Pollen dispersal, the main component of overall plant gene flow, generally decreases with increasing distance from the pollen source, but the pattern of this relationship may differ among sites. Although site-based differences in pollen dispersal may lead to over- or underestimation of gene flow, no studies have investigated pollen dispersal patterns among differing urban site types, despite the incongruent range of habitats in urban areas. We used paternity assignment to assess pollen dispersal patterns in a wind-pollinated species (waterhemp; Amaranthus tuberculatus) and in an insect-pollinated species (tomato; Solanum lycopersicum) in experimental arrays at four disparate sites (two roof-level sites, two ground-level sites) in the New York (New York, USA) metropolitan area. The number of seeds or fruits, a proxy for the number of flowers pollinated, decreased with increasing distance from the pollen donors at all sites for both species. However, the mean number of Amaranthus tuberculatusseeds produced at a given distance differed two-fold among sites, while the slope of the relationship between Solanum lycopersicumfruit production and distance differed by a factor of four among sites. Pollen dispersal patterns may differ substantially among sites, both in the amount of pollen dispersed at a given distance and in the proportional decrease in pollen dispersal with increasing distance, and these effects may act independently. Accordingly, the capacity of plant species to adapt to climate change and other selection pressures may be different from predictions based on pollen dispersal patterns at a single location.