基于多普勒天气雷达的两种垂直风廓线反演方法的对比分析

Intercomparison of Vertical Wind Profiles Retrieved Using Two Techniques from Doppler Weather Radar

利用2010年阳江高精度探空仪国际比对试验中观测的探空测风资料、CINRAD-SA型多普勒天气雷达体扫资料及基于PUP-VAD方法反演的垂直风廓线产品资料,使用三种不同复杂程度的分层VVP方法(VVP1、VVP2和VVP3)反演了阳江雷达站上空的垂直风廓线,以探空测风为参考,以平均风向、风速均方差和相关系数作为评估因子,将雷达站周围降水回波分布分为较均匀和不均匀两种情况,对比分析了基于分层VVP方法和PUP-VAD方法反演的垂直风廓线。结果表明,相比PUPVAD方法,分层VVP方法都能提供更实用、质量更好的垂直风廓线,特别是在雷达周围降水回波分布不太均匀的情况下。主要原因是在某一高度,相比PUP-VAD方法只分析某仰角某距离圈上的径向速度资料来说,分层VVP方法是通过对该高度层10~100 km范围内的所有有效径向速度资料进行多元线性回归来求解该高度层的平均水平风向、风速,受资料覆盖率的影响程度要小于PUP-VAD方法。

The vertical wind profile product from Doppler weather radar has important role for the weather analysis. This paper retrieved vertical wind profiles using three different layered VVP algorithms with an increasing level of complexity( VVP1,VVP2 and VVP3). Radiosonde data,radar volume scan data and vertical wind profile product data retrieved using PUP- VAD algorithmfrom CINRAD- SA w ere collected,w hich w ere measured during Yangjiang 8thinternational radiosondeintercomparisontest sponsored by WM O from 12 July to 1 August2010. Take soundingwind as reference,the mean-square deviation and correlation coefficient of wind directions /speeds betw een sounding wind and retrieved wind w ere calculated. Take the mean-square deviation and correlation coefficient as evaluation factors,the vertical wind profiles retrieved using layered VVP and PUP- VAD algorithms w ere intercomparedin tw o different conditions of relatively uniform and uneven precipitation echo distribution around radar. Results show that,in the tw o cases,the layered VVP retrieval algorithm provides vertical wind profiles with a higher availability and a better quality than the PUP- VAD algorithm,especially in the case of uneven precipitation echo distribution around radar. Compared to analyze the causes,at a certain height layer,in contrast to a PUP- VAD retrieval,w here only those data from a range circle is used,all available scan data in the height layer within 10 ~ 100 km range is processed using a multi-dimensional linear regression. The impact of the data coverage ratio on layered VVP retrieval is less than that on PUP- VAD retrieval.