Integrating Self-Organizing Maps, Positive Matrix Factorization and Time-Series Decomposition for Urban Air Pollution...

Receptor modeling of ambient pollutant concentrations plays a central role in urban air quality assessments. This study proposes an integrated framework combining Self-Organizing Maps (SOM), Positive Matrix Factorization (PMF), and Time-Series Analysis (TSA) for a comprehensive evaluation of urban air pollution patterns and source dynamics. The methodology was applied to multi-annual air quality and meteorological datasets (2009-2018) from two major Bulgarian cities, Plovdiv and Varna. The SOM was used for assessing the overall parameter patterns of the cities, leading to a clear clustering of the site samples on the map. Thus, PMF was run separately for the two sites, identifying a different number of sources (three and four, respectively). Traffic-related and sulfur-rich combustion sources were identified in both cities, while a crustal/resuspended dust factor was observed only in Varna. TSA revealed distinct temporal behaviors among source types. Traffic-related aerosol contributions decreased in both cities (-5.14% yr^-1 in Plovdiv; -9.30% yr^-1 in Varna), whereas sulfur-rich combustion factors showed increasing trends (+4.64% yr^-1 and +2.97% yr^-1, respectively). Traffic fresh exhaust factors exhibited pronounced seasonal variability and significant weekday-weekend differences in both cities. The integrated SOM-PMF-TSA framework enhanced source interpretability and temporal characterization, providing a robust approach for urban air quality assessment and supporting targeted air pollution management strategies.