Spatial Correlation-based Passenger Flow Characteristics And Mechanism Analysis For Subway Network

As the network scale of urban rail transit increase gradually, the space-time distributions of passenger flows on the rail network become more complex. Complex ridership interaction exists on the network. Therefore, the change of land use density around certain station not only affects the passenger flows of the station, but also the passenger flows of other stations in the network. It is necessary to analyze the spatial characteristics of passenger flows and its mechanism with land use to provide a basis for rail planning and operations and further improve passenger service level.This paper proposes quantitative indexes to analyze the spatial correlation of passenger flows, and establishes a spatial autoregressive model to reflect the spatial effect of land use on station passenger flows. To evaluate the performance of the proposed indexes and the model, a case study on the Beijing subway network is undertaken based on with the passenger flow data of April 2013. The main contents and conclusions are as follows:(1) From generative and attractive perspectives, the global and local indexes for the spatial correlation of subway passenger flows are proposed according to the general form of spatial autocorrelation index Moron’s I. The expected value and variance for the proposed spatial correlation index is calculated. And then, the Z-statistics is puts forward to test the significance on spatial correlation of subway passenger flows.(2) Based on OD matrix of Beijing Subway passenger flow during the morning peak hour of April 2013, the proposed indexes are applied to test the effectiveness in practical application. The case study shows that the spatial correlation of subway passenger flows is significant within the trip distance between 2km to 40km. The spatial correlation of subway passenger flows is ruled by exponential function. The average travel distance of passengers is about 15.5km. The average travel distance of departure passengers in downtown stations is almost less than 10km, and that of departure passengers in suburban stations exceeds lOkm.On the contrary, the average travel distance of arrival passengers in downtown stations and development zone almost exceeds 15km, and that of arrival passengers in suburban stations is less than 10km.(3) According to the local indexes for spatial correlation of subway passenger flows, the trip distance of passengers of subway stations can be determined and the subway stations are classified into global point, generative point, attractive point and local point. Based on the classification, the corresponding organization schemes for passenger flows and train operational plan can be put forward to improve the service level.(4) According to the spatial correlation of subway passenger flows, an improved spatial autoregressive model (SAR) is developed based on OD matrix instead of spatial adjacency matrix. The maximum likelihood is used to estimate the parameters of the improved SAR model. Then, the effect of land use on subway passenger flows can be measured. By comparing with the traditional SAR model, it can be found that the improved SAR model is more accurate to measure the spatial effect of land use on subway passenger flows.(5) The improved SAR model is applied on Beijing subway network to test its performance. The change of land use density around certain station lead to the change of passenger inflow and outflow of other stations on the network. The effect degree of land use on passenger flow varies with different stations. The passenger flows will change significantly when the land use around the stations in downtown and residential area, large-scale transportation hub and transfer stations. Compared with employed population, the increase of resident population has a more significant influence on passenger flow.(6) The case study shows that the improved SAR model can analyze the impact on network flows which is caused by the changes of land use around certain station and forecast the passenger inflow and outflow. The operator can improve train operation schemes and station facilities in order to satisfy the demands of passenger. On the other hand, the improved SAR model can also be applied on the situation that the new line operates. The effect of the new line on the network can be evaluated.