【作者】 刘一；

【导师】 陈德辉； 胡江林；

【作者基本信息】 中国气象科学研究院 ， 气象学， 2009， 硕士

【摘要】 动能谱分析表明,给定数值预报模式存在对实际大气运动的最高有效分辨尺度的问题。同样地,给定数值预报模式也存在对实际地形的最高有效分辨尺度的问题,也即模式有效地形问题。模式地形中若包含有模式无法正确“分辨”的次网格地形,不但对数值模式预报无益处,而且还会激发出“噪音”,对模式预报结果造成“有害”的影响。通常地,采用某种滤波平滑的方法对高分辨率的地形资料进行平滑,构造模式的有效地形。针对GRAPES模式来说,何种尺度的模式地形是最有效的?模式有效地形与模式格距存在何种关系?采用何种地形平滑的方法最有效?这些都是在GRAPES模式设计和实际应用中需要回答的基本问题。本文通过大量的数值对比试验,采用理想数值试验与实际个例敏感性试验相结合的方法,系统的研究分析了以下几个方面的问题:GRAPES模式最高可分辨地形尺度;五点平滑滤波方案和隐式切变滤波器的对比试验研究;GRAPES模式最有效模式地形的构造;模式有效地形应用对实际个例预报的改进等。并取得了如下主要结论:模式预报结果对模式地形尺度非常敏感。不同分辨率的数值预报模式需要选择恰当尺度的模式地形,才能减少地形“噪音”引起的虚假地形降雨,进而有效地提高模式的预报性能。理想数值试验结果表明,GRAPES模式使用10倍格距尺度的模式地形可以成功地获得过山气流的模拟试验结果,所模拟出的过山气流场的演变与采用大尺度模式地形所模拟出的结果一致,两者模拟出的地形地表拖曳力之比高达80%左右,两者几乎完全一致;若采用6倍格距尺度的模式地形,也能模拟出大尺度地形60%-80%的地表拖曳力,两者差异较小。可以近似地认为6倍格距的模式地形对GRAPES模式是有效的。当继续降低模式地形尺度时,结果差异很大,模拟结果急剧变差,所以,可以认为6倍格距是GRAPES模式的最高可分辨地形尺度。6倍以下格距的小尺度地形则需要过滤平滑去除,其影响作用需要通过次地形参数化技术加以考虑。对平滑滤波方案的研究结果显示,高阶低通隐式切变滤波器的滤波效果非常好,可以根据不同的需要,调节滤波参数,进而滤除模式无法辨别的小尺度地形。针对GRAPES模式,高阶低通隐式切变滤波器的滤波参数设置为p=5、ε=10时,就可以有效地滤除GRAPES模式无法分辨的小尺度地形(5Δx及其以下尺度的地形)。并利用高阶低通隐式切变滤波器对业务模式地形进行平滑滤波处理,将得到的高阶低通滤波地形与业务模式地形和五点平滑方案处理的模式地形(以下简称“五点平滑地形”)对比发现,高阶低通滤波平滑方案可以有效地滤除模式无法辨别的小尺度地形,对于模式能够分辨的地形尺度影响较小,而且较大程度的保存了业务模式地形复杂的地形变化特征。所以,高阶低通滤波方案可作为模式有效地形构造的通用方法,高阶低通滤波地形可以作为GRAPES模式的有效地形。对于这种构造GRAPES模式有效地形的通用方法也进行了连续一个月实际连续的试验。采用高阶低通滤波地形(即模式的有效地形)相对于业务模式地形预报的月平均降水将一些零星的降水区域滤除,预报结果更符合实况的降水分布。TS和BS评分也显示,高阶低通滤波地形预报结果的TS评分和BS评分均优于业务模式地形预报结果。高阶低通滤波地形使GRAPES中尺度区域模式对降水的预报能力得到了明显的改善。同时,对其他主要气象要素的预报结果也有了不同程度的改进。更多还原

【Abstract】 For a given numerical model, the kinetic energy spectra analysis shows that the model is limited to effectively simulate the atmospheric motion in certain scales. Similarly, a numerical model is limited to effectively represent the topography. If an inappropriate topography is used in a numerical model, it would cause the noises to harm the model predictions. Generally, some filter could be used to smooth the high resolution topography, in order to generate the model effective topography. For GRAPES, which scale of topography is effective? What is the relationship between the model effective topography and the model resolution? What is the good method to generate the effective topography for GRAPES_Meso model? Those are the basic questions to be answered for the scientific design and application of GRAPES model.By intercomparing a series of numerical simulation tests with ideal data sets and real data sets, this paper is mainly focused on the issues of the highest effective scale of topography for GRAPES and the methods to generate the model effective topography. The five points smoothing or high-order low-pass implicit tangent filters have been tested. The following main remarks were concluded:The model prediction is very sensitive to the scale of model topography. The numerical prediction models, which have different resolutions, need to choose suitable model topography. It is important to reduce the noises, which often cause false terrain rainfall, for improving the prediction performances of the models.The ideal data sets show that GRAPES can reproduce the airflow across the hill effectively using ten times grid scale of the model topography. This result is consistent with the best resolved runs. The GRAPES can still simulate the airflow motion by using six times grid scale of the model topography. Simulations with six times grid scale of the model topography capture about 60-80% of the drag in the best resolved runs. When GRAPES uses smaller scale of the model topography than six times grid scale, the result is very poor. Therefore, the six times grid scale of the model topography is the highest effective scale for GRAPES. The scales, which are smaller than the six times scale, need to be filtered. These influences are considered to be reduced by sub-grid topography parameterizations.By intercomparing the results of the different smoothing methods, the high-order low-pass implicit tangent filter was a better smoothing to deal with model topographies. We can adjust filter parameters to filter off the scale of model topographies for GRAPES according the different needs. The ideal tests show that the six times grid scale of the model topography (6Δx) is the highest effective scales. And we configure the filter parameters (p=5,ε=10) to smooth the five times grid scale and smaller than that scales off. We have made the model effective topography for GRAPES in this way to deal with the original topography.In the numerical simulation experiments, we get the smoother results by using the smoothed topography. Comparing the results with the observation, the prediction by using high-order low-pass filtered topography is more consistent with the observation. The impacts of different model topographies on the results of precipitation forecast are very significant. The precipitation forecast by model topography has stronger rainfall and extended rainfall areas. And the precipitation forecast is the best by using high-order low-pass filtered topography, which is most consistent with the observed precipitation. The whole rainfall reduces and the majority of sporadic rainfall completely disappears. TS and BS score also show the same result .At the same time, other major elements of weather forecasting has been the result of varying degrees of improvement.更多还原