面向复杂云图变化的遥感降水临近预报方法及可预报性研究

Remote sensing quantitative precipitation nowcasting (QPN) plays a critical role in the world weather research program (WWRP). Because of complex change of cloud, most QPN approaches are limited in real-time hydro-meteorological applications due to low accuracy of cloud moving tracking, failure of capturing clouds dissipation and extension process, as well as short effective forcast time, and so on. To deal with this issue, we aim to establish a complete set of QPN methodological system for complicate changing cloud, based on combining the advantages of geostationary meteorological satellites (FY-2F/G and FY-4A) and the weather radars. The methodological system involves the following four aspects, cloud motion vector calculation, remote sensing parameter extrapolation, combination of precipitation estimation and parameter nowcasting, as well as reconstruction of forecasted precipitation field from satellite and radar. It is expected that more effective cloud tracking would be made by an improved pyramid optical flow method. Moreover, a new remote sensing parameter extrapolation method would be established by removing the near-rigid assumption of cloud and considering the spatio-temporal continuity of cloud patches. In addition, new precipitation fields can be obtained by integrating precipitation estimation and parameter extrapolation, as well as reconstructing multi-source and multi-scale precipitation forecast field. Last but not the least, the predictability of QPN would be explored thoroughly from precipitation system characteristics, QPN methods, and remote sensing sensors, and so on. Therefore, this project is expected to propose new ideas and new vias to provide timely, effective, and large-scale precipitation nowcasting information by developing a new methodological system and understanding the predictability of QPN, and thus to meet the demand of fine weather forecast.

遥感降水临近预报是当今“世界天气研究计划(WWRP)”所列的重要研究课题之一。由于云图变化十分复杂，目前遥感降水临近预报面临云运动矢量计算精度低、云团消散扩张过程捕捉困难、降水预报时效短等问题，本项目拟面向复杂云图变化，结合静止气象卫星(FY-2F/G和FY-4A)及天气雷达优势，建立并验证一套完整的云运动矢量计算—遥感参量外推—降水估算及预报—卫星与雷达降水预报场重构的遥感降水临近预报理论与方法体系，实现：基于金字塔光流法的高精度云运动矢量计算，去除云团近刚性假设的遥感参量外推，降水估算与参量外推一体化的降水预报，融合多源多尺度信息的降水预报场重构，并基于此系统探索遥感降水临近预报可预报性规律。通过研究，有望创立一种新型遥感降水临近预报方法体系，摸清影响遥感降水可预报性的因素及机制，为增强降水临近预报能力，获取及时、大范围、高精度降水预报场提供新思路和方法，满足精细化天气预报的迫切需求。

及时、准确、大范围的遥感降水临近预报是国家防灾减灾、重大社会活动和精细化天气预报的迫切需求。由于云图变化十分复杂，目前遥感降水临近预报面临云运动矢量计算精度低、云团消散扩张过程捕捉困难、降水预报时效短等问题，本项目面向复杂云图变化，结合静止气象卫星及天气雷达优势，建立并验证了一套完整的云运动矢量计算—遥感参量外推—降水估算及预报—卫星与雷达降水预报场重构的遥感降水临近预报理论与方法体系。实现了基于金字塔光流法的高精度云运动矢量计算，去除了云团近刚性假设的遥感参量外推，开展了降水估算与参量外推一体化的降水预报，融合了多源多尺度信息的降水预报场重构，摸清了影响遥感降水可预报性的因素及机制。本项目还开发降水临近预报系统一套，该系统具有：实时数据下载、雷达数据处理、降水反演与预报、预报结果画图四个功能模块。该系统在北京市、湖北武汉、贵州、淮河流域得到了应用。通过本项目研究，为增强降水临近预报能力，获取及时、大范围、高精度降水预报场提供新思路，为满足精细化天气预报的迫切需求提供了新方法。