Unraveling the impact of COVID-19 on urban mobility: A Causal Machine Learning Analysis of Beijing's Subway System

Abstract

The COVID-19 pandemic has drastically altered urban travel patterns, particularly in public transportation systems like subways. This study examines the effects of the pandemic on subway ridership in Beijing by analyzing the influence of 19 factors, including demographics, land use, network metrics, and weather conditions, before and during the pandemic. Data was collected from June 2019 and June 2020, covering 335 subway stations and over 258 million trips. Using a three-stage analytical framework - comprising Light Gradient Boosting Machine (LightGBM) for fitting, Meta-Learners for causal analysis, and SHapley Additive exPlanations (SHAP) for interpretation - we observed a substantial decline in ridership, with approximately 10,000 fewer passengers per station daily, especially in densely populated areas. Our findings reveal significant shifts in influential factors such as centrality, housing prices, and restaurant density. The spatiotemporal analysis highlights the dynamic nature of these changes. This study underscores the need for adaptive urban planning and provides insights for public health strategies to enhance urban resilience in future pandemics.