On thresholds for controlling negative particle (PM2.5) readings in air quality reporting

Abstract

Ambient PM_2.5 (particles less than 2.5 μm in diameter) is monitored in many countries including Australia. Occasionally PM_2.5 instruments may report negative measurements, although in realty the ambient air can never contain negative amounts of particles. Some negative readings are caused by instrument faults or procedural errors, thus can be simply invalidated from air quality reporting. There are occasions, however, when negative readings occur due to other factors including technological or procedural limitations. Treatment of such negative data requires consideration of factors such as measurement uncertainty, instrument noise and risk for significant bias in air quality reporting. There is very limited documentation on handling negative PM_2.5 data in the literature. This paper demonstrates how a threshold is determined for controlling negative hourly PM_2.5 readings in the New South Wales (NSW) air quality data system. The investigation involved a review of thresholds used in different data systems and an assessment of instrument measurement uncertainties, zero air test data and impacts on key reporting statistics when applying different thresholds to historical datasets. The results show that a threshold of -10.0 μg/m³ appears optimal for controlling negative PM_2.5 data in public reporting. This choice is consistent with the measurement uncertainty estimates and the zero air test data statistics calculated for the NSW Air Quality Monitoring Network, and is expected not to have significant impacts on key compliance reporting statistics such as data availability and annual average pollution levels. The analysis can be useful for air quality monitoring in other Australian jurisdictions or wider context.

Keywords: Beta Attenuation Monitor (BAM); Compliance reporting; Data validation; Fine particle (PM2.5) monitoring; Negative reading.

Fig. 1

Locations of monitoring stations in the New South Wales Air Quality Monitoring Network, as of mid-2020. Standard-compliant method instruments for PM_2.5 are applied at the standard, industry-funded and roadside monitoring stations

Fig. 2

Schematic of the analytic framework for the study

Fig. 3

Expanded uncertainty (a) and 95% confidence interval (b) for hourly PM_2.5 measurements

Fig. 4

Histogram of hourly zero (clean) air stability test data, pooled for all sites and all years

Fig. 5

Box plot of annual data loss rates (%) for individual stations and years by threshold option. Bold line inside box: median; box bottom and top sides: 25th and 75th percentiles; circle: outliers; lines outside the box: 9th and 91st percentiles

Fig. 6

The box plot of annual averages for individual stations and years by threshold option. Bold line inside box: median; box bottom and top sides: 25th and 75th percentiles; circle: outliers; lines outside the box: 9th and 91st percentiles