For example, nitric acid and organic vapors may adsorb on the filter used for the collection of atmospheric particles (positive artifacts), and ammonium nitrate or organic PM may evaporate from the filter during sampling (negative artifacts). Many studies have been completed or are being conducted to develop and test more reliable monitoring methods and analytical procedures to determine the chemical composition of PM (see NARSTO  for a review). A promising new technology based on automated semicontinuous sampling technology has been developed in the supersites programs and could be used in the routine Speciation Monitoring Network.
There is an emerging interest in bioaerosols—fine PM of biological origin, which may include allergens, viruses, bacteria, or fungi. They can enter the atmosphere inadvertently as a result of animal-feeding operations (NRC 2003c) or intentionally as a result of bioterrorism (NRC 2002d). Identification of these materials will be more complex than simple chemical or elemental analysis and may involve microbiological tests.
Because the primary focus of monitoring strategies has been on documenting NAAQS attainment, there is little motivation to develop methods to measure ambient concentrations that are substantially below the standards. Such measurements are needed to provide critical data for understanding atmospheric processes and providing input for air quality models.
A variety of techniques can be used to determine long-term trends in ambient pollutant concentrations. The method most commonly used by EPA is to compute yearly averages of concentrations at all stations within a metropolitan statistical area (MSA). Values for missing yearly averages are linearly interpolated if in a middle point; missing end points are replaced with the nearest year of valid data. Linear regression analysis is done, and a method known as the Theil test (Hollander and Wolfe 1973) is applied to detect the significance of any trend. This approach is a fairly simple, straightforward method of trend detection, but it has several important limitations. As noted earlier, the monitoring stations may not be, and often are not, spatially distributed to document the average air quality over the region of interest, and thus the trends derived might not be representative of the region. The results of a linear regression analysis can be highly sensitive to such factors as the time interval chosen for analysis and the frequency of observations. In some cases, sparse data make it difficult to calculate trends. For instance, samples of PM are generally collected at a frequency that is longer than most PM episodes. Consequently, several years of data collection will be needed to characterize seasonal and spatial patterns accurately and an even longer period of sampling will be needed to detect long-term temporal trends.