A scale-adjusted measure of “Urban sprawl” using nighttime satellite imagery
Introduction
The issue of what is commonly referred to as “Urban sprawl” is gaining increasing attention and concern from citizens, environmental organizations, and governments (http://www.sierraclub.org/sprawl/; http://www.vtsprawl.org/index3.htm; (Benfield et al., 2001)). Concerns are raised about the impact urban sprawl has on the loss of open space, traffic congestion, and energy consumption. Nonetheless, specific, measurable, and generally accepted definitions of urban sprawl are difficult to find. William Whyte's 1958 definition of urban sprawl referred to patterns of urban development (“…the leapfrog nature of urban growth…”) (Whyte, 1958). Others have defined “Urban Sprawl” based simply on the aggregate population density of a given urban area Fulton et al., 2001, Kolankiewicz & Beck, 2001. It is very likely that “Urban Sprawl” happens to some extent in specific areas of most cities. It could be argued that “Urban Sprawl” is similar to pornography in that it is difficult to define but ‘You know it when you see it’. It could be argued that “Urban Sprawl” is a multi-dimensional phenomenon that needs to be characterized with several variables. Nonetheless, this research focuses on providing a single, scale-adjusted (population corrected), aggregate indicator of “Urban Sprawl” for all urban areas of population greater than 50,000 in the conterminous United States.
Studies using aggregate population density as an indicator of “Urban Sprawl” have typically used ‘urban area’ designations of the U.S. Census along with corresponding population figures to determine an average population density for urbanized areas within the US Fulton et al., 2001, Kolankiewicz & Beck, 2001. These aggregate measures of sprawl suffer from two problems: (1) problems associated with measurements of the areal extent of an urban area, and (2) the nonlinear variation of the aggregate population density of urban areas as a function of total population. Remotely sensed images of urban environments have great potential for delineating urban areas. GIS coverages of urban environments suffer from arbitrary administrative boundaries used in conjunction with housing unit density or population density thresholds. Nighttime imagery has some advantages over daytime imagery in that it is measuring emitted rather than reflected radiation, this avoids some classification problems in separating developed vs. nondeveloped land cover. This research utilizes a ‘scale-adjusted’ measure of “Urban Sprawl” that addressees the nonlinearity problem and uses two ‘thresholds’ of nighttime satellite imagery as a means of measuring the areal extent of urban areas in the United States.
The urban extent of cities varies as a nonlinear function of their total population Nordbeck, 1965, Stewart & Warntz, 1958, Tobler, 1969. This has also been demonstrated using nighttime satellite imagery as a proxy measure of urban areal extent both nationally and globally Sutton et al., 1997, Sutton et al., 2001. Typically, as cities grow their aggregate population density increases; consequently, the aggregate population density of large cities like Los Angeles and Chicago will be higher than the aggregate population density of smaller cities such as Portland and Kansas City. However, this does not imply that Los Angeles and Chicago suffer less from “Urban Sprawl” than Portland or Kansas City. Any aggregate measure of “Urban Sprawl” for an urban area should be scale-adjusted by the total population of that urban area.
The U.S. Census defines the urban area (UA) of a city each time a census takes place. A UA is designated for all central cities with a population in excess of 50,000. The urban areas designated by the census do not always correspond with land cover maps derived from satellite imagery (Vogelmann et al., 1998). This study uses nighttime satellite imagery provided by the Defense Meteorological Satellite Program's Operational Linescan System (DMSP OLS) to measure the areal extent of the urban areas of the conterminous United States. Imhoff et al. (1997) have used the DMSP OLS imagery in similar ways. The DMSP OLS imagery is compared to a gridded population density dataset derived from the 1990 U.S. Census (Meij, 1995). This comparison results in measures of both the areal extent and the population of all the urban areas in the conterminous United States. These numbers are then used to calculate improved aggregate measures of urban sprawl.
Section snippets
Data and methods
The data required to develop a ‘scale-adjusted’ measure of urban sprawl are simply: (1) the areal extent of urban areas, (2) the corresponding population of those urban areas, and (3) a formula describing the relationship between the population and areal extent of these urban areas. The data used to obtain areal extent and population are: (1) a radiance calibrated DMSP OLS image of the United States (Elvidge et al., 1998), and (2) a grid of population density derived from the U.S. Census (Meij,
Results and analysis
The regression line on the scatterplots of the Ln(Area) vs. Ln(Population) relationship represents a scale-adjusted “Sprawl Line”. The line itself represents the average relationship between the areal extent and population of urban areas in the conterminous United States. It should be noted that the nature of this sprawl line is specific to the United States. A ‘Sprawl Line’ for other countries will generally have a higher intercept for countries with lower GDP per capita (in general poorer
Discussion
These aggregate indicators of “Urban Sprawl” for the urban areas of the United States do not suggest that there are no areas of “Urban Sprawl” within urban areas above the “Sprawl Line”; however, these results are comparable yet different than previously determined aggregate numbers because they are scale-adjusted. It is very likely that the green urban areas in Fig. 5, Fig. 6 that are well above the “Sprawl Line” contain smaller areas within them that most people would characterize as “Urban
Conclusion
Clearly, measuring “Urban Sprawl” is a daunting task. Many people decide “Urban Sprawl” is happening in their backyard based on perceived negative experiences such as traffic congestion, changing demographics, or overall population growth that is correctly or incorrectly attributed to “Urban Sprawl”. This investigation presents a measure of “Urban Sprawl” that is scale-adjusted to the total population of an urban area and uses nighttime satellite imagery as an objective and uniform measure of
Acknowledgements
The author would like to gratefully acknowledge insightful suggestions, comments, and criticisms from the anonymous reviewers of this article.
References (16)
- et al.
Solving sprawl
(2001) Sprawl-weary Los Angeles builds up and in
(2002, March 10)- et al.
Radiance calibration of DMSP-OLS low-light imaging data of human settlements
Remote Sensing of Environment
(1998) - et al.
Who sprawls the most? How growth patterns differ across the united states
(2001) - et al.
A technique for using composite DMSP/OLS “city lights” satellite data to map urban area
Remote Sensing of Environment
(1997) Jobs, migration, and emerging urban mismatches
- et al.
Weighing sprawl factors in large U.S. cities
(2001) - Meij, H. (1995). Integrated Datasets for the USA, Consortium for International Earth Science Information...
Cited by (257)
Defining urban sprawl: A systematic review of 130 definitions
2024, Habitat InternationalUrban intensive land use and enterprise emission reduction: New micro-evidence from China towards COP26 targets
2022, Resources PolicyCitation Excerpt :Recently, the increasing abundance of remote sensing information data has provided a new direction for accurately measuring intensive land use, one that is more focused on describing the land distribution pattern of intensive land use. Sutton (2003) identifies built-up land and non-built-up land with the help of DMSP/OLS night light data and population data from the same period and constructs an elasticity indicator of land-population growth to quantitatively measure urban intensive land use. Based on US satellite images, Burchfield et al. (2006) measure the proportion of undeveloped land within 1 km of urban built-up areas and use this to measure the degree of intensive land use.
Estimating freight production/attraction using nighttime satellite imagery
2022, Communications in Transportation ResearchUrban footprint detection from night light, optical and SAR imageries: A comparison study
2022, Remote Sensing Applications: Society and EnvironmentImpact of temporal compositing on nighttime light data and its applications
2022, Remote Sensing of EnvironmentExploring the effect of urban sprawl on carbon dioxide emissions: An urban sprawl model analysis from remotely sensed nighttime light data
2022, Environmental Impact Assessment Review