Research Article: Comprehensive framework for visualizing and analyzing spatio-temporal dynamics of racial diversity in the entire United States

Date Published: March 30, 2017

Publisher: Public Library of Science

Author(s): Anna Dmowska, Tomasz F. Stepinski, Pawel Netzel, Luís A. Nunes Amaral.


The United States is increasingly becoming a multi-racial society. To understand multiple consequences of this overall trend to our neighborhoods we need a methodology capable of spatio-temporal analysis of racial diversity at the local level but also across the entire U.S. Furthermore, such methodology should be accessible to stakeholders ranging from analysts to decision makers. In this paper we present a comprehensive framework for visualizing and analyzing diversity data that fulfills such requirements. The first component of our framework is a U.S.-wide, multi-year database of race sub-population grids which is freely available for download. These 30 m resolution grids have being developed using dasymetric modeling and are available for 1990-2000-2010. We summarize numerous advantages of gridded population data over commonly used Census tract-aggregated data. Using these grids frees analysts from constructing their own and allows them to focus on diversity analysis. The second component of our framework is a set of U.S.-wide, multi-year diversity maps at 30 m resolution. A diversity map is our product that classifies the gridded population into 39 communities based on their degrees of diversity, dominant race, and population density. It provides spatial information on diversity in a single, easy-to-understand map that can be utilized by analysts and end users alike. Maps based on subsequent Censuses provide information about spatio-temporal dynamics of diversity. Diversity maps are accessible through the GeoWeb application SocScape ( for an immediate online exploration. The third component of our framework is a proposal to quantitatively analyze diversity maps using a set of landscape metrics. Because of its form, a grid-based diversity map could be thought of as a diversity “landscape” and analyzed quantitatively using landscape metrics. We give a brief summary of most pertinent metrics and demonstrate how they can be applied to diversity maps.

Partial Text

Census Bureau population projections [1] indicate that the racial dynamic in the U.S. will steer the country toward a society with no absolute racial majority by 2044. How this overall prediction translates to a change in the racial makeup of local neighborhoods is of great interest to academics, as well as to policy makers and to the general public. Racial makeup is quantitatively analyzed in terms of segregation or diversity. A classic definition of racial segregation is the physical separation of two or more groups into different neighborhoods [2]. From a spatial point of view, segregation is the spatial pattern reflecting varying contributions of two races to local sub-populations. When the population is multi-racial rather than bi-racial, its makeup is studied in terms of diversity rather then segregation; diversity is a spatial pattern reflecting varying contributions of all races to local sub-populations. Analysis of segregation and/or diversity capable of yielding results which are lucid and useful to all stakeholders is important for a better understanding of the processes responsible for the observed segregation/diversity and to better inform the decision makers.

The Census collects race data at the ultimate resolution of an individual household (individual level). In Geographic Information System (GIS) terminology such data is referred to as point data. However, this data is not released to the public in its original format due to privacy concerns. In some other countries, most notably in Sweden, individual level data is made available to scientific organizations under strict supervision [27], but this is not the case in the U.S. There are two GIS formats in which the Census could release the data, vector (shapefile) which aggregates point data to population counts within Census units, or raster (grid) which models the point data to obtain cell-based population density. For predominantly historical reasons [28] the Census releases data in aggregated (vector) format. As this has always been the case, this format may appear as the natural choice for providing information about population characteristics, but, in reality, it has multiple shortcomings in comparison to the grid format. Table 1 summarizes the major shortcomings of the aggregated format of population data and contrasts them with the advantages of using gridded data.

The advantages of gridded population data over aggregated data are purely academic in the absence of complete and accurate gridded data. Until now no such data has been available. The Socioeconomic Data and Application Center (SEDAC) ( provides 1 km resolution (250 m for selected metropolitan areas) demographic grids. However, in addition to having a resolution which is too coarse for segregation studies and being the result of an oversimplified gridding model (areal weighting), these grids are only available for the years 1990 and 2000. A higher resolution (90 m) US–wide demographic grid, presumably based on the most recent Census data, is under development by the Oak Ridge National Laboratory [42]. This project, called LandScan–USA, aims at providing both nighttime (residential) as well as daytime population densities, but it is not currently available, nor is it expected to be in the public domain once it becomes available.

All our grids (total population, race sub-populations, and racial diversity) are freely available for download. Data can be accessed by two different methods: (1) using our GeoWeb mapping application SocScape, and (2) using county-by-county or MSA-by-MSA download. Table 2 summarizes what gridded data is available for download using the two methods. It also provides detailed information on original Census data used in our models.

To demonstrate the process of analyzing racial diversity using a diversity grid, we use the greater Chicago area as an example. Using the download tool in SocScape we selected an area as shown in the upper row of Fig 3 and downloaded MYC diversity maps for this area for 1990, 2000, and 2010. These maps open in GIS software without any additional processing; they are grids containing 1788 × 1749 cells; each cell’s color corresponds to a specific community (see Fig 2 for legend). To better show details we also enlarged the central area of Chicago and displayed the enlarged maps in the lower row of Fig 3.

A diversity grid not only provides a compelling visualization of the overall racial/diversity condition in an area of interest (hereafter referred to as a site), but it also can be used as an input (instead of several race sub-population grids) for quantitative analysis of “diversity landscape.” First, that a diversity map is a categorical raster. Closer inspection (see the lower row of panels in Fig 3) reveals that a diversity map consists of a large number of patches, where a patch is defined as a contiguous group of same-color (same community category) cells. All patches of the same color indicate an area inhabited by a given community. Thus, the diversity grid has the same data format as a landscape in the field of landscape ecology [54]. Quantitative analysis of ecological landscapes is a highly developed topic [32, 33] with decades of development and an established software package (see below) for quantitative analysis. All this experience and methods of analysis can be carried over to diversity landscapes without modification.

Our goal in this paper was to present a comprehensive framework for visualizing and analyzing spatio-temporal dynamics of racial diversity in the entire conterminous U.S. We believe that all parts of the presented framework lower barriers for investigating issues related to segregation and diversity in the United States.




0 0 vote
Article Rating
Notify of
Inline Feedbacks
View all comments