Long-term field calibration of low-cost metal oxide VOC sensor: Meteorological and interference gas effects

This study aimed to evaluate the performance of the Sensirion SGP30 MOS (metal oxide semiconductor) VOC sensors (SMVSs) for a long sampling period (10–12 months) at different ambient air monitoring stations (AQMSs). In the field tests, temperature (T), relative humidity (RH), carbon monoxide (CO) concentration, and non-methane hydrocarbons (NMHC) concentrations were found to be the crucial parameters influencing the sensor readings. A multiple linear regression (MLR) calibration model using hourly T, RH, CO, and VOC values as parameters was developed at Zhongming (ZM) station and validated at Hsinchu (HC) and Tainan (TN) stations. After calibration, the SMVS performance was improved significantly with the decrease of mean normalized bias (MNBs) and mean normalized errors (MNEs) of hourly-average data from +228 to 473% to ±25.9% and 231–473%–47.2%, respectively based on the reference NMHC values in the AQMSs. For ambient NMHC concentrations higher than 100 ppbv, the MLR-calibrated results were further improved with MNBs (MNEs) less than ±20% (28.3%). The correlation between calibrated sensor data and reference data was enhanced significantly with the R² values increasing from 0.27 to 0.50 to 0.65–0.78 at three stations. Therefore, these well-calibrated sensors can become viable tools for many purposes including hotspot identification and characterization, and personal exposure study (MNE<30%).