Spatiotemporal modeling of PM2.5 data with missing values

this paper is the analysis of PM 2.5 data (particulate matter of aerodynamic diameter 2.5 mm or less) from a network set up by the U.S. Environmental Protection Agency (EPA). Under the Clean Air Act and amendments, the EPA is responsible for maintaining standards on air pollutants that are "requisite to protect human health" with "an adequate margin of safety." Among the currently regulated pollutants is PM 10 (particulate matter of aerodynamic diameter 10 mm or less). Under a new standard, proposed in 1997 but not yet implemented, this is extended to include PM 2.5 . Specifically, the proposed standard requires that (1) the 3-year average of the 98th percentile of PM 2.5 should not exceed 65 mgm #3 and that (2) the arithmetic mean (over all monitors within a given region) of the 3-year average of daily PM 2.5 levels should not exceed 15 mgm #3 . Further information about the background of these standards and the epidemiological basis behind them is given by Cox [2000]. [4] In 1999, the EPA established a network of some 800 PM 2.5 monitors to supplement the much longer established PM 10 monitoring network. This new network has been used to gather information about the spatial distribution of PM 2.5 , which is needed to help design a long-term PM 2.5 network and which can also be used to answer other research questions, such as whether spatial interpolation could help in determining individual exposure to air pollution [National Research Council, 1998, p. 96]. Satisfactory answers to this and related questions, however, require methodological research on spatiotemporal analysis of PM 2.5 dat