Characterizing the complex influence of the urban built environment on the dynamic population distribution of Shenzhen, China, using geographically and temporally weighted regression

Abstract

Big data can provide new insights for smart city planning. This study exploits mobile-phone locating-request (MPLR) data as a proxy for real-time intra-urban population distribution. It models the relationship between the dynamic population distribution and the urban built environment using geographically and temporally weighted regression (GTWR), which can account for spatial and temporal non-stationarity simultaneously. A case study is undertaken based on MPLR records in Shenzhen, China and points of interest-based urban environment data aggregated to grid zones. Compared with previous models, GTWR yields a better result. Furthermore, the spatiotemporal coefficients are analyzed and compared. The results suggest that the patterns of urban population distribution are more complex during weekends than during weekdays. The coefficients of the company density variable are significantly higher during weekdays than weekends, while the coefficients associated with residential buildings are lower during weekday afternoons. Hence, the urban built environment plays an important role in the dynamic distribution of the population at different times. The findings show that the GTWR model in combination with MPLR and points of interest-based urban environment data can assist urban planners in gaining a better understanding of the spatiotemporal dynamics of the population distribution, thereby providing potential inputs to the rational allocation of public resources over space and time.