<p>A comprehensive understanding of the effects of meteorology, emission and chemistry on severe haze is critical in the mitigation of air pollution. However, such understanding is largely hindered by the nonlinearity of atmospheric chemistry systems. Here, we developed a novel quantitative decoupling analysis (QDA) method to quantify the effects of emission, meteorology, chemical reaction, and their nonlinear interactions on the fine particulate matter (PM<sub>2.5</sub>) pollution based on the accompanying simulations for different atmospheric processes. Via embedding the QDA method into the Weather Research and Forecasting-Nested Air Quality Prediction Modeling System (WRF-NAQPMS) model, we first employed this method into a typical heavy haze episode in Beijing. Different from the previously sensitive simulation method, which usually linked to a certain period, the QDA achieves the fully decomposing analysis of PM<sub>2.5</sub> concentration during any pollution event into seven different parts, including meteorological contribution (M), emission contribution (E), chemical contribution (C), and interactions among these drivers (i.e., ME, MC, EC and MCE). The results show that the meteorology contribution varied significantly at different stages of episode, from 0.21 µg·m<sup>−3</sup>·h<sup>−1</sup> during accumulation period to −11.82 µg·m<sup>−3</sup>·h<sup>−1</sup> during the removal period, dominating the hourly changes of PM<sub>2.5</sub> concentrations. The chemical contributions were shown to increase with the level of haze, which become largest (0.37 µg·m<sup>−3</sup>·h<sup>−1</sup>) at the maintenance period, 25 % higher than that during the clean period. The contribution of primary emission is relatively stable in all stages due to the use of fixed emission during the simulation. Besides, the QDA method highlights that there exist nonnegligible coupling effects of meteorology, emission and chemistry on PM<sub>2.5</sub> concentrations (−1.83 to 2.44 µg·m<sup>−3</sup>·h<sup>−1</sup>), which were commonly ignored in previous studies and the development of heavy-pollution control strategies. These results indicate that the QDA method can not only provide researchers and policy makers with valuable information for understanding of key factors to heavy pollution, but also help the modelers to find out the sources of uncertainties among numerical models.</p>