Empty spaces: neighbourhood change and the greening of Detroit, 1975–2005

Abstract

This paper investigates the disappearing residential geography of Detroit, Michigan, between 1975 and 2005 by examining the relationship between the ‘greenness’ of the urban landscape and the structural thinning of residential areas via satellite imagery and census data. The study uses normalized difference vegetation index (NDVI) and key housing variables as a proxy for observed changes in neighbourhood structure that correspond to the neighbourhood life cycle. Ordinary least squares and geographically weighted regression (GWR) were used to visualize the observed trends and performance of the models across space. Results from GWR analyses suggest the shifting residential geography of Detroit has changed from uniformly developed residential blocks to neighbourhoods that have experienced severe structural thinning across an urban landscape characterized by uneven development. The performance of the study models and parameters demonstrate how the relationships among NDVI and housing indicators, though significant, have diminished over time; this trend runs counter to green models applied to other urban landscapes, particularly those that follow the standard neighbourhood life cycle. Based on the empirical results, the study demonstrates the importance of understanding local histories and the broader socio-spatial context of cities when designing and implementing socio-spatial applications of remote sensing technologies.

Keywords:

• residential thinning
• Detroit
• geographically weighted regression
• NDVI
• greenness
• neighbourhood life cycle

Acknowledgements

The authors acknowledge the technical insights of Ryan Jensen (Brigham Young University) and Eric LaFary (University of Auckland). Additionally, the authors appreciate the constructive critique of Lynn Patterson (Kennesaw State University). As indicated in the methods section, census and spatial data were downloaded from the Minnesota Population Center's National Historical Geographic Information System: (http://www.nhgis.org). Finally, travel support for this project was provided by Indiana State University.

Notes

1. The maps demonstrate how the individual model performance varies across space visually and are not intended to be comparative per se. As such, the legends are distinct.