382 / 2019-02-25 22:27:52

Identification and extraction of “ghost town” in urban area

ghost town,multi-source data,Urban vitality

全体主题 > Spatial data collection, Analysis, and Visualization

Ghost town, which has not a currently recognized definition, originally is proposed by economist David McWilliams for describing empty or unfinished housing developments in Ireland. Shepard defined “ghost city” as “a new development that is running at severe under-capacity, a place with drastically fewer people and businesses than there is available space for.” Nie and Liu attribute ghost towns to natural disasters, wars, the abandonment of houses and high vacancy rates. After the phenomenon of ghost city in Ordos City was reported by Time magazine, “China” with “ghost town” received extensive attention. In China, with the acceleration of urbanization, urban construction, especially the real estate projects, has been over promoted, the construction of many new areas far exceeds the actual needs. Therefore, the vast majority of ghost towns of China are due to high vacancy rates.
At present, about the study of phenomenon of ghost town is mainly based on the national level, some scholars have sorted some cities by indicator of phenomenon of ghost town, and identified a series of typical cities. An exception is that Chi using the location based service data from Baidu, derive the distribution of ghost town in urban with 500 by 500-m grid. Hence, in order to map the distribution of ghost town on the city level and explore what factors it is related to in the context of accelerating the construction of urbanization, we propose a high-resolution map-ping method of distribution of ghost town based on random forests in urban areas. Wuhan, as the research object, is a good case because it has developed rapidly in recent years.
The nighttime light data of Luojia-1A satellite whose resolution was higher than DMSP-OLS and NPP-VIIR, reflecting the intensity of nocturnal activity, can be used to judge the existence of empty houses. On the other hand, In SAR interferometry, coherence is the criteria of phase stability, stabilized objects in urban areas can be identified as buildings according to the coherence, so the ratio of coherence to the intensity of nighttime light can be used to construct a ghost city index. Generally, regions in urban with high coherence and low intensity of nighttime light have a high ghost city index, its actual ground is with some buildings but lack of nocturnal activities, which is in line with the public's cognition of ghost town. At the same time, in order to research the cause of the appearance of ghost town, or the relationship with other factors of urban, we borrow the theory of urban vitality of Jacobs J, in which the principle of urban space vitality is proposed. Table 1 divides the factors that affect the city's vitality into three categories. Based on the available data, we quantify the three types of impact factors for urban vitality.
Furthermore, Random Forest Regression was applied to simulate the distribution of ghost city index, taking various impact factors for urban vitality as independent variables. Random forest has a good effect in solving the multicollinearity caused by multi-dimensional feature combination. That the importance order of influencing factors is obtained by comparing the out-of-bag error of random ordering of one of variables with the original out-of-bag error, can initially determine the cause of the phenomenon of ghost city. On the other hand, the mapping resolution of ghost city index can be optimized from 170m(determined by the resolution of nighttime light data) to 25m(determined by the resolution of in-dependent variable) by fitted with random forest regression. Finally, the phenomenon of ghost town was verified by manually selecting samples.