Short-term Climate Forecast Experiment of Winter Temperature in China Using Canonical Correlation Analysis

By using the mulativariate linear statistical climate forecast model developed from canonical correlation analysis (CCA), a forecast experiment for winter temperature in China is carried out. Forecast skill is assessed using the test method of historical data independent samples. The results show that there is specified statistical predictive skill for short-term climatic prediction of winter temperature in China using CCA method. In most areas the best correlation coefficient between forecasts and observed values at lead times of 0~2 seasons is more than 0.5. The mean seasonal prediction is more effective than mean monthly one in general. The sea surface temperature is the most effective predictor field. The SST as predictor field can get higher skill score than other single one, and SST combined with other predictors would get much higher skill score