Statistically Based Air-Quality Indices

Oak Ridge National Laboratory is, under sponsorship of the NSF, studying the development of air quality or air pollution indices. We have examined a number of ways for combining measured information about several pollutants into a single number that reflects the overall state of the air quality within a given area. A number of these index systems have been strictly mathematical relationships that tended to give the index itself desirable qualities, that is, the desired response as values of the measurements changed. As this work progressed, however, we have come to believe that some sort of statistically based index is probably the ultimate answer. We feel this way for several reasons, one of which is that there seems to be some truth in the belief that air quality measurements follow a definable statistical distribution. More importantly, the hazard or harm that results from air pollution is usually measured — when it is measured — in statistical terms, that is, the frequency of effect. So it seems that if we could devise a valid statistically based indexing system, it should be compatible with the damage information or data that is available or that will become available in the future. Accordingly, such an index may have something to say about how standards are set or at least point the way to a valid assessment of standards. In this paper, we wish to present some of the assumptions and rationale and flaws associated with this approach. This work is not complete.