Complex air pollutant sources and distinct meteorological conditions resulted in unique wintertime haze pollution in the Harbin-Changchun (HC) metropolitan area, China's only national-level city cluster located in the severe cold climate region. In this study, field observation and air quality modeling were combined to investigate fine particulate matter (PM2.5) pollution during a six-month long heating season in HC's central city (Harbin). The model significantly underpredicted PM2.5 and organic carbon (by up to ∼230 μg/m3 and 110 μgC/m3, respectively, in terms of daily average) when levoglucosan concentrations were above 0.5 μg/m3. Based on a synthesis of levoglucosan concentrations and fire counts, the large gaps were attributed to underestimation of open burning emissions by the model. However, the model tended to overpredict elemental carbon (more significantly at higher NO2), likely pointing to an overestimation of vehicle emissions. With increasing levoglucosan, the difference between observed and simulated nitrate (nitrateobs ‒ nitratemod, i.e., Δnitrate) showed a transition from negative to positive values. The positive Δnitrate were attributed to underprediction of the open-burning related nitrate, whereas the negative Δnitrate were likely caused by overprediction of nitrate from other sources (presumably vehicle emissions). The dependence of Δnitrate on levoglucosan indicated that with stronger impact of open burning, the overprediction effect was gradually offset and finally overwhelmed. Influence of open burning on sulfate formation was evident as well, but less apparent compared to nitrate. This study illustrates how the uncertainties in open burning emissions will influence PM2.5 simulation, on not only primary components but also secondary species.