Urban Air Pollution Mapping Using Fleet Vehicles as Mobile Monitors and Machine Learning

Area of Science:

• Environmental Science
• Data Science
• Urban Planning

Background:

• Spatially explicit urban air quality data is crucial for effective pollution control.
• Traditional stationary monitors offer limited spatial coverage and availability.
• Mobile monitoring presents an alternative but lacks fleet-level validation.

Purpose of the Study:

• To demonstrate the feasibility of using a ride-hailing fleet for large-scale mobile air quality monitoring.
• To develop a model for high-resolution mapping of fine particulate matter (PM2.5) concentrations.
• To assess the performance improvement over traditional monitoring methods.

Main Methods:

• Equipped 260 electric vehicles in Beijing with low-cost sensors for real-time PM2.5 data collection.
• Developed a decision tree model to infer PM2.5 distribution at 1 km by 1 km and 1 h resolution.
• Utilized machine learning techniques to augment sensor data for enhanced pollution mapping.

Main Results:

• Collected the largest dataset of mobile sensor data for urban air quality monitoring to date.
• The developed model achieved a coefficient of determination of 0.80, a significant improvement over the benchmark's 0.56.
• Root mean square error decreased from 12.6 to 8.1 μg/m³ compared to the benchmark model.
• Successfully identified short-term and long-term variations and local air pollution hotspots.

Conclusions:

• Fleet vehicles can serve as routine mobile sensors for comprehensive urban air quality monitoring.
• Advanced data science methods combined with mobile sensing offer a powerful solution for high-resolution air pollution mapping.
• This approach is essential for understanding and managing urban air quality effectively.