Resumen
To better understand the persistence of residential racial segregation in U.S. cities, it is essential to develop testable, spatially explicit models of racial dynamics. However, the original census data are not formatted in a way that facilitates the testing of such models. In this study, we propose a novel geoprocessing pipeline that transforms census data into image-like geovisualization of urban segregation. The process consists of converting original data to a grid format, transforming subpopulation shares in each grid cell into a CMYK color, and multiyear clustering of CMYK values followed by classification and smoothing of images. When applied to data from multiple censuses, our methodology generates a time series of snapshots that depict the spatio-temporal dynamics of racial change. As a demonstration, we have utilized our methodology to generate racial images for a period spanning from 1990 to 2020 in seven major cities. Detailed results are shown for Chicago and Atlanta. By visually examining these time series, we have identified patterns that challenge the conventional Schelling model of racial change. Instead, our findings point towards a model that incorporates the interplay of preferential growth and diffusion as significant factors that shape the complex dynamics of racial composition in these urban areas.