基于亚像素的金字塔光流法的雷达回波外推预报

Extrapolation Prediction of Radar Echo by the Pyramid Optical Flow Method Based on Sub-Pixels

该文采用基于亚像素的金字塔光流法获取广西的4次典型的降水雷达回波的速度和方向,然后根据线性外推方法对雷达回波进行2小时外推,获得未来的雷达回波的强度.采用命中率、临界成功指数及误报率等统计指标对预报的效果进行检验评估.研究结果表明,相比起基于像素的金字塔光流法,基于亚像素的金字塔光流法的精确度和临界成功指数更高,而空报率显著降低.这表明后者更适用于广西地区复杂多变的天气条件.

In this paper we adopt the pyramid optical flow method based on sub-pixels to acquire the speeds and directions of four typical precipitation radar echoes in Guangxi,then extrapolate the radar echoes up to two hours by the linear extrapolation method to obtain the future radar echo intensity.We check the effect of prediction by evaluating statistical indicators such as probability of detection,critical success index and false alarm rate.The research results show that,compared with the pyramid optical flow method based on pixels,the pyramid optical flow method based on sub-pixels has higher precision and critical success index,with the false alarm rate significantly reduced.This indicates that the latter is more fit for the complex and changeable weather conditions in Guangxi.