Comparative Correction of Air Temperature Forecast from ECMWF Model by the Decaying Averaging and the Simple Linear Regression Methods
Article
Figures
Metrics
Preview PDF
Reference
Related
Cited by
Materials
Abstract:
The decaying averaging and the simple linear regression methods were used to correct air temperature forecast in a fine-mesh grid point forecast system of Shaanxi Meteorological Service. Based on the dataset of daily 2 m maximum and minimum air temperature forecasts of 99 national meteorological stations in Shaanxi Province from ECMWF high resolution model in 2017, the abilities of the two methods to correct temperature prediction errors were analyzed and compared. The results showed that the prediction accuracies of daily 2 m maximum and minimum air temperatures are improved significantly by the two methods, whosecorrection abilities are gradually weakened with the extension of prediction time. There is a significant negative correlation between accuracy of the temperature forecast and skill-score of the two methods. The accuracies are all low for the daily 2 m maximum air temperature in Qinling Mountains and the area south to it and daily 2 m minimum air temperature in the north to Qinling Mountains, in which the skill-score is usually more than 40% and its maximum value is even larger than 70%. The systematic deficiencies of daily 2 m maximum and minimum air temperature forecasts are effectively reduced. As a result, the frequency with a smaller error range is increased, while the frequency with a larger error range is decreased. More advantages of the two methods are attained when the absolute errors are less than 2℃ for daily 2 m maximum air temperature forecast and more than 3℃ for daily 2 m minimum air temperature forecast. The ability of the simple linear regression method to correct daily 2 m maximum air temperature forecast is slightly better than that of the decaying averaging method, whose ability to correct daily 2 m minimum air temperature forecast is better than the simple linear regression method. The mixed correction of temperature forecast by the two methods is more effective.