有地形模式中气压梯度力误差扣除法

：在现有有地形的数值模式中，陡峭地形区气压梯度力的计算存在一个普遍问题，即计算精度较高的格式比较繁琐而费时，简单省时的格式又不精确和不稳定。为解决这个问题，作者等在最近提出了一种方法，称为气压梯度力的误差扣除法。该方法假定：气压梯度力的计算误差主要来自地形而与气压形势关系较小。用理想场对该方法进行检验后表明，这一方法是成功的。本文则用气候模式对作者提出的有地形数值模式中气压梯度力的误差扣除法进行了模拟检验。所用模式是作者等使用多年的Ｐ－混合坐标系５层模式，选用了四种气压梯度力的计算格式，即ＤＤＤ格式、Ｃｏｒｂｙ格式、平均温度格式和经典中央差格式。比较了这四种格式在有无误差扣除时的模拟结果，发现：对于计算精度较高的格式，如ＤＤＤ格式、Ｃｏｒｂｙ格式及平均温度格式，有无误差扣除的结果相差不大，但误差扣除法仍可在一定程度上改善模拟效果。对于计算精度差的格式，如经典中央差格式，在无误差扣除时计算不稳定，得不到模拟结果，进行误差扣除后，从根本上提高了其计算精度，因而也提高了计算的稳定性，达到了较满意的模拟效果。而且与其它格式的模拟结果相当接近。本文提出的误差扣除法可同时用于格点模式和谱模式。

MODELLING TESTS OF THE ERRORSUBTRACTION SCHEME FOR THE PRESSURE GRADIENT FORCE IN MODELS WITH TOPOGRAPHY

In the present numerical models with topography , there is a common problem of computation of the pressure gradient force in the steep mountainous areas. That is, the, computational formats with high accuracy are complex and time-consuming and those which are simple and time-saving arc not precise and stable. In order to solve the problem, the authurs recently developed a scheme called the error subtraction method of the pressure gradient force which assumes that the computational errors of the pressure gradient force are mainly resulted from the topog- rapby and have less relations to the pressure patterns. The tests made by use of ideal fields show that the method is successful. Modelling tests are made in this paper to check the error subtraction scheme of the pressure gradient force in models with topography developed by the authors. Four computational formats are selected, the modelling results of the four formates with and without the subtraction scheme are compared with one another. It is found that the modelling results are not very much different between the schemes with and without sub- tractions when formats with higher accuracy are used. However, the formats with error subtractions can still improve the modelling results to some extent. For formats with low accuracy the error subtraction method can fundamentally improve the accuracy and stabillty of computation. and get satisfactory efficiency of modellings.