A novel calibration method for continuous airborne metal measurements: Implications for aerosol source apportionment.

Continuous metal monitors have been widely used in environmental monitoring due to the high temporal resolution, high detection limit, and necessity for near real-time source apportionment. However, the reliability of the conventional calibration method, the deviation caused by uncalibrated monitoring data, and the subsequent impact on source identification results are rarely discussed. In this study, a reliable multi-point calibration approach by Primary Standard Aerosol Mass Concentration Calibration System (PAMAS) for the Xact625i Ambient Metals Monitor was developed and applied. The measured data was almost meaningless in the low-concentration range with bias even exceeding 100 % by using the conventional single-point calibration method based on thin-film standards. PAMAS was utilized to generate aerosols with known concentrations of the 20 metal elements including Al, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Ga, As, Sr, Cd, Sn, Sb, Ba, Tl, Pb, and Bi, in two concentration ranges of 150-1200 ng m-3 and 2.5-30 ng m-3 to validate the Xact625i Monitor. The results showed that the elemental concentrations were underestimated, especially in the low-concentration range, only for Cr, As, and Sr with slopes close to unity (1.00 ± 0.03). After calibration by PAMAS, the slopes of the linear relationships between measured and standard concentrations were all unity for the 19 elements in the high-concentration range, and close to unity for the 15 elements in the low-concentration range, and the accuracy of the remaining elements was also improved. After considering the calibration of aerosol metal data, it was found the number of source factors and their contributions to metals and PM2.5 in Chongming Dongtan, China, based on the PMF model significantly changed. This study highlighted the need of developing reliable calibration methods for online aerosol monitoring instruments and implied that the source apportionment results could be biased without careful data calibration.