Predictive signals

• November signal 
• May signal

The linear modelling approach is based on identification of joint modes of variability linking predictor and predictand variables.

• A subset of CCA pairs representing two fields is used to build a regression model (Von Storch, 1995) to estimate the anomalies of predictand (SLP, geopotential height at 500 hPa) from the anomalies of the predictor (May SST, November air surface temperature).
• The performance of the model is sensitively dependent on the number of EOFs and CCA pairs used in the regression model.
• Here, the CCA based models have been been trained for the interval 1961-1986 and were independently tested in the interval 1987-2001. Both models use 4 CCA modes derived from 6 EOFs from the paired fields.

Results

• November/DJFM model
• May/DJFM model