Abstract Diesel particulate matter (DPM) is carcinogenic to humans. DPM concentrations in underground mines are much higher than other working environments, thus pose substantial health threats to miners due to overexposure. Computational fluid dynamics is commonly used to study the DPM dispersion and assess the concentration distribution in various working environments. However, most such studies for underground mines treated DPM as a continuous phase (gas phase) in the model. DPM is a solid discrete phase, and its behaviours could be quite different from that of gaseous contaminants. This study compared DPM concentration distributions by using three modelling methods: the Eulerian-Lagrangian method and the Eulerian-Eulerian method that treats DPM as discrete phase particles, and the species transport method that treats DPM as a continuous phase gas. The model was based on a typical underground mine development face with a forcing auxiliary ventilation setup. It was found that the general DPM concentration distribution for the three numerical methods was similar for simple geometry with more uniform flow regions. However, large discrepancies existed in the development heading with complex geometry and flow features. The findings suggest that when simulating DPM, although the species transport method can provide relatively accurate results with much less computational time, the parameters of the modelled gas need to be carefully calibrated to get a better simulation result. For key areas where the diesel machinery and miners are usually located, the Eulerian-Lagrangian method should be used for more accurate analysis.