资源环境科技发展态势分析平台

The current dynamical climate models have limited skills in predicting winter temperature in China. The present study uses physics-based empirical models (PEMs) to explore the sources and limits of the seasonal predictability in the total number of extremely cold days (NECD) over China. A combined cluster-rotated EOF analysis reveals two sub-regions of homogeneous variability among hundreds of stations, namely the Northeast China (NE) and Main China (MC). This reduces the large-number of predictands to only two indices, the NCED-NE and NCED-MC, which facilitates detection of the common sources of predictability for all stations. The circulation anomalies associated with the NECD-NE exhibit a zonally symmetric Arctic Oscillation-like pattern, whereas those associated with the NECD-MC feature a North-South dipolar pattern over Asia. The predictability of the NECD originates from SST and snow cover anomalies in the preceding September and October. However, the two regions have different SST predictors: The NE predictor is in the western Eurasian Arctic while the MC predictor is over the tropical-North Pacific. The October snow cover predictors also differ: The NE predictor primarily resides in the central Eurasia while the MC predictor is over the western and eastern Eurasia. The PEM prediction results suggest that about 60% (55%) of the total variance of winter NECD over the NE (Main) China are likely predictable 1 month in advance. The NECD at each station can also be predicted by using the four predictors that were detected for the two indices. The cross-validated temporal correlation skills exceed 0.70 at most stations. The physical mechanisms by which the autumn Arctic sea ice, snow cover, and tropical-North Pacific SST anomalies affect winter NECD over the NE and Main China are discussed.