基于卫星资料观测和反演降水是当前监测全球尺度降水的主要方式, 而其中一大难题就是如何将降水云与非降水云在像素尺度进行有效分离, 这也是准确反演地表降水量的基本前提。为建立一套适用于常见星载可见光/红外探测仪器的降水云识别方法, 本文利用热带测雨卫星 (TRMM) 可见光/红外辐射计 (VIRS) 和测雨雷达 (PR) 的融合观测资料, 针对选定的代表性区域, 统计分析了较长时间尺度上降水云与非降水云的典型云属性差异。在此基础上, 提出了一种基于云光学厚度和云滴有效半径的白天降水云识别方案 (IPCτRe)。由于用来获取上述云参数的可见光/红外信号无法透过降水性云层, 此方案不受下垫面条件的影响, 适用于陆地和海洋区域。为验证IPCτRe方案的降水云识别效果, 本文以PR瞬时降水探测结果为真值, 采用三种二元预报评价因子对识别结果进行了定量评估, 并与Inoue and Aonashi (2000) 和Nauss and Kokhanovsky (2006) 提出的降水云识别方案进行了比较。研究表明, IPCτRe方案的降水云识别性能均高于其它两种方案。特别是在洋面上, 降水云识别比例达到84%, 而对非降水云的误判率只有6%, 到达了降水卫星监测和预报业务所要求的精度。
Satellite remote sensing is currently the most important way of global-scale precipitation observations. The identification of precipitating clouds based on the satellite-borne measurements is still one of the most challengable problems. In order to get a universal precipitating-cloud identification method available for common optical satellite measurements, the relationship between cloud parameters and precipitating-cloud pixels is analyzed by using matched TRMM Visible and Infrared Scanner (VIRS) and Precipitation Radar (PR) long time scale measurements in the selected regions. According to the derived characteristic cloud parameters of precipitating clouds that is contrast to non-precipitating clouds, a daytime precipitating clouds detection scheme, called Identification of Precipitating Clouds from Optical Thickness and Effective Radius (IPCτRe), is proposed relying on both cloud optical thickness and effective radius. As the cloud parameters are retrieved from the visible and infrared signals that cannot penetrate the precipitating clouds, the IPCτRe scheme can be used operationally over both ocean and land areas. Comparison to PR standard rain products is conducted to verify the IPCτRe results, in which three dimidiate-forecast factors are utilized and two other precipitating-cloud identification schemes are also evaluated, with one proposed by Inoue and Aonashi (2000) and the other proposed by Nauss and Kokhanovsky (2006). The study proves that IPCτRe scheme gives better spatial depiction of precipitating clouds. Especially, in oceanic areas, precipitating and nonprecipitating clouds are well separated by current method, with the probability of detection near 0.84 and probability of false detection remaining just 0.06, indicating a satisfying accuracy for satellite monitoring and forecasting of precipitation operations.