An invasive exotic insect, the hemlock woolly adelgid (HWA), Adelges tsguae , threatens to kill all eastern hemlock trees, Truga Canadensis , in North American forests. Hemlocks provide important structure, habitat, economic benefits, and aesthetic values in northeastern mixed forests. HWA and hemlock mortality are spreading quickly through eastern North America. Lack of research has forced the simplifying assumption that HWA populations and range expansion are uniform, which has led to potentially inaccurate crown sampling methods, obscured possibilities for mitigation, and driven unnecessary pre-emptive salvage logging. This dissertation tests the heterogeneity of HWA population density and range expansion at multiple scales by designing appropriate sampling methods, measuring HWA in four stands over three years, studying seedling infestation, and modeling the geographic variability of spread.

The results of this dissertation demonstrate the utility of randomized branch sampling in forest entomology and the power of advanced statistical models for the study of invasive species. Multiple scales of analysis unveil ecologically significant patterns not obvious at any single scale and highlight the importance of heterogeneity in exotic invasions. Randomized branch sampling provides design unbiased and detailed estimates of HWA density from individual branch samples. Branch samples show that the tops of tree crowns have larger HWA populations than lower portions of the crown under certain conditions. In addition, some forest stands maintain significantly lower HWA populations than others, which presents the opportunity for silviculture to encourage hemlock. Samples from multiple years reveal that HWA infestations change over time, but the relative densities across the forest remain consistent. Field planting document that the infestation rate of seedlings and the HWA density on infested seedlings is heterogeneous across stands. Plantings suggest that T. chinensis is a better candidate for reforesting hemlock stands than T. heterophylla . Models of range expansion demonstrate that HWA is spreading at a median of 12.5 km per year and much more slowly in areas in areas with an average annual minimum temperature less than -26°C. The heteroscedasticity and spatial correlation of the data require modeling techniques, such as classification trees and quantile regression, that are appropriate to these types of data.