Quantifying the contribution of environmental variables to cyclists' exposure to PM2.5 using machine learning
Martín Rodríguez Núñez1,2, Iván Tavera Busso1,2, Hebe Alejandra Carreras1,2
1Instituto Multidisciplinario de Biología Vegetal (IMBIV), Consejo Nacional de Investigaciones Científicas y Técnicas, Argentina.
Heliyon
|February 1, 2024
Summary
Cyclists face significant air pollution exposure. This study developed a machine learning model identifying temporal and environmental factors influencing fine particulate matter (PM2.5) exposure for urban cyclists.
Area of Science:
- Environmental Science
- Urban Planning
- Data Science
Background:
- Cyclists are highly susceptible to urban air pollution.
- Understanding exposure factors is key for healthier city environments.
- Machine learning offers predictive power but can lack interpretability.
Purpose of the Study:
- To develop a predictive model for cyclist exposure to fine particulate matter (PM2.5).
- To identify key factors influencing PM2.5 exposure for cyclists in urban settings.
Main Methods:
- Utilized geo-temporally referenced data and machine learning techniques.
- Developed and evaluated predictive models, selecting gradient boosting for best fit.
- Analyzed variable importance including temporal, meteorological, and spatial attributes.
Main Results:
- The gradient boosting model achieved a minimum root mean square error of 5.62 μg m⁻³ for PM2.5 prediction.
- Temporal factors (month, day, time) were most influential, followed by meteorological variables.
- Street typology, vegetation, and vehicle flow also impacted exposure; dedicated bike paths showed lower PM2.5 than roadside lanes.
Conclusions:
- Temporal and environmental factors significantly influence cyclist PM2.5 exposure.
- Urban planning should integrate bicycle pathways with transportation routes to mitigate air pollution.
- Strategic design of urban infrastructure can enhance air quality for cyclists.
