We examined the distribution and abundance of the invasive Hieracium species (hawkweeds) in the dry grasslands of the Upper Waitaki Basin (Canterbury) and Otago, using measures of Hieracium species frequency and hawkweed cover from 301 vegetation plots. Average hawkweed cover was significantly less in Otago than Canterbury. Hawkweed cover was also lower on drier sites, with hawkweeds having less cover at lower elevation, on more xeric sites and, in Canterbury, on soils with a lower moisture holding capacity. Environmental variables (elevation, moisture index, soil class, and measures of rabbit activity and the amount of recent pasture development) explained 43% of the variation in hawkweed cover in Canterbury, but only 25% in Otago. Furthermore, having controlled for environmental differences among plots, hawkweed cover in Otago was significantly spatially autocorrelated such that plots on the same property had similar hawkweed cover, but tended to differ from plots on other properties. This pattern is consistent with the hypothesis that variation in property management influenced the degree of hawkweed infestation. However, the same pattern was not found in Canterbury. We propose a general model to explain the observed patterns. The probability that a site will be invaded by hawkweeds is a function of: (1) the suitability of the site for hawkweed establishment (a function of the environment and past management), and (2) the size of the hawkweed propagule rain. The probability of hawkweed invasion is high when one or both of these variables have high values. Following initial invasion, there will be more local propagules coming from newly established plants and hawkweed cover at a site will increase, eventually stabilising at a level predictable from the environmental conditions at the site. This model predicts that the sites most suited for hawkweed establishment would be invaded first and fill the quickest. These susceptible sites would then serve as foci, providing propagules for subsequent hawkweed spread into adjacent, less invasion prone areas. The model predicts that the initial increase in hawkweed cover at a site should be exponential, and that there should be a lag in hawkweed spread whereby less susceptible sites are invaded later or at a slower rate. Both predictions are supported by data from monitoring trials. The model can explain the lower abundance of hawkweeds, the less predictable distribution of hawkweeds in relation to environmental variables, and the significant spatial autocorrelation in hawkweed cover in Otago compared to Canterbury, if hawkweed invasion in Otago is less complete than in Canterbury.