起讫点建成环境对轨道交通高峰时段网络客流的作用机制分析

Mechanism of impact of the built environment of urban rail transit originand destination stations on network ridership during peak hours

建成环境与轨道交通客流之间联系密切。然而既有研究主要致力于探究建成环境对轨道交通站点客流的影响,对网络客流的影响分析较少。与站点客流不同,网络客流的矢量性受起讫点建成环境的双重作用。论文基于多源地理空间数据构建“5D+N”维度的建成环境指标体系,采用可解释轻量级梯度提升机机器学习模型(LightGBM-SHAP)从特征贡献度、单变量非线性关系、双变量交互效应三个方面解析起讫点建成环境对高峰时段网络客流的影响。针对天津的案例研究表明:①对早高峰网络客流重要度排序前5位的建成环境因素综合贡献达到13.22,其中讫点距市中心距离贡献度最大,从起讫点双重视角来看,讫点建成环境贡献度(20.86)大于起点建成环境贡献度(18.30);②建成环境因素与早高峰网络客流的非线性关系差异显著,存在非线性正相关、非线性负相关等多种关系,阈值效应也存在单端阈值、多段阈值等多种情况;③对于早高峰网络客流,起点或讫点公交站点密度分别与讫点中介中心性有显著交互作用,公交站点密度在两者交互效应中处于主导地位。研究结果对引导站域建成环境更新及提升轨道交通客流效能提供了精准有效的策略指引。

There exists a significant link between the built environment and rail transit ridership.However,the existing studies are mainly devoted to investigating the impact of the built environment on ridership at rail transit stations,and there are fewer analyses on the impact of network ridership.Unlike station ridership,the vectoral nature of network ridership is subject to the dual effects of the built environment at the origin and destination.In this study,we constructed a built environment indicator system with"5D+N"dimensions based on multi-source geospatial data,and used the LightGBM-SHAP model to analyze the impact of the built environment of the origin and destination stations on network ridership during peak hours from three aspects:the contribution of different characteristics,univariate nonlinear relationship,and bivariate interaction effect.The case study of Tianjin showed that:1)The combined contribution of the top five built environment factors to the importance of network ridership in morning peak reached 13.22,among which the distance of the destination from the city center contributed the most,and the contribution of the built environment of the destination(20.86)was greater than the contribution of the built environment of the starting point(18.30).2)The nonlinear relationship between built environment factors and morning peak network ridership varied significantly,with nonlinear positive correlation,nonlinear negative correlation,and so on,and the threshold effect also existed in a variety of cases,such as single-end threshold and multi-segment threshold.3)For morning peak network ridership,origin or destination bus stop density significantly interacts with destination betweenness centrality,respectively,with bus stop density dominating the interaction effect.The results of the study provide a precise and effective strategy to guide the renewal of the built environment of the station area and to enhance the effectiveness of rail transit ridership.