Field Calibration of a Low-Cost Air Quality Monitoring Device in an Urban Background Site Using Machine Learning Models

Field calibration of low-cost air quality (AQ) monitoring sensors is essential for their successful operation. Low-cost often exhibit non-linear responses to pollutants and signals may be affected by the presence multiple compounds making challenging. We investigate different approaches field an AQ device named ENSENSIA, developed in Institute Chemical Engineering Sciences Greece. The present study focuses on measurements two most important measured ENSENSIA: NO2 O3. measurement site located center Patras, third biggest city Reference instrumentation used regulatory purposes Region Western Greece was as evaluation standard. were installed years at same locations. Measurements from first year (2021) seven ENSENSIA (NO2, NO, O3, CO, PM2.5, temperature relative humidity) train several Machine Learning (ML) Deep (DL) algorithms. resulting algorithms assessed using data second (2022). Random Forest algorithm exhibited best performance correcting O3 NO2. For mean error reduced 9.4 ppb 3 ppb, whilst R2 improved 0.22 0.86. Similar results obtained wherein 13 4.3 increased 0.52 0.69. Long-Short Term Memory Network (LSTM) also showed good pollutants.