Disentangling Multiannual Air Quality Profiles Aided by Self-Organizing Map and Positive Matrix Factorization

Area of Science:

• Environmental Science
• Data Science
• Public Health

Background:

• Air pollution poses significant risks to human health, necessitating robust evaluation methods.
• Analyzing large, complex air quality datasets with missing data and noise presents a major research challenge.
• Multivariate data analysis, including neural networks and chemometrics, is increasingly employed for air pollution source identification.

Purpose of the Study:

• To develop and validate a unified method for analyzing multiannual, multisite air pollution data.
• To effectively identify pollutant sources, their temporal variations, and site-specific characteristics.
• To overcome limitations of traditional methods by processing diverse datasets in a single step.

Main Methods:

• Integration of Self-Organizing Map (SOM) for data clustering.
• Application of Hierarchical Clustering Analysis (HCA) for grouping similar patterns.
• Utilization of Positive Matrix Factorization (PMF) for source apportionment.

Main Results:

• The combined SOM-HCA-PMF approach successfully disentangled complex air pollution data.
• Site-specific pollutant source profiles and their yearly variations were accurately identified.
• The method demonstrated robustness in handling noisy and incomplete datasets, revealing outliers.

Conclusions:

• The integrated approach offers a powerful tool for comprehensive air pollution assessment.
• It enables reliable interpretation of multiannual, multisite air quality data, improving source identification.
• This method facilitates a deeper understanding of pollution dynamics and health impacts.