雷达观测对应模式变量非线性特征及对四维变分同化的影响

四维变分同化（4DVar）中切线性模式和伴随模式的时间积分长度即为同化时间窗的长度。为理解线性模式时间积分长度对4DVar的具体影响，在雷达观测对应变量非线性分析的基础上，进行了一系列不同时间窗（10 min、20 min和30 min）4DVar单点观测试验和一次降雨的实际雷达同化和预报试验。从径向风同化来看:短时间窗（10 min）的风场增量更大、更局地；长时间窗(20 min、30 min)的风场增量则更具系统性特征，但会丢失一些小尺度信息，导致暴雨预报能力降低。从反射率同化来看:短时间窗对6 h内强降水预报有较明显的改善，较长时间窗甚至会降低降水预报效果。研究旨在为合理设置4DVar的同化时间窗提供参考，以有效利用高时空分辨率的雷达观测资料，又尽量减小线性化造成的误差，进而快速有效地同化雷达信息。 

NONLINEAR CHARACTERISTICS OF MODEL VARIABLES CORRESPONDING TO RADAR OBSERVATIONS AND ITS EFFECTS ON 4D-VAR ASSIMILATION

The time integral length of tangent-linear and adjoint model in 4DVar is the length of the assimilation window. To understand the effect of time integral length on 4DVar, a series of 4DVar single observation experiments with different time windows (10 min, 20 min, 30 min) and a real rainfall case had been carried out based on the nonlinearity analysis of radar observations. In terms of radial velocity assimilation, the wind increments of the short-time window (10 min) is larger and more localized; while the long-time window brings broader and more systemic features, but some small-scale information will be lost, which leads to the decrease of the ability of heavy rain prediction. As to reflectivity, The 10min-window experiment has a distinct improvement on the forecast of heavy rainfall in 6 hours while the long window (20 min, 30 min) even degrades the precipitation forecasts. The purpose of this study is to provide a reference for the reasonable setting of the 4DVar assimilation window so as to make good use of radar observations with high spatial and temporal resolutions, and to minimize the error caused by linearization so as to assimilate the radar information quickly and effectively.