Using Spatial Entropy of Urban Vegetation to Measure Neighborhood Stability in Shrinking Cities

Abstract

Previous studies on land-cover change have focused on urban growth and its consequences. However, urban shrinkage has also occurred as a consequence of global economic transformations. Urban shrinkage can have profound consequences and change the spatial patterns of urban vegetation. To detect and predict urban shrinkage is important for better urban planning and policy making. This study works on 1) determining the possible roles of spatial entropy, which represents the spatial configuration of urban vegetation, in combination with other socioeconomic variables, in predicting neighborhood stability and urban shrinkage, and 2) how the scale of defined neighborhoods may affect the relationship between spatial entropy and neighborhood stability. For the City of Detroit, MI, I adopted spectral mixture analysis of Landsat-8 imagery to yield moderate-resolution maps of urban vegetation proportion. I calculated spatial entropy for defined neighborhoods based on the vegetation information. Controlling for socioeconomic variables from parcel data and U.S. Census Data, I developed spatial models of the relationships between no-structure rate with neighorhoods, an indicator of urban shrinkage, and vegetation spatial entropy. Models were performed on two levels of neighborhoods: census block groups and census tracts. The results show that spatial entropy has the largest (negative) association with the no structure rate compared with other predictors on both levels of neighborhoods. While high-resolution imagery or parcel-based data were not readily available, this study shows that moderate-resolution imagery can be an effective source for detecting and predicting urban shrinkage.