3. 多普勒速度数据的质量问题以及它的影响
论文在2. Data and methods描述了它使用的多普勒速度数据存在的问题,我们在这里翻译关键的几段,当然这里我可能只提炼其中的主要信息,不追求翻译的准确度:
We used the radar data that were routinely sent to PAGASA headquarters by communication satellites after some data processing at the radar site. Unfortunately, raw data stored at the radar site did not survive the destruction of the radar by Haiyan's extremely violent winds.
论文使用的雷达数据是通过通讯卫星传给PAGASA的,这些数据在雷达站已经被预先处理过,原数据由于雷达被摧毁已经丢失。
In addition, in the processing at the radar site, the Doppler velocity data were spatially smoothed with a local nine-point filter to reduce noise such as sea clutter and folded velocities. Unfortunately, this smoothing was done without first dealiasing the Doppler velocities and, as a result, it was impossible to dealias Doppler velocities in our processing of the data in two areas: area I, the narrow boundary regions between aliased and nonaliased Doppler velocities, as shown on the 6° plan position indicator (PPI) display at 2023 UTC (bands marked by Xs in Fig. 3a); and area II, inside the RMW, where the horizontal wind shear was very strong(oval surrounded by the dashed line in Fig. 3a). Area I included outliers because of the averaging of aliased and nonaliased velocities, and area II included outliers because of dual-PRF velocity errors in the presence of strong horizontal wind shear.
在雷达站的预处理中,多普勒速度数据在空间上被平滑了。平滑的方法是九数据点过滤,应该是一种3*3数据点的平滑方式。这个平滑操作在多普勒速度退模糊处理之前进行,
因此我们无法对如下两个区域的多普勒速度进行退模糊:
I. 出现速度模糊的边界,即图(a)中用×标示的部分。速度模糊的边界正常来说是没有过渡的,比如直接从-70的负速度转为+70的正速度这样,然而在平滑处理中,这些边界上的速度数据被平滑了,因此我们能看到速度模糊的边界有明显的“负--0--正”这样的过渡。速度模糊边界上的数据难以被退模糊,需要被舍弃。
II. 海燕RMW(最大风速半径)内部的水平风切变非常大,导致双PRF雷达多普勒速度出现异常值,在图(a)中用黑色虚线椭圆标示。这些异常值被平滑以后,相邻的速度数据也受到影响,需要被舍弃。比如在黑色虚线椭圆中,出现了接近0的多普勒速度,如果直接退模糊,得出的多普勒速度会达到147m/s左右

。这里“水平风切变”与我们熟悉的,阻碍台风发展的“垂直风切变”不同,指的是风向风速在水平方向上的变化。
