Urbanization greatly alters plant and pollinator communities and can affect pollinator movement and subsequent gene flow. Plants persisting in urban areas must adjust to local environmental conditions often different from those in which they naturally evolved, and cultivation techniques for landscaping species have developed traits suitable for existence in urban habitats. Cultivated varieties and native conspecifics often exist in geographic proximity, and if pollinator movement is not blocked by urban structural components, functional differences may negatively impact spatially proximate native populations. I used spatial analysis of successful pollination of Cornus florida to estimate how pollinator movement is affected by urban features. My results suggest that buildings and canopy are the most important components which influence reproductive success in urban habitats. Additionally, I compared functional responses of both adult and offspring C. florida cultivars and native plants to differential light environments in the urban and natural understory habitats, and we found differences in physiology and morphology that could lead to negative fitness consequences for native populations should gene escape from urban cultivar to native populations occur via pollinator movement.

This study examined aspects of local reproductive variation in the flowering dogwood (Cornus florida L.) coincident with recent differences in primary canopy architecture. The dogwood trees in this study were impacted by a hurricane that created numerous treefall gaps, which created fine scale heterogeneity in the primary canopy. Fine scale disturbances in a forest can result in changes for multiple members of the forest community, including the reproductive responses of the trees and interspecific pollination mutualisms. Previously determined differences in offspring genetic structure suggested that pollen movement among genetically unstructured maternal individuals was significantly impacted between open, or disturbed, canopy and closed, or undisturbed, types. To further understand mechanisms by which this nonrandom mating occurred, this study examined both intrinsic and extrinsic reproductive factors for C. florida. The first chapter examines several parameters of the reproductive phenology of sample groups of dogwood between the canopy types as intrinsic factors. The parameters observed included initiation, course and termination of flowering, seed set and seed germination ratios. The results showed some significant differences between the sampling areas in flowering intensity, although the majority of the variation was between maternal individuals throughout the entire sample. The second chapter examines insect community composition across areas of differential canopy disturbance. Community analysis showed that the open canopy contained a larger and more diverse assemblage of insects than the closed canopy or the field, which represented an area of complete canopy removal. This shift in insect community composition may have created functional differences in the local pattern of pollen flow by altering the functional composition of local potential pollinator assemblages. The major finding of this study was that the impact of intermediate natural disturbance on the mating systems of understory plant populations may be more indirect than direct. In conclusion, the observed differences in insect community structure within these habitat types in this study were consistent with predetermined patterns of pollen flow; this structure can explain some of the previously observed genetic structure within locally proximate understory tree populations.

This study examined the impact intervening environment has on gene flow in the insect pollinated understory tree, Cornus florida L., by combining GIS and landscape genetic techniques (Least Cost Path Analysis, Circuit Theory, and Conditional Genetic Distance). Traditional population genetic analysis indicated pair-wise relatedness was significantly correlated to distance (Pearson; r = -0.312, P < 0.001) suggesting a spatial component to offspring relatedness. Dispersal throughout the study site was non-random, exhibiting a high degree of pollen pool structure due to restricted gene flow (Two-Generation Analysis; Φft = 0.161, P = 0.001). Forest structure was quantified in GIS layers representing coniferous canopies, mixed hardwood canopies, C. florida canopies, open understory (roads), and open understory/canopy due to tree removal. Of these layers, landscape isolation for the roads layer provided the best-fit model for describing genetic differentiation among sampled pollen pools (Mantel; r = 0.542, P = 0.001). These data also suggest that improved biological inferences can be gained by examining a range of landscape isolation models.

Introduction of invasive species can alter seed fate predictions made by the Janzen-Connell Escape Hypothesis (JCEH). The JCEH states that there is a suitable region around a plant that is ideal for seed germination, growth, and recruitment. Seeds dispersed too close to the maternal plant are subject to competition from the maternal individual and perhaps density-dependent predation, whereas seeds dispersed further away may end up in suboptimal habitats. Invasive species may change the amount of these suitable habitats for native plants by creating unsuitable light environments and as a result, may influence the size of the ideal recruitment zone surrounding a parent plant. This study examines the extent to which the invasive grass species, Microstegium vimineum, influences recruitment of the understory tree, Cornus florida. In general, M. vimineum was found to reduce both germination and early seedling success and may have significant consequences for future forest structure.

Flowering dogwood (Cornus florida L.) populations have experienced severe declines caused by dogwood anthracnose in the past three decades. Mortality has ranged from 48 to 98%, raising the concern that genetic diversity of this native tree has been reduced significantly. Microsatellite data were used to evaluate the level and distribution of genetic variation throughout much of the native range of the tree. In the first conducted study, we found that genetic variation in areas affected by anthracnose was as high as or higher than areas without die-offs. We found evidence of four widespread, spatially contiguous genetic clusters. However, there was little relationship between geographic distance and genetic difference. These observations suggest that high dispersal rates and large effective population sizes have so far prevented rapid loss of genetic diversity. The effects of anthracnose on demography and community structure are likely to be far more consequential than short-term genetic effects. The second study examined levels and distribution of genetic variation of C. florida throughout Great Smoky Mountains National Park (GSMNP). Significant genetic structure at both landscape and local levels were found. We infer that two genetic clusters exist within the park, mostly separated by the main dividing ridge of the Great Smoky Mountains. The differentiation is statistically significant, but subtle, with gene flow evident through low-elevation corridors. It seems unlikely that recent demographic dynamics have resulted in a depletion of genetic variation in flowering dogwoods.

Part 1: Visitation of potential pollinators to Cornus florida and C. kousa flowers was assessed in East Tennessee in 2008 and 2009. Data regarding insect visitation rates to multiple trees per location were gathered throughout the flowering period. Diurnal and seasonal variations in visitation were recorded. Pollen coverage was assessed on portions of captured insect exoskeletons that were most likely to contact the stigma. Eleven families in four insect orders were collected from C. florida and 26 families in five orders from C. kousa. The most important pollinators in eastern Tennessee were bees in the Andrenidae and Halictidae. The most common visitors to C. kousa flowers were scarab, cerambycid and cantharid beetles. Halictid bees were also frequent visitors to C. kousa. Part 2: Dispersion of flowering dogwood pollen in an orchard was evaluated by performing parentage analysis on open pollinated seedlings collected from a single maternal tree. Pollen source for 45 seedling trees were established using three polymorphic simple sequence repeat (SSR) loci. The expected leptokurtic distribution was not seen. Although the majority of paternal trees were within the 12 m radius of the study area, the trees most likely to donate pollen were some of the furthest away. Unusual pollen movement may be a product of lack of synchronization of flowering times between the mother tree and potential pollen donors in the area. Additionally relative proportions of certain pollinating insects (andrenid and halictid bees) may have moved pollen further than expected. Appendix 1: Pollen from 6 species in the genus Cornus was analyzed with Fourier Transform Infrared (FTIR) Spectroscopy to determine the utility of this tool to identify pollen grains to species. Although there were differences visible in the spectra, principal components analysis coupled with cluster analysis could not consistently identify pollen grains. Appendix 2: Floral volatile emissions from 6 species of Cornus were collected in a headspace chamber and analyzed with gas chromatography-mass spectrometry to determine if differences between emitted volatiles of C. florida and C. kousa explain differences in floral insect visitation. The relationship between floral volatile emission and the phylogeny of Cornus was also examined.

Part 1: Visitation of potential pollinators to Cornus florida and C. kousa flowers was assessed in East Tennessee in 2008 and 2009. Data regarding insect visitation rates to multiple trees per location were gathered throughout the flowering period. Diurnal and seasonal variations in visitation were recorded. Pollen coverage was assessed on portions of captured insect exoskeletons that were most likely to contact the stigma. Eleven families in four insect orders were collected from C. florida and 26 families in five orders from C. kousa. The most important pollinators in eastern Tennessee were bees in the Andrenidae and Halictidae. The most common visitors to C. kousa flowers were scarab, cerambycid and cantharid beetles. Halictid bees were also frequent visitors to C. kousa.Part 2: Dispersion of flowering dogwood pollen in an orchard was evaluated by performing parentage analysis on open pollinated seedlings collected from a single maternal tree. Pollen source for 45 seedling trees were established using three polymorphic simple sequence repeat (SSR) loci. The expected leptokurtic distribution was not seen. Although the majority of paternal trees were within the 12 m radius of the study area, the trees most likely to donate pollen were some of the furthest away. Unusual pollen movement may be a product of lack of synchronization of flowering times between the mother tree and potential pollen donors in the area. Additionally relative proportions of certain pollinating insects (andrenid and halictid bees) may have moved pollen further than expected.Appendix 1: Pollen from 6 species in the genus Cornus was analyzed with Fourier Transform Infrared (FTIR) Spectroscopy to determine the utility of this tool to identify pollen grains to species. Although there were differences visible in the spectra, principal components analysis coupled with cluster analysis could not consistently identify pollen grains.Appendix 2: Floral volatile emissions from 6 species of Cornus were collected in a headspace chamber and analyzed with gas chromatography-mass spectrometry to determine if differences between emitted volatiles of C. florida and C. kousa explain differences in floral insect visitation. The relationship between floral volatile emission and the phylogeny of Cornus was also examined.