论文

基于GRAPES-MESO模式的非静力三维变分同化方案性能分析

  • 王叶红 ,
  • 赖安伟 ,
  • 林春泽 ,
  • 王明欢
展开
  • 中国气象局武汉暴雨研究所 暴雨监测预警湖北省重点实验室, 湖北 武汉430074

网络出版日期: 2013-06-28

Evaluation on Nonhydrostatic Three-Dimensional Variational Data Assimilation Scheme of GRAPES-MESO Model

Expand

Online published: 2013-06-28

摘要

利用国家气象中心中尺度业务模式GRAPES-MESO V3.0, 以2009年6—8月为例, 用T213模式12 h预报场为大尺度模式背景场, 进行了非静力GRAPES-3DVAR系统同化(3DVAR试验)与不同化(CNTL试验)探空资料的两组数值试验, 并对降水、 高度、 温度、 风场和相对湿度场进行了统计检验和典型个例分析, 探讨了非静力三维变分同化系统的性能。结果表明: (1)除700 hPa相对湿度场外, 3DVAR试验850, 700和500 hPa初始温、 压、 湿和风场均方根误差均比CNTL试验有所增加, 其中3DVAR试验对500 hPa初始温、 压和风场的负贡献最为显著。随着预报时间延长, 3DVAR试验对500 hPa风场预报的负贡献较为明显; (2)对整个模式积分区域平均状况而言, 3DVAR试验对模式0~24 h和12~36 h降水预报中大多数降水量级的TS评分都小于CNTL试验; (3)CNTL试验对小雨、 中雨的预报范围比较适中, 对大雨、 暴雨及大暴雨的预报范围均小于实况, 且随着降水量级的增大, 预报范围较实况偏小的程度加剧; 3DVAR试验则对各量级降水范围预报偏小的程度均大于CNTL试验, 特别是中雨以上强度的降水; (4)CNTL试验较3DVAR试验能更好地模拟雨带的分布、 演变特征和降水强度的变化; (5)CNTL试验较好地再现了长江中下游、 华南、 华北、 东北、 西南地区东部和华东平均降水率的逐日演变及峰值、 谷值模拟, 但强度普遍偏弱;  3DVAR试验模拟的各区域日平均降水率的演变趋势与CNTL试验类似, 但强度弱于CNTL试验, 因而其与实况的差异进一步加大; (6)两个试验与观测均有较好的一致性, CNTL试验的一致性略好于3DVAR试验; 对<1.5 mm·d-1的降水, 两个试验都倾向于过多地预报了降水量; 对>1.5 mm·d-1的降水, 两个试验预报的都偏弱; 对1.5 mm·d-1以下的降水, 两个试验的预报水平相当; 对1.5 mm·d-1以上的降水, CNTL试验的预报优于3DVAR试验; (7)对2009年6—8月逐日降水预报结果进行了考察, 整体而言, CNTL试验能较好地预报出逐日雨带的位置, 强度也与实况基本接近, 对西风带低槽、 台风、 低涡、 切变线和局地降水都有较好的表现。3DVAR试验对逐日雨带的预报大致与CNTL试验类似, 但预报的降水强度普遍弱于CNTL试验。

本文引用格式

王叶红 , 赖安伟 , 林春泽 , 王明欢 . 基于GRAPES-MESO模式的非静力三维变分同化方案性能分析[J]. 高原气象, 2013 , 32(3) : 689 -706 . DOI: 10.7522/j.issn.1000-0534.2012.00065

Abstract

With the mesoscale operational numerical model GRAPES-MESO V3.0 from  National Meteorological Center of China Meteorological Administration and 12 h T213 forecasts as background field, two numerical experiments with (Exp. 3DVAR) and without (Exp. CNTL) assimilating radio-soundings using the nonhydrostatics GRAPES-3DVAR are conducted  from June to August 2009 to investigate the performance of GRAPES-3DVAR. The statistical verifications of precipitation, geopotential height, temperature, wind and relative humidity and typical weather events are conducted. The results indicate that: (1) The root mean square errors of initial temperature, geopotential height, relative humidity and wind field on 850, 700 and 500 hPa in Exp. 3DVAR are morer than that of in Exp. CNTL, except for 700 hPa relative humidity. Exp. 3DVAR has a significant negative contribution to initial temperature, geopotential height and relative humidity on 500 hPa. And Exp. 3DVAR has an obvious negative contribution to 500 hPa wind forecast with forecasting time going on. (2) As to the area mean of whole model domain, TS scores at different orders in 0~24 h and 12~36 h rainfall forecasts in Exp. 3DVAR are mostly less than that of in Exp. CNTL. (3) The coverage of light rain and moderate rain forecasted by Exp. CNTL are close to the observation, while that of heavy rain, torrential rain, heavy torrential rain are less than that of the observation. Moreover, with the increase of rainfall, the changing small degree of forecast range is more than the observation. The forecast area of rainfall at different orders in Exp. 3DVAR are worse than that of in Exp. CNTL, especially for those above moderate rain. (4) The distribution, evolution and intensity variations of rain region in Exp. CNTL are better than that of Exp. 3DVAR. (5) The daily evolution, peak and valley values of the simulated average rainfall rate in the mid-lower reaches of Yangtze River, South China, North China, Northeast China, east of Southwest China and East China, can be almost simulated in Exp. CNTL, but the average rain rate is weaker than the observation. The forecast of Exp. 3DVAR is similar to that of Exp. CNTL, but the rainfall is weaker than in Exp. CNTL. Accordingly, the difference between Exp. 3DVAR and the observation is increased. (6) Rainfall forecasts in Exp. CNTL and 3DVAR are both in a good consistency with the observation, with the former is a little better than the latter. For rainfall more (lower) than 1.5 mm·d-1, the two experiments tend to forecast more (less) rainfall than the observation, respectively. The forecasts of two experiments are nearly in a same level for rainfall lower than 1.5 mm·d-1. However, Exp. CNTL gives a better forecast than Exp. 3DVAR for rainfall above 1\^5 mm·d-1. (7) In the period of June to August 2009, the daily rain band location and rainfall intensity in Exp. CNTL are well forecasted and nearly close to the observation. This experiment gives a good simulation for the rainfall produced by different weather systems, such as westerly trough, typhoon, low vortex, wind shear and local system. The results of Exp. 3DVAR are roughly similar to that of Exp. CNTL, but its simulated rainfall intensity is generally weaker than that in Exp. CNTL.

参考文献

[1]Parrish D F, Derber J C. The national meteorological center’s spectral statistical interpolation analysis system[J]. Mon Wea Rev, 1992, 120: 1747-1763.
[2]Courtier P, Andersson E, Heckley W, et al. The ECMWF implementation of three-dimensional variational assimilation (3D-Var): Formulation[J]. Quart J Roy Meteor Soc, 1998, 124: 1783-1807.
[3]Lorence A C, Ballard S P, Bell R S, et al. The met office global 3-dimensional variational data assimilation[J]. Quart J Roy Meteor Soc, 2000, 126: 2991-3012.
[4]朱宗申, 汪厚君, 张跃堂. 国家气象中心准业务有限区域客观分析方案[J]. 应用气象学报, 1992, 3(4): 459-467.
[5]庄世宇, 薛纪善, 朱国富, 等. GRAPES全球三维变分同化系统——基本设计方案与理想试验[J]. 大气科学, 2005, 29(6): 872-884.
[6]张华, 薛纪善, 庄世宇, 等. GRAPeS三维变分同化系统的理想试验[J]. 气象学报, 2004, 62(1): 31-41.
[7]曾智华, 马雷鸣, 梁旭东, 等. MM5数值预报引入GRAPES三维变分同化技术在上海地区的预报和检验[J]. 应用气象学报, 2004, 15(5): 534-542.
[8]万齐林, 薛纪善, 陈子通, 等. 雷达TREC风的三维变分同化应用与试验[J]. 热带气象学报, 2005, 21(5): 449-457.
[9]Wang Yehong, Zhao Yuchun, Cui Chunguang. Numerical research on effects upon precipitation forecast of Doppler-radar estimated precipitation and retrieved wind field under different model initial schemes [J]. Acta Meteor Sinica, 2007, 21(1): 98-113.
[10]张万诚, 肖子牛, 郑建萌. 三维变分同化在一η坐标中尺度模式预报中的应用[J]. 高原气象, 2005, 24(1): 70-76.
[11]陈德辉, 沈学顺. 新一代数值预报系统GRAPES研究进展[J]. 应用气象学报, 2006, 17(6): 773-777.
[12]Zhang Hua, Xue Ji Shan, Zhu Guo Fu, et al. Application of direct assimilation of ATOVS microwave radiances to typhoon track prediction[J]. Adv Atmos Sci, 2004, 21(2): 283-290.
[13]谭晓伟, 陈德辉, 张庆红. 一种新型的适应性观测资料应用对台风预报的影响试验研究[J]. 热带气象学报, 2006, 22(1): 18-25.
[14]梁科, 万齐林, 丁伟钰, 等. 飞机报资料在0506华南致灾暴雨过程模拟中的应用[J]. 热带气象学报, 2007, 23(4): 313-325.
[15]王叶红, 赵玉春, 张兵, 等. “7·8”特大暴雨过程雷达风场资料的三维变分同化[J]. 气象科学, 2008, 28(3): 301-307.
[16]朱立娟, 龚建东, 李泽椿. 资料同化中雷达VAD风质量控制研究[J]. 高原气象, 2012, 31(6): 1731-1738.
[17]李华宏, 曹杰, 杞明辉, 等. 雷达风廓线反演在云南强降水预报中的应用[J]. 高原气象, 2012, 31(6): 1739-1745.
[18]冯文, 万齐林, 陈子通, 等. 逐时云迹风资料同化对暴雨预报的模拟试验[J]. 气象学报, 2008, 66(4): 500-512.
[19]丁杨, 庄世宇, 顾建峰. GRAPES地面风场同化方案研究[J]. 热带气象学报, 2008, 24(6): 629-640.
[20]李媛, 董佩明, 刘艳, 等. 多普勒雷达基本观测量直接同化对模式物理量场的影响[J]. 气象与减灾研究, 2009, 32(3): 19-24.
[21]王遂缠, 胡向军, 张新荣, 等. 雷达资料同化在甘肃局地暴雨天气个例中的应用[J]. 高原气象, 2011, 30(3): 711-718.
[22]唐细坝, 薛纪善. COSMIC资料在GRAPES全球三维变分同化系统的初步研究[J]. 热带气象学报, 2009, 25(5): 521-531.
[23]任强, 董佩明, 薛纪善. 台风数值预报中受云影响微波卫星资料的同化试验[J]. 应用气象学报, 2009, 20(2): 137-146.
[24]丁伟钰, 万齐林, 张诚忠, 等. 有云环境下MODIS亮温资料的变分同化I——方案设计与理想试验[J]. 热带气象学报, 2009, 25(6): 649-659.
[25]郝民, 徐枝芳, 陶士伟, 等. GRAPES RUC系统模拟研究及应用试验[J]. 高原气象, 2011, 30(6): 1573-1583.
[26]薛纪善, 陈德辉, 等. 数值预报系统GRAPES的科学设计与应用[M]. 北京: 科学出版社, 2008.
[27]叶成志, 欧阳里程, 李象玉, 等. GRAPES中尺度模式对2005年长江流域重大灾害性降水天气过程预报性能的检验分析[J]. 热带气象学报, 2006, 22(4): 393-399.
[28]徐双柱, 张兵, 谌伟. GRAPES模式对长江流域天气预报的检验分析[J]. 气象, 2007, 33(11): 65-71.
[29]张兵, 钟敏. GRAPES模式对湖北省汛期强降水预报的分类检验分析[J]. 暴雨灾害, 2009, 28(2): 137-142.
[30]彭新东, 常燕, 王式功. GRAPES模式对2008年两次强降水过程的数值预报检验[J]. 高原气象, 2010, 29(2): 321-330.
[31]王雨, 李莉. GRAPES_Meso V3.0模式预报效果检验[J]. 应用气象学报, 2010, 21(5): 524-534.
[32]马旭林, 庄照荣, 薛纪善, 等. GRAPES非静力数值预报模式的三维变分资料同化系统的发展[J]. 气象学报, 2009, 67(1): 50-60.
[33]王雨. 2002年主汛期国家气象中心主客观降水预报对比检验[J]. 气象, 2003, 29(5): 21-25.
[34]王雨, 闫之辉. 降水检验方案变化对降水检验评估效果的影响分析[J]. 气象, 2007, 33(12): 53-61.
[35]王叶红, 赵玉春, 张兵. 改进的AREMS对2003年汛期降水预报及对比检验[J]. 热带气象学报, 2005, 21(6): 597-604.
[36]公颖. 2007年汛期AREM模式降水预报效果检验分析[J]. 暴雨灾害, 2007, 26(4): 372-380.
[37]廖移山, 闵爱荣, 杨荆安, 等. 2009年4-9月我国暴雨天气概述及重要过程浅析[J]. 暴雨灾害, 2010, 29(1): 96-103.
[38]王叶红, 赖安伟, 赵玉春. 降水资料同化在梅雨锋特大暴雨个例模拟中的应用研究[J]. 气象学报, 2012, 70(3): 402-417.
文章导航

/