福建省ECMWF细网格模式2 m温度预报误差分析与订正

Analysis and Correction of 2 m Temperature Forecast Error ofECMWF Fine Grid Model in Fujian Province

选取2022年1月1日—12月31日ECMWF细网格模式2 m温度预报24 h以内预报时效产品和对应时次的福建省70个国家站观测资料进行分析,采用ARIMA(差分自回归移动平均)模型和双权重ARIMA模型分别对2 m温度预报产品进行偏差订正,并对订正前后的结果进行对比分析。结果表明:1)ECMWF模式2 m温度预报在福建省主要呈现冷偏差,随着预报时效的增加,均方根误差和准确率随之变差;分别用两种模型进行订正,平均绝对误差由2.1℃以内减小到1.6℃以内,均方根误差从2.5℃以内降低到2.1℃以内,且偏差越大,订正效果越明显。2)ECMWF模式2 m温度逐月预报效果差异较大,订正后各评价指标均有显著改进,各月平均误差在-0.5—0.5℃。3)ECMWF模式2 m温度预报偏差主要表现为福建东部沿海小、中西部较大;订正后平均绝对误差和均方根误差减小至2℃以内,且对高海拔地区的站点改善效果更加明显。与ARIMA模型相比,双权重ARIMA模型订正后平均绝对误差与均方根误差更小、准确率更高,订正效果更好。

Based on the ECMWF fine grid model 2 m temperature prediction within 24 hours of forecast time products and corresponding observation data from 70 national stations in Fujian Province,the deviation of 2 m temperature prediction product was corrected by the ARIMA(Differential Autoregressive Moving Average)model and the dual weight ARIMA model,and the results before and after the correction were compared and analyzed.The results showed that:1)2 m temperature prediction of ECMWF model was mainly cold deviation in Fujian Province.With the increase of prediction time,the root mean square error and accuracy became worse.Corrected by two models,the mean absolute error decreased from less than 2.1℃to less than 1.6℃,and the root mean square error decreased from less than 2.5℃to less than 2.1℃.2)There was a significant difference in the monthly 2 m temperature prediction effect of ECMWF model.All evaluation indicators were significantly improved after revision,with the monthly mean error of-0.5-0.5℃.3)The deviation of 2 m temperature forecast of ECMWF model was mainly small in the east coast and large in the central and western regions in Fujian.After the correction of two models,the mean absolute error and root mean square error can be reduced within 2℃,the bias correction was more effective in the high altitude area.4)Compared with ARIMA,the bi-weight ARIMA presented better performance,with smaller mean absolute error and root mean square error and higher accuracy.