| description abstract | nsembles of climate models are commonly used to improve decadal climate predictions and assess the uncertainties associated with them. Weighting the models according to their performances holds the promise of further improving their predictions. Here, an ensemble of decadal climate predictions is used to demonstrate the ability of sequential learning algorithms (SLAs) to reduce the forecast errors and reduce the uncertainties. Three different SLAs are considered, and their performances are compared with those of an equally weighted ensemble, a linear regression, and the climatology. Predictions of four different variables?the surface temperature, the zonal and meridional wind, and pressure?are considered. The spatial distributions of the performances are presented, and the statistical significance of the improvements achieved by the SLAs is tested. The reliability of the SLAs is also tested, and the advantages and limitations of the different measures of the performance are discussed. It was found that the best performances of the SLAs are achieved when the learning period is comparable to the prediction period. The spatial distribution of the SLAs performance showed that they are skillful and better than the other forecasting methods over large continuous regions. This finding suggests that, despite the fact that each of the ensemble models is not skillful, they were able to capture some physical processes that resulted in deviations from the climatology and that the SLAs enabled the extraction of this additional information. | |
