End-user perspective of low-cost sensors for outdour air pollution monitoring

Sensors & R² values PM sensors OPC-N2 : Laboratory 0.95-1 Dylos: Laboratory 0.65-1 ; Field: 0.4-1.0 PMS 1003: Laboratory : 0.7-1.0; Field : 0.8-0.9 PMS 3003: Laboratory 0.75-0.95 DSM501A: Laboratory 0.5-0.95 ; Field: 0.1)0.5 DN7C3CA006 : Laboratory : 0.98-1.00 GP2Y1010: Laoratory 0.4-1.0 ; Field: 0.7-1.0 PPD42NS: Laboratory 0.7-0.99; Field: <0.0 - 0.95 O3 Sensors EC Sensors : > 0.9 (laboratory), 0.0-0.9 (field) MOS sensors: Laboratory 0.85-0.95 ; Field: 0.15-0.90 NO2 Sensors EC Sensors: ; 0.95-1.00 (laboratory) ; <0.0 - 0.9 (field) MOS sensor: 0.95-1.0 (laboratory) ; 0.05-0.75 (field)

Low-cost sensor technology can potentially revolutionise the area of air pollution monitoring by providing high-density spatiotemporal pollution data. Such data can be utilised for supplementing traditional pollution monitor-ing, improving exposure estimates, and raising community awareness about air pollution. However, data qualityremains a major concern that hinders the widespread adoption of low-cost sensor technology. Unreliable datamay mislead unsuspecting users and potentially lead to alarming consequences such as reporting acceptableair pollutant levels when they are above the limits deemed safe for human health. This article provides scientificguidance to the end-users for effectively deploying low-cost sensors for monitoring air pollution and people's ex-posure, while ensuring reasonable data quality. We review the performance characteristics of several low-costparticle and gas monitoring sensors and provide recommendations to end-users for making proper sensor selec-tion by summarizing the capabilities and limitations of such sensors. The challenges, best practices, and future outlook for effectively deploying low-cost sensors, and maintaining data quality are also discussed. For data qual-ity assurance, a two-stage sensor calibration process is recommended, which includes laboratory calibrationunder controlled conditions by the manufacturer supplemented with routine calibration checks performed bythe end-user underfinal deployment conditions. For large sensor networks where routine calibration checksare impractical, statistical techniques for data quality assurance should be utilised. Further advancements andadoption of sophisticated mathematical and statistical techniques for sensor calibration, fault detection, anddataquality assurance canindeed help to realise the promised benefits of a low-costair pollution sensor network.