Control charts for dynamic process monitoring with an application to air pollution surveillance

Abstract

Air pollution is a major global public health risk factor. Among all air pollutants, PM${}_{2.5}$ is especially harmful. It has been well demonstrated that chronic exposure to PM${}_{2.5}$ can cause many health problems, including asthma, lung cancer and cardiovascular diseases. To tackle problems caused by air pollution, governments have put a huge amount of resources to improve air quality and reduce the impact of air pollution on public health. In this effort it is extremely important to develop an air pollution surveillance system to constantly monitor the air quality over time and to give a signal promptly once the air quality is found to deteriorate so that a timely government intervention can be implemented. To monitor a sequential process, a major statistical tool is the statistical process control (SPC) chart. However, traditional SPC charts are based on the assumptions that process observations at different time points are independent and identically distributed. These assumptions are rarely valid in environmental data because seasonality and serial correlation are common in such data. To overcome this difficulty, we suggest a new control chart in this paper, which can properly accommodate dynamic temporal pattern and serial correlation in a sequential process. Thus, it can be used for effective air pollution surveillance. This method is demonstrated by an application to monitor the daily average PM${}_{2.5}$ levels in Beijing and shown to be effective and reliable in detecting the increase of PM${}_{2.5}$ levels.