An aggregate AQI: Comparing different standardizations and introducing a variability index

Many studies demonstrate a strong relationship between air pollution and respiratory and cardiovascular diseases. For this reason, assessing air pollution, and conveying information about its possible adverse health effects, may encourage population and policy makers to reduce those activities increasing pollution levels. In this paper a relative index of variability, to be associated with the aggregate Air Quality Index (AQI) among pollutants proposed by Ruggieri and Plaia (2011), is developed in order to better investigate air pollution conditions for the whole area of a city/region. The most widely-used and up to date pollution indices, based mainly on AQI computed by the US Environmental Protection Agency (EPA) and often defined by the value of the pollutant with the highest (opportunely standardized) concentration on a given day, aim at warning the people for short term health impact. An aggregate AQI, taking into account the combined effects of all the considered pollutants, gives emphasis to possible chronic health effects and long term damages on environment caused by air pollution. The proposed index of variability adds precious information to the aggregate AQI, as it allows one to know whether the value assumed by the AQI is influenced by one or more pollutants. The two indices are jointly used on simulated data, considering different possible scenarios. Applications to real air pollution data are also reported. Before applying the two indices, the effects of different standardizations on data are evaluated from a theoretical point of view.

► A relative index of variability to be associated with an aggregate AQI is proposed.
► An aggregate AQI accounts for the combined effects of all the considered pollutants.
► The variability index clarifies if the AQI value depends on one or more pollutants.
► The effects of different types of standardization on data are also evaluated.
► We show why the standardization by segmented linear function is to be preferred.