criterion dictates that greater than 95% of the measurements fall within + 25% of the true value at 95% confidence in the method’s validation experiments. The method was considered unbiased (i.e., considered the reference method), and the overall precision reflected method accuracy. Sample generation and collection, not analysis, was the main source of variability. When only combustion-source EC is present, different samplers can be expected to give comparable EC results because particles from combustion sources are generally less than 1 :m (diameter). As such, the particles are evenly deposited on the filter and collected with the same efficiency (near 100%). To confirm this assumption, seven different sampler types (open-faced 25-mm and 37-mm cassettes; 298 personal cascade impactor [7 stages, 0.9-:m cutpoint]; 4 prototype impactors) were used to collect diesel aerosol at the loading dock of an express mail facility. The RSD for the mean EC concentration was 5.6% [39]. Based on the 95% confidence limit (19%; 13 degrees of freedom, n =14) on the accuracy, results of this experiment also indicated that the NIOSH accuracy criterion [61] was fulfilled. The amount of EC collected (240 :g per sample) would have been equivalent to sampling an air concentration of 250 :g/m3 for 8 h at 2 L/min. Variability in the OC results was higher (RSD = 12.3%), which is to be expected when different samplers are used to collect aerosols that contain semivolatile (and volatile) components because these can have a filter face velocity dependence. Similar performance was obtained from collected samples in an underground molybdenum mine. Five different sampler types were used (closed-face 25-mm and 37-mm cassettes; 298 cascade impactor [7 stages, 0.9-:m cutpoint]; cyclone with filter; in-house impactor). The RSD for the EC results (mean EC = 297 :g/m3 ) was 7%. The EC deposits obtained with all five sampler types were homogeneous, even when the ore deposit was visually heavier in the center of the filter (e.g., with the closed-face 37-mm cassette). Although the dust loading was higher in the center of the filter, portions taken from the center gave equivalent EC results, indicating the ore contained no EC component. The TC results for the center portions were only slightly higher, so this particular ore was mostly inorganic. EC concentrations found with three different sampler types (nylon cyclone, open-faced cassette, and impactor with submicrometer cut) also were comparable in a study of railroad workers [42]. If high levels of other dusts are present, a size classifier (e.g., impactor and/or cyclone) should be used to prevent filter overloading, particularly if the dust is carbonaceous. In the latter case, a size classifier provides a more selective measure of the diesel-source OC. It also provides a better measure of the diesel-source EC if the dust contains an EC component, which is less common. A finely ground sample of the bulk material can be analyzed to determine whether a specific dust poses potential interference [34]. Depending on the dust concentration, size distribution, and target analyte, an impactor may be required.
For mines, the Mine Safety and Health Administration (MSHA) recommends a specialized impactor to minimize collection of carbonates and other carbonaceous dusts [62]. An impactor can greatly improve the selectivity of the TC measurement in some cases, but it may exclude a small amount of the diesel particulate matter. Then, too, some OC interferences cannot be excluded on the basis of size (e.g., condensation aerosols, fumes, wood and cigarette smokes). If present in the sampling environment, these materials can positively bias
3/15/03
235
NIOSH Manuual of Analytical Methods
