Regimes in the wintertime circulation over northern extratropics. II: Consequences for dynamical predictability
Regimes in the wintertime circulation over northern extratropics. II: Consequences for dynamical predictability
复制标题
北温带冬季环流状况。
DOI：
10.1002/qj.49711649602
复制
发表时间：
1990
影响因子：
8.9
通讯作者：
S. Tibaldi
中科院分区：
文献类型：
--
作者：
F. Molteni;S. Tibaldi
将半球流型按环流形势分类，分析了ECMWF模式在冬季中期预报的位势场误差。结果表明，这两种模型的概率分布都是一致的。北半球的误差和大尺度行星波的振幅在相空间区域是双峰的，在该区域，场有负的或小的投影到部分类似于太平洋-北美模式的EOF(经验正交函数)上，并且星系团之间的分离最强。
除了真正的系统误差外，误差场中还存在一个与流型有关的分量，其模式与观测流型的异常相反。事实上，ECMWF模式(尤其是其早期版本)往往会放松流向其人口最稠密地区的流动。这些结果得到了不同团簇之间跃迁频率的分析的支持，并基于随机扰动动力系统的Fokker-Planck方程的简单解析解得到了解释。
对不同ECMWF模式计算的统计数据表明，T106模式在中期误差的系统分量和与体制相关的分量方面都有相当大的改善，这主要是由于它改进了物理参数。然而，即使在T106模式中，对低波幅状态的偏差已经大大减少，可预报性的丧失仍然与向行星波幅大的区域的转变有关。
A classification of hemispheric flow patterns into clusters corresponding to circulation regimes is used to analyse the errors in geopotential fields predicted by the ECMWF models in the medium range during winter. It is found that the probability distributions of both the r.m.s. error over the northern hemisphere and the amplitude of large-scale planetary waves are bimodal in the region of phase space where the fields have negative or small projection onto an EOF (empirical orthogonal function) that partially resembles the Pacific-North-American pattern, and where the separation between clusters is strongest.
In addition to a true systematic error, a regime-dependent component exists in the error fields, whose pattern is opposite to the anomaly of the observed flow regime. In fact, the ECMWF model (especially its earlier versions) tends to relax the flow towards its most densely populated regime. These results are supported by an analysis of the frequencies of transitions between different clusters, and are interpreted on the basis of simple analytical solutions of the Fokker-Planck equation for stochastically-perturbed dynamical systems.
Statistics computed for different ECMWF models show that a considerable improvement in both the systematic and the regime-dependent component of the medium-range error occurred with the T106 model, owing principally to its improved physical parametrization. However, even though the bias towards states with low wave amplitude has been strongly reduced in the T106 model, a loss of predictability is still associated with transitions to regimes with large amplitude of planetary waves.