Abstract:
The measurement of ambient particulate concentration (PM10 and PM2.5) using active monitors requires expensive or labour-intensive apparatus or both. Although PM is a priority pollutant, this has precluded widespread or intensive ambient particulate monitoring networks to be set up in South Africa, except in some priority areas. In this study, locally manufactured passive PM monitoring devices (samplers), based on a design initially published by researchers at the University of North Carolina (UNC), were calibrated for the measurement of PM10-2.5. Duplicate samplers of each type (local and UNC) were co-located at eight reference stations equipped with TEOM (Tapered element oscillating microbalance) or BAM (Beta radiation attenuation monitor) particulate monitors in three networks. Imaging of the substrate was carried out using an optical microscope to limit the cost of analysis. The images (10 to 49 per sample) were analysed using both proprietary (Zeiss AxioVision®) and open-source (ImageJ) software at 100X and 200X magnification. Considerable variation exists between the four co-located local and UNC samplers at all stations, although the latter show lesser discrepancies when analysed using the AxioVision software. Greater agreement of the local samplers and the continuous monitors is noted at a 200X and 100X magnification using the AxioVision software with R2 = 0.81 and R2 = 0.79 respectively. The precision of PM10-2.5 measured with the passive samplers was highly variable with calculated CVs ranging from 10.4% to 73.3%. 82% of the CVs were less than 40%. The average CV for all samplers was 34.6%. Passive samplers analysed using the AxioVision software recorded smaller average discrepancies of 45.3% at 100X and 37.3% at 200X magnification. Samplers analysed using ImageJ at 100X magnification exhibited the highest percentage difference from the reference values (81.2%). Using a two factor ANOVA we can show that (at 95% confidence) the analysis software and the imaging magnification have the most significant effect on the calculated sampler concentration results. A disadvantage of passive samplers is large number of days (2-7) required to produce statistically significant values hence disqualifying it for use as a reference method. The device is, therefore, more suitable for screening-level, high spatial density sampling, but some non-regulatory applications are pointed out.
