Assessment and spatial characterization of ground-level ozone exposure using a low-cost sensor network

Area of Science:

• Environmental Science
• Air Quality Monitoring
• Public Health

Background:

• Ground-level ozone (O3) is a significant air pollutant impacting public health, especially in areas with complex emissions and limited monitoring.
• Understanding O3's spatial variability is crucial for targeted mitigation and health protection.

Purpose of the Study:

• To investigate the hyperlocal spatial variability of ambient O3 in the Greater Springfield area, Massachusetts.
• To develop and validate a machine learning model for O3 estimation using low-cost sensors.
• To differentiate contributions of regional and local sources to O3 concentrations.

Main Methods:

• Deployment of a calibrated network of 13 metal oxide semiconductor-based low-cost sensors (LCS).
• Development of a random forest regression model for O3 calibration using a one-year dataset.
• Analysis of sensor-derived O3 data to attribute variance to photochemical processes and local emission sources.

Main Results:

• The machine learning model demonstrated robust O3 predictive performance (R2 = 0.82) using only sensor data.
• Regional photochemical processes accounted for 37.7% of O3 variance, while local sources (industry, traffic) contributed 22.3%.
• Spatial O3 patterns aligned with known emission sources and land use features, indicating model effectiveness.

Conclusions:

• Low-cost sensor networks, when calibrated with machine learning, can effectively capture spatially resolved O3 dynamics.
• This approach enhances air quality data accessibility, particularly in resource-limited communities.
• Findings support targeted interventions by distinguishing between regional and local O3 influences.