Abstract:

Based on the observation data from 65 stations in Liaoning province, the accuracy of the prediction products in different seasons was evaluated using the 2-meter temperature products within 12-36 hours from 2016 to 2018 predicted by the EC (ECMWF) model.The effect of the fixed error correction and optimal sliding period correction methods on improving the accuracy was analyzed.The results show that the accuracy of temperature in Liaoning province predicted by the EC model is as follows:that of the maximum temperature is the best in winter with the accuracy in the urban sites of 81.5%, and that of the minimum temperature is the best in summer with the accuracy in the urban sites of 84.3%.After the adoption of the optimal sliding cycle correction, the accuracy of the maximum and minimum temperature in Liaoning province from 2016 to 2018 is improved 19.7% and 20.5%, respectively, compared with the EC model.The prediction accuracy in the minimum temperature is higher than that in the maximum temperature.In whole spatial distribution, the prediction accuracy of the EC model for the maximum or minimum temperature in the central plain of Liaoning province is higher than that in the east and west parts, especially in the northeast, southwest, and southeast of Liaoning province.The accuracy in the areas affected by Changbai Mountain is significantly lower than that in other regions.At the same time, the correction prediction ability of the maximum and minimum temperature in summer is better than that in other seasons.Under the rainy and sunshine weather conditions, the correction test in temperature forecasted in Liaoning province is carried out.It is concluded that the test results have some supplementary effect on the correction of the maximum and minimum temperature in Liaoning province, especially when precipitation occurs in winter, the supplementary correction effect for the maximum temperature forecast is most significant.

Key words: Maximum or minimum temperature, Error correction, Seasonal integration, ECMWF model