Using multi-source geospatial big data to identify the structure of polycentric cities

Abstract

Identifying the structure of a polycentric city is vital to various studies, such as urban sprawl and population movement dynamics. This paper presents an efficient and reliable method that uses multi-source geospatial big data, including nighttime light imagery and social media check-in maps, to locate the main center and subcenters of a polycentric city. Unlike traditional methods that rely on statistical data categorized by administrative units, the proposed method can effectively identify the boundaries of urban centers, and the data source guarantees a timely monitoring and update. Four main procedures are involved: 1) a new observation unit is developed using object-oriented segmentation; 2) main centers are located using cluster analysis (Local Moran's I); 3) subcenter candidates are selected using significant positive residuals from geographically weighted regression (GWR); and 4) final centers are filtered using global natural breaks classification (NBC). These steps can be reproduced in different regions. To evaluate the effectiveness, the method was applied to three rapidly developing Chinese cities: Beijing, Shanghai, and Chongqing with different natural and economic characteristics. The performance of the proposed method has been carefully evaluated with qualitative and quantitative analyses. Comparative experiments were also conducted across different datasets to prove the benefits of combining a social media check-in map with remotely sensed imagery in a human environment study.