The 2014 fire in the Hazelwood open-cut mine of brown coal, located in the State of Victoria (Australia), burned for 45 days. The fire sent dense smoke over the nearby town of Morwell and beyond, resulting in one of the worst air quality incidents in Victoria. Precision air monitoring of PM2.5 (particulate matter 2.5 μm or less in diameter) and carbon monoxide (CO), which has been reported previously, started a few days after the fire at two locations in Morwell and measured PM2.5 levels up to 19 times higher than the 24-h Australian air quality standard. Because of the sparseness of the monitors and the fact that the fire was most intense prior to the start of the air monitoring, it is likely that the smoke concentrations in Morwell were even greater than measured. Thus, the concentration measurements are insufficient in time and space for a comprehensive study of exposure and health impacts due to the smoke. We reconstruct the hourly spatial distributions of smoke (as represented by PM2.5 and CO) in and around Morwell by first developing a rigorous methodology for estimating the fire emissions, and then using them in a high-resolution prognostic meteorological and dispersion model with local wind data assimilation and an appropriate plume rise mechanism. Larger-scale modelling is also conducted to estimate the background concentrations without the mine fire. The model simulates the number of observed exceedances and the observed maximum concentrations exceeding the air quality standards for both PM2.5 and CO within a factor of 2. At the monitor south of Morwell near the fire, the model predicts hourly PM2.5 and CO concentrations as high as 3730 μg m−3 and 58.6 ppm, respectively, in the early phase of the fire; these levels are much higher than those recorded by the subsequent air monitoring. The modelled PM2.5 fields are being used by other researchers to estimate the impact of smoke exposure on health outcomes in the local community.