Evaluating Trends in Historical PM2.5 Element Concentrations by Reanalyzing a 15-Year Sample Archive

Abstract:

The IMPROVE (Interagency Monitoring of PROtected Visual Environments) network monitors aerosol concentrations at 170 remote sites throughout the United States. Twenty-four-hour filter samples of particulate matter are collected every third day and analyzed for chemical composition. About 30 of the sites have operated continuously since 1988, and the sustained data record (http://views.cira.colostate.edu/web/) offers a unique window on regional aerosol trends.

All elemental analyses have been performed by Crocker Nuclear Laboratory at the University of California in Davis, and sample filters collected since 1995 are archived on campus. The suite of reported elements has remained constant, but the analytical methods employed for their determination have evolved. For example, the elements Na – Mn were determined by PIXE until November 2001, then by XRF analysis in a He-flushed atmosphere through 2004, and by XRF analysis in vacuum since January 2005. In addition to these fundamental changes, incompletely-documented operational factors such as detector performance and calibration details have introduced variations in the measurements.

Because the past analytical methods were non-destructive, the archived filters can be re-analyzed with the current analytical systems and protocols. The 15-year sample archives from Great Smoky Mountains (GRSM), Mount Rainier (MORA), and Point Reyes National Parks (PORE) were selected for reanalysis. The agreement between the new analyses and original determinations varies with element and analytical era. The graph below compares the trend estimates for all the elements measured by IMPROVE based on the original and repeat analyses; the elements identified in color are measured above the detection limit more than 90% of the time. The trend estimates are sensitive to the treatment of non-detect data. The original and reanalysis trends are indistinguishable (have overlapping confidence intervals) for most of the well-detected elements.