Systematic error correction of dynamical seasonal prediction of sea surface temperature using a stepwise pattern project method

Every dynamical climate prediction model has significant errors in its mean state and anomaly field, thus degrading its performance in climate prediction. In addition to correcting the model's systematic errors in the mean state, it is also possible to correct systematic errors in the predicted anomalies by means of dynamical or statistical postprocessing. In this study, a new statistical model has been developed based on the pattern projection method in order to empirically correct the dynamical seasonal climate prediction. The strength of the present model lies in the objective and automatic selection of optimal predictor grid points. The statistical model was applied to systematic error correction of SST anomalies predicted by Seoul National University's (SNU) coupled GCM and evaluated in terms of temporal correlation skill and standardized root-mean-square error. It turns out that the statistical error correction improves the SST prediction over most regions of the global ocean with most forecast lead times up to 6 months. In particular, the SST predictions over the western Pacific and Indian Ocean are improved significantly, where the SNU coupled GCM shows a large error.