Abstract: Based on the observation data of automatic weather stations in Guangyuan, the refined grid data of the European Centre for Medium-Range Weather Forecasts (ECMWF), 5 km grid meteorological element forecasting products (national grid guidance and forecasting SCMOC, provincial correction SPCC and provincial guidance SPCO), three interpolation methods, such as nearest neighbor method, linear interpolation and inverse distance weight, were used to test the applicability of precipitation and temperature products in Guangyuan area. The results show that for the stations at an altitude of 0~1000 m, the SPCC is the best for sunny and rainy forecast, the EC model is the best for medium to heavy rain forecast, the SPCC is the best for heavy rain forecast, and the SCMOC and EC models are the best for the forecast of heavy rain and above. For stations at an altitude above 1000 m, SPCC and SCMOC are the best for sunny and rainy forecast, EC model is the best for medium to heavy rain forecast, SCMOC and SPCC are the best for rthe forecast of rainstorm and above. The accuracy of sunny and rainy forecast at an altitude of 0~1000 m is between 62% and 68%, and the hit rate at an altitude of above 1000 m is lower than that at an altitude of 0~1000 m. On the whole, the daily minimum temperature forecast of all kinds of products is better than the daily maximum temperature forecast, and the daily maximum temperature forecast is lower than the actual situation. Among them, the forecast effect of SPCO and SPCC is better than that of EC model and SCMOC, especially the EC model is the most low, and the reference is relatively poor.