Smart card data derived from automatic fare collection (AFC) systems of public transit enable us to study resident movement from a macro perspective. The rhythms of traffic generated by different land uses differ, reflecting differences in human activity patterns. Thus, an understanding of daily ridership and mobility patterns requires an understanding of the relationship between daily ridership patterns and characteristics of stations and their direct environment. Unfortunately, few studies have investigated this relationship. This study aims to propose a framework of identifying urban mobility patterns and urban dynamics from a spatiotemporal perspective and pointing out the linkages between mobility and land cover/land use (LCLU). Relying on 1 month’s transactions data from the AFC system of Nanjing metro, the 110 metro stations are classified into 7 clusters named as employment-oriented stations, residential-oriented stations, spatial mismatched stations, etc., each characterized by a distinct ridership pattern (combining boarding and alighting). A comparison of the peak hourly ridership of the seven clusters is conducted to verify whether the clustering results are reasonable or not. Finally, a multinomial logit model is used to estimate the relationship between characteristics of the local environment and cluster membership. Results show that the classification based on ridership patterns leads to meaningful interpretable clusters and that significant associations exist between local LCLU characteristics, distance to the city center and cluster membership. The analytical framework and findings may be beneficial for improving service efficiency of public transportation and urban planning.

中文翻译：

---

从时空角度理解城市交通模式：地铁站的每日乘客量分布

来自公共交通自动收费（AFC）系统的智能卡数据使我们能够从宏观角度研究居民运动。不同土地利用产生的交通节奏不同，反映了人类活动模式的差异。因此，了解每日乘客量和移动模式需要了解每日乘客模式与车站特征及其直接环境之间的关系。不幸的是，很少有研究调查这种关系。本研究旨在提出一个框架，从时空角度识别城市流动模式和城市动态，并指出流动性与土地覆盖/土地利用 (LCLU) 之间的联系。依托南京地铁AFC系统1个月交易数据，110个地铁站分为就业型站、居住型站、空间错配站等7个群，每个群都有不同的乘车模式（上下车相结合）。对七个集群的每小时高峰客流量进行比较，以验证集群结果是否合理。最后，使用多项 logit 模型来估计本地环境特征与集群成员之间的关系。结果表明，基于乘客模式的分类导致有意义的可解释集群，并且本地 LCLU 特征、到市中心的距离和集群成员之间存在显着关联。