Pinnacles and pitfalls for source apportionment of potential health effects from airborne particle exposure

Since its origins in the 1970s, source apportionment using receptor modeling has improved to a point where both the chemical mass balance and various methods of factor analysis have been applied to many urban and regional data sets to infer major sources or source classes influencing airborne particle concentrations. Recently the factors from the latter analyses have been combined with regression techniques using human health endpoints to infer source influence on health effects. This approach is attractive for air quality management when the composition of particles is known, since it provides, in principle, a means of quantifying major source influence on health consequences. The factor-based analyses have been used for both epidemiological and toxicological studies with some success. While the method is useful in many ways, it also has important limitations that include failing to identify specific sources, misidentification from comingled source factors, and inconsistency or unreasonableness of results from the same locations using different factor techniques. Examples of ambiguities evolving from these limitations are cited in this article. Ambiguity found in the literature is fostered by loosely worded terminology that does not distinguish statistically based factors from actual sources, and from health impacts inferred by single centrally located air monitors, which are assumed to represent actual exposure or dosage to airborne particles.