Sandstorm is a serious natural disaster in north China. It is of great significance to carry out relevant research to improve the forecast level of this kind of catastrophic weather. Based on the RegCM-dust model, an extended period numerical prediction analysis of a typical severe sandstorm process in north China is conducted, and the results are compared with NCEP reanalysis data and other analysis results. The results show that the regions with high sediment discharge simulated by the model are mainly located in southern Xinjiang, Mongolia and western Inner Mongolia. The model has a certain forecasting ability for 10 m wind speed, but the simulated wind speed is smaller than the reanalysis data. The changes of dust column content and total sedimentation simulated by the model can reflect the characteristics of the dust storm weather process. The simulated sand-dust mixing ratio has a certain correspondence with the urban pollution index, which indicates that the model has certain forecasting ability for the pollution weather caused by sand-dust.

Based on 2-meter temperature data from the second generation monthly dynamic extended range forecast (DERF2.0) model, observational temperature data at 69 weather stations in Gansu Province, reanalysis data of NCEP/DOE and sea surface temperature data of NOAA，the forecast errors of January temperature in Gansu Province by DERF2.0 model from 1992 to 2013 and their relationship with external forcing were analyzed. The results are as follows: (1) The simulated effects of January temperature by DERF2.0 model in eastern Yellow River of Gansu (known as Hedong for short) were better than that in most regions of western Yellow River of Gansu (known as Hexi for short), especially  in Gannan, Linxia, Lanzhou, Dingxi, Pingliang and Qingyang, the average errors between forecast and observation were small and stable, and the linear tendency rates of forecasted temperature in January were consistent with observation from 1992 to 2013, while the average errors were bigger and unstable in most regions of Hexi, and the change trends of forecasted temperature were contrary to actual observation. (2) Although the model could well reflect the inter-annual variation and spatial distribution pattern of January temperature in Gansu, the abnormal centers and values of temperature change were significantly different from the observation. (3) The EOF1 of error field reflected consistent overestimate or underestimate to  January temperature, the EOF2 presented an opposite distribution pattern in Hedong and Hexi, while the EOF3 appeared a reverse phase distribution pattern in Gannan Plateau and other parts of Gansu Province. (4) The main modes of forecast error field were significantly correlated with circulation and sea surface temperature (SST) in key areas, which indicated that the response of model to circulation and SST anomalies was deficient. Therefore, it was partially possible to reduce forecast errors of January temperature in Gansu by adjusting the response ability of DERF2.0 model to circulation and SST in key areas.