Page:NIOSH Manual of Analytical Methods - Chapter Q.pdf/6

 2% or better [60]. The total carbon (TC) in the sample is the sum of OC and EC. If carbonate is present, the carbon in it is quantified as OC, unless a carbonate-subtracted value is requested by the client. Additional details about carbonates are given in a following section. c. Standards Currently, a reference material is not available for determining the OC and EC content of a complex carbonaceous aerosol. For this reason, only the accuracy of the method in the determination of TC could be examined. No discernable differences in the responses of five different organic compounds were noted. Linear regression of the data (43 analyses total) for all five compounds gave a slope and correlation coefficient (r) near unity [slope = 0.99 (± 0.01), r2 = 0.999, n = 43]. In addition to the OC standards, eight different carbonaceous materials were analyzed. Three different methods (including the thermal-optical method) were used, and laboratories reported the TC contents of the samples. The samples analyzed included diesel particulate matter and other types of carbonaceous matter (coals, urban dust, humic acid). Thermal-optical results agreed well with those reported by the two other laboratories. The variability in TC results for the three laboratories ranged from about 1%–7%. These findings [34, 59] indicate that the carbon in a sample is accurately quantified, irrespective of sample type. To estimate the method’s limit of detection (LOD), a set of low-level calibration standards (ethylenediaminetetraacetic acid [EDTA]) was analyzed [34, 59]. The standards covered a loading range from 0.23 to 2.82 :g C (or from 0.15 to 1.83 :g C per cm2 of filter). Results of linear regression of the low-level calibration data were then used to calculate a LOD as 3 Fy/m, where Fy is the standard error of the regression and m is the slope of the regression line. The LOD estimated through the linear regression results was 0.24 :g C, or 0.15 :g/cm2. This value showed good agreement with the LOD estimated as 3Fblank (three times the standard deviation for blanks), which gave a value of about 0.3 :g C. The mean (n = 40) instrumental blank was 0.03 ± 0.1 :g C. With a 960-L air sample collected on a 37-mm filter and use of a 1.5 cm2 sample portion, this LOD translates to an air concentration of about 2 µg/m3. As with all analytical methods, the LOD is a varying number. Other values can be expected, depending on the individual instrument and means by which the LOD is calculated.

d. Air Sampling In the initial evaluation of the thermal-optical method, a set of laboratory-generated air samples was analyzed. A dilution tunnel equipped with a dynamometer was used for generation of diesel particulate samples. Four EC concentrations, ranging from 23 to 240
 * g/m3 (EC loadings from 2.7 to 27 :g/cm2 ), were generated. The analytical results [59]

indicated that the method met the NIOSH accuracy criterion [61]. The variance was roughly proportional to the mean concentration; therefore, the relative standard deviation (RSD) decreased with increasing concentration. The accuracy was calculated accordingly. The accuracy was ± 16.7% at the lowest loading (2.7 :g/cm2 ), with an overall precision (RSD) of 8.5%. On the basis of a method evaluation, the NIOSH accuracy criterion requires a confidence limit on the accuracy less than 25% at the 95% confidence level. Restated, the

3/15/03

234

NIOSH Manuual of Analytical Methods