大气科学  2016, Vol. 40 Issue (6): 1127-1142 PDF
EnSRF雷达资料同化在一次飑线过程中的应用研究

1 南京信息工程大学气象灾害预报预警与评估协同创新中心, 南京 210044
2 南京信息工程大学气象灾害教育部重点实验室, 南京 210044

The Simulation of a Squall Line with Doppler Radar Data Assimilation Using the EnSRF Method
GAO Shibo1,2, MIN Jinzhong1,2, HUANG Danlian1,2
1 Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters, Nanjing University of Information Science & Technology, Nanjing 210044
2 Key Laboratory of Meteorological Disaster of Ministry of Education, Nanjing University of Information Science & Technology, Nanjing 210044
Abstract: An ensemble forecast and a deterministic forecast of a squall line that occurred in southern China on 23 April 2007 have been conducted using the Weather Research and Forecasting (WRF) model with microphysical schemes that include complex ice and snow processes.It is found that the deterministic forecast can capture the main characteristics of the squall line, but the simulated squall line is inaccurate, especially in the back stratus cloud region.The ensemble forecast technique can reduce the uncertainty in the model simulation and the majority of the members in the ensemble show a better performance than the deterministic forecast.The analysis members, which are obtained from radar data assimilation using the EnSRF (Ensemble Square Root Filter) method with outputs of the 40 members in the ensemble experiment as backgrounds, are used to provide initial conditions for the ensemble forecast.Differences in results among the ensemble members with and without radar data assimilation reflect the impact of EnSRF radar data assimilation on the simulation of the squall line.The analysis members with radar data assimilation provide more mesoscale and microscale information of the convective cells in the squall line system.Most members can capture the thermal-dynamical structure of the squall line system and successfully simulate the suqall line in the back stratus cloud region.Analysis of the simulations in the ensemble forecast with radar data assimilation indicates that most members perform better than that without radar data assimilation.The ETS (Equitable Threat Score) of the ensemble forecast with radar data assimilation is higher than that without radar data assimilation, and the ETS of the deterministic forecast is lower than that of the ensemble forecast.
Key words: EnSRF method      Data assimilation      Deterministic forecast      Ensemble forecast      Squall line
1 引言

2 资料和方法 2.1 资料

2.2 集合均方根滤波方法

 $\mathit{\boldsymbol{X}}_i^{\rm{b}}(t) = \mathit{\boldsymbol{MX}}_i^{\rm{a}}(t - 1),$ (1)
 $\mathit{\boldsymbol{X}}_i^{\rm{a}} = \mathit{\boldsymbol{X}}_i^{\rm{b}} + \mathit{\boldsymbol{K}}(y_j^{\rm{o}} - \mathit{\boldsymbol{HX}}_i^{\rm{b}}),$ (2)
 $\mathit{\boldsymbol{K}} = {\mathit{\boldsymbol{P}}^{\rm{b}}}{\mathit{\boldsymbol{H}}^{\rm{T}}}{(\mathit{\boldsymbol{H}}{\mathit{\boldsymbol{P}}^{\rm{b}}}\mathit{\boldsymbol{H}} + \mathit{\boldsymbol{R}})^{ - 1}},$ (3)
 $\mathit{\boldsymbol{H}}{\mathit{\boldsymbol{P}}^{\rm{b}}}{\mathit{\boldsymbol{H}}^{\rm{T}}} \approx [(\mathit{\boldsymbol{HX}}_i^{\rm{b}} - \overline {\mathit{\boldsymbol{H}}{\mathit{\boldsymbol{X}}^{\rm{b}}}} ){(\mathit{\boldsymbol{HX}}_i^{\rm{b}} - \overline {\mathit{\boldsymbol{H}}{\mathit{\boldsymbol{X}}^{\rm{b}}}} )^{\rm{T}}}],$ (4)

 $\overline {{\mathit{\boldsymbol{X}}^{\rm{a}}}} = \overline {{\mathit{\boldsymbol{X}}^{\rm{b}}}} + \mathit{\boldsymbol{K}}\left( {{y^{\rm{o}}} - \overline {\mathit{\boldsymbol{H}}{\mathit{\boldsymbol{X}}^{\rm{b}}}} } \right),$ (5)
 $\mathit{\boldsymbol{X'}}_i^{\rm{a}} = \mathit{\boldsymbol{X'}}_i^{\rm{b}} - \alpha \mathit{\boldsymbol{KHX'}}_i^{\rm{b}},$ (6)

2.3 雷达资料质量控制及观测算子

3 试验设计

 图 1 模拟区域与雷达位置[图中黑色矩形框代表模拟范围、倒三角代表雷达位置，大圆圈代表半径为230 km雷达观测的范围，其中9部雷达分别为厦门(XMRD)、福州(FZRD)、建阳(JYRD)、广州(GZRD)、深圳(SZRD)、汕头(STRD)、阳江(YJRD)、韶关(SGRD)以及桂林(GLRD)雷达] Figure 1 The model domain and radar locations. Black rectangle: area of simulation; circles: the coverage of 230 km radar observations; inverted triangles: locations of the 9 radar stations at Xiamen (XMRD), Fuzhou (FZRD), Jianyang (JYRD), Guangzhou (GZRD), Shenzhen (SZRD), Shantou (STRD), Yangjiang (YJRD), Shaoguan (SGRD), and Guilin (GLRD)

4 飑线过程及环境背景

2007年4月23日，受中层的低压槽，地面准静止锋和夜间低空急流的影响，华南地区发生了一次典型的飑线过程。飑线系统在20:00进入广西和广东省交界处，随后以17 m s-1向东南方向移动，24日02:00飑线强度达到最大，水平尺度达800 km左右，基本覆盖了整个广东省，一直到24日03:00逐渐消散，维持了约7 h。经统计，广东省累计降水超过50 mm，部分县市雷雨大风为8~9级，强降水超过100 mm。局部地区发生冰雹灾害，最大风速可达24 m s-1

 图 2 2007年4月（a）23日12:00（协调世界时，下同）和（b）23日18:00基于NCEP再分析资料的850 hPa环境场变量。阴影为可降水量（单位：kg m−2），黑实线为等高线（单位：dagpm），红虚线为等位温线（单位：K），矢量为风场（单位：m s−1） Figure 2 Environmental variables at 850 hPa extracted from the NCEP reanalysis data, including the column-integrated precipitable water (shaded, units:kg m−2), geopotential height (black solid lines, units: dagpm), potential temperature (red dashed lines, units: K), and wind vectors (blue vectors, units: m s−1) at (a) 1200 UTC and (b) 1800 UTC, 23 April 2007
5 试验结果 5.1 控制试验模拟能力评估

 图 3 雷达观测组合反射率因子（单位：dBZ）2007年4月：（a）23日22:00；（b）23日23:00；（c）24日00:00；（d）24日01:00；（e）24日02:00 Figure 3 Observed composite radar reflectivity (units: dBZ) in the squall line at (a) 2200 UTC 23, (b) 2300 UTC 23, (c) 0000 UTC 24, (d) 0100 UTC 24, and (e) 0200 UTC 24 April 2007

 图 4 同图 3，但为确定性预报试验模拟结果 Figure 4 Same as Fig. 3, but for the deterministic forecast experiment
5.2 集合预报模拟结果

 图 5 集合预报试验模拟的2007年4月24日00:00雷达组合反射率因子(单位：dBZ)：(a)观测(OBS)；(b-l)分别是第2、3、12、13、16、17、19、21、22、36和37个集合成员 Figure 5 The composite radar reflectivity (units: dBZ) in the squall line from observations and simulated by the ensemble forecast experiment at 0000 UTC 24 April 2007: (a) Observations (OBS); (b-l) simulations of member 2, member 3, member 12, member 13, member 16, member 17, member 19, member 21, member 22, member 36, and member 37
5.3 同化结果

5.2节以雷达观测为标准，分别对比了确定性预报和集合预报试验的组合回波分布特征，结果表明集合预报试验能够更加细致地描述飑线的中尺度结构。下面我们分别在23日23:00和24日00:00，利用EnSRF方法同化9部多普勒雷达资料。为了和同化雷达资料前的模拟结果对比，这里同样以24日00:00(同化结束时刻)的11个集合成员为例。

 图 6 同图 5，但为EnSRF同化试验结果 Figure 6 Same as Fig. 5, but for EnSRF final analysis results

 图 7 三组试验2007年4月24日00:00雷达反射率因子的垂直剖面(单位：dBZ)：(a)确定性预报；(b)未同化雷达资料的第13个集合成员；(c)同化雷达资料的第13个集合成员的分析场 Figure 7 Vertical cross sections of radar reflectivity in the squall line at 0000 UTC 24 April 2007 (units: dBZ): (a) Deterministic forecast, (b) member 13 forecast without radar data assimilation, (c) member 13 analysis with radar data assimilation

 图 8 三组试验2007年4月24日00:00，地表扰动位温(阴影，单位：K)和水平速度(矢量，单位：m s-1)：(a)确定性预报；(b)未同化雷达资料的第13个集合成员；(c)同化雷达资料的第13个集合成员的分析场；(d)第13个集合成员同化试验与集合预报试验地表面扰动位温的差 Figure 8 Surface potential temperature perturbations (shaded, units: K) and horizontal winds (vector, units: m s-1) in the squall line at 0000 UTC 24 April 2007: (a) Deterministic forecast, (b) member 13 forecast without radar data assimilation, (c) member 13 analysis with radar data assimilation, (d) the difference in surface potential temperature perturbation between simulations of member 13 with and without radar data assimilation

 图 9 同图 8，但为地表面气压(阴影，单位：hPa)和水平速度 Figure 9 Same as Fig. 8, but for surface pressure (shaded, units: hPa) and horizontal winds

 图 10 同图 8，但为2 km水凝物总量(阴影，单位：g kg-1)和水平速度 Figure 10 Same as Fig. 8, but for water mixing ratio (shaded, units: g kg-1) and horizontal winds (vector, units: m s-1) in the squall line at 2 km
5.4 预报结果

 图 11 同图 5，但为02:00 Figure 11 Same as Fig. 5, but for 0200 UTC

 图 12 EnSRF雷达资料同化试验模拟的2007年4月24日02:00雷达组合反射率因子(单位：dBZ)：(a)观测；(b-l)分别是第2、3、12、13、16、17、19、21、22、36和37个集合成员 Figure 12 The composite radar reflectivity (unit: dBZ) in the squall line at 0200 UTC 24 April 2007 from observations and from simulations based on the EnSRF final analysis field: (a) Observations, (b-l) results of member 2, member 3, member 12, member 13, member 16, member 17, member 19, member 21, member 22, member 36, member 37

 图 13 三组试验2007年4月24日02:00模拟的雷达反射率因子的垂直剖面(单位：dBZ)：(a)确定性预报；(b)未同化雷达资料的第13个集合成员；(c)同化雷达资料的第13个集合成员的预报场 Figure 13 The simulated cross sections of reflectivity of the squall line at 02:00 UTC on April 24, 2007(unit: dBZ): (a) Deterministic; (b) member 13 forecast with no radar data; (c) member 13 forecast with radar data

 图 14 2007年4月23日22:00至24日02:00，确定性预报试验、集合预报试验和同化试验的第2、3、12、13、16、17、19、21、22、36和37个集合成员的雷达组合反射率因子(20 dBZ)的ETS评分 Figure 14 The ETS scores for the composite reflectivity (20 dBZ) simulated by member 2, member 3, member 12, member 13, member 16, member 17, member 19, member 21, member 22, member 36 and member 37 for the period from 2200 UTC 23 April 2007 to 0200 UTC 24 April 2007
6 结论

(1)相比确定性预报，集合预报具有更好的模拟能力。确定性预报能大致捕捉到飑线系统的发生发展过程，但模拟的飑线强对流区域范围偏大，“V”型结构不明显，飑线系统后部的层云结构也未能模拟出来。集合预报通过对多个集合成员分别积分，能最大限度地模拟出飑线的可能状态。大部分集合成员均模拟出飑线的主要特征，其中有11个集合成员均方根误差最低，模拟效果最佳。

(2) EnSRF能有效地同化实际雷达资料，增加模式初始场的中小尺度信息，使各个集合成员的分析场在飑线强对流回波区的范围和强度上更加准确，且同化试验分析出了飑线后部的层云回波和过渡区。相比确定性预报和集合预报，同化后得到的冷池面积增大，强度变强，“V”型结构更明显，飑线后部雷暴高压强度增大，水凝物高值中心的位置更加准确，且在层云区存在次高值区。

(3)利用集合预报技术，通过对同化后的集合成员进行模拟发现，大部分集合成员较未同化雷达资料时模拟效果有明显改善，模拟的雷达回波更接近观测，强对流回波区的线状回波和“V”型结构明显，并且较好地模拟出飑线后部的层云区。同化后的集合预报ETS评分最高，其次是未同化的集合预报，确定性预报最低。