NCEP/CFS模式对东亚夏季延伸预报的检验评估  

Assessing the Extended Range Forecast Error of NCEP/CFS in the Summer of East Asia

作　　者：陈官军[1,2] 魏凤英[2] 巩远发[1] 

机构地区：[1]成都信息工程学院,成都610225 [2]中国气象科学研究院灾害天气国家重点实验室,北京100081

出　　处：《应用气象学报》2010年第6期659-670,共12页Journal of Applied Meteorological Science

基　　金：国家科技支撑计划项目(2009BAC51B04);灾害天气国家重点实验室自主研究课题(2009LASWZF01)共同资助

摘　　要：利用NCEP的气候预报系统(Climate Forecast System,CFS)所提供的1981 2004年历史回报试验结果,检验和评估了该系统对夏季东亚地区大气环流的预报技巧和系统误差;在此基础上通过提取模式预报和观测的10～20 d及30～60 d低频振荡分量,重点对我国南方3次典型持续性暴雨过程的预报技巧进行检验和诊断分析。结果表明:CFS系统对东亚整体大气环流逐日预报的可靠时效为5 d左右,60°N以北的对流层中高层高度场预报系统性偏低,而在40°～60°N则为系统性偏高。系统性误差随预报时间的延长而增加,但10 d以上预报的系统性误差大小和空间分布逐渐趋于稳定;CFS系统对低频分量的延伸期顶报技巧好于对其整体大气环流的预报技巧,并且在典型持续性暴雨过程中,CFS系统对影响强降水过程的主要环流系统低频振荡特征有一定预报能力。Regional and persistent heavy rain in the southern parts of China can lead to severe flooding,and abnormality of key circulation system in East Asia is the most important inducement for the abnormal precipitation in target area.Furthermore,the low frequency oscillation（LFO） has special impact on the extended range forecast of general circulation.By use of fully coupled retrospective forecasts covering a 24-year period （1981—2004） provided by NCEP Climate Forecast System（CFS）,the 1 to 30 days predictive skill and errors between forecast and observations（NCEP/DOE Atmospheric Model Intercomparison Project （AMIP）ⅡReanalysis） are diagnosed and the predictive skill of three typical cases of persistent heavy rain,in the Yangtze-Huaihe River Valley,the southern parts to the Yangtze River and South China are examined respectively.The influence of LFO on extended range forecast skill is revealed. The forecast skill of 200 hPa,500 hPa and 850 hPa heights is obtained,by calculating the anomaly correlation coefficient（ACC） between predicted and observed fields over a typical region of the East Asia during the summer.Based on the 24-year climatology,the ACC skill maintains greater than 0.6 at a lead time of less than 6 days,which is the average level of current numerical weather prediction models.The ACC drops rapidly with increasing lead time but when it comes to 10—19 days,the ACC is relatively stable and ranges from 0.05 to 0.15.The errors also keep stable at the lead time of more than 10 days with the predicted fields lower than the observations in areas north of 60°N and higher than the observations in areas between 40°N and 60°N. By use of 30—60 days component data which is obtained by Butterworth Band-pass Filter,the forecast exhibits more skillful with the ACC to observed 30—60 days component data of about 0.15 at lead time of 15 days. From the typical cases,it＇s found that the low frequency precipitation with quasi two weeks and 30—60 days fluctuation is an important characteri

关 键 词：气候预报系统 历史回报 稳定误差 诊断 低频振荡 

分 类 号：P459.9[天文地球—大气科学及气象学]