基于多尺度混合滤波的GRAPES_3Dvar及其在实际暴雨预报中的应用分析

Improvement of GRAPES_3Dvar with a New Multi-Scale Filtering and Its Application in Heavy Rain Forecasting

为提高暴雨的数值模拟和预报效果,在生成初值的同化分析中保证中小尺度信息的引入十分重要。为改善GRAPES(Global/Regional Assimilation and Prediction System)区域三维变分同化系统对中小尺度信息的引入,本文利用气候态背景误差样本估计变量误差的水平协相关尺度,进而利用三个不同水平特征尺度的递归滤波器对统计结果予以拟合,以替代原有单一尺度的递归滤波方案。利用新方案对2015年6月1—2日江汉平原的暴雨过程进行同化和预报,研究结果表明:新方案的功率谱衰减较慢,通过单点试验和分析场分析发现,新同化方案引入了更多的α中尺度信息;在暴雨个例的预报中发现,利用新同化方案后湿度场、散度场和涡度场在分析场和预报场中更加接近实际情况,降水预报技巧明显提高。通过能量谱分析发现,新方案与原方案相比反映出更多的α中尺度信息;新方案对江汉平原地区暴雨预报具有正效果,在α中尺度内出现了低层辐合、高层辐散及湿度增加等有利于降水的要素。在暴雨个例的基础上,本文进行16d的批量试验,结果显示新方案对降水的预报技巧明显提高,与个例结果一致。

To improve the effect of the numerical simulation and forecasting of heavy rains,it is very important to introduce the meso-and small-scale information in the assimilation analysis of initial values.For enhancing the introduction of meso-and small-scale information in the regional GRAPES 3 Dvar system,climatic background error sample was used in this paper to estimate the level covariate correlation scale of variable error,and then statistical result fitting was performed with recursive filter of the feature scales of three different levels,thus replacing the original single-scaled recursive filtering.The new scheme was used to assimilate and forecast the rainstorms in the Jianghan Plain during 1-2 June,2015,and the research results showed that power spectrum attenuation in the new scheme is slower.Through single point test and field analysis,we found that the new assimilation scheme introduces more meso-αscale information.In the report of the rainstorm,it was found that,with the adoption of the new scheme,the moisture field,divergence field and vorticity field are much closer to the observation values when measured in the analysis field and forecasting field.So the precipitation forecasting skill is improved obviously.By analyzing the energy spectrum,it was learned that the new scheme could reflect more meso-αscale information and the new scheme has positive effect on the forecasting of rainstorms in the Jianghan Plain area.In meso-αscale,there are some favorable factors for rainfall,such as lower convergence,upper divergence and increased humidity.Based on individual cases of rainstorm,batch experiments for 16 days were completed,and the result showed that the new scheme could improve precipitation forecasting skill,which is consistent with the results of cases study.