基于增长模繁殖法的集合预报试验  

ENSEMBLE FORECASTING EXPERIMENTS BASED ON THE BGM METHOD

作　　者：关吉平[1] 张立凤[1] 

机构地区：[1]解放军理工大学气象学院,南京211101

出　　处：《气象科学》2006年第3期272-278,共7页Journal of the Meteorological Sciences

基　　金：国家重点基础研究发展规划项目(编号:2004CB18300);中国科学院大气物理研究所大气科学和地球流体力学数值模拟国家重点实验室开放课题(编号:2004-01-01)共同资助

摘　　要：利用微机版本的T63L9全球谱模式，首先进行了用增长模繁殖法（BGM）制作集合预报的方案设计，然后用确定的方案进行了中期数值预报的集合预报试验。繁殖循环的初始扰动为T63L9谱模式各层24h预报的均方根误差乘随机数，该随机数服从[-1，＋1]区间上的均匀分布，繁殖循环每12h 1次，繁殖循环的周期为3d，集合成员为9个。结果表明：在1～8d的预报中，集合预报可以比控制预报的时效延长1d，集合预报无论是距平相关系数还是预报均方根误差都明显优于控制预报，集合预报对形势场的预报较控制预报更为接近实况分析场。Firstly, a scheme of ensemble forecasting using the method of breeding of growing mode （BGM） is designed, using the global spectral model T63L9 based on microcomputer. Secondly, some experiments of ensemble forecast about medium-range numerical forecasting are done using the scheme. The first day＇s forecasting error of the model times random number is regarded as the initial disturbance of the breeding cycles and these random numbers meet uniform distribution between 1 and ＋1. The breeding period is 12h, so twice one day, the total time of breeding is 3 days and the number of ensemble members is 9. Ensemble forecast is made by the equivalent weight average with 9 members. Results show that from the 1st day to 8th day＇s forecasting, the forecast term from ensemble forecasting is one day longer than that from control forecast. Within 8 days, whether for anomaly correlation coefficient or forecasting rmse error, the ensemble forecasting is better than control forecasting and the posture field from ensemble forecasting is more similar to analysis field than that from control forecast.

关 键 词：中期数值预报 集合预报 增长模繁殖法 

分 类 号：P456.7[天文地球—大气科学及气象学]