Diagnosing Dominant Drivers of Sulfate Simulation Uncertainties with the EPICC Model: An Integrated Assessment of Chemical Mechanisms and Emission Inventories

Sulfate aerosol is a key driver of severe haze pollution in China, yet atmospheric chemical transport models (CTMs) typically underestimate its concentrations, and the dominant formation pathways remain highly controversial. Clarifying the specific model uncertainties responsible for this is critical for accurate haze forecasting and effective control measures. Here, we apply a diagnostic framework, integrating the EPICC model with Bayesian Monte Carlo simulations and a Reduced-Form Model, to diagnose the sources of sulfate simulations in China. Our results reveal a seasonal shift in the dominant uncertainty sources of sulfate simulations. Wintertime uncertainties are driven by Mn-catalyzed heterogeneous pathway (MnHET,45%–60%) and NH3 emissions (11%–27%). These winter uncertainties are modulated by meteorology, peaking under high-RH and moderate-temperature conditions, explaining why models fail specifically during severe haze episodes. In contrast, summertime uncertainties associated with current model parameters are dominated by inaccuracies in emission inventories, particularly for condensable particulate matter(CPM,25%–50%) and sulfur dioxide (SO2,14%–36%), while an underestimated sulfur oxidation ratio (SOR) also point to the contribution of missing chemical mechanisms. Incorporating the MnHET pathway and correcting underestimated NH3 and missing CPM emissions largely resolves the wintertime sulfate simulation bias. We demonstrate that underestimating NH3 causes a misattribution of the dominant sulfate formation pathways by suppressing the Mn-catalyzed pathway. Furthermore, incorporating the MnHET pathway reveals that NH3 emission reduction is a more effective haze mitigation strategy than conventional models suggest. This study highlights the critical need for accurate emission inventories and provides a robust scientific basis for developing the EPICC model and other CTMs.