Spatial Cluster Analysis Of Urban Landscape Pattern Using Stable Nighttime Light Satellite Images

Under the circumstance of economic globalization, urban spatial cluster (USC) is a result of spatial integration of diverse industries and gradually becomes a main area for the redistribution and transferring of essential production factors. It plays a crucial role in the country’s economic development. How closely the cooperation and interaction between cities within an USC can be carried out directly determines the economic competitiveness of this region as a whole when competing with other regions or countries. The better understanding of this kind of urban phenomena is critical for better management and governance and for the maintenance of the economic competitiveness of these cities. Traditionally, in the field of urban geography, USC are defined and identified based on statistical analysis of social-economic variables, such as urban area, population size and density, economic interaction, and commuter rate. This puts forward a high requirement for the veracity and integrality of these social-economic datasets. And also, for traditional studies, there is a shortage of methods to investigate the spatial pattern and morphology of an USC and to extract the spatial relationship between cities within this cluster. The introduction of Remote Sensing (RS) and Geographic Information Science (GIS) into urban geography offers a new way to obtain spatial information of USC. The existing studies have proved of the great advantage of Remote Sensing technologies on acquiring and utilizing urban spatial information, such as urban extent delineation, urban impervious surface mapping, urban growth monitoring etc. However, most of them focus on the analysis of urban expansion in a small area by using high resolution (HR) or very high resolution (VHR) remotely sensed images. And few are conducted to identify USC at national or continental scale. In order to solve these problems, based on stable nighttime light satellite image, this paper establishes a theoretical foundation for the USC analysis from the perspective of spatial proximity and Gestalt theory, investigates the implementing process of the method we proposed based on Graph theory and Object-oriented method and at last applies it to the identification of USC in the mainland China.The main findings of this paper are as follows: 1. The essential principle that underlies the identification of USC is discussed. USC results from the spatial expansion of cities within a specific region and is a spatial integration of the functions possessed and offered by these cities. The distribution of these cities obeys to a basic rule-proximity, which means new city tends to develop at the location near the other cities in this area. Therefore the identification of USC is in fact a compartmentalization of the geographical space and then to differentiates these cities from the others around them.2. The theoretical foundation and framework of methodology for USC identification is established using satellite images with a comparatively coarse spatial resolution. From the perspective of human spatial cognition, the identification and regionalization of USC is actually a Gestalt-resembling issue. There is a common rule underlies both of them-the proximity. Thus, the methods widely utilized in the field of Gestalt Theory can be adopted for our USC analysis. With this recognition, this paper divides the USC identification into three detailed questions:1) how to acquire the spatial distribution of cities;2) how to represent the spatial proximal relationships of cities; and3) how to group the cities into clusters based on the spatial proximity. To conform to the reality, in our analysis urban built-up area is treated as two-dimensional urban object instead of point. And a minimum spanning tree (MST) is then generated to extract and represent the proximal relationship between urban objects. After that, a set of rules consistent with Gestalt Theory is applied to partition this MST and obtain the targeted USC.3. Algorithms for identifying urban objects, generating and partitioning MST, identifying USC are implemented and improved. The algorithms implemented in this paper are recursive connected-region identification and marking, morphological operation (filling, closing), two-dimensional object attributes computing algorithm, object’s inner boundary tracing algorithm and USC identifying algorithm. The algorithm of generating MST for two-dimensional objects is improved in our research, which is a combination of inner boundary tracing algorithm,d∞-propagation Euclidean distance transformation algorithm and the common MST generating algorithm.4. The methods presented in this paper are successfully applied to the identification of USC in mainland China. In this application, we numerically and automatically identified urban spatial clusters of mainland China at the national level for the first time. In addition, the constituent structure attributes of urban spatial cluster are defined to quantify the inner links of urban objects in a cluster. Three levels of urban spatial unit, including urban agglomeration, urban cluster and urban cluster influencing region are defined and used to represent the relative linking strength between the cities. Our analysis results are consistent with previous studies, but provide much more detailed quantitative information.