Field Calibration and Evaluation of an Internet-of-Things-Based Particulate Matter Sensor

This paper presents a field evaluation of IoT-enabled Plantower PMS5003 particulate matter sensors in Birmingham, United Kingdom. Commercial, off the shelf, sensors were adapted to utilise Low Power Wide Area Network (LPWAN) IoT technology enabling batteries to be used as a power source. The devices are capable of measuring and communicating data to an online platform with a battery life of ∼2 months, at a measurement interval of 15 min, allowing for automated air quality monitoring for extended periods at high density. The sensors demonstrate success at being integrated into a wireless sensor network, with a high presence of readings. The average correlation coefficients (r2) between raw PMS device data and reference instrumentation are 0.718, 0.703, and 0.543 for PM1, PM2.5, and PM10, respectively. The devices also demonstrate good intersensor consistency, with Pearson’s r values between pairs ranging from 0.92 to 0.99 across all size ranges. Relative humidity (RH) clearly influences the response of the sensors, especially for RH >85%, in keeping with previous laboratory evaluations and evaluations of similar devices. The development of a multi-linear correction factor that accounts for humidity effects on the performance of the sensors is described; using this model, Pearson’s r values range from 0.81 to 0.91 compared to 0.73–0.85 from uncorrected values. There is also some evidence of drift at high humidity over an 8-week period, suggesting that such sensors will (at least currently) need recalibration approximately bimonthly. The limit of detection (LoD) (1.60–4.75 μg m−3) calculated from this study also demonstrates that the sensors are suitable for capturing concentrations typical of a moderately polluted United Kingdom urban environment—LoDs of PM2.5 in this study would have allowed for capture of 94.7% of the concentrations recorded at a typical United Kingdom urban roadside monitoring site between 2017 and 2020.