“Ghost cities” identification using multi-source remote sensing datasets: A case study in Yangtze River Delta

- A feasible framework for identifying “ghost cities” is proposed using multi-source remote sensing datasets.
- The spatial characteristics of “ghost cities” are analyzed.
- Counties are vulnerable to the phenomenon, and highly developed cities lower “ghost city” intensity for ambient regions.
- The key contributing factor is explored by using the ternary contour.

Drastic urbanization has taken place in China during the last two decades. Recently, a considerable number of “ghost cities” have emerged due to the extensive and unreasonable urban expansion which far exceeds the practical demand. In order to investigate “ghost cities”, we proposed a feasible framework by utilizing multi-source remote sensing datasets, including nighttime light imagery, land cover type products and population grid. After eliminating blooming effect of nighttime imagery by a proposed modified optimal threshold method (MOTM) and extracting built-up area from land cover type products, we developed a “ghost city” index (GCI) to quantify and evaluate the intensity of “ghost city” phenomenon in Yangtze River Delta at county/district level. The GCI was established according to the intrinsic features of “ghost cities”, comprising three criteria: consistency of lit area and built-up area, illumination intensity and population density. Then, we explored the spatial pattern of “ghost cities” of different GCI categories and the ternary contour was applied to demonstrate the key factor among three criteria. Our finding implies that “ghost cities” are prominently spatial clustered. Meanwhile, counties and new development zones have higher risk of suffering from the phenomenon, while capital cities and municipal cities have an alleviative effect for ambient regions. Besides, regions with higher intensity of the phenomenon tend to have less balanced composition among three criteria. Our results show good consistency with previous reports and studies, providing a more objective and spatial explicit insight into the “ghost city” phenomenon. Our findings do not only prove the capability of monitoring “ghost cities” using remote sensing data, but would also be beneficial to urban planning and regional sustainable development.

267