A dependent Bayesian Dirichlet process model for source apportionment of particle number size distribution

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Public Health

Background:

• Particulate matter (PM) exposure poses significant health risks.
• Identifying diverse PM sources is crucial for effective air quality policies.
• Existing source identification methods have limitations, including pre-specifying source numbers and not incorporating covariates.

Purpose of the Study:

• To develop a novel Bayesian nonparametric approach for particulate matter source identification.
• To overcome limitations of existing methods by estimating the number of sources and including covariates.
• To improve the characterization and identification of PM sources for better air quality management.

Main Methods:

• Utilized a Bayesian nonparametric approach with a Dirichlet process prior for source profiles.
• Modeled source contributions, allowing the number of sources to be estimated rather than fixed.
• Incorporated meteorological variables (wind speed and direction) as covariates using a Gaussian kernel for enhanced source characterization.

Main Results:

• Successfully applied the model to particle number size distribution data near London Gatwick Airport.
• Identified common particulate matter sources.
• Discovered new PM sources not detectable by conventional methods.

Conclusions:

• The proposed Bayesian nonparametric method offers a flexible and powerful tool for particulate matter source apportionment.
• This approach enhances the ability to identify a wider range of PM sources, including previously unknown ones.
• The findings support the development of more targeted and effective air quality improvement strategies.