Validation of urban boundaries derived from global night-time satellite imagery

Abstract

Night-time imagery from the Defense Meteorological Satellite Program (DMSP) Operational Linescan System (OLS) has been proposed as a useful tool for monitoring urban expansion around the world, but determining appropriate light thresholds for delineating cities remains a challenge. In this paper we present a new approach. We used DMSP stable lights and radiance-calibrated images to delimit urban boundaries for San Francisco, Beijing and Lhasa, cities with different levels of urbanization and economic development, and compared the results against boundaries derived from high-resolution Landsat Thematic Mapper (TM) imagery. Unthresholded DMSP images exaggerate and shift the extent of these urban areas. We then calculated light thresholds that minimized the discrepancies between the DMSP- and TM-derived urban boundaries for each city. Our comparison highlights the difficulty of using DMSP data across areas with disparate urban characteristics, but suggests the possibility of calibrating this data source for monitoring growth of cities at comparable levels of development.