基于雷达、卫星资料的台风多尺度混合资料同化技术发展及应用研究

近年来TC路径预报水平稳步提高，但强度预报几乎没有进步。实现TC强度的准确预报，模式的水平分辨率须达到云分辨尺度。然而，单纯提高分辨率对描述TC内核结构和快速增强过程效果并不理想，高分辨率模式还需要高分辨率的观测资料及有效的同化方法加以配合。本项目拟在已有的GSI-ETKF同化系统基础上，发展能协同融合雷达、台风报文MSLP、卫星辐射率资料，包含云分辨尺度在内的多尺度混合同化系统；针对混合同化计算量大、动力强非线性、多尺度适应性参数等开展技术攻关；综合评价各种资料的作用，揭示其中的影响机理；在此基础上寻找对TC预报尤其是强度预报最为有效的观测资料及合理的同化方案，提升系统对区域高分辨率台风模式的初始化性能和预报水平。

Even though there has been a steady improvement in tropical cyclone (TC) track forecast over the past 20 years, TC intensity forecast has remained a major challenge with virtually no improvement. Studies suggest that a cloud-resolving resolution is essential to capture inner-core structure and rapid intensity changes, but increasing resolution alone does not lead to systematic improvements which indicates requirement of high-resolution observations and advanced data assimilation techniques to efficiently use these observations. This proposal project will address TC prediction improvement problem by a) developing a multi-scale including cloud-resolving scale GSI-ETKF hybrid data assimilation system capable of efficiently assimilating radar, MSLP and satellite observations; b) addressing several issues related to hybrid analysis, such as high computational cost, strong nonlinearity and adaptive technique for multi-scale assimilation; c) evaluating the impact of these observations on typhoon forecasting especially intensity forecast; all of which will help to seek the most efficient observations and optimal assimilation strategy to improve typhoon initialization and prediction.