How does the structure of an agricultural landscape affect the dispersal of transgenes coming from genetically modified crops –by wind, volunteers or mechanical transportation ? Which landscapes do minimize the risks ? What are the effects of landscape structures on other environmental problems such as the diffusion of disease spores, the development of invasive species or the erosion phenomena ? We present a statistical approach and a tool which have been elaborated for answering those questions, within the framework of a multidisciplinary project initiated by the French ministry of Research. We have implemented a software for simulating 2-dimensional agricultural landscapes, named GenExP (Adamczyk et al., 2007), which generates a land subdivision by using a traditional algorithmic geometry (Chassery et Montanvert, 1991). For tessellation methods based on points (e.g. Voronoï tessellation), the spatial distribution of germs can be estimated from real lanscapes, allowing GenExP to build «realistic» landscapes. Then, crops (rapeseed, maize, wheat, ..., whether they are GM or not) are allocated to the fields, either randomly or according to spatial and temporal rotation rules. Indeed, GenExP can use the capabilities of the software R (Baddeley et Turner, 2005) –which allows to estimate and simulate the germs distribution of a tessellation-- on the one hand, and the rotation models built by the data-mining software Carottage (Mari et al., 2002), on the other hand. Finally, based on real or realistic field-patterns, GenExP provides multiannual maps of agricultural landscapes, which can be used for example by the softwares Mapod-maize (Angevin et al., 2007) and Genesys-rape (Colbach et al., 2005), which were developed by agronomists and biologists to simulate the dispersal of GM pollen grains and seeds at various scales (from fields to collecting areas)