In-situ baseline calibration approach for enhanced data quality of large-scale air sensor monitoring networks

As dense sensor networks for air quality monitoring become increasingly prevalent, effective calibration remains a critical yet challenging component of their operation, particularly for large-scale networks. Conventional calibration methods, which rely heavily on co-locating sensors with reference monitors for learning and training, often face significant scalability challenges, rendering them impractical for post-deployment recalibration. To address this limitation, we propose an in-situ baseline calibration method (b-SBS) that calibrates sensors remotely without the need for direct co-location. This approach is grounded in the physical characteristics of electrochemical sensors and is informed by statistical analyses of calibration coefficients across a large group of similar sensors. Through preliminary field tests conducted on a batch of sensors for NO2, NO, CO, and O3, two key linear calibration coefficients, sensitivity and baseline, were systematically investigated. Sensitivity analysis of over 100 short-term calibration samples for each gas revealed coefficients clustered within 20% variation, enabling universal parameterization. Long-term baseline drift remained stable within ±5 ppb for NO2, NO, and O3, and ±100 ppb for CO over 6 months, supporting semi-annual recalibration. Applying the b-SBS calibration approach to 73 NO2 sensors in a large-scale Shanghai network yielded pronounced data quality improvements compared to their original measurements (initially calibrated before deployment): the median R2 increased by 45.8% (from 0.48 to 0.70), and RMSE decreased by 52.6% (from 16.02 to 7.59 ppb), as validated against nearby reference stations. The Shanghai application, while showing the method’s potential for large-scale deployments, awaits further real-time validation to confirm its robustness under diverse operational conditions. This study is a valuable advancement in calibration strategies, offering a cost-effective solution that reduces operational costs while ensuring accurate measurements across numerous sensors and long-term network deployments.