Identifying the Central Business Districts of global megacities using nighttime light remote sensing data

The Central Business Districts (CBDs) are important hubs for urban economic activities. In the context of globalization, a unified CBD boundary would greatly facilitate the study of global CBD comparative analysis, urban socio-economic development, urban transport and commuting mode. However, previous studies have encountered challenges in developing a practical method for global CBD identification due to limitations in data sources and methodologies. In this study, we selected 32 global megacities as research objects and used the open-access Black Marble nighttime light (NTL) products to construct indicators with the intensity and angular effects of NTL. Clustering and decision tree were then employed to derive rules for CBD identification. Results show that combining Z-score indicators and the strategy of clustering 32 cities before decision tree classification could improve the accuracy of CBD identification, which achieved a producer accuracy of 85%. The 32 cities were clustered into three types, i.e. U.S.A.-like, the mixed type, and China-like. The rules for CBD identification became more complex in the above order, but the accuracy decreased in turn. This study provides a new CBD identification method for cities lacking reference data, allowing for the delineation of unified and comparable CBD boundaries on a large scale.