梅雨期降水卫星高精度连续反演和现时预报研究

The objective of this research is to deepen and continue the National Natural Science Foundation project finished by us..On the basis of the previous work, we hope to address the bottleneck problem of influence for improving the accuracy of satellite precipitation retrieval, in two ways, to further fundamentally improve the method of satellite precipitation retrieval. 1) By 10-minute rain gauge rainfall as the true value, we are to establish corresponding satellite precipitation retrieval algorithm of the day, night and twilight respectively. 2) Based on Cb-TRAM or cross-correlation methods, we are to develop the cloud movement trend tracking algorithm and to build the cloud image series of ten minutes interval. Combined with 10-minite rainfall rate retrieval algorithm and cloud movement trend tracking algorithm, the immediate tracking and continuous accumulation technique (ITCAT) for precipitation inversion is to be established. .Because the technology improves the reliability of precipitation inversion, and at the same time also gives full consideration to the cloud cluster motion process, precipitation retrieval can be dynamically estimated of intensity changes and duration time of cloud coverage, so it can significantly improve the retrieval accuracy of precipitation intensity. Especially to strong convective cloud cluster moving fast and changing greatly, the technology is more effective. The development of this technology provides a new practical idea for improving the accuracy of the precipitation monitoring and nowcasting. The consecutive forecast experiment of daytime precipitation field of half hour internal has been completed recently. The experiment results show that the technology has an important and positive significance to improve monitoring and nowcasting of precipitation, especially for strong precipitation of convective cloud.

本研究是已结题的国家自然科学基金项目的深化和继续。.期望能在前期工作的基础上，针对影响提高静止卫星降水反演精度的瓶颈问题，从两方面着手，进一步从根本上改进方法，1）以10分钟雨量计雨量作真值建立相应的卫星降水反演算法；2）在互相关方法基础上发展云团移动趋势跟踪算法，构建十分钟间隔云图系列。结合应用十分钟降水强度反演算法和云团移动趋势跟踪算法，建立即时跟踪和连续累计降水反演技术。因为该技术在提高反演可靠性的同时，又充分考虑了云团的移动过程，使降水反演能够动态的估计降水云覆盖测站的持续时间和强度变化，故能显著改善降水强度反演精度。特别是对移动快、变化大的强对流云团，更为有效。该技术的实现为提高降水监测和临近预报准确率的探索提供了新的可行的思路。我们目前完成的白天半小时降水场连续预报初步试验也证明了该思路对改进降水临近预报，特别是对流云强降水的临近预报，具有重要的积极意义。

在本项目的支持下，提出了全新的三时段各自独立的自训练半监督支持向量机云分类算法。针对卫星降水反演需求，该算法能够正确判识往往造成中尺度灾害性天气的积雨云与几乎不降水的厚卷云和产生持续稳定降水的雨层云及其他云类。为保证分类的客观性，所有训练过程无人工干预，人工监督只在于事先确定云类分类标准。分类过程，以30天为单位滚动训练，即今天的分类器来自最近30天样本训练，分类器将根据季节变化自行缓慢调整，不断更新。诸云类的分类判据随季节、地域变化，也随之动态跟踪变化。.在本项目的支持下，提出了夜间降雨率的卫星反演方法。该方法包括雨区划分和雨量估计两个步骤。雨区划分方法建立在降水概率判识矩阵RPIM基础上。夜间降水概率判识矩阵RPIM在IR1-BTD41二维光谱空间构建，任一同名像素的亮温Tb1和亮温差Tb4-1组合，由此即可直接确定出其相应的降水发生。雨量估计则针对落在二维光谱空间雨区的同名像素通过雨量估计方程进行进一步分析。雨量估计方法所获得的雨量估计值，能够与实际的降水值具有较好的相关性及较小的偏差，即该雨量估计值与实测值具有较好的一致性。致使本研究结果在多光谱卫星图像夜间降雨反演及临近预报中，具有更好的应用优势。结合前期已经完成的白天降水反演，即可实现全天时的降水连续反演和短时预报。.在本课题的支持下，开展了针对积雨云强降水的卫星反演。结合云顶亮温和过冲指数，构建了加权亮温差参数(WBTD)，提出了监测对流云强度变化的实时监测跟踪方法。在此基础上，进一步提出了一个可用于西北太平洋热带气旋强度估计的新卫星指标，定义为WIRa（WV-IRW to IRW Ratio），该指标在TC强度估计方面表现出了更高的反演精度。.以上这些重要成果，都申请了专利和软件著作权，发表了相关的重要SCI和EI论文及核心期刊论文，并在一些气象和民航部门得到转化和应用，获得高度评价。

内容获取失败，请点击重试