- Low-cost air pollution sensors have the potential to revolutionize air pollution measurement by enabling more widespread monitoring. However, several challenges must be addressed including sensor-to-sensor variability, complex interference from other pollutants, and factory calibrations not being applicable to real-world conditions.
- New calibration methods using machine learning algorithms that account for multivariate responses show promise in addressing these challenges. One study used support vector regression and random forest to calibrate NO and NO2 sensors, achieving accurate results with errors below 5 ppb after deployment in urban areas.
- For low-cost sensors to provide reliable data, calibration and evaluation methods must ensure data quality is sufficient for the intended application. Significant progress has been made