The impacts of weather conditions on metro ridership: An empirical study from three mega cities in China

Understanding the impacts of weather conditions on metro ridership can provide insights to improve the resil-ience of metro systems. However, there are few studies to examine the weather-metro ridership relationship in multiple cities with a framework to obtain generalizable and robust results. This study aims to fill this gap by establishing reliable models to investigate the impacts of weather conditions on metro ridership in three mega cities with different weather conditions in China, namely Beijing, Shanghai and Shenzhen. We collect daily metro ridership and weather parameters of three cities for a 12 month period in 2019, and the multiple linear regression models with 9-term moving average are utilized to investigate the weather-metro ridership rela-tionship in each city. The metro ridership residual is selected as the dependent variable, and independent var-iables include weather variables, seasonal variables, comprehensive comfort index and air quality index. In addition, socioeconomic characteristics are also considered in the generalized linear mixed-effect models with multiple cities. Results highlight that independent variables have significant impacts on metro ridership in all models. Although significances of variables are various in different models, most variables have the same in-fluence direction in different models on both weekdays and weekends. As a whole, temperature residual, spring, summer, and fall have a positive impact on metro ridership, while relative humidity residual, wind speed re-sidual, and precipitation are negatively associated with metro ridership. Besides, the ridership on weekends is more severely affected by weather conditions compared to that on weekdays. In the overall models, socioeco-nomic characteristics are significant on weekdays, but they are insignificant on weekends. It should be noted that the comprehensive comfort index and air quality index are significantly associated with metro ridership in specific models, and they should be considered in further weather-ridership modeling. The findings can deepen our understandings about weather-metro ridership relationship and the differences among cities, which would provide valuable information for scheduling and management of metro systems.