Modeling Particulate Matter Using Uncertain Measurements: Case Study for PM-10 in North Carolina
Marc Serre, University of North Carolina
Monitoring atmospheric air pollutants is an issue of primary importance in our society. Toxic air pollutants are poisonous substances in the air that come from natural or manmade sources, and can harm the environment or our health. Two key issues need to be addresses when mapping air pollutants: a) the distribution of air pollutants exhibits a high variability in both space and time, and b) the measurements of air pollutants at monitoring stations may be uncertain. The Bayesian Maximum Entropy (BME) method of modern Geostatistics provides an adequate and rigorous framework for this type of analysis. BME uses the theory of Space/Time Random Field S/TRF that accounts for the space/time variability of air pollutants. Additionally BME incorporates uncertain information (also called soft data) in a rigorous manner, providing more accurate mapping estimates than the classical kriging methods when a combination of hard (exact) and soft information is used. This approach is demonstrated on a case study concerned with mapping ambient PM10 in the state of North Carolina.
Abstract Author(s): Marc Serre