Long-term field comparison of multiple low-cost particulate matter sensors in an outdoor urban environment

Exposure to ambient particulate matter (PM) air pollution is a leading risk factor for morbidity and mortality, associated with up to 8.9 million deaths/year worldwide.  Measurement of personal exposure to PM is hindered by poor spatial resolution of monitoring networks. Low-cost PM sensors may improve monitoring resolution in a cost-effective manner but there are doubts regarding data reliability. PM sensor boxes were constructed using four low-cost PM micro-sensor models. Three boxes were deployed at each of two schools in Southampton, UK, for ~1 year and sensor performance was analysed. Comparison of sensor readings with a nearby background station showed moderate to good correlation (0.61<r<0.88, p<0.0001), but indicated that low-cost sensor performance varies with different PM sources and background concentrations, and to a lesser extent relative humidity and temperature.  This may have implications for their potential use in different locations. Data also indicates that these sensors can track short-lived events of pollution, especially in conjunction with wind data. We conclude that, with appropriate consideration of potential confounding factors, low-cost PM sensors may be suitable for PM monitoring where reference-standard equipment is not available or feasible, and that they may be useful in studying spatially localised airborne PM concentrations.

This dataset contains:

1. sensor_data.Rds (R format) containing the data from the sensors averaged (median) per minute over the period of the study (13/03/18 until 28/02/19)
2. winddata.csv containing the data from the meteorological station over the period of the study (13/03/18 until 28/02/19) taken from http://www.southamptonweather.co.uk/
3. CV_for_ICC.csv containing the coefficients of variation calculated and used for the Intra Class Correlation (ICC) analysis
4. Figure5.zip (zip file) containing the CSV files underlying the data presented in Figure 5 of the paper and underlying Supplementary Figure S15:
   • Figure 5:
     • month.csv: correlation per month with the background reference station per sensor per AQM
     • NoAugust.BG.csv: correlation per quartiles of background PM25 concentrations per sensor per AQM excluding August 2018
     • NoAugust.RH.csv: correlation per quartiles of relative humidity per sensor per AQM excluding August 2018
     • NoAugust.Temperature.csv: correlation per quartiles of temperature per sensor per AQM excluding August 2018
     • NoAugust.wd.csv: correlation per wind direction per sensor per AQM excluding August 2018
   • Supplementary Figure S15:
     • BG.csv: correlation per quartiles of background PM25 concentrations per sensor per AQM
     • RH.csv: correlation per quartiles of relative humidity per sensor per AQM
     • Temperature.csv: correlation per quartiles of temperature per sensor per AQM
     • wd.csv: correlation per wind direction per sensor per AQM