小世界网络特征识别的城市交通状态网络自相关分析

摘要

大量研究表明,考虑了空间多维性的小世界网络模型对于研究交通网络特性对交通流的影响很有必要.但现有相关研究普遍未能有效找出小世界网络特征产生的临界阈值进而判断网络结构特征对城市交通运行的影响.以伦敦市主干道路网络为案例,通过网络环境下的Moran's I和Getis-Ord's G两个统计量,找出道路网络结构的小世界特征及其网络动力产生的临界阈值,并利用基于交通网络距离的半变异(方差)函数分析该临界阈值是否合理,然后运用网络自相关分析小世界网络结构特征及其网络动力对道路交通流的影响.结果表明:①伦敦市主干道路网络的全局Moran's I和全局Getis-Ord's G的收敛值是0.34,小世界网络现象或动力出现的临界阈值是17000 m;②半方差图显示估算的小世界网络特征出现的阈值距离远小于理论半变异函数模型估算的空间相关性作用范围,且在该阈值附近半变异值呈明显上升趋势,表明该阈值是合理的;③在此阈值影响范围下,伦敦市主干道路段间交通流网络自相关性总体上呈显著正自相关特征,其局域空间关联格局呈中心区低-低和外围高-高的分布模式.%A number of studies have shown that the small world network model plays an important role in the effects of the characteristics of traffic network on traffic flow. However, the previous researches existed few effective approaches to find the threshold which was generated by the small world network features and explored the influence of network structure characteristics on traffic flow. In this study, taking the London trunk road as an example, we tried to find out the critical threshold of the small world characteristics and network dynamic of the road network structure by using the network autocorrelation statistics of Global Moran's I and Global Getis-Ord's G, and judged whether or not the threshold was reasonable by using the semi-variable function which was based on the traffic network distance. Then, the influences of the structure characteristics and network dynamic of the small world network on traffic flow were analyzed by network autocorrelation methods. The results show that: (1) The convergence value of global Moran's I and global Getis-Ord's G for the London trunk road is 0.34, and the threshold of the small world network phenomenon or dynamic is 17000 meters; (2) Through the semi-variogram, the estimated threshold distance of small world networks is far less than the spatial autocorrelation effect distance acquired from the theoretical semi-variable function model, and the semi-variable value presents an obvious rising trend, which indicates the threshold is reasonable; (3) Under the effect of the threshold, the traffic flow of the London trunk way shows a significantly positive autocorrelation as a whole, and the local spatial association patterns present the low-low agglomeration in the center and high-high agglomeration in the edge of London.