8 . 3 VERIFICATION OF MESOSCALE FEATURES IN NWP MODELS
8 . 3 VERIFICATION OF MESOSCALE FEATURES IN NWP MODELS
复制标题
8.
DOI：
--
复制
发表时间：
2001
期刊：
影响因子：
--
通讯作者：
J. Kain
中科院分区：
文献类型：
--
作者：
M. Baldwin;S. Lakshmivarahan;J. Kain
预测验证的主要目标是回答这个问题：“这个预测有多好？”或“这组预测有多好？”与大多数看似简单的问题一样，简单的答案通常是不够的。特别是，与验证中小型模型预测相关的一些实际困难尚未得到令人满意的解决。随着计算能力的增强，运营天气预报中心能够运行分辨率越来越高的数值模型。由于这些模型的预报特征（例如降水最大值）的幅度往往会随着水平网格间距的减小而增加，因此空间中相对较小的误差可能会导致特定位置的预报值与观测值之间存在很大差异。因此，当比较包含小规模、高幅度特征的预测场和观测场时，传统验证方法获得的性能统计测量结果看起来很差。也许通过一个例子可以最好地说明这一点。
The primary goal of forecast verification is to answer the question: “how good is this forecast?” or “how good is this set of forecasts?” As is the case for most deceptively simple questions, a simple answer is usually not sufficient. In particular, some of the practical difficulties associated with verifying forecasts from mesoor smaller-scale models have not yet been satisfactorily resolved. As computing power increases, operational weather forecasting centers obtain the capability to run numerical models that contain increasingly higher resolution. Since the amplitude of forecast features (e.g., precipitation maxima) from these models tends to increase as the horizontal grid spacing decreases, relatively small errors in space can cause very large differences between forecast and observed values at a specific location. As a result, statistical measures of performance obtained by traditional verification approaches will look poor when forecast and observed fields containing small-scale, high-amplitude features are compared. This is perhaps best illustrated by an example.