(The calibration of two low-cost particulate matter sensors (PurpleAir and MicroPEM) in Glasgow (UK) and the evaluation of accounting for the influence of relative humidity (in PurpleAir) and baseline drift (in MicroPEMs)

Abstract

Low-cost particulate matter sensors are promising technologies due to their small size, rapid-response capabilities, and affordability. The distribution of such sensors can enable the creation of more extensive and intensive PM2.5 monitoring network improving the spatiotemporal resolution. Low-cost PM sensors have been reported by many literatures of having lower and inconsistent accuracy as compared to highly reliable PM2.5 monitors. The reason for this inconsistency is thought to be due to their sensitivity to the changes in atmospheric conditions. For more optimized performance, these sensors need to be calibrated under realistic conditions. Therefore, PurpleAir and MicroPEM which were selected and collocated in duplicate with the reference analyser located in Townhead District in Glasgow city. The measurements from PurpleAir and MicroPEM were hourly averaged and evaluated against that of the reference analyser. PurpleAir was empirically corrected for relative humidity (RH). The RH correction increased the accuracy of PurpleAir duplicate with the reference analyser as the normalised mean bias error (NMBE) reduced from (+19%, +20%) to (-2%, -1%) and the normalised mean absolute error (NMAE) reduced from (73%,71%) to (58%, 57%). However, the effects of RH on the PM2.5 estimated by PurpleAir was not found to be predicable. On the other hand, MicroPEMs were corrected for baseline drift. MicroPEMs were exposed to extremely varied PM2.5 concentrations range 147-0.4 µg m-3 due to a firework event happened during the deployment. This resulted in very high linear agreement and accuracy with the reference analyser as shown by the coefficient of determination (R2), NMBE and NMAE obtained from the duplicate (R2= 0.97, 0.99) and NMBE (23%, 10%) NMAE (34%, 19%) demonstrating that field calibrating at high PM2.5 levels can produce more reliable calibration. The study was limited to one geographical location.