Assessing long-term spatiotemporal patterns of PM2.5 using hybrid geostatistical modeling

Air pollution and fine particulate matter (PM2.5) pose significant environmental and public health challenges, particularly in rapidly urbanizing regions. Long-term assessment of spatiotemporal patterns of PM2.5 is essential for effective air quality management and pollution mitigation. This study examines the long-term spatiotemporal distribution of PM2.5 concentrations across multiple decades using advanced geostatistical and hybrid modeling approaches. Pakistan, ranked as one of the most polluted countries worldwide according to recent global air quality assessments, is selected as the study region to investigate persistent patterns of particulate pollution. A hybrid spatial interpolation (HSI) framework integrating machine learning algorithms with regression kriging is employed to improve prediction accuracy and capture complex spatial trends. A decade-by-decade analysis was carried out to capture changes in PM2.5 patterns over successive decades. The results indicate that PM2.5 concentrations remained consistently elevated across most regions of Pakistan, frequently exceeding the WHO annual guideline of $5\mu {\text{g,m}}^{-3}$ and, in recent assessments, reaching levels multiple times higher than this threshold, with evidence of increasing extremes and spatial expansion rather than a uniform rise in average levels. Spatial analysis revealed a persistent southwest to northeast orientation along the Indus corridor, covering Karachi, Hyderabad, and Sukkur, and extending toward the central plains of Multan, Lahore, Faisalabad, and Gujranwala. The proposed HSI framework achieved improved interpolation performance and provided a more refined representation of spatial PM2.5 patterns. Overall, the findings highlight the persistence and spatial expansion of PM2.5 concentrations and provide a quantitative basis for improved air quality management and policy formulation.