Spatially adaptive calibrations of airbox PM_2.5 data.

Area of Science:

• Environmental Science
• Data Science
• Sensor Networks

Background:

• Taiwan utilizes the Taiwan Air Quality Monitoring Network (TAQMN) and the AirBox network for PM2.5 monitoring.
• The TAQMN offers high-quality data, while the AirBox network provides extensive spatial coverage using low-cost IoT sensors.
• AirBox sensor data is unreliable and requires location-specific calibration due to spatial variations in PM2.5 chemical composition.

Purpose of the Study:

• To develop a method for accurate PM2.5 estimation using spatially adaptive calibrations for AirBox sensors.
• To address the challenge of misaligned AirBox and EPA monitoring locations.
• To create a robust calibration model that adapts to local environmental factors.

Main Methods:

• Proposed a spatial model with spatially varying coefficients to handle data heterogeneity.
• Implemented spatially adaptive calibrations for AirBox sensors.
• Incorporated both TAQMN and AirBox data for improved PM2.5 concentration estimates.

Main Results:

• Achieved accurate PM2.5 concentration estimates with error bars at any location.
• Demonstrated significant improvement in PM2.5 prediction, reducing root-mean-squared prediction error by 38%-68%.
• The model is robust to outliers and can provide calibration formulas for new sensors.

Conclusions:

• The proposed spatial model effectively calibrates AirBox sensors, enhancing PM2.5 monitoring accuracy in Taiwan.
• Reliable PM2.5 values can be obtained using calibrated AirBox data, independent of EPA data.
• This approach offers a scalable solution for improving air quality monitoring networks globally.