Selection procedure to study ecological impact of GM plants: a case study of Bt-maize in Kenya A.N.E Birch 1 , S. Arpaia 2 , G.L. Lövei 3* , M. Sétamou 4 , S. Sithanantham 4 , R. Wheatley 2 , A. Hilbeck 5 , D. A. Andow 6 1 - Scottish Crop Research Institute, Invergowrie, Dundee, DD2 5NQ, Scotland, U.K. 2 - ENEA –Research Centre Trisaia, I-75012 Rotondella (MT) Italy; 3 - Department of Crop Protection, Danish Institute of Agricultural Sciences, Flakkebjerg Research Centre, DK-4200 Slagelse, Denmark; 4 - International Centre for Insect Physiology and Ecology, P.O. Box 30772, Nairobi, Kenya; 5 - Swiss Federal Institute of Technology, Geobotanical Institute, Zurichbergstr. 38, CH-8044 Zurich, Switzerland; 6 - Department of Entomology and Center for Community Genetics, University of Minnesota, St. Paul, MN 55108, USA * Presenting author Conceptional and methodological uncertainties of studying ecological effects of GM crop plants on non-target arthropods have raised several intriguing general problems..What species or ecosystem functions should be chosen to test? By what routes might these species or functions be exposed directly or indirectly to GM crop plant products? How can meaningful scientific hypotheses be constructed to provide rapid assessments of the magnitude of the potential risks? In contrast to ecotoxicological methods for addressing these problems, assessment of the impacts of GM crop plants must be case specific and contextualized to the environment in which they will be used. In the framework of the IOBC Global Working Group on transgenic organisms, the “Guidelines Project” (www.gmo-guidelines.info), we developed an“ecosystem representative approach”for selecting species and ecosystem function as foci for testing.This approach combines ideas and methods from a“community approach”, which emphasizes analysis of biodiversity, a ‘functional approach’, which emphasises community reactions, a “key species approach”, which emphasizes the individuality of species,and an “indicator species approach”,which is central to ecotoxicological testing.We used classic qualitative methods of risk assessment (NRC, 1983), formalized in selection matrices and directed questions, which provide transparent summaries of scientific data and expert judgement. The process of ranking and species selection in the above-ground functional groups (herbivores, detritivores, natural enemies and pollinators), proved useful and allowed the identification and prioritisation of non-target species for some key ecological groups; it also reflected the current state of knowledge and expertise available. An example for ranking parasitoids is on Table 1. Identification of the significance or function of the non-target species was important to prioritisation and identifying gaps in present knowledge.Species with a ranking of 1 are suggested for further testing.It was also important to consider the process of exposure (Table 2) as part of overall species selection, as this identified missing information about the expression of Bt toxin in different plant tissues was also crucial for the above-ground exposure analysis. Similarly, the ranking and selection matrix for soil ecosystem functions proved useful.The species selection matrices could bwere modified appropriately to rank and select ecosystem functions. We concluded that sufficient ecological information existed to use qualitative ecological expertise to identify and rank soil ecosystem functions as candidates for non-target risk assessment.This selection tool was useful in identifying key interactions in a systematic and transparent way that might be most likely to be affected by Bt maize or those functions significant for maize production. • 31 •