使用天气学检验方法，对中国气象局兰州干旱气象研究所目前使用的GRAPES-SDM沙尘暴预报业务模式在2008—2011年春季沙尘天气预报情况进行检验评估。结果表明：①自2008年以来，GRAPES-SDM沙尘暴模式对中国北方区域沙尘天气的模拟预报能力较好，TS评分和预报效率保持较高的水平；②模式对内蒙古地区、河套地区及甘肃河西地区的预报效果最好，但常有空报或预报沙尘强度偏强的现象；模式能预报出南疆盆地的沙尘天气，但常有预报范围偏小、强度偏弱的现象；对青海地区的沙尘天气常有漏报现象；③模式对沙尘暴频发地区的预报效果较好，对沙尘天气偶发地区容易漏报，模式对新疆东部、内蒙古中西部地区空报较多；④模式对大范围沙尘天气过程的预报能力较好，对零星沙尘天气预报能力较差。通过检验，我们还提出了改进和完善GRAPES-SDM沙尘暴预报系统的一些建议。

By comparing the estimated forecasts by the GRAPES-SDM model and the observed data on sandstorm weather in North of China in the springs of 2008-2011, we tested the forecast product with synoptic verification method. The results showed that: (1) The forecasting ability of the GRAPES-SDM model was accurate in North of China in the spring of 2008, with high TS scores and prediction efficiency. (2) The forecasting effect in the Hetao region of Inner Mongolia, Hexi corridor of Gansu province was the best in spite of some failure prediction or over-intensified prediction happened, the model could forecast dust weathers in South Xinjiang Basin with less narrow belt range and less intensity than that observed, but often failed to predict the sandstorms in Qinghai. (3) The successful forecasts were in relation to those areas with more sandstorms, while the missing predictions were usually limited to those areas with occasional sandstorms, the failure prediction often appeared in eastern Xinjiang and middle-west Inner Mongolia. (4) The forecast ability of the GRAPES-SDM model performed well in predicting those dust weathers with large range, but poor in those scattered dust weathers. We also proposed some suggestions on improving the GRAPES-SDM model in regard to its accuracy.