Field calibration protocol for optical-based air quality sensors in Indian conditions

Abstract

The calibration process of the sensor/instrument governs sensor accuracy and needs to be dynamic. Optical air quality sensors (PM2.5) were collocated with reference monitoring stations for two weeks. The sensors were installed at different distances from the reference instrument and rotated twice. The performance parameters, such as inter-node variability and correlation between concentration data, were evaluated using statistical descriptors. Finally, the data was calibrated by the reverse regression method. The results are in close agreement with the reference values. Initial uncalibrated measurements ranged from 52.43 to 59.58 μg/m3. After calibration, LR models achieved MAE improvements from 22.63 to 71.36 μg/m3, NMSE from 0.16 to 2.69 μg/m3, and bias from 1.74 % to 30.42 %. MLR models showed MAE enhancements from 31.10 to 67.39 μg/m3, NMSE from 0.25 to 2.58 μg/m3, and bias from 18.22 % to 42.08 %. However, the correlation (R2) with the reference monitor is poor and sensitive to slight deviations indicating independence of sensor performance of correlation values.