基于雷达资料四维变分同化及云模式的中尺度对流系统数值临近预报试验

Numerical nowcasting experiments for the simulation of a mesoscale convective system using a cloud model and radar data assimilation with 4DVar

  • 摘要: 基于雷达资料快速刷新四维变分同化(RR4DVar)初始化的三维数值云模式,利用京津冀6部新一代多普勒天气雷达和区域自动气象站观测资料,针对2013年7月4日出现在京津冀平原地区的中尺度对流系统(MCS),开展了数值临近预报试验。研究结果表明,充分考虑雷达观测信息的对流尺度数值临近预报具有很大的优势,但也存在不足:(1)模式能够较好地把握中尺度对流系统的组织发展和移动演变特征,对风暴回波带的走向和尺度特征有较好的预报,但对强回波的强度和位置预报存在一定偏差;(2)模式预报可以反映风暴系统的中小尺度扰动特征,对风暴冷池和出流边界(阵风锋)的发展变化均有较为合理的预报;(3)模式对强降水中心和雨带位置的预报有很大优势,能较好地预报弱降水雨带的分布形势和雨量,但对强降水落区的预报偏大;(4)模式对风暴造成的对流性强降水的预报准确率较高,对0.5—10 mm阈值的降水范围预报偏差比较合理,对10 mm以上降水范围的预报偏大,但是对弱降水风暴的弱回波较强回波的预报性能要好;(5)由于三维数值云模式对京津冀复杂地形的处理不够完善,对山前风场预报偏差较大,造成对山前风暴的发展演变和山前降水的预报偏差较大。

     

    Abstract: The paper focuses on numerical nowcasting experiments for the simulation of a mesoscale convective system occurred in the Beijing-Tianjin-Hebei plain on 4 July 2013. A three-dimensional numerical cloud model is applied in this study. The rapid-refresh radar data of 6 CINRAD radar observations combined with regional automatic weather stations are assimilated with 4DVar to provide initial condition for model simulation. Results indicate that convective-scale numerical nowcasting has been improved greatly when radar observations are considered fully; however, several problems are also detected. It is found that: 1) the model is able to capture the characteristics of the organization and evolution of the moving storm system, and well forecast the moving direction and scale of the echo band, but cannot realistically simulate the strength and location of strong echoes; 2) the model can reproduce small-scale disturbances in the storm system, and provides reasonable forecast of changes in cold pool and outflow boundary (gust fronts); 3) the model has a great advantage of forecasting the heavy rain center and rain-belt location, and can well predict the distribution and total rainfall related to weak rain belts, but the area of heavy rain is overestimated; 4) the model forecast is more accurate for convective rainfall, and the forecast bias for areas of precipitation within 0.5-10 mm is reasonable , but areas of precipitation above 10 mm is overestimated. Besides, for a weak rain storm, the model performs better in weak echo forecast than in strong echo forecast; 5) large biases are found in the forecast of Piedmont wind, rainfall, and the development and evolution of the storm system due to the poor representation of the complex terrain in Beijing-Tianjin-Hebei region in the three-dimensional numerical cloud model.

     

/

返回文章
返回