Modeling of Gene Flow by a Bayesian Approach : A New Perspective for Decision Support

Maize is the major crop in Europe and the second most widely-cultivated genetically modified (GM) crop in the world after soybeans. Maize is one of the only GM crops commercially grown in Europe (along with potato). Since maize is a cross-pollinated crop relying on wind for the dispersal of its pollen, pollen flow between neighboring maize fields is one of the major potential on-farm sources of adventitious mixing between GM and non-GM material (Devos, Reheul, & De Schrijver, 2005). The cross-fertilization between GM and non-GM crops has been widely studied through measurements of pollen concentration and levels of cross-fertilization. Experimental data on gene flow for maize were collated and synthesized within the SIGMEA (Sustainable Introduction of GMOs into European Agriculture) European research project (Messéan et al., 2009). As stated before, GM maize is commercially grown in Europe (except in some countries, as in France), thus coexistence situations may occur. Coexistence refers to the ability of farmers and consumers to make a practical choice between conventional and GM products based on compliance with the legal obligation for labeling and/or purity standards (European Commission, 2003a). In Europe, up to 0.9% of GM material in non-GM food and feed is authorized, provided these traces of genetically modified organisms (GMOs) are adventitious or otherwise technically unavoidable (European Commission 2003b). Above this threshold, in order to allow consumers to make a practical choice about the product, it must be labeled as consisting of, containing, or being produced from a GMO. In order to meet regulatory requirements, accurate prediction of maize gene flow is thus needed to assess risk of commingling between GM and non-GM crops. Moreover, tools are needed to help stakeholders of the maize supply chain to manage coexistence between GM and non-GM maize. In this context, considerable efforts have been made to model maize pollen dispersal with different modeling approaches. Spatially explicit and quasi-mechanistic models were defined (Angevin et al., 2008; Colbach, Clermont-Dauphin, & Meynard, 2001; Klein, Lavigne, Foueillassar, Gouyon, & Larédo, 2003) and tested in order to determine legal separation distances between GM and non-GM maize fields. HowArnaud Bensadoun and Hervé Monod National Institute for Agricultural Research (INRA), Research Unit (UR), France