Improving the UNC Passive Aerosol Sampler Model Based on Comparison with Commonly Used Aerosol Sampling Methods

Abstract

Objectives: In an occupational environment, passive sampling could be an alternative to active sampling with pumps for sampling of dust. One passive sampler is the University of North Carolina passive aerosol sampler (UNC sampler). It is often analysed by microscopic imaging. Promising results have been shown for particles above 2.5 µm, but indicate large underestimations for PM2.5. The aim of this study was to evaluate, and possibly improve, the UNC sampler for stationary sampling in a working environment.

Methods: Sampling was carried out at 8-h intervals during 24 h in four locations in an open pit mine with UNC samplers, respirable cyclones, PM10 and PM2.5 impactors, and an aerodynamic particle sizer (APS). The wind was minimal. For quantification, two modifications of the UNC sampler analysis model, UNC sampler with hybrid model and UNC sampler with area factor, were compared with the original one, UNC sampler with mesh factor derived from wind tunnel experiments. The effect of increased resolution for the microscopic imaging was examined.

Results: Use of the area factor and a higher resolution eliminated the underestimation for PM10 and PM2.5. The model with area factor had the overall lowest deviation versus the impactor and the cyclone. The intraclass correlation (ICC) showed that the UNC sampler had a higher precision and better ability to distinguish between different exposure levels compared to the cyclone (ICC: 0.51 versus 0.24), but lower precision compared to the impactor (PM10: 0.79 versus 0.99; PM2.5: 0.30 versus 0.45). The particle size distributions as calculated from the different UNC sampler analysis models were visually compared with the distributions determined by APS. The distributions were obviously different when the UNC sampler with mesh factor was used but came to a reasonable agreement when the area factor was used.

Conclusions: High resolution combined with a factor based on area only, results in no underestimation of small particles compared to impactors and cyclones and a better agreement with the APS's particle size distributions. The UNC sampler had lower precision than the impactors, but higher than the respirable cyclone. The UNC sampler with area factor could be used for PM2.5, PM10 and respirable fraction measurements in this working environment without wind.

Figure 1.

Particle mass concentrations for the UNC passive aerosol sampler versus PM₁₀ impactor, respirable cyclone and PM_2.5 impactor. The black cross is the mean for both samplers and the grey dots are each individual observation from the samplers. The value of the slope of the fitted linear regression model and R-squared are noted for each analysis model and particle mass concentration. (a) UNC sampler with mesh factor versus SKC impact sampler for PM₁₀. (b) UNC sampler with hybrid model versus SKC impact sampler for PM₁₀. (c) UNC sampler with area factor versus SKC impact sampler for PM₁₀. (d) UNC sampler with mesh factor versus cyclone for respirable mass fraction. (e) UNC sampler with hybrid model versus cyclone for respirable mass fraction. (f) UNC sampler with area factor versus cyclone for respirable mass fraction. (g) UNC sampler with mesh factor versus SKC impact sampler for PM_2.5. (h) UNC sampler with hybrid model versus SKC impact sampler for PM_2.5. (i) UNC sampler with area factor versus SKC impact sampler for PM_2.5.

Figure 2.

Mean particle mass concentrations at all locations for respirable cyclone versus PM_2.5 impactor.

Figure 3.

Normalised mass concentration distributions (dM/dlogd_a) for the UNC sampler with mesh factor, hybrid model and area factor versus APS at the different locations. Crushing station: (a) mesh factor; (b) hybrid model; (c) area factor. Drive station: (d) mesh factor; (e) hybrid model; (f) area factor. Concentrator: (g) mesh factor; (h) hybrid model; (i) area factor. Concentrate terminal: (j) mesh factor; (k) hybrid model; (l) area factor. Note that the APS was not intended or calibrated for fully quantitative measurements but for characterisation of relative distributions. To best illustrate distributions we therefore used different scales (see left and right hand side of the plots) for the different samplers.