Extended range forecast experiments of persistent winter low temperature indexes based on intra-seasonal oscillation over southern China

On the basis of daily NCEP/NCAR reanalysis product and observational data for 1961-2009, this study investigates the low frequency oscillation signals of regional persistent low temperature events (RPLTEs) to the south of 36°N in China and identifies indexes that can be used to characterize RPLTEs. These indexes are then used as predictands in extended range forecast experiments based on the DERF2.0 hindcasts. Results show that the RPLTEs can be classified into three types, i.e. North of Yangtze River, South of Yangtze River, and the entire region. The types of North of and South of Yangtze River have their own key common circulation features that are distinguished by latitudes of anomalous circulation centers and characterized by low-frequency wave trains propagating from northwest to southeast in Asia. 10-30 d low-frequency components of the daily minimum temperature series of North of Yangtze River (T₁) and South of Yangtze River (T₂) are defined as the persistent low temperature indexes (RPLTIs). The phase and amplitude of the RPLTIs have a close relationship with the RPLTEs and are used as the predictands in extended range forecast experiments. EOF1 of the 850 hPa temperature anomalies between 100°-120°E coincides with the low-frequency mode of T₁ while EOF2 coincides with that of T₂. Projection of daily data onto the pair of leading EOFs of 850 hPa temperature anomalies yields time series of principal components that can serve as an effective filter for low-frequency oscillation without the need for bandpass filtering and makes the time series of the two principal components effective predictors for real-time application. DERF2.0 hindcasts and stepwise regression statistical method are employed to explore extended range forecast (ERF) of RPLTIs. The forecast skill of this statistical-dynamical prediction for 2-m temperature is better than that of DERF2.0 (direct model output) in real-time experiments.