Spatial models of pollen dispersal in the forage grass meadow fescue

Abstract

Several bivariate probability distributions, generated by different underlying dispersal mechanisms, are fitted to the observed frequencies of an isozyme marker gene using a maximum likelihood approach. The pollen dispersal data were generated using two experimental populations of meadow fescue (Festuca pratensis Huds.), homozygous for different allozymes at the (Pgi-2) locus, arranged in a circular donor–acceptor field design. The contribution of a plant depends on plant position, fecundity and flowering time, factors which are taken into account when fitting the different models. Several approximate likelihood-ratio tests are done between alternative nested models, and a wind threshold model with bimodality in the wind direction is selected. The evolutionarily important variances and expectations of gene displacement under the selected model are calculated. It is also shown that the underlying probability distribution is significantly more than exponentially leptokurtic. By fitting a distribution of deposition in all three dimensions to the data, taking into account differences in plant height, separate estimates of additional physical parameters are obtained, showing that gravity and vertical random movements are more important than intervening vegetation in limiting pollen dispersal in meadow fescue. According to the model, plants with a high seed yield contribute pollen over-proportionally to neighbouring plants.