华北干旱的多时间尺度组合预测模型

根据华北地区干旱具有显著的年代际和年际变化特性，该文提出了建立多时间尺度预测模型的新构想。利用奇异谱动力学重构的方法将干旱序列的年代际和年际时间尺度变化进行分离，然后分别建立两种时间尺度变化的预测模型，最后将两者进行组合。在建立年际变化预测模型时，使用信噪比的方法将前期大气和海温出现显著异常变化区域的强信号引进到预测模型中。1996~2002年跨季度的预测试验表明，这一建模方案可以较好地反映华北干旱的变化趋势，并具有一定的预测技巧。

A Predicting Drought Model with an Integration of Multi-scale in North China

Based on the significant interdecadal and interannual variations of the drought in North China, a new modeling technique of the drought prediction is proposed, i.e. the drought change is regarded as a composition of the interdecadal variation, the interannual variation and a noise. The technique is composed of modeling both the climate trend, to be represented with the interdecadal variation, and the interannual variation plus the antecedent strong signal. The sum of predictions made with two models is taken as a final prediction of the drought. The prediction experiments for winter, spring, summer and autumn with one to two seasons ahead are made. The results show that the modeling technique can catch the change in drought well in North China. The prediction model differs from the previous modeling, which based on computing correlation between the drought index and the atmospheric and oceanic elements, then considering the correlation be invariable. Hereby the concept of dynamic strong signal is suggested, it is the observation series in some regions where the remarkable difference between the recent atmospheric and oceanic anomaly and the multi year mean occurs. The strong signals varies with different years in which the serious drought occurs, and their spatial, positions and intensities varies too. An extra season hind cast for 1996 to 2002 shows that the modeling is capable of fitting the drought trend and exhibits a higher prediction skill.