中期数值天气预报的集合预报试验

The Experiment of Ensemble Forecasting about Numerical Medium-Range Forecasting

利用中国科学院大气物理研究所研制的Ｔ４２Ｌ９ 谱模式， 在中期数值预报领域中引入集合预报的概念和方法， 初始扰动场取为Ｔ４２Ｌ９ 谱模式２４ 小时预报误差的平均均方差乘随机数， 再综合利用蒙特卡洛预报（ ＭＣＦ） 和落后平均预报（ＬＡＦ） 两种方法作集合预报试验， 试验结果表明： 各成员预报的等权平均或不等权平均的集合预报明显优于单一的控制预报； 不等权平均与等权平均的集合预报结果相比较， 不等权平均的集合预报优势较明显； 在不等权平均的集合预报中， 区域性不等权平均又比全球性不等权平均的预报稍好。

The conception and method of ensemble forecasting is introduced the field of numerical medium-range forecasting by using the global spectral model T42L9 developed by IAP. The first day′s forecasting error of the model T42L9 times random number is regarded as the initial error′s disturbance field. Then, the experiment of ensemble forecasting is made by using the methods of Monte Carlo forecasting and lagged average forecasting comprehensively. The results show that the ensemble forecasting of the equivalent and differential weight average, made by each ensemble member, is better distinctly than only control forecasting, and in the ensemble forecasting of differential weight average, the regional differential weight average forecasting is better than the global differential weight average.