Previous studies explore the impact of rail transit on land development and identify the size of its impact area. However, many pre-specify their relationship as linear or quadratic. This restrictive assumption biases the estimate for rail transit impact. To relax this assumption, this study employs gradient boosting decision trees to the 2008 and 2014 building census data in Shenzhen, China, and examines the nonlinear relationship between distance to metro stations and land development intensity, i.e., changes in floor area. We find that metro transit is more important in predicting changes in floor areas than other transportation infrastructures. Distance to metro station shows a threshold effect on development intensity, with the effect extending to 1.25 km from the station. Metro stations and land availability have a synergistic effect, strengthening the development of adjacent lands. Overall, metro transit and supportive policies are important instruments to shape urban form in urbanizing China.
Bibliographical noteFunding Information:
This research was funded by the National Natural Science Foundation of China ( 51678004 ).
© 2020 Elsevier Ltd
- Decision trees
- Nonlinear effect
- Rail transit
- Transportation investment
- Urban form