用1979/1980~2017/2018年冬季逐日长江下游气温资料研究长江下游冬季低温日数与温度低频振荡的联系。结果表明,冬季长江下游逐日气温主要有15~25, 25~40 和50~70 d 的周期振荡,其中长江下游气温的25~40 d 振荡强度年际变化和12~2月低温日数之间有显著的正相关。基于2001~2018年逐日长江下游气温实时25~40 d低频分量和东亚地区850hPa 低频温度主成分,建立了长江下游冬季温度低频分量的延伸期预测的时变扩展复数自回归模型(ECAR)。其中,采用基于T-EOF延拓的实时奇异谱(SSA)滤波,较好地抑制经典SSA滤波的边界效应,得到稳定的实时低频振荡信号。对2001/2002~2017/2018年12~2月长江下游温度低频分量进行独立的实时延伸期预报试验的结果表明, 这种数据驱动的简化的复数预测模型对25~40 d时间尺度的长江下游冬季低频温度分量的预测时效可达26 d左右, 预报能力显著优于经典自回归模型(AR), 能为提前3~4周预报长江下游地区冬季持续低温过程提供有价值的预测背景信息。
Based on the observational data, the variations of intraseasonal oscillation (ISO) of the daily temperatures and its relationships to the low temperature in December－February over the lower reaches of the Yangtze River valley (LYRV) were studied for the period of 1979/1980—2017/2018. It is found that the daily temperatures over the LYRV in December－February is mainly of periodic oscillations of 15－25, 25－40 and 50－70 days, and the interannual variation of the intensity of its 25－40-day oscillation has a strongly positive correlation with the number of low temperature days in December－February. A real-time low frequency components of daily temperature in the LYRV, and the principal components of the Eastern Asian 850 hPa low frequency temperature, over a time period ranging from 2001 to 2018, are used to establish the time-varying extended complex autoregressive model (ECAR) on an extended-range forecast of the 25－40-day low frequency temperature over the LYRV in winter. Using the real-time SSA filtering with the T-EOF extension, it can effectively inhibit the end effects of the traditional SSA and make a better real-time signal of ISO. A 17-year independent real-time extended-range forecast was conducted on the extended-range forecast of low frequency component of the temperature over the LYRV in December－February, for the period ranging from 2001/2002 to 2017/2018. These experimental results show that this ECAR model, which is based on a data-driven model, has a good forecast skill at the lead time of approximately 26 days, with a forecast ability superior to the traditional autoregressive (AR) model. Hence, the development and variation of the leading 25－40-day modes for the Eastern Asian 850 hPa low frequency temperatures and temporal evolutions of their relationships to low frequency components of the temperature over the LYRV in winter supplied the valuable predicting background in determination of extended-range weather process in the persistent low temperature over the LYRV at the 3－4-week leads.