Research On Seasonal And Intra-seasonal Variation Characteristics And Prediction Methods Of Winter Temperature In Xinjian

The spatiotemporal characteristics of the seasonal and intra-seasonal surface air temperature(SAT)variations in winter over Xinjiang are heterogeneous,which bring great uncertainty to local short-term climate prediction.In this study,the temporal and spatial variation characteristics of seasonal and intra-seasonal SAT are analyzed,and the physical relationships between SAT and the corresponding signals are diagnosed.Focused on the above aims,by the highly skilled information of climate models and the preceding external forcing factors of significance,the downscaling prediction models for seasonal and intra-seasonal SAT in Xinjiang are established,which significantly enhance the prediction skills.The conclusions are as follows:(1)The interannual SAT variation of winter in Xinjiang is the most prominent signal of all time scales,which variance contribution is the order of up to 46%.The sea ice over Barents-East Siberian Sea(BES＿SIC)in September-October has significant effects on the following winter SAT in Xinjiang,which can be used as a precursor signal.Based on the BES＿SIC and the 500 h Pa meridional wind in winter over East Greenland-East Siberia predicted by CFSv2 in October,the single-factor and multi-factor schemes of the downscaling prediction models for winter SAT are established through year-to-year increment strategy.All the SAT downscaling hindcast schemes in 1983-2020 winter perform better than CFSv2 outputs,among which the multi-factor scheme being the best.(2)The monthly SAT anomalies in winter in Xinjiang have undergone a transition from cold to warm period during 1980 s.From cold to warm period,the synergistic variability of January and February gets weakened,and intra-seasonal SAT evolution of the first mode shifts from an anti-phase to an in-phase.The main manifestation of the anti-phase mode in the cold period is the anti-phase of December and January-February,while the main manifestation of the anti-phase mode in the warm period is the anti-phase of December-January and February.(3)During the warm period,the intra-seasonal SAT evolution in winter is affected by the in-phase or anti-phase changes of the sea ice over Kara-East Siberian Sea(K-ES)and the sea ice over Barents Sea(BS)in earlier October.Moverever,there is complementarity in the high-impact time of sea ice anomalies in different regions on the intra-seasonal SAT variations.The sea ice anomaly in K-ES can modulate the formation and maintenance of the Ural blockings through stratosphere-troposphere interactions,and the wave energy is enhanced in propagation to the Central Asian region in December and January,which is conducive to forming the meridional dipole wave train from the Ural area to Central Asia,leading to SAT anomalies in December and January.The sea ice anomaly in BS is conducive to the formation and maintenance of the NAO-like phases,and the wave energy is enhanced in propagating downstream in February,which stimulates the circulation anomaly of the SCAND-like teleconnection over Xinjiang,resulting SAT anomaly in February.(4)The winter monthly SAT prediction model is established through year-to-year increment strategy by using the preceding predictor sea ice(SIC)and the contemporaneous predictor sea surface temperature(SST)by CFSv2 which predicts from November,that acts to enhance the prediction skills.The concurrent strong SST signals which may cause winter monthly SAT anomalies are North Pacific(NP)SST,Equatorial Central Pacific(CP)SST,and Kuroshio(KS)SST.Based on the December NP SST(January CP SST,February KS SST)from CFSv2 and the K-ES SIC(K-ES SIC,BS SIC)characteristic in earlier October,the downscaling model for December(January,February)SAT is established.The downscaling hindcast on the December(January,February)SAT during the winter period of 1991-2020 has better prediction skills than CFSv2 direct outputs,with the regional average TCC of 0.42(0.48,0.53),the time average ACC of 0.3(0.24,0.34),and the independent experiment(2015-2020)ACC average of 0.5(0.31,0.37).