交叉相关外推算法的改进及其在对流临近预报中的应用

Improvement and Application Test of TREC Algorithm for Convective Storm Nowcast

交叉相关外推算法是目前对流临近预报的主要算法之一。它主要是通过计算雷达回波等资料在连续时次的空间最优相关,得到对流系统不同位置的移动矢量特征,并基于这些获得的移动矢量对雷达回波等进行外推,从而达到预报的目的。因为算法既考虑了回波移动矢量大小和方向的变化,也考虑了整个回波在移动过程中的形变,所以,基于该算法的外推临近预报具有一定的物理意义。该文介绍了交叉相关外推预报算法的基本原理,对算法进行了一系列的质量控制,包括抑制地物杂波和噪声的影响、对"丢失"计算的点进行矢量补充以及对追踪的矢量进行平滑处理等多个步骤。个例对比分析表明:算法在通过质量控制和优化处理后,追踪的回波移动矢量质量明显改善。利用优化后的算法,对2004年夏季和2005年夏季发生在京津冀地区的4个强对流个例,包括2次飑线过程、1次雹暴过程和1次强雷暴过程,进行了外推临近预报试验,并且对算法在对流临近预报中的效果做了初步的检验评估。结果表明:该算法对强对流天气过程的临近预报具有一定效果,算法可以给出未来30 min或者60 min内雷达回波和雷暴位置以及形状的外推预报,预报结果与实况比较接近。

At present, cross-correlation extrapolation is one of the main algorithms for convective storm nowcast. Motion vectors of convective storm for every divided equal-sized two-dimensional arrays of radar echo or other data measured at two times several minutes apart by calculating optimal spatial cross-correlation are obtained in the algorithm. The obtained motion vectors are customarily called TREC （tracking radar echoes by correlation） vectors or TREC winds. And then, storm nowcast can be achieved by extrapolating radar echoes or other data based on the obtained TREC vectors. The algorithm results involve not only the changing characteristics of magnitude and direction of the motion vectors, but also shape varieties of the whole echoes in the course of their movement. So the result of storm nowcast based on the algorithm is assuredly reasonable and significant in meteorology. The basic principle of TREC is introduced firstly. And a number of methods to improve the algorithm result are presented, including noisy vector restriction and clutter contamination removal, discarded or missing motion vector supplement, vector smoothing, and so on. Analysis results of two cases indicate that the tracked motion vectors can be markedly improved after quality control and optimization processes to TREC algorithm. Finally, based on the optimized TREC algorithm and Tianjin radar data, storm nowcast tests and verifications of four intense convective storms that occur in Beijing-Tianiin-Hebei areas during 2004 and 2005 summertime, including two squall line cases, a hailstorm case and a strong thunderstorm case, are described in detail. The results indicate that the improved algorithm is available for convective storm nowcast. The algorithm can automatically produce 30- minute or 60-minute forecast of location and shape of radar echoes or storm characteristics based on the extrapolated vectors. The forecast results are close to what is actually happening. Because the algorithm can automatically produce forecast results in real time mode, it is helpful for convective storm nowcast and warning. The forecast results are also clear at a glance, so abilities of forecasters for strong convective storm identification and forecast can be enhanced by the algorithm. An expectation is that the improved algorithm can be used for operational storm nowcast in the near future.