To inform the station-area planning, previous studies use direct ridership models to examine the relationship between the built environment around stations and transit ridership. Based on this framework, this study innovatively applies gradient boosting decision trees to investigate the non-linear effects of built environment variables on station boarding. Using the Metrorail data in the Washington metropolitan area, we found that station-area built environment characteristics collectively contribute to 34% of the predictive power for Metrorail ridership, after controlling for transit service factors and demographics. Built environment variables show threshold effects on Metrorail ridership. We further identified their effective ranges, guiding land use planning around stations. This study highlights the roles of compact and mixed land use development, the number of bus stops, and car ownership in determining the station-level ridership.
Bibliographical noteFunding Information:
This work is supported by the National Natural Science Foundation of China ( 71874010 ), and Young Elite Scientist Sponsorship Program by the China Association for Science and Technology ( 2017QNRC001 ).
- Direct ridership model
- Land use
- Transit ridership
- Transit-oriented development
- Travel demand