Based on coherent Doppler wind lidar and ground-based conventional observational data, a typical dust weather process occurring in Hulunbuir was analyzed using machine learning and the HYSPLIT model. The study revealed that the dust event started with a sudden increase in southerly wind. Subsequently, the wind direction shifted to south-southwest, resulting in a reduction in wind speed and a weakening of dust transport. However, when the wind shifted to westerlies, the dust transport intensified again. The transport of dust ceased after a decrease in westerly wind speeds. During the dust transport period, turbulence was relatively weak, and the mixing layer height remained limited. Machine learning particle size calculations indicated PM10 dominating the early transport and both PM10 and PM2.5 showing substantial growth in the later phase. This divergence in particle sizes across different transport periods suggests a potential change in dust sources. HYSPLIT revealed that in the early phase of dust transport originated from northwestern Mongolia, passing through Xilingol League in China before reaching Hulunbuir. In the later phase, dust transport directly entered Hulunbuir from the southern regions of Russia, resulting in an escalation of dust pollution. Finally, using total mass flux analysis, it was determined that the response to dust occurred earlier in the period from the pre-dust period until the beginning, compared to ground-level particulate concentration. The total mass flux threshold for this dust event was established using box-and-whisker plot. Variations in total mass flux and the establishment of thresholds could serve as novel indicators for dust event early warning systems.