Building and Evaluating Models to Estimate Ambient Population Density

The concept of human population density is quite simple: the number of persons occupying a given area. Nonetheless, practical representations of population density must use appropriate spatial and temporal scales of measurement to be useful. The 11 September 2001 attack on the World Trade Center in New York is a poignant example: “How many people were in the two World Trade Center buildings at 0830 local time?” Population density data derived from mostnational censuses is a residential measure of population density and consequently does not capture non-residential population density. Human mobility suggests that a nonresidential or ambient measure of population density may be a more useful representation for some applications. Ambient population density in this sense is a temporally averaged measure of population density that takes into account where people work, sleep, eat, drive, shop, etc. Short of implanting a GPS reciever into everyone’s skull and tracking their spatio-temporal behavior, it is extremely difficult to make direct measurements of ambient population density. This paper explores some theoretical and empirical efforts at estimating ambient population density and proposes a quantitative means for evaluating their validity. The three models of population density examined are LandScan, Gridded Population of the World (GPW), and a simple empirical model derived from nighttime satellite imagery provided by the Defense Meteorological Satellite Program’s Operational Linescan System (DMSP OLS). These measures are compared to both residential and employment-based measures of population density in the Los Angeles metropolitan area. The GPW, LandScan, and DMSP OLS models of ambient population density described here all make foundational contributions to future efforts at filling the gap in social, economic, and demographic information for parts of the world where such data are unavailable. The proxy measures of population density described here show promise for many applications, including improved mapping of population distribution and as a supplement to census enumerations in many parts of the world.