热带测雨卫星(Tropical Rainfall Measuring Mission, TRMM)积累的从可见光到微波的多光谱云和降水辐射信息,为全球 尺度云和降水参数反演研究提供了宝贵资料。云/晴空识别作为反演流程中的关键一环,对反演结果准确度起到决定性的 作用。利用可见光/红外扫描仪(visible and infrared scanner, VIRS)提供的五通道辐射通量观测,提出一种 可依靠TRMM平台观测资料独立实现的白天云检测方案。此方案由地形识别、二维表云检测和决策树云检测三个基本 步骤构成,检测结果分为无云覆盖、部分云覆盖和完全云覆盖三种。与MODIS云产品的对比结果表明,非冰雪下垫面 云检测精度较高,云量均值差异低于5%。特别对于冰雪下垫面,结合使用1.6 m反 射率以及3.7 m和10.8 m亮温信息,利用阈值法对冰雪覆盖的晴空像素和真实云像素实现了较好的分离。 利用该云检测方案,仅借助TRMM平台可见光和红外多通道资料即获得可信的云/晴空识别,这将为基于TRMM卫星多 仪器融合资料的云和降水参数反演等相关研究带来方便。

The Tropical Rainfall Measuring Mission (TRMM) satellite has collected a plenty of radiances from global clouds and precipitation, providing invaluable datasets for parameter retrieval of clouds and precipitation, in which cloud detection undoubtedly plays a key role. A cloud detection scheme, available during daytime, was proposed relying on five-channel measurements from TRMM visible and infrared scanner (VIRS). Three kinds of pixels including clear-sky, mixed-cloudy and cloudy ones are separated by such a scheme, which contains three major steps, i.e., terrain recognition, two-dimensional table search and decision tree determination. A comparison was conducted specially with MODIS standard cloud products to verify the cloud detection. It was demonstrated that the accuracy of cloud detection is satisfied on the whole, especially for grounds that are not covered by snow and ice, with mean cloud fraction bias less than 5%. When applied to cases companied by snow and ice in winter and or over plateau area, the 1.6 m channel reflectivity information, combined with 3.7 m and 10.8 m channel BT, were used to identify snow and ice covered sky in this cloud detection scheme, which made snow-covered sky pixels and cloud pixels well separated. Therefore, the proposed cloud detection scheme can be used to TRMM VIRS with considerable reliance, which leads to great convenience for further application of associated TRMM multiple-sensor measurements.