GRAPES-EPS系统的初值生成方法与对比试验研究

介绍了增长模繁殖法（BGM）和集合转化卡尔曼滤波法（ETKF）两种不同的模式初始扰动方法的基本原理，并以GRAPES中尺度模式为基础，利用两种初始扰动方法构建了两套中尺度集合预报系统。通过圣帕台风的个例试验，对比两种初始扰动生成方法对降水预报结果的影响。结果表明：两种方法均可以很好的捕捉到中尺度强降水的过程信息，集合平均结果优于控制预报，并在一定程度上改善了对强降水的落区和强度的预报；从邮票图和对集合预报系统的检验参数上来看，ETKF的集合离散度和特征值分布好于BGM方法，但对于降水结果TS等的评分的比较上来看，BGM的预报结果要优于ETKF的预报结果；另外，BGM方法和原理更简单，易于实现业务应用。

Two methods of initial perturbation generation and both comparison in the GRAPES-EPS system

The basic principles of two methods of a Breeding of growing mode （BGM） and an ensemble transform Kalman filter （ETKF） for the generation of initial perturbation value were introduced and compared with each oth- er. Based on the GRAPES-Meso model, two mecoscale ensemble forecast systems were established, i. e. a GRAPES-BGM and a GRAPES-ETKF. Using the Sepat typhoon event as a case study, the prediction accuracy of two methods for precipitation was compared. The results indicate that two systems can catch the information of pre- cipitation. The accuracy of ensemble forecast is higher than that of control forecast, especially for falling area and intensity of the heavy rain to some extent. The talagrand distribution of the ETKF scheme is better than that of the BGM scheme in terms of the poststamps and other verification, while the reverse is true in terms of the TS score of precipitation. Furthermore, the BGM scheme is easy and convenient for weather forecast service.