Social-ecological gradients and city-regional geospatial structure across the United States

Urbanization diversly manifests across social and ecological dimensions. Gradient analysis, a technique capable of systematically assessing relationships across space, provides insight into urban geospatial structure. Through a combination of geo-computation and automation, we construct urban gradients for 370 US city-regions across 13 metrics that span social, physical, landscape, and emissions domains. To acknowledge urban shape irregularity, modeling leverages an absolute- and relative-distance measure that normalizes distance using the urban periphery. These two approaches, augmented by an urban-only model, yield more than 14 k linear regression models that demonstrate how most urban social-ecological phenomena follow an outward gradient. Model fits are loose, likely due to urban heterogeneity and methodological challenges, but only one metric (Social Vulnerability) had a gradient that was not linear. Importantly, model significance and fit both benefit from implementation of the relative-distance predictor. To synthesize model outputs (slope and y-intercept), 165 urban regions are processed using k-means and hierarchical cluster analysis. Results identify four to six distinct geospatial structures across the US. Hierarchical clustering is driven by variables corresponding to urban core intensity (population density, imperviousness) and functioning (housing type, GHG emissions), while the more flexible k-means clustering produced geographically representative divisions based on tree canopy, steep slopes, air quality, and household income. This work represents both a novel analysis of geo-spatial structure and synthesis useful to academics and policymakers alike. The identified clusters support a context-sensitive approach to sustainability or resilience and aid in identification of peer-cities useful for transfer of policy and planning solutions.