Ensemble Canonical Correlation Prediction Method Of Winter Temperature Over China

The predictors are the monthly mean data of the surface temperature over thecontinent of Europe and Asia, the height over 500hPa of Northern Hemisphere, SST(sea surface temperature) over tropical Indian Ocean and SST over North Pacificin summer from NCEP/NCAR reanalysis datasets. The sample data are the monthly meantemperature in winter of 160 stations over China provided by Beijing MeteorologyCenter. The canonical correlation analysis prediction is used to establishforecasting relationship. Then the ensemble canonical correlation predictionmethod is used to predict winter temperature over China. The forecast skill isanalyzed and independent samples are validated. The results show that differentpredictors have different forecast skill in various regions. When the trainingperiod is 53 years, among the four predictors tested in this text, the surfacetemperature over the continent of Europe and Asia gives the highest skill, theheight over 500hPa of Northern Hemisphere takes second place, SST over tropicalIndian Ocean takes third place, SST over North Pacific takes lastly place. The testshows that ECC method which ensembles several predictors can recognize weeklyforcing in local regions. It ensembles various predictors' skill, so it can providemore general and substantial foundation than the prediction by single predictor.ECC used by regression can exhibition the importance of various predictorsforecasting various regions, so it improves skill substantially in comparison withsimple equal ensemble mean. In the independent samples-validation, SST overtropical Indian Ocean gives the highest general skill, the surface temperature overthe continent of Europe and Asia takes second place, SST over North Pacific takesthird place, the height over 500hPa of Northern Hemisphere takes lastly place. The independent samples-validation also shows that ECC can offset single predictor' slack. Regression ensemble can give more substantial skill. ECC can also predictthat the winter temperature over China has been increasing approximately.