Covariables influencing effective pollen dispersal from source populations to single-plant isolates of Centaurea jacea.

Summary of generalized linear model analyses using quasi-Poisson errors and F-tests to explore the effects of covariables (number of species and number of individuals of flower visitor groups visiting potted plants of C. jacea during 60 min and habitat variables) on mean seed set of single-plant isolates (main effect of covariable) and the slope of the decline of this seed set (interaction of covariable with log(distance) from source population). Source populations were restricted to restored meadows and single-plant isolates were set up in surrounding intensively managed meadows at 25, 50, 100 and 200 m distance from source populations. A plus (+) indicates a positive effect, a minus (–) a negative effect. A positive effect of a covariable on the mean seed set of single-plant isolates indicates that more pollen was effectively transferred from source populations to single-plant isolates. A negative covariable x log(distance) interaction indicates a less steep decline of effectively transferred pollen with distance from source populations. In all analyses, the mean seed set in 2004 was used as the dependent variable, since the covariables were sampled in 2004 only. Main effects of the covariables population size, species richness of flowering plants in restored meadows and flower abundance in restored meadows were tested at the site level, all other covariables were tested at the single-plant isolate level. The fitted models are described in the “Statistical analysis” section.