多普勒天气雷达资料的反演及其在中尺度模式中的四维同化试验
|
摘要
数值天气预报是现代天气预报的重要手段，中尺度模式MM5在很多气象台得到了广泛应用，随着我国多普勒天气雷达的普及，如何在数值预报模式中使用雷达资料以提高对中小尺度天气现象的预报准确率是摆在广大气象工作者面前的一个重要课题。四维变分同化方法作为一种提高数值天气预报的有效方法受到国内外专家的广泛关注，其重要特点是一个时次的数据可以影响到它以前时间的分析结果，因此不但可以为数值预报模式提供初始场，而且可以对观测资料不足的地区加以弥补。本文的工作就是多普勒雷达资料在MM5模式中的四维变分同化试验。文中首先在利用模拟风场比较各种雷达资料反演方法的优缺点后，选择变分分析法从雷达资料中反演风场，并根据雷达回波强度得到湿度场。接着介绍了气象资料同化的发展历史及主要方法，并重点强调了四维变分同化理论。然后详细阐述了本文同化试验所用的MM5四维变分同化系统中构造伴随模式的基本方法—伴随码技术以及伴随模式中目标函数的构造，及其权重、尺度因子、下降算法的选取。最后把雷达风场和雷达湿度场同化到MM5四维变分同化系统中。同化试验结果表明：同化空间分辨率很高的雷达风场后，能改善中小尺度降水的预报效果，并且能够得到常规观测资料所不能得到的中小尺度信息，对分析中小尺度系统结构具有重要意义；同化雷达湿度场效果不明显，可能与降水类型和同化时刻有关；而直接在初始时刻加入雷达湿度场，补充了常规资料在反映中小尺度系统方面的不足，增强了初始场中的水汽，有利于降水量的增加；同时同化雷达风场和雷达湿度场的试验表明，水汽的输送和局地的水汽辐合对于产生特大暴雨的贡献远大于仅有高湿中心的贡献。本文的试验结果还表明：要想提高客观定量的模式降水预报，仅有一个成熟的中尺度模式是不够的，资料的时空精度至关重要，而多普勒天气雷达资料在中尺度模式中的四维变分同化是解决资料时空精度问题的有效手段。
Numerical weather prediction is an important means in present-day weather prediction. Mesoscale model MM5 has been used widely in many observatories. Along with a large number of Doppler radar stations have been built in our country, it is a crucial problem for the meteorologists how to use Doppler radar data in numerical weather prediction model. As an efficient approach to improve the accuracy of numerical weather prediction, Four-dimensional Variational Assimilation is widely focused on by the domestic and overseas experts. One of its important characteristics is that the data in one time can influence previousanalysis results, which can not only provide the optimized initial condition but also make up observation datum's absence in some areas . This paper does the experiments of Doppler radar's Four-dimensional Variational Assimilation in MM5 model. Firstly after comparing advantages and disadvantages of several methods by using the data of simulative wind, the paper chooses variational analysis method t
o retrive three-dimensional wind field. Also, the humility field is obtained from radar echo intensity. Secondly meteorologic datum assimilation history and major ways are introduced. Especially four-dimensional Variational Assimilation is emphasized. Thirdly we expound the basic methods how to construct the adjoint model including adjoint code technique, objective function formation, and how to adopt the weighting coefficient, scaling factor, descent arithmetic. Lastly radar wind field and radar humility field are assimilated in the MM5 4D Variational Assimilation System. The assimilation experimentation results indicate that after assimilating radar wind field of small spatial scale, mesoscale and small-scale precipitation prediction can be improved and mesoscale and small-scale information which can't
appear by tradition datum can be gained, which is valuable to analyze the mesoscale and small-scale system structure ; the effect assimilating radar humidity field isn't obvious, which is perhaps correlation with precipitation types and assimilating time. The results also show that adding radar humidity field to initial condition at initial time can supply the gap of the regular data in reflecting the mesoscale and small-scale systems, strengthen the humidity in the initial field, and eventually help to improve precipitation. The experiment of assimilating radar wind field and radar humility field at the same time shows that vapor transportation and local vapor divergence play more significant role in causing excessively heavy rain than only high wet center. The paper's experiments also reveal that it is far beyond the need of improving objective and quantitative precipitation prediction to have only a mature mesoscale model, in effect, the temporal and spatial accuracy of data is crucial, and that radar dat
a's 4D variational assimilation in mesoscale model is an efficient way to solve the problem of the temporal and spatial accuracy of data.
Numerical weather prediction is an important means in present-day weather prediction. Mesoscale model MM5 has been used widely in many observatories. Along with a large number of Doppler radar stations have been built in our country, it is a crucial problem for the meteorologists how to use Doppler radar data in numerical weather prediction model. As an efficient approach to improve the accuracy of numerical weather prediction, Four-dimensional Variational Assimilation is widely focused on by the domestic and overseas experts. One of its important characteristics is that the data in one time can influence previousanalysis results, which can not only provide the optimized initial condition but also make up observation datum's absence in some areas . This paper does the experiments of Doppler radar's Four-dimensional Variational Assimilation in MM5 model. Firstly after comparing advantages and disadvantages of several methods by using the data of simulative wind, the paper chooses variational analysis method t
o retrive three-dimensional wind field. Also, the humility field is obtained from radar echo intensity. Secondly meteorologic datum assimilation history and major ways are introduced. Especially four-dimensional Variational Assimilation is emphasized. Thirdly we expound the basic methods how to construct the adjoint model including adjoint code technique, objective function formation, and how to adopt the weighting coefficient, scaling factor, descent arithmetic. Lastly radar wind field and radar humility field are assimilated in the MM5 4D Variational Assimilation System. The assimilation experimentation results indicate that after assimilating radar wind field of small spatial scale, mesoscale and small-scale precipitation prediction can be improved and mesoscale and small-scale information which can't
appear by tradition datum can be gained, which is valuable to analyze the mesoscale and small-scale system structure ; the effect assimilating radar humidity field isn't obvious, which is perhaps correlation with precipitation types and assimilating time. The results also show that adding radar humidity field to initial condition at initial time can supply the gap of the regular data in reflecting the mesoscale and small-scale systems, strengthen the humidity in the initial field, and eventually help to improve precipitation. The experiment of assimilating radar wind field and radar humility field at the same time shows that vapor transportation and local vapor divergence play more significant role in causing excessively heavy rain than only high wet center. The paper's experiments also reveal that it is far beyond the need of improving objective and quantitative precipitation prediction to have only a mature mesoscale model, in effect, the temporal and spatial accuracy of data is crucial, and that radar dat
a's 4D variational assimilation in mesoscale model is an efficient way to solve the problem of the temporal and spatial accuracy of data.
引文
[1] Wang W, Wamer T T. Use of four-dimensional data assimilation by newtoinon relaxation and latent-heat forming to improve a mesoscale model precipitation forecast a case study [J]. Mon Wea Rev, 1988, 116(12):2593～2613.
[2] 葛文忠，党人庆，江敦春。Application of radar and satellite data for numerical simulation of heavy rainfall. In: International Symposium Asia Nagoya, Japan, Feb 2～4, 1998, 123～126.
[3] 郭霞，党人庆，葛文忠．雷达资料在江淮暴雨数值模拟中的应用[J]．热带气象学报．1999，15(4)：356～362．
[4] 徐玉貌，江敦春，夏文梅．雷达和卫星资料对江淮暴雨数值模拟的影响[J]．南京气象学院学报．2001，24(4)：545～552．
[5] 马清云，李泽椿，陶士伟．单部多普勒雷达风场反演及其在数值预报中的应用试验[J]．应用气象学报．2001，12(4)：488～493．
[6] 徐枝芳，徐玉貌，葛文忠．雷达和卫星资料在中尺度模式中的初步应用[J]．气象科学．2002，22(2)：167～174．
[7] 王峰云，刑谦．民航上海多普勒雷达风场反演UW方法的试验[J]．气象科学．2002，22(2)：230～235．
[8] 刘晓阳，刘桂馥．差分反演多普勒速度水平风场[J]．南京气象学院学报．2000，23(4)：549～554．
[9] 陶祖钰．从单多普勒速度场反演风矢量场的VAP方法[J]．气象学报．1992，50(1)：80～90．
[10] 邱崇践，余金香，Qin Xu．多普勒雷达资料对中尺度短期预报的改进[J]．气象学报，2000，58(2)：244～249．
[11] 陈列，寿绍文等．应用单多普勒雷达资料反演风场作暴雨中尺度分析[J]．南京气象学院学报．2003，26(3)：358～363．
[12] Sasak i. Some basic formalisms in numerical variational analysis[J]. Mon Wea Rev, 1970, 98(12):875～883.
[13] 陈列．利用单多普勒天气雷达观测资料反演风场进行暴雨研究．南京气象学院硕士学位
论文．2003：19～20．
[14] Hoke. J. E and R. A. Anthes. The initialization of numerical models by dynamical initialization technique. [J] Mon. Wea. Rev. 1976, 104:1551～1556.
[15] Stauffer. D. R. and N. L. Seaman. Use of four-dimentional data assimilation in a limited-area mesoscale. Part Ⅰ: Experiments with synoptic-scale data. [J]Mon. Wea. Rev. 1990, 118:1250～1277.
[16] Stauffer. D. R., T. T. Warner and N. L. Seaman, A Newtonian "Nudging" Approach to four-dimensional data Assimilation: Use of SEAME-Ⅳ data in a meso-scale model. Preprints, Seventh Conference on Numerical Weather Prediction. Montreal, Amer Soc. 1985, 77～85.
[17] Bell. R. S. The meteorological office fine-mesh data assimilation scheme. Meteor Mag.[J]1986. 115:161～177.
[18] 孟智勇，徐祥德，陈联寿．卫星亮温资料四维同化方案及其对“7．20”武汉特大暴雨的模拟试验．[J] 大气科学．2002，26(5)：663～676．
[19] Gandin. L. S. Objective Analysis of Meteorological fields.[M] (Translated by Israel program for scientific Translation, Jerusalem, 242pp.), 1963.
[20] 沈桐立，田永祥，葛效贞等．数值天气预报 [M] 北京：气象出版社，2003：237～240．
[21] 罗乔林，纪立人，朱宗申．四维同化的一些方法．[J] 气象科技，1993(4)：38～44．
[22] 杨晓霞，沈桐立．最优插值客观分析方法．[J] 南京气象学院学报，1991，14(4)：566～574．
[23] 李晓莉，沈桐立．四维伴随模式变分同化简介．[J]气象教育与科技，2000(1)：1～5．
[24] 高山红，吴增茂，谢红琴．Kalman滤波在气象数据同化中的发展与应用．[J]地球科学进展．2000，15(5)：571～575．
[25] Lorenc A C. Anaylysis methods for numerical weather prediction.[J]. Quart JR Met Soc, 1986, 112:1177～1194.
[26] Ehrendorfer M. Four-dimensional data assimilation: comparison of variational and sequential algorithms.[J]. Quart JR Met Soc, 1992, 118:673～713.
[27] Sasaki Y. An objective analysis based on the variational method.[J] J Meteor Soc Japan, 1958, 36(1):77～88.
[28] Sasaki Y Some basic formalisms in numerical variational analysis. Mon Wea Rev. 1970, 98(10):875～910
[29] 李华宏．伴随模式同化系统的设计及其应用研究．南京气象学院硕士学位论文，2003：27～31．
[30] LeDimet F X, Talagrand O. Variational algorithms for analysis and assimilation of meteorological observations: theoretical aspects. [J] Tellus, 1986, 38(A2):97～110.
[31] 周毅，刘宇迪等．现代数值天气预报[M] 北京：气象出版社，2002：203～205．
[32] Navon I M, Zou X, Derber J et al. Variational data assimilation with an adiabatic version of the NMC spectral model. Mon Wea Rev, 1992, 120(7):1433～1446.
[33] 蒲朝霞，丑纪范。对中尺度遥感资料进行四维同化的共轭方法及数值研究。[J]高原气象，1994，13(4)：419～429．
[34] Zou X, Kuo H, Guo Y R. Assimilation of atmosphere radio refractivity using a nonhydrostatic adjoint model. Mon Wea Rev, 1995, 123(7): 2229～2247.
[35] Zou X, Navon I M, Sela J G. Variational data assimilation with moist threshold process using the NMC spectral model. Tellus, 1993, 45A(5): 370～387.
[36] Zou X. Rainfall assimilation through an optimal control of initial and boundary conditions in a limited-area mesoscale model. Mon Wea Rev, 1996, 124(12):2859～2882.
[37] Rostaling N, et al. Automatic differentiation in Odysses. Tellus, 1993, 45A (1): 558-568.
[38] Zou X, et al. Introduction to adjoint techniques and the MM5 adjoint modeling system. NCAR Technical Note, 1997, 71～99.
[2] 葛文忠，党人庆，江敦春。Application of radar and satellite data for numerical simulation of heavy rainfall. In: International Symposium Asia Nagoya, Japan, Feb 2～4, 1998, 123～126.
[3] 郭霞，党人庆，葛文忠．雷达资料在江淮暴雨数值模拟中的应用[J]．热带气象学报．1999，15(4)：356～362．
[4] 徐玉貌，江敦春，夏文梅．雷达和卫星资料对江淮暴雨数值模拟的影响[J]．南京气象学院学报．2001，24(4)：545～552．
[5] 马清云，李泽椿，陶士伟．单部多普勒雷达风场反演及其在数值预报中的应用试验[J]．应用气象学报．2001，12(4)：488～493．
[6] 徐枝芳，徐玉貌，葛文忠．雷达和卫星资料在中尺度模式中的初步应用[J]．气象科学．2002，22(2)：167～174．
[7] 王峰云，刑谦．民航上海多普勒雷达风场反演UW方法的试验[J]．气象科学．2002，22(2)：230～235．
[8] 刘晓阳，刘桂馥．差分反演多普勒速度水平风场[J]．南京气象学院学报．2000，23(4)：549～554．
[9] 陶祖钰．从单多普勒速度场反演风矢量场的VAP方法[J]．气象学报．1992，50(1)：80～90．
[10] 邱崇践，余金香，Qin Xu．多普勒雷达资料对中尺度短期预报的改进[J]．气象学报，2000，58(2)：244～249．
[11] 陈列，寿绍文等．应用单多普勒雷达资料反演风场作暴雨中尺度分析[J]．南京气象学院学报．2003，26(3)：358～363．
[12] Sasak i. Some basic formalisms in numerical variational analysis[J]. Mon Wea Rev, 1970, 98(12):875～883.
[13] 陈列．利用单多普勒天气雷达观测资料反演风场进行暴雨研究．南京气象学院硕士学位
论文．2003：19～20．
[14] Hoke. J. E and R. A. Anthes. The initialization of numerical models by dynamical initialization technique. [J] Mon. Wea. Rev. 1976, 104:1551～1556.
[15] Stauffer. D. R. and N. L. Seaman. Use of four-dimentional data assimilation in a limited-area mesoscale. Part Ⅰ: Experiments with synoptic-scale data. [J]Mon. Wea. Rev. 1990, 118:1250～1277.
[16] Stauffer. D. R., T. T. Warner and N. L. Seaman, A Newtonian "Nudging" Approach to four-dimensional data Assimilation: Use of SEAME-Ⅳ data in a meso-scale model. Preprints, Seventh Conference on Numerical Weather Prediction. Montreal, Amer Soc. 1985, 77～85.
[17] Bell. R. S. The meteorological office fine-mesh data assimilation scheme. Meteor Mag.[J]1986. 115:161～177.
[18] 孟智勇，徐祥德，陈联寿．卫星亮温资料四维同化方案及其对“7．20”武汉特大暴雨的模拟试验．[J] 大气科学．2002，26(5)：663～676．
[19] Gandin. L. S. Objective Analysis of Meteorological fields.[M] (Translated by Israel program for scientific Translation, Jerusalem, 242pp.), 1963.
[20] 沈桐立，田永祥，葛效贞等．数值天气预报 [M] 北京：气象出版社，2003：237～240．
[21] 罗乔林，纪立人，朱宗申．四维同化的一些方法．[J] 气象科技，1993(4)：38～44．
[22] 杨晓霞，沈桐立．最优插值客观分析方法．[J] 南京气象学院学报，1991，14(4)：566～574．
[23] 李晓莉，沈桐立．四维伴随模式变分同化简介．[J]气象教育与科技，2000(1)：1～5．
[24] 高山红，吴增茂，谢红琴．Kalman滤波在气象数据同化中的发展与应用．[J]地球科学进展．2000，15(5)：571～575．
[25] Lorenc A C. Anaylysis methods for numerical weather prediction.[J]. Quart JR Met Soc, 1986, 112:1177～1194.
[26] Ehrendorfer M. Four-dimensional data assimilation: comparison of variational and sequential algorithms.[J]. Quart JR Met Soc, 1992, 118:673～713.
[27] Sasaki Y. An objective analysis based on the variational method.[J] J Meteor Soc Japan, 1958, 36(1):77～88.
[28] Sasaki Y Some basic formalisms in numerical variational analysis. Mon Wea Rev. 1970, 98(10):875～910
[29] 李华宏．伴随模式同化系统的设计及其应用研究．南京气象学院硕士学位论文，2003：27～31．
[30] LeDimet F X, Talagrand O. Variational algorithms for analysis and assimilation of meteorological observations: theoretical aspects. [J] Tellus, 1986, 38(A2):97～110.
[31] 周毅，刘宇迪等．现代数值天气预报[M] 北京：气象出版社，2002：203～205．
[32] Navon I M, Zou X, Derber J et al. Variational data assimilation with an adiabatic version of the NMC spectral model. Mon Wea Rev, 1992, 120(7):1433～1446.
[33] 蒲朝霞，丑纪范。对中尺度遥感资料进行四维同化的共轭方法及数值研究。[J]高原气象，1994，13(4)：419～429．
[34] Zou X, Kuo H, Guo Y R. Assimilation of atmosphere radio refractivity using a nonhydrostatic adjoint model. Mon Wea Rev, 1995, 123(7): 2229～2247.
[35] Zou X, Navon I M, Sela J G. Variational data assimilation with moist threshold process using the NMC spectral model. Tellus, 1993, 45A(5): 370～387.
[36] Zou X. Rainfall assimilation through an optimal control of initial and boundary conditions in a limited-area mesoscale model. Mon Wea Rev, 1996, 124(12):2859～2882.
[37] Rostaling N, et al. Automatic differentiation in Odysses. Tellus, 1993, 45A (1): 558-568.
[38] Zou X, et al. Introduction to adjoint techniques and the MM5 adjoint modeling system. NCAR Technical Note, 1997, 71～99.