Scientific air pollution mapping across space and time: Dealing with data uncertainties and the integration of physical laws.

The study of the spatiotemporal distribution of air pollutants is an important issue due to the health risks associated with these pollutants. In the U.S., criteria air pollutants are measured throughout the country by means of an extensive network of monitoring stations. The high variability of air pollutants across space and time and the varying levels of data accuracy introduce major sources of uncertainty in the study of their spatiotemporal distribution. The last two decades the BME modelling has offered a powerful epistemic framework for integrating various knowledge bases (physical laws, scientific theories, primitive equations, uncertain data sources, secondary information etc.) and producing realistic pictures of air pollutant distribution in a composite space-time domain. In this work we use BME to integrate physical knowledge bases about air pollution variability and to map efficiently the annual arithmetic averages of particulate matter (PM) across the U.S. during the 1984-2000 time period. BME rigorously processes probabilistic (soft) data describing different accuracy levels in the PM measurements and produces informative representations of the spatiotemporal air pollutant distribution and its associated mapping uncertainty. Several applications can be found in the literature in which BME has provided a mathematically rigorously and physically meaningful framework for integrating several kinds of physical laws (in the form of partial differential equations, algebraic equations etc.). This work presents recent developments involving an advection-dominated air pollutant transport equation in space-time. It is shown that the integration of the specified physical law incorporates valuable knowledge about wind and pollution sources in the mapping process and leads to improved predictions of air pollution concentrations.

Key words: Spatiotemporal, BME, random fields, air pollution, particulate matter