Characterization of low-cost sensors via simultaneous field measurements: a case study

The use of low-cost sensors is continuously increasing as more evidence appears of their benefits: from the wide range of applications to the ease of use and accessible costs. Still, they are extensively studied to rightfully understand their range of use and the right way to handle their data. The first part of this work focuses on one month of simultaneous measurements of temperature, relative humidity, pressure, CO, NO, NO₂, O_3, O_X, VOCs, PM_2.5 and PM₁₀ by three different low-cost sensors and a Pollution PyxisGC BTEX. Dataset acquired by ARTA (Agenzia Regionale per la Tutela dell’Ambiente) and ITAF (Italian Air Force) have been used as reference for the comparison. Vaisala AQT530, while also being part of the characterized instruments, has been used as reference for those quantities that were not present in the mentioned dataset and a 2B Ozone Monitor has also been used for O₃ characterization.

Aside from a discrepancy in temperature and RH for AirSensEUR, the meteorological quantities for all the sensors show high correlation values (R ∼ 0.9). NO, NO₂ and VOCs have high correlation with the same compounds observed by ARTA instruments (R ≥ 0.8), except for Libelium Smart Environment PRO NO₂ that has a lower value (R = 0.348). Pollution PyxisGC BTEX VOCs comparison shows low slope values (∼ 0.16 against 1 hoped). CO measurements have a high similarity between the three low-cost sensors, differently O_X values show a generally lower similarity with the reference instruments (with R ∼ 0.8 for Vaisala while being R ∼ 0.5 for the others). Libelium Smart Environment PRO measurement of NO, NO₂ and O₃ are affected by an extremely high bias (with values ≥ 100), a peculiar result considering how the sensors mounted where all factory-new and already calibrated by the manufacturer. PM values have only been compared averaging over the entire day (due to the kind of reference data available), showing a general matching with ARTA measurements only for AirSensEUR, which, at the same time, has the highest standard deviation.

Moreover, the dataset has also been used as a case study to investigate the ability of these instruments to catch signals from different near sources: 1) the local regional airport, 2) heavily used highway and roads. Analysis during rush hours, weekdays vs weekends, during the Christmas Holiday and with the help of wind data have been conducted. The results prove the ability of these low-cost sensors to detect rush hours measurements as well as the contribution to CO, NO, NO₂, VOCs and PM emission due to the presence of the near highway.