Assessment of air quality sensor system performance after relocation
Stefano Zauli Sajani, Stefano Marchesi, Claudia Pironi, Carla Barbieri, Vanes Poluzzi, Annamaria Colacci
Abstract
Sensor based monitoring systems have been indicated as a promising tool to increase information on spatio-temporal distribution of air pollution but several issues have been raised about the accuracy of such monitors when used in the field. The study aimed at assessing the performance of sensor based systems after multiple relocation in different seasons and sites. The systems included electrochemical sensors to measure NO2 and O3 concentrations. The approach consisted in two-week field calibration of each device at a reference monitoring station and the test of the calibrated device at different reference station sites. The main specific goal was a comparison of sensor performance considering site-specific (SS) and no site-specific (no_SS) calibration, i.e. calibration and testing carried out or not at sites with similar characteristics. Calibration was performed by season using random forest (RF) models. Very good performance was found for calibrated O3 sensors with R2 ≥ 0.82 regardless of seasons and sites. Mean normalized root mean square error (nRMSE) was around 7% and 6% in winter and summer tests, respectively. Very good performance of sensor systems was observed also for NO2 during winter (R2 ≥ 0.84) with much better accuracy for SS compared to no_SS calibration (nRMSE equal to 6% and 17%, respectively). A marked decrease of performance was observed for NO2 sensors during summer. Our results show a good potential of sensor based systems after SS field calibration in increasing information on the distribution of air pollution at high spatial and temporal resolution.