Cost-effective indoor air quality monitoring in schools: in-field calibration of PM low-cost sensor

The availability of low-cost sensors (LCS) devices for indoor air monitoring has boosted air pollution field. This study aims to calibrate particulate matter (PM) LCS (PM 1 , PM 2.5 and PM 10 ) in four age groups/types of rooms (infants, preschoolers, schoolers and lunchroom) during occupancy and non-occupancy period in nursery and primary schools in Porto, considering the purpose of being used as a tool to empower schools to apply indoor air pollution (IAP) mitigation measures. Sixteen LCS devices (AirVisual Pro) and three research-grade instruments (DustTrak DRX 8534/8533) were used to monitor PM in around 130 different samplings. Before calibration, a methodology approach based on local maxima was applied to all PM fractions of LCS data, since various error events were found. Thus, after identifying and removing these events, a merged dataset was created using 1-min mean of LCS and reference data. Calibration models were applied, such as simple and multiple linear regressions (LR and MLR), linear and non-linear support vector regression (SVR) and gradient boosting regression (GBR). A strong linear relationship was observed between LCS device and reference data, especially for non-occupancy period and in finer PM fractions (Pearson’s correlation reached 0.94 for PM 1 and PM 2.5 ). While PM 10 exhibited a slightly weaker correlation than the other PM fractions. The calibration models, particularly SVR and GBR models, significantly improved the results depending on the PM fraction, age group/type of room and occupancy pattern. Overall, results indicated that LCS devices are an effective tool for managing IAQ in schools, based on PM sensor. • Strong linear relationship between low-cost and reference data before calibration • Finer particles showed better correlation with reference data than PM 10 • Pearson’s correlation was higher for non-occupation period in all age groups • Non-linear calibration models (SVR and GBR) significantly improved the results