Maximum feasible distance of windborne cross-pollination in Brassica napus: A ‘mass budget’ model

A theory of gene dispersal by wind pollination can make an important contribution to understanding the viability and evolution of important plant groups in the Earth's changing landscape and it can be applied to evaluate concerns about the spread of engineered genes from genetically modified (GM) crops into conventional varieties via windborne pollen. Here, we present a model of cross-pollination between plant populations due to the wind. We perform a ‘mass budget’ of pollen by accounting for the number of pollen grains from release in the source population, dispersal from the source to the sink population by the wind, and deposition on receptive surfaces in the sink population. Our model can be parameterised for any wind-pollinated species, but we apply it to Brassica napus (oilseed rape or canola) to investigate the threat posed by wind pollination to GM confinement in agriculture. Specifically, we calculate the maximum feasible distance at which a particular level of windborne gene dispersal could be attained. This is equivalent to the separation distance between populations or fields required to achieve a given threshold of gene dispersal or adventitious GM presence. As required, model predictions of the upper bounds on levels of wind-mediated gene dispersal exceed observations from a wide range of published studies. For a level of gene dispersal below 0.9%, which is the EU threshold for GM adventitious presence, we predict that the maximum feasible distance for agricultural fields of B. napus is 1000m, regardless of field shape and direction of prevailing winds. For fields closer than 1000m, our model results do not necessarily imply that the 0.9% threshold is likely to be breached, because in this instance we have conservatively set the values of parameters where current knowledge is limited. We also predict that gene dispersal is reduced by 50% when the lag in peak flowering between the source and sink populations is 13 days, and reduced by 90% when the lag is 24 days. We identify further measurements necessary to improve the accuracy of the model predictions.