A severe dust event during 19-22 March 2002 over East Asia was simulated by using the Weather Research and Forecasting/Chemistry (WRF/Chem) model coupled with two different dust emission schemes (the GOCART and Shao04 schemes) to investigate the impact of dust emission schemes on model performance. A comparison with observational data shows that the main observed dust emission regions, the temporal variation of dust emission intensity, and the evolution of surface dust concentration are all quite well reproduced. Both near the dust source and in its downstream regions, the model-simulated temporal variation in surface dust concentration agrees closely with station observations. Generally, the WRF/Chem model with the Shao04 scheme shows better performance, especially in simulating the occurrence and intensity of dust emission and the surface dust concentration distribution. Further analysis shows that the dust emission regions simulated by the GOCART scheme are broader than those from observational data, which might be ascribed to the lower criteria value of the threshold wind velocity at the surface. In addition, the GOCART scheme could not resolve the potential dust source in the southeast of Mongolia and the middle-east of Inner Mongolia, which leads to the underestimation of dust concentration in this area and in downwind regions. However, the simulated dust emission flux in Taklimakan with the Shao04 scheme is lower compared with observations. This could be ascribed to the fact that the effect of aerodynamic lift in the dust emission is not considered in the Shao04 scheme. Moreover, the inaccurate parent soil texture for the Taklimakan Desert used in the Shao04 scheme might be another possible reason for the discrepancy.