Sources of genetic structure in the woody perennial Betula occidentalis

We designed a transect-based survey of allozyme variation in water birch, Betula occidentalis, to investigate causes of genetic structure common to many temperate woody perennials: large-scale Holocene range shifts, landscape barriers to gene flow, and introgressive hybridization. We sampled subpopulations at low, medium, and high elevations at approximately every degree of latitude along the eastern Rocky Mountain front. Genetic diversity was relatively even across the latitudinal range of the species (H-e = 0.19), but this was a result of two opposing forces. Introgressive hybridization with the more northerly distributed hexaploid, Betula papyrifera, has apparently enriched variation in that direction. But native variation decreased in the same direction because of a loss of rare alleles along this known postglacial migration route. Two latitudinal frequency clines were attributed to an inter- and an intraspecific source, but clinal variation was not a feature at other loci. High levels of gene flow were detected at scales of up to 660 km in a hierarchical analysis of gene flow. This is a pattern expected of highly outcrossing species with wind-dispersed pollen and seed. In riparian-distributed species such as water birch, wind dispersal is probably much more common within mountain valleys than between mountain valleys. Instead, pairwise estimates of historical gene flow,(M) over cap followed the opposite pattern. Elevational barriers within valleys seem to be stronger than topographical barriers between valleys. We tested for isolation by distance using subsets of the data that either removed or maximized the effect of elevation. When population pairs were restricted to the same elevation, the isolation by distance model explained 58% more variation than that of the entire data set, but when pairs were selected from maximally different elevations, the slope of the regression was not significantly different from zero. Thus, systematic sampling across a predictable barrier to gene flow introduced noise into our data set. Such noise may be present in many other data sets, in which researchers have haphazardly sampled across ecological barriers to gene flow.