Abstract: Based on coherent Doppler wind lidar and ground-based conventional observational data, a typical dust transport event occurring in Hulunbuir, China, was analyzed herein using machine learning and the HYSPLIT (Hybrid Single-Particle LagrangianIntegrated Trajectory) model. The study revealed that the dust transport event started with a sudden increase in southerly wind. Subsequently, the wind direction shifted to south–southwest, resulting in wind speed and dust transport reductions. However, when the wind shifted to westerlies, dust transport again intensified. The transport of dust ceased after the westerly wind speed decreased. During the dust transport period, turbulence was relatively weak, and the mixing-layer height remained limited, particularly when compared to conditions without the influence of dust. Machine learning–based particle size calculations indicated that PM10 dominated the early dust transport phase, and both PM10 and PM2.5 showed substantial growth in the later phase. This divergence in particle sizes across different transport periods suggested a potential change in dust sources. The HYSPLIT model revealed that the early phase of dust transport started from northwestern Mongolia, passing through the Xilingol League in China before reaching Hulunbuir. In the later phase, dust directly entered Hulunbuir from the southern regions of Russia, escalating dust pollution. Finally, the study found that the response of total mass flux to dust occurred earlier than the change in ground-level particulate concentration from the pre-dust phase to the onset. Furthermore, total mass flux values during dust events were significantly higher than those before and after the dust events. Therefore, total mass flux variations and threshold establishment can serve as novel indicators for dust events in early warning systems.