of Cambridge, United Kingdom) University of Graz, Austria), Jutta Vollmann (Institute of The parasitic mite Varroa destructor is the major pest Zoology University of Graz, Austria), Karl Crailsheim (Institute of the honeybee Apis mellifera. Chemical methods for of Zoology University of Graz, Austria) its control have targeted members of the pentameric A xenobiotic is a chemical compound which can be found in (Cys-loop) ligand gated ion channel (pLGIC) superfamily, an organism but which is not normally produced or expected however the Varroa pLGIC family has not been to be present in it. The study aimed at visualizing the internal characterized. Here we report our studies on the Varroa distribution pattern of indigo carmine within a honeybee ortholog of the insect GABA-gated RDL receptor. We colony as an example of a distribution scenario for any water- have identified four differences in the amino acid residues soluble substance, e.g., a pesticide. Food containing the blue that constitute the pore lining (M2) region of Varroa water-soluble dye indigo carmine was presented to a nucleus RDL when compared to those in Drosophila and other colony ad libitum. The status of cells was visually evaluated insects. We have investigated these by replacement after 12 and 24 h. After 24 h, the colony was transferred to of the corresponding Drosophila RDL residues, new wax foundations and the color of the newly formed cells expression of these receptors in Xenopus oocytes, and was evaluated after 4 d. Already 12 h after feeding, indigo their characterization by two-electrode voltage clamp carmine could be detected in honey, pollen and brood cells. electrophysiology. We find that N-7’H, T6’M, A20’S The majority of dye was stored in open honey cells. Even and A21’Q mutants have lower GABA EC50 s than WT after 4 d on new comb foundations, indigo carmine could be receptors, and T6’M mutants are resistant to picrotoxin. detected in all newly built honey cells, which demonstrates The T6’ residue has previously been shown to be important the prevalence of the compound in the social stomach. For for the binding of many insecticides, and a mutation the experiments, a relatively high dose (c. 300 mg/colony) of here (T6’L) has been found in the southern cattle tick marker substance was used. Our results indicate that, upon (Rhipicephalus microplus) where it is associated with scaling down, the distribution pattern of very small amounts resistance to the cyclodiene insecticides (Hope et al. will be similar. Therefore, a hydrophilic xenobiotic will reach 2010). Further characterization of the Varroa RDL receptor all cell categories and will stay with the colony even when the may reveal new approaches towards the development of colony leaves the nest site. selective acaricides. Email: [email protected] Hope, M, Menzies, M, Kemp, D (2010) J. Econ. Entomol. 103 B14.35 Thiacloprid levels in Brassica rapa pollen samples in a (4) 1355-9. Finnish field study Email: [email protected] Mari Kukkola (Finnish Food Safety Authority Evira, Finland), B14.33 Double whammy? Impact of pesticides and parasites on Kati S Hakala (Finnish Food Safety Authority Evira, Finland), bumblebee fitness Lauri Ruottinen (Agrifood Research Finland MTT, Finland), Gemma Baron (Royal Holloway University of London, United Sakari Raiskio (Agrifood Research Finland MTT, Finland), Kingdom), Nigel E Raine (Royal Holloway University of Jarmo Ketola (Agrifood Research Finland MTT, Finland), London, United Kingdom), Mark J F Brown (Royal Holloway Kimmo Peltonen (Finnish Food Safety Authority Evira, Finland) University of London, United Kingdom) Thiacloprid is a systemic insecticide used in Finnish turnip Bees foraging in agricultural landscapes are exposed to rape (Brassica rapa subsp. oleifera) cultivation. It differs numerous stressors such as pesticides and parasites, that from most of the other neonicotinoid compounds as it is can negatively impact individual and colony-level fitness. applied as a spray treatment from budding to blooming Pyrethroid pesticides are widely used throughout Europe, periods. During blooming of the turnip rape, honey bees are and the use of these compounds has nearly doubled exposed to thiacloprid residues. Thiacloprid is less toxic since the 1990’s. Here we present data on the impacts of than some other neonicotinoids e.g. clothianidin. However, a commonly used pyrethroid, -cyhalothrin, on the colony due to the application period and use levels it may have an success of a widespread and abundant native pollinator, the adverse impact on honey bees. The aim of this study was to bumblebee Bombus terrestris, under laboratory conditions. develop an analytical method for residue determination and In order to fully understand the impacts of pesticides on to investigate thiacloprid residues in pollen samples. Samples bees, it is vital to consider interactions with other stressors were collected from hives nearby turnip rape fields treated as well as individual effects. We investigated the effect of with 0,3-0,4 l/ha thiacloprid (240 g/l). The sample preparation pesticide exposure on the susceptibility of worker bees was based on modifications of QuEChERS method (Quick, to a prevalent trypanosome parasite, Crithidia bombi, and Easy, Cheap, Effective, Rugged, Safe). Analysis was the combined impacts of pesticide and parasite exposure performed by ultra performance liquid chromatography- on individual worker survival. Under laboratory conditions, tandem mass spectrometry (UPLC-MS/MS). The determined -cyhalothrin affects an important aspect of colony function, thiacloprid levels will be presented and the suitability of worker size, which may have important implications under the method will be discussed. The results of this study can field conditions. However, in our study, colonies were able to be exploited in assessing the exposure of honey bees to compensate for this effect, and no significant impact on the thiacloprid residues in Finnish turnip rape production. production of sexual offspring was found. Additionally, no Email: mari.kukkola@evira.fi effect of pesticide exposure was found on worker infection rates or survival, even when bees were also challenged with the parasite. In light of the recent policy discussion on the use of pesticides, we discuss the implications of our study for the management of a key group of wild pollinators. Email: [email protected]

9 Poster Session: Wednesday 22 January

B14.36 Are bee colonies affected by a chronic exposure to sub- B14.37 A sublethal dose of the neonicotinoid thiacloprid may lethal doses of thiacloprid? affect social communication and foraging behavior in Reinhold Siede (Landesbetrieb Landwirtschaft Hessen honeybees Bieneninstitut Kirchhain, Germany), Lena Faust (Institut für Léa Tison (Institut für Biologie-Neurobiologie Freie Universität Bienenkunde Polytechnische Gesellschaft Goethe Universität Berlin, Germany), Marie-Luise Hahn (Institut für Biologie- Frankfurt, Germany), (Bayer CropScience Neurobiologie Freie Universität Berlin, Germany), Uwe Aktiengesellschaft, Germany), Marina Meixner (Landesbetrieb Greggers (Institut für Biologie-Neurobiologie Freie Universität Landwirtschaft Hessen Bieneninstitut Kirchhain, Germany), Berlin, Germany), Randolf Menzel (Institut für Biologie- Bernd Grönewald (Institut für Bienenkunde Polytechnische Neurobiologie Freie Universität Berlin, Germany) Gesellschaft Goethe Universität Frankfurt, Germany), Ralph Honeybee foragers may be exposed to neonicotinoids during Büchler (Landesbetrieb Landwirtschaft Hessen Bieneninstitut their foraging flights via pollen, nectar and guttation drops. Kirchhain, Germany) It has been observed that neonicotinoids may compromise Traces of thiacloprid are found in honey and pollen collected navigation in honeybees. Yet, the colony relies on the from rape. Effects of this insecticide on individual bees foragers’ ability to locate food sources and bring pollen are known but its relevance for colonies is controversially and nectar back to the hive. The successfully returning debated. With the goal to assess its risk potential on the foragers will deposit pesticide containing substances in colony level a three-year study was initiated in Germany. In the hive which may accumulate over time. The waggle July 2011, 2012 and 2013, 30 colonies were started from dance may also be affected by sublethal doses of these shook swarms, divided into three groups with ten colonies pesticides since neonicotinoids interfere with the nicotinic each, and migrated to an experimental yard. Five times per synaptic transmission in the central nervous system of autumn they were provided with 5 l sugar syrup containing insects possibly altering social communication processes. either 200 ppb or 2000 ppb thiacloprid or syrup alone We trained a group of about 30 foragers from two colonies (control). The colonies were scored four times in autumn of honeybees Apis mellifera carnica in observation hives to and four times in spring till the succeeding May every 3 two separate feeders located 340 meters from the respective weeks. Food stores were sampled before and after wintering hive. One group (from the experimental colony) foraged 4 and analyzed for thiacloprid. All three groups performed weeks on a sucrose solution containing a sub-lethal dose similarly. Mean-numbers of bees per colony were 12599 for (0.03 mM) of thiacloprid, and another group (from the control controls, 12598 for 200 ppb and 11705 for 2000 ppb and colony) foraged over the same time at a feeder containing 10784, 10545 and 9977 for brood resp. (N observations = only sucrose solution. Navigation abilities of these bees were 200 per group, current state 11/ 2013). Differences were not tested in a catch-and-release design using the harmonic significant (glm rep. measurements, spss, p bees ≥ 0.092; radar technique. In addition, bees´ waggle dances were p brood ≥ 0.547) despite high recovery rates for thiacloprid video-recorded in both control and treated hives and the in March (≤ 0.01mg/kg for controls, ≤ 0.13mg/kg for group electric fields of the dancers were measured via electrodes. 200ppb and ≤ 0.83 mg/kg for group 2000 ppb). So far our The traffic at each feeder was estimated using light-sensitive study does not indicate an impact from chronic exposure detectors at the feeders’ entrances. We found that chronic to sub-lethal concentrations of thiacloprid on colonies. This exposure of a colony to a field-relevant dose of thiacloprid study is supported by funds of the German BMELV (via BLE in field-realistic conditions may affect honeybees’ social under the innovation support program). communication and foraging behavior. Email: [email protected] Email: [email protected] B14.38 Cholinergic pesticides cause mushroom body neuronal inactivation in honeybees Mary J Palmer (University of Dundee, United Kingdom), Chris Moffat (University of Dundee, United Kingdom), Nastja Saranzewa (University of Dundee, United Kingdom), Jenni Harvey (University of Dundee, United Kingdom), Christopher Connolly (University of Dundee, United Kingdom) Pesticides that target cholinergic neurotransmission are highly effective but their use has been implicated in insect pollinator population decline. Honeybees are exposed to two widely-used classes of cholinergic pesticide: neonicotinoids (nicotinic receptor agonists), and organophosphate miticides (acetylcholinesterase inhibitors). Although sub-lethal levels of neonicotinoids are known to disrupt honeybee learning and foraging behaviour, the neurophysiological basis of these effects is unclear. Here, using recordings from mushroom body Kenyon cells in acutely-isolated honeybee brain, we show that the neonicotinoids imidacloprid and clothianidin, and the miticide coumaphos oxon, cause depolarisation- block of neuronal firing and inhibit nicotinic responses. These effects are observed at nanomolar concentrations and are additive with combined application. Our findings demonstrate a neuronal mechanism for cognitive impairments caused by neonicotinoids, and predict that exposure to multiple pesticides that target cholinergic signalling will cause enhanced toxicity to pollinators. Email: [email protected]

10 Poster Session: Wednesday 22 January

B14.39 Towards direct monitoring of sub-lethal effects of neonicotinoids in the honeybee brain. Albrecht Haase (University of Trento Department of Physics Center for BrainMind Sciences, Italy), Mara Andrione (University of Trento Department of Physics Center for BrainMind Sciences, Italy), Marco Paoli (University of Trento BIOtech center, Italy), Elisa Rigosi (University of Trento BIOtech Center Center for BrainMind Sciences, Italy), Giorgio Vallortigara (University of Trento Center for BrainMind Sciences, Italy), Renzo Antolini (University of Trento Department of Physics Center for BrainMind Sciences, Italy) Alarming reports on a decline of the honeybee population are coming also from Trentino, a province in northern Italy dominated by fruit and wine cultivation. Our laboratory for nonlinear bioimaging developed an experimental platform for in-vivo morpho-functional imaging of the honeybee brain via two-photon microscopy. We are going to apply this method in an optical study on the influence of pesticides to the honeybee brain. We will look at neuroplastic changes on different scales from entire brain regions, to single neurons and synaptic densities under the influence of neonicotinoids. Besides morphological changes, also the antennal lobe activation patterns as response to specific odour stimuli will be directly monitored via calcium imaging to individuate changes under neonicotinoid administration. With these studies we hope to identify neural dysfunctions and give thresholds for the onset of these sub-lethal effects. The achieved results will support the local agricultural industry to redesign their pest control strategies avoiding damage to the pollinators. Email: [email protected]

11 Programme & Talks: Thursday 23 January Chair: Geraldine Wright 09:00 Prof Francesco Pennacchio 15:00 (University of Napoli Federico II, Italy) (Syngenta, United Kingdom) Effect of clothianidin on insect immunity and Operation Pollinator 2001-2013 honeybee health [B14.17] [B14.9] 15:30 Dr Geraldine A Wright 09:30 Mr Yves Le Conte (Newcastle University, United Kingdom) (INRA UMR INRAUAPV Abeilles et Interaction of dietary protein and neonicotinoids Environnement, France) on the survival and nutrient balancing of the buff- Stresses interactions and honey bee losses tailed bumblebee, Bombus terrestris [B14.10] [B14.18] 10:00 Dr Jeffery S Pettis 16:00 Miss Elizabeth J Collison (United States Dept. of Agriculture, United (University of Exeter, United Kingdom) States) Investigating the effects of neonicotinoids on Crop pollination exposes honey bees to pesticides glucose oxidase- a honey bee hypopharyngeal which alters their susceptibility to the gut pathogen gland secretion nosema ceranae [B14.19] [B14.11] 16:15 Paula M. Garrido 10:30 Refreshment Break (Laboratorio de Artrópodos Universidad Nacional de Mar del Plata-CONICET, Argentina) 11:00 Ian Boyd (DEFRA) How apicultural practices and Nosema ceranae 11:30 Dr Louisa A Hooven affect individual honey bee health? (Oregon State University, United States) [B14.20] Fungicides affect development of honey bee 16:30 Questions and Discussion colonies [B14.12] 16:45 Posters and Wine Reception 11:45 Dr Elaine C. M. Silva-Zacarin 19:30 Conference Dinner at The Kitchin (see page 24 (Universidade Federal de São Carlos, Brazil) for directions) Early side-effects of non-lethal doses of thiamethoxam in the midgut of Africanized honeybee and its interaction with Nosema spores inoculation. [B14.13] 12:00 Lunch Break 13:00 Prof May R Berenbaum (University of Illinois at Urbana-Champaign, United States) Honeybee cytochrome P450s: how a 6-million- year-old genome copes with pesticide-intensive modern agriculture [B14.14] 13:30 Dr Reinhard Stöger (University of Nottingham, United Kingdom) Exposure to low levels of imidacloprid affects metabolic network of honeybee larvae [B14.15] 14:00 Dr Christopher N Connolly (University of Dundee, United Kingdom) Sub-lethal effects of neonicotinoids on neuronal function and dysfunction. [B14.16] 14:30 Refreshment Break

12 Talks: Thursday 23 January

B14.9 B14.11 09:00 Effect of clothianidin on insect immunity and honeybee 10:00 Crop pollination exposes honey bees to pesticides which health alters their susceptibility to the gut pathogen nosema Francesco Pennacchio (University of Napoli Federico II, Italy), ceranae Gennaro Di Prisco (University of Napoli Federico II, Italy), Jeffery S Pettis (United States Dept. of Agriculture, United Valeria Cavaliere (University of Bologna, Italy), Desiderato States), Elinor Lichtenberg (University of Maryland, United Annoscia (University of Udine, Italy), Paola Varricchio States), Dennis Van Engelsdorp (University of Maryland, (University of Napoli Federico II, Italy), Emilio Caprio United States) (University of Napoli Federico II, Italy), Francesco Nazzi Recent declines in honey bee populations and increasing (University of Udine, Italy), Giuseppe Gargiulo (University of demand for insect-pollinated crops raise concerns about Bologna, Italy) pollinator shortages. Pesticide exposure and pathogens Biotic and abiotic stress factors are both involved in the may interact to have strong negative effects on managed honeybee colony decline and eventual collapse, which honey bee colonies. Such findings are of great concern are often associated with high loads of parasites and given the large numbers and high levels of pesticides found pathogens. This suggests that they are able to interfere with in honey bee colonies. Thus it is crucial to determine how the immune defense barriers of honeybees. In particular, an field-relevant combinations and loads of pesticides affect immunodepression triggered by neonicotinoid insecticides bee health. We collected pollen from bee hives in seven has been hypothesized on the basis of their capacity to major crops to determine 1) what types of pesticides bees enhance pathogen impact on honeybees. Here we report are exposed to when rented for pollination of various crops the molecular mechanism underlying the neonicotinoid and 2) how field-relevant pesticide blends affect bees’ induced replication of the Deformed wing virus (DWV) in susceptibility to the gut parasite Nosema ceranae. Our honeybees bearing covert infections. We have identified samples represent pollen collected by foragers for use by a negative regulatory pathway of NF-B immune signaling the colony, and do not necessarily indicate foragers’ roles as in insects, which is activated by clothianidin. Indeed, pollinators. In blueberry, cranberry, cucumber, pumpkin and exposure to field realistic sub-lethal doses of this insecticide watermelon bees collected pollen almost exclusively from and of imidacloprid promotes viral replication in infected weeds and wildflowers during our sampling. We detected 35 honeybees. This study will likely contribute to the definition different pesticides in the sampled pollen, and found high of additional guidelines for testing chronic or sub-lethal fungicide loads. The insecticides esfenvalerate and phosmet effects of pesticides, which will take into consideration were at a concentration higher than their median lethal dose the immunomodulating effects of neuroactive substances. in at least one pollen sample. While fungicides are typically Moreover, we predict that it will foster new research on neural seen as fairly safe for honey bees, we found an increased modulation of immunity in insects. probability of Nosema infection in bees that consumed pollen Email: [email protected] with a higher fungicide load. Our results highlight a need for research on sub-lethal effects of fungicides and other B14.10 chemicals that bees placed in an agricultural setting are 09:30 Stresses interactions and honey bee losses exposed to. Yves Le Conte (INRA UMR INRAUAPV Abeilles et Email: [email protected] Environnement, France), Claudia Dussaubat (INRA UMR B14.12 INRAUAPV Abeilles et Environnement, France), Belzunces Luc (INRA UMR INRAUAPV Abeilles et Environnement, 11:30 Fungicides affect development of honey bee colonies France), Jean-Luc Brunet (INRA UMR INRAUAPV Abeilles et Louisa A Hooven (Oregon State University, United States) Environnement, France), Cédric Alaux (INRA UMR INRAUAPV Fungicides are applied during bloom to control a variety of Abeilles et Environnement, France) agricultural pathogens, based on the assertion that they have Massive honeybees losses have been reported in many negligible toxicity to pollinators. However, beekeepers have places in the world, and usually the specific causes are observed effects on larvae weeks after their application. unknown. Single factors have not yet explained this Laboratory studies have also suggested that certain global decline, leading to the hypothesis of multifactorial fungicides exhibit larval toxicity. Long-term, colony level syndromes. Consequently, testing the integrative effects studies are needed to decipher whether specific fungicide of more than one stress is an interesting approach to affect colony health or growth. Using a semi-field approach, understand colony losses. We tested the effects of an we introduced pollen to honey bee colonies containing field- infectious organism and an insecticide on honeybee relevant concentrations of iprodione, chlorothalonil, ziram, or health. We demonstrated that a synergistic effect between a mixture of boscalid/pyraclostrobin. Our initial experiments both agents, at concentrations encountered in nature, suggested that iprodione and chlorothalonil treatments significantly weakened honeybees.Nosemain combination exhibit delayed effects on brood, ziram treatment resulted in with imidacloprid caused in the short term a higher rate loss of queens, while boscalid/pyraclostrobin had no effects of mortality and energetic stress than either agent alone. compared to controls. We expanded our study of iprodione, A measure of immunity, glucose oxidase activity, was a dicarboximide fungicide used on many crops. Our previous significantly decreased only by the combined treatments, results were confirmed, and we found less increase in larvae emphasizing their synergistic effects. We demonstrated an and capped brood compared to controls several weeks after effect ofNosemaon worker pheromone production which treatment. As commercial honey bees are moved from crop shows that a pheromonal disruption related to different to crop during the growing season, they are directly sprayed stresses could be involved in weakening colonies. We did and collect pollen contaminated with fungicides. Many not found an effect of imidacloprid on worker pheromone fungicides also accumulate in beeswax. The significance of production yet. We also showed that the quality and diversity our results, demonstrating deleterious effects of iprodione on of pollen can affect honey bee health and demonstrated colony development, must be considered in the context of that diet diversity increase the immune-competence of repeated and prolonged fungicide exposures to honey bees honey bees, which suggest a link between protein nutrition and other pollinators. and immunity. We will also present data on synergetic Email: [email protected] effect ofNosemain combination with imidacloprid on queen survival in natural conditions. We, thus, provide evidence for integrative effects of different agents and stress on honeybee health, both in the short and long term. Email: [email protected] 13 Talks: Thursday 23 January

B14.13 B14.14 11:45 Early side-effects of non-lethal doses of thiamethoxam 13:00 Honeybee cytochrome P450s: how a 6-million-year- in the midgut of Africanized honeybee and its interaction old genome copes with pesticide-intensive modern with Nosema spores inoculation. agriculture Elaine C. M. Silva-Zacarin (Universidade Federal de São May R Berenbaum (University of Illinois at Urbana- Carlos, Brazil), Ales Gregorc (KIS, Slovenia), Roberta Champaign, United States) C.F. Nocelli (Universidade Federal de São Carlos, Brazil), As a managed pollinator, the honeybeeApismelliferais Stephan M. Carvalho (Universidade Federal de Uberlândia, critical to the American agricultural enterprise. Persistent Brazil), Thaisa Roat (UNESP Campus de Rio Claro, Brazil), colony losses are thus of continuing concern; possible Daiana A. Tavares (UNESP Campus de Rio Claro, Brazil), explanations for bee decline include nutritional deficiencies Hellen M. Soares (UNESP Campus de Rio Claro, Brazil), and exposures to pesticide and pathogens. Genome Érica W. Teixeira (APTA Pindamoguangaba, Brazil), Fbio C. sequencing revealed in 2006 the honeybee genome Abdalla (Universidade Federal de São Carlos, Brazil), Osmar contains far fewer cytochrome P450 genes associated with Malaspina (UNESP Campus de Rio Claro, Brazil) xenobiotic metabolism than do most other insect genomes. Histopathological analyses of the workers midgut were The dominance of the CYP6AS family appears to relate performed in order to establish the synergistic effect of to its role in processing phytochemicals encountered by the insecticide thiamethoxam and Nosema infection in honeybees in their distinctive diet of (processed) honey and Africanized honeybees. Midgut presents the first contact with beebread.Toxicological studies indicate that competition for any orally administered insecticide and it is also the site of access to catalytic sites of CYP9Q enzymes that detoxify Nosema infection. For this purpose, newly emerged workers pesticides may result in synergistic interactions among were divided in four treatment groups receiving: 0.0856 ng pesticides, a phenomenon likely exacerbated by honeybee thiamethoxam/bee (LC50/50); 0.00856 ng thiamethoxam/bee exposures to agricultural pesticides and in-hive acaricides, (LC50/500); Nosema+0.0856ng/bee; Nosema+0.00856ng/ over and above dietary exposures to naturally occurring bee. All bees were individually treated ‘per os’ with 4µL dietary phytochemicals. Moreover, dietary phytochemicals sucrose solution containing thiamethoxam and/or 60.000 play a hitherto unrecognized role in upregulation of genes Nosema spores. Seventy-two hours after the exposure, five associated with both immunity and defense, a finding linking individuals were sampled randomly from each treatment nutrition to the ability to withstand both pesticides and group and processed for histopathological diagnosis. pathogens. Morphological alterations in the digestive cells of the Email: [email protected] midgut’s posterior region were observed in dose-dependent B14.15 thiamethoxam-exposed bees. The increase of both the nuclear chromatin compaction and cytoplasm vacuolization, 13:30 Exposure to low levels of imidacloprid affects metabolic as well as the increase of the apocrine secretion in digestive network of honeybee larvae cells and the cell elimination to the lumen were evident. Reinhard Stöger (University of Nottingham, United Kingdom), Some regenerative cells of the midgut’s posterior region Kamila Derecka (University of Nottingham, United Kingdom), presented morphological alterations only at the highest dose Martin J. Blythe (University of Nottingham, United Kingdom), of thiamethoxam. Midgut’s anterior region kept intact in Sunir Malla (University of Nottingham, United Kingdom), bees exposed to both thiamethoxam treatments. Nosema- Diane Genereux (Westfield State University, United States), treated bees presented the midgut epithelium similar to the Alessandro Guffanti (Genomnia srl, Italy), Paolo Pavan untreated bees. In bees exposed to either thiamethoxam (Genomnia srl, Italy), Anna Moles (Genomnia srl, Italy), doses and simultaneously to Nosema, an increase of the Charles Snart (University of Nottingham, United Kingdom), apocrine secretion in digestive cells and the cell elimination Thomas Ryder (Parks Apiaries, United Kingdom), Thomas to the lumen was found. In our report we further discuss Ryder (Parks Apiaries, United Kingdom), Catharine A. Ortori about morphological alterations in midgut tissue after (University of Nottingham, United Kingdom), David A. Barrett thiamethoxam and Nosema spores synergism. (University of Nottingham, United Kingdom), Eugene Schuster Email: [email protected] (Versity College London, United Kingdom) Evolutionarily new environmental stressors such as the neonicotinoid class of crop-protecting agents have been implicated in the population declines of pollinating insects, including honeybees (Apis mellifera). We asked if field- relevant levels of the neonicotinoid insecticide imidacloprid could influence the physiology of worker bee larvae. Over a period of 15 days, we provided syrup tainted with low levels (2µg/L-1 ) of imidacloprid to beehives located in the field. We measured transcript levels by RNA sequencing and established lipid profiles using liquid chromatography coupled with mass spectrometry from larvae of imidacloprid- exposed (IE) and unexposed, control (C) hives. Within a catalogue of 300 differentially expressed transcripts in larvae from IE hives, we detect significant enrichment of genes functioning in lipid-carbohydrate-mitochondrial metabolic networks. Altered metabolism is also implied by our lipid profiling results, where we observed significant differences in ratios of around 15% of the sampled lipid metabolites between IE and C larvae. Collectively, we identify a multifaceted, physiological response of worker bee larvae to an evolutionarily novel stress factor. We discuss how pesticide exposure in early life could lead to persistent changes in gene expression patterns that are mediated by epigenetic programming mechanisms. Email: [email protected]

14 Talks: Thursday 23 January

B14.16 protein and carbohydrates. We also we tested nutrition the bumblebee’s susceptibility to the neonicotinoid 14:00 Sub-lethal effects of neonicotinoids on neuronal function pesticide, imidacloprid. Using a paired-diet design, our and dysfunction. experiments identified the intake target (IT), or optimal ratio Christopher N Connolly (University of Dundee, United of protein-to-carbohydrate (P:C) to be 1:75. When bees Kingdom) were exposed to imidacloprid, they ate significantly less Insect pollinators are exposed to neonicotinoids in the food. As expected if pesticides increased the demand for environment at very low (ppb) doses but the effects of nutrients, the bumblebees’ IT shifted towards a diet higher this exposure are still hotly debated. We have determined in protein. Contrary to our expectations, however, bees fed the delivery rate of ingested imidacloprid to the brains diets containing protein had a significantly greater risk of of Apis mellifera and Bombus terrestris and related this dying than bees fed sucrose alone. We predict that bees exposure level to neuronal function. At field-relevant doses, that consume neonicotinoids face a trade-off between imidacloprid in bumblebees reaches neuroactive levels consumption of protein that they need for reproduction and and can induce neuronal damage. At even lower levels in somatic maintenance and their survival. Our data identify bumblebees, imidacloprid can increase neuronal vulnerability a possible mechanism for the previously observed slow to other, normally innocuous, insults. Field studies indicate declines of wild bee populations exposed to neonicotinoids that the level of Varroa infestation does not correlate with Email: [email protected] overwintering failures in honeybees suggesting that most beekeepers are successfully controlling Varroa destructor. B14.19 Furthermore, miticide treatment itself does not appear to 16:00 Investigating the effects of neonicotinoids on glucose have a major impact on colony survival rates. Therefore, the oxidase - a honey bee hypopharyngeal gland secretion impact of pesticides and how they interact with nutritional Elizabeth J Collison (University of Exeter, United Kingdom), deficits and disease remain critical questions in the Heather J Hird (Food and Environment Research Agency, understanding of insect pollinator declines. United Kingdom), Charles R Tyler (University of Exeter, United Email: [email protected] Kingdom), James C Cresswell (University of Exeter, United B14.17 Kingdom) The honey bee hypopharyngeal gland (HPG) synthesises 15:00 Operation Pollinator 2001-2013 several proteins important for colony function. Major Royal (Syngenta, United Kingdom) Jelly Proteins (MRJPs) are secreted from the HPG of nurse In 2001 the Buzz project sponsored by Syngneta Unilever bees to feed the brood, whilst secretions from older foragers and DEFRA commisioned reseach into the Greening of the are dominated by Glucose Oxidase (GOX) to sterilise colony CAP ( Common Ag Policy)after concerns form the farming honey supplies. Honey bee workers experimentally exposed community about how it would effect the businesses and to the neonicotinoid imidacloprid and the microsporidian landscape. Buzz showed that these measures could deliver pathogen Nosema spp in combination have been shown to the greening and biodiversity to the landscape whilst have a reduced GOX activity (Alaux et al., 2010). This was continuing to farm in the same field . From these results in not the case when exposed to these two factors separately. 2005 Operation Bumblebee was formed and over 1200ha Recent studies have also found that oral exposure to of vital Pollen & Nectar habitat were established in the UK. sublethal doses of imidacloprid resulted in a reduction in In 2009 the project moved to Europe and now exists in 16 the diameter of the acini in the honey bee HPG (Heylen et countries , there are 15 different seed mixes to suit climate al., 2010; Smodis Skerl and Gregorc, 2010; Hatjina et al., soil type and pollinator. And the value as yield and quality 2013). Following these findings, the European Food Safety of the Pollination services that it brings to the landscape are Authority (EFSA) new guidance document (July 2013) being developed as ecosystem services to mass flowering outlined that an assessment of honey bee HPGs should be crops. There are a number of research projects undertaking included in pesticide risk assessment. Here we investigated this work in academic institutions across Europe. Operation the effect of imidacloprid and thiamethoxam on GOX activity Pollinator demonstrates that commercial sustainable food in honey bees. We observed an increase in GOX activity in production and positive environmental management can co- imidacloprid-exposed honey bees, but found no effect of exist in our landscape – after all its food for us and food for thiamethoxam exposure. We hypothesise that this increase wildlife in the same field” may result from a reduction in HPG size, leading to a shift Email: @syngenta.com from MRJP-secretion to GOX-secretion, but this needs B14.18 investigation for verification. These findings support the hypothesis that imidacloprid may lead to a premature shift 15:30 Interaction of dietary protein and neonicotinoids on from nurse to forager role in worker honey bee development. the survival and nutrient balancing of the buff-tailed Email: [email protected] bumblebee, Bombus terrestris Geraldine A Wright (Newcastle University, United Kingdom), Sophie Derveau (Newcastle University, United Kingdom), Daniel Stabler (Newcastle University, United Kingdom), Jessica Mitchell (Newcastle University, United Kingdom) Poor nutrition, diseases and exposure to pesticides in modern agriculture are all like to contribute to the decline of wild pollinators. We know surprisingly little about the nutritional needs of wild pollinators and whether or not they obtain sufficient nutrients from modern agricultural landscapes. Several controversial studies have shown that colonies of buff-tailed bumblebees, Bombus terrestris, exposed to neonicotinoids in food have impaired brood rearing, compromised foraging, and poor worker survival. Nutrition could be an important mitigator of the survival of wild pollinators when they are exposed to low doses of pesticides. Here, we used the Geometric Framework for nutrition to identify the optimal nutrition of adult foraging workers of Bombus terrestris for diets composed of

15 Talks & Posters: Thursday 23 January

B14.20 B14.31 Relation between treatment of bee colonies with coumaphos or amitraz on their residue levels in honey, 16:15 How apicultural practices and Nosema ceranae affect bee brood and beeswax individual honey bee health? Tomaz Snoj (Veterinary faculty University in Ljubljana, Paula M. Garrido (Laboratorio de Artrópodos Universidad Slovenia), Blanka Premrov Bajuk (Veterinary faculty University Nacional de Mar del Plata-CONICET, Argentina), Karina in Ljubljana, Slovenia), Katarina Babnik (Veterinary faculty Antúnez (Departamento de Microbiología Instituto de University in Ljubljana, Slovenia), Luka Milcinski (Veterinary Investigaciones Biológicas Clemente Estable, Uruguay), faculty University in Ljubljana, Slovenia), Vlasta Jencic Martín P. Porrini (Laboratorio de Artrópodos Universidad (Veterinary faculty University in Ljubljana, Slovenia), Metka Nacional de Mar del Plata-CONICET, Argentina), María B. Pislak Ocepek (Veterinary faculty University in Ljubljana, Branchiccela (Departamento de Microbiología Instituto de Slovenia), Martina Skof (Veterinary faculty University in Investigaciones Biológicas Clemente Estable, Uruguay), Ljubljana, Slovenia), Silvestra Kobal (Veterinary faculty Giselle M. Martínez Noël (Instituto de Investigaciones en University in Ljubljana, Slovenia) Biodiversidad y Biotecnología, Argentina), Pablo Zunino The aim of the study was to assess the accumulation of (Departamento de Microbiología Instituto de Investigaciones coumaphos and amitraz residua in honey, bee brood and Biológicas Clemente Estable, Uruguay), Martín J. Eguaras beeswax after the treatment of honeybee colonies against (Laboratorio de Artrópodos Universidad Nacional de Mar del varosis (Varroa destrucor). The study was conducted in Plata-CONICET, Argentina) two apiaries on two different locations. In the first location Honey bee colonies are exposed to pesticides used in ten bee colonies of Apis melifica carnica were treated with agriculture or within bee hives by beekeeper intervention. coumaphos (CheckMite, Bayer, Germany) and on other The organophosphate coumaphos and the pyrethroid tau- location five bee colonies were treated with amitraz (Apivar, fluvalinate are widely used to control Varroa destructor. These Veto-Pharma, France). Non-treated colonies served as acaricides are applied directly to bee hives, accumulate in controls. Honey, wax and brood samples were collected wax and had been detected even in commercial bee wax before and six weeks after the treatment. Detection of foundation. Nosemosis caused by Nosema ceranae is one coumaphos and amitraz and its metabolites DPMF, DMF of the most prevalent and pathogenic disease that affect and DMA in honey and brood was performed by HPLC with adult honeybees. Interactive effects between N. ceranae UV detection, while wax coumaphos and amitraz and its and sublethal dosis of these acaricides (at concentrations metabolites were determined using GC. Coumaphos levels found in honey) on immune related genes were assessed. in honey from treated and non-treated bee colonies were In order to allow honeybee development under a free- found bellow MRL (100 µg/kg). In the brood from treated acaricide environment, plastic foundation was used and colonies coumaphos levels ranged between 49.0 and 784.1 bees drawn out the foundation. Gene expression changes µg/kg. Interestingly, coumaphos in wax was found in treated in nurse bees were measured using qPCR. This work and non-treated bee colonies. The levels of coumaphos in demonstrates that chronic exposure with tau-fluvalinate the wax were high and ranged between 0.19 and 36 mg/kg. significantly reduced the transcription of genes encoding High coumaphos residua accumulation in wax in both treated the antimicrobian peptides abaecin and hymenoptaecin. and untreated colonies are probably the results of the bee Coumaphos decreased vitellogenin and lysozyme expression colonies treatments with coumaphos before the reasearch. and, in combination with N. ceranae infection, reduced levels However, the levels of amitraz and its metabolites were found of abaecin and enhance phenoloxidase transcripts. Only bellow limit of detection. Our study shows that coumaphos defensin and glucose dehydrogenase genes were not altered accumulates in wax and brood, while amitraz and its by the different treatments. Immune response at individual metabolites do not accumulate in honey, brood nor wax. level and susceptibility to pathogens may be compromised Email: [email protected] when honeybees are exposed not only to sublethal doses of acaricides but N. ceranae infection and their interactive B14.40 A novel assay to assess the impact of chemicals on effects. learning behavior and locomotion in honey bees Email: [email protected] Nicholas Kirkerud (University of Konstanz Neurobioloy, Germany), David Gustav (University of Konstanz Neurobioloy, Germany), Giovanni Galizia (University of Konstanz Neurobioloy, Germany) Recent results suggest that pesticides, especially neonicotinoids have impacts on learning, memory and locomotion of honey bees even at sub-lethal doses. So far assessment of these effects has been accomplished through time-consuming, expensive and/or technically challenging PER (Proboscis Extension Response)-conditioning, RFID- tagging or video analysis of movement assays. We recently developed a flexible conditioning device, APIS (Automatic Performance Index System), where effects of chemicals such as pesticides can be quantified in a standardized and convenient way: Bees are introduced into a walking chamber where they learn to associate injected odors with mild electric shocks. The bees’ movement is continuously sampled by infrared photosensors, enabling us to quantify learning and locomotion by analyzing different movement parameters. All controlling elements are integrated in a single device, and operated through a single program. This allows us to use a variety of training protocols where the bee’s behavior governs the triggering of stimuli (operant conditioning). To demonstrate APIS as a diagnostic tool, we present results from two different experiments: One where the acute effect of the neonicotinoid Acetamiprid is tested in an operant differential conditioning-paradigm and another where the short-term effect of Thiamethoxam is tested in an operant absolute conditioning-paradigm. Our results show that bees treated with sub-lethal doses of these two 16 commonly used neonicotinoids have impaired performance Poster Session: Thursday 23 January

compared to untreated bees, in that they fail to recognize the because of their role as pollinators. It is as yet equivocal reinforced odor from the neutral odor. as to whether pesticides such as OPs contribute to the Email: [email protected] existing problem of bee population decline, but the need to understand how OPs directly affect bees, and from B14.41 An examination of targets of organophosphorus their environmental persistence affect other organisms pesticides in human and honey bee brains. as well as humans, is a crucial research topic. We have John W Grzeskowiak (School of Medicine University of examined post-mortem human brain tissue and brains Nottingham, United Kingdom), Fryni Drizou (School of from honey bees to assess the presence of secondary Medicine University of Nottingham, United Kingdom), Ian OP targets. Human and bee brain tissue was fractionated Mellor (School of Life Sciences University of Nottingham, by differential centrifugation and proteins radiolabelled United Kingdom), Amaia M Erdozian (Department of with the broad serine hydrolase inhibitor tritiated- Pharmacology University of the Basque Country, Spain), diisopropylfluorophosphate (3 H-DFP). From quantitation of Benito Morentin (Department of Pharmacology University radiolabelling, and protein resolution by one dimensional and of the Basque Country, Spain), Luis F Callado (Department two dimensional gel electrophoresis, we aim to characterise of Pharmacology University of the Basque Country, secondary OP targets in both tissues. Once identified their Spain), Wayne G Carter (School of Medicine University of influence on cellular and homeostatic mechanisms can begin Nottingham, United Kingdom) to be dissected. Organophosphorus (OP) compounds are widely used Email: [email protected] as commercial and domestic pesticides. Toxicity of OP compounds arises from the targeted inhibition of B14.43 Nutrition and immunity in honeybees: a lab study acetylcholinesterase (AChE), a serine hydrolase active on the Gennaro Di Prisco (University of Naples, Italy), Francesco post synaptic membrane of neurons. Pesticides that target Pennacchio (University of Naples, Italy), Geraldine Wright cholinergic neurotransmission have been shown to be highly (Newcastle University, United Kingdom) effective in controlling populations of pest insects, however Honeybees (Apis mellifera L.) play a key-role in the both organophosphate and neonicotinoid pesticides have environment and are essential for pollination of many crop been implicated in the decline of honey bee populations. In plants. Honeybee immunity can be adversely affected by addition to cholinesterases OP pesticides have been shown a number of biotic and abiotic stress factors, which can to bind and adduct other secondary protein targets. This synergistically interact. In the past few years, an increasing promiscuous binding remains a health concern to humans occurrence of colony decline and eventual collapse and animals, with an association between low-level exposure has been reported globally. These declines are often to OPs and impaired neurobehavioral function being recently associated with colonies containing high loads of pathogens suggested (Ross et al., 2013). Furthermore a recent letter and parasites, with bees exhibiting clear signs of bee in nature communications indicates that sublethal levels of immunosuppression, but are also associated with increased neonicotinoids are able to disrupt honeybee learning and exposure to pesticides and acaricides. One critical factor behaviour, principally through postsynaptic depolarization that could improve bee resistance to pesticides, parasites meditating block of neuronal firing (Palmer et al., 2013). and pathogens is colony nutrition. Here we tested whether We have examined post-mortem human brain tissue and access to essential amino acids in diet influenced adult brains from honey bees to assess the presence of secondary worker bee susceptibility to the neonicotinoid pesticide, OP targets. Human and bee brain tissue was fractionated imidacloprid. We also measured immunocompetence in by differential centrifugation and proteins radiolabelled this population and pathogen load of deformed wing virus with the broad serine hydrolase inhibitor tritiated- (DWV). Our results indicate that diets high in amino acids diisopropylfluorophosphate (3 H-DFP). From quantitation of reduce survival and promote a more intense proliferation radiolabelling, and protein resolution by one dimensional and of DWV. The effect of diet on survival was amplified by two dimensional gel electrophoresis, we aim to characterise exposure to imidacloprid. Our study shows that there are secondary OP targets in both tissues. potentially severe and unpredictable interactions between Once identified their influence on cellular and homeostatic dietary components and pesticides on immunocompetence mechanisms can begin to be dissected. PALMER, M. J. et al. and survival that are likely to account for honeybee colony 2013. Nat Commun, 4, 1634. collapse. ROSS, S. M. et al. 2013. Crit Rev Toxicol, 43, 21-44. Email: [email protected] Email: [email protected] B14.44 Understanding and managing honey bee health in the B14.42 An examination of targets of organophosphorus UK: beekeeper knowledge and engagement with science pesticides in human and honey bee brains and policy Fryni Drizou (School of Medicine Nottingham University, Emily Adams (Lancaster University, United Kingdom), United Kingdom), John Grzeskowiak (School of Medicine Rebecca Ellis (Lancaster University, United Kingdom), Ken Nottingham University, United Kingdom), Ian Mellor (School Wilson (Lancaster University, United Kingdom) of Life Sciences University of Nottingham, United Kingdom), Insect pollinators have become well-recognised symbols of Amaia M. Erdozian (Department of Pharmacology University environmental issues and biological decline for the media of the Basque Country, Spain), Benito Morentin (Department and general public. Much attention has focused on honey of Pharmacology University of the Basque Country Spain, bees. In the UK, beekeeping has become very popular, Spain), Luis F. Callado (Department of Pharmacology with membership of beekeeping organisations rising University of the Basque Country, Spain), Wayne G. Carter rapidly. Simultaneously, there has been a rapid expansion (School of Medicine Nottingham University, United Kingdom) in academic research on insect pollinators in the UK and Organophosphorus compounds (OPs) have been widely elsewhere. However, little of this has focused directly on used as commercial and domestic pesticides for many years, beekeeping communities, despite their role in supporting although historically some OPs have been banned because honey bee populations during a period when they have been of their toxicity to non-target organisms. Toxicity of OPs challenged by multiple, interacting stressors (including pests arises from the targeted inhibition of acetylcholinesterase, and diseases, habitat loss and fragmentation, and climatic a serine hydrolase active within nerve synapses. However, variation). Using semi-structured interviews, participant OP pesticides also bind and adduct other secondary protein observation and experimental work, we investigated targets. This promiscuous binding remains a health concern beekeeping knowledge and engagement with science and to humans and non-target organisms such as bees. Honey policy in beekeeping communities in north west England. bees are essential insects for life existence on the planet We found that whilst many beekeepers are passionate about

17 Poster Session: Thursday 23 January

honey bees, expending time and resources on ensuring B14.47 Investigating sublethal effects of a neonicotinoid the health of their colonies, this does not always translate pesticide on bumblebee navigation and foraging into an interest in wider environmental issues. We reflect on behaviour what implications this could have for the future, whether it is Dara A Stanley (Royal Holloway University of London, United reasonable to expect hobby beekeepers to become stewards Kingdom), Nigel E Raine (Royal Holloway University of for a key ecosystem service like pollination, and how London, United Kingdom) future engagement between academia and the beekeeping Bumblebees are essential pollinators of many important community could be carried out. agricultural crops and wild plants. While foraging in Email: [email protected] agricultural farmland bees are likely to be exposed to pesticides applied for crop protection. Although bees B14.45 Effects of field relevant concentrations of imidacloprid typically encounter these pesticides at sublethal levels, and clothianidin on bee neuronal function exposure may still have impacts on factors such as behaviour Christopher Moffat (University of Dundee, United Kingdom), or reproduction with potential consequences for colony Mary J Palmer (University of Dundee, United Kingdom), fitness. Here we examined the impact of field realistic doses Christopher N Connolly (University of Dundee, United of a neonicotinoid pesticide, thiamethoxam, on foraging and Kingdom) navigation in a common bumblebee Bombus terrestris. We Neonicotinoids are nicotinic acetylcholine receptor used Radio Frequency IDentification (RFID) tag technology agonists displaying high affinity in insects. Worldwide, to monitor colonies in a semi-field experiment. Colonies were these insecticides are used in veterinary products and as located in the lab but had free access to forage for nectar agricultural pesticides. Their high affinity and specificity and pollen outside. Our results indicate varying effects of to insects has underpinned their commercial success but pesticide exposure on both bumblebee foraging and homing there are increasing concerns regarding their effects on ability. important pollinators. While there is mounting evidence that Email: [email protected] sub-lethal doses of neonicotinoids can have detrimental effects on bees; there is continuing controversy over the B14.48 The levels of neonicotinoids that pollinators are exposed precise levels of pesticides to which bees are exposed in to whilst foraging on oilseed rape the field and stemming from this the levels that reach bee Kristopher D Wisniewski (Keele University, United Kingdom), brains. This study investigated the levels of imidacloprid: Dr Falko Drijfhout (Keele University, United Kingdom), Dr the archetypal neonicotinoid reaching bee brains and the William D.J Kirk (Keele University, United Kingdom) effects of concentrations within this range on neuronal Honey bees, bumble bees and stingless bees only represent health and function. 18 hour to 8 day feeding experiments a small proportion of insect pollinators; however, they provide were conducted with Apis melliferra and Bombus terrestris important and economically valuable pollination services using food laced with environmentally relevant levels of to agriculture and many terrestrial ecosystems. The recent imidacloprid. Experiments revealed low levels in the brains documented decline of social pollinators has been a central of both species that were sufficient to cause rapid neuronal concern to the scientific and wider communities, with damage in Bombus terrestris only. Chronic exposure of pesticides being identified as a key contributing factor to the cultured Bombus Kenyon cells to low (nM) imidacloprid falling numbers. produced vulnerability to normally sub-toxic insults. Chronic exposure to low levels of pesticides, whereby direct Electrophysiological recordings from Bombus Kenyon mortality does not occur, has been shown to induce sublethal cells revealed that sub-nanomolar concentrations of effects, which can have profound effects on both the clothianidin had an apparent desensitising effect on nicotinic individual and ultimately the overall functioning and survival acetylcholine receptors. In conclusion, exposure of bees of a colony. Various methods of pesticide exposure have to field-relevant levels of neonicotinoids may result in brain been identified; including contaminated pollen and nectar concentrations that affect neuronal function. collected during foraging; but at what levels? In this poster Email: [email protected] we present the levels of three neonicotinoids found in the pollen and nectar collected from oilseed rape. B14.46 The herbizide Paraquat induces neuronal cell death and Email: [email protected] locomotor dysfunction in a Tribolium castaneum model of Parkinson´s disease B14.49 Modelling the impacts of various pesticide effects on the Annely Brandt (Bieneninstitut Kirchhain, Germany), Ralph honeybee colony Büchler (Bieneninstitut Kirchhain, Germany), Andreas Jack CO Rumkee (University Of Exeter, United Kingdom), Vilcinskas (IME Fraunhofer Institut, Germany) Matthias A Becher (University of Exeter, United Kingdom), Paraquat (N,N-dimethyl-4-4-4-bypiridinium) is a widely used Juliet L Osborne (University of Exeter, United Kingdom) herbicide in agriculture. Epidemiological and experimental Studies have shown that the sublethal effects of pesticides studies point Paraquat as an etiological agent for Parkinson´s may have a significant effect on the behaviour or disease. The damage done by Paraquat is caused by development of the honeybee (Dai et al. 2010, Schneider oxidative stress that leads to the damage of lipids, proteins, et al. 2012 for example) and that this may lead to adverse RNA, and DNA. Paraquat induces Parkinson-like pathologies, effects on the colony as a whole. We have been using the e.g. degeneration of dopaminergic neurons in the brain and BEEHAVE model (Becher et al. currently under reveiw), locomotor dysfunction in rodents as well as in insect-model to investigate the effects of brief periods of increased organisms. We established a Paraquat-based Parkinson- foraging mortality, reduced egg-laying and disturbed larval model in the red flour beetle Tribolium castaneum with the development on the health of the colony, with an intent to aid aim to screen for neuroprotective substances. Paraquat further pesticide risk-assessment, as it has been suggested treated beetles showed a significant increase in the number that a more solid understanding is required before sublethal of apoptotic neurons in the brain, impaired climbing ability, effects can be confidently included in the risk assessment and increased tonic immobility. We found that the known (Thompson and Maus 2007). BEEHAVE is an individual-based Parkinson-drug L-DOPA, but also Curcumin, hempseed flour, model that includes multiple stressors acting on a honey Ascorbic acid and tea extracts of Uncariae ramulus had bee colony, allowing a more realistic model of the reaction of neuroprotective effects. the colony to, for example, the introduction of a pesticide to Email: [email protected] the landscape. The use of modelling in this investigation is of great value as to conduct field experiments with enough power to answer the same questions has already proven to be difficult (Cresswell 2010), and to continue the trials over

18 Poster Session: Thursday 23 January

long periods of time is a difficult task. The results produced B14.53 Spatial variation of metal concentrations found in honeys show that the effects are very dependent on the timing of the around Greater Manchester effects through the year and that sublethat effects can have a David De Peña (Manchester Metropolitan University, United significant effect on the colony health.” Kingdom) Email: [email protected] Honey bees (Apis mellifera) forage up to 4km, and can access around 50km² from their apiaries, thus sampling a B14.50 Ectopic expression of Apis mellifera voltage-gated huge number of individual points, from all environmental Ca2+ channel subunit reveals an original phospholipid- compartments such as water, air and soil. As honey is dependent regulation of Ca2+ currents considered a composited random sample, it has the Matthieu Rousset (CNRS, France), Thierry Cens (CNRS, potential of providing the most representative values for France), Michel Bellis (CNRS, France), Claude Collet (INRA, the average concentrations of bio available elements in an France), Raymond Valérie (Université d’Angers, France), areas environment. In this study, the contents of Al, As, Ba, Pierre Charnet (CNRS, France) Ca, Cd, Co, Cu,Fe, K, Mg, Mn, Mo, Na, Ni, P, Pb, S, Sr, and Insecticides have been suspected to participate in Zn were determined by inductively coupled plasma optical abnormally elevated mortalities occurring in domestic spectrophotometry (ICP-OES) on honey samples collected honeybee apiaries worldwide. Among the currently from members of the Manchester and District Beekeepers commercialized pesticides, Pyrethroids are synthetic Association (MDBKA) around Manchester during 2012 and insecticides broadly used for their major neurotoxic 2013. Multidimensional statistical analysis of the honey action on the insect nervous system. These chemically samples showed that there were correlating factors present engineered compounds, mainly described as targeting Na+ in the distribution of the metals determined within the honey channels, have also been suspected to inhibit invertebrate samples, and relationships between geographical location and vertebrate voltage-gated Ca2+ channels. However, and metal content were clearly seen. Confirmation of this the molecular tools needed to precisely study the impact was also seen from the GIS contour maps which showed of these drugs on the physiology and toxicology of the the spatial distribution of the heavy metals determined from honeybee are still lacking. We have identified in the genomic the ICP-OES analysis, with lead being distributed mainly honeybee database three pore-forming CaV subunits that amongst the south of the study area and arsenic around the could belong to the CaV1, CaV2 and CaV3 families, a single north. The results of this study indicate that honey can serve CaV and three CaV2- subunits. We report here the cloning, in the detection of spatial patterns of metal concentrations expression and characterization of the unique CaV subunit within the environment, even at relatively low levels of (AmCaV). AmCaV is expressed in thorax and leg muscles, pollution. and in different structures of the honeybee brain. When Email: [email protected] expressed in Xenopus oocytes with the mammalian CaV2.2 or CaV2.3 Ca2+ channel subunit, the Apis subunit increases the current amplitude and shifts the current-voltage curve toward hyperpolarized potentials. Interestingly, the AmCaV subunit also slows the inactivation kinetics of the current. In the presence of AmCaV subunit, the slow inactivation is regulated by drugs that affect the level of phosphoinositides. This regulation is also found in honeybee neurons in primary culture, thus suggesting that the AmCaV subunit can regulate channel inactivation via dynamic interactions with the plasma membrane. These results may help to understand the functional diversity of voltage-gated Ca2+ currents recorded in honeybee muscle and neurons. Email: [email protected] B14.52 Can honeybees sense neonicotinoids? Reception of Imidacloprid in Apis mellifera Matthias Schott (Justus Liebig University Giessen, Germany), Annely Brandt (LLH Bieneninstitut Kirchhain and Fraunhofer IME Giessen, Germany), Andreas Vilcinskas (Justus Liebig University Giessen and bLLH Bieneninstitut Kirchhain, Germany) Neonicotinoids are regularly discussed in the context of the honey bee colony losses. The amount of applied neonicotinoids in agriculture has rapidly increased since 2006. Up to now, it is not known if honey bees are able to sense neonicotinoids. If they are able to detect these chemicals that are relatively new to their environment, this could be one reason for unusual behavior. As Imidacloprid is one of the top-selling pesticides worldwide, we started our electroantennographic investigation with this neonicotinoid, to answer the question if they could smell it. We measured the reaction ofApis melliferato different concentrations of Imidacloprid by connecting dissected foraging worker antennae in an antenna holder chip. Thereupon we recorded the summed voltage emitted by the antennae sensilla after air puffs loaded with specific amounts of Imidacloprid. Our first data indicate thatA. melliferais able to sense Imidacloprid in environmentally significant concentrations. We will discuss our findings in the context of the infochemical effect. They are a starting point for further investigations that may prove the honey bee’s misinterpretation of the neonicotinoid signal. Email: [email protected] 19 Programme & Talks: Friday 24 January Chair: Chris Connolly 09:00 Richard Schmuck B14.21 (Bayer CropScience, Germany) 09:00 A causal analysis of the role of neonicotinoid insecticides A causal analysis of the role of neonicotinoid in the reported declines of bee colonies (apis mellifera) insecticides in the reported declines of bee Richard Schmuck (Bayer CropScience, Germany), colonies (Apis mellifera) (Bayer CropScience, Germany) Reported declines in overwinter survival of bee colonies B14.21 (Apis mellifera) in Western Europe, and the onset of colony 09.30 Peter Neumann collapse disorder in 2006 in North America encouraged many researchers to investigate the causative factors behind the (University of Bern) bee health problem. Throughout many study reports on this The neonicotinoids in solitary bees subject, there is a peculiar hierarchy of reasoning: first that B14.22 the increased numbers of bee losses are due to multiple causes, including stress caused by unspecified pesticides, 10:00 Maj Rundlöf then that pesticides must be given priority in the investigation (Lund University, Sweden) of the effects, and then that neonicotinoids are the pesticides A replicated landscape scale field study of impacts of concern. There have been incidents of acute toxicity of clothianidin seed dressing in oilseed rape on that appear to be linked to exposure of bees to dust from neonicotinoid seed treatments, but these are unrelated to the wild and managed bees kind of chronic toxicity and effects from sublethal exposure B14.23 that have been postulated. The authors performed a causal analysis of reported declines in bee colonies with a special 10:15 Coffee/tea break emphasis on the role of neonicotinoid insecticides. 10:45 Matthias Becher Email: [email protected] (University of Exeter, UK) B14.22 BEEHAVE: An integrated honey bee model and its 09:30 The neonicotinoids in solitary bees application to pesticide scenarios Peter Neumann (University of Bern) B14.24 B14.23 11:15 Discussion: Knowledge gaps, policy changes 10:00 A replicated landscape scale field study of impacts of and future risks clothianidin seed dressing in oilseed rape on wild and managed bees 12:00 Questions Maj Rundlöf (Lund University, Sweden), Georg KS Andersson (Lund University, Sweden), Riccardo Bommarco (Swedish University of Agricultural Sciences, Sweden), Ingemar Fries (Swedish University of Agricultural Sciences, Sweden), Veronica Hederström (Lund University, Sweden), Lina Herbertsson (Lund University, Sweden), Björn K Klatt (Lund University, Sweden), Thorsten R Pedersen (Swedish Board of Agriculture, Sweden), Johanna Yourstone (Lund University, Sweden), Henrik G Smith (Lund University, Sweden) Sublethal doses of neonicotinoids have been shown to negatively impact the health of bees. However, studies to date have only artificially fed bees with low doses and no well designed and replicated study has examined if such impacts are observed for bees foraging under field conditions. We used a study system of 16 spatially separated (>4 km) spring oilseed rape fields, where eight of the fields were randomly assigned to be sown with clotianidin dressed seeds and the other eight as controls, to assess neonicotinoid residues and impacts on wild and managed bees in Sweden. Six equally sized Apis mellifera colonies, from controlled queen origin, six commercially bred Bombus terrestris colonies, and 27 cocoons of Osmia bicornis were placed at each field. Colony development was monitored for honey bees and bumble bees, foraging trip duration and survival for bumble bees, reproduction for bumble bees and solitary bees and density of all foraging bees in fields and borders. Samples to detect insecticide residues were taken from field border vegetation during sowing and of pollen and nectar from honey bees foraging in the experimental fields and of bees at the hives during rape flowering. Five neonicotinoids, including clothianidin, were detected in the samples. Samples from the treated fields generally had an order of magnitude higher clothianidin concentrations compared to those from control fields. The study can reveal the level of neonicotinoid exposure in agricultural landscapes, as well as the field- realistic impacts of clothianidin seed dressing on bees foraging in such landscapes. Email: [email protected]

20 Talks: Friday 24 January

B14.24 10:45 BEEHAVE: An integrated honey bee model and its application to pesticide scenarios Matthias Becher (Environment Sustainability Institute, University of Exeter, United Kingdom), P. J. Kennedy (Environment Sustainability Institute, University of Exeter, United Kingdom), J. L. Osborne (Environment Sustainability Institute, University of Exeter, United Kingdom) Notable losses of managed honeybee colonies have been reported, mainly in the Northern Hemispere, which are presumed to be caused by a combination of stressors. Among these stressors, varroa mites and varroa transmitted viruses (esp. Deformed Wing Virus), forage quantity and quality, and exposure to pesticides are regularly considered as being important. To improve our understanding of the complex interactions within a colony and its environment, we developed BEEHAVE, a honeybee model that integrates colony dynamics, agent-based foraging in realistic landscapes and population dynamics of varroa mites, acting as vectors for viruses (Deformed Wing Virus and Acute Paralysis Virus). Pesticide exposure scenarios can be addressed with this model and we present results from simulations of increased forager mortality. The output of the model suggests that the impact of pesticide exposure on colony dynamics can depend on the season and on the availability of food in the landscape and may increase with repeated exposure over several years. BEEHAVE is currently under review and will be publicly available for download and use in the near future. Email: [email protected]

21 Author Index

Adams, E., B14.44 Grzeskowiak, J. W., B14.41 Rousset, M., B14.50 Baron, G., B14.33 Haase, A., B14.39 Rumkee, J. C. O., B14.49 Becher, M., B14.24 Holder, P. J., B14.25 Rundlöf, M., B14.23 Berenbaum, M. R., B14.14 Hooven, L. A., B14.12 Samson-Robert, O., B14.7 Brandt, A., B14.46 Kirkerud, N., B14.40 Schmuck, R., B14.21 Campbell, P. J., B14.6 Kukkola, M., B14.35 Schott, M., B14.52 Coates, G., B14.17 Le Conte, Y., B14.10 Siede, R., B14.36 Collison, E. J., B14.19 Lummis, S., B14.32 Silva-Zacarin, E. C. M., B14.13 Connolly, C. N., B14.16 Menzel, R., B14.1 Snoj, T., B14.31 Cutler, G. C., B14.4 Moffat, C., B14.45 Stanley, D. A., B14.47 De Peña, D., B14.53 Neumann, P. B14.22 Stöger, R., B14.15 Di Prisco, G., B14.43 Nora, C., B14.27 Tiedeken, E. J., B14.30 Dively, G. P., B14.5 Palmer, M. J., B14.38 Tison, L., B14.37 Drizou, F., B14.42 Pennacchio, F., B14.9 Willmer, P., B14.8 Garrido, P. M., B14.20 Pettis, J. S., B14.11 Wisniewski, K. D., B14.48 Goulson, D., B14.2 Raine, N. E., B14.3 Wright, E. L., B14.26 Greco, M. K., B14.29 Riessberger-Gallé, U., B14.34 Wright, G. A., B14.18

22 Delegate Information Venue Lectures and registration will take place at: Charles Darwin House 12 Roger Street London WC1N 2JU Tel: 020 7685 2400

Venue Map

KING'S CROSS THE KITCHIN NATIONAL RAIL STATIONS OSSULSTON STREET PANCRAS ROAD CALEDONIAN ST. ONIAN ROAD8 Caledonian Street UNDERGROUND STATIONS ST. PANCRAS PENTONVILLE ROAD CALED 1 UNIVERSITY COLLEGE LONDON 9 KING'S 2 HOTEL RUSSELL GRAY'S INN ROAD

CROSS 3 BRITISH MUSEUM 8 JUDD STREET 4 NATIONAL GALLERY 7 5 TRAFALGAR SQUARE EUSTON EUSTON ROAD ROAD 6 BONNINGTON HOTEL

7 NOVOTEL EUSTON EUSTON ROAD FARRINGDON ROAD 8 BRITISH LIBRARY

9 SCALA WOBURN PLACE CALTHORPE ST.

BROWNLOW 10 YORKSHIRE GREY PH DOUGHTY ST. RUSSELL DOUGHTY MWS SQUARE GOWER STREET GRAY'S INN R GUILFORD STREET OGER ST. 1 2 MWS.

JOHN ST. CLERKENWELL RUSSELL SQ. CHARLES DARWIN HOUSE 10 ROAD SOUTHAMPTON ROW

PORTLAND PLACE GOODGE THEOBALD'S ROAD STREET TOTTENHAM COURT ROAD 6 FARRINGDON DRAKE ST.

BLOOMSBURY STREET3 CHANCERY PROCTER ST. LANE

HOLBORN HIGH HOLBORN HOLBORN FARRINGDON ROADVIADUCT TOTTENHAM NEW OXFORD STREETHIGH HOLBORN COURT ROAD OXFORD STREET

KINGSWAY ST. STREET OXFORD STREET CHARING CROSS ROAD GILES HIGH OXFORD CIRCUS FLEET ST.

COVENT GARDEN STRAND REGENT STREET BLACKFRIARS

TEMPLE LEICESTER SQUARE STRAND SHAFTESBURY AVENUE EMBANKMENT

VICTORIA BLACKFRIARS BRIDGE PICCADILLY CIRCUS STRAND WATERLOO BRIDGE HAYMARKET RIVER THAMES REGENT STREET 4

PICCADILLY CHARING CROSS JERMYN STREET TRAFALGAR5 ST. JAMES STREET

BURY ST WAR HOUSE ST. EMBANKMENT PALL MALL

Registration Registration will take place from 11:00 to 13:00 on Wednesday 22 October at Charles Darwin House reception. Staff will be on hand throughout the meeting should you need any assistance. Badges must be worn for the duration of the meeting, both for security purposes and for entry to the lectures and social events.

23 Delegate Information

Meals and Refreshments Social Programme The registration fee includes lunches, We have organised a number of informal refreshments throughout the meeting and the networking and social events throughout conference dinner on Thursday 23 January. the meeting; these are all optional, but we Lunch and refreshments will be served encourage all delegates to attend to get the at the following times: most from the conference. Drinks receptions will be held in the break out area of Charles Darwin House, where lunch Wednesday 22 January and Wednesday’s buffet dinner will be served. Lunch 12:00 – 13:00 The conference dinner will be held at The Kitchin, the price of which is included in the Coffee/tea 14:30 – 15:00 registration fee. Directions to the restaurant Drinks reception 17:00 – 19:00 are below. and buffet dinner

Wednesday 22 January Thursday 23 January Poster session 17:00 – 19:00 Coffee/tea 10:30 – 11:00 with drinks reception and buffet dinner Lunch 12:00 – 13:00 Coffee/tea 14:30 – 15:00 Thursday 23 January Drinks reception 16:45 – 19.00 Poster session 16:45 – 19.00 Conference Dinner 19.30 – 21.30 with drinks reception Conference Dinner 19.30 – 21.30

Friday 24 January Coffee/tea 10:15 – 10:45 It takes approximately 15 minutes to walk from Charles Darwin House to the conference dinner venue, ‘The Kitchin’. Please follow the instructions below and refer to the map on page 23. The Kitchin is located on 8 Caledonia Street, Kings Cross, London, N1 9AA. When you exit Charles Darwin House, go straight ahead to the main street (Gray’s Inn Road) and turn left. Follow Gray’s Inn Road for about 10 minutes then bear right onto Caledonia Road. At the junction of Pentonville Road (the Scala will be on your right), cross straight over and take the ﬁrst left onto Caledonian Road. The Kitchin is on your left.

24 Delegate Information

Information for Speakers Poster Sessions All lectures will take place in the Charles Poster presenters are requested to stand Darwin Lecture Theatre. alongside their posters during their session. Speakers should contact a member of staff Velcro will be provided at the registration in advance of their presentation to upload desk. Poster display has been split into two their talk; it is recommended that talks are groups: uploaded as soon as possible to avoid Group 1: queues. abstracts numbered B14.25 – B14.39 Authors with posters in this group should hang their posters upon arrival on morning of Information for 22 January; they will present their posters at Poster Presenters lunchtime and on the evening of 22 January. Posters should be removed after that poster At this conference, we encourage cross- session. disciplinary interactions as much as possible and want every delegate to have the Group 2: opportunity to look at as many posters as abstracts numbered B14.40 – B14.53 feasible – so we encourage all delegates to Authors with posters in this group should attend both poster sessions. hang their posters on the morning of 23 January; they will present their posters at lunchtime and on the evening of 23 January. Posters should be removed at the end of the meeting. The Societies cannot be held responsible for lost or damaged posters.

25 Facilities Information

Accommodation Delegate List Accommodation is not included in the A delegate list will be sent by email to all registration fees; all delegates are advised attendees after the conference. Please to make their own accommodation note that this list is intended for use only to arrangements. promote networking between scientists. You do not have permission to use this list for Parking any other purpose, and any other use may There is no parking available at Charles infringe the Data Protection Act 1998. The Darwin House; delegates are advised to use list contains the names and affiliations of all public transport services where possible as attendees. Contact details are included only parking places in central London are limited for attendees who gave their permission and subject to high charges. during the registration process. Bars Liability There are many pubs and bars available in The Biochemical Society, British Ecological close proximity to Charles Darwin House; for Society and Society for Experimental Biology further details please visit the Registration will assume no responsibility whatsoever Desk. for damage or injury to persons or property during the meeting. Participants are advised Medical Services to arrange their own personal travel and Please contact the Registration Desk in the health insurance. case of a minor incident; if it is an emergency, Tweeting and Blogging please call 999 directly and inform staff at the Registration Desk. The societies encourage the discussion of its conferences via Twitter, Facebook and Internet Access similar social networks. In order to promote Wireless internet access is available free of discussion and the exchange of information, charge in all areas of Charles Darwin House. delegates who wish to tweet are asked to use To access our wireless network, within your the joint meeting hash tag: #Bees2014. available wireless networks select SSID: Speakers will be made aware of this policy CDH; when prompted for a password, and have the right to ask delegates not to enter time2work. In case of any connection disseminate their research via the internet; difficulties, please visit the Registration Desk. if a Speaker makes this request, delegates are asked not to discuss the relevant work Shopping and Banking Facilities in this way. Delegates are respectfully asked There are numerous shopping and banking to refrain from communicating using mobile facilities available in close proximity to devices whilst lectures are in progress. Charles Darwin House, for further details please visit the Registration Desk. Photography Please note that photographs taken at this Certificates of Attendance event may be used for promotional purposes Attendees requiring a Certificate of by any of the three societies, e.g. by inclusion Attendance for the meeting should contact on our websites and/or marketing materials. the Registration Desk. If you have any concerns or queries regarding this, please visit the Registration Desk.

26 Useful Contacts

Charles Darwin House Reception +44(0) 20 7685 2400

Airports Heathrow Airport (www.HeathrowAirport.com) +44 (0) 870 000 0123 Lost Property +44 (0) 20 8745 7727 Heathrow Express (www.HeathrowExpress.com) 0845 600 1515 (UK only)

Gatwick Airport (www.GatwickAirport.com +44 (0) 870 000 2468 Lost Property +44 (0) 870 000 2468 Gatwick Express (www.GatwickExpress.co.uk) +44 (0) 845 850 1530 Gatwick Airport Travel Line 0870 574 7777 (UK only)

Stansted Airport (www.StanstedAirport.com) +44 (0) 870 000 0303 Lost Property +44 (0) 1279 663 213 Stansted Express (www.StanstedExpress.com) +44 (0) 845 850 0 150

London City Airport (www.LondonCityAirport.com) +44 (0) 20 7646 0088 Lost Property +44 (0) 20 7646 0088

London Luton (www.London-Luton.co.uk) +44 (0) 1582 405 100 Lost Property +44 (0) 1582 405 100

Manchester Airport +44 (0) 161 489 3000

Local Taxi Services Bloomsbury Cars +44 (0) 20 7379 0555 Taxis are very common around the venue and can be hailed from the pavement; please only use licensed taxis.

Other Transport Services Transport for London (www.tﬂ.gov.uk) +44 (0) 20 7222 1234 National Express Coaches (www.NationalExpress.com) +44 (0) 990 80 80 80 National Rail Enquiries (www.NationalRail.co.uk) +44 (0) 8457 48 49 50 Eurostar Reservations (www.Eurostar.com) +44 (0) 20 8882 4412

27 From: Pywell, Richard F. Sent: 21 March 2014 15:47 To: @syngenta.com; @bayer.com Cc: @bayer.com; @syngenta.com'; @SYNGENTA.COM'; ; Bullock, James M.; @bayer.com Subject: Final experimental design scenarios for the study of impacts of Neonicotinoid seed treatments on bees foraging on oilseed rape

Dear ,

Re: Final experimental design scenarios for the study of impacts of Neonicotinoid seed treatments on bees foraging on oilseed rape

Summary: following the meeting of the 17/3/2014 we outline below a series of potential experimental design scenarios for single and multiple countries intended to assess the impacts of Neonicotinoids on bees foraging on oilseed rape. This will ultimately be linked to scenario specific cost assessments and power analysis, however, following on from a request from we will provide an initial overview of the assumptions and outlines of each scenario to limit the scope of this process.

Action Bayer/Syngenta: Please could you urgently review these design scenarios and let us know if you are happy for us to proceed with the statistical power analysis and cost-feasibility assessment by COP Monday 24 March. If we hear nothing we shall assume you are happy for us to proceed.

Note: To date we estimate the project feasibility analysis has taken in excess of 40 staff days and we would consider the work outlined below to be the final exercise in this process.

Basis of scenario development:

1. Sub-contracting: Field trials in countries other than the UK will not be undertaken by CEH for logistical reasons and will need to be sub-contracted to a pan-European specialist field trials company (e.g. Eurofins). 2. Site costs: The average cost of establishing and monitoring a single site will be raised to to account for the higher than anticipated costs associated with sub-contracting sampling to third parties (per diem costs were thought to be 20% higher than the CEH field team at the 17/3/14 meeting). These figures will be refined, if possible, by incorporating accurate per diem costs from Eurofins (action for ). However, some degree of uncertainty as to the actual cost of the project will remain until the experimental design is finalised and negotiations with contractors have been concluded 3. Artificial Bombus terrestris colonies: Syngenta and Bayer have suggested that the study should focus on just Honeybees to reduce overall costs. CEH felt that this considerably reduced the scope of the study while making only a minimal saving in overall costs that would not in themselves translate into additional replication. In the final scenario document we will present scenario costs based on a) Apis only and b) Apis and Bombus colonies.. 4. Independent Scientific Advisory Group (SAG): The SAG (Chair Prof. Charles Godfray, FRS) was set up to ensure both scientific rigour and independence and their involvement is crucial to the credibility of the study in the wider scientific community. We must stress that the final decision on the experimental design will have to be made following consultation with the SAG. 5. Effect size: Following on from the meeting of the 17/3/2014 it is currently accepted that the 7 % effect size detection rate (probability of detection at β=0.8) as proposed by EFSA (EFSA Journal 2013, 11:3295:266.) is not practical. However, discussions within CEH have defined an acceptable effect size for detecting population level changes to be no less than 20 % (subject to discussions with the SAG). We consider this to be the minimum defendable effect size for detecting changes in bee populations at field scales. We emphasise however that there is no universally accepted effect size for this kind of study and that what we propose is based on practical experience of running similar landscape scale studies. 6. Population response parameters: We will produce a two tier power analysis considering different subsets of response variables describing bee populations. The first tier will focus on core measures of honeybee populations that we consider to be the basic response variables of the study. These response variables will be used for assessing whether replication allows a minimum 20 % effect size detection rate. These response variables will be: 1) Overwintering colony strength; 2) Peak honeybee colony strength / rate of increase in colony strength; 3) colony weight. A second tier will include remaining commonly measured response variables for honeybees and bumblebees considered to be of secondary importance, e.g. percentage mean area of nectar cells or egg stage on combs. These will be listed in a table of samples replication × response variable and can be used to judge the suitability of specific scenarios for detecting a wider range of bee population measures. 7. Power analyses for multiple countries: There is no detailed information on between country variation in bee population parameters derived using comparable methods over comparable time periods. For this reason the power analysis will give a range of required replication for scenarios 2 and 3 which consider multiple countries. This will be based on a worst case scenario (where each country must be analysed independently) and a best case scenario (where each replicate block from all countries can be included in the same analysis).

Scenarios

We outline three scenarios. In each case we propose to provide detail of the degree of replication the current funding of will allow and to state whether this reaches the minimum 20 % effect size detection rate. Note that if a 20 % effect size cannot be detected the CEH position is that the study is not scientifically robust. In addition, we will provide an estimate of the replication required to reach the 20% detection threshold and the costs of this.

Scenario 1 (current design): Single country comparing Control, Thiamethoxam and Clothianidin (a) Apis only; b) Apis+Bombus)  To be undertaken in UK only.  Replicate blocks will comprise three sites separated by c. 10 km treated respectively with Control, Thiamethoxam and Clothianidin treated oilseed rape at c. 50 ha per farm.  Replicate blocks sufficient to detect a 20 % effect size will be determined and linked overall project costs.  Due to reduced running costs linked with a single country replication will be split across two years.

Scenario 2: Three countries comparing Control, Thiamethoxam and Clothianidin (a) Apis only; b) Apis+Bombus)  To be undertaken in three countries, one of which will be the UK.  Replicate blocks will comprise three sites separated by c. 10 km treated respectively with Control, Thiamethoxam and Clothianidin treated oilseed rape at c. 50 ha per farm.  These blocks will be replicated within each country at a level to be determined by i) the current budget ( , and ii) to achieve the 20% detection threshold.  Due to increased costs linked with multiple countries this study will be limited to a single year.

Scenario 3: Four countries comparing Control vs one neonicotinoid in each country (Thiamethoxam or Clothianidin) (a) Apis only; b) Apis+Bombus)  To be undertaken in four countries, one of which will be the UK.  Replicate blocks will comprise two sites separated by c. 10 km treated respectively with Control or either Thiamethoxam or Clothianidin treated oilseed rape at c. 50 ha per farm.  Two countries will be randomly allocated to testing Thiamethoxam and the remaining two to testing Clothianidin. The aim of this approach is to maximise the number of replicate blocks and so increase the statistical power of the study.  These blocks will be replicated within each country at a level to be determined by i) the current budget ( ), and ii) to achieve the 20% detection threshold.  Due to increased costs linked with multiple countries this study will be limited to a single year.

We look forward to hearing from you in due course.

Best wishes

Richard

:01491 692356

: [email protected]

From: Pywell, Richard F. Sent: 28 March 2014 15:01 To: @syngenta.com; @bayer.com; @bayer.com; @SYNGENTA.COM'; @syngenta.com' Cc: Bullock, James M.; @bayer.com Subject: NNI Experiment final power analysis and feasibility study

Dear ,

As promised I attach a final power analysis and feasibility study of the three NNI experimental design options discussed at our meeting on 17 March 2014.

Please also be aware that once we have agreed the final design there are a number of factors that will affect the start date of the experiment proper, including:

1) Consultation with the Scientific Advisory Group 2) Obtaining permits for the experimental treatment of crops from host countries 3) Finding suitable subcontractors and agreeing terms 4) Finding suitable sites etc

I look forward to your response.

Best wishes

Richard

:01491 692356

: [email protected]

PA(s):

Re: Final experimental design scenarios for the study of impacts of Neonicotinoid seed treatments on bees foraging on oilseed rape

Note: To date we estimate the project feasibility analysis has taken in excess of 40 staff days and we would consider the work outlined below to be the final exercise in this process.

Justification of a 20 % effect size detection threshold The specific protection goals defined by the European Food Standards Agency in their bee risk assessment of plant protection products require that studies should have sufficient replication to identify a 7 % detrimental effect on bee colony sizes (to be detected with an 80 % confidence in a one-tailed test) (EFSA 2013). To put this into perspective, small, medium, and large effect sizes have conventionally been defined as 20, 50 and 80 % by Cohen (1992) – a key work on power analysis – and so the likelihood of detection of 7% changes in colony sizes are small. The exact origin of the requirement to be able to detect a 7 % effect size is unclear from the EFSA report. However, in the comprehensive meta-analysis of both lethal and sub-lethal impacts of the Neonicotinoid Imidacloprid, Cresswell (2011) demonstrated that field realistic exposure rates for honeybees in oilseed rape (acute dose applied at a single time point = 0.023 – 0.03 ng; chronic dose applied over several time periods = 0.7 1.3 μg L-1) would have sub-lethal effects on honeybees that reduced performance by between 6 and 11 %. These sub-lethal effects included impacts on gustatory thresholds, success in returning to colonies and learning and memory. Based on this, Cresswell (2011) makes the reasonable assertion that the replication of studies aiming to detect non-lethal effects of Neonicotinoids on bees should be sufficient to detect such small effect sizes, i.e. in the range of 6 – 11 %. While the nature of sub-lethal effects of Neonicotinoids on honeybees may be relatively subtle and so warrant such a low detection rate, the inherent variability in honeybee colony sizes under real world conditions may mean that the replication to achieve this detection rate is unfeasible.

Current cost estimates for field scale studies under a limited sampling strategy are c. per site. For many basic measures of honeybee colony size detecting a 7 % effect size as proposed by EFSA would require replication of pairs of control and Neonicotinoid treated sites in excess of 50 replicate blocks, with a cost of the trials of at least However, the effect sizes for sub-lethal effects identified by Cresswell (2011) focus principally on immediate detrimental impacts on performance. The propagation of these sub-lethal effects throughout the exposure period of honeybees foraging on oilseed rape crops in a season would be expected to result in much greater impacts at the colony level. Indeed, the importance of considering longer-term consequences of sub lethal effects is highlighted by a recent influential study by Henry et al. (2012). In this work, sublethal impacts on honeybees exposed to realistic field doses of Neonicotinoids (Thiamethoxam) were demonstrated to cause a doubling of mortality rates when workers were foraging on oilseed rape, an impact attributed to homing failure. By modelling this increase in forager mortality, Henry et al. showed that these sub-lethal effects would result in a 22% reduction in honeybee colony size by the end of the oilseed rape flowering period, even under ideal conditions of low worker exposure rates and high queen fecundity. Where worker exposure rates to Neonicotinoids are higher and queen reproductive capacity lower, colony size could fall by as much as 72 % over this same period. Such propagation of sub-lethal effects over longer exposure periods found under field conditions are likely to result in differences in colony size that are detectable by the more realistic levels of replication which can be applied in field scale studies. Based on the Henry et al. study we conclude that the ability to detect a <20 % effect size on honeybee colony size is a realistic level for identifying field scale impacts of Neonicotinoids on honeybees.

References: Cohen, J. 1992. A power primer. Psychological Bulletin 112 155-159; Cresswell, J. E. 2011. A meta-analysis of experiments testing the effects of a neonicotinoid insecticide (imidacloprid) on honey bees. Ecotoxicology 20:149-157; EFSA. 2013. Guidance on the risk assessment of plant protection products on bees (Apis mellifera, Bombus spp. and solitary bees). EFSA Journal 11:3295:266; Henry, M., et al 2012. A common pesticide decreases foraging success and survival in honey bees. Science 336:348-350.

Basis of scenario development:

1) Sub-contracting: Field trials in countries other than the UK will not be undertaken by CEH for logistical reasons and will need to be sub-contracted to a pan-European specialist field trials company (e.g. Eurofins). 2) Site costs: The average cost of establishing and monitoring a single site has been estimated at based on sub-contracting sampling to third parties (assuming per diem Eurofin staff costs). However, some degree of uncertainty as to the actual cost of the project will remain until the experimental design is finalised and negotiations with contractors have been concluded. For this reason some modifications of experimental design may occur. 3) Artificial Bombus terrestris colonies: Syngenta and Bayer have suggested that the study should focus on just Honeybees to reduce overall costs. Following our original experimental design we describe all scenarios without B. terrestris sampling (set at per site). CEH believe that this reduces considerably the scientific scope of the study and while we appreciate the potential saving in overall costs, we are concerned about the impact on the merit of the experiment. For this reason we also present scenario 2c (our preferred option) which considers reduced sampling effort for the B. terrestris colonies. Whilst reducing the scope of the original experimental design this retains both honeybees and B. terrestris in the study for a minimal additional site cost ( per site). 4) Independent Scientific Advisory Group (SAG): The SAG (Chair Prof. Charles Godfray, FRS) will be set up to ensure both scientific rigour and independence and their involvement is crucial to the credibility of the study in the wider scientific community. It is vitally important to consult with the SAG. Their input may result in changes to the experimental design. 5) Effect size: Following on from the meeting of the 17/3/2014, and the argument presented above, it is currently accepted that the 7 % effect size detection rate (probability of detection at β=0.8) as proposed by EFSA (EFSA Journal 2013, 11:3295:266.) is not practical. However, discussions within CEH have defined an acceptable effect size for detecting population level changes to be no less than 20 % (subject to discussions with the SAG). We consider this to be a key defendable threshold effect size for detecting changes in bee colony strengths at field scales. We emphasise, however, that there is no universally accepted effect size for this kind of study and that what we propose is based on practical experience of running similar landscape scale studies and the justification detailed above. 6) Population response parameters: We produce a two tier power analysis considering different subsets of response variables describing bee populations. The first tier focuses on core measures of honeybee populations that we consider to be the basic response variables of the study. These response variables are used to assess whether replication allows a minimum 20 % effect size detection rate. These response variables are: 1) Overwintering colony strength; 2) Peak honeybee colony strength / rate of increase in colony strength; 3) colony weight. A second tier includes other commonly measured response variables for honeybees and bumblebees considered to be of secondary importance, e.g. percentage mean area of nectar cells or egg stage on combs. These are listed in a table of sample replication × response variable and can be used to judge the suitability of specific scenarios for detecting a wider range of bee population measures. Note that some response parameters are inherently variable and so. For example rate of increase in colony strength over the first 5 weeks of monitoring (considered to be a core response variable) is hugely variable between hives, with some hives even showing decreases in colony strength over this time period. This variability is reflected in the very high degrees of replication required to identify 20 % effect sizes. The same issue is also seen for several other response variable considered for both honeybees and B. terrestris colonies. 7) Power analyses for multiple countries: There is no detailed information on between-country variation in bee population parameters derived using comparable methods over comparable time periods. For this reason it is not possible to do a formal power analysis for a multi-country, multi-treatment experiment. However, it is possible to give a range of required replication for scenarios 2 and 3 which consider multiple countries. This is based on a worst case scenario (where each country must be analysed independently) and a best case scenario (where each replicate block from all countries can be included in the same analysis).

Scenarios

We outline three scenarios. In each case we provide detail of the degree of replication, which population level parameters for honeybees and bumblebees can be detected for a <20 % effect size with the replication possible for the current funding Table 1 provides an overview of the potential for different scenarios to detect a minimum 20% effect size, and summarises the costs for each strategy.

SCENARIO 1 (current proposed design): Single country comparing Control, Thiamethoxam and Clothianidin a) Honeybee & Bombus; b) Honeybee only.

TWO YEARS OF SAMPLING

 To be undertaken in UK only.  Replicate blocks will comprise three sites separated by c. 10 km treated respectively with Control, Thiamethoxam and Clothianidin treated oilseed rape at c. 50 ha per farm.  Due to reduced running costs linked with a single country replication will be split across two years.

This scenario is included for reference, although following the request of Syngenta / Bayer the study needs to be undertaken in at least three countries.

UK only S1a: Honeybees & Bombus S1b: Honeybees only Cloth Thiam Con PLUS

S1a: Honeybees and B. terrestris colonies Replication: 10 replicate blocks of three sites. Note this is for a single sampling year, or could be split across two years. Response variables detectable with < 20% effect sizes: The core Tier 1 response variables of honeybee peak colony strength, overwintering colony weight and overwintering colony strength can be detected with a <20 % effect size. In addition, Tier 2 response variables of mean colony strength, mean colony weight and overwintering % empty cells for honeybees and estimated worker numbers for the B. terrestris colonies could also be detected at this level. Costs:

S1b: Honeybees only Replication: 12 replicate blocks of three sites. Note this is for a single sampling year, or could be split across two years. Response variables detectable with < 20% effect sizes: The core Tier 1 response variables of honeybee peak colony strength, overwintering colony weight and overwintering colony strength can be detected with a <20 % effect size. Tier 2 response variables also detectable at this level are mean colony strength, mean colony weight, overwintering % of nectar cells and overwintering % empty cells for the honeybees, and estimated worker numbers for the B. terrestris colonies. Costs:

SCENARIO 2: Three countries comparing Control, Thiamethoxam and Clothianidin a) Honeybee & Bombus; b) Honeybee only.

ONE YEAR OF SAMPLING ONLY

To be undertaken in three countries, one of which will be the UK.  Replicate blocks will comprise three sites separated by c. 10 km treated respectively with Control, Thiamethoxam and Clothianidin treated oilseed rape at c. 50 ha per farm.  Due to increased costs linked with multiple countries this study will be limited to a single year.

UK S2a: Honeybees & S2b: Honeybees FR S2a: Honeybees & S2b: Honeybees DE S2a: Honeybees & S2b: Honeybees Bombus only Bombus only Bombus only

PLUS PLUS PLUS

S2a: Honeybees and B. terrestris colonies Replication: in each of the three countries three replicate blocks of three sites, with an additional replicate block in one country only (a total of 10 replicate blocks). Note this is for a single sampling year. Response variables detectable with < 20% effect sizes (blocks within a single country n=3): For the honeybees the only core Tier 1 detectable with a <20% effect size is overwintering colony weight. Tier 2 response variables of mean colony weight and overwintering % empty cells for the honeybees and estimated worker numbers for the B. terrestris colonies are also detectable with a <20 % effect size. Response variables detectable with < 20% effect sizes (across all blocks in all countries n=10): For the honeybees Tier 1 response variables of peak colony strength, mean colony weight, overwintering colony weight and overwintering colony strength can be detected with a <20 % effect size. Additional Tier 2 response variables are mean colony strength and overwintering % empty cells for the honeybees and estimated worker numbers for the B. terrestris colonies. Costs:

S2b: Honeybees only Replication: in each of the three countries four replicate blocks of three sites (a total of 12 replicate blocks). Note this is for a single sampling year. Response variables detectable with < 20% effect sizes (blocks within a single country n=4): For the honeybees Tier 1 response variables of overwintering colony weight as well as Tier 2 variables of mean colony weight and overwintering % empty cells could be detected at a 20 % effect size. Response variables detectable with < 20% effect sizes (across all blocks in all countries n=12): For the honeybees core Tier 1 response variables of peak colony strength, overwintering colony weight, overwintering colony strength could be detected at a <20% effect size. In addition the Tier 2 variables of mean colony weight, mean colony strength, overwintering % nectar cells and overwintering % empty cells could also be detected at this level. Costs:

S2c: Honeybees and limited B. terrestris sampling.

We propose a third alternative which includes a reduced level of sampling effort for the B. terrestris colonies. This will have a single site cost of , only slightly higher than the Honeybee only sampling strategy with a site cost of Field work for B. terrestris would involve initial colony weights, locating hives on sites and freezing and posting to CEH in the UK only. This dramatically reduces overall costs while significantly increasing the scientific scope of the study over a honeybee only approach.

THIS IS THE PREFERED OPTION FOR CEH AS: 1) IT REMAINS WITHIN BUDGET; 2) ENSURES GOOD REPLICATION WITHOUT COMPROMISING THE TRIPLICATE (control, Thiamethoxam and Clothianidin) EXPERIMENTAL DESIGN; AND 3) INCREASES THE TAXONOMIC BREADTH OF STUDY AND SO THE SCIENTIFIC SCOPE AND RELEVANCE OF OUTPUTS.

Replication: in each of the three countries four replicate blocks of three sites (a total of 12 replicate blocks). Note this is for a single sampling year. Response variables detectable with < 20% effect sizes (blocks within a single country n=4): For the honeybees Tier 1 response variables of overwintering colony weight as well as Tier 2 variables of mean colony weight and overwintering % empty cells could be detected at a 20 % effect size. For the B. terrestris colonies we would only be able to detect estimated worker numbers with at least a 20 % effect size. Response variables detectable with < 20% effect sizes (across all blocks in all countries n=12): For the honeybees core Tier 1 response variables of peak colony strength, overwintering colony weight, overwintering colony strength could be detected at a <20% effect size. In addition the Tier 2 variables of mean colony weight, mean colony strength, overwintering % nectar cells and overwintering % empty cells could also be detected at this level. For the B. terrestris colonies we would only be able to detect estimated worker numbers with at least a 20 % effect size. Costs:

SCENARIO 3: Four countries comparing Control vs one neonicotinoid in each country (Thiamethoxam or Clothianidin) a) Honeybee & Bombus; b) Honeybee only.

ONE YEAR OF SAMPLING ONLY

To be undertaken in four countries, one of which will be the UK.  Replicate blocks will comprise two sites separated by c. 10 km treated respectively with Control or either Thiamethoxam or Clothianidin treated oilseed rape at c. 50 ha per farm.  Two countries will be randomly allocated to testing Thiamethoxam and the remaining two to testing Clothianidin. The aim of this approach is to maximise the number of replicate blocks and so increase the statistical power of the study.  Due to increased costs linked with multiple countries this study will be limited to a single year.

Note on experimental design: this experimental design has some advantages in that it allows for greater overall replication of blocks and would be applied across four countries. However, as each chemical will be applied in only two of the four countries there will be a confounding effect of country and chemical. This means that the advantages of overall increased replication are likely to be lost.

UK S3a: Honeybees & S3b: Honeybees FR S3a: Honeybees & S3b: Honeybees DE S3a: Honeybees & S3b: Honeybees PL S3a: Honeybees & S3b: Honeybees Bombus only Bombus only Bombus only Bombus only

Cloth Thiam

Con PLUS Con PLUS PLUS PLUS

S3a: Honeybees and B. terrestris colonies Replication: In three of the four countries four replicate blocks of two sites, with one country with only three replicate blocks (a total of 15 replicate blocks). Note this is for a single sampling year and assumes two countries would be treated with Clothianidin, and two countries would be treated with Thiamethoxam. Response variables detectable with < 20% effect sizes (blocks within a single country n=4: The only Tier 1 response variables detectable at a <20% effect size is overwintering colony weight, although the Tier 2 response variables of honeybees mean colony weight and overwintering % empty cells as well as B. terrestris estimated worker numbers can also be detected at this level. Response variables detectable with < 20% effect sizes (across all blocks in all countries n=15): Tier 1 response variables detectable with a <20% effect size are honeybee peak colony strength, overwintering colony weight and overwintering colony strength. Tier 2 response variables are honeybee mean colony strength, mean colony weight, overwintering % nectar cells and overwintering % empty cells. Also detectable with a < 20 % effect size are maximum cumulative worker count and estimated worker numbers for the B. terrestris colonies. Costs:

S3b: Honeybees only Replication: in each of the four countries five replicate blocks of two sites (a total of 20 replicate blocks). Note this is for a single sampling year and assumes two countries would be treated with Clothianidin, and two countries would be treated with Thiamethoxam. Response variables detectable with < 20% effect sizes (blocks within a single country n=5): For the honeybees Tier1 response variables of overwintering colony weight could be detected at for a <20% effect size, as well as the Tier 2 variables of mean colony weight, mean colony strength, and overwintering % empty cells. Response variables detectable with < 20% effect sizes (across all blocks in all countries n=20): For the honeybees Tier 1 response variable detectable for a <20% effect size are peak colony strength, overwintering colony weight and overwintering colony strength. Tier 2 response variables also detectable with this same threshold effect size are mean colony strength, mean colony weight, overwintering % nectar cells, overwintering % pollen cells and overwintering % empty cells could be detected with a < 20 % effect size. Costs: (Note this is over budget but may be viable). Otherwise one replicate block would need to be lost in one country bringing costs to £ and an overall number of replicate blocks of n=19 across all four countries.

Number of replicate blocks (n) n for n for Variance 3 4 5 6 7 8 9 10 12 15 20 7 % ES 20 % ES τ2 σ2 Honeybees Peak colony strength (T1) >20 >20 >20 19.9 18.6 17.5 16.5 15.8 14.5 13.1 11.5 55 6 0.013 0.053 Rate of colony strength increase (T1) >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 765 80 0.178 0.742 Overwintering colony weight (T1) 15.9 13.9 12.5 11.5 10.7 10 9.5 9 8.3 7.5 6.5 17 2 * 0.004 0.016 Overwintering colony strength (T1) >20 >20 >20 >20 >20 20 18.9 18 16.6 15.0 13.2 75 20 0.017 0.075 Mean colony strength >20 >20 20 18.5 17.4 16.3 15.5 14.7 13.5 12.2 10.7 51 5 0.012 0.049 Dead bees >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 242 26 0.076 0.137 Counts on OSR >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 264 28 0.096 0.083 Colony weight average 17.2 15.2 13.7 12.6 11.7 11 10.4 9.9 9.1 8.2 7.1 20 3 0.004 0.023 Overwintering % nectar cell area >20 >20 >20 >20 >20 >20 >20 >20 19.8 17.9 15.7 112 12 0.040 0.038 Overwintering % pollen cell area >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 348 37 0.089 0.298 Overwintering % empty cell area 18.5 16.3 14.7 13.5 12.5 11.8 11.2 10.6 9.7 7.6 7.6 24.8 3 0.009 0.008 Bumblebee colonies (B. terrestris) Maximum cumulative worker count >20 >20 >20 >20 >20 >20 >20 >20 18.9 17.0 14.9 99 11 0.019 0.163 Worker gain >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 388 41 0.112 0.373 Proportional worker gain >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 205 22 0.068 0.136 Estimated worker numbers 12.5 11 9.9 9.1 8.5 8 7.5 7.1 <7 <7 <7 10 2* 0.003 0.009 Days till queen death >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 >20 413 44 0.107 0.484

Key expenditure costs

* Note a min. of three replicate required; †Assuming 5 × honeybee and 7 × B. terrestris colonies; n = 20% is the number of blocks to find 20% effect size. Table 1. Power analyses indicating the effect size that can be detected for a range of commonly-measured population-level parameters recorded as part of honeybees and artificial Bombus terrestris colony monitoring for a given number (n) of replicate blocks. A replicate block may contain three sites for scenarios 1 and 2 (control, Clothianidin and Thiamethoxam) or two sites for scenario 3 (control and either Clothianidin or Thiamethoxam). We summarise total cost (CEH costs and individual site costs) for the difference levels of replication (n) assuming either two or three sites in a block with sampling considering all bees (honeybees and B. terrestris) or honeybees only. Where total costs exceed the £ budget cells are shaded. Variance parameters are determined from data within Pilling et al. (2013: PLoS ONE 8:e77193) and Carvell et al. (2008: Ecological Entomology 33:321-327) and focus on oilseed rape associated bee colonies only, where τ2 = between site variance in response parameter, equivalent to the CV2 and σ2 = within site between colony variance, equivalent to ln(1+CV2). Power analysis is based on a modification of that proposed by EFSA (2013: EFSA Journal 11:3295:266). Where effect sizes exceed the minimum 20 % detection rate (indicated by >20) the required number of replicate blocks to detect both a 7 (column ‘n for 7 % ES’) and 20 % effect size (column ‘n for 20 % ES’) are given. Note that all measures of colony strength for honeybees are recorded using the Liebefeld method and the rate of increase in colony strength is based on a Pearsons correlation coefficient of the increase in Liebefeld colony strength over the first 5 weeks of monitoring in a given year. Variables with the suffix T1 are considered to be core Tier 1 parameters of key importance in assessing honeybee responses to Neonicotinoids. From: Pywell, Richard F. Sent: 30 April 2014 18:12 To: @syngenta.com Cc: ; ; @SYNGENTA.COM; ' '; @syngenta.com'; ; Bullock, James M. Subject: NNI Experiment - update on meeting with and other actions

Dear ,

James and I met over dinner on 28th April following his return from China the day before.

Below is a summary of our discussion:

1) was unaware of any issues/conflicts raised by Defra regarding him chairing the Defra Pollinator Expert Group and chairing the NNI Project Scientific Advisory Group; 2) He is supportive of our stance that the 20% effect detection achieved by the current experimental design is realistic/reasonable; 3) He regards studying the impacts of NNIs on native pollinators to be of equal importance to that of honeybees – he is concerned that this is not omitted from the study but accepts that other sources of funding may have to be sought to achieve this; 4) He would like to now convene a small Scientific Advisory Group and make formal comment on the experimental design. However, he is very busy at present so not much will happen for 2-3 weeks.

Other actions:

and I will engage with Eurofins regarding the feasibility of establishing the experiment in three EU countries this year. We will also ask them to provide a firm quotation for undertaking the field experiment as outlined in the recent project documents. Once we have had a response we will set up a face-to-face meeting with them. At the same time we will also seek a quote from a rival company.

Best wishes

Richard

':01491 692356

*: [email protected]

______

From: Pywell, Richard F. Sent: 06 June 2014 17:47 To: @SYNGENTA.COM; @bayer.com Cc: @SYNGENTA.COM'; ; Bullock, James M.; @SYNGENTA.COM Subject: NNI experiment - update on costs and sites

Dear ,

We now have written quotations from Eurofins for locating sites and monitoring the NNI experiment.

1) Based on these we estimate there is a shortfall on the monitoring budget. We consider this reasonable given we did not know the experimental parameters when the budget was set back in August 2013! 2) Eurofins also recommend a contingency fund is provided for farmer compensation of around (e.g. for additional management or to pay a small premium for the crop to encourage participation). This further increases the budget shortfall. 3) consider finding suitable sites for the project to be a major challenge. We therefore recommend that we review the situation weekly through July and make a final decision on viability of the trial in early August. 4) CEH are in the process of providing targeting maps to narrow the search area in each country based on land cover, proportion of oilseed rape and wheat grown. The maps will be ready next week. CEH will also provide a protocol for assessing site suitability. 5) To support Eurofins in site location we would ask that Bayer and Syngenta field teams are mobilised in the three countries. Thank you already for the contacts you have sent. These have been forwarded to Eurofins.

Best wishes

Richard

:01491 692356

: [email protected]

From: Pywell, Richard F. Sent: 11 June 2014 11:52 To: @SYNGENTA.COM; ; @SYNGENTA.COM; @SYNGENTA.COM Cc: '; @SYNGENTA.COM; @bayer.com'; @bayer.com'; @bayer.com'; .; Bullock, James M.; Subject: Locating sites for the NNI experiment in Scotland, Germany and Hungary

Dear ,

CEH have now developed a protocol for the broad targeting of regions suitable for the NNI field experiment in Scotland, Germany and Hungary.

The criteria used for selecting regions are as follows:

 Scottish regions are the most limiting in terms of total number and number with enough cover of arable land (but also have the known estates growing OSR), so these formed the baseline with which to find matching regions in Hungary and Germany  >30% cover of arable across region – so not the most intensively farmed landscapes  > Mean 3% wheat cover per ~10km cell in region  Mean 2.5% - 5% OSR cover per ~10km cell in region  <10% cover of urban across region  Less than 3% cover of permanent crops (orchards, vines etc) across region  <25% cover of forest across region

Regions (or sub-regions) which match these criteria and look similar in landscape configuration:

COUNTRY REGION NOTES Scotland Aberdeenshire (NE) Plentiful arable land on coastal plain, avoid moorland inland Scotland East/Mid Lothian Plentiful arable land on coastal plain, avoid moorland inland Scotland Fife Known OSR estate at Cupar Scotland Perthshire Below criteria for %arable land but known OSR estate at Balbeggie; Arable in valleys, avoid high moorland to West. Scotland Scottish borders Plentiful arable land on coastal plain, avoid moorland inland. Known OSR estate at Coldstream Hungary Veszprem Arable land around Pápa. Avoid forests and grassland further south Hungary Vas Hungary Tolna Avoid orchard/vineyard regions where possible Hungary Somogy Avoid orchard/vineyard and wetland along shores of lake Balaton Hungary Gyor-Moson-Sopron Hungary Fejer Germany Hannover Arable land to North nearer Bremen. More forested and built up around Hannover itself Germany Luneberg Very mixed arable/pasture/forest but good equivalent to Scottish regions as coastal. Germany Weser-Ems Some arable on coast, and more further inland Germany Magdeburg Forested in centre but big tracts of arable to North and around Magdeburg itself

We have produced maps detailing the specific areas of the chosen regions which we believe meet the criteria – I will send this large file to you using FTP (this works like Dropbox and you will receive an e-mail link to the file).

PLEASE COULD YOUR LOCAL FIELD TEAMS LOOK AT THE CRITERIA AND REGIONS AND COMMENT ON THEIR SUITABILITY AND PRACTICALITY FOR EACH COUNTRY.

The amended maps and targeting criteria will form the basis of the first stage selection protocol we will send to Eurofins.

Obvious we would prefer it if Bayer/Syngenta teams find as many sites as possible as it will save the project a considerable amount of money.

Best wishes

Richard

:01491 692356 : : [email protected]

HOW FAR DO HONEY BEES FLY TO FIELDS OF BRASSICA NAPUS (OILSEED RAPE)?

1Osborne, J.L., Carreck, N.L. and Williams, I.H.

Plant and Invertebrate Ecology Division, IACR Rothamsted, Harpenden, Herts, AL5 2JQ, UK. 1Tel: 01582 763133 ext 2738, fax: 01582 760981, email: [email protected]

Introduction Herbicide-tolerant genetically-modified (HTGM) oilseed rape (Brassica napus L.) is one of the first GM crops to be evaluated in farmscale trials in the UK (Firbank et al. 1999). Oilseed rape self pollinates, but also benefits from cross-pollination effected by wind and insects, primarily bees (Williams et al. 1986; 1987). Small quantities of pollen will be transported by wind or bees over hundreds of metres, perhaps kilometres (Thompson et al. 1999). Debate on pollen movement from GM oilseed rape crops has focussed on a) the extent of gene flow into conventional crops and wild plants (Crawley et al. 1993; Raybould & Gray 1993; Scheffler et al. 1993; Champolivier et al. 1999; Downey 1999; Simpson et al. 1999; Ingram 2000) and b) the chances of GM pollen entering honey bee colonies and being incorporated into honey and other hive products (Ramsey et al. 1999; Treu & Emberlin 2000).

Patterns of bee movement are fundamental to predicting the extent of bee-mediated pollen movement away from oilseed rape crops. Maximum distances flown by bees from colony to forage have been used to predict pollen flow, and consequently make inferences on gene flow away from a crop (Emberlin et al. 1999). However, this is not the most important parameter, since most cross- pollination is likely to occur when individual bees, carrying pollen, fly between fields of oilseed rape on a single foraging trip. Distances flown by bees on individual trips, constancy to crops, and viability of pollen on the bees’ bodies are more important parameters in determining the extent of gene flow away from a field. In addition, there may be some in-hive pollen transfer between nest- mates leading to low levels of cross-pollination between fields (Vaissière et al. 1994; Ramsay et al. 1999), and this will depend on the colony-to-forage distance. The “colony-to-forage” distance is also important in determining the presence or absence of pollen in the honey and hive products.

We made use of a field scale trial of HTGM oilseed rape at Rothamsted to study the following: 1. Do honey bees show a preference for HTGM oilseed rape or conventional oilseed rape? 2. What distances do honey bees fly from their colonies to oilseed rape fields (i.e. mean colony-to-forage distance) in an arable landscape, typical of Eastern England?

Methods Over four weeks, we examined the spatial and temporal distribution of mass-marked honey bees, from a transect of colonies, as they foraged on flowering winter-sown oilseed rape fields, on the arable farm of Rothamsted Estate (Hertfordshire, UK). One 5 ha field, called Black Horse (BH), contained a trial of HTGM oilseed rape, which formed part of a SAPPIO LINK programme entitled “Botanical and Rotational Implications of Genetically-modified Herbicide Tolerance” (BRIGHT). The trial was testing four cultivars of winter oilseed rape and was situated over 200 m from the nearest conventional oilseed rape field (Fig 1). It had sixteen plots (24 m x 120 m), arranged in four randomised blocks. Each block contained four plots, one of each cultivar of oilseed rape. The cultivars were genetically-modified glufosinate-tolerant rape (Liberty Link, Aventis), genetically-

Proceedings of the 37th International Apicultural Congress, 28 October – 1 November 2001, Durban, South Africa APIMONDIA 2001 To be referenced as: Proc. 37th Int. Apic. Congr., 28 Oct – 1 Nov 2001, Durban, South Africa ISBN: 0-620-27768-8 Produced by: Document Transformation Technologies Organised by: Conference Planners modified glyphosate-tolerant rape (Roundup-Ready, Monsanto), non-GM Imazamox-tolerant rape (Cyanamid/Pioneer) and a conventional oilseed rape cultivar called Apex.

Four fields of conventional (non-GM) winter-sown oilseed rape, at varying distances from the GM trial and from the transect of honey bee colonies, were also surveyed (Table 1).

Table 1 Details of surveyed oilseed rape fields

Field name Approx. distance Perpendicular (and abbrev.) (m) from HTGM distance (m) from trial hive transect *Black Horse (BH) 0 12 White Horse (WH) 650 10 Great Knott (GK) 990 500-600 Appletree (AT) 1160 10 Long Hoos (LH) 1650 600 *HTGM trial site

Three similar-sized honey bee (Apis mellifera L.) colonies were placed at each of six positions (A – F) along a transect (total = 18), running East from the GM trial (Fig 1; Table 1). Each colony was fitted with a pollen trap to sample pollen loads from returning foragers, and a powder dye dispenser to automatically mark bees leaving the colony. Six fluorescent, non-toxic powder dyes (from Radiant Color NV, Belgium and Sterling Industrial Colours, London, UK) were used to mark the bees. One colour of powder dye was used for all three colonies at each position along the transect (Table 2) so that a bee’s origin was apparent when it was observed foraging on a field.

Table 2 Details of colony sites along transect (see also Figure 1)

Colony Distance (m) from HTGM Dye colour site trial used A 12 magenta B 304 green C 484 red D 662 pale orange E 892 dark orange F 1302 yellow

The pollen traps collected pollen for 24 h from Monday to Tuesday morning, each week for four weeks. On each Tuesday morning, the pollen from each trap was emptied into a clean plastic bag and frozen until analysis. Using a microscope for identification, the percentage of oilseed rape pollen present in each sample was determined.

After each pollen trap had been emptied, approximately 8 g of powder dye were placed in the dye dispenser on the colony. Standard searches were made for marked bees, along fixed transects, on the GM trial (Black Horse) and the four conventional rape fields, on Tuesday and Wednesday of each week. The order in which fields were surveyed for marked bees was random, as was the order of plot surveys on Black Horse. Bee surveys were performed in warm, sunny conditions, between 10.00h–16.30h, when bee activity was at its peak. On Black Horse, observers walked, at a steady pace, down the central tractor-wheelings on each 120 m long plot, examining a one-metre-wide strip to one side. The numbers of honey bees marked with each powder dye colour were scored. On each of the four conventional oilseed rape fields, eight transects, each of 50 m in length, were spread out across the field, and walked as described above in search of marked bees. To estimate forager density, all foraging bees (marked and unmarked honey bees, bumble bees and solitary bees) were counted on each transect, whilst searching for marked bees, every Tuesday. The presence/absence of oilseed rape pollen in the bees’ corbiculae was also recorded.

Disposable overalls and gloves were worn to minimise observer-mediated pollen transfer between fields and between colonies.

Results The four weeks over which the experiment was conducted (26 April – 23 May 1999) were generally warm and bright with variable wind conditions (0-10 mph).

Pollen trap samples Oilseed rape pollen accounted for between 0.5% and 50.6% of the pollen collected by a colony (average per trap = 12±2% oilseed rape pollen). Even colonies positioned next to an oilseed rape field, with no obscuring vegetation between colony and crop (position A, D and F), had low percentages of oilseed rape in the pollen traps. Other pollen present was identified to be most probably from Crataegus monogyna Jacq. (hawthorn) with a small percentage from Vicia faba L. (field bean).

Forager density There were between zero and 5.3 foraging bees per 20 m2. Most were honey bees (marked and unmarked), and there were a few bumble bees (Bombus lapidarius, B.pascuorum, B.terrestris or B.lucorum and B.pratorum) and some unidentified solitary bees. Forty of the 573 honey bees counted during these density counts (7%) had oilseed rape pollen in their corbiculae. Forager densities were highest on Black Horse (1.2 ± 0.2 bees/20 m2), and on White Horse (1.7 ± 0.2 bees/20 m2), and low on the other fields, particularly Great Knott (0.6 ± 0.1 bees/20 m2) and Long Hoos (0.4 ± 0.1 bees/20 m2). Bee density increased in weeks 3 and 4, particularly on Black Horse. This reflected the flowering stages of the crops. Great Knott and Long Hoos were in full flower during week 1 of the experiment and had almost finished flowering by week 4. Apple Tree and White Horse were in full flower during week 2. Black Horse was in full flower during weeks 3 and 4, when the other fields had almost completed flowering.

On Black Horse, the total number of bees per 120 m2 transect varied from 0 to 32. A two-way anova (with randomized blocking) was performed on the transformed transect counts (log (x+1)) to compare bee density between different cultivars of oilseed rape and between weeks. The total number of bees per transect varied significantly between weeks (F3,45 = 58.6, P<0.001) but not between crop cultivars (F3,45 = 1.65, P=0.191) and there was no interaction between week and cultivar (F9,45 = 1.25, P=0.290). Considering only honey bees, density varied significantly between weeks (F3,45 = 70.42, P<0.001) and between crop cultivars (F3,45 = 4.10, P=0.012), with the highest density on the GM Liberty Link cultivar, and the lowest forager density on the GM Roundup-Ready cultivar (Fig 2). There was no interaction between week and cultivar (F9,45 = 1.42, P=0.209).

Marked bee surveys At least 90% of the bees exiting the colony were dusted with dye powder, which remained visible on the dorsal thorax, in wing joints and between abdominal segments, even when the bee groomed. Marked bees from site D (pale orange) and E (dark orange) became indistinguishable, so they were recorded together as “orange”. There were large differences in the numbers of bees of each colour observed (max = 374 magenta; min = 19 yellow), because of the different sampling effort expended on different fields, and because of differences in colony activity.

A total of 736 marked bees were observed over the four weeks. They constituted 67% of the foraging population on the oilseed rape fields. The majority of marked bees (558) were seen on Black Horse, where the sampling effort was greatest. Sixteen x 120 m transects were walked each day on Black Horse, compared to eight x 50 m transects per day on the other fields. 157 marked bees were seen on White Horse, but only 19 were seen on Apple Tree despite similar sampling efforts. No marked bees were seen on Long Hoos and only two bees, one green and one yellow, were seen on Great Knott. These two fields were therefore excluded from analysis.

When considering how far bees fly to a field of oilseed rape, one can consider either the distribution at each field and where they came from, or one can consider each colony site and which field the bees were going to. We shall do both.

Marked bee distributions varied over the four weeks , and the majority of bees came from the colony sites nearest to the field (Fig 3). 89% of marked bees observed on Black Horse were magenta or green, from sites A and B, but there was an influx of red bees, from site C, during week 4 (Fig 3a). On White Horse, 95% of marked bees were from sites C, D and E. Magenta and green bees, from sites A and B, were only seen on the field during week 2. On Apple Tree, only yellow bees (from site F) were seen in weeks 1-3, but six orange bees were seen in week 4.

Focussing on the colony sites, the mean density of each colour of marked bees per 100 m2 of transect was calculated for each field, and plotted against distance of the field from the colonies. The number of marked bees per 100 m2 observed foraging on crops declined sharply with distance from their colonies (Fig 4). Overall, 89.5% of the observed marked bees were foraging on the field closest to their colony, and 9.8% were foraging on the next closest field. Only 0.7% were observed further afield. The average observed colony-to-forage distance was 127 m. The maximum observed foraging distance was for a green marked bee, seen 955 m from her colony, on Great Knott. The maximum potential observable distance of 1770 m (the distance between colony site A and Long Hoos).

Discussion Placing honey bee colonies along a transect, and mass-marking the foragers as they left their colonies, proved to be a very effective technique for determining the distribution of bees on oilseed rape fields in the surrounding landscape. However, as with most mark-recapture studies, the area surveyed was limited, restricting the observable range of movement. Collecting pollen from traps, in combination with mark - re-observation, gave a clearer picture of colony foraging patterns. The high percentage of non-oilseed rape pollen in the traps suggests that many marked bees did not go to the oilseed rape and were, unobserved, foraging on uncropped vegetation. The high densities of honey bees on the oilseed rape were primarily collecting nectar. This may affect the amount of cross-pollination occurring, and will limit the quantity of pollen entering a honey bee colony, and subsequently the honey.

Although the density of honey bees differed significantly between cultivars of oilseed rape on Black Horse, the differences did not provide evidence that bees avoided GM cultivars. Picard-Nizou et al. (1995) examined bee behaviour on insect-resistant GM oilseed rape, and found no difference in visitation patterns. The lower density of bees on the Roundup-Ready cultivar in our experiment may be linked to this cultivar suffering low flower production due to sulphur deficiency, because of plot positions on the field.

Forager density and marked bee distribution were influenced by the change in relative attractiveness of the different fields over time. For example, in week 4, when most fields were finishing flowering and Black Horse was still in full flower, forager density on this field increased, as did the proportion of marked bees coming from more distant colonies. As the number of oilseed rape flowers declined over time, the bees were travelling further from their colonies to forage.

The oilseed rape fields attracted honey bees from colonies several hundred metres away, but the vast majority of marked bees (90%) foraged on the field closest to their colony. The foraging site that a bee chooses on leaving a colony will depend both on distance to be flown (Cresswell et al. 2000), competitive forage in the vicinity and probably on visibility of the foraging site. Only two marked bees were observed foraging on Great Knott and Long Hoos, which were distant from the colony transect. Bees from most of the colonies would have had to fly over a large beech wood to reach these fields. Apple Tree had suprisingly few marked bees, perhaps because it was adjacent to a large, mixed woodland, and garden, containing several flowering tree and herb species.

Although the bees had a choice of oilseed rape fields within foraging range, at different distances from their colonies, they were all observed within 1km of their colonies. Maximum observable distance was 1.7 km. These results do not provide information on pollen movement between fields, but they do have implications for beekeepers with colonies positioned in the vicinity of GM crops. If there are several oilseed rape crops in the area, our results suggest that it is very unlikely that bees will travel beyond the nearest one or two fields to forage. However, a tiny proportion of bees may travel long distances to feed, even if they have comparable forage nearer to home. No pale orange or yellow bees were seen foraging on Black Horse, so we predict that there is only likely to be a discernible quantity of GM material in the pollen samples from colonies at sites A-D (situated at a maximum of 662 m from the GM trial). This prediction will be tested by analysing the DNA of the pollen collected in the traps.

If a beekeeper has colonies nearer to conventional oilseed rape crops, than to a GM crop, then the quantity of GM pollen returned to the colony is likely to be minimal, although it may increase as flowering finishes and bees spread further afield.

Ackowledgements We thank Jeremy Sweet and Peter Lutman for use of the BRIGHT trial. IACR Rothamsted recieves grant-aided support from the Biotechnology and Biological Sciences Research Council of the UK.

References Champolivier, J., Gasquez, J., Messean, A. & Richard-Molard, M. (1999) Management of transgenic crops within the cropping system. Gene flow and agriculture:relevance for transgenic crops. BCPC Symposium proceedings No. 72. (ed. P.J.W. Lutman). pp. 233-240. BCPC, London. Crawley, M.J., Hails, R.S., Rees, M., Kohn, D. & Buxton, J. (1993) Ecology of transgenic oilseed rape in natural habitats. Nature, 363, 620-623. Cresswell, J.E., Osborne, J.L. & Goulson, D. (2000) An economic model of the limits to foraging range in central place foragers with numerical solutions for bumble bees. Ecological Entomology, 25, 249-255. Downey, R.K. (1999) Gene flow and rape - the Canadian experience. Gene flow and agriculture:relevance for transgenic crops. BCPC Symposium proceedings No. 72. (ed. P.J.W. Lutman) pp. 109-116. BCPC, London. Emberlin, J., Adams-Groom, B. & Tidmarsh, J. (1999) A report on the dispersal of maize pollen. Soil Association Report. Firbank, L., Dewar, A., Hill, M., May, M., Perry, J., Rothery, P., Squire, G. & Woiwod, I.H. (1999) Farm-scale evaluation of GM crops explained. Nature, 399, 727-728. Ingram, J. (2000) The separation distances required to ensure cross-pollination is below specified limits in non-seed crops of sugar beet, maize and oilseed rape. Plant Varieties and Seeds, 13, 181-199. Picard-Nizou, A.L., Pham-Delegue, M.H., Kerguelen, V., Doualt, P., Marilleau, R., Olsen, L., Grison, R., Toppan, A. & Masson, C. (1995) Foraging behaviour of honey bees (Apis mellifera L.) on transgenic oilseed rape (Brassica napus L. var. oleifera). Trangenic Research, 4, 270-276. Ramsay, G., Thompson, C.E., Neilson, S. & Mackay, G.R. (1999) Honey bees as vectors of GM oilseed rape pollen. Gene flow and agriculture:relevance for transgenic crops. BCPC Symposium proceedings No. 72. (ed. P.J.W. Lutman). pp. 209-214. BCPC, London. Raybould, A.F. & Gray, A.J. (1993) Genetically modified crops and hybridization with wild relatives: a UK perspective. Journal of Applied Ecology, 30, 199-219. Scheffler, J.A., Parkinson, R. & Dale, P.J. (1993) Frequency and distance of pollen dispersal from transgenic oilseed rape (Brassica napus). Transgenic research, 2, 356-364. Simpson, E., Norris, C., Law, J., Thomas, J. & Sweet, J. (1999) Gene flow in genetically modifed herbicide tolerant oilseed rape (Brassica napus) in the UK. Gene flow and agriculture:relevance for transgenic crops. BCPC Symposium proceedings No. 72. (ed. P.J.W. Lutman). pp. 75-81. BCPC, London. Thompson, C.E., Squire, G., Mackay, G.R., Bradshaw, J.E., Crawford, J. & Ramsay, G. (1999) Regional patterns of gene flow and its consequence for GM oilseed rape. Gene flow and agriculture:relevance for transgenic crops. BCPC Symposium proceedings No. 72. (ed. P.J.W. Lutman). pp. 95-100. BCPC, London. Treu, R. & Emberlin, J. (2000) Pollen dispersal in the crops maize (Zea mays), oil seed rape (Brassica napus ssp. oleifera), potatoes (Solanum tuberosum), sugar beet (Beta vulgaris ssp. vulgaris) and wheat (Triticum aestivum) Soil Association Report Vaissière, B.E., Torre Grossa, J., Rodet, G. & Malabœuf, F. (1994) Sociality and pollen flow: direct evidence for effective in-hive pollen transfer. Les Insectes Sociaux, 12ème Congress de l'Union Internationale pour l"Etude des Insectes Sociaux UIEIS, (eds A. Lenoir, G. Arnold & M. Lepage), pp. 290. Université Paris Nord, Paris, Sorbonne Williams, I.H., Martin, A.P. & White, R.P. (1986) The pollination requirements of oil-seed rape (Brassica napus L.). Journal of Agricultural Science, Cambridge, 106, 27-30. Williams, I.H., Martin, A.P. & White, R.P. (1987) The effect of insect pollination on plant development and seed production in winter oilseed rape (Brassica napus L.). Journal of Agricultural Science, Cambridge. 109, 135-139.

GK LH

A B C D E F BH WH AT

500m

Figure 1 Outline of fields on Rothamsted estate. Hatched block (BH) = Black Horse, field of HTGM trial site. Grey blocks with initials (WH, AT, GK and LH) = surveyed fields of conventional oilseed rape (Table 1). Dotted line = transect of 18 honeybee colonies at sites A-F. Three colonies were placed at each coloured dot, and colour indicates powder dye added to colony (Table 2).

8 7 6 5 4 3 2 1 Mean no bees per 120m 0 CILR Crop variety

Figure 2 Back-transformed mean number of honey bees (± s.e.m) on crop cultivars on Black Horse, combining weeks (n=16 for each column i.e. 4 weeks x 4 plots). C = conventional variety (Apex); I = non-GM Imazamox-tolerant; L = GM Liberty Link, R = GM Roundup-Ready.

a) Black Horse, n = 558 b) White Horse, n = 157 c) Apple Tree, n = 19

350 yellow 300 orange red 250 green 200 magenta 60 50 150 40 No. marked bees 100 30 8 20 6 50 4 10 2 0 0 0 1234 1234 1234 Week Week Week

Figure 3 Numbers of marked bees of each colour seen each week on each field a) Black Horse b) White Horse c) Apple Tree (refer to Fig 1 for relative positions of fields and colonies). n = total number of marked bees observed on that field.

3 magenta 2 green red 2 orange yellow

1 Mean no. of bees per 100m 0 0 200 400 600 800 1000 1200 1400 1600 1800 Distance of surveyed field from hives (m)

Figure 4 Mean number of marked bees per 100m2 observed at different distances from their colonies. Different colours represent bees from different colony sites (Table 2). HOW FAR DO HONEY BEES FLY TO FIELDS OF BRASSICA NAPUS (OILSEED RAPE)?

1Osborne, J.L., Carreck, N.L. and Williams, I.H.

Plant and Invertebrate Ecology Division, IACR Rothamsted, Harpenden, Herts, AL5 2JQ, UK. 1Tel: 01582 763133 ext 2738, fax: 01582 760981, email: [email protected]

Curriculum vitae for J L Osborne

Making presentation in “Pollination and Bee flora” session on GM crops, at Apimondia 2001

Degrees 1989 BA Cambridge University; BA(Hons), Natural Sciences, 1st class

1994 PhD Cambridge University; PhD Thesis “Evaluating a pollination system: Borago officinalis and bees”

Relevant posts 1990-1991 Cambridge University; review for the European Parliament entitled “Bees and the pollination of crops and wild flowers: changes in the European community”

1995-1997 Postdoctoral researcher in “Insect Pest and Pollinator Ecology” Group, Dept of Entomology & Nematology, IACR Rothamsted; funded by The Leverhulme Trust

1998-present Project leader of Insect Behaviour Programme, Plant and Invertebrate Ecology Division, IACR Rothamsted, Harpenden; BBSRC funded

Current research area: Ecology and movement of bees, and the interactions between pollinators and the plants dependent on them.

From: @syngenta.com Sent: 20 June 2014 16:46 To: Pywell, Richard F. Cc: @bayer.com Subject: CEH_BCS_SYT.docx Attachments: CEH_BCS_SYT.docx

Dear Richard,

In order to proceed with the signature of the agreement, I would like to ask you a final review of what was agreed and consolidated during our discussions.

If everything resonate with you, we agree with that we will try to have this signed and sent at your office by the end of next week.

I wait for your green light.

Thank you.

Have a good weekend,

Syngenta Schwarzwaldallee 215 4058 Basel www.syngenta.com

This message may contain confidential information. If you are not the designated recipient, please notify the sender immediately, and delete the original and any copies. Any use of the message by you is prohibited. From: Pywell, Richard F. Sent: 27 June 2014 14:32 To: @bayer.com'; @bayer.com'; @bayer.com'; @bayer.com'; @syngenta.com'; @syngenta.com'; @syngenta.com'; @syngenta.com'; ; '; '; Bullock, James M.; Shore, Richard F.; ; ; .; @syngenta.com' Subject: NNI experiment - notes from meeting 25 June Attachments: Osborne_oilseed_rape_bees_fly_2001.pdf; NNI_Expt_meeting 25 June_Actions_rpywell.docx

Dear all, thanks again for attending the meeting at Wallingford on 25 June.

I attach notes and actions kindly compiled by .

I also attach a paper from which supports the exclusion of non-experiment oilseed rape in a 1km radius around each site.

...... see the attached paper by Juliet Osborne et al giving recorded foraging distances of honey bees from hive to OSR fields - see the last main paragraph of discussion in particular "Although bees had a choice of rape fields within foraging range, at different distances from their colonies, they were all observed witin 1km of their colonies. Max observable distance was 1.7km." ...I think this validates our decision to aim for no additional OSR fields within 1km radius of hives......

Best wishes

Richard

':01491 692356

*: [email protected]

Field Trial to Measure the Effects of Neonicotinoid Seed Dressings on Honeybees

Site selection criteria Draft 1.0; 26 June 2014

Selection process

1. Search Bayer/Syngenta/Eurofins databases for growers ≥45 ha oilseed rape 2. Telephone interview (points 1-6) 3. Site inspection (points 7-11) 4. Submit site assessment form for approval 5. Contract with grower

Telephone interview (complete attached form – Annex 1)

1. Is the grower interested in participating in the field trial based on the summary description and offer below?

Summary field trial description for grower • This is an independent field trial to measure any effects of commercial neonicotinoid (NNI) seed dressings on honeybees. • The trial will involve growing oilseed rape treated with either Cruiser OSR® or Modesto® seed treatments and comparing this with a rape crop receiving just a fungicide seed treatment (untreated control). • Treated/untreated seed will be supplied free to the grower. • The crop will be grown under a full experimental licence allowing for it to be sold on the open market. • Depending on the size of the farm we may ask you to grow one, two or all three treated rape crops on different parts of your farm. • The crop will be grown using Good Agricultural Practice specified by the project team. • The project team will need to make frequent visits to the crop and place honeybee hives in the centre of the crop. • The location of the field trials will remain anonymous. Offer to the grower • We will provide free seed of the selected study variety (treated with NNI or not treated) to sow 45-75 ha. • [depending on negotiation] we will pay a premium of 5% to 10% of the crop value for the treated and untreated field(s) up to a maximum of 10000 € per treated/untreated crop of 45-75 ha.

2. Do they grow oilseed rape in discrete/separate patches of 45-75ha (comprising either a single field or group of adjacent fields)? 3. Agronomic history: i) have they grown crops treated with Imidacloprid seed dressing (e.g. Tripod PLUS® on cereal; Gaucho® on sugar beet, Chinook® on rape) on this land in the last three years? ii) Is the current crop treated with a Clothianidin or Thiamethoxam seed treatment (e.g. Redigo Deter, NipsIT on winter wheat)?

4. Are they prepared to establish and manage the crop variety supplied free according to a Good Agricultural Practice protocol. This will include drilling the crop at a specified seed rate, the use of some specified pesticide products, and keeping good records of inputs. 5. Is the grower prepared to have honeybee hives in the middle of the oilseed rape crop? (this may require leaving a central strip through the crop c.3m wide left unsown and cut to allow access). Is there good vehicle access to the study field(s)? 6. Does the grower have a crop plan or good knowledge of what crops that will be grown around the proposed study field(s) next season (approximately 1km radius)? We need to exclude other oilseed rape crops and crops attractive to bees in spring (e.g. winter beans) from a 1km radius around the study field(s).

Site inspection (complete attached form – Annex 1)

7. Description of the study field(s): number of fields, accurate location (lat. Long. – use Google earth - http://www.google.co.uk/intl/en_uk/earth/) current crop & condition. Soil type: make a simple field assessment of soil type using the 13 categories below:

8. Soil sample for NNI residue analysis: take a single bulked soil sample from each of the proposed study fields for analysis of NNI residues. From each field collect 20 soil samples using a gauge auger (e.g. Pürkhauer) to plough depth (c. 20cm) across the whole field. Place the samples in a labelled bag (site name, coordinates, project GTZW20140201-01). Thoroughly mix the soil and reduce to 100 g for analysis (keep the remaining soil in a labelled bag for re-analysis). Within one week send the sample by express air freight to: SOFIA GmbH Rudower Chaussee 29 D-12489 Berlin

Label each sample as follows: SOFIA GmbH

Rudower Chaussee 29 D-12489 Berlin Project: GTZW20140201-01 Test: PSFSF (soil analysis for imidacloprid, thiamethoxam and clothianodin (LOQ 0.005 mg/kg) Turnaround time: 2-3 days Sponsor contact. /Eurofins Agroscience Services GmbH Site name: Site coordinates:

9. Landscape description: using the categories on the form characterise the surround land use. We are looking for predominantly arable land within each block (>50%).

10. Surrounding crops. Please attach a crop plan (autumn 2014 / spring 2015 sowing) for the land surrounding the study field(s). We particularly wish to avoid oilseed rape crops within a 1km radius of centre of study field(s) (hive location):

1 km

11. Separation: confirm there is sufficient separation ( ≥4km) between treated/untreated study field(s) (‘Treatments’) and ≥10km separation between clusters of three treatments (‘Blocks’) (see Annex 2 – glossary of terms):

Block 1 Untreated control Clothianidin Hives ≥4km Hives ≥10km Block 2

≥4km Untreated control ≥4km Clothianidin Hives ≥4km Hives

Hives

Thiamethoxam ≥4km ≥4km

Hives Thiamethoxam

Annex 1: Site assessment record

Telephone interview Grower contact details: Name: Address:

Telephone: E-mail:

1. Interested in taking part in the trial 2. Number of discrete/separate patches growing oilseed rape of 45-75ha Area of patch(s) (approximately) Patch No. fields Total area (ha)

3. Agronomic history: i) have they grown If so, what product and when crops treated with Imidacloprid seed dressing (e.g. Chinook®) on this land in the last three years? ii) Is the current crop treated with a If so, what product and when Clothianidin or Thiamethoxam seed treatment (e.g.....).

4. Prepared to establish & manage crop Comments: according to a Good Agricultural Practice protocol. 5. i) Prepared to have honeybee hives in the Comments: middle of the oilseed rape crop? Comments: ii) Prepared to establish/maintain access path in centre of crop 6. Crops surrounding trial field(s) 2014/15

Site inspection 7. Detailed site description:

Patch identifier No. fields Total area Location (centre of Condition Soil type (ha) field(s) Lat. Long.) (e.g. wheat (categories stubble) 1-13)

8. Soil sample(s) taken according to Comments: guidelines for NNI residue analysis 9. General landscape description around Land use Approx. % study field(s) 1-2km Arable Improved grassland Unimproved grassland Forest Orchards Urban Other (name) 10. Surrounding crops: Comments: Confirm no untreated oilseed rape crops within 1km radius of centre of study field(s) for 2014/15 Mark fields where crop unknown on map Add contact details of neighbouring farmers on map (name, phone number) 11. Separation distances: Comments: i) Confirm separation ( ≥4km) between treated/untreated study field(s) (‘Treatments’) Comments: ii) Confirm separation between blocks ( ≥4km) (if applicable)

12. Number of sites approved Comments: 13. Proposed compensation € 14. Estimated drilling date(s)

15. Recommendation (to be completed by CEH)

Please attach site maps / cropping plans etc

E-mail Site Assessment Form to Richard Pywell at Centre for Ecology and Hydrology: [email protected]

Annex 2: Glossary of terms

Treatment = a field or group of fields with a total area of between 45-75ha growing oilseed rape. There are three separated treatments (1. no NNI; 2. Clothianidin; 3. Thiamethoxam). The experiment comprises a total of 36 treated fields/groups of fields.

Block = a replicate of the three treatments situated within a similar landscape. There are FOUR blocks per country (Total = 12)

From: Pywell, Richard F. Sent: 03 July 2014 18:42 To: ; ' @syngenta.com'; @syngenta.com' Cc: @SYNGENTA.COM'; ; ' '; @SYNGENTA.COM; Bullock, James M.; @syngenta.com' Subject: NNI crop agronomy

Dear ,

We need to agree the crop agronomy protocol for the NNI experiment soon as this will be part of the contract with each grower.

To summarise the issue:

As discussed at the 25 Jun meeting there are three approaches to the agronomic management of the crop depending upon the question we wish to answer: 1) Regulatory-type trial of NNI effects in isolation Pros i. a fundamental scientific question, in theory should be straight forward to assign causality Cons i. impossible to achieve this level of simplicity under field conditions (e.g. you cannot grow oilseed rape without pesticides or compare this with a crop protected by NNIs alone) ii. alternatively, you could keep all other pesticide inputs the same and just vary +/- NNI. However, given the scale and dispersed nature of experiment it will be practically very difficult and costly to keep achieve this (e.g. many prescribed inputs could be purely prophylactic or may be insufficient to control serious pest outbreaks specific to just one region, causing unwanted variation in the trial) – either way they will not reflect the real world iii. this approach would tend to repeat the previous small-scale field trials (e.g. Pilling et al. 2014)?

2) A real world test of NNI effects under field conditions compared with the alternative agronomy without NNI (restriction on use of some products to protect bees) Pros i. Compares two realistic crop management strategies (one with NNI and one without) ii. No one has looked at the impacts of the impacts of the alternative (non-NNI) agronomy on bees iii. Relative freedom of choice (with some limits) is practical and growers will buy into the experiment iv. Accepted approach in other large-scale field trials (e.g. Farm-scale Evaluations of GM crops) Cons i. There will be variation in inputs based on local pest pressure and farmer behaviour etc ii. Need to ensure alternative products are available locally iii. Compensation must cover any losses

3) A compromise between 1) and 2) with initially divergent agronomy coming in spring together after the NNI protective effect has worn off Pros i. Marginally less variable than 2) Cons i. Still difficult to police and has the same issue of unnecessary or insufficient application of pesticides ii. Not sure if farmers will accept the constraints on management iii. A compromise that may not reflect the real world

Recommendation: CEH believe the key is to grow all crops (treated/untreated) to a uniformly high standard USING OPTION 2) REAL WORLD/BEST PRACTICE AGRONOMY. The key strength of the experiment is the real world exposure hence the size of the study fields etc. Evenness in crop quality between control and NNI means the bees are exposed equally the crop regardless of treatment. The only constraints should be to avoid applying pesticides that are harmful to bees at flowering time and avoid applying other NNI insecticides (where viable alternatives exist) as this will cloud the issue.

The Bayer UK technical team suggested the following agronomic management which is along these lines:

Sowing /autumn 50seeds m2

Fungicide - Apply Hy-Pro Duet @ 9ml/kg seed Modesto @ 12.5ml/kg seed OSR Cruiser @***ml/kg seed

An even crop 20-30 plants m2 coming through winter is ideal.

For autumn treatments (fungicide/ molluscicide /insecticide) allow the growers to apply inputs with aim of getting the best possible crop establishment.

Spray according to pest thresholds ( CSFB pyrethroid, aphids Plenum).

Spring up to flower bud

For fungicide/PGR inputs in spring up to flower bud allow the growers to apply inputs with aim of getting the best possible crop yield.

For insecticide sprays it would be easier to adopt the local agronomy for the plots but without any “Neonic sprays” ( ie thiacloprid).

Spring Flowering For fungicides again no restriction of product use but always apply according to best stewardship practices with respect to sprays in the flowering period.

There are likely to be differences in the need to control Pollen Beetle between the three countries, but depending on thresholds Plenum WG (Syngenta) – ( pymetrozine non neonic). Likewise for seed weevil and blossom midge.

Note – this protocol will need adapting for Ger, Hun (e.g. there are issues with some products routinely applied in Hungary – we will need to exclude these and suggest readily available alternatives) (technical input needed).

Please can I have you comments.

Best wishes

Richard

:01491 692356

: [email protected]

From: Pywell, Richard F. Sent: 03 July 2014 13:11 To: @syngenta.com' Cc: @syngenta.com'; Bullock, James M. Subject: Resources for the NNI Expt

Dear ,

Following our discussions yesterday I enclose a revised table of costs for the NNI experiment.

1) Clarification of end date of monitoring costs To clarify, the end date for monitoring will be early 2016 as we have to measure overwinter survival of the honeybees. However, most of the spend on monitoring will be in early to mid-2015

2) Budget This has been updated with a quote for overseeing site establishment and a revised quote for site location received from Eurofins today (Note I have had to estimate the mileage charges in both cases)

Also, I have not yet include a small cost for CEH supporting BCS/SYN field teams in selecting sites for the UK (telecom tomorrow will provide an estimate).

The bottom line is that neither has made much difference to the previous estimated total cost:

Experiment = Seed costs + farmer compensation =

3) Options Following discussions with the CEH team there is nothing we can suggest to reduce these costs by the margin required at this stage. Our view is that we must offer grows a reasonable incentive to join the trial and do a good job.

One option would be to terminate the Demo Farm Network at put any remaining resources into the experiment. A further option would be to profile the project costs such that SYN could request additional resources in 2015 and 2016.

Best wishes

Richard

Net cost Gross cost Gross cost Gross cost Activity Organisation Delivery (Euros) Tax (%) (Euros) (Dollars) (Pounds) Comment 1. Experimental design and feasibility CEH Mar-14 66,830 € 0 66,830 € $91,062 £53,492 2. Site NNI residue analysis Eurofins Sep-14 16,200 € 19 19,278 € $26,268 £15,431 Assume 3 fields per site × 36 = 108 samples @ 150 euro per sample = 16200 euros 3. Location of experimental sites (Ger, Hun) Eurofins Sep-14 61,000 € 19 72,590 € $98,911 £58,103 Estimated cost, including bonuses and mileage 4. Overseeing establisment of experiment Eurofins Sep-14 33,000 € 19 39,270 € $53,509 £31,433 Estimated mileage 5. Seed costs, treatment and shipping (c.2160 ha oilseed rape) Eurofins Sep-14 160,000 € 20 192,000 € $261,619 £153,683 Estimated cost based median of 45-75ha = 60 ha per site × 36 = 2160 ha @ £80 ha-1 = £172800; shipping assumed to be included; EXCLUDES costs of seed treatment and product 6. Farmer compensation payments (c.36 farms) Eurofins Sep-15 300,000 € 20 360,000 € $490,536 £288,156 Maximum estimated cost, includes 10% supplement to market proven value of crop = 10000 euro per farm 7. Monitoring field experiment (including overwinter survival to 2016) Eurofins Mar-16 1,025,640 € 0 1,025,640 € $1,397,537 £820,955 8. Honeybee pollen analysis and hive disease analysis Eurofins Mar-16 635,040 € 0 635,040 € $865,306 £508,306 9. Chemical residue analysis CEH Mar-16 203,244 € 0 203,244 € $276,940 £162,683 10. Project management and co-ordination, QA, data management, analysis, reporting CEH Mar-16 334,148 € 0 334,148 € $455,309 £267,462

TOTAL 2,835,101 € 2,948,039 € $4,016,997 £2,359,704

:01491 692356

: [email protected]

______