Calculating Change in Global Urbanization Using Nighttime Lights

Abstract:

Measuring urban change is important for policy makers at the municipal, regional, national, and international levels. However, obtaining urban metrics from countries is often difficult, and data-poor, developing countries have little to no capacity to measure urban change over long time periods. The goals of this analysis were to develop a standard methodology for extracting urban footprints from remotely sensed data, and to determine what environmental and economic conditions control urban brightness. For example, an urban area in a country with less development may have darker urban areas than their more developed counterparts, thus, country specific thresholds for separating urban from non-urban areas are necessary to appropriately analyze urbanization. Nighttime lights data, from the DMSP-OLS constellation, are well suited for this task, due to their global coverage, historical archive, and ease of accessibility for data-poor countries. In this project, we used ESA’s GlobCover dataset as our validation data, and developed a country-specific classification of urban vs. non urban cover using the nighttime lights data. Classification accuracies varied by country, but ranged from ~90-99%. Preliminary results show that brightness thresholds vary substantially both between countries, and as a function of geographic location. Finally, brightness thresholds and their associated accuracies were compared to environmental and economic indices through regression analysis, in an attempt to determine the drivers of urban brightness.