复杂地形下地面观测资料同化Ⅲ.两种解决模式地形与观测站地形高度差异方法的对比分析

A Study of Assimilation of Surface Observational Data in Complex Terrain Part Ⅲ: Comparison Analysis of Two Methods on Solving the Problem of Elevation Difference between Model Surface and Observation Sites

模式地形与观测站地形高度差异是地面资料同化方案设计中的一大难点。本文通过个例和单点试验对第I和第II部分(徐枝芳等,2007a,2007b)中涉及的两种解决模式地形与观测站地形高度差异的地面资料同化方法(即增加温度地形代表性误差和温度订正方法)进行对比分析,并将这两种方法应用于WRF-3DVAR,进行3个月的连续数值试验。研究结果表明:随着地形高度差异的增加,采用增加观测误差方法得到的估计值向其它资料同化分析值(背景场值)靠近,在一定的高度差异下则失去了同化分析地面资料的意义;温度订正方法对温度递减率的取值较为敏感,在有探空资料参与同化分析时,温度递减率取值敏感性相对减弱。当采用的订正值较为准确时,采用温度订正方法较增加观测误差方法能更好地处理两种地形高度差异,地面资料信息应用更充分,得到的估计值最有可能接近真实值。当模式地形与观测站地形高度差异较小(小于100m)时,两种解决模式地形与观测站地形高度差异的方法达到的效果基本一致。单点及个例试验表明,有探空资料等高质量观测参与同化分析时,采用增加观测误差方法得到分析场更接近真实场。三个月连续试验也表明有探空资料参与同化分析时,采用增加观测误差方法比温度订正法改进的郭永润同化方案(Guoetal.,2002)同化地面资料效果好,且加入地面资料同化对所有量级降水预报都有所改善。当地面资料同化方案没处理好时,加入地面资料同化的效果反而不如不加地面资料同化。

Elevation difference between surface observation sites and numerical model surface is a big problem of Surface Observation Data Assimilation （SODA）. In Part_I （Xu et al, 2007a）, results have shown that the elevation difference between observation sites and numerical model surface can impact the results of SODA. And better results can be obtained when surface observation data are assimilated into a numerical model with temperature observation reduced from actual station elevation to model terrain height. In Part_II （Xu et al, 2007b）, better results also can be obtained when surface observation data are assimilated into a numerical model with Terrain Error of Representativeness （TER）, which is related to the elevation difference between model surface and observation sites. In this paper, the two methods to solve negative influence of the difference between observation station and model elevation are analyzed and compared, by making three-month experiments in WRF_3DVAR. One is that the TER is added into the surface observation error, another is that temperature observation is reduced, via the temperature lapse rate, from actual station elevation to model terrain height. Results show that, when the difference between observation station and model elevation increases to a certain value, effects of the former method are close to the analysis field of other data assimilation, while the latter method is more robust when the value of temperature lapse rate is accurate. The latter method is sensitive to the value of temperature lapse rate, and the sensitivity decreases when radiosonde data are also assimilated. The effects of the two methods are the same when the model surface is close to observation sites, while the former method is better when radiosonde data are also assimilated. The scheme of Guo Yongrun＇s surface data assimilation （Guo et al., 2002） with TER is the best scheme in this paper. The SODA scheme can improve the results of rainfall simulation. The surface data assimilation approach without consideration of the difference between observation station and model elevation is not suitable to be used.