摘要:采用基于本征正交分解的四维集合变分同化（POD-4DEnVar）方法，利用梅州站的多普勒天气雷达资料和NCEP资料，对2015年12月9日一次华南冬季暴雨过程进行同化试验，探讨了同化不同的雷达观测要素对暴雨模拟的影响。结果表明：同化多普勒天气雷达资料有利于削弱控制试验偏强降水的模拟结果，改善降水分布结构；同化不同的雷达观测要素得到的模拟结果不同，同时同化径向风和反射率的降水模拟结果最好。同化试验对降水模拟结果的改善主要通过调整初始时刻的风场和水汽条件来实现，一方面减弱偏南风和偏东风在暴雨区的辐合，阻碍海上暖湿气流对暴雨区的水汽输送，另一方面直接削弱暴雨区的水汽条件，大幅降低水汽混合比。同化试验相对于控制试验的同化增量远大于不同雷达观测要素的同化试验之间的分析场差异，这表明同化不同的雷达观测要素对初始风场和水汽条件的调整呈现类似的特征。虽然同化试验的初始场存在较小的差异，但随着模式积分，16 h后模拟降水出现了明显差异。分析同化试验之间的初始偏差演变发现，850~700 hPa的平均垂直速度偏差和雨水混合比偏差在模式积分至16 h开始急剧增长，这种变量偏差的急剧增长与逐时降水偏差的迅速增加一致，是降水偏差增长的直接原因。另外，这两个变量偏差的增大，也伴随着偏差能量的增大，变量偏差增长最明显的时段为偏差能量增幅最大的时段，且偏差能量迅速增长早于变量偏差和降水偏差的迅速增长，变量偏差增长最明显的区域为偏差能量梯度较大的区域。
关键词:四维集合变分同化/
雷达资料/
冬季暴雨/
初始偏差演变
Abstract:Using the POD-4DEnVar (Proper Orthogonal Decomposition-based four-dimensional ensemble variational assimilation) method, the impact of assimilation of different radar data on the simulation of a heavy rainfall is discussed. A suite of experiments to simulate a heavy rainfall process that occurred in Guangdong Province on 9 December 2015 have been conducted using NCEP reanalysis data and next-generation weather radar data collected at Meizhou. The results show that assimilation of the Doppler radar data is helpful to reduce the overestimation of precipitation simulated by the control experiment and improve the simulation of the precipitation structure. The simulation results obtained by assimilating different types of radar data are different, and the assimilation of both radial velocity and reflectivity yields the best result for precipitation simulation. The improvement in the simulation by the assimilation experiment is mainly achieved by adjusting winds and water vapor condition at the initial time. On the one hand, radar data assimilation weakens the convergence of the southerly and easterly flow in the heavy rain zone, which indirectly hinders water vapor transport associated with the warm moist flow to the storm area; on the other hand, radar data assimilation directly affects the water vapor condition related to the heavy rainfall by reducing the water vapor mixing ratio. The assimilation increments are much larger than the differences between data used in different assimilation experiments, which shows that assimilating different types of radar data has similar adjustment of the initial wind field and water vapor condition. Although there exist slight differences in the initial fields among the assimilation experiments, significant differences in precipitation simulation appear after about 16 hours of integration. The evolution of the initial deviations in different assimilation experiments is analyzed. It is found that average deviations of the 850-700 hPa vertical velocity and rainwater mixing ratio begin to increase rapidly when the model integration time reaches the 16th hour, and the rapid increases in these deviations are consistent with the rapid increase of precipitation deviation, indicating they are directly responsible for the increase in precipitation deviation. At the same time, with the increase in the deviations of the two variables, the difference total energy also develops. The deviations of the two variables grow the fastest when the difference total energy develops most rapidly, and the rapid growth of the difference total energy is preceded by that of the deviations of the variables and precipitation. Also, the region where the two variable deviations grow the most evidently is the area where the gradient of difference total energy is large.
Key words:Four-dimensional ensemble variational assimilation/
Radar data/
Heavy winter rainfall/
Initial deviation growth evolution
PDF全文下载地址:
http://www.iapjournals.ac.cn/dqkx/article/exportPdf?id=20180511
