Quantifying the driving factors of particulate matter variabilities in the Beijing-Tianjin-Hebei and Yangtze River Delta regions from 2015 to 2020 by machine learning approach

Particulate matter (PM) pollution is a critical air quality challenge in China. This study quantifies meteorological versus anthropogenic contributions to PM variations in Beijing-Tianjin-Hebei (BTH) and Yangtze River Delta (YRD) (2015–2020) using ground observations, meteorological assimilated data, emission inventories, and a LightGBM model. Observations show significant PM2.5 and PM10 declines (e.g., BTH PM2.5: −0.07 ± 0.03 μg m⁻³ yr⁻¹; PM10: −0.11 ± 0.04 21 μg m⁻³ yr⁻¹). Model decomposition identifies anthropogenic emission reductions as the primary driver (PM2.5 decrease: 7.19–24.76 μg m⁻³; PM10 decrease: 0.40–27.12 μg m⁻³). Key meteorological drivers differ: 2-m specific humidity (QV2M), sea-level pressure (SLP), 2-m temperature (T2M), and 10-m meridional (V10M) collectively explain 15 % of PM2.5 variance; precipitation flux (PRECTOT) is critical for PM10. PM2.5 concentrations are primarily governed by PM10, CO, NO2, and SO2 (cumulative contribution 37.60 %), while PM10 variations center on PM2.5, interacting with NO2, CO, and SO2 (explaining 34 % variance). PM2.5 shows stronger correlation with CO than PM10 (regional difference +0.07–+0.08), linked to combustion/SOA. SO₂/NO₂ exhibit comparable PM correlations but divergent mechanisms: NO₂ with traffic/nitrate, SO₂ with stationary sources/sulfate, both via "co-emission-chemical transformation-meteorological synergy". Our research support optimizing region-specific control strategies.