XGBoost-based analysis of short- and long-term COVID-19 air quality impacts in Tehran, Iran.

This study examines the air quality trends associated with COVID-19 lockdowns and the subsequent recovery on air quality in Tehran, employing an established XGBoost framework for de-weathering with daily data from 22 monitoring stations from 2019 to 2025. A comprehensive analysis was conducted on six principal pollutants (PM₂.₅, PM₁₀, NO₂, O₃, SO₂, and CO) across four distinct temporal phases: Pre-pandemic, Pandemic, Post-COVID, and After Termination. While the model effectively reduced meteorological bias, exhibiting strong efficacy for particulate matter (R = 0.78-0.88; IOA = 0.68-0.88; FAC2 > 0.95), we explicitly acknowledge less consistent outcomes for reactive gases such as O₃ (R < 0.75; NMGE > 0.30). The imposition of lockdown measures coincided with significant reductions in pollutants associated with vehicular traffic: NO₂ exhibited a decline of - 30 to - 36%, PM₂.₅ decreased by - 18 to - 22%, and CO diminished by - 20%. In contrast, O₃ experienced an increase of + 20-30%, indicative of diminished NO titration. Principal Component Analysis (PCA) corroborated these trends, with PC1 (NO₂, PM₂.₅, CO) accounting for approximately 60% of the variance throughout the pandemic. In subsequent phases, partial recoveries were noted: PM₂.₅ escalated by + 15-20%, NO₂ stabilized at around -10%, SO₂ consistently decreased (-30% by 2025), whereas O₃ surged by + 20% by 2025, clustering with PM₁₀ in PC2, thereby highlighting the impacts of dust and secondary formation processes. Lockdowns yielded transient enhancements in air quality, with restricted long-term advantages. Sustainable progress necessitates the adoption of comprehensive policies targeting traffic, industrial emissions, VOCs, and dust management.