Statistical Detection of Abnormal Ozone Levels Using Principal Component Analysis

Ozone is one of the most severe air pollution problems in the world. The concentration of ozone in the troposphere is of great interest because of its negative influence on the human health, vegetation and materials. The complexity of ozone (O3) formation mechanisms in the troposphere, the complexity of meteorological conditions in urban areas and the uncertainty in the measurements of all the parameters involved, make fast and accurate modeling of ozone a challenging task. In the absence of a process model, multivariate statistical techniques, such as principal component analysis (PCA) have been successfully used in fault detection (FD) of highly correlated process variables. This paper presents an application of PCA in detecting abnormalities in ozone measurements, which are caused by air pollution or any incoherence between the different network sensors or sensor faults. Practical data from various ozone surveillance network stations in Upper Normandy, France, are used in this analysis. Index Term— Ozone pollution, fault detection, principal