Long distance seed dispersal by wind: measuring and modelling the tail of the curve.

The size and shape of the tail of the seed dispersal curve is important in determining the spatial dynamics of plants, but is difficult to quantify. We devised an experimental protocol to measure long-distance dispersal which involved measuring dispersal by wind from isolated individuals at a range of distances from the source, but maintaining a large and constant sampling intensity at each distance. Seeds were trapped up to 80 m from the plants, the furthest a dispersal curve for an individual plant has been measured for a non-tree species. Standard empirical negative exponential and inverse power models were fitted using likelihood methods. The latter always had a better fit than the former, but in most cases neither described the data well, and strongly under-estimated the tail of the dispersal curve. An alternative model formulation with two kernel components had a much better fit in most cases and described the tail data more accurately. Mechanistic models provide an alternative to direct measurement of dispersal. However, while a previous mechanistic model accurately predicted the modal dispersal distance, it always under-predicted the measured tail. Long-distance dispersal may be caused by rare extremes in horizontal wind speed or turbulence. Therefore, under-estimation of the tail by standard empirical models and mechanistic models may indicate a lack of flexibility to take account of such extremes. Future studies should examine carefully whether the widely used exponential and power models are, in fact, valid, and investigate alternative models.