随机参数扰动方法对中国冬季降水集合预报的影响

Influence of Stochastically Perturbed Parameterization on Ensemble Forecasting of Winter Precipitation in China

降水数值预报有很大的不确定性,与降水预报密切相关的物理过程参数化方案中关键参数的不确定性是降水数值预报误差来源之一,对这些参数引入随机扰动的随机参数扰动方法(StochasticallyPerturbed Parameterization,简称SPP方法)可以代表模式降水预报的不确定性,是国际集合预报前沿研究领域。为了认识该方法能否代表中国冬季降水数值预报的不确定性,为业务应用提供科学依据,基于中国气象局中尺度区域集合预报模式(Global/Regional Assimilation and Prediction System-Regional Ensemble Prediciton System,简称GRAPESREPS),从对模式降水预报不确定性有较大影响的积云对流、云微物理、边界层及近地面层等四个参数化方案中选取了16个与降水密切相关的关键参数,引入了随机参数扰动方法,并通过2018年12月12日至2019年1月12日总计31天的冬季集合预报试验,对比分析了SPP方法对等压面要素及降水的集合预报效果。结果显示:在冬季应用SPP方法时,等压面要素的概率预报技巧总体来说优于无SPP方法扰动的对比试验,且对于低层、近地面要素的改进效果优于对中高层等压面要素的改进;但对降水概率预报而言,尽管检验评分数值略优于对比预报试验,但并未通过显著性检验,这表明,在东亚冬季风影响下,随机参数扰动方法对中国冬季降水概率预报技巧没有明显的改进。究其原因,可能是由于SPP方法主要代表对流性降水预报的不确定性,而中国冬季降水过程主要与斜压不稳定发生发展有关,模式降水以大尺度格点降水为主,对流性降水较少,故对冬季降水预报改进不明显,这为业务集合预报模式中应用随机参数扰动方法提供了科学依据。

Precipitation ensemble forecasting is characterized by great uncertainty, and the uncertainty of the parameters in the physical that is closely related to the precipitation forecast is one of the sources of its numerical prediction error. As a frontier research field in international ensemble forecasting, the stochastically perturbed parameterization(SPP) method has been developed to address the uncertainty of representative model precipitation forecasts. To determine whether this method can reflect the uncertainty of numerical predictions of winter precipitation in China and provide a scientific basis for business applications, we used the China Meteorological Administration’s Global/Regional Assimilation and Prediction System(GRAPES) mesoscale regional ensemble prediction model and selected 16 key parameters from four parameterization schemes. These parameters, e.g., cumulus convection, cloud microphysics, boundary layer, and nearsurface layer, greatly influence the uncertainty of model precipitation forecasts. In this paper, we introduce the stochastically perturbed parameterization(SPP) method and describe the results of an ensemble prediction experiment conducted from December 12, 2018 to January 12, 2019, a total of 31 days. We compare and analyze the effect of the SPP method on the winter weather situation and precipitation ensemble prediction. The results show that with the addition of a test for the SPP method, the results of probability prediction techniques for precipitation and isobaric elements are better than the control predictions without the SPP method, and the improvement of low-level and near-surface elements is better that of the iso-surface elements in the middle or upper floors. The precipitation prediction results obtained superior scores to those of the control prediction test, but because the improvement did not pass the test of significance, the differences were not statistically significant. The above results indicate that under the influence of the East Asian winter monsoon, the SPP method demonstrates no obvious improvement on the current prediction technique used for winter precipitation in China. The reason for this may be that the SPP method mainly represents the uncertainty of convective precipitation forecasting, whereas the winter precipitation process in China is mainly one characterized by the development of baroclinic instability. Because model precipitation is dominated by large-scale grid precipitation, and less convective precipitation, improvement in the winter precipitation forecast is not obvious. Thus, there is a scientific basis for applying the SPP method to the operation ensemble forecasting model.