Anthropogenic stressors on the environment: Identifying interactions between demographic and environmental factors in the conterminous United States

I demonstrate here that spatially extensive analysis can reveal considerable detail as to human-environment interactions of regional concern. In chapter 1, I used regression-trees to model human population density and relative change in population density in relation to environmental variables across the United States.; In chapter 2, I considered the utility of regression-tree analysis in identifying relevant spatial scales when relating the distribution of human populations to the biophysical environment. Our results imply that no single scale is optimal for socio-demographic analysis over a continental extent. Instead, initial global models need to be locally refined in a recursive but geographically specific manner using hierarchical models.; In chapter 3, I compared the efficiency of agricultural census versus remotely sensed data in explaining continental-scale patterns of bird species richness. The implications of using each database for predicting avian species richness are discussed.; In chapter 4, I described an approach to modelling the spatial distribution of human-environment interactions using principal component analysis (PCA) on a set of demographic variables. I modelled the environmental determinants of these axes using regression-tree analysis.; In chapter 5, I further developed the chapter 4 methodology but concentrated on the second principal component axis (β-settlement) which described relative population growth correlated with recent construction in non-agricultural areas. β-settlement was greatest in coastal and desert areas, and coincided with national concentrations of threatened and endangered species.; In chapter 6, I used regression-trees to separately model: (1) breeding bird species richness, (2) actual evapotranspiration, and (3) population density in relation to environmental variables. Our three models used the environmental variables as predictors to identify areas of interaction, as end nodes for each regression-tree. All locations in a given end node shared a unique combination of environmental constraints on the dependent variable and were concentrated regionally. These end nodes were numbered uniquely within trees and then overlaid in a GIS to identify areas of spatial overlap. These overlap areas thus defined the co-occurrence of individual patterns of the respective interactions of biodiversity, ecosystem function, and human activity with the biophysical environment; simplifying and defining the domain of process models of global change.