基于MET检验的青海大范围降水个例参数化方案优选

Parameterization Schemes Optimization of a Large Scale Precipitation Case in Qinghai Based on MET

利用WRF模式不同积云对流和微物理过程参数化方案对2015年8月1—3日青海省大范围降水过程进行了模拟,并利用MET(数值模式评估系统)对本次模拟结果进行了检验,结果表明:(1)此次过程,模式存在“早报”现象,尽管大部分方案预报降水量均偏多,但从降水过程中心位置和强度来看,BMJ(积云对流)/Thompson(微物理)组合方案模拟效果较为理想,Grell/WSM5和KF/Kessler次之;Grell/WSM5从降水极值点的空间分布特征和降水量级上模拟结果较为理想。(2)就空报率和漏报率的空间分布而言,空报率远高于漏报率,空报率高值区主要分布在玉树南部和海西东北部,模式对玉树地区的预报存在较大的误差。(3)用SEDS(对称极值依赖评分)评估极端降水,Grell/WSM5对极端降水的预报效果较好。(4)主观判断降水空间分布特征无法量化预报的质量,且单一的评分指数因侧重点不同无法综合评价模式的预报能力,故选用8种评分或指标进行定量评估,其结果表明,WSM5方案(微物理)的R(相关系数)较大且RMSE(均方根误差)较小,Grell/WSM5对此次降水预报效果最好,Grell/Kessler次之。综合来看,对此次降水个例Grell/WSM5组合方案为最优组合方案,参数化方案的优选有利于客观把握模式的预报能力与预报技巧,有利于提高区域数值模式在青海高原的适用性。

A large scale precipitation case in Qinghai province from August 1st to August 3rd,2015 were simulated by using different cumulus convection and microphysical process parameterization schemes in WRF.The forecast results were evaluated by Model Evaluate Tool(MET).The results showed:(1)There was a wide range phenomenon of premature forecast in the model,most of the combination schemes forecasted more precipitation,but as far as central position and total precipitation were concerned,the simulation effect of BMJ(cumulus convection)/Thompson(microphysical process)combination scheme was more ideal,Grell/WSM5 and KF/Kessler came second.Grell/WSM5 was ideal for simulating the extreme point position and amount of precipitation.(2)Judging from the spatial distribution of false alarm rate and miss rate,the false alarm rate was higher than miss rate,the area of higher false alarm rate located at the south of Yushu and northeast of Haixi.There were great errors in the prediction of precipitation of Yushu in the model.(3)Extreme precipitation was evaluated by symmetric extreme dependency score(SEDS),and scheme of Grell/WSM5 performed well.(4)The subjective judgment of spatial distribution cannot quantify the quality of prediction,and the single scoring index cannot comprehensively evaluate the model's forecasting ability because of different emphasis,so 8 indices or scores were selected to quantitatively assessment the simulation results,WSM5 scheme had a higher correlation coefficient and a lower RMSE(Root mean square error).Comprehensively considering 8 indices,Grell/Kessler were obviously higher than the other combination schemes,Grell/Kessler came second.Generally speaking,the Grell/WSM5 combination scheme was the best for this precipitation case.Integrating multiple scores or indices was beneficial to objectively grasp the forecasting ability and forecasting skills of the model,and evaluate the applicability of the parameterization scheme of the study area.