Abstract: Based on the observed data and the data forecasted by ECMWF (European Centre for Medium-Range Weather Forecasts) fine grid model at 15 meteorological stations in the coastal areas of Shandong Peninsula from 2017 to 2019, 2-meter air temperature on the ground is verified by using daily forecast accuracy and monthly average error methods. The forecast effectiveness of ECMWF model is verified respectively for the processes of cold air, cold-flow snow and sea fog in Weihai region. The results show that when the error of forecast accuracy is less than 2 ℃ within 24 h time limit the forecast accuracy is 60%-95% for low temperature and 50%-89% for high temperature. In the low-temperture forecasts the monthly average forecast error is smaller than 1 ℃ in the northern and southern coastal areas from March to September. The ratio of positive to negative errors is comparable fot the high temperature forecast in the cold-air process. The positive error of the low temperature forecast has a larger proportion, being up to 96% in maximum. In the cold-flow snow process the forecasts of high and low air temperature are generally dominated with positive errors. In the coastal areas where heavy sea fog occurs often, the low air temperature forecasted by the ECMWF model is relatively higher than that observed actually. The results from the ECMWF model can provide a good reference for the low temperature forcast from March to September.