The genetics of geographically structured populations
Gene-flow has been studied in this research from an analytical, theoretical, and practical angle. While simple models of restricted gene-flow are tractable analytically and can produce very accurate predictions when compared with the results of computer simulations, models of discrete populations with geographical structure and models of continuous populations need further research. In particular, models of isolation by distance in a continuum are very difficult to relate to concepts familiar to the population geneticist since the basic concept linking continuous populations to discrete ones, the neighbourhood size, is shown to be flawed. Inferring gene-flow from indirect methods implies obtaining unbiased estimators of quantities such as F-statistics. The framework for estimation presented in this research can be used to derive unbiased estimators in different situations, and can also help to clarify the underlying assumptions made when making these estimates. In particular the conditions are specified under which Nei and Chesser's (1983) and Weir and Cockerham's (1984) estimators are most appropriate. While analytical treatment of geographically structured populations is difficult, F-statistics can be used to unravel levels of genetic structuring in these populations. Methods are presented which yield ways of discriminating between samples taken within and among breeding units, a necessary distinction if levels of gene-flow are to be inferred. Calculations of pairwise Fat, even in continuous populations, provide a picture of the geography of gene-flow in the population investigated. The methods are applied to data sets of three species, Brassica oleracea ssp. oleracea, Beta vulgaris ssp. maritima and Nucella lapillus and lead to new insights in the population biology of these species.