Design of an ozone and nitrogen dioxide sensor unit and its long-term operation within a sensor network in the city of Zurich
Abstract. This study focuses on the investigation and quantification of low-cost sensor performance in application fields such as the extension of traditional air quality monitoring networks or the replacement of diffusion tubes. For this, sensor units consisting of two boxes featuring NO2 and O3 low-cost sensors and wireless data transfer were engineered. The sensor units were initially operated at air quality monitoring sites for 3 months for performance analysis and initial calibration. Afterwards, they were relocated and operated within a sensor network consisting of six locations for more than 1 year. Our analyses show that the employed O3 and NO2 sensors can be accurate to 2–5 and 5–7 ppb, respectively, during the first 3 months of operation. This accuracy, however, could not be maintained during their operation within the sensor network related to changes in sensor behaviour. For most of the O3 sensors a decrease in sensitivity was encountered over time, clearly impacting the data quality. The NO2 low-cost sensors in our configuration exhibited better performance but did not reach the accuracy level of NO2 diffusion tubes (∼ 2 ppb for uncorrected 14-day average concentrations). Tests in the laboratory revealed that changes in relative humidity can impact the signal of the employed NO2 sensors similarly to changes in ambient NO2 concentration. All the employed low-cost sensors need to be individually calibrated. Best performance of NO2 sensors is achieved when the calibration models also include time-dependent parameters accounting for changes in sensor response over time. Accordingly, an effective procedure for continuous data control and correction is essential for obtaining meaningful data. It is demonstrated that linking the measurements from low-cost sensors to the high-quality measurements from routine air quality monitoring stations is an effective procedure for both tasks provided that time periods can be identified when pollutant concentrations can be accurately predicted at sensor locations.