我国东南沿海FY-2C卫星图像上多光谱云分析技术

A TECHNIQUE OF ANALYZING MULTI-SPECTRAL CLOUDS FOR SOUTHEAST COAST OF CHINA BASED ON DAYTIME SATELLITE IMAGERY

为了对我国东南沿海地区的云类型/下垫面进行客观定量检测,从云的光学特性出发,利用卫星图像上各种云型/下垫面呈现出的光谱特征,结合白天正午样本集,分析云的光谱特征差异,提出一种利用通道亮温和通道间亮温差及反照率等,识别发展较强的对流云、高中低混合多层云、中低层水云、薄高云和晴空的方法。通过样本采集、绘制散点图、检测试验,建立FY-2C云类型/下垫面识别流程,给出白天及夜间适合的检测判据。结果表明:强对流云团的光谱特征最显著,其判别相对简单,采用单通道或通道间亮温差即可获得较好的判别效果,但也存在对密实卷云错判的情形;与天气过程相联系的高中低混合多层云光谱特征也比较明显,较易识别,但高中低混合多层云和厚卷云以及发展较强的对流云光谱特征存在相似之处,在一些云层边界或交界处还是存在一些错判;暖水云和薄卷云的判识对可见光通道依赖性较强。经过多光谱云分析,可间接确定云相态,并可为其它云参数(云顶高、光学厚度及有效粒子半径等)的反演提供有利信息。

In order to conduct objective and quantitative measurement of cloud type classification and underlying surface over the region of Southeast China,a multi-spectral cloud analysis technique is presented in this paper.It distinguishes multiple cloud types,including deep convective clouds,high-middle-low multilayered clouds,warm water clouds,and thin high clouds and clear sky,from the underlying surface by investigating multi-spectral properties of imagery during the daytime.Cloud detection thresholds are derived by analyzing the multi-spectral characteristics of the five cloud category cases collected and drawing scatter figures and conducting a series of detection tests.The results show that the deep convective clouds are clearly identified utilizing single-channel thresholds or the differences in brightness temperature between the channels because of significant optical features,while there are some misclassifications with the optically thick high clouds.The multilayered clouds associated with mid-latitude fronts and cyclones can be recognized based on thresholds of blackbody brightness temperature and the brightness temperature,which are different between the infrared channel and the water vapor channel,while the cloud edges of fronts and cyclones are relatively difficult to identify due to similar spectral features of the optically thick high clouds and the convective clouds.The separation of warm-water clouds from the other clouds depends on the reflectance threshold of the visible channel,especially for the warm liquid water clouds over land and ocean surfaces.The optically thin cirrus clouds are distinguished with the split window methods while the visible reflectance is not applied.By analyzing the multi-spectral cloud characteristics the cloud phases are indirectly determined and the cloud top information is obtained while retrieving the cloud top parameters （the height,optical thickness,effective radius of cloud top）.