A post-processing technique is applied to statistically correct the rainfall forecasts done by numerical climate models available at CPTEC. The statistical algorithm, called Climate Forecast Combining and Correcting Statistical Scheme (ECCOCLIM), is deterministic and based on the Linear Multiple Regression (LMR) method. The skill of the ECCOCLIM is verified with the cross-validation method. Although the forecast improvements using the new method are localized in certain regions and vary throughout the year, they are remarkable over South America, mainly in those areas where climatic previsibility is commonly low as the case of Southeast and Central-West regions. ECCOCLIM shows encouraging results over South America reducing the errors when compared to the traditional arithmetic mean of forecasts of different models.
Climate Forecast; Multimodel Superensemble; ECCOCLIM
