2005JD006400-text1.txt

      In order determine if the results from XRF and ICPMS are equivalent, the ratio of XRF to ICPMS and the 95% confidence interval based on measured uncertainties is compared to 1, the value of XRF/ICPMS if the two methods are exactly equal.  If the confidence interval overlaps with 1, then the methods are not significantly different from each other.  SuppFigure1.eps shows the XRF/ICPMS ratio and 95% confidence interval for all seven elements for PM2.5 and TSP along with the mass concentration for each event.  

      For PM2.5, concentrations measured by XRF and ICPMS are not significantly different for four of the six events, March 22, 24, 30 and April 1, for all seven elements and for all 42 parallel PM2.5 concentrations, 78% were not significantly different from each other.  It is possible, because the filters used for the ICP-MS were half filters, that there was an error in filter cutting for these two events in which the XRF and ICP-MS concentrations were significantly different.  For the cut filters, one half was used for ICP-MS and one half was used for IC.  Had the filter been cut in half in such a way that the smaller half was used for ICP-MS, then the S and Cl concentration obtained by IC would be higher than the S and Cl obtained by XRF.  This is not the case for either event, indicating that there was not a cutting error.  The events in which the XRF and ICPMS give significantly different results are low PM2.5 concentrations events.  Therefore, difference in the two methods may be due to incomplete representation of the error at low concentrations of elements, although this is inconclusive since a third low concentration event yields XRF/ICPMS ratios very close to 1.

      For TSP, four of the six events, March 26, 28, 30 and April 1, the XRF/ICP-MS ratio is within the 95% confidence interval for all elements except for Ti, Fe and Rb on March 26.  Over all TSP samples, 74% of the comparisons of element concentrations were not significantly different from each other.  It is possible that the difference in concentrations on March 22 and 24 are due to a cutting the filters, however, like the PM2.5 samples, S and Cl from IC and S and Cl from XRF were not significantly different for both events suggesting that these differences are not likely due to filter cutting error.  The majority of the samples for which the XRF concentration is higher than the ICP-MS concentration are on March 22 and 24 when the TSP concentration is high.  XRF mass results are dependent on particle size and a correction factor to account for presence of large particles is used (XRF manual).  The correction factor was developed for PM10 samples and may not accurately account for the presence of very large particles during dust storms.  As discussed below, March 22 and 24, the two events with the most discrepancies, were dust storm events.