随机物理倾向扰动在风暴尺度集合预报中的影响研究

Impact of stochastically perturbed parameterization tendencies on storm-scale ensemble forecast

为深入探究随机物理倾向扰动(Stochastically Perturbed Parameterization Tendencies,SPPT)方案在风暴尺度集合预报中的影响,基于WRF模式利用FNL资料对SPPT方案中的3个参量分别进行敏感性试验,得到SPPT方案的最佳参数配置,并在此基础上分析SPPT方案模拟的降水分布特征。结果表明:SPPT方案敏感性试验中,去相关时间选择6 h时构造的集合成员可信度更高,逐时降水评分效果在积分中后期较高,对于暴雨及以上量级的评分技巧最优;造成降水主要天气系统的维持时间对该变量的选取有较大的影响。去相关空间尺度选择100 km的集合试验更为可靠,对降水预报技巧较高;同时该变量的选取与天气过程中的大尺度信息、中小尺度系统的活跃以及模式的空间分辨率有密切联系。通过对离散度和离群值分析认为扰动振幅选择0.525最为合理。SPPT方案集合成员在局部地区可以较大幅度地改变降水量,对降水落区的准确模拟存在一定的局限性。

The present study further explores the influence of Stochastically Perturbed Parameterization Tendencies （SPPT） scheme on the storm-scale ensemble forecast. Three sensitivity experiments are performed using the Weather Research Forecast （WRF） model to investigate impacts of three parameters in the SPPT scheme. The NCEP FNI. analysis product is used to pro vide initial and boundary conditions for WRF. Optimal parameter configuration of the SPPT scheme is obtained. Precipitation distribution characteristics simulated with the SPPT scheme are analyzed. Results show that in the SPPT scheme sensitivity ex periments, when the decorrelation time scale is 6 h, simulations of the ensemble members are the most reliable and the hourly precipitation score is the best in the middle and later periods of integration. The precipitation scores are also the best for simulations of rainstorms and big rainstorms. The duration of the weather system that can lead to precipitation has great influences on the selection of decorrelation time scale. The experiment that chooses 100 km as the decorrelation spatial scale performs best and the precipitation forecasting skill scores are also the highest. At the same time, the selection of this variable is closely related to large scale information and active small and medium scale systems as well as the model spatial resolution. Analysis of spread and outliers indicates that choosing 0. 525 as the disturbance amplitude is the most reasonable. The ensemble members of SPPT can significantly improve precipitation simulation in some local areas, hut there still exist some limitations on accurate simulation of the precipitation area.