地面观测资料同化初步研究

Preliminary Study on Surface Observational Data Assimilation

通过对比分析当前两种地面观测资料同化方案(Ruggiero方案和郭永润方案),对其优缺点进行初步分析研究。结果表明:两种同化方案同化地面观测资料后都是从模式低层到高层对分析场产生影响,中低层高度影响比较大,高层影响较小,并且对24 h模拟结果产生一定的影响。Ruggiero方案考虑了模式地形与观测站地形高度差异,但资料剔除较多,且模式分辨率越高,资料剔除对分析场的影响越明显。郭永润方案不考虑模式地形与观测站地形高度差异,假定所有地面观测资料位于模式面,因此无论模式分辨率如何,资料都能得到充分利用,但模式地形与实际观测站地形高度差异大小在地面观测资料同化分析中的影响不可忽略,因此郭永润方案有待改进,适当考虑这两种地形高度差异时效果更好。

With the development of numerical weather simulation and forecasting, the spatial and temporal resolution of numerical model is much higher and the data assimilation updating cycle is much shorter. Surface Observation Data Assimilation （SODA） becomes more and more important in data assimilation system. But designing a SODA scheme is a tough work. The problem is mainly resulted from two aspects. At first, Surface Observation Data （SOD） is related to land surface process and the atmospheric boundary-layer. Meanwhile, the topographic distribution of China is very complex and there exists different elevation between surface observation station measurement and numerical model surface, which also presents many challenges. Elevation difference between surface observation sites and numerical model surface is one big problem of SODA. Surface observation data are assimilated into mesoscale model MM5 using MM5-3DVAR system, and a case study of a heavy rainfall event occurred on 11--12 May 2004 is simulated. At the same time, two surface data assimilation approaches （Ruggiero＇ s approach and Guo Yongrun＇s approach） are compared and studied, and the strongpoint and shortcoming of these two assimilation approaches are discussed. The preliminary study shows that using these assimilation approaches to assimilate surface observed data can impact the analysis field of the numerical model from bottom to top, and bring more contribution to the mid-lower layers＇ analysis field while less to the higher layers＇. Ruggiero＇s approach considers the different elevation between model surface and surface observational sites, but discards too many surface observational data. When the resolution of numerical model is low, the problem of discarding data doesn＇t impact the analysis field much, but when the resolution of numerical model is higher, the problem becomes sharper, and both the analysis field and 24-hour simulation are impacted negatively. The actual difference in elevation between the observation sites and the model surface （DEOM） is not considered in Guo Yongrun＇s approach. Whether the resolution of numerical model is high or low, most of the surface observed data can be used with this approach. But the impact caused by DEOM to the analysis field can not be neglected. When DEOM is put into consideration, better results can be obtained. Because of the complexity consists in the topography distribution of China, these two SODA schemes can not be applied directly. Much work needs to be done to improve these two SODA schemes or to design a new reasonable and effective SODA scheme.