Date of Award
Doctor of Philosophy (PhD)
Forestry and Natural Resources
Committee Member 1
Committee Member 2
Ecological data is inherently spatial; however, it is still the norm to model ecological data as spatially invariant. Failure to account for spatial structure in response variables and modeled relationships can result in inflated coefficient values, shifts in the relative importance and sign of predictors, cross-scale contradictions in relationships, and reduced predictive power due to the averaging of modeled relationships. When ecological models are used to support conservation decision-making, model error can be costly leading to both misallocation of limited resources and distrust of science-based management.
My dissertation focuses on developing methods to account for spatial structure in two models commonly used to inform conservation decisions. Both chapters focus on the imperiled hellbender salamander (Cryptobranchus alleganiensis) and were designed to provide guidance on the conservation and management of a species that is facing precipitous declines throughout much of its range. In chapter one, I modeled the relationship between the hellbender genome and climate and stream variables across the range of the species. I extended multiple matrix regression into a mixed modeling framework to account for strong spatial population structuring. The approach improved model fits, shrunk coefficient estimates, and increased the concordance of model results with an independent analysis of locus-specific environmental associations. The results of the model were used to forecast genomic vulnerability across the range of the species and the resulting map suggested a potential genetic mismatch between current and future conditions in portions of the range that accommodate stable populations.
In chapter two, I developed a species distribution model to help target sampling and translocation locations for eastern hellbenders (Cryptobranchus alleganiensis alleganiensis). It extended presence-only modeling into a mixed modeling framework to help account for autocorrelation and nonstationarity in the intensity of hellbender occurrences and unexplained environmental heterogeneity across physiographic provinces. The spatially explicit approach improves overall model discrimination and dramatically improves model performance in regions most in need of conservation guidance. Taken together, the chapters provide flexible methods to improve the performance of common ecological models and tangible products to support hellbender conservation.
McCallen, Emily Boersma, "Spatial Genetic and Distribution Modeling for the Conservation of Hellbender Salamanders" (2018). Open Access Dissertations. 2021.