Modeling air quality level with a flexible categorical autoregression

Area of Science:

• Environmental Science
• Statistics
• Data Science

Background:

• Urban air quality is a critical environmental concern.
• Dynamic and systemic features of air quality require advanced modeling techniques.

Purpose of the Study:

• To propose a novel categorical time series model for urban air quality analysis.
• To describe the dynamic and systemic features of air quality using probabilistic distributions.

Main Methods:

• Developed a model combining bounded Poisson and discrete distributions.
• Analyzed daily air quality level data from Beijing, Shanghai, and Guangzhou.
• Employed an adaptive Bayesian Markov chain Monte Carlo (MCMC) sampling scheme for parameter estimation.

Main Results:

• Beijing exhibits the poorest air quality among the three cities, though it is improving.
• Air quality dynamics are most pronounced in Beijing.
• The model demonstrated satisfactory finite sample performance in simulation studies.

Conclusions:

• The proposed model offers a flexible probabilistic structure with a dynamic feedback mechanism.
• The model is computationally efficient.
• The findings provide insights into urban air quality trends and dynamics in major Chinese cities.