Developing rational policies regarding coexistence of GM and conventional crops that are open pollinated requires reliable information about pollen dispersal, particularly over complex terrain where high value crops might be grown.  To this end, we combined a high-resolution atmospheric dynamical model with a Lagrangian dispersion model to create a framework for predicting long-distance pollen dispersion over complex terrain.  A published case study on GM bentgrass reported cross-pollination in complex terrain at distances more than 20 km from the nearest GM source.  As a test for our model, we simulated six days of pollen dispersal from the GM bentgrass sites implicated as most likely being responsible for the reported incidents of cross-pollination.  Model output confirmed the potential for bentgrass pollen dispersion beyond 20 km from its source.  And pollen deposition was predicted in locations where out-crossing was observed.  Predicted patterns of pollen deposition, however, were far more complex than those predicted over simpler terrains.  These results suggest that simpler methods often used to simulate pollen dispersion can produce misleading results if local effects of complex topography on atmospheric turbulence are not taken into account.  As such, distance, temporal separation, and topography need to be considered in establishing isolation distances between GM and non-GM crops.