Air-mediated pollen dispersal from genetically modified (GM) wind pollinated crops can cause cross-pollination between GM and conventional fields, and so result in the presence of the transgene in non-GM crops. The demand to keep crops separated requires farm scale changes in agricultural practices and special methods and regulations to prevent the gene flow. In order to define sufficient but cost-effective methods to enable successful coexistence, reliable modeling tools for estimating gene flow are needed. To describe and predict the pollen dispersal, a mechanistic simulation model for air-mediated pollen dispersal was developed. The model was used for a simulation study, to assess the risk of gene flow from a GM to a conventional maize field. Simulations were done under various weather conditions. From the simulated pollen dispersal, the amount of contamination of the whole conventional crops was estimated. The effects of several factors like barriers, isolation distance and the size of the GM field and the spatial configuration of the fields to the gene flow were assessed. Of the landscape related factors, in addition to isolation distance, GM field size and the spatial configuration of the fields turned out to have effect on the average gene flow. However, the simulation study also showed that even though the amount of gene flow on average was relatively low, much higher contaminations were also observed due to the variation in weather between the pollination periods. Due to this large variation, gene flow cannot be described by the average amount of contamination only, but information about the frequency of extreme events is also needed. When considering the co-existence of GM and conventional crops, the large variation in contaminations should be taken into account, and rather than trying to prevent all the extreme cases, an acceptable risk for these should be decided upon.