Urban development is known to alter both habitat connectivity and, consequently, dispersal, as well as local environmental conditions, two important determinants of plant community structure. However, few studies have investigated the relative importance of dispersal and environmental filtering in urban landscapes. We used data on plant community composition from 50 riparian forest sites in the Research Triangle area of North Carolina, an urban landscape with seemingly well-connected remnant forest patches, to ask: 1) Is there evidence for seed dispersal limitation across the Research Triangle area? And 2) what is the relative importance of dispersal and environmental conditions for community composition? We used redundancy analysis and variance partitioning to determine the variance in plant species composition explained by environmental variables, measures of urban development (e.g. impervious surface cover, population density), and spatial structure (Moran’s Eigenvector Maps). We analyzed spatial structure using three measures of spatial distances between sites: Euclidean distance, distance along stream networks, and distance of least cost paths ("LCPs") calculated from resistance surfaces based on land cover. We expected that the degree of dispersal limitation would vary across modes of dispersal (unassisted, vertebrate-, wind-, or water-dispersed), and that LCPs and stream networks are better models of habitat connectivity than Euclidean distance, particularly for vertebrate-dispersed and water-dispersed species, respectively.
Results/Conclusions
We found little evidence for dispersal limitation across all plant species, and generally found that environmental conditions (soil texture, elevation, soil pH, and air temperature) explained more variation in species composition across sites than did spatial models or measures of urbanization. However, as expected, plant species with unassisted dispersal showed evidence of dispersal limitation. Unexpectedly, unassisted plant species composition was better explained by spatial structures related to LCPs (R2=0.23) than Euclidean distance (R2=0.10), suggesting that species with unassisted dispersal may be more affected by fragmentation of habitat by urban development than other species. Horticultural species also showed strong spatial structure in community composition related to LCPs rather than Euclidean distance. We had expected the most important determinant of horticultural species composition to be urbanization measures, but instead it appears that dispersal between forest patches, rather than only from urban areas, influences the presence and abundance of horticultural species across this landscape. These results demonstrate the importance of habitat connectivity for maintaining the composition of some species in urban landscapes, but that environmental conditions determine the distributions of most species across landscapes with relatively well-connected remnant habitats.