数字滤波方法在月尺度数值预报中的应用 II. 保留不同低频过程的效果比较

The Application of the Digital Filtering Method in Monthly Numerical Weather Prediction Part II: The Comparison of the Forecast Skill with the Different Low Frequency Processes

用数字滤波方法对观测资料序列进行处理，得到初始场用于Ｔ４２Ｌ９全球谱模式的月预报，以去除误差增长较快的高频扰动对低频过程的影响，并且利用多时刻的观测资料提取低频过程的信息。对冬季和夏季两个不同个例进行了实验，并比较了取不同长度的观测序列，截取不同周期的过程作为初值对预报效果的影响。结果说明，经过滤波后对低频分量和平均场的预报都有较显著的改进。而且对于较长时效的预报，应保留更低频的过程（比如１０ｄ以上周期）。最显著的改进是在第２旬。冬季个例经过滤波的初始场在第２旬对北太平洋阻塞形势的预报能力有较明显提高。

The digital filtering method is used to process the observational series and to provide initial fields for the monthly extended range forecasting of the T42L9 global model The aim is to suppress the “noise” of the high frequency (HF) perturbations with the fast error growth rate and to reduce their impact on the low frequency (LF) processes and to extract information of LF modes from the multi time observations The experiments of both winter and summer cases have been completed and their impact on forecast skill of the length of observation series and the cut-off period of the filter are compared The results show that after filtering, the forecast skill for LF modes and for mean fields are significantly improved For the longer forecast rangs, it is preferable to use longer observational series and to filter out more HF modes (such as shorter than 10 day period) The most obvious improvement is for the second 10-day forecast In the winter case, the filtered experiments predict the onset of blocking event over the North Pacific with the reasonable accuracy