Air Pollution Monitoring Using Cost-Effective Devices Enhanced by Machine Learning.

Area of Science:

• Environmental Science
• Data Science
• Sensor Technology

Background:

• Air pollution poses significant global health risks, necessitating comprehensive air quality monitoring.
• Current monitoring networks are often inadequate due to the high cost of reference-grade stations.
• Low-cost sensors offer a potential solution for widespread air quality assessment.

Purpose of the Study:

• To investigate the effectiveness of machine learning in enhancing the accuracy of low-cost air quality sensors.
• To address challenges related to sensor cross-sensitivity, environmental factors, and manufacturing variations.
• To determine if individual sensor calibration models outperform generalized correction approaches.

Main Methods:

• Deployed cost-effective air pollution monitoring devices alongside a reference-grade air quality station.
• Collected sensor response data under various conditions.
• Trained and evaluated machine learning models, including individually tailored algorithms, to correct sensor readings.

Main Results:

• Machine learning models significantly improved the accuracy of low-cost sensor measurements.
• Individually tailored models enhanced the correlation with reference instruments by up to 10%.
• Performance inconsistencies between similar sensor units were observed, hindering the development of a single correction model for a sensor type.

Conclusions:

• Machine learning, particularly with individualized sensor models, is a viable strategy to improve low-cost air quality monitoring accuracy.
• Addressing sensor-specific variations is critical for reliable and widespread deployment of affordable air quality monitoring solutions.
• Further research is needed to overcome inter-sensor variability for robust, unified correction models in air quality networks.