Study on 2m Temperature Bias Correction of EC Model in Sichuan Province

Based on the 2m temperature 168h forecast of ECMWF during the period from January to December 2017 and the observation data of the same period,the systematic deviation characteristics of 2m temperature prediction are analyzed.Then the bi-weight moving mean method with a 15 day sliding period is conducted to the model to acquire a correction field,which is compared with the existing subjective and objective forecast products in Sichuan Meteorological Observatory.The results show that:(1) The prediction accuracy of lowest temperature is much higher than that of highest temperature for EC model,and both of them have been obviously improved after the correction.On average the lowest temperature accuracy has increased by 20.5% while the highest temperature accuracy has increased by 31.2%.The average absolute errors have decreased about 1.1℃ and 2.9℃ respectively.(2) The monthly difference of accuracy for the highest temperature,which has been markedly reduced after correction,is significantly greater than that for the lowest temperature.The accuracy for both the highest and lowest temperature have been significantly increased,while the mean absolute error for all months are within 2℃.(3) The forecasts of the lowest temperature and highest temperature both present negative systematic bias in the whole province,and the systematic bias of the highest temperature is obviously larger than that of the lowest temperature.The systematic biases have been significantly decreased after the correction,maintaining between ± 1℃ in most parts of the province.(4) The comparison with the existing subjective and objective forecast products in Sichuan Meteorological Observatory shows that the highest and lowest temperature correction field has the highest accuracy and the minimum average absolute error.