Measuring and modelling anthropogenic secondary seed dispersal along roadverges for feral oilseed rape

Seed dispersal of feral crop plants along roadverges is likely to be influenced by numerous anthropogenic vectors in the agroecosystem. Within the context of introducing genetically modified (GM) cultivars, long-distance dispersal of feral seeds associated with the growth of GM feral populations (via a selective advantage due to transgene expression) could make these populations become invasive. Their expansion could then favour the spread of transgenes and modify the composition of roadverge plant communities. Because quantitative data on anthropogenic seed dispersal along roadverges were few, we estimated effective secondary dispersal for oilseed rape, the seeds of which are not adapted to dispersal by wind or biotic agents. A seed deposition experiment showed that secondary dispersal did not systematically occur along roadverges, was correlated with traffic intensity and was local. Low traffic intensity and anthropogenic disturbances (covering of seeds by mown grasses and burial by farming machinery) prevented dispersal on three of the experimental sites. Along a road with higher traffic, secondary dispersal occurred (d_max=21.5 m), probably induced by wind turbulence behind vehicles. The best-fitting dispersal kernel was a mixture of two components: 20% of seeds dispersing over a few metres on average and 80% remaining in the original place. Expansion rates of feral populations of GM herbicide-tolerant oilseed rape were computed using an invasion model and this kernel. They were low (1–4 m yr⁻¹) when only ballistic and/or secondary dispersal were included but higher (4–20 m yr⁻¹) when theoretically rare events of long-distance dispersal by verge mowers were added. This study suggests that secondary seed dispersal is unlikely to have a significant impact on the spread of GM feral oilseed rape populations in highly disturbed and dynamic habitats such as roadverges. Detecting long-distance dispersal events induced by other vectors (e.g. mowers) would require integrative approaches based on genetic and spatial data.